Teduglutide significantly reduced the need for parenteral support in patients with short bowel syndrome and intestinal failure, based on data from 85 adults in a randomized, controlled multicenter trial. The findings were published in the December issue of Gastroenterology.

Patients with short bowel syndrome and intestinal failure (SBS-IF) have inadequate intestinal absorption and require parenteral support (PS) to maintain fluids, electrolytes, trace elements, vitamins, and nutrient balances, said Dr. Palle Bekker Jeppesen of Rigshospitalet in Copenhagen and colleagues.

Source: American Gastroenterological Association

Data from previous open-label studies suggest an association between teduglutide and clinically meaningful reductions in wet weight and energy, which may reduce the need for PS in these patients, the investigators noted.

The researchers randomized 86 adults with SBS-IF to either 0.05 mg/kg per day of teduglutide or a placebo. One patient was randomized in error; complete data were available for 42 teduglutide patients and 43 placebo patients.

Significantly more patients in the teduglutide group responded to treatment, compared with the placebo group (63% vs. 30%). This response was defined as sustaining a 20%-100% reduction from baseline in weekly PS volume during weeks 20-24. "Small bowel length did not appear to be a predictor of response," the researchers noted.

The high placebo response may be explained by examining the fluid composite effect, a measure of the combined effects of teduglutide on PS volume reduction as well as the ability to reduce oral fluid intake and increase urine output volume, the researchers noted.

"In the current study, where protocol modifications encouraged earlier and more aggressive PS reductions, significantly larger PS reductions were also achieved in patients receiving placebo, but subsequently these patients had to increase their oral fluid intake significantly to maintain urine production and hydration constant," they said.

After 24 weeks, overall PS volume was reduced by 32% from baseline in teduglutide patients, compared with 21% in placebo patients. Although no patients in either group were completely weaned from parenteral support at 24 weeks, the difference in PS volume reduction was significantly greater in the teduglutide group.

The average weekly PS volume in teduglutide patients decreased significantly from 12.5 L/wk at baseline to 8.1 L/wk at week 24. The placebo patients also had a significant decrease in average weekly PS volume, from 13.4 L/wk at baseline to 11.1 L/wk at week 24.

Treatment-ending adverse events were similar between the two groups; 5% of teduglutide patients and 7% of placebo patients discontinued treatment because of such events during the study period. The most frequently reported treatment-emergent adverse events included abdominal pain, abdominal distension, nausea, and gastrointestinal stoma complications.

Although the study did not specifically assess quality of life measures, significantly more teduglutide patients had at least 1 day off PS, compared with placebo patients, which could help to "liberate considerable time for unhindered daytime activities or undisturbed sleep," the researchers said.

The study did not address the possible benefit of teduglutide therapy earlier in the course of SBS, or the duration of effect after patients discontinued teduglutide, the researchers added.

However, the findings indicate that teduglutide was safe and well tolerated, and "could positively add to the limited treatment armamentarium" for patients with SBS-IF.

Dr. Jeppesen and several coauthors have served on the advisory board of and as consultants to NPS Pharmaceuticals, the company that funded the study. One author is an employee of NPS Pharmaceuticals.

Teduglutide significantly reduced the need for parenteral support in patients with short bowel syndrome and intestinal failure, based on data from 85 adults in a randomized, controlled multicenter trial. The findings were published in the December issue of Gastroenterology.

Patients with short bowel syndrome and intestinal failure (SBS-IF) have inadequate intestinal absorption and require parenteral support (PS) to maintain fluids, electrolytes, trace elements, vitamins, and nutrient balances, said Dr. Palle Bekker Jeppesen of Rigshospitalet in Copenhagen and colleagues.

Source: American Gastroenterological Association

Data from previous open-label studies suggest an association between teduglutide and clinically meaningful reductions in wet weight and energy, which may reduce the need for PS in these patients, the investigators noted.

The researchers randomized 86 adults with SBS-IF to either 0.05 mg/kg per day of teduglutide or a placebo. One patient was randomized in error; complete data were available for 42 teduglutide patients and 43 placebo patients.

Significantly more patients in the teduglutide group responded to treatment, compared with the placebo group (63% vs. 30%). This response was defined as sustaining a 20%-100% reduction from baseline in weekly PS volume during weeks 20-24. "Small bowel length did not appear to be a predictor of response," the researchers noted.

The high placebo response may be explained by examining the fluid composite effect, a measure of the combined effects of teduglutide on PS volume reduction as well as the ability to reduce oral fluid intake and increase urine output volume, the researchers noted.

"In the current study, where protocol modifications encouraged earlier and more aggressive PS reductions, significantly larger PS reductions were also achieved in patients receiving placebo, but subsequently these patients had to increase their oral fluid intake significantly to maintain urine production and hydration constant," they said.

After 24 weeks, overall PS volume was reduced by 32% from baseline in teduglutide patients, compared with 21% in placebo patients. Although no patients in either group were completely weaned from parenteral support at 24 weeks, the difference in PS volume reduction was significantly greater in the teduglutide group.

The average weekly PS volume in teduglutide patients decreased significantly from 12.5 L/wk at baseline to 8.1 L/wk at week 24. The placebo patients also had a significant decrease in average weekly PS volume, from 13.4 L/wk at baseline to 11.1 L/wk at week 24.

Treatment-ending adverse events were similar between the two groups; 5% of teduglutide patients and 7% of placebo patients discontinued treatment because of such events during the study period. The most frequently reported treatment-emergent adverse events included abdominal pain, abdominal distension, nausea, and gastrointestinal stoma complications.

Although the study did not specifically assess quality of life measures, significantly more teduglutide patients had at least 1 day off PS, compared with placebo patients, which could help to "liberate considerable time for unhindered daytime activities or undisturbed sleep," the researchers said.

The study did not address the possible benefit of teduglutide therapy earlier in the course of SBS, or the duration of effect after patients discontinued teduglutide, the researchers added.

However, the findings indicate that teduglutide was safe and well tolerated, and "could positively add to the limited treatment armamentarium" for patients with SBS-IF.

Dr. Jeppesen and several coauthors have served on the advisory board of and as consultants to NPS Pharmaceuticals, the company that funded the study. One author is an employee of NPS Pharmaceuticals.

Teduglutide significantly reduced the need for parenteral support in patients with short bowel syndrome and intestinal failure, based on data from 85 adults in a randomized, controlled multicenter trial. The findings were published in the December issue of Gastroenterology.

Patients with short bowel syndrome and intestinal failure (SBS-IF) have inadequate intestinal absorption and require parenteral support (PS) to maintain fluids, electrolytes, trace elements, vitamins, and nutrient balances, said Dr. Palle Bekker Jeppesen of Rigshospitalet in Copenhagen and colleagues.

Source: American Gastroenterological Association

Data from previous open-label studies suggest an association between teduglutide and clinically meaningful reductions in wet weight and energy, which may reduce the need for PS in these patients, the investigators noted.

The researchers randomized 86 adults with SBS-IF to either 0.05 mg/kg per day of teduglutide or a placebo. One patient was randomized in error; complete data were available for 42 teduglutide patients and 43 placebo patients.

Significantly more patients in the teduglutide group responded to treatment, compared with the placebo group (63% vs. 30%). This response was defined as sustaining a 20%-100% reduction from baseline in weekly PS volume during weeks 20-24. "Small bowel length did not appear to be a predictor of response," the researchers noted.

The high placebo response may be explained by examining the fluid composite effect, a measure of the combined effects of teduglutide on PS volume reduction as well as the ability to reduce oral fluid intake and increase urine output volume, the researchers noted.

"In the current study, where protocol modifications encouraged earlier and more aggressive PS reductions, significantly larger PS reductions were also achieved in patients receiving placebo, but subsequently these patients had to increase their oral fluid intake significantly to maintain urine production and hydration constant," they said.

After 24 weeks, overall PS volume was reduced by 32% from baseline in teduglutide patients, compared with 21% in placebo patients. Although no patients in either group were completely weaned from parenteral support at 24 weeks, the difference in PS volume reduction was significantly greater in the teduglutide group.

The average weekly PS volume in teduglutide patients decreased significantly from 12.5 L/wk at baseline to 8.1 L/wk at week 24. The placebo patients also had a significant decrease in average weekly PS volume, from 13.4 L/wk at baseline to 11.1 L/wk at week 24.

Treatment-ending adverse events were similar between the two groups; 5% of teduglutide patients and 7% of placebo patients discontinued treatment because of such events during the study period. The most frequently reported treatment-emergent adverse events included abdominal pain, abdominal distension, nausea, and gastrointestinal stoma complications.

Although the study did not specifically assess quality of life measures, significantly more teduglutide patients had at least 1 day off PS, compared with placebo patients, which could help to "liberate considerable time for unhindered daytime activities or undisturbed sleep," the researchers said.

The study did not address the possible benefit of teduglutide therapy earlier in the course of SBS, or the duration of effect after patients discontinued teduglutide, the researchers added.

However, the findings indicate that teduglutide was safe and well tolerated, and "could positively add to the limited treatment armamentarium" for patients with SBS-IF.

Dr. Jeppesen and several coauthors have served on the advisory board of and as consultants to NPS Pharmaceuticals, the company that funded the study. One author is an employee of NPS Pharmaceuticals.

On february 28, 2012, the US Food and Drug Administration (FDA) updated its labeling requirements for statins. In addition to revising its recommendations for monitoring liver function and its alerts about reports of memory loss, the FDA also warned of the possibility of new-onset diabetes mellitus and worse glycemic control in patients taking statin drugs.¹

This change stoked an ongoing debate about the risk of diabetes with statin use and the implications of such an effect. To understand the clinical consequences of this alert and its effect on treatment decisions, we need to consider the degree to which statins lower the risk of cardiovascular disease in patients at high risk (including diabetic patients), the magnitude of the risk of developing new diabetes while on statin therapy, and the ratio of risk to benefit in treated populations.

This review will discuss the evidence for this possible adverse effect and the implications for clinical practice.

DO STATINS CAUSE DIABETES?

Individual controlled trials dating back more than a decade have had conflicting results about new diabetes and poorer diabetic control in patients taking statins.

The West of Scotland Coronary Prevention Study (WOSCOPS)² suggested that the incidence of diabetes was 30% lower in patients taking pravastatin (Pravachol) 40 mg/day than with placebo. However, this was not observed with atorvastatin (Lipitor) 10 mg/day in the Anglo-Scandinavian Cardiac Outcomes Trial–Lipid-Lowering Arm (ASCOT-LLA)³ in hypertensive patients or in the Collaborative Atorvastatin Diabetes Study (CARDS)⁴ in diabetic patients,⁴ nor was it noted with simvastatin (Zocor) 40 mg/day in the Heart Protection Study (HPS).⁵

The Justification for the Use of Statins in Primary Prevention: An Intervention Trial Evaluating Rosuvastatin (JUPITER),⁶ using the more potent agent rosuvastatin (Crestor) 20 mg/day in patients with elevated levels of C-reactive protein (CRP), was stopped early when an interim analysis found a 44% lower incidence of the primary end point. However, the trial also reported a 26% higher incidence of diabetes in follow-up of less than 2 years.

In the Prospective Study of Pravastatin in the Elderly at Risk (PROSPER),⁷ with a mean age at entry of 75, there was a 32% higher incidence of diabetes with pravastatin therapy.⁷

Results of meta-analyses

Several meta-analyses have addressed these differences.

Rajpathak et al⁸ performed a meta-analysis, published in 2009, of six trials—WOSCOPS,² ASCOT-LLA,³ JUPITER,⁶ HPS,⁵ the Long-term Intervention With Pravastatin in Ischaemic Disease (LIPID) study,⁹ and the Controlled Rosuvastatin Multinational Study in Heart Failure (CORONA),¹⁰ with a total of 57,593 patients. They calculated that the incidence of diabetes was 13% higher (an absolute difference of 0.5%) in statin recipients, which was statistically significant. In their initial analysis, the authors excluded WOSCOPS, describing it as hypothesis-generating. The relative increase in risk was less—6%—and was not statistically significant when WOSCOPS was included.

Sattar et al,¹¹ in a larger meta-analysis published in 2010, included 91,140 participants in 13 major statin trials conducted between 1994 and 2009; each trial had more than 1,000 patients and more than 1 year of follow-up.^{2,3,5–7,9,10,12–17} New diabetes was defined as physician reporting of new diabetes, new diabetic medication use, or a fasting glucose greater than 7 mmol/L (126 mg/dL).

New diabetes occurred in 2,226 (4.89%) of the statin recipients and in 2,052 (4.5%) of the placebo recipients, an absolute difference of 0.39%, or 9% more (odds ratio [OR] 1.09; 95% confidence interval [CI] 1.02–1.17) (Figure 1).

The incidence of diabetes varied substantially among the 13 trials, with only JUPITER⁶ and PROSPER⁷ finding statistically significant increases in rates (26% and 32%, respectively). Of the other 11 trials, 4 had nonsignificant trends toward lower incidence,^2,9,13,17 while the 7 others had nonsignificant trends toward higher incidence.

Does the specific statin make a difference?

Questions have been raised as to whether the type of statin used, the intensity of therapy, or the population studied contributed to these differences. Various studies suggest that factors such as using hydrophilic vs lipophilic statins (hydrophilic statins include pravastatin and rosuvastatin; lipophilic statins include atorvastatin, lovastatin, and simvastatin), the dose, the extent of lowering of low-density lipoprotein cholesterol (LDL-C), and the age or clinical characteristics of the population studied may influence this relationship.^18–20

Yamakawa et al¹⁸ examined the effect of atorvastatin 10 mg/day, pravastatin 10 mg/day, and pitavastatin (Livalo) 2 mg/day on glycemic control over 3 months in a retrospective analysis. Random blood glucose and hemoglobin A_1c levels were increased in the atorvastatin group but not in the other two.¹⁸

A prospective comparison of atorvastatin 20 mg vs pitavastatin 4 mg in patients with type 2 diabetes, presented at the American College of Cardiology’s 2011 annual meeting, reported a significant increase in fasting glucose levels with atorvastatin, particularly in women, but not with pitavastatin.¹⁹

In the Compare the Effect of Rosuvastatin With Atorvastatin on Apo B/Apo A-1 Ratio in Patients With Type 2 Diabetes Mellitus and Dyslipidaemia (CORALL) study,²⁰ both high-dose rosuvastatin (40 mg) and high-dose atorvastatin (80 mg) were associated with significant increases in hemoglobin A_1c, although the mean fasting glucose levels were not significantly different at 18 weeks of therapy.

A meta-analysis by Sattar et al¹¹ did not find a clear difference between lipophilic statins (OR 1.10 vs placebo) and hydrophilic statins (OR 1.08). In analysis by statin type, the combined rosuvastatin trials were statistically significant in favor of a higher diabetes risk (OR 1.18, 95% CI 1.04–1.44). Nonsignificant trends were noted for atorvastatin trials (OR 1.14) and simvastatin trials (OR 1.11) and less so for pravastatin (OR 1.03); the OR for lovastatin was 0.98. This may suggest that there is a stronger effect with more potent statins or with greater lowering of LDL-C.

Meta-regression analysis in this study demonstrated that diabetes risk with statins was higher in older patients but was not influenced by body mass index or by the extent that LDL-C was lowered.

Intensive-dose statin therapy has been shown to reduce cardiovascular risk more than low-dose or moderate-dose therapy, thus supporting more aggressive treatment of LDL-C in higher-risk patients. However, some controlled studies comparing more-potent with less-potent statin regimens suggest that there may also be a higher risk of incident diabetes at higher doses.^21–24

In a post hoc analysis of the Pravastatin or Atorvastatin Evaluation and Infection Therapy– Thrombolysis in Myocardial Infarction 22 (PROVE-IT TIMI 22) trial,²¹ patients who had experienced an acute coronary syndrome had a greater increase in hemoglobin A_1c if treated with atorvastatin 80 mg/day than with pravastatin 40 mg/day.

Waters et al²³ reported a higher risk of new diabetes with atorvastatin 80 mg than with placebo and a trend toward a higher risk with atorvastatin 80 mg than with atorvastatin 10 mg or simvastatin 20 mg.

In contrast, a review by Yousef et al²⁴ of the data from the Enhanced Feedback for Effective Cardiac Treatment (EFFECT) study did not find a higher diabetes risk with more intensive statin therapy based on the magnitude of LDL-C reduction. A propensity-matched examination of deaths, recurrent acute ischemic events, or new diabetes in patients previously hospitalized with myocardial infarction found no differences in these end points each year out to 5 years. The risk of diabetes was in fact lower (but the difference was not statistically significant) in the high-dose groups out to 5 years. The risk of myocardial infarction or death was numerically different in the high-dose groups, but the difference was not statistically significant.

Preiss et al²⁵ in 2011 performed a meta-analysis of the impact of intensity of statin therapy on diabetes risk. They examined data from 32,752 participants without diabetes at baseline in five randomized controlled trials with more than 1,000 participants and more than 1 year of follow-up, comparing high-dose therapy against moderate-dose statin therapy.^{21,22,26–28} New diabetes was considered present if there was an adverse event report of diabetes, if glucose-lowering drugs were started, or if two fasting plasma glucose measurements were higher than 7 mmol/L (126 mg/dL).

Diabetes developed in 1,449 (8.8%) of the intensive-therapy group and 1,300 (8.0%) of the moderate-therapy group (OR 1.12, 95% CI 1.04–1.22). In contrast, incident cardiovascular disease occurred in 3,134 (19.1%) of the intensive-therapy group and 3,550 (21.7%) of the moderate-therapy group (OR 0.84, 95% CI 0.75–0.94). Therefore, there was an 0.8% absolute increase in diabetes cases on high-dose statins and a 2.6% absolute reduction in adverse cardiovascular events.

CAUTION IN INTERPRETING THESE DATA

There are many reasons for caution in interpreting these studies.

The trials were not designed to look for diabetes

The data supporting the relationship between statin therapy and higher risk of diabetes are primarily from observational studies. These studies were not prospectively designed to address this question, and we therefore need to view this as association and not as causation.

The definition of diabetes varied between trials, and new-onset diabetes was often not rigorously screened for. In many trials the outcome of diabetes was at least partially based on nonstandardized, nonadjudicated physician reporting.

Consequently, if statins reduce the risk of diabetes, the results from WOSCOPS may overstate the reduction, since this study used a non-standard definition of incident diabetes (fasting plasma glucose > 126 mg/dL plus a > 36 mg/dL increase from baseline). When Sattar et al¹¹ reanalyzed WOSCOPS data using a more standard definition, they found a smaller effect.

On the other hand, nonstandardized physician reporting may overstate an adverse effect. Sattar et al¹¹ also found that when fasting plasma glucose levels alone were used as the definition for diabetes, the overall risk was attenuated and was no longer statistically significant (OR 1.07, 95% CI 0.97–1.17).

Perhaps statin therapy uncovers diabetes only in people at risk of diabetes

Perhaps statin therapy uncovers diabetes only in people at higher baseline risk of developing diabetes. Therefore, this adverse effect may be restricted to certain groups and not applicable to the general population.

In JUPITER, one of the two trials in which, on independent analysis, statin use was associated with new diabetes, 77% of patients in the rosuvastatin group who developed diabetes had impaired fasting glucose at entry and therefore were at higher risk of developing diabetes.⁶

Possibly, the relationship is driven by preexisting metabolic syndrome or other risk factors for diabetes. In the two studies that reported a statistically significantly higher incidence of new diabetes, more than 40% of patients in JUPITER met the criteria for metabolic syndrome, and metabolic syndrome, which increases in prevalence with age, was likely more prominent in the elderly population in PROSPER.

Waters et al²³ grouped patients according to whether they had risk factors for diabetes (impaired fasting glucose, obesity, elevated triglycerides, and hypertension) and found that those who had none or one of these risk factors had no difference in the rate of new-onset diabetes with either moderate or intensive statin therapy, but the risk was pronounced in those who had three or four risk factors.

Ridker et al²⁹ reanalyzed the JUPITER data from patients who did not have cardiovascular disease at baseline. Overall, for every 54 new cases of diabetes in follow-up, 134 cardiovascular events or deaths were prevented. In subgroup analysis, those who had one or more risk factors for diabetes at baseline (metabolic syndrome, impaired fasting glucose, obesity, or hemoglobin A_1c > 6%) had a 39% reduction in the primary end point and a 28% increase in new diabetes. Those who had none of these risk factors had a 52% lower rate of cardiovascular events but no increase in diabetes.

Other confounding factors

Bias and confounding factors are difficult to control for in studies without prospectively defined, recognized, and analyzed outcomes.

Although it may be a bit of a stretch, residual confounding factors such as myalgia side effects while on statins may reduce exercise in the statin-treatment groups. Perhaps a change to a healthier lifestyle after cardiovascular events may be more common in placebo groups. Improved survival with statins may allow more people at risk of diabetes to live longer and present with the diagnosis.³⁰

POSSIBLE EXPLANATIONS, BUT NO UNIFYING MECHANISM

If mechanisms could be identified to explain the association between statins and diabetes, this would strengthen the argument that it is a cause-and-effect relationship. Many explanations have been proposed as to how statins may influence glucose metabolism and insulin sensitivity.^31–34 These are possible explanations based on other observations.

In theory, statins may improve insulin sensitivity via their anti-inflammatory effect, since inflammatory markers and proinflammatory cytokines have been linked with insulin resistance. However, other effects of statins may adversely affect glycemic control.

In vivo analysis has shown that some but not all statins increase insulin levels and decrease insulin sensitivity in a dose-dependent fashion. Some statins decrease adiponectin and may worsen glycemic control through loss of adiponectin’s proposed protective anti-proliferative and antiangiogenic properties. In vitro studies and animal studies have demonstrated a decrease in expression of insulin-responsive glucose transporter 4 (GLUT4) with atorvastatin, and an increase in GLUT1. It has been hypothesized that reduction in isoprenoid biosynthesis or decreased insulin signaling may explain these effects and that changes in glucose transport in adipocytes may cause insulin resistance. Other studies suggest that dysregulation of cellular cholesterol may attenuate beta-cell function. Impaired biosynthesis of ubiquinones may result in delayed production of adenosine triphosphate and consequently diminish insulin release.

But different effects have been reported for atorvastatin, simvastatin, and pravastatin, arguing against a unifying explanation or, alternatively, suggesting that differences in lipophilicity and potency among statins are important. Hydrophilic statins may be less likely to be taken up by extrahepatic cells such as pancreatic cells and adipocytes, possibly lessening these effects. However, the strong association between rosuvastatin (which is hydrophilic) and new diabetes would not support this hypothesis.

Despite these speculations, lack of conformity in response to different statins and discrepancies in the clinical outcomes noted in trials fail to clearly identify a common causative mechanism.

OTHER COMMON THERAPIES MAY INFLUENCE GLYCEMIC CONTROL

Statins are not the first drugs for reducing cardiovascular risk that have been shown to affect glucose levels during treatment.

Niacin

Niacin has been known to increase glucose levels but has long been used as a treatment for dyslipidemia despite this caution. Reduced glycemic control during niacin treatment in diabetic patients does not seem to alter the beneficial effects of treatment.^35–37

In a post hoc analysis of the Coronary Drug Project (CDP), in patient subgroups defined by baseline fasting plasma glucose and compared with placebo, niacin reduced the 6-year risk of recurrent myocardial infarction and the combined end point of coronary heart disease death or nonfatal myocardial infarction similarly (interactive P value nonsignificant) across all levels of baseline fasting plasma glucose, including levels of 126 mg/dL or higher at study entry.³⁶

In another post hoc analysis of CDP patient subgroups defined by the change in glycemic status from baseline to 1 year, niacin reduced the 6-year risk of the same end points similarly (interactive P value nonsignificant) across all levels of change in fasting plasma glucose from baseline to year 1, whether baseline fasting plasma glucose levels decreased, stayed the same, or increased to 10 mg/dL or higher on niacin therapy.³⁶

Therefore, the beneficial effect of niacin of reducing the rate of recurrent nonfatal myocardial infarction and coronary heart disease events was not significantly diminished when impaired fasting glucose or diabetes was present when therapy was started or by on-therapy increases from baseline fasting plasma glucose.

In addition, on-therapy changes in glycemic control may be dose-related and minimized by surveillance and therapy adjustments. The Assessment of Diabetes Control and Evaluation of the Efficacy of Niaspan Trial (ADVENT)³⁸ found that changes in glycemic control were minimal as measured by fasting glucose and hemoglobin A_1c; were associated with a higher niacin dose (1.5 g/day vs 1 g/day); and, when present, were successfully managed by adjusting the diabetes treatment regimen.

Antihypertensive drugs

Diuretics as well as beta-blockers have been reported to increase the incidence of diabetes in patients with hypertension.^15,38–40

A retrospective longitudinal cohort study⁴⁰ in 2009 examined the development of new-onset diabetes (defined as a new ICD-9 code for diabetes or initiation of diabetes treatment) in 24,688 treated hypertensive patients without diabetes at baseline; 4,385 (17.8%) of the patients developed diabetes. After adjusting for sex and age, the risk of new diabetes was significant in users of diuretics (OR 1.10), beta-blockers (OR 1.12), and calcium channel blockers (OR 1.10) compared with users of angiotensin-converting enzyme inhibitors, (OR 0.92), angiotensin receptor blockers (OR 0.90), or alpha-blockers (OR 0.88).

However, the increase in blood glucose does not seem to attenuate the beneficial effects of reducing cardiovascular events. In the Antihypertensive and Lipid-lowering Treatment to Prevent Heart Attack Trial (ALLHAT),¹⁵ a long-term follow-up of those developing new-onset diabetes while taking chlorthalidone (Hygroton) found no difference in the risk of death from cardiovascular disease or from any cause (hazard ratio = 1.04).¹⁵

CLINICAL IMPLICATIONS

Balancing the benefits and risks of statins

It is important to examine how the 0.4% increase in absolute risk of new-onset diabetes as calculated in meta-analyses compares with the benefits of statin treatment in terms of cardiovascular risk reduction.

Using data from the Cholesterol Treatment Trialists (CTT) meta-analysis of statin trials in 71,370 participants, Sattar et al¹¹ estimated that statin treatment is associated with 5.4 fewer deaths from coronary heart disease and cases of nonfatal myocardial infarction per 255 patients treated over 4 years for each 1-mmol/L (39 mg/dL) reduction in LDL-C compared with controls. The benefit would be even greater if stroke, revascularization, and hospitalization are included as end points. This benefit is contrasted with the risk of developing 1 additional case of diabetes for every 255 patients treated with statins over the same period.

Preiss et al²⁵ calculated that there were 2 more cases of diabetes per 1,000 patient-years in patients receiving intensive doses than in those receiving moderate doses (18.9 vs 16.9), corresponding to 1 additional case of diabetes for every 498 patients treated per year. However, there were 6.5 fewer first major cardiovascular events per 1,000 patient-years (44.5 vs 51.0), corresponding to a number needed to treat per year to prevent 1 cardiovascular event of 155. Most of the benefit was due to fewer revascularizations, followed by nonfatal myocardial infarctions. The 12% increase in new diabetes with high-dose therapy contrasted with a 16% reduction in new cardiovascular disease combined events (OR 0.84, 95% CI 0.75–0.94).

As previously noted, in the JUPITER trial, the benefits of preventing cardiovascular events with statin therapy outweighed the risk of new diabetes in people both with and without baseline risk factors for diabetes.²⁹ Similar to the observations with niacin and some antihypertensive drugs, the increase in blood glucose with statins does not appear to reduce the benefits of cardiovascular risk reduction in these patients at moderate to high risk, even when used at high doses.

Diabetes is a coronary heart disease risk equivalent and is associated with high risk of cardiovascular events.^41–46 Overall, the risk for these adverse events is two to four times greater in people with diabetes than without. Atherosclerosis-related events account for approximately 65% to 75% of all deaths in people with diabetes, and 75% of these events are coronary. Lipid abnormalities are strongly correlated with the risk of cardiovascular disease in people with diabetes, and aggressive treatment of risk factors, particularly lipid abnormalities, has been shown to reduce this risk.^47–49 And data from multiple clinical trials support the use of statins to lower LDL-C as the first-line therapy for dyslipidemia in people with diabetes, just as it is in the general population.^{3–7,9,13,23,50–61}

Analyses of diabetic subgroups encompassing 18,000 to 20,000 patients in the large statin trials have clearly demonstrated the benefits of statin therapy. A recent metaanalysis of 10 placebo-controlled trials that included approximately 16,000 patients with diabetes and 54,000 without diabetes demonstrated a 30% reduction in coronary heart disease, a 19% reduction in strokes, and a 12% reduction in mortality.⁵⁴ Furthermore, in another meta-analysis of 14 trials, a similar 22% reduction in coronary heart disease was noted in people with diabetes whether or not they had a history of cardiovascular disease.⁵⁵

Therefore, aggressive treatment of lipid abnormalities with statins as primary treatment has generally been adopted as a standard of care in diabetic patients, particularly those with clinical cardiovascular disease or one or more risk factors. The Adult Treatment Panel III guidelines recommend a minimum LDL-C goal of less than 100 mg/dL and a goal of less than 70 mg/dL as an option for patients with diabetes (Table 1).^41,62 Similar recommendations have been issued by the American Diabetes Association together with the American College of Cardiology (Table 2),³⁰ the American Diabetes Association by itself,⁶³ and the American Academy of Pediatrics.⁶

Is new-onset diabetes as dangerous as established diabetes?

In studies to date, there did not appear to be more events in those who developed new-onset diabetes.

Waters et al,²⁴ evaluating three trials of high-dose atorvastatin therapy, found that major cardiovascular events occurred in 11.3% of those with new-onset diabetes, 10.8% of those without new-onset diabetes (HR 1.02, 95% CI 0.77–1.35), and 17.5% of those who had diabetes at baseline.

Therefore, it may not be appropriate to extrapolate the glucose changes seen on statin therapy to an equivalent increase in adverse cardiovascular events as seen in other diabetic patients. The beneficial reduction in cardiovascular events does not appear to be diminished in those developing diabetes. It is not clear that the increase in glucose on statins has the same implications of a new diagnosis of diabetes. Does this elevation in glucose represent true diabetes or some downstream effect? For example, thiazide diuretics have been known to increase blood glucose levels, but the levels drop when these drugs are discontinued, even after many years of treatment.

On the other hand, it is possible that follow-up of 5 years or less in clinical trials has not allowed sufficient time to examine the influence of the increase in new-onset diabetes on future cardiovascular events. In addition, because of the widespread use of statins across a broad range of cardiovascular risk, even if the effect is small in absolute terms, the potential adverse effects are magnified, particularly in a low-risk population in which the cardiovascular benefits are smaller.

The association is real, but questions remain

In view of the evidence, it is difficult to refute that an association exits between statin use and new-onset diabetes, at least in some subgroups. The dose response noted in some studies further reinforces the conclusion that the association is real. However, many questions remain unanswered regarding mechanism of effect, whether there are differences depending on the particular statin or dose used, or differential effects in the populations treated (such as patients with metabolic syndrome or the elderly).

Until the contradictory observations can be resolved and plausible mechanisms of action elucidated, causality cannot be established. From a clinical standpoint there is no current evidence suggesting that the elevations in blood glucose seen while on lipid-lowering or blood-pressure-lowering therapy are associated with an increased risk of cardiovascular events or that they attenuate the beneficial effects of the therapy.

Statins should continue to be used in patients at high risk

Until further studies are done, statins should continue to be used, after assessing the risks and the benefits.

Primary prevention patients at moderate to high risk and secondary prevention patients stand to gain from statin therapy, and it should not be denied or doses reduced on the basis of concerns about the development of new-onset diabetes. The recognized modest risk of developing diabetes does not appear to blunt the cardioprotective effects of statin therapy in these moderate-to high-risk groups.

Rather than stop statins in patients at risk of diabetes such as the elderly or those with prediabetes, insulin resistance, or metabolic syndrome who are on therapy for appropriate reasons, it is reasonable to continue these drugs, monitoring glucose more closely and emphasizing the importance of weight reduction, diet, and aerobic exercise for preventing diabetes. The Diabetes Prevention Program Research Group, for example, reduced the incidence of diabetes by 58% over 2.8 years of follow-up with intensive lifestyle interventions (a low-calorie, low-fat diet plus moderate physical activity 150 minutes per week) vs usual care in at-risk populations.⁶⁵

Should statins be used more cautiously in patients at lower risk?

The most recent Cholesterol Treatment Trialists meta-analysis of 27 randomized clinical trials (22 placebo-controlled, 134,537 people; 5 high-dose vs low-dose, 39,612 people) reported that reducing LDL-C with statins lowered cardiovascular risk even in low-risk patients.⁶⁶ Overall, there were 21% fewer major cardiovascular events (coronary heart disease, stroke, or coronary revascularization) for every 1-mmol/L reduction in LDL-C.

The proportional reduction in events was at least as large in the two lowest-risk groups (estimated 5-year risk of < 5% and 5% to < 10%, 53,152 people) as in the higher-risk groups. This was reflected mainly in fewer nonfatal myocardial infarctions and coronary revascularizations. In these groups, the absolute reduction in risk for each 1-mmol/L reduction in LDL-C was 11 per 1,000 patients over 5 years. Even in this low-risk population, the reduction in cardiovascular risk seems to compare favorably with the small estimated increase risk of diabetes.

However, even in the lowest-risk group studied, the average baseline LDL-C level was greater than 130 mg/dL.

Therefore, in groups in which the benefits of statins on cardiovascular risk reduction are less robust (eg, low-risk primary prevention groups without significant elevations in LDLC, particularly the elderly), it would not be difficult to justify the case for more cautious use of statin therapy. If statins are used in these low-risk groups, restricting their use to those with at least moderate LDL-C elevation, using less aggressive LDL-C-lowering targets, and regular monitoring of fasting glucose seem reasonable until further information is available.

On february 28, 2012, the US Food and Drug Administration (FDA) updated its labeling requirements for statins. In addition to revising its recommendations for monitoring liver function and its alerts about reports of memory loss, the FDA also warned of the possibility of new-onset diabetes mellitus and worse glycemic control in patients taking statin drugs.¹

This change stoked an ongoing debate about the risk of diabetes with statin use and the implications of such an effect. To understand the clinical consequences of this alert and its effect on treatment decisions, we need to consider the degree to which statins lower the risk of cardiovascular disease in patients at high risk (including diabetic patients), the magnitude of the risk of developing new diabetes while on statin therapy, and the ratio of risk to benefit in treated populations.

This review will discuss the evidence for this possible adverse effect and the implications for clinical practice.

DO STATINS CAUSE DIABETES?

Individual controlled trials dating back more than a decade have had conflicting results about new diabetes and poorer diabetic control in patients taking statins.

The West of Scotland Coronary Prevention Study (WOSCOPS)² suggested that the incidence of diabetes was 30% lower in patients taking pravastatin (Pravachol) 40 mg/day than with placebo. However, this was not observed with atorvastatin (Lipitor) 10 mg/day in the Anglo-Scandinavian Cardiac Outcomes Trial–Lipid-Lowering Arm (ASCOT-LLA)³ in hypertensive patients or in the Collaborative Atorvastatin Diabetes Study (CARDS)⁴ in diabetic patients,⁴ nor was it noted with simvastatin (Zocor) 40 mg/day in the Heart Protection Study (HPS).⁵

The Justification for the Use of Statins in Primary Prevention: An Intervention Trial Evaluating Rosuvastatin (JUPITER),⁶ using the more potent agent rosuvastatin (Crestor) 20 mg/day in patients with elevated levels of C-reactive protein (CRP), was stopped early when an interim analysis found a 44% lower incidence of the primary end point. However, the trial also reported a 26% higher incidence of diabetes in follow-up of less than 2 years.

In the Prospective Study of Pravastatin in the Elderly at Risk (PROSPER),⁷ with a mean age at entry of 75, there was a 32% higher incidence of diabetes with pravastatin therapy.⁷

Results of meta-analyses

Several meta-analyses have addressed these differences.

Rajpathak et al⁸ performed a meta-analysis, published in 2009, of six trials—WOSCOPS,² ASCOT-LLA,³ JUPITER,⁶ HPS,⁵ the Long-term Intervention With Pravastatin in Ischaemic Disease (LIPID) study,⁹ and the Controlled Rosuvastatin Multinational Study in Heart Failure (CORONA),¹⁰ with a total of 57,593 patients. They calculated that the incidence of diabetes was 13% higher (an absolute difference of 0.5%) in statin recipients, which was statistically significant. In their initial analysis, the authors excluded WOSCOPS, describing it as hypothesis-generating. The relative increase in risk was less—6%—and was not statistically significant when WOSCOPS was included.

Sattar et al,¹¹ in a larger meta-analysis published in 2010, included 91,140 participants in 13 major statin trials conducted between 1994 and 2009; each trial had more than 1,000 patients and more than 1 year of follow-up.^{2,3,5–7,9,10,12–17} New diabetes was defined as physician reporting of new diabetes, new diabetic medication use, or a fasting glucose greater than 7 mmol/L (126 mg/dL).

New diabetes occurred in 2,226 (4.89%) of the statin recipients and in 2,052 (4.5%) of the placebo recipients, an absolute difference of 0.39%, or 9% more (odds ratio [OR] 1.09; 95% confidence interval [CI] 1.02–1.17) (Figure 1).

The incidence of diabetes varied substantially among the 13 trials, with only JUPITER⁶ and PROSPER⁷ finding statistically significant increases in rates (26% and 32%, respectively). Of the other 11 trials, 4 had nonsignificant trends toward lower incidence,^2,9,13,17 while the 7 others had nonsignificant trends toward higher incidence.

Does the specific statin make a difference?

Questions have been raised as to whether the type of statin used, the intensity of therapy, or the population studied contributed to these differences. Various studies suggest that factors such as using hydrophilic vs lipophilic statins (hydrophilic statins include pravastatin and rosuvastatin; lipophilic statins include atorvastatin, lovastatin, and simvastatin), the dose, the extent of lowering of low-density lipoprotein cholesterol (LDL-C), and the age or clinical characteristics of the population studied may influence this relationship.^18–20

Yamakawa et al¹⁸ examined the effect of atorvastatin 10 mg/day, pravastatin 10 mg/day, and pitavastatin (Livalo) 2 mg/day on glycemic control over 3 months in a retrospective analysis. Random blood glucose and hemoglobin A_1c levels were increased in the atorvastatin group but not in the other two.¹⁸

A prospective comparison of atorvastatin 20 mg vs pitavastatin 4 mg in patients with type 2 diabetes, presented at the American College of Cardiology’s 2011 annual meeting, reported a significant increase in fasting glucose levels with atorvastatin, particularly in women, but not with pitavastatin.¹⁹

In the Compare the Effect of Rosuvastatin With Atorvastatin on Apo B/Apo A-1 Ratio in Patients With Type 2 Diabetes Mellitus and Dyslipidaemia (CORALL) study,²⁰ both high-dose rosuvastatin (40 mg) and high-dose atorvastatin (80 mg) were associated with significant increases in hemoglobin A_1c, although the mean fasting glucose levels were not significantly different at 18 weeks of therapy.

A meta-analysis by Sattar et al¹¹ did not find a clear difference between lipophilic statins (OR 1.10 vs placebo) and hydrophilic statins (OR 1.08). In analysis by statin type, the combined rosuvastatin trials were statistically significant in favor of a higher diabetes risk (OR 1.18, 95% CI 1.04–1.44). Nonsignificant trends were noted for atorvastatin trials (OR 1.14) and simvastatin trials (OR 1.11) and less so for pravastatin (OR 1.03); the OR for lovastatin was 0.98. This may suggest that there is a stronger effect with more potent statins or with greater lowering of LDL-C.

Meta-regression analysis in this study demonstrated that diabetes risk with statins was higher in older patients but was not influenced by body mass index or by the extent that LDL-C was lowered.

Intensive-dose statin therapy has been shown to reduce cardiovascular risk more than low-dose or moderate-dose therapy, thus supporting more aggressive treatment of LDL-C in higher-risk patients. However, some controlled studies comparing more-potent with less-potent statin regimens suggest that there may also be a higher risk of incident diabetes at higher doses.^21–24

In a post hoc analysis of the Pravastatin or Atorvastatin Evaluation and Infection Therapy– Thrombolysis in Myocardial Infarction 22 (PROVE-IT TIMI 22) trial,²¹ patients who had experienced an acute coronary syndrome had a greater increase in hemoglobin A_1c if treated with atorvastatin 80 mg/day than with pravastatin 40 mg/day.

Waters et al²³ reported a higher risk of new diabetes with atorvastatin 80 mg than with placebo and a trend toward a higher risk with atorvastatin 80 mg than with atorvastatin 10 mg or simvastatin 20 mg.

In contrast, a review by Yousef et al²⁴ of the data from the Enhanced Feedback for Effective Cardiac Treatment (EFFECT) study did not find a higher diabetes risk with more intensive statin therapy based on the magnitude of LDL-C reduction. A propensity-matched examination of deaths, recurrent acute ischemic events, or new diabetes in patients previously hospitalized with myocardial infarction found no differences in these end points each year out to 5 years. The risk of diabetes was in fact lower (but the difference was not statistically significant) in the high-dose groups out to 5 years. The risk of myocardial infarction or death was numerically different in the high-dose groups, but the difference was not statistically significant.

Preiss et al²⁵ in 2011 performed a meta-analysis of the impact of intensity of statin therapy on diabetes risk. They examined data from 32,752 participants without diabetes at baseline in five randomized controlled trials with more than 1,000 participants and more than 1 year of follow-up, comparing high-dose therapy against moderate-dose statin therapy.^{21,22,26–28} New diabetes was considered present if there was an adverse event report of diabetes, if glucose-lowering drugs were started, or if two fasting plasma glucose measurements were higher than 7 mmol/L (126 mg/dL).

Diabetes developed in 1,449 (8.8%) of the intensive-therapy group and 1,300 (8.0%) of the moderate-therapy group (OR 1.12, 95% CI 1.04–1.22). In contrast, incident cardiovascular disease occurred in 3,134 (19.1%) of the intensive-therapy group and 3,550 (21.7%) of the moderate-therapy group (OR 0.84, 95% CI 0.75–0.94). Therefore, there was an 0.8% absolute increase in diabetes cases on high-dose statins and a 2.6% absolute reduction in adverse cardiovascular events.

CAUTION IN INTERPRETING THESE DATA

There are many reasons for caution in interpreting these studies.

The trials were not designed to look for diabetes

The data supporting the relationship between statin therapy and higher risk of diabetes are primarily from observational studies. These studies were not prospectively designed to address this question, and we therefore need to view this as association and not as causation.

The definition of diabetes varied between trials, and new-onset diabetes was often not rigorously screened for. In many trials the outcome of diabetes was at least partially based on nonstandardized, nonadjudicated physician reporting.

Consequently, if statins reduce the risk of diabetes, the results from WOSCOPS may overstate the reduction, since this study used a non-standard definition of incident diabetes (fasting plasma glucose > 126 mg/dL plus a > 36 mg/dL increase from baseline). When Sattar et al¹¹ reanalyzed WOSCOPS data using a more standard definition, they found a smaller effect.

On the other hand, nonstandardized physician reporting may overstate an adverse effect. Sattar et al¹¹ also found that when fasting plasma glucose levels alone were used as the definition for diabetes, the overall risk was attenuated and was no longer statistically significant (OR 1.07, 95% CI 0.97–1.17).

Perhaps statin therapy uncovers diabetes only in people at risk of diabetes

Perhaps statin therapy uncovers diabetes only in people at higher baseline risk of developing diabetes. Therefore, this adverse effect may be restricted to certain groups and not applicable to the general population.

In JUPITER, one of the two trials in which, on independent analysis, statin use was associated with new diabetes, 77% of patients in the rosuvastatin group who developed diabetes had impaired fasting glucose at entry and therefore were at higher risk of developing diabetes.⁶

Possibly, the relationship is driven by preexisting metabolic syndrome or other risk factors for diabetes. In the two studies that reported a statistically significantly higher incidence of new diabetes, more than 40% of patients in JUPITER met the criteria for metabolic syndrome, and metabolic syndrome, which increases in prevalence with age, was likely more prominent in the elderly population in PROSPER.

Waters et al²³ grouped patients according to whether they had risk factors for diabetes (impaired fasting glucose, obesity, elevated triglycerides, and hypertension) and found that those who had none or one of these risk factors had no difference in the rate of new-onset diabetes with either moderate or intensive statin therapy, but the risk was pronounced in those who had three or four risk factors.

Ridker et al²⁹ reanalyzed the JUPITER data from patients who did not have cardiovascular disease at baseline. Overall, for every 54 new cases of diabetes in follow-up, 134 cardiovascular events or deaths were prevented. In subgroup analysis, those who had one or more risk factors for diabetes at baseline (metabolic syndrome, impaired fasting glucose, obesity, or hemoglobin A_1c > 6%) had a 39% reduction in the primary end point and a 28% increase in new diabetes. Those who had none of these risk factors had a 52% lower rate of cardiovascular events but no increase in diabetes.

Other confounding factors

Bias and confounding factors are difficult to control for in studies without prospectively defined, recognized, and analyzed outcomes.

Although it may be a bit of a stretch, residual confounding factors such as myalgia side effects while on statins may reduce exercise in the statin-treatment groups. Perhaps a change to a healthier lifestyle after cardiovascular events may be more common in placebo groups. Improved survival with statins may allow more people at risk of diabetes to live longer and present with the diagnosis.³⁰

POSSIBLE EXPLANATIONS, BUT NO UNIFYING MECHANISM

If mechanisms could be identified to explain the association between statins and diabetes, this would strengthen the argument that it is a cause-and-effect relationship. Many explanations have been proposed as to how statins may influence glucose metabolism and insulin sensitivity.^31–34 These are possible explanations based on other observations.

In theory, statins may improve insulin sensitivity via their anti-inflammatory effect, since inflammatory markers and proinflammatory cytokines have been linked with insulin resistance. However, other effects of statins may adversely affect glycemic control.

In vivo analysis has shown that some but not all statins increase insulin levels and decrease insulin sensitivity in a dose-dependent fashion. Some statins decrease adiponectin and may worsen glycemic control through loss of adiponectin’s proposed protective anti-proliferative and antiangiogenic properties. In vitro studies and animal studies have demonstrated a decrease in expression of insulin-responsive glucose transporter 4 (GLUT4) with atorvastatin, and an increase in GLUT1. It has been hypothesized that reduction in isoprenoid biosynthesis or decreased insulin signaling may explain these effects and that changes in glucose transport in adipocytes may cause insulin resistance. Other studies suggest that dysregulation of cellular cholesterol may attenuate beta-cell function. Impaired biosynthesis of ubiquinones may result in delayed production of adenosine triphosphate and consequently diminish insulin release.

But different effects have been reported for atorvastatin, simvastatin, and pravastatin, arguing against a unifying explanation or, alternatively, suggesting that differences in lipophilicity and potency among statins are important. Hydrophilic statins may be less likely to be taken up by extrahepatic cells such as pancreatic cells and adipocytes, possibly lessening these effects. However, the strong association between rosuvastatin (which is hydrophilic) and new diabetes would not support this hypothesis.

Despite these speculations, lack of conformity in response to different statins and discrepancies in the clinical outcomes noted in trials fail to clearly identify a common causative mechanism.

OTHER COMMON THERAPIES MAY INFLUENCE GLYCEMIC CONTROL

Statins are not the first drugs for reducing cardiovascular risk that have been shown to affect glucose levels during treatment.

Niacin

Niacin has been known to increase glucose levels but has long been used as a treatment for dyslipidemia despite this caution. Reduced glycemic control during niacin treatment in diabetic patients does not seem to alter the beneficial effects of treatment.^35–37

In a post hoc analysis of the Coronary Drug Project (CDP), in patient subgroups defined by baseline fasting plasma glucose and compared with placebo, niacin reduced the 6-year risk of recurrent myocardial infarction and the combined end point of coronary heart disease death or nonfatal myocardial infarction similarly (interactive P value nonsignificant) across all levels of baseline fasting plasma glucose, including levels of 126 mg/dL or higher at study entry.³⁶

In another post hoc analysis of CDP patient subgroups defined by the change in glycemic status from baseline to 1 year, niacin reduced the 6-year risk of the same end points similarly (interactive P value nonsignificant) across all levels of change in fasting plasma glucose from baseline to year 1, whether baseline fasting plasma glucose levels decreased, stayed the same, or increased to 10 mg/dL or higher on niacin therapy.³⁶

Therefore, the beneficial effect of niacin of reducing the rate of recurrent nonfatal myocardial infarction and coronary heart disease events was not significantly diminished when impaired fasting glucose or diabetes was present when therapy was started or by on-therapy increases from baseline fasting plasma glucose.

In addition, on-therapy changes in glycemic control may be dose-related and minimized by surveillance and therapy adjustments. The Assessment of Diabetes Control and Evaluation of the Efficacy of Niaspan Trial (ADVENT)³⁸ found that changes in glycemic control were minimal as measured by fasting glucose and hemoglobin A_1c; were associated with a higher niacin dose (1.5 g/day vs 1 g/day); and, when present, were successfully managed by adjusting the diabetes treatment regimen.

Antihypertensive drugs

Diuretics as well as beta-blockers have been reported to increase the incidence of diabetes in patients with hypertension.^15,38–40

A retrospective longitudinal cohort study⁴⁰ in 2009 examined the development of new-onset diabetes (defined as a new ICD-9 code for diabetes or initiation of diabetes treatment) in 24,688 treated hypertensive patients without diabetes at baseline; 4,385 (17.8%) of the patients developed diabetes. After adjusting for sex and age, the risk of new diabetes was significant in users of diuretics (OR 1.10), beta-blockers (OR 1.12), and calcium channel blockers (OR 1.10) compared with users of angiotensin-converting enzyme inhibitors, (OR 0.92), angiotensin receptor blockers (OR 0.90), or alpha-blockers (OR 0.88).

However, the increase in blood glucose does not seem to attenuate the beneficial effects of reducing cardiovascular events. In the Antihypertensive and Lipid-lowering Treatment to Prevent Heart Attack Trial (ALLHAT),¹⁵ a long-term follow-up of those developing new-onset diabetes while taking chlorthalidone (Hygroton) found no difference in the risk of death from cardiovascular disease or from any cause (hazard ratio = 1.04).¹⁵

CLINICAL IMPLICATIONS

Balancing the benefits and risks of statins

It is important to examine how the 0.4% increase in absolute risk of new-onset diabetes as calculated in meta-analyses compares with the benefits of statin treatment in terms of cardiovascular risk reduction.

Using data from the Cholesterol Treatment Trialists (CTT) meta-analysis of statin trials in 71,370 participants, Sattar et al¹¹ estimated that statin treatment is associated with 5.4 fewer deaths from coronary heart disease and cases of nonfatal myocardial infarction per 255 patients treated over 4 years for each 1-mmol/L (39 mg/dL) reduction in LDL-C compared with controls. The benefit would be even greater if stroke, revascularization, and hospitalization are included as end points. This benefit is contrasted with the risk of developing 1 additional case of diabetes for every 255 patients treated with statins over the same period.

Preiss et al²⁵ calculated that there were 2 more cases of diabetes per 1,000 patient-years in patients receiving intensive doses than in those receiving moderate doses (18.9 vs 16.9), corresponding to 1 additional case of diabetes for every 498 patients treated per year. However, there were 6.5 fewer first major cardiovascular events per 1,000 patient-years (44.5 vs 51.0), corresponding to a number needed to treat per year to prevent 1 cardiovascular event of 155. Most of the benefit was due to fewer revascularizations, followed by nonfatal myocardial infarctions. The 12% increase in new diabetes with high-dose therapy contrasted with a 16% reduction in new cardiovascular disease combined events (OR 0.84, 95% CI 0.75–0.94).

As previously noted, in the JUPITER trial, the benefits of preventing cardiovascular events with statin therapy outweighed the risk of new diabetes in people both with and without baseline risk factors for diabetes.²⁹ Similar to the observations with niacin and some antihypertensive drugs, the increase in blood glucose with statins does not appear to reduce the benefits of cardiovascular risk reduction in these patients at moderate to high risk, even when used at high doses.

Diabetes is a coronary heart disease risk equivalent and is associated with high risk of cardiovascular events.^41–46 Overall, the risk for these adverse events is two to four times greater in people with diabetes than without. Atherosclerosis-related events account for approximately 65% to 75% of all deaths in people with diabetes, and 75% of these events are coronary. Lipid abnormalities are strongly correlated with the risk of cardiovascular disease in people with diabetes, and aggressive treatment of risk factors, particularly lipid abnormalities, has been shown to reduce this risk.^47–49 And data from multiple clinical trials support the use of statins to lower LDL-C as the first-line therapy for dyslipidemia in people with diabetes, just as it is in the general population.^{3–7,9,13,23,50–61}

Analyses of diabetic subgroups encompassing 18,000 to 20,000 patients in the large statin trials have clearly demonstrated the benefits of statin therapy. A recent metaanalysis of 10 placebo-controlled trials that included approximately 16,000 patients with diabetes and 54,000 without diabetes demonstrated a 30% reduction in coronary heart disease, a 19% reduction in strokes, and a 12% reduction in mortality.⁵⁴ Furthermore, in another meta-analysis of 14 trials, a similar 22% reduction in coronary heart disease was noted in people with diabetes whether or not they had a history of cardiovascular disease.⁵⁵

Therefore, aggressive treatment of lipid abnormalities with statins as primary treatment has generally been adopted as a standard of care in diabetic patients, particularly those with clinical cardiovascular disease or one or more risk factors. The Adult Treatment Panel III guidelines recommend a minimum LDL-C goal of less than 100 mg/dL and a goal of less than 70 mg/dL as an option for patients with diabetes (Table 1).^41,62 Similar recommendations have been issued by the American Diabetes Association together with the American College of Cardiology (Table 2),³⁰ the American Diabetes Association by itself,⁶³ and the American Academy of Pediatrics.⁶

Is new-onset diabetes as dangerous as established diabetes?

In studies to date, there did not appear to be more events in those who developed new-onset diabetes.

Waters et al,²⁴ evaluating three trials of high-dose atorvastatin therapy, found that major cardiovascular events occurred in 11.3% of those with new-onset diabetes, 10.8% of those without new-onset diabetes (HR 1.02, 95% CI 0.77–1.35), and 17.5% of those who had diabetes at baseline.

Therefore, it may not be appropriate to extrapolate the glucose changes seen on statin therapy to an equivalent increase in adverse cardiovascular events as seen in other diabetic patients. The beneficial reduction in cardiovascular events does not appear to be diminished in those developing diabetes. It is not clear that the increase in glucose on statins has the same implications of a new diagnosis of diabetes. Does this elevation in glucose represent true diabetes or some downstream effect? For example, thiazide diuretics have been known to increase blood glucose levels, but the levels drop when these drugs are discontinued, even after many years of treatment.

On the other hand, it is possible that follow-up of 5 years or less in clinical trials has not allowed sufficient time to examine the influence of the increase in new-onset diabetes on future cardiovascular events. In addition, because of the widespread use of statins across a broad range of cardiovascular risk, even if the effect is small in absolute terms, the potential adverse effects are magnified, particularly in a low-risk population in which the cardiovascular benefits are smaller.

The association is real, but questions remain

In view of the evidence, it is difficult to refute that an association exits between statin use and new-onset diabetes, at least in some subgroups. The dose response noted in some studies further reinforces the conclusion that the association is real. However, many questions remain unanswered regarding mechanism of effect, whether there are differences depending on the particular statin or dose used, or differential effects in the populations treated (such as patients with metabolic syndrome or the elderly).

Until the contradictory observations can be resolved and plausible mechanisms of action elucidated, causality cannot be established. From a clinical standpoint there is no current evidence suggesting that the elevations in blood glucose seen while on lipid-lowering or blood-pressure-lowering therapy are associated with an increased risk of cardiovascular events or that they attenuate the beneficial effects of the therapy.

Statins should continue to be used in patients at high risk

Until further studies are done, statins should continue to be used, after assessing the risks and the benefits.

Primary prevention patients at moderate to high risk and secondary prevention patients stand to gain from statin therapy, and it should not be denied or doses reduced on the basis of concerns about the development of new-onset diabetes. The recognized modest risk of developing diabetes does not appear to blunt the cardioprotective effects of statin therapy in these moderate-to high-risk groups.

Rather than stop statins in patients at risk of diabetes such as the elderly or those with prediabetes, insulin resistance, or metabolic syndrome who are on therapy for appropriate reasons, it is reasonable to continue these drugs, monitoring glucose more closely and emphasizing the importance of weight reduction, diet, and aerobic exercise for preventing diabetes. The Diabetes Prevention Program Research Group, for example, reduced the incidence of diabetes by 58% over 2.8 years of follow-up with intensive lifestyle interventions (a low-calorie, low-fat diet plus moderate physical activity 150 minutes per week) vs usual care in at-risk populations.⁶⁵

Should statins be used more cautiously in patients at lower risk?

The most recent Cholesterol Treatment Trialists meta-analysis of 27 randomized clinical trials (22 placebo-controlled, 134,537 people; 5 high-dose vs low-dose, 39,612 people) reported that reducing LDL-C with statins lowered cardiovascular risk even in low-risk patients.⁶⁶ Overall, there were 21% fewer major cardiovascular events (coronary heart disease, stroke, or coronary revascularization) for every 1-mmol/L reduction in LDL-C.

The proportional reduction in events was at least as large in the two lowest-risk groups (estimated 5-year risk of < 5% and 5% to < 10%, 53,152 people) as in the higher-risk groups. This was reflected mainly in fewer nonfatal myocardial infarctions and coronary revascularizations. In these groups, the absolute reduction in risk for each 1-mmol/L reduction in LDL-C was 11 per 1,000 patients over 5 years. Even in this low-risk population, the reduction in cardiovascular risk seems to compare favorably with the small estimated increase risk of diabetes.

However, even in the lowest-risk group studied, the average baseline LDL-C level was greater than 130 mg/dL.

Therefore, in groups in which the benefits of statins on cardiovascular risk reduction are less robust (eg, low-risk primary prevention groups without significant elevations in LDLC, particularly the elderly), it would not be difficult to justify the case for more cautious use of statin therapy. If statins are used in these low-risk groups, restricting their use to those with at least moderate LDL-C elevation, using less aggressive LDL-C-lowering targets, and regular monitoring of fasting glucose seem reasonable until further information is available.

On february 28, 2012, the US Food and Drug Administration (FDA) updated its labeling requirements for statins. In addition to revising its recommendations for monitoring liver function and its alerts about reports of memory loss, the FDA also warned of the possibility of new-onset diabetes mellitus and worse glycemic control in patients taking statin drugs.¹

This change stoked an ongoing debate about the risk of diabetes with statin use and the implications of such an effect. To understand the clinical consequences of this alert and its effect on treatment decisions, we need to consider the degree to which statins lower the risk of cardiovascular disease in patients at high risk (including diabetic patients), the magnitude of the risk of developing new diabetes while on statin therapy, and the ratio of risk to benefit in treated populations.

This review will discuss the evidence for this possible adverse effect and the implications for clinical practice.

DO STATINS CAUSE DIABETES?

Individual controlled trials dating back more than a decade have had conflicting results about new diabetes and poorer diabetic control in patients taking statins.

The West of Scotland Coronary Prevention Study (WOSCOPS)² suggested that the incidence of diabetes was 30% lower in patients taking pravastatin (Pravachol) 40 mg/day than with placebo. However, this was not observed with atorvastatin (Lipitor) 10 mg/day in the Anglo-Scandinavian Cardiac Outcomes Trial–Lipid-Lowering Arm (ASCOT-LLA)³ in hypertensive patients or in the Collaborative Atorvastatin Diabetes Study (CARDS)⁴ in diabetic patients,⁴ nor was it noted with simvastatin (Zocor) 40 mg/day in the Heart Protection Study (HPS).⁵

The Justification for the Use of Statins in Primary Prevention: An Intervention Trial Evaluating Rosuvastatin (JUPITER),⁶ using the more potent agent rosuvastatin (Crestor) 20 mg/day in patients with elevated levels of C-reactive protein (CRP), was stopped early when an interim analysis found a 44% lower incidence of the primary end point. However, the trial also reported a 26% higher incidence of diabetes in follow-up of less than 2 years.

In the Prospective Study of Pravastatin in the Elderly at Risk (PROSPER),⁷ with a mean age at entry of 75, there was a 32% higher incidence of diabetes with pravastatin therapy.⁷

Results of meta-analyses

Several meta-analyses have addressed these differences.

Rajpathak et al⁸ performed a meta-analysis, published in 2009, of six trials—WOSCOPS,² ASCOT-LLA,³ JUPITER,⁶ HPS,⁵ the Long-term Intervention With Pravastatin in Ischaemic Disease (LIPID) study,⁹ and the Controlled Rosuvastatin Multinational Study in Heart Failure (CORONA),¹⁰ with a total of 57,593 patients. They calculated that the incidence of diabetes was 13% higher (an absolute difference of 0.5%) in statin recipients, which was statistically significant. In their initial analysis, the authors excluded WOSCOPS, describing it as hypothesis-generating. The relative increase in risk was less—6%—and was not statistically significant when WOSCOPS was included.

Sattar et al,¹¹ in a larger meta-analysis published in 2010, included 91,140 participants in 13 major statin trials conducted between 1994 and 2009; each trial had more than 1,000 patients and more than 1 year of follow-up.^{2,3,5–7,9,10,12–17} New diabetes was defined as physician reporting of new diabetes, new diabetic medication use, or a fasting glucose greater than 7 mmol/L (126 mg/dL).

New diabetes occurred in 2,226 (4.89%) of the statin recipients and in 2,052 (4.5%) of the placebo recipients, an absolute difference of 0.39%, or 9% more (odds ratio [OR] 1.09; 95% confidence interval [CI] 1.02–1.17) (Figure 1).

The incidence of diabetes varied substantially among the 13 trials, with only JUPITER⁶ and PROSPER⁷ finding statistically significant increases in rates (26% and 32%, respectively). Of the other 11 trials, 4 had nonsignificant trends toward lower incidence,^2,9,13,17 while the 7 others had nonsignificant trends toward higher incidence.

Does the specific statin make a difference?

Questions have been raised as to whether the type of statin used, the intensity of therapy, or the population studied contributed to these differences. Various studies suggest that factors such as using hydrophilic vs lipophilic statins (hydrophilic statins include pravastatin and rosuvastatin; lipophilic statins include atorvastatin, lovastatin, and simvastatin), the dose, the extent of lowering of low-density lipoprotein cholesterol (LDL-C), and the age or clinical characteristics of the population studied may influence this relationship.^18–20

Yamakawa et al¹⁸ examined the effect of atorvastatin 10 mg/day, pravastatin 10 mg/day, and pitavastatin (Livalo) 2 mg/day on glycemic control over 3 months in a retrospective analysis. Random blood glucose and hemoglobin A_1c levels were increased in the atorvastatin group but not in the other two.¹⁸

A prospective comparison of atorvastatin 20 mg vs pitavastatin 4 mg in patients with type 2 diabetes, presented at the American College of Cardiology’s 2011 annual meeting, reported a significant increase in fasting glucose levels with atorvastatin, particularly in women, but not with pitavastatin.¹⁹

In the Compare the Effect of Rosuvastatin With Atorvastatin on Apo B/Apo A-1 Ratio in Patients With Type 2 Diabetes Mellitus and Dyslipidaemia (CORALL) study,²⁰ both high-dose rosuvastatin (40 mg) and high-dose atorvastatin (80 mg) were associated with significant increases in hemoglobin A_1c, although the mean fasting glucose levels were not significantly different at 18 weeks of therapy.

A meta-analysis by Sattar et al¹¹ did not find a clear difference between lipophilic statins (OR 1.10 vs placebo) and hydrophilic statins (OR 1.08). In analysis by statin type, the combined rosuvastatin trials were statistically significant in favor of a higher diabetes risk (OR 1.18, 95% CI 1.04–1.44). Nonsignificant trends were noted for atorvastatin trials (OR 1.14) and simvastatin trials (OR 1.11) and less so for pravastatin (OR 1.03); the OR for lovastatin was 0.98. This may suggest that there is a stronger effect with more potent statins or with greater lowering of LDL-C.

Meta-regression analysis in this study demonstrated that diabetes risk with statins was higher in older patients but was not influenced by body mass index or by the extent that LDL-C was lowered.

Intensive-dose statin therapy has been shown to reduce cardiovascular risk more than low-dose or moderate-dose therapy, thus supporting more aggressive treatment of LDL-C in higher-risk patients. However, some controlled studies comparing more-potent with less-potent statin regimens suggest that there may also be a higher risk of incident diabetes at higher doses.^21–24

In a post hoc analysis of the Pravastatin or Atorvastatin Evaluation and Infection Therapy– Thrombolysis in Myocardial Infarction 22 (PROVE-IT TIMI 22) trial,²¹ patients who had experienced an acute coronary syndrome had a greater increase in hemoglobin A_1c if treated with atorvastatin 80 mg/day than with pravastatin 40 mg/day.

Waters et al²³ reported a higher risk of new diabetes with atorvastatin 80 mg than with placebo and a trend toward a higher risk with atorvastatin 80 mg than with atorvastatin 10 mg or simvastatin 20 mg.

In contrast, a review by Yousef et al²⁴ of the data from the Enhanced Feedback for Effective Cardiac Treatment (EFFECT) study did not find a higher diabetes risk with more intensive statin therapy based on the magnitude of LDL-C reduction. A propensity-matched examination of deaths, recurrent acute ischemic events, or new diabetes in patients previously hospitalized with myocardial infarction found no differences in these end points each year out to 5 years. The risk of diabetes was in fact lower (but the difference was not statistically significant) in the high-dose groups out to 5 years. The risk of myocardial infarction or death was numerically different in the high-dose groups, but the difference was not statistically significant.

Preiss et al²⁵ in 2011 performed a meta-analysis of the impact of intensity of statin therapy on diabetes risk. They examined data from 32,752 participants without diabetes at baseline in five randomized controlled trials with more than 1,000 participants and more than 1 year of follow-up, comparing high-dose therapy against moderate-dose statin therapy.^{21,22,26–28} New diabetes was considered present if there was an adverse event report of diabetes, if glucose-lowering drugs were started, or if two fasting plasma glucose measurements were higher than 7 mmol/L (126 mg/dL).

Diabetes developed in 1,449 (8.8%) of the intensive-therapy group and 1,300 (8.0%) of the moderate-therapy group (OR 1.12, 95% CI 1.04–1.22). In contrast, incident cardiovascular disease occurred in 3,134 (19.1%) of the intensive-therapy group and 3,550 (21.7%) of the moderate-therapy group (OR 0.84, 95% CI 0.75–0.94). Therefore, there was an 0.8% absolute increase in diabetes cases on high-dose statins and a 2.6% absolute reduction in adverse cardiovascular events.

CAUTION IN INTERPRETING THESE DATA

There are many reasons for caution in interpreting these studies.

The trials were not designed to look for diabetes

The data supporting the relationship between statin therapy and higher risk of diabetes are primarily from observational studies. These studies were not prospectively designed to address this question, and we therefore need to view this as association and not as causation.

The definition of diabetes varied between trials, and new-onset diabetes was often not rigorously screened for. In many trials the outcome of diabetes was at least partially based on nonstandardized, nonadjudicated physician reporting.

Consequently, if statins reduce the risk of diabetes, the results from WOSCOPS may overstate the reduction, since this study used a non-standard definition of incident diabetes (fasting plasma glucose > 126 mg/dL plus a > 36 mg/dL increase from baseline). When Sattar et al¹¹ reanalyzed WOSCOPS data using a more standard definition, they found a smaller effect.

On the other hand, nonstandardized physician reporting may overstate an adverse effect. Sattar et al¹¹ also found that when fasting plasma glucose levels alone were used as the definition for diabetes, the overall risk was attenuated and was no longer statistically significant (OR 1.07, 95% CI 0.97–1.17).

Perhaps statin therapy uncovers diabetes only in people at risk of diabetes

Perhaps statin therapy uncovers diabetes only in people at higher baseline risk of developing diabetes. Therefore, this adverse effect may be restricted to certain groups and not applicable to the general population.

In JUPITER, one of the two trials in which, on independent analysis, statin use was associated with new diabetes, 77% of patients in the rosuvastatin group who developed diabetes had impaired fasting glucose at entry and therefore were at higher risk of developing diabetes.⁶

Possibly, the relationship is driven by preexisting metabolic syndrome or other risk factors for diabetes. In the two studies that reported a statistically significantly higher incidence of new diabetes, more than 40% of patients in JUPITER met the criteria for metabolic syndrome, and metabolic syndrome, which increases in prevalence with age, was likely more prominent in the elderly population in PROSPER.

Waters et al²³ grouped patients according to whether they had risk factors for diabetes (impaired fasting glucose, obesity, elevated triglycerides, and hypertension) and found that those who had none or one of these risk factors had no difference in the rate of new-onset diabetes with either moderate or intensive statin therapy, but the risk was pronounced in those who had three or four risk factors.

Ridker et al²⁹ reanalyzed the JUPITER data from patients who did not have cardiovascular disease at baseline. Overall, for every 54 new cases of diabetes in follow-up, 134 cardiovascular events or deaths were prevented. In subgroup analysis, those who had one or more risk factors for diabetes at baseline (metabolic syndrome, impaired fasting glucose, obesity, or hemoglobin A_1c > 6%) had a 39% reduction in the primary end point and a 28% increase in new diabetes. Those who had none of these risk factors had a 52% lower rate of cardiovascular events but no increase in diabetes.

Other confounding factors

Bias and confounding factors are difficult to control for in studies without prospectively defined, recognized, and analyzed outcomes.

Although it may be a bit of a stretch, residual confounding factors such as myalgia side effects while on statins may reduce exercise in the statin-treatment groups. Perhaps a change to a healthier lifestyle after cardiovascular events may be more common in placebo groups. Improved survival with statins may allow more people at risk of diabetes to live longer and present with the diagnosis.³⁰

POSSIBLE EXPLANATIONS, BUT NO UNIFYING MECHANISM

If mechanisms could be identified to explain the association between statins and diabetes, this would strengthen the argument that it is a cause-and-effect relationship. Many explanations have been proposed as to how statins may influence glucose metabolism and insulin sensitivity.^31–34 These are possible explanations based on other observations.

In theory, statins may improve insulin sensitivity via their anti-inflammatory effect, since inflammatory markers and proinflammatory cytokines have been linked with insulin resistance. However, other effects of statins may adversely affect glycemic control.

In vivo analysis has shown that some but not all statins increase insulin levels and decrease insulin sensitivity in a dose-dependent fashion. Some statins decrease adiponectin and may worsen glycemic control through loss of adiponectin’s proposed protective anti-proliferative and antiangiogenic properties. In vitro studies and animal studies have demonstrated a decrease in expression of insulin-responsive glucose transporter 4 (GLUT4) with atorvastatin, and an increase in GLUT1. It has been hypothesized that reduction in isoprenoid biosynthesis or decreased insulin signaling may explain these effects and that changes in glucose transport in adipocytes may cause insulin resistance. Other studies suggest that dysregulation of cellular cholesterol may attenuate beta-cell function. Impaired biosynthesis of ubiquinones may result in delayed production of adenosine triphosphate and consequently diminish insulin release.

But different effects have been reported for atorvastatin, simvastatin, and pravastatin, arguing against a unifying explanation or, alternatively, suggesting that differences in lipophilicity and potency among statins are important. Hydrophilic statins may be less likely to be taken up by extrahepatic cells such as pancreatic cells and adipocytes, possibly lessening these effects. However, the strong association between rosuvastatin (which is hydrophilic) and new diabetes would not support this hypothesis.

Despite these speculations, lack of conformity in response to different statins and discrepancies in the clinical outcomes noted in trials fail to clearly identify a common causative mechanism.

OTHER COMMON THERAPIES MAY INFLUENCE GLYCEMIC CONTROL

Statins are not the first drugs for reducing cardiovascular risk that have been shown to affect glucose levels during treatment.

Niacin

Niacin has been known to increase glucose levels but has long been used as a treatment for dyslipidemia despite this caution. Reduced glycemic control during niacin treatment in diabetic patients does not seem to alter the beneficial effects of treatment.^35–37

In a post hoc analysis of the Coronary Drug Project (CDP), in patient subgroups defined by baseline fasting plasma glucose and compared with placebo, niacin reduced the 6-year risk of recurrent myocardial infarction and the combined end point of coronary heart disease death or nonfatal myocardial infarction similarly (interactive P value nonsignificant) across all levels of baseline fasting plasma glucose, including levels of 126 mg/dL or higher at study entry.³⁶

In another post hoc analysis of CDP patient subgroups defined by the change in glycemic status from baseline to 1 year, niacin reduced the 6-year risk of the same end points similarly (interactive P value nonsignificant) across all levels of change in fasting plasma glucose from baseline to year 1, whether baseline fasting plasma glucose levels decreased, stayed the same, or increased to 10 mg/dL or higher on niacin therapy.³⁶

Therefore, the beneficial effect of niacin of reducing the rate of recurrent nonfatal myocardial infarction and coronary heart disease events was not significantly diminished when impaired fasting glucose or diabetes was present when therapy was started or by on-therapy increases from baseline fasting plasma glucose.

In addition, on-therapy changes in glycemic control may be dose-related and minimized by surveillance and therapy adjustments. The Assessment of Diabetes Control and Evaluation of the Efficacy of Niaspan Trial (ADVENT)³⁸ found that changes in glycemic control were minimal as measured by fasting glucose and hemoglobin A_1c; were associated with a higher niacin dose (1.5 g/day vs 1 g/day); and, when present, were successfully managed by adjusting the diabetes treatment regimen.

Antihypertensive drugs

Diuretics as well as beta-blockers have been reported to increase the incidence of diabetes in patients with hypertension.^15,38–40

A retrospective longitudinal cohort study⁴⁰ in 2009 examined the development of new-onset diabetes (defined as a new ICD-9 code for diabetes or initiation of diabetes treatment) in 24,688 treated hypertensive patients without diabetes at baseline; 4,385 (17.8%) of the patients developed diabetes. After adjusting for sex and age, the risk of new diabetes was significant in users of diuretics (OR 1.10), beta-blockers (OR 1.12), and calcium channel blockers (OR 1.10) compared with users of angiotensin-converting enzyme inhibitors, (OR 0.92), angiotensin receptor blockers (OR 0.90), or alpha-blockers (OR 0.88).

However, the increase in blood glucose does not seem to attenuate the beneficial effects of reducing cardiovascular events. In the Antihypertensive and Lipid-lowering Treatment to Prevent Heart Attack Trial (ALLHAT),¹⁵ a long-term follow-up of those developing new-onset diabetes while taking chlorthalidone (Hygroton) found no difference in the risk of death from cardiovascular disease or from any cause (hazard ratio = 1.04).¹⁵

CLINICAL IMPLICATIONS

Balancing the benefits and risks of statins

It is important to examine how the 0.4% increase in absolute risk of new-onset diabetes as calculated in meta-analyses compares with the benefits of statin treatment in terms of cardiovascular risk reduction.

Using data from the Cholesterol Treatment Trialists (CTT) meta-analysis of statin trials in 71,370 participants, Sattar et al¹¹ estimated that statin treatment is associated with 5.4 fewer deaths from coronary heart disease and cases of nonfatal myocardial infarction per 255 patients treated over 4 years for each 1-mmol/L (39 mg/dL) reduction in LDL-C compared with controls. The benefit would be even greater if stroke, revascularization, and hospitalization are included as end points. This benefit is contrasted with the risk of developing 1 additional case of diabetes for every 255 patients treated with statins over the same period.

Preiss et al²⁵ calculated that there were 2 more cases of diabetes per 1,000 patient-years in patients receiving intensive doses than in those receiving moderate doses (18.9 vs 16.9), corresponding to 1 additional case of diabetes for every 498 patients treated per year. However, there were 6.5 fewer first major cardiovascular events per 1,000 patient-years (44.5 vs 51.0), corresponding to a number needed to treat per year to prevent 1 cardiovascular event of 155. Most of the benefit was due to fewer revascularizations, followed by nonfatal myocardial infarctions. The 12% increase in new diabetes with high-dose therapy contrasted with a 16% reduction in new cardiovascular disease combined events (OR 0.84, 95% CI 0.75–0.94).

As previously noted, in the JUPITER trial, the benefits of preventing cardiovascular events with statin therapy outweighed the risk of new diabetes in people both with and without baseline risk factors for diabetes.²⁹ Similar to the observations with niacin and some antihypertensive drugs, the increase in blood glucose with statins does not appear to reduce the benefits of cardiovascular risk reduction in these patients at moderate to high risk, even when used at high doses.

Diabetes is a coronary heart disease risk equivalent and is associated with high risk of cardiovascular events.^41–46 Overall, the risk for these adverse events is two to four times greater in people with diabetes than without. Atherosclerosis-related events account for approximately 65% to 75% of all deaths in people with diabetes, and 75% of these events are coronary. Lipid abnormalities are strongly correlated with the risk of cardiovascular disease in people with diabetes, and aggressive treatment of risk factors, particularly lipid abnormalities, has been shown to reduce this risk.^47–49 And data from multiple clinical trials support the use of statins to lower LDL-C as the first-line therapy for dyslipidemia in people with diabetes, just as it is in the general population.^{3–7,9,13,23,50–61}

Analyses of diabetic subgroups encompassing 18,000 to 20,000 patients in the large statin trials have clearly demonstrated the benefits of statin therapy. A recent metaanalysis of 10 placebo-controlled trials that included approximately 16,000 patients with diabetes and 54,000 without diabetes demonstrated a 30% reduction in coronary heart disease, a 19% reduction in strokes, and a 12% reduction in mortality.⁵⁴ Furthermore, in another meta-analysis of 14 trials, a similar 22% reduction in coronary heart disease was noted in people with diabetes whether or not they had a history of cardiovascular disease.⁵⁵

Therefore, aggressive treatment of lipid abnormalities with statins as primary treatment has generally been adopted as a standard of care in diabetic patients, particularly those with clinical cardiovascular disease or one or more risk factors. The Adult Treatment Panel III guidelines recommend a minimum LDL-C goal of less than 100 mg/dL and a goal of less than 70 mg/dL as an option for patients with diabetes (Table 1).^41,62 Similar recommendations have been issued by the American Diabetes Association together with the American College of Cardiology (Table 2),³⁰ the American Diabetes Association by itself,⁶³ and the American Academy of Pediatrics.⁶

Is new-onset diabetes as dangerous as established diabetes?

In studies to date, there did not appear to be more events in those who developed new-onset diabetes.

Waters et al,²⁴ evaluating three trials of high-dose atorvastatin therapy, found that major cardiovascular events occurred in 11.3% of those with new-onset diabetes, 10.8% of those without new-onset diabetes (HR 1.02, 95% CI 0.77–1.35), and 17.5% of those who had diabetes at baseline.

Therefore, it may not be appropriate to extrapolate the glucose changes seen on statin therapy to an equivalent increase in adverse cardiovascular events as seen in other diabetic patients. The beneficial reduction in cardiovascular events does not appear to be diminished in those developing diabetes. It is not clear that the increase in glucose on statins has the same implications of a new diagnosis of diabetes. Does this elevation in glucose represent true diabetes or some downstream effect? For example, thiazide diuretics have been known to increase blood glucose levels, but the levels drop when these drugs are discontinued, even after many years of treatment.

On the other hand, it is possible that follow-up of 5 years or less in clinical trials has not allowed sufficient time to examine the influence of the increase in new-onset diabetes on future cardiovascular events. In addition, because of the widespread use of statins across a broad range of cardiovascular risk, even if the effect is small in absolute terms, the potential adverse effects are magnified, particularly in a low-risk population in which the cardiovascular benefits are smaller.

The association is real, but questions remain

In view of the evidence, it is difficult to refute that an association exits between statin use and new-onset diabetes, at least in some subgroups. The dose response noted in some studies further reinforces the conclusion that the association is real. However, many questions remain unanswered regarding mechanism of effect, whether there are differences depending on the particular statin or dose used, or differential effects in the populations treated (such as patients with metabolic syndrome or the elderly).

Until the contradictory observations can be resolved and plausible mechanisms of action elucidated, causality cannot be established. From a clinical standpoint there is no current evidence suggesting that the elevations in blood glucose seen while on lipid-lowering or blood-pressure-lowering therapy are associated with an increased risk of cardiovascular events or that they attenuate the beneficial effects of the therapy.

Statins should continue to be used in patients at high risk

Until further studies are done, statins should continue to be used, after assessing the risks and the benefits.

Primary prevention patients at moderate to high risk and secondary prevention patients stand to gain from statin therapy, and it should not be denied or doses reduced on the basis of concerns about the development of new-onset diabetes. The recognized modest risk of developing diabetes does not appear to blunt the cardioprotective effects of statin therapy in these moderate-to high-risk groups.

Rather than stop statins in patients at risk of diabetes such as the elderly or those with prediabetes, insulin resistance, or metabolic syndrome who are on therapy for appropriate reasons, it is reasonable to continue these drugs, monitoring glucose more closely and emphasizing the importance of weight reduction, diet, and aerobic exercise for preventing diabetes. The Diabetes Prevention Program Research Group, for example, reduced the incidence of diabetes by 58% over 2.8 years of follow-up with intensive lifestyle interventions (a low-calorie, low-fat diet plus moderate physical activity 150 minutes per week) vs usual care in at-risk populations.⁶⁵

Should statins be used more cautiously in patients at lower risk?

The most recent Cholesterol Treatment Trialists meta-analysis of 27 randomized clinical trials (22 placebo-controlled, 134,537 people; 5 high-dose vs low-dose, 39,612 people) reported that reducing LDL-C with statins lowered cardiovascular risk even in low-risk patients.⁶⁶ Overall, there were 21% fewer major cardiovascular events (coronary heart disease, stroke, or coronary revascularization) for every 1-mmol/L reduction in LDL-C.

The proportional reduction in events was at least as large in the two lowest-risk groups (estimated 5-year risk of < 5% and 5% to < 10%, 53,152 people) as in the higher-risk groups. This was reflected mainly in fewer nonfatal myocardial infarctions and coronary revascularizations. In these groups, the absolute reduction in risk for each 1-mmol/L reduction in LDL-C was 11 per 1,000 patients over 5 years. Even in this low-risk population, the reduction in cardiovascular risk seems to compare favorably with the small estimated increase risk of diabetes.

However, even in the lowest-risk group studied, the average baseline LDL-C level was greater than 130 mg/dL.

Therefore, in groups in which the benefits of statins on cardiovascular risk reduction are less robust (eg, low-risk primary prevention groups without significant elevations in LDLC, particularly the elderly), it would not be difficult to justify the case for more cautious use of statin therapy. If statins are used in these low-risk groups, restricting their use to those with at least moderate LDL-C elevation, using less aggressive LDL-C-lowering targets, and regular monitoring of fasting glucose seem reasonable until further information is available.

Sessile serrated polyps are a type of polyp recently recognized to be a precursor of colorectal cancer. They arise from a pathway of genetic alterations different from the pathway that causes the more common and well-understood conventional adenomas (also called tubular adenomas, tubulovillous adenomas, and villous adenomas).

We do not yet know enough about the lifetime colorectal cancer risk for individuals with sessile serrated polyps, nor do we know the optimal surveillance interval for patients who have these polyps on colonoscopy. It is believed that sessile serrated polyps may be the cause of a substantial number of “interval” colorectal cancers—ie, cancers that occur after colonoscopy but before the next scheduled examination.

Serrated polyps get their name from their jagged appearance on microscopy. In the past, all serrated colorectal lesions were called hyperplastic polyps. But with the advent of molecular and genetic diagnostics and with the ability to recognize the subtle morphologic differences of serrated lesions, they have been reclassified into those without malignant potential (hyperplastic polyps) and those that are neoplastic (sessile serrated polyps and traditional serrated adenomas) (Table 1).

In this article, we discuss the evolving understanding of the different types of serrated polyps, and we offer our thoughts on a reasonable postpolypectomy surveillance plan in patients with these lesions. We focus on sessile serrated polyps, the most common form of serrated polyp with cancerous potential, since it may be one of our greatest challenges in optimal colorectal cancer prevention.

CLINICAL SCENARIO

A 65-year-old woman with no family history of colorectal cancer undergoes screening colonoscopy, during which three polyps are found and removed—a 3-mm tubular adenoma in the sigmoid colon, an 8-mm sessile serrated polyp at the hepatic flexure, and a 2-mm hyperplastic polyp in the rectum. When should she undergo follow-up colonoscopy?

Based on the number, size, and pathologic makeup of the polyps in this patient, we would recommend follow-up surveillance colonoscopy in 5 years.

THE SERRATED POLYP PATHWAY: A DIFFERENT PATH TO COLORECTAL CANCER

Colorectal cancer is the third most common cancer in the United States.¹ From 70% to 80% of these cancers arise from adenomatous polyps via the adenoma-carcinoma pathway. This molecular pathway develops through chromosomal instability (CIN) and involves the loss of heterozygosity (the loss of function of one allele). This leads to the progressive accumulation of mutations in tumor-suppressor genes such as adenomatous polyposis coli (APC) and p53, and oncogenes such as KRAS. The result of these mutations is the development of adenomatous polyps that lead to microsatellite-stable colorectal cancers (Figure 1).²

More recently, studies have shown that the other 20% to 30% of colorectal cancers likely arise through a separate pathway, called the serrated polyp pathway or serrated neoplasia pathway. In contrast to CIN, this pathway is characterized by methylation of CpG islands (CIMP–CpG island methylation phenotype, CIMP) in the promoter regions of specific genes.³ Central to the serrated polyp pathway is progressive methylation in colonic mucosa; mutation in the BRAF oncogene, activating cell proliferation leading to a sessile serrated polyp; and epigenetic silencing of the DNA mismatch repair gene hMLH1, which is a key step in the progression to a sessile serrated polyp with dysplasia, which may rapidly become a microsatellite-unstable colorectal cancer.⁴

Histologically, serrated polyps have a serrated or sawtooth appearance from the folding in of the crypt epithelium, and they include hyperplastic polyps, traditional serrated adenomas, and sessile serrated polyps (sessile serrated adenomas).

Sessile serrated polyps and traditional serrated adenomas (which are rare) are thought to be precancerous, whereas hyperplastic polyps do not have malignant potential.

COMMON, BUT PREVALENCE IS NOT CLEARLY ESTABLISHED

The histologic criteria for sessile serrated polyps and traditional serrated adenomas have been elucidated,^4–7 but the epidemiology of these serrated polyps is not clear. Small studies have shown that sessile serrated polyps account for 2% to 9% of all polyps removed at colonoscopy^8–10; however, larger studies are needed to determine the prevalence because detection by an endoscopist and pathologic diagnosis of these polyps are both operator-dependent.

Traditional serrated adenomas are the least common type of serrated polyp, with a reported prevalence of 0.3%.⁷ Hyperplastic polyps are by far the most common, accounting for 20% to 30% of all polyps removed at colonoscopy.^9,11 Sessile serrated polyps have a predilection for the proximal colon and are associated with female sex and with smoking, ^12,13 but no consistent effect of other factors on their formation has been reported. In contrast, Wallace et al¹³ found that obesity, cigarette smoking, dietary fat intake, total caloric intake, and the consumption of red meat were associated with an increased risk of distal (but not proximal) serrated polyps, including hyperplastic polyps, sessile serrated polyps, and traditional serrated adenomas.

HYPERPLASTIC POLYPS

Hyperplastic polyps usually occur in the rectosigmoid colon. They appear as slightly elevated, whitish lesions with a diameter less than 5 mm (Figure 2). Microscopically, the serrated architecture is present in the upper half of their crypts (Figure 3). The proliferative zone is more or less normally located in the basal half of the crypt (the nonserrated portion), with nuclei that are small, uniform, and basally located.¹⁴ The bases of the crypts have a rounded contour and do not grow laterally along the muscularis mucosae.

SESSILE SERRATED POLYPS

Endoscopically, sessile serrated polyps are often subtle, appear flat or slightly elevated, and can be covered by yellow mucus (Figure 4). They are typically found in the proximal colon and are usually larger than typical adenomas, with 50% being larger than 10 mm.¹⁰

Histologically, the serrations are more prominent than those of hyperplastic polyps and involve the entire length of the crypt (Figure 5). The crypt bases are often dilated and display lateral growth along the lamina muscularis mucosae, resembling a letter t or l. The lamina muscularis mucosae is often thinner than normal. Crypts from sessile serrated polyps are occasionally found beneath the muscularis mucosae, a condition called pseudoinvasion.⁷

TRADITIONAL SERRATED ADENOMAS

Traditional serrated adenomas are usually left-sided. In contrast to the other types of serrated polyps, they are histologically often villiform and are lined by cells with elongated nuclei and abundant eosinophilic cytoplasm (Figure 6). Unlike those in sessile serrated polyps, the crypt bases do not display an abnormal architecture; rather, traditional serrated adenomas have abundant ectopic crypts (“budding crypts”) in the long, slender villi.⁷

Traditional serrated adenomas also appear to be genetically distinct from sessile serrated polyps. They are most often characterized by a KRAS (or less commonly, BRAF) mutation and commonly have methylation of the DNA repair gene MGMT (O-6-methylguanine-DNA methyltransferase) rather than hMLH1.

CHALLENGES TO EFFECTIVE COLONOSCOPY

Colonoscopic polypectomy of adenomatous polyps reduces the incidence of colorectal cancer and the rate of death from it.^15,16 However, recent data show that colonoscopy may not be as effective as once thought. As many as 9% of patients with colorectal cancer have had a “normal” colonoscopic examination in the preceding 3 years.^17,18 In addition, the reduction in incidence and mortality rates was less for cancers in the proximal colon than for cancers in the distal colon.^19,20

Possible explanations for this discrepancy include the skill of the endoscopist, technical limitations of the examination, incomplete removal of polyps, and inadequate bowel preparation. Several studies have shown that interval colorectal cancers are more likely to be found in the proximal colon and to have the same molecular characteristics as sessile serrated polyps and the serrated colorectal cancer pathway (CIMP-high and MSI-H).^21,22 Therefore, it is now thought that sessile serrated polyps may account for a substantial portion of “postcolonoscopy cancers” (ie, interval cancers) that arise in the proximal colon.

Two large studies of screening colonoscopy confirmed that the ability to detect sessile serrated polyps depends greatly on the skill of the endoscopist. Hetzel et al⁹ studied the differences in the rates of polyp detection among endoscopists performing more than 7,000 colonoscopies. Detection rates varied significantly for adenomas, hyperplastic polyps, and sessile serrated polyps, with the greatest variability noted in the detection of sessile serrated polyps. Significant variability was also noted in the ability of the pathologist to diagnose sessile serrated polyps.⁹

In the other study, a strong correlation was found between physicians who are “high detectors” of adenomas and their detection rates for proximal serrated polyps.²³ There is widespread acceptance that screening colonoscopy in average-risk patients age 50 and older should detect adenomas in more than 25% of men and more than 15% of women. There is no current minimum recommended detection rate for sessile serrated polyps, but some have suggested 1.5%.⁸

POLYPS AS PREDICTORS OF CANCER RISK

Certain polyp characteristics predict the risk of metachronous, advanced neoplasia. Advanced neoplasms are defined as invasive carcinomas, adenomas 10 mm or larger, or adenomas with any villous histology or high-grade dysplasia. Patients with one or two small tubular adenomas have a much lower risk of metachronous advanced neoplasia than do patients with more than two adenomas or advanced neoplasms.²⁴ Current recommended surveillance intervals vary on that basis (Table 2).²⁵

People who harbor serrated neoplasms are at high risk of synchronous serrated polyps and advanced adenomatous neoplasia. Pai et al²⁶ found that patients with one sessile serrated polyp were four times more likely to have additional serrated polyps at the same time than an unselected population. The authors suggested that this indicates a strong colonic mucosal-field defect in patients with sessile serrated polyps, thereby predisposing them to the development of synchronous serrated polyps.

Li et al²⁷ found that large serrated polyps (ie, > 10 mm) are associated with a risk of synchronous advanced neoplasia that is three times higher than in patients without adenomas. Schreiner et al²⁸ determined that patients with either a proximal or a large serrated polyp were at higher risk of synchronous advanced neoplasia compared with patients who did not have those lesions. Vu et al²⁹ found that patients who have both sessile serrated polyps and conventional adenomas have significantly larger and more numerous lesions of both types.²⁹ In addition, these lesions are more likely to be pathologically advanced when compared with people with only one or the other type.

In the only study of the risk of advanced neoplasia on follow-up colonoscopy,²⁸ patients with advanced neoplasia and proximal serrated polyps at baseline examination were twice as likely to have advanced neoplasia during subsequent surveillance than those with only advanced neoplasia at baseline examination.²⁸

Therefore, it seems clear that the presence of large or proximal serrated polyps or serrated neoplasms predicts the presence of synchronous and likely metachronous advanced neoplasms.

Guidelines for postpolypectomy surveillance for individuals with serrated lesions of the colon have recently been published.²⁵ Patients with large serrated lesions (≥ 10 mm) or an advanced serrated lesion (a sessile serrated polyp with or without cytologic dysplasia or a traditional serrated adenoma) should be followed closely. Patients with small (< 10-mm) rectosigmoid hyperplastic polyps should be followed as average-risk patients. If a patient with a sessile serrated polyp also has adenomas, the surveillance interval should be the shortest interval recommended for either lesion.²⁹

SURVEILLANCE FOR OUR PATIENT

In our patient, given the number, size, and histologic features of the polyps found, surveillance colonoscopy should be considered in 5 years. Although the clinical significance of the serrated pathway to colorectal cancer cannot be argued, further study is required to understand the lifetime risk to patients with serrated neoplasms and the optimal surveillance interval.

Sessile serrated polyps are a type of polyp recently recognized to be a precursor of colorectal cancer. They arise from a pathway of genetic alterations different from the pathway that causes the more common and well-understood conventional adenomas (also called tubular adenomas, tubulovillous adenomas, and villous adenomas).

We do not yet know enough about the lifetime colorectal cancer risk for individuals with sessile serrated polyps, nor do we know the optimal surveillance interval for patients who have these polyps on colonoscopy. It is believed that sessile serrated polyps may be the cause of a substantial number of “interval” colorectal cancers—ie, cancers that occur after colonoscopy but before the next scheduled examination.

Serrated polyps get their name from their jagged appearance on microscopy. In the past, all serrated colorectal lesions were called hyperplastic polyps. But with the advent of molecular and genetic diagnostics and with the ability to recognize the subtle morphologic differences of serrated lesions, they have been reclassified into those without malignant potential (hyperplastic polyps) and those that are neoplastic (sessile serrated polyps and traditional serrated adenomas) (Table 1).

In this article, we discuss the evolving understanding of the different types of serrated polyps, and we offer our thoughts on a reasonable postpolypectomy surveillance plan in patients with these lesions. We focus on sessile serrated polyps, the most common form of serrated polyp with cancerous potential, since it may be one of our greatest challenges in optimal colorectal cancer prevention.

CLINICAL SCENARIO

A 65-year-old woman with no family history of colorectal cancer undergoes screening colonoscopy, during which three polyps are found and removed—a 3-mm tubular adenoma in the sigmoid colon, an 8-mm sessile serrated polyp at the hepatic flexure, and a 2-mm hyperplastic polyp in the rectum. When should she undergo follow-up colonoscopy?

Based on the number, size, and pathologic makeup of the polyps in this patient, we would recommend follow-up surveillance colonoscopy in 5 years.

THE SERRATED POLYP PATHWAY: A DIFFERENT PATH TO COLORECTAL CANCER

Colorectal cancer is the third most common cancer in the United States.¹ From 70% to 80% of these cancers arise from adenomatous polyps via the adenoma-carcinoma pathway. This molecular pathway develops through chromosomal instability (CIN) and involves the loss of heterozygosity (the loss of function of one allele). This leads to the progressive accumulation of mutations in tumor-suppressor genes such as adenomatous polyposis coli (APC) and p53, and oncogenes such as KRAS. The result of these mutations is the development of adenomatous polyps that lead to microsatellite-stable colorectal cancers (Figure 1).²

More recently, studies have shown that the other 20% to 30% of colorectal cancers likely arise through a separate pathway, called the serrated polyp pathway or serrated neoplasia pathway. In contrast to CIN, this pathway is characterized by methylation of CpG islands (CIMP–CpG island methylation phenotype, CIMP) in the promoter regions of specific genes.³ Central to the serrated polyp pathway is progressive methylation in colonic mucosa; mutation in the BRAF oncogene, activating cell proliferation leading to a sessile serrated polyp; and epigenetic silencing of the DNA mismatch repair gene hMLH1, which is a key step in the progression to a sessile serrated polyp with dysplasia, which may rapidly become a microsatellite-unstable colorectal cancer.⁴

Histologically, serrated polyps have a serrated or sawtooth appearance from the folding in of the crypt epithelium, and they include hyperplastic polyps, traditional serrated adenomas, and sessile serrated polyps (sessile serrated adenomas).

Sessile serrated polyps and traditional serrated adenomas (which are rare) are thought to be precancerous, whereas hyperplastic polyps do not have malignant potential.

COMMON, BUT PREVALENCE IS NOT CLEARLY ESTABLISHED

The histologic criteria for sessile serrated polyps and traditional serrated adenomas have been elucidated,^4–7 but the epidemiology of these serrated polyps is not clear. Small studies have shown that sessile serrated polyps account for 2% to 9% of all polyps removed at colonoscopy^8–10; however, larger studies are needed to determine the prevalence because detection by an endoscopist and pathologic diagnosis of these polyps are both operator-dependent.

Traditional serrated adenomas are the least common type of serrated polyp, with a reported prevalence of 0.3%.⁷ Hyperplastic polyps are by far the most common, accounting for 20% to 30% of all polyps removed at colonoscopy.^9,11 Sessile serrated polyps have a predilection for the proximal colon and are associated with female sex and with smoking, ^12,13 but no consistent effect of other factors on their formation has been reported. In contrast, Wallace et al¹³ found that obesity, cigarette smoking, dietary fat intake, total caloric intake, and the consumption of red meat were associated with an increased risk of distal (but not proximal) serrated polyps, including hyperplastic polyps, sessile serrated polyps, and traditional serrated adenomas.

HYPERPLASTIC POLYPS

Hyperplastic polyps usually occur in the rectosigmoid colon. They appear as slightly elevated, whitish lesions with a diameter less than 5 mm (Figure 2). Microscopically, the serrated architecture is present in the upper half of their crypts (Figure 3). The proliferative zone is more or less normally located in the basal half of the crypt (the nonserrated portion), with nuclei that are small, uniform, and basally located.¹⁴ The bases of the crypts have a rounded contour and do not grow laterally along the muscularis mucosae.

SESSILE SERRATED POLYPS

Endoscopically, sessile serrated polyps are often subtle, appear flat or slightly elevated, and can be covered by yellow mucus (Figure 4). They are typically found in the proximal colon and are usually larger than typical adenomas, with 50% being larger than 10 mm.¹⁰

Histologically, the serrations are more prominent than those of hyperplastic polyps and involve the entire length of the crypt (Figure 5). The crypt bases are often dilated and display lateral growth along the lamina muscularis mucosae, resembling a letter t or l. The lamina muscularis mucosae is often thinner than normal. Crypts from sessile serrated polyps are occasionally found beneath the muscularis mucosae, a condition called pseudoinvasion.⁷

TRADITIONAL SERRATED ADENOMAS

Traditional serrated adenomas are usually left-sided. In contrast to the other types of serrated polyps, they are histologically often villiform and are lined by cells with elongated nuclei and abundant eosinophilic cytoplasm (Figure 6). Unlike those in sessile serrated polyps, the crypt bases do not display an abnormal architecture; rather, traditional serrated adenomas have abundant ectopic crypts (“budding crypts”) in the long, slender villi.⁷

Traditional serrated adenomas also appear to be genetically distinct from sessile serrated polyps. They are most often characterized by a KRAS (or less commonly, BRAF) mutation and commonly have methylation of the DNA repair gene MGMT (O-6-methylguanine-DNA methyltransferase) rather than hMLH1.

CHALLENGES TO EFFECTIVE COLONOSCOPY

Colonoscopic polypectomy of adenomatous polyps reduces the incidence of colorectal cancer and the rate of death from it.^15,16 However, recent data show that colonoscopy may not be as effective as once thought. As many as 9% of patients with colorectal cancer have had a “normal” colonoscopic examination in the preceding 3 years.^17,18 In addition, the reduction in incidence and mortality rates was less for cancers in the proximal colon than for cancers in the distal colon.^19,20

Possible explanations for this discrepancy include the skill of the endoscopist, technical limitations of the examination, incomplete removal of polyps, and inadequate bowel preparation. Several studies have shown that interval colorectal cancers are more likely to be found in the proximal colon and to have the same molecular characteristics as sessile serrated polyps and the serrated colorectal cancer pathway (CIMP-high and MSI-H).^21,22 Therefore, it is now thought that sessile serrated polyps may account for a substantial portion of “postcolonoscopy cancers” (ie, interval cancers) that arise in the proximal colon.

Two large studies of screening colonoscopy confirmed that the ability to detect sessile serrated polyps depends greatly on the skill of the endoscopist. Hetzel et al⁹ studied the differences in the rates of polyp detection among endoscopists performing more than 7,000 colonoscopies. Detection rates varied significantly for adenomas, hyperplastic polyps, and sessile serrated polyps, with the greatest variability noted in the detection of sessile serrated polyps. Significant variability was also noted in the ability of the pathologist to diagnose sessile serrated polyps.⁹

In the other study, a strong correlation was found between physicians who are “high detectors” of adenomas and their detection rates for proximal serrated polyps.²³ There is widespread acceptance that screening colonoscopy in average-risk patients age 50 and older should detect adenomas in more than 25% of men and more than 15% of women. There is no current minimum recommended detection rate for sessile serrated polyps, but some have suggested 1.5%.⁸

POLYPS AS PREDICTORS OF CANCER RISK

Certain polyp characteristics predict the risk of metachronous, advanced neoplasia. Advanced neoplasms are defined as invasive carcinomas, adenomas 10 mm or larger, or adenomas with any villous histology or high-grade dysplasia. Patients with one or two small tubular adenomas have a much lower risk of metachronous advanced neoplasia than do patients with more than two adenomas or advanced neoplasms.²⁴ Current recommended surveillance intervals vary on that basis (Table 2).²⁵

People who harbor serrated neoplasms are at high risk of synchronous serrated polyps and advanced adenomatous neoplasia. Pai et al²⁶ found that patients with one sessile serrated polyp were four times more likely to have additional serrated polyps at the same time than an unselected population. The authors suggested that this indicates a strong colonic mucosal-field defect in patients with sessile serrated polyps, thereby predisposing them to the development of synchronous serrated polyps.

Li et al²⁷ found that large serrated polyps (ie, > 10 mm) are associated with a risk of synchronous advanced neoplasia that is three times higher than in patients without adenomas. Schreiner et al²⁸ determined that patients with either a proximal or a large serrated polyp were at higher risk of synchronous advanced neoplasia compared with patients who did not have those lesions. Vu et al²⁹ found that patients who have both sessile serrated polyps and conventional adenomas have significantly larger and more numerous lesions of both types.²⁹ In addition, these lesions are more likely to be pathologically advanced when compared with people with only one or the other type.

In the only study of the risk of advanced neoplasia on follow-up colonoscopy,²⁸ patients with advanced neoplasia and proximal serrated polyps at baseline examination were twice as likely to have advanced neoplasia during subsequent surveillance than those with only advanced neoplasia at baseline examination.²⁸

Therefore, it seems clear that the presence of large or proximal serrated polyps or serrated neoplasms predicts the presence of synchronous and likely metachronous advanced neoplasms.

Guidelines for postpolypectomy surveillance for individuals with serrated lesions of the colon have recently been published.²⁵ Patients with large serrated lesions (≥ 10 mm) or an advanced serrated lesion (a sessile serrated polyp with or without cytologic dysplasia or a traditional serrated adenoma) should be followed closely. Patients with small (< 10-mm) rectosigmoid hyperplastic polyps should be followed as average-risk patients. If a patient with a sessile serrated polyp also has adenomas, the surveillance interval should be the shortest interval recommended for either lesion.²⁹

SURVEILLANCE FOR OUR PATIENT

In our patient, given the number, size, and histologic features of the polyps found, surveillance colonoscopy should be considered in 5 years. Although the clinical significance of the serrated pathway to colorectal cancer cannot be argued, further study is required to understand the lifetime risk to patients with serrated neoplasms and the optimal surveillance interval.

Sessile serrated polyps are a type of polyp recently recognized to be a precursor of colorectal cancer. They arise from a pathway of genetic alterations different from the pathway that causes the more common and well-understood conventional adenomas (also called tubular adenomas, tubulovillous adenomas, and villous adenomas).

We do not yet know enough about the lifetime colorectal cancer risk for individuals with sessile serrated polyps, nor do we know the optimal surveillance interval for patients who have these polyps on colonoscopy. It is believed that sessile serrated polyps may be the cause of a substantial number of “interval” colorectal cancers—ie, cancers that occur after colonoscopy but before the next scheduled examination.

Serrated polyps get their name from their jagged appearance on microscopy. In the past, all serrated colorectal lesions were called hyperplastic polyps. But with the advent of molecular and genetic diagnostics and with the ability to recognize the subtle morphologic differences of serrated lesions, they have been reclassified into those without malignant potential (hyperplastic polyps) and those that are neoplastic (sessile serrated polyps and traditional serrated adenomas) (Table 1).

In this article, we discuss the evolving understanding of the different types of serrated polyps, and we offer our thoughts on a reasonable postpolypectomy surveillance plan in patients with these lesions. We focus on sessile serrated polyps, the most common form of serrated polyp with cancerous potential, since it may be one of our greatest challenges in optimal colorectal cancer prevention.

CLINICAL SCENARIO

A 65-year-old woman with no family history of colorectal cancer undergoes screening colonoscopy, during which three polyps are found and removed—a 3-mm tubular adenoma in the sigmoid colon, an 8-mm sessile serrated polyp at the hepatic flexure, and a 2-mm hyperplastic polyp in the rectum. When should she undergo follow-up colonoscopy?

Based on the number, size, and pathologic makeup of the polyps in this patient, we would recommend follow-up surveillance colonoscopy in 5 years.

THE SERRATED POLYP PATHWAY: A DIFFERENT PATH TO COLORECTAL CANCER

Colorectal cancer is the third most common cancer in the United States.¹ From 70% to 80% of these cancers arise from adenomatous polyps via the adenoma-carcinoma pathway. This molecular pathway develops through chromosomal instability (CIN) and involves the loss of heterozygosity (the loss of function of one allele). This leads to the progressive accumulation of mutations in tumor-suppressor genes such as adenomatous polyposis coli (APC) and p53, and oncogenes such as KRAS. The result of these mutations is the development of adenomatous polyps that lead to microsatellite-stable colorectal cancers (Figure 1).²

More recently, studies have shown that the other 20% to 30% of colorectal cancers likely arise through a separate pathway, called the serrated polyp pathway or serrated neoplasia pathway. In contrast to CIN, this pathway is characterized by methylation of CpG islands (CIMP–CpG island methylation phenotype, CIMP) in the promoter regions of specific genes.³ Central to the serrated polyp pathway is progressive methylation in colonic mucosa; mutation in the BRAF oncogene, activating cell proliferation leading to a sessile serrated polyp; and epigenetic silencing of the DNA mismatch repair gene hMLH1, which is a key step in the progression to a sessile serrated polyp with dysplasia, which may rapidly become a microsatellite-unstable colorectal cancer.⁴

Histologically, serrated polyps have a serrated or sawtooth appearance from the folding in of the crypt epithelium, and they include hyperplastic polyps, traditional serrated adenomas, and sessile serrated polyps (sessile serrated adenomas).

Sessile serrated polyps and traditional serrated adenomas (which are rare) are thought to be precancerous, whereas hyperplastic polyps do not have malignant potential.

COMMON, BUT PREVALENCE IS NOT CLEARLY ESTABLISHED

The histologic criteria for sessile serrated polyps and traditional serrated adenomas have been elucidated,^4–7 but the epidemiology of these serrated polyps is not clear. Small studies have shown that sessile serrated polyps account for 2% to 9% of all polyps removed at colonoscopy^8–10; however, larger studies are needed to determine the prevalence because detection by an endoscopist and pathologic diagnosis of these polyps are both operator-dependent.

Traditional serrated adenomas are the least common type of serrated polyp, with a reported prevalence of 0.3%.⁷ Hyperplastic polyps are by far the most common, accounting for 20% to 30% of all polyps removed at colonoscopy.^9,11 Sessile serrated polyps have a predilection for the proximal colon and are associated with female sex and with smoking, ^12,13 but no consistent effect of other factors on their formation has been reported. In contrast, Wallace et al¹³ found that obesity, cigarette smoking, dietary fat intake, total caloric intake, and the consumption of red meat were associated with an increased risk of distal (but not proximal) serrated polyps, including hyperplastic polyps, sessile serrated polyps, and traditional serrated adenomas.

HYPERPLASTIC POLYPS

Hyperplastic polyps usually occur in the rectosigmoid colon. They appear as slightly elevated, whitish lesions with a diameter less than 5 mm (Figure 2). Microscopically, the serrated architecture is present in the upper half of their crypts (Figure 3). The proliferative zone is more or less normally located in the basal half of the crypt (the nonserrated portion), with nuclei that are small, uniform, and basally located.¹⁴ The bases of the crypts have a rounded contour and do not grow laterally along the muscularis mucosae.

SESSILE SERRATED POLYPS

Endoscopically, sessile serrated polyps are often subtle, appear flat or slightly elevated, and can be covered by yellow mucus (Figure 4). They are typically found in the proximal colon and are usually larger than typical adenomas, with 50% being larger than 10 mm.¹⁰

Histologically, the serrations are more prominent than those of hyperplastic polyps and involve the entire length of the crypt (Figure 5). The crypt bases are often dilated and display lateral growth along the lamina muscularis mucosae, resembling a letter t or l. The lamina muscularis mucosae is often thinner than normal. Crypts from sessile serrated polyps are occasionally found beneath the muscularis mucosae, a condition called pseudoinvasion.⁷

TRADITIONAL SERRATED ADENOMAS

Traditional serrated adenomas are usually left-sided. In contrast to the other types of serrated polyps, they are histologically often villiform and are lined by cells with elongated nuclei and abundant eosinophilic cytoplasm (Figure 6). Unlike those in sessile serrated polyps, the crypt bases do not display an abnormal architecture; rather, traditional serrated adenomas have abundant ectopic crypts (“budding crypts”) in the long, slender villi.⁷

Traditional serrated adenomas also appear to be genetically distinct from sessile serrated polyps. They are most often characterized by a KRAS (or less commonly, BRAF) mutation and commonly have methylation of the DNA repair gene MGMT (O-6-methylguanine-DNA methyltransferase) rather than hMLH1.

CHALLENGES TO EFFECTIVE COLONOSCOPY

Colonoscopic polypectomy of adenomatous polyps reduces the incidence of colorectal cancer and the rate of death from it.^15,16 However, recent data show that colonoscopy may not be as effective as once thought. As many as 9% of patients with colorectal cancer have had a “normal” colonoscopic examination in the preceding 3 years.^17,18 In addition, the reduction in incidence and mortality rates was less for cancers in the proximal colon than for cancers in the distal colon.^19,20

Possible explanations for this discrepancy include the skill of the endoscopist, technical limitations of the examination, incomplete removal of polyps, and inadequate bowel preparation. Several studies have shown that interval colorectal cancers are more likely to be found in the proximal colon and to have the same molecular characteristics as sessile serrated polyps and the serrated colorectal cancer pathway (CIMP-high and MSI-H).^21,22 Therefore, it is now thought that sessile serrated polyps may account for a substantial portion of “postcolonoscopy cancers” (ie, interval cancers) that arise in the proximal colon.

Two large studies of screening colonoscopy confirmed that the ability to detect sessile serrated polyps depends greatly on the skill of the endoscopist. Hetzel et al⁹ studied the differences in the rates of polyp detection among endoscopists performing more than 7,000 colonoscopies. Detection rates varied significantly for adenomas, hyperplastic polyps, and sessile serrated polyps, with the greatest variability noted in the detection of sessile serrated polyps. Significant variability was also noted in the ability of the pathologist to diagnose sessile serrated polyps.⁹

In the other study, a strong correlation was found between physicians who are “high detectors” of adenomas and their detection rates for proximal serrated polyps.²³ There is widespread acceptance that screening colonoscopy in average-risk patients age 50 and older should detect adenomas in more than 25% of men and more than 15% of women. There is no current minimum recommended detection rate for sessile serrated polyps, but some have suggested 1.5%.⁸

POLYPS AS PREDICTORS OF CANCER RISK

Certain polyp characteristics predict the risk of metachronous, advanced neoplasia. Advanced neoplasms are defined as invasive carcinomas, adenomas 10 mm or larger, or adenomas with any villous histology or high-grade dysplasia. Patients with one or two small tubular adenomas have a much lower risk of metachronous advanced neoplasia than do patients with more than two adenomas or advanced neoplasms.²⁴ Current recommended surveillance intervals vary on that basis (Table 2).²⁵

People who harbor serrated neoplasms are at high risk of synchronous serrated polyps and advanced adenomatous neoplasia. Pai et al²⁶ found that patients with one sessile serrated polyp were four times more likely to have additional serrated polyps at the same time than an unselected population. The authors suggested that this indicates a strong colonic mucosal-field defect in patients with sessile serrated polyps, thereby predisposing them to the development of synchronous serrated polyps.

Li et al²⁷ found that large serrated polyps (ie, > 10 mm) are associated with a risk of synchronous advanced neoplasia that is three times higher than in patients without adenomas. Schreiner et al²⁸ determined that patients with either a proximal or a large serrated polyp were at higher risk of synchronous advanced neoplasia compared with patients who did not have those lesions. Vu et al²⁹ found that patients who have both sessile serrated polyps and conventional adenomas have significantly larger and more numerous lesions of both types.²⁹ In addition, these lesions are more likely to be pathologically advanced when compared with people with only one or the other type.

In the only study of the risk of advanced neoplasia on follow-up colonoscopy,²⁸ patients with advanced neoplasia and proximal serrated polyps at baseline examination were twice as likely to have advanced neoplasia during subsequent surveillance than those with only advanced neoplasia at baseline examination.²⁸

Therefore, it seems clear that the presence of large or proximal serrated polyps or serrated neoplasms predicts the presence of synchronous and likely metachronous advanced neoplasms.

Guidelines for postpolypectomy surveillance for individuals with serrated lesions of the colon have recently been published.²⁵ Patients with large serrated lesions (≥ 10 mm) or an advanced serrated lesion (a sessile serrated polyp with or without cytologic dysplasia or a traditional serrated adenoma) should be followed closely. Patients with small (< 10-mm) rectosigmoid hyperplastic polyps should be followed as average-risk patients. If a patient with a sessile serrated polyp also has adenomas, the surveillance interval should be the shortest interval recommended for either lesion.²⁹

SURVEILLANCE FOR OUR PATIENT

In our patient, given the number, size, and histologic features of the polyps found, surveillance colonoscopy should be considered in 5 years. Although the clinical significance of the serrated pathway to colorectal cancer cannot be argued, further study is required to understand the lifetime risk to patients with serrated neoplasms and the optimal surveillance interval.

The integrity of cognitive function is a reliable indicator of healthy aging. But the progression of cognitive changes from normal aging to dementia is often insidious and easily underrecognized. Consequently, mild cognitive impairment (MCI)—the entity that characterizes this transition—has become an area of intense research. Since 1999, the number of research publications related to MCI has exploded, with more than 1,000 peer-reviewed studies in 2010 alone.

Controversy remains over the definition, diagnosis, prognosis, and management of MCI. However, in an evidence-based review of the literature,¹ the American Academy of Neurology concluded that MCI is a useful clinical entity and that patients with MCI should be identified and monitored because of the increased risk of progression to dementia.

See related article

Early studies appeared to indicate that patients with MCI were at high risk of further cognitive decline and progression to Alzheimer dementia.¹ But subsequent research found that not all were, leading to the recognition of two subtypes of MCI: amnestic, which mainly involves memory loss, and nonamnestic, which involves impairment of other cognitive domains. Patients with the amnestic type were determined to be more likely to eventually develop Alzheimer disease.² The amnestic subtype is being considered for inclusion in the next revision of the Diagnostic and Statistical Manual of Mental Disorders, ie, the fifth edition (DSM-V).³

MCI varies with each person affected. Neither its clinical nor its neuropathologic course follows a predictable, linear path, making its study especially challenging. The pathologic and molecular mechanisms of MCI are not well established. In the amnestic type, the distribution of cortical amyloid deposits appears transitional to the pathologic changes seen in Alzheimer disease.⁴ But postmortem brain tissues⁵ and clinical imaging studies⁶ reveal that some normal controls have a degree of amyloid deposition similar to that in patients with MCI. These findings limit the use of amyloid lesions as a robust pathologic marker for distinguishing normal aging from MCI.

MCI is diagnosed clinically, and clinicians should be able to diagnose most cases of MCI in the office. The first step is cognitive concern (ie, a change from the patient’s baseline cognitive status) raised by the patient, by an informant, or by a clinician. Often, in amnestic MCI, the earliest symptom is memory loss. Once persistent memory loss is documented, the patient is assessed for the ability to perform activities of daily living. To fulfill the criteria for the diagnosis of MCI, patients need to have intact function in the activities of daily living and no features of neurologic and psychiatric diseases that affect cognition. Further office-based cognitive testing helps to determine whether MCI is the amnestic or the nonamnestic type. A brief neuropsychological test such as the Montreal Cognitive Assessment often supports the diagnosis of MCI, although accurate characterization of cognitive dysfunction is enhanced with thorough neuropsychological testing.

MCI remains a clinical diagnosis with an imprecise prognosis. Although the amnestic MCI criteria are reasonably specific, they do not always predict progression to Alzheimer disease. Growing evidence suggests that neuropsychiatric symptoms, including depression, apathy, and anxiety, are clinical predictors of the progression of MCI to Alzheimer disease, and that the added risk can be substantial. For example, in one study, the risk of incident dementia was seven times higher if apathy was present.⁷ As such, a careful psychiatric evaluation of patients with MCI is strongly recommended and should be part of a comprehensive workup.

The study of MCI touches on almost all aspects of aging and dementia investigation. A great deal of research is focusing on the development of cerebrospinal fluid or imaging biomarkers of amyloid deposition, structural magnetic resonance imaging markers of neuronal loss, and genetic predisposition to detect the earliest signs of the disease in people who may be at risk. The rationale for the intense study of MCI is that the sooner the intervention in a degenerative process is started, the more likely that further cognitive and functional decline can be prevented: early diagnosis is paramount in trying to prevent subsequent disability. Clinical trials are needed to determine whether early detection of MCI or the detection of biomarkers in asymptomatic individuals alters the incidence of dementia or its prognosis.

In this issue of the Cleveland Clinic Journal of Medicine, Patel and Holland⁸ present a comprehensive overview of MCI and highlight the issues related to its diagnosis and management. The treatment of MCI is another area that is unclear. At this time, prescription of cognition-enhancing medications is not indicated. No pharmacologic agent is approved by the US Food and Drug Administration for treating MCI, although cholinesterase inhibitors have been studied. At the pathologic level, there is no clear consensus on whether presynaptic or postsynaptic (or both) cholinergic receptors are defective in MCI.⁹ There is some evidence of increased choline acetyltransferase activity in the hippocampus and the superior frontal cortex.¹⁰ Selected hippocampal and cortical cholinergic systems may be capable of compensatory responses in MCI. This may help explain why cholinesterase inhibitors are ineffective in preventing dementia in patients with MCI in therapeutic trials.

Patel and Holland recommend a reasonable multidisciplinary approach for managing MCI, although supporting evidence for such recommendations from clinical trials is lacking. Realizing that not all patients with MCI progress to Alzheimer disease and that some cases are reversible is cause for recommending close follow-up and monitoring of neuropsychiatric and cognitive symptoms in older patients.

MCI is now a clinical reality for all physicians dealing with older patients. Thus, MCI is of more than merely research interest to clinicians, who will come to recognize and diagnose this condition frequently in the aging population.

The integrity of cognitive function is a reliable indicator of healthy aging. But the progression of cognitive changes from normal aging to dementia is often insidious and easily underrecognized. Consequently, mild cognitive impairment (MCI)—the entity that characterizes this transition—has become an area of intense research. Since 1999, the number of research publications related to MCI has exploded, with more than 1,000 peer-reviewed studies in 2010 alone.

Controversy remains over the definition, diagnosis, prognosis, and management of MCI. However, in an evidence-based review of the literature,¹ the American Academy of Neurology concluded that MCI is a useful clinical entity and that patients with MCI should be identified and monitored because of the increased risk of progression to dementia.

See related article

Early studies appeared to indicate that patients with MCI were at high risk of further cognitive decline and progression to Alzheimer dementia.¹ But subsequent research found that not all were, leading to the recognition of two subtypes of MCI: amnestic, which mainly involves memory loss, and nonamnestic, which involves impairment of other cognitive domains. Patients with the amnestic type were determined to be more likely to eventually develop Alzheimer disease.² The amnestic subtype is being considered for inclusion in the next revision of the Diagnostic and Statistical Manual of Mental Disorders, ie, the fifth edition (DSM-V).³

MCI varies with each person affected. Neither its clinical nor its neuropathologic course follows a predictable, linear path, making its study especially challenging. The pathologic and molecular mechanisms of MCI are not well established. In the amnestic type, the distribution of cortical amyloid deposits appears transitional to the pathologic changes seen in Alzheimer disease.⁴ But postmortem brain tissues⁵ and clinical imaging studies⁶ reveal that some normal controls have a degree of amyloid deposition similar to that in patients with MCI. These findings limit the use of amyloid lesions as a robust pathologic marker for distinguishing normal aging from MCI.

MCI is diagnosed clinically, and clinicians should be able to diagnose most cases of MCI in the office. The first step is cognitive concern (ie, a change from the patient’s baseline cognitive status) raised by the patient, by an informant, or by a clinician. Often, in amnestic MCI, the earliest symptom is memory loss. Once persistent memory loss is documented, the patient is assessed for the ability to perform activities of daily living. To fulfill the criteria for the diagnosis of MCI, patients need to have intact function in the activities of daily living and no features of neurologic and psychiatric diseases that affect cognition. Further office-based cognitive testing helps to determine whether MCI is the amnestic or the nonamnestic type. A brief neuropsychological test such as the Montreal Cognitive Assessment often supports the diagnosis of MCI, although accurate characterization of cognitive dysfunction is enhanced with thorough neuropsychological testing.

MCI remains a clinical diagnosis with an imprecise prognosis. Although the amnestic MCI criteria are reasonably specific, they do not always predict progression to Alzheimer disease. Growing evidence suggests that neuropsychiatric symptoms, including depression, apathy, and anxiety, are clinical predictors of the progression of MCI to Alzheimer disease, and that the added risk can be substantial. For example, in one study, the risk of incident dementia was seven times higher if apathy was present.⁷ As such, a careful psychiatric evaluation of patients with MCI is strongly recommended and should be part of a comprehensive workup.

The study of MCI touches on almost all aspects of aging and dementia investigation. A great deal of research is focusing on the development of cerebrospinal fluid or imaging biomarkers of amyloid deposition, structural magnetic resonance imaging markers of neuronal loss, and genetic predisposition to detect the earliest signs of the disease in people who may be at risk. The rationale for the intense study of MCI is that the sooner the intervention in a degenerative process is started, the more likely that further cognitive and functional decline can be prevented: early diagnosis is paramount in trying to prevent subsequent disability. Clinical trials are needed to determine whether early detection of MCI or the detection of biomarkers in asymptomatic individuals alters the incidence of dementia or its prognosis.

In this issue of the Cleveland Clinic Journal of Medicine, Patel and Holland⁸ present a comprehensive overview of MCI and highlight the issues related to its diagnosis and management. The treatment of MCI is another area that is unclear. At this time, prescription of cognition-enhancing medications is not indicated. No pharmacologic agent is approved by the US Food and Drug Administration for treating MCI, although cholinesterase inhibitors have been studied. At the pathologic level, there is no clear consensus on whether presynaptic or postsynaptic (or both) cholinergic receptors are defective in MCI.⁹ There is some evidence of increased choline acetyltransferase activity in the hippocampus and the superior frontal cortex.¹⁰ Selected hippocampal and cortical cholinergic systems may be capable of compensatory responses in MCI. This may help explain why cholinesterase inhibitors are ineffective in preventing dementia in patients with MCI in therapeutic trials.

Patel and Holland recommend a reasonable multidisciplinary approach for managing MCI, although supporting evidence for such recommendations from clinical trials is lacking. Realizing that not all patients with MCI progress to Alzheimer disease and that some cases are reversible is cause for recommending close follow-up and monitoring of neuropsychiatric and cognitive symptoms in older patients.

MCI is now a clinical reality for all physicians dealing with older patients. Thus, MCI is of more than merely research interest to clinicians, who will come to recognize and diagnose this condition frequently in the aging population.

The integrity of cognitive function is a reliable indicator of healthy aging. But the progression of cognitive changes from normal aging to dementia is often insidious and easily underrecognized. Consequently, mild cognitive impairment (MCI)—the entity that characterizes this transition—has become an area of intense research. Since 1999, the number of research publications related to MCI has exploded, with more than 1,000 peer-reviewed studies in 2010 alone.

Controversy remains over the definition, diagnosis, prognosis, and management of MCI. However, in an evidence-based review of the literature,¹ the American Academy of Neurology concluded that MCI is a useful clinical entity and that patients with MCI should be identified and monitored because of the increased risk of progression to dementia.

See related article

Early studies appeared to indicate that patients with MCI were at high risk of further cognitive decline and progression to Alzheimer dementia.¹ But subsequent research found that not all were, leading to the recognition of two subtypes of MCI: amnestic, which mainly involves memory loss, and nonamnestic, which involves impairment of other cognitive domains. Patients with the amnestic type were determined to be more likely to eventually develop Alzheimer disease.² The amnestic subtype is being considered for inclusion in the next revision of the Diagnostic and Statistical Manual of Mental Disorders, ie, the fifth edition (DSM-V).³

MCI varies with each person affected. Neither its clinical nor its neuropathologic course follows a predictable, linear path, making its study especially challenging. The pathologic and molecular mechanisms of MCI are not well established. In the amnestic type, the distribution of cortical amyloid deposits appears transitional to the pathologic changes seen in Alzheimer disease.⁴ But postmortem brain tissues⁵ and clinical imaging studies⁶ reveal that some normal controls have a degree of amyloid deposition similar to that in patients with MCI. These findings limit the use of amyloid lesions as a robust pathologic marker for distinguishing normal aging from MCI.

MCI is diagnosed clinically, and clinicians should be able to diagnose most cases of MCI in the office. The first step is cognitive concern (ie, a change from the patient’s baseline cognitive status) raised by the patient, by an informant, or by a clinician. Often, in amnestic MCI, the earliest symptom is memory loss. Once persistent memory loss is documented, the patient is assessed for the ability to perform activities of daily living. To fulfill the criteria for the diagnosis of MCI, patients need to have intact function in the activities of daily living and no features of neurologic and psychiatric diseases that affect cognition. Further office-based cognitive testing helps to determine whether MCI is the amnestic or the nonamnestic type. A brief neuropsychological test such as the Montreal Cognitive Assessment often supports the diagnosis of MCI, although accurate characterization of cognitive dysfunction is enhanced with thorough neuropsychological testing.

MCI remains a clinical diagnosis with an imprecise prognosis. Although the amnestic MCI criteria are reasonably specific, they do not always predict progression to Alzheimer disease. Growing evidence suggests that neuropsychiatric symptoms, including depression, apathy, and anxiety, are clinical predictors of the progression of MCI to Alzheimer disease, and that the added risk can be substantial. For example, in one study, the risk of incident dementia was seven times higher if apathy was present.⁷ As such, a careful psychiatric evaluation of patients with MCI is strongly recommended and should be part of a comprehensive workup.

The study of MCI touches on almost all aspects of aging and dementia investigation. A great deal of research is focusing on the development of cerebrospinal fluid or imaging biomarkers of amyloid deposition, structural magnetic resonance imaging markers of neuronal loss, and genetic predisposition to detect the earliest signs of the disease in people who may be at risk. The rationale for the intense study of MCI is that the sooner the intervention in a degenerative process is started, the more likely that further cognitive and functional decline can be prevented: early diagnosis is paramount in trying to prevent subsequent disability. Clinical trials are needed to determine whether early detection of MCI or the detection of biomarkers in asymptomatic individuals alters the incidence of dementia or its prognosis.

In this issue of the Cleveland Clinic Journal of Medicine, Patel and Holland⁸ present a comprehensive overview of MCI and highlight the issues related to its diagnosis and management. The treatment of MCI is another area that is unclear. At this time, prescription of cognition-enhancing medications is not indicated. No pharmacologic agent is approved by the US Food and Drug Administration for treating MCI, although cholinesterase inhibitors have been studied. At the pathologic level, there is no clear consensus on whether presynaptic or postsynaptic (or both) cholinergic receptors are defective in MCI.⁹ There is some evidence of increased choline acetyltransferase activity in the hippocampus and the superior frontal cortex.¹⁰ Selected hippocampal and cortical cholinergic systems may be capable of compensatory responses in MCI. This may help explain why cholinesterase inhibitors are ineffective in preventing dementia in patients with MCI in therapeutic trials.

Patel and Holland recommend a reasonable multidisciplinary approach for managing MCI, although supporting evidence for such recommendations from clinical trials is lacking. Realizing that not all patients with MCI progress to Alzheimer disease and that some cases are reversible is cause for recommending close follow-up and monitoring of neuropsychiatric and cognitive symptoms in older patients.

MCI is now a clinical reality for all physicians dealing with older patients. Thus, MCI is of more than merely research interest to clinicians, who will come to recognize and diagnose this condition frequently in the aging population.

As our population ages, people are thinking more about preserving their quality of life, especially with regard to maintaining their cognitive and functional abilities. Older patients and caregivers often raise concerns about cognitive issues to their primary care providers: many patients have memory complaints, are worried about whether these are merely part of normal aging or symptoms of early dementia, and want strategies to forestall the progression of cognitive impairment.

Mild cognitive impairment (MCI) is a heterogeneous syndrome that in some cases represents a transition between normal aging and dementia. However, this condition is not yet well understood. Although some patients progress to dementia, others remain stable, or even improve. This article will review the current definitions and the underlying physiology of MCI, as well as diagnostic and management strategies.

See related editorial

COGNITIVE CHANGES OCCUR WITH NORMAL AGING

Cognition is defined as a means of acquiring and processing information about ourselves and our world. It includes memory as well as other domains such as attention, visuospatial skills, mental processing speed, language, and executive function. Cognitive abilities typically peak between ages 30 and 40, plateau in our 50s and 60s, and decline in our late 70s.

With age come detectable changes in the brain: brain weight declines by 10% by age 80, blood flow diminishes, neurons are lost throughout life, and nerve conduction slows. Despite these changes, the brain has a great deal of functional reserve capacity.

Table 1 compares the signs of normal aging, MCI, and dementia. Normally, cognitive abilities decline gradually with age without affecting overall function or activities of daily living. Even in normal aging, the processing of new information (new learning) is reduced. Mental processing becomes less efficient and slower. Visuospatial skills gradually decline, recall slows, and ultimately, the speed of performance slows as well. Additionally, distractibility increases. On the other hand, normal aging does not affect recognition, intelligence, or long-term memory.¹

The line between the normal effects of aging on cognition and true pathologic cognitive decline is blurry. In a busy clinical practice, it is often difficult to determine whether problems with memory and cognition that elderly patients and their family members describe represent true pathologic decline. In general, the clinical presentation of MCI is more profound than that of age-associated cognitive impairment: whereas normal aging may involve forgetting names and words and misplacing things, MCI frequently involves forgetting conversations, information that one would ordinarily remember, appointments, and planned events.

BETWEEN NORMAL AGING AND DEMENTIA

MCI is a transitional state between normal cognition and dementia. But the course is not inevitably downward: on follow-up, patients with MCI may be better, stable, or worse (see PROGNOSIS VARIES, below).

On autopsy studies, the brains of people with MCI appear intermediate between normal brains and brains of people with Alzheimer-type dementia, which have neurofibrillary tangles, amyloid senile plaques, and neuronal degeneration.

Definitions of MCI vary

True cognitive decline that is more profound than normal aging was named and defined differently in different studies, making comparisons difficult. The concept of MCI arose from the term “benign senescent forgetfulness,” used by Kral in 1962.² Other early terms include “cognitive impairment no dementia,” “memory impairment,” “mild cognitive disorder,” and “mild neurocognitive disorder.”^3,4

MCI was first defined as a precursor to Alzheimer dementia. The term later described a sometimes reversible but abnormal state. It is a heterogeneous syndrome in terms of etiology, incidence, prevalence, presentation, and overall prognosis.

Most recently, MCI has been defined as^5,6:

• Subjective memory complaints, preferably qualified by another person
• Memory impairment, with consideration for age and education
• Preserved general cognitive function
• Intact activities of daily living
• Absence of overt dementia.

MCI may arise from vascular, neurodegenerative, traumatic, metabolic, psychiatric, and other underlying medical disorders.^7–9

The prevalence of MCI is difficult to determine because of the various definitions, populations studied (eg, clinic-based vs community-dwelling), and evaluation techniques. Published rates vary from 2% to 4% in all patients to 10% to 20% in the elderly. Incidence rates in the elderly vary from 14 to 75 per 1,000 patient-years.^10–14

EARLY RECOGNITION ALLOWS PROMPT EVALUATION AND PLANNING

Pathologic cognitive decline is best detected early, for many reasons. Early recognition and intervention may help delay further decline. Establishing a diagnosis can also lessen family and caregiver stress and misunderstanding. Education of caregivers is important so that they can prepare for likely behavioral changes and plan for future care. Advance care planning, including advance directives, power of attorney, and designation of proxy for decision-making, is extremely important and is best considered before cognitive impairment becomes severe.

The diagnosis of MCI also provides the opportunity to assess safety concerns related to driving, working, medication compliance, the home environment, and firearms. Because patients with MCI are still highly functional, these issues need not be fully evaluated and should be handled on a case-by-case basis, depending on concerns raised. For example, if depression is an active concern, firearms safety should be addressed.

MEMORY LOSS MAY NOT BE THE PRIMARY CONCERN

MCI is categorized into two types based on whether memory loss is the primary cognitive deficit.

The amnestic type predominantly involves memory problems and is more common. Generally, several years elapse between initial memory concerns and a clinical diagnosis of MCI. Patients with amnestic MCI that progresses to dementia are more likely to develop Alzheimer disease.^2,15

Nonamnestic types involve domains of cognition other than memory, such as executive function, attention, visuospatial ability, and language. Nonamnestic MCI can be subcategorized through extensive neuropsychological evaluation as involving single or multiple impaired domains.^16,17 Such categorization is particularly important in determining prognosis, as patients with involvement of multiple domains are at higher risk of progressing to dementia.

Patients with nonamnestic MCI who progress to dementia are more likely to have non-Alzheimer types of dementia, such as Lewy body dementia and frontotemporal dementias.¹⁰

HISTORY SHOULD FOCUS ON FUNCTION, MEDICATIONS, AND DEPRESSION

Cognitive impairment should be clinically evaluated within the context of cognition, function, and behavior. Clinicians should focus on the time course of cognitive concerns, the specifics of the concerns, and their impact on day-to-day living and functioning. In assessing functional capacity, it is important to determine the level of assistance the patient needs to perform specific activities of daily living and instrumental activities of daily living (ie, the more advanced skills needed to live independently) (Table 2).

A thorough history includes consideration of baseline education, intellect, and previous learning disabilities; sensory impairments with emphasis on sight and hearing impairments; uncontrolled pain; head trauma; sleep disorders; concurrent medical and psychosocial illnesses such as depression and anxiety; substance abuse; and polypharmacy.

Depression, delirium, and the use of anticholinergic drugs are particularly important to evaluate, as these can result in cognitive deficits associated with MCI. The cognitive deficits may resolve with treatment or with stopping the drug.

Behavioral concerns such as wandering, agitation, and anger and sleep concerns, eating habits, and social etiquette are also important to evaluate.

PHYSICAL EVALUATION: RULE OUT REVERSIBLE CONDITIONS

The differential diagnosis of MCI includes delirium, depression, dementia, possibly reversible conditions affecting cognition (vitamin B₁₂ deficiency, hypothyroidism, effects of anticholinergic drugs), and uncommonly, central nervous system conditions (normal pressure hydrocephalus, subdural hematoma, tumor, stroke), and others (Table 3).¹⁸

A thorough physical examination should include neurologic, cardiovascular, hearing, and vision examinations, as well as an evaluation of functional status.

Laboratory studies. Although evidence is lacking to support a laboratory diagnostic workup for MCI, a selective evaluation including a comprehensive metabolic profile, complete blood count, thyroid studies, and a vitamin B₁₂ level can be useful. Occasionally, a treatable cause of impaired cognition such as vitamin B₁₂ deficiency or thyroid disease can be identified and resolved. A further comprehensive laboratory evaluation should be obtained if a patient progresses to dementia.

Imaging can be used in conjunction with other supportive evidence but should not be used solely to establish a diagnosis of MCI. Magnetic resonance imaging (MRI) can detect metastatic disease, normal pressure hydrocephalus, and subdural hematoma, in addition to traumatic, inflammatory, infectious, and vascular causes of cognitive impairment. MRI can also determine focal areas of atrophy; temporal lobe atrophy is a risk factor for progression to dementia.

Other studies. Structural MRI using techniques to evaluate the hippocampus, functional imaging, genetic testing for ApoE4 alleles, and biomarkers in cerebrospinal fluid are currently under evaluation to identify those at risk of progression to dementia. Recently published guidelines by the Alzheimer’s Association and the National Institute on Aging indicate that pathophysiologic findings in MCI that may predict future Alzheimer disease are meant to guide research and are not part of clinical practice at this time.¹⁹

COGNITIVE AND NEUROLOGIC TESTING IDENTIFIES DEFICITS

A number of global measures of cognition can be used in the office in clinical practice to help in evaluating significant cognitive concerns and to determine areas and severity of deficits at presentation. These include the Mini-Mental State Examination, the Montreal Cognitive Assessment, the Saint Louis University Mental Status, and many others (Table 4).²⁰

Caveats about interpreting the results: each of these tests has different sensitivities and specificities for detecting MCI. Also, we need to take into account the patient’s level of education, as highly educated people tend to do better on these tests.^21–23 It is important to note that some patients with MCI have normal results or only minimally abnormal results on these tests.

Neuropsychological testing is reserved for patients needing further evaluation, eg, those with atypical or complex cases, and those in whom the specific domains of cognition involved need to be identified. It can also provide additional insight into the contribution of depression to cognitive deficits. Neuropsychological testing is usually very time-intensive and requires patients to be able to perform complicated cognitive tasks. Not all patients are good candidates for this testing; sensory and motor impairments must be considered to determine if patients can adequately participate in testing. The cost of neuropsychological testing for MCI may not be covered by insurance and should be discussed with patients before referral. Specific concerns about cognitive problems that need further evaluation should be stated in the referral.

No one test should be used to make a diagnosis of MCI or dementia; clinical judgment is also necessary. The need for referral to a neurologist, geriatrician, or psychiatrist depends on the nature of the cognitive and behavioral concerns, the complexity of making a diagnosis, the need for further assessment of functional ability, and the need for evaluation of risk of progression to dementia.

MEDICATIONS HAVE LITTLE ROLE IN MANAGEMENT

No drug has yet been approved by the US Food and Drug Administration for treating MCI.

The acetylcholinesterase inhibitors donepezil (Aricept), galantamine (Razadyne), and rivastigmine (Exelon) have undergone clinical trials for treatment of MCI but have not been definitely shown to significantly reduce the risk of progression to dementia.²⁴

On the other hand, Diniz et al²⁵ performed a meta-analysis of the use of cholinesterase inhibitors in patients with MCI as a means of delaying the progression to Alzheimer disease.²⁵ They calculated that 15.4% of patients who received these drugs progressed to dementia, compared with 20.4% of those who received placebo, for a relative risk of 0.75 (95% confidence interval 0.66–0.87, P < .001). They concluded that the use of these drugs in patients with MCI “may attenuate the risk of progression” to Alzheimer disease and dementia.

In addition to not being approved for this indication and showing mixed evidence of efficacy, these drugs have well-known side effects such as diarrhea, nausea, vomiting, anorexia, and rhinitis, as well as significant but lesser-known side effects such as syncope, bradycardia, gastrointestinal bleeding, and vivid dreams.²⁶

Nevertheless, some patients with MCI, particularly those at high risk with amnestic MCI, may still want to try these medications. In these cases, the risks and possible benefits (or lack of them) should be reviewed thoroughly with the patient and family, and the discussion should be documented before starting therapy. The lowest starting dose of acetylcholinesterase inhibitor should be used to determine tolerability; generally, the dose is increased after 4 weeks to a maintenance dosage, with particular consideration of side effects.

Other agents have also been evaluated for MCI but have shown no evidence of benefit. Nonsteroidal anti-inflammatory drugs have not been found to either improve symptoms or delay progression to dementia. Ginkgo biloba has shown unclear benefit in achieving important treatment goals for MCI,²⁷ and it increases the risk of bleeding in the elderly. Vitamin E was evaluated in one study and did not slow progression to dementia.²⁸

STAYING HEALTHY AND ACTIVE MAY HELP

We recommend optimizing vascular risk factors such as diabetes, blood pressure, smoking, and lipid levels in managing MCI, given that uncontrolled vascular risk factors may lead to progression to dementia. However, we can point to no research to support this recommendation.

Cognitive rehabilitation involves training in deficient domains and developing strategies to compensate for deficits. Different interventions are used, including computerized simulation exercises, memory aids, organizational techniques, personal digital assistants, crossword puzzles, mind games, and other mentally engaging activities.²⁹

Increasing physical activity is another aspect of treatment. Some studies have shown that it improves cognitive performance in MCI, at least in the short term.^30,31

Optimizing mood and emotions is also important. If present, depression should be identified and optimally treated. Social activity can be useful and leads to less emotional stress and to better coping mechanisms.

A multidisciplinary approach may help patients and may also help relieve the burden on the caregiver. Periodic reassessment of cognitive and functional symptoms may be warranted.

Maintaining disease-specific registries of patients who have MCI may be useful to longitudinally follow patients and ensure that they get the care they need.

PROGNOSIS VARIES

MCI is a heterogeneous condition that often does not predictably progress to dementia. Patients and families should be told that having MCI does not mean that the patient will necessarily get dementia.

Several studies have shown that the annual risk of progression to dementia for patients with MCI is 5% to 10% in community-dwelling populations and up to 15% in specialty-clinic patients.^24,32 In comparison, the incidence of dementia in the general elderly population is 1% to 3% per year.

On the other hand, a number of studies show that MCI improves significantly in up to 15% to 40% of patients and sometimes reverts to a normal cognitive state.^33,34 But prospective studies of patients with clinically diagnosed MCI usually find a low rate of reversion to a normal state.^35,36 Many are short-term follow-up studies of different populations, making generalizations difficult.¹⁴

Patients with impairment in instrumental activities of daily living may be more likely to have nonreversible MCI and may be at higher risk of progressing to dementia.³⁷

PATIENT AND FAMILY EDUCATION AND FOLLOW-UP CONSIDERATIONS

Caregiver education and stress management are important components of managing patients with MCI. Formally assessing caregiver stress is useful. Steps to prevent caregiver burnout include making use of respite care, counseling, education, and community resources such as adult day care and those offered by the Alzheimer’s Association.

Clinicians should follow patients with MCI closely to evaluate progression, address specific concerns, minimize risks, emphasize healthy habits, manage concurrent illnesses, and evaluate management.

Functional status, as demonstrated by activities of daily living, is the most important determinant of progression of MCI to dementia and should be evaluated at each visit. Repeat cognitive testing should be done on patients who have significant loss of functional status. Changes in work habits also warrant further attention.

Patients diagnosed with MCI or those who have persistent cognitive concerns should be considered for neuropsychological evaluation after 1 year to assess specific deficits and progression of cognitive impairment.

Finally, consideration should be given to current clinical research, and referrals should be made to research centers that focus on MCI management and treatment.

As our population ages, people are thinking more about preserving their quality of life, especially with regard to maintaining their cognitive and functional abilities. Older patients and caregivers often raise concerns about cognitive issues to their primary care providers: many patients have memory complaints, are worried about whether these are merely part of normal aging or symptoms of early dementia, and want strategies to forestall the progression of cognitive impairment.

Mild cognitive impairment (MCI) is a heterogeneous syndrome that in some cases represents a transition between normal aging and dementia. However, this condition is not yet well understood. Although some patients progress to dementia, others remain stable, or even improve. This article will review the current definitions and the underlying physiology of MCI, as well as diagnostic and management strategies.

See related editorial

COGNITIVE CHANGES OCCUR WITH NORMAL AGING

Cognition is defined as a means of acquiring and processing information about ourselves and our world. It includes memory as well as other domains such as attention, visuospatial skills, mental processing speed, language, and executive function. Cognitive abilities typically peak between ages 30 and 40, plateau in our 50s and 60s, and decline in our late 70s.

With age come detectable changes in the brain: brain weight declines by 10% by age 80, blood flow diminishes, neurons are lost throughout life, and nerve conduction slows. Despite these changes, the brain has a great deal of functional reserve capacity.

Table 1 compares the signs of normal aging, MCI, and dementia. Normally, cognitive abilities decline gradually with age without affecting overall function or activities of daily living. Even in normal aging, the processing of new information (new learning) is reduced. Mental processing becomes less efficient and slower. Visuospatial skills gradually decline, recall slows, and ultimately, the speed of performance slows as well. Additionally, distractibility increases. On the other hand, normal aging does not affect recognition, intelligence, or long-term memory.¹

The line between the normal effects of aging on cognition and true pathologic cognitive decline is blurry. In a busy clinical practice, it is often difficult to determine whether problems with memory and cognition that elderly patients and their family members describe represent true pathologic decline. In general, the clinical presentation of MCI is more profound than that of age-associated cognitive impairment: whereas normal aging may involve forgetting names and words and misplacing things, MCI frequently involves forgetting conversations, information that one would ordinarily remember, appointments, and planned events.

BETWEEN NORMAL AGING AND DEMENTIA

MCI is a transitional state between normal cognition and dementia. But the course is not inevitably downward: on follow-up, patients with MCI may be better, stable, or worse (see PROGNOSIS VARIES, below).

On autopsy studies, the brains of people with MCI appear intermediate between normal brains and brains of people with Alzheimer-type dementia, which have neurofibrillary tangles, amyloid senile plaques, and neuronal degeneration.

Definitions of MCI vary

True cognitive decline that is more profound than normal aging was named and defined differently in different studies, making comparisons difficult. The concept of MCI arose from the term “benign senescent forgetfulness,” used by Kral in 1962.² Other early terms include “cognitive impairment no dementia,” “memory impairment,” “mild cognitive disorder,” and “mild neurocognitive disorder.”^3,4

MCI was first defined as a precursor to Alzheimer dementia. The term later described a sometimes reversible but abnormal state. It is a heterogeneous syndrome in terms of etiology, incidence, prevalence, presentation, and overall prognosis.

Most recently, MCI has been defined as^5,6:

• Subjective memory complaints, preferably qualified by another person
• Memory impairment, with consideration for age and education
• Preserved general cognitive function
• Intact activities of daily living
• Absence of overt dementia.

MCI may arise from vascular, neurodegenerative, traumatic, metabolic, psychiatric, and other underlying medical disorders.^7–9

The prevalence of MCI is difficult to determine because of the various definitions, populations studied (eg, clinic-based vs community-dwelling), and evaluation techniques. Published rates vary from 2% to 4% in all patients to 10% to 20% in the elderly. Incidence rates in the elderly vary from 14 to 75 per 1,000 patient-years.^10–14

EARLY RECOGNITION ALLOWS PROMPT EVALUATION AND PLANNING

Pathologic cognitive decline is best detected early, for many reasons. Early recognition and intervention may help delay further decline. Establishing a diagnosis can also lessen family and caregiver stress and misunderstanding. Education of caregivers is important so that they can prepare for likely behavioral changes and plan for future care. Advance care planning, including advance directives, power of attorney, and designation of proxy for decision-making, is extremely important and is best considered before cognitive impairment becomes severe.

The diagnosis of MCI also provides the opportunity to assess safety concerns related to driving, working, medication compliance, the home environment, and firearms. Because patients with MCI are still highly functional, these issues need not be fully evaluated and should be handled on a case-by-case basis, depending on concerns raised. For example, if depression is an active concern, firearms safety should be addressed.

MEMORY LOSS MAY NOT BE THE PRIMARY CONCERN

MCI is categorized into two types based on whether memory loss is the primary cognitive deficit.

The amnestic type predominantly involves memory problems and is more common. Generally, several years elapse between initial memory concerns and a clinical diagnosis of MCI. Patients with amnestic MCI that progresses to dementia are more likely to develop Alzheimer disease.^2,15

Nonamnestic types involve domains of cognition other than memory, such as executive function, attention, visuospatial ability, and language. Nonamnestic MCI can be subcategorized through extensive neuropsychological evaluation as involving single or multiple impaired domains.^16,17 Such categorization is particularly important in determining prognosis, as patients with involvement of multiple domains are at higher risk of progressing to dementia.

Patients with nonamnestic MCI who progress to dementia are more likely to have non-Alzheimer types of dementia, such as Lewy body dementia and frontotemporal dementias.¹⁰

HISTORY SHOULD FOCUS ON FUNCTION, MEDICATIONS, AND DEPRESSION

Cognitive impairment should be clinically evaluated within the context of cognition, function, and behavior. Clinicians should focus on the time course of cognitive concerns, the specifics of the concerns, and their impact on day-to-day living and functioning. In assessing functional capacity, it is important to determine the level of assistance the patient needs to perform specific activities of daily living and instrumental activities of daily living (ie, the more advanced skills needed to live independently) (Table 2).

A thorough history includes consideration of baseline education, intellect, and previous learning disabilities; sensory impairments with emphasis on sight and hearing impairments; uncontrolled pain; head trauma; sleep disorders; concurrent medical and psychosocial illnesses such as depression and anxiety; substance abuse; and polypharmacy.

Depression, delirium, and the use of anticholinergic drugs are particularly important to evaluate, as these can result in cognitive deficits associated with MCI. The cognitive deficits may resolve with treatment or with stopping the drug.

Behavioral concerns such as wandering, agitation, and anger and sleep concerns, eating habits, and social etiquette are also important to evaluate.

PHYSICAL EVALUATION: RULE OUT REVERSIBLE CONDITIONS

The differential diagnosis of MCI includes delirium, depression, dementia, possibly reversible conditions affecting cognition (vitamin B₁₂ deficiency, hypothyroidism, effects of anticholinergic drugs), and uncommonly, central nervous system conditions (normal pressure hydrocephalus, subdural hematoma, tumor, stroke), and others (Table 3).¹⁸

A thorough physical examination should include neurologic, cardiovascular, hearing, and vision examinations, as well as an evaluation of functional status.

Laboratory studies. Although evidence is lacking to support a laboratory diagnostic workup for MCI, a selective evaluation including a comprehensive metabolic profile, complete blood count, thyroid studies, and a vitamin B₁₂ level can be useful. Occasionally, a treatable cause of impaired cognition such as vitamin B₁₂ deficiency or thyroid disease can be identified and resolved. A further comprehensive laboratory evaluation should be obtained if a patient progresses to dementia.

Imaging can be used in conjunction with other supportive evidence but should not be used solely to establish a diagnosis of MCI. Magnetic resonance imaging (MRI) can detect metastatic disease, normal pressure hydrocephalus, and subdural hematoma, in addition to traumatic, inflammatory, infectious, and vascular causes of cognitive impairment. MRI can also determine focal areas of atrophy; temporal lobe atrophy is a risk factor for progression to dementia.

Other studies. Structural MRI using techniques to evaluate the hippocampus, functional imaging, genetic testing for ApoE4 alleles, and biomarkers in cerebrospinal fluid are currently under evaluation to identify those at risk of progression to dementia. Recently published guidelines by the Alzheimer’s Association and the National Institute on Aging indicate that pathophysiologic findings in MCI that may predict future Alzheimer disease are meant to guide research and are not part of clinical practice at this time.¹⁹

COGNITIVE AND NEUROLOGIC TESTING IDENTIFIES DEFICITS

A number of global measures of cognition can be used in the office in clinical practice to help in evaluating significant cognitive concerns and to determine areas and severity of deficits at presentation. These include the Mini-Mental State Examination, the Montreal Cognitive Assessment, the Saint Louis University Mental Status, and many others (Table 4).²⁰

Caveats about interpreting the results: each of these tests has different sensitivities and specificities for detecting MCI. Also, we need to take into account the patient’s level of education, as highly educated people tend to do better on these tests.^21–23 It is important to note that some patients with MCI have normal results or only minimally abnormal results on these tests.

Neuropsychological testing is reserved for patients needing further evaluation, eg, those with atypical or complex cases, and those in whom the specific domains of cognition involved need to be identified. It can also provide additional insight into the contribution of depression to cognitive deficits. Neuropsychological testing is usually very time-intensive and requires patients to be able to perform complicated cognitive tasks. Not all patients are good candidates for this testing; sensory and motor impairments must be considered to determine if patients can adequately participate in testing. The cost of neuropsychological testing for MCI may not be covered by insurance and should be discussed with patients before referral. Specific concerns about cognitive problems that need further evaluation should be stated in the referral.

No one test should be used to make a diagnosis of MCI or dementia; clinical judgment is also necessary. The need for referral to a neurologist, geriatrician, or psychiatrist depends on the nature of the cognitive and behavioral concerns, the complexity of making a diagnosis, the need for further assessment of functional ability, and the need for evaluation of risk of progression to dementia.

MEDICATIONS HAVE LITTLE ROLE IN MANAGEMENT

No drug has yet been approved by the US Food and Drug Administration for treating MCI.

The acetylcholinesterase inhibitors donepezil (Aricept), galantamine (Razadyne), and rivastigmine (Exelon) have undergone clinical trials for treatment of MCI but have not been definitely shown to significantly reduce the risk of progression to dementia.²⁴

On the other hand, Diniz et al²⁵ performed a meta-analysis of the use of cholinesterase inhibitors in patients with MCI as a means of delaying the progression to Alzheimer disease.²⁵ They calculated that 15.4% of patients who received these drugs progressed to dementia, compared with 20.4% of those who received placebo, for a relative risk of 0.75 (95% confidence interval 0.66–0.87, P < .001). They concluded that the use of these drugs in patients with MCI “may attenuate the risk of progression” to Alzheimer disease and dementia.

In addition to not being approved for this indication and showing mixed evidence of efficacy, these drugs have well-known side effects such as diarrhea, nausea, vomiting, anorexia, and rhinitis, as well as significant but lesser-known side effects such as syncope, bradycardia, gastrointestinal bleeding, and vivid dreams.²⁶

Nevertheless, some patients with MCI, particularly those at high risk with amnestic MCI, may still want to try these medications. In these cases, the risks and possible benefits (or lack of them) should be reviewed thoroughly with the patient and family, and the discussion should be documented before starting therapy. The lowest starting dose of acetylcholinesterase inhibitor should be used to determine tolerability; generally, the dose is increased after 4 weeks to a maintenance dosage, with particular consideration of side effects.

Other agents have also been evaluated for MCI but have shown no evidence of benefit. Nonsteroidal anti-inflammatory drugs have not been found to either improve symptoms or delay progression to dementia. Ginkgo biloba has shown unclear benefit in achieving important treatment goals for MCI,²⁷ and it increases the risk of bleeding in the elderly. Vitamin E was evaluated in one study and did not slow progression to dementia.²⁸

STAYING HEALTHY AND ACTIVE MAY HELP

We recommend optimizing vascular risk factors such as diabetes, blood pressure, smoking, and lipid levels in managing MCI, given that uncontrolled vascular risk factors may lead to progression to dementia. However, we can point to no research to support this recommendation.

Cognitive rehabilitation involves training in deficient domains and developing strategies to compensate for deficits. Different interventions are used, including computerized simulation exercises, memory aids, organizational techniques, personal digital assistants, crossword puzzles, mind games, and other mentally engaging activities.²⁹

Increasing physical activity is another aspect of treatment. Some studies have shown that it improves cognitive performance in MCI, at least in the short term.^30,31

Optimizing mood and emotions is also important. If present, depression should be identified and optimally treated. Social activity can be useful and leads to less emotional stress and to better coping mechanisms.

A multidisciplinary approach may help patients and may also help relieve the burden on the caregiver. Periodic reassessment of cognitive and functional symptoms may be warranted.

Maintaining disease-specific registries of patients who have MCI may be useful to longitudinally follow patients and ensure that they get the care they need.

PROGNOSIS VARIES

MCI is a heterogeneous condition that often does not predictably progress to dementia. Patients and families should be told that having MCI does not mean that the patient will necessarily get dementia.

Several studies have shown that the annual risk of progression to dementia for patients with MCI is 5% to 10% in community-dwelling populations and up to 15% in specialty-clinic patients.^24,32 In comparison, the incidence of dementia in the general elderly population is 1% to 3% per year.

On the other hand, a number of studies show that MCI improves significantly in up to 15% to 40% of patients and sometimes reverts to a normal cognitive state.^33,34 But prospective studies of patients with clinically diagnosed MCI usually find a low rate of reversion to a normal state.^35,36 Many are short-term follow-up studies of different populations, making generalizations difficult.¹⁴

Patients with impairment in instrumental activities of daily living may be more likely to have nonreversible MCI and may be at higher risk of progressing to dementia.³⁷

PATIENT AND FAMILY EDUCATION AND FOLLOW-UP CONSIDERATIONS

Caregiver education and stress management are important components of managing patients with MCI. Formally assessing caregiver stress is useful. Steps to prevent caregiver burnout include making use of respite care, counseling, education, and community resources such as adult day care and those offered by the Alzheimer’s Association.

Clinicians should follow patients with MCI closely to evaluate progression, address specific concerns, minimize risks, emphasize healthy habits, manage concurrent illnesses, and evaluate management.

Functional status, as demonstrated by activities of daily living, is the most important determinant of progression of MCI to dementia and should be evaluated at each visit. Repeat cognitive testing should be done on patients who have significant loss of functional status. Changes in work habits also warrant further attention.

Patients diagnosed with MCI or those who have persistent cognitive concerns should be considered for neuropsychological evaluation after 1 year to assess specific deficits and progression of cognitive impairment.

Finally, consideration should be given to current clinical research, and referrals should be made to research centers that focus on MCI management and treatment.

As our population ages, people are thinking more about preserving their quality of life, especially with regard to maintaining their cognitive and functional abilities. Older patients and caregivers often raise concerns about cognitive issues to their primary care providers: many patients have memory complaints, are worried about whether these are merely part of normal aging or symptoms of early dementia, and want strategies to forestall the progression of cognitive impairment.

Mild cognitive impairment (MCI) is a heterogeneous syndrome that in some cases represents a transition between normal aging and dementia. However, this condition is not yet well understood. Although some patients progress to dementia, others remain stable, or even improve. This article will review the current definitions and the underlying physiology of MCI, as well as diagnostic and management strategies.

See related editorial

COGNITIVE CHANGES OCCUR WITH NORMAL AGING

Cognition is defined as a means of acquiring and processing information about ourselves and our world. It includes memory as well as other domains such as attention, visuospatial skills, mental processing speed, language, and executive function. Cognitive abilities typically peak between ages 30 and 40, plateau in our 50s and 60s, and decline in our late 70s.

With age come detectable changes in the brain: brain weight declines by 10% by age 80, blood flow diminishes, neurons are lost throughout life, and nerve conduction slows. Despite these changes, the brain has a great deal of functional reserve capacity.

Table 1 compares the signs of normal aging, MCI, and dementia. Normally, cognitive abilities decline gradually with age without affecting overall function or activities of daily living. Even in normal aging, the processing of new information (new learning) is reduced. Mental processing becomes less efficient and slower. Visuospatial skills gradually decline, recall slows, and ultimately, the speed of performance slows as well. Additionally, distractibility increases. On the other hand, normal aging does not affect recognition, intelligence, or long-term memory.¹

The line between the normal effects of aging on cognition and true pathologic cognitive decline is blurry. In a busy clinical practice, it is often difficult to determine whether problems with memory and cognition that elderly patients and their family members describe represent true pathologic decline. In general, the clinical presentation of MCI is more profound than that of age-associated cognitive impairment: whereas normal aging may involve forgetting names and words and misplacing things, MCI frequently involves forgetting conversations, information that one would ordinarily remember, appointments, and planned events.

BETWEEN NORMAL AGING AND DEMENTIA

MCI is a transitional state between normal cognition and dementia. But the course is not inevitably downward: on follow-up, patients with MCI may be better, stable, or worse (see PROGNOSIS VARIES, below).

On autopsy studies, the brains of people with MCI appear intermediate between normal brains and brains of people with Alzheimer-type dementia, which have neurofibrillary tangles, amyloid senile plaques, and neuronal degeneration.

Definitions of MCI vary

True cognitive decline that is more profound than normal aging was named and defined differently in different studies, making comparisons difficult. The concept of MCI arose from the term “benign senescent forgetfulness,” used by Kral in 1962.² Other early terms include “cognitive impairment no dementia,” “memory impairment,” “mild cognitive disorder,” and “mild neurocognitive disorder.”^3,4

MCI was first defined as a precursor to Alzheimer dementia. The term later described a sometimes reversible but abnormal state. It is a heterogeneous syndrome in terms of etiology, incidence, prevalence, presentation, and overall prognosis.

Most recently, MCI has been defined as^5,6:

• Subjective memory complaints, preferably qualified by another person
• Memory impairment, with consideration for age and education
• Preserved general cognitive function
• Intact activities of daily living
• Absence of overt dementia.

MCI may arise from vascular, neurodegenerative, traumatic, metabolic, psychiatric, and other underlying medical disorders.^7–9

The prevalence of MCI is difficult to determine because of the various definitions, populations studied (eg, clinic-based vs community-dwelling), and evaluation techniques. Published rates vary from 2% to 4% in all patients to 10% to 20% in the elderly. Incidence rates in the elderly vary from 14 to 75 per 1,000 patient-years.^10–14

EARLY RECOGNITION ALLOWS PROMPT EVALUATION AND PLANNING

Pathologic cognitive decline is best detected early, for many reasons. Early recognition and intervention may help delay further decline. Establishing a diagnosis can also lessen family and caregiver stress and misunderstanding. Education of caregivers is important so that they can prepare for likely behavioral changes and plan for future care. Advance care planning, including advance directives, power of attorney, and designation of proxy for decision-making, is extremely important and is best considered before cognitive impairment becomes severe.

The diagnosis of MCI also provides the opportunity to assess safety concerns related to driving, working, medication compliance, the home environment, and firearms. Because patients with MCI are still highly functional, these issues need not be fully evaluated and should be handled on a case-by-case basis, depending on concerns raised. For example, if depression is an active concern, firearms safety should be addressed.

MEMORY LOSS MAY NOT BE THE PRIMARY CONCERN

MCI is categorized into two types based on whether memory loss is the primary cognitive deficit.

The amnestic type predominantly involves memory problems and is more common. Generally, several years elapse between initial memory concerns and a clinical diagnosis of MCI. Patients with amnestic MCI that progresses to dementia are more likely to develop Alzheimer disease.^2,15

Nonamnestic types involve domains of cognition other than memory, such as executive function, attention, visuospatial ability, and language. Nonamnestic MCI can be subcategorized through extensive neuropsychological evaluation as involving single or multiple impaired domains.^16,17 Such categorization is particularly important in determining prognosis, as patients with involvement of multiple domains are at higher risk of progressing to dementia.

Patients with nonamnestic MCI who progress to dementia are more likely to have non-Alzheimer types of dementia, such as Lewy body dementia and frontotemporal dementias.¹⁰

HISTORY SHOULD FOCUS ON FUNCTION, MEDICATIONS, AND DEPRESSION

Cognitive impairment should be clinically evaluated within the context of cognition, function, and behavior. Clinicians should focus on the time course of cognitive concerns, the specifics of the concerns, and their impact on day-to-day living and functioning. In assessing functional capacity, it is important to determine the level of assistance the patient needs to perform specific activities of daily living and instrumental activities of daily living (ie, the more advanced skills needed to live independently) (Table 2).

A thorough history includes consideration of baseline education, intellect, and previous learning disabilities; sensory impairments with emphasis on sight and hearing impairments; uncontrolled pain; head trauma; sleep disorders; concurrent medical and psychosocial illnesses such as depression and anxiety; substance abuse; and polypharmacy.

Depression, delirium, and the use of anticholinergic drugs are particularly important to evaluate, as these can result in cognitive deficits associated with MCI. The cognitive deficits may resolve with treatment or with stopping the drug.

Behavioral concerns such as wandering, agitation, and anger and sleep concerns, eating habits, and social etiquette are also important to evaluate.

PHYSICAL EVALUATION: RULE OUT REVERSIBLE CONDITIONS

The differential diagnosis of MCI includes delirium, depression, dementia, possibly reversible conditions affecting cognition (vitamin B₁₂ deficiency, hypothyroidism, effects of anticholinergic drugs), and uncommonly, central nervous system conditions (normal pressure hydrocephalus, subdural hematoma, tumor, stroke), and others (Table 3).¹⁸

A thorough physical examination should include neurologic, cardiovascular, hearing, and vision examinations, as well as an evaluation of functional status.

Laboratory studies. Although evidence is lacking to support a laboratory diagnostic workup for MCI, a selective evaluation including a comprehensive metabolic profile, complete blood count, thyroid studies, and a vitamin B₁₂ level can be useful. Occasionally, a treatable cause of impaired cognition such as vitamin B₁₂ deficiency or thyroid disease can be identified and resolved. A further comprehensive laboratory evaluation should be obtained if a patient progresses to dementia.

Imaging can be used in conjunction with other supportive evidence but should not be used solely to establish a diagnosis of MCI. Magnetic resonance imaging (MRI) can detect metastatic disease, normal pressure hydrocephalus, and subdural hematoma, in addition to traumatic, inflammatory, infectious, and vascular causes of cognitive impairment. MRI can also determine focal areas of atrophy; temporal lobe atrophy is a risk factor for progression to dementia.

Other studies. Structural MRI using techniques to evaluate the hippocampus, functional imaging, genetic testing for ApoE4 alleles, and biomarkers in cerebrospinal fluid are currently under evaluation to identify those at risk of progression to dementia. Recently published guidelines by the Alzheimer’s Association and the National Institute on Aging indicate that pathophysiologic findings in MCI that may predict future Alzheimer disease are meant to guide research and are not part of clinical practice at this time.¹⁹

COGNITIVE AND NEUROLOGIC TESTING IDENTIFIES DEFICITS

A number of global measures of cognition can be used in the office in clinical practice to help in evaluating significant cognitive concerns and to determine areas and severity of deficits at presentation. These include the Mini-Mental State Examination, the Montreal Cognitive Assessment, the Saint Louis University Mental Status, and many others (Table 4).²⁰

Caveats about interpreting the results: each of these tests has different sensitivities and specificities for detecting MCI. Also, we need to take into account the patient’s level of education, as highly educated people tend to do better on these tests.^21–23 It is important to note that some patients with MCI have normal results or only minimally abnormal results on these tests.

Neuropsychological testing is reserved for patients needing further evaluation, eg, those with atypical or complex cases, and those in whom the specific domains of cognition involved need to be identified. It can also provide additional insight into the contribution of depression to cognitive deficits. Neuropsychological testing is usually very time-intensive and requires patients to be able to perform complicated cognitive tasks. Not all patients are good candidates for this testing; sensory and motor impairments must be considered to determine if patients can adequately participate in testing. The cost of neuropsychological testing for MCI may not be covered by insurance and should be discussed with patients before referral. Specific concerns about cognitive problems that need further evaluation should be stated in the referral.

No one test should be used to make a diagnosis of MCI or dementia; clinical judgment is also necessary. The need for referral to a neurologist, geriatrician, or psychiatrist depends on the nature of the cognitive and behavioral concerns, the complexity of making a diagnosis, the need for further assessment of functional ability, and the need for evaluation of risk of progression to dementia.

MEDICATIONS HAVE LITTLE ROLE IN MANAGEMENT

No drug has yet been approved by the US Food and Drug Administration for treating MCI.

The acetylcholinesterase inhibitors donepezil (Aricept), galantamine (Razadyne), and rivastigmine (Exelon) have undergone clinical trials for treatment of MCI but have not been definitely shown to significantly reduce the risk of progression to dementia.²⁴

On the other hand, Diniz et al²⁵ performed a meta-analysis of the use of cholinesterase inhibitors in patients with MCI as a means of delaying the progression to Alzheimer disease.²⁵ They calculated that 15.4% of patients who received these drugs progressed to dementia, compared with 20.4% of those who received placebo, for a relative risk of 0.75 (95% confidence interval 0.66–0.87, P < .001). They concluded that the use of these drugs in patients with MCI “may attenuate the risk of progression” to Alzheimer disease and dementia.

In addition to not being approved for this indication and showing mixed evidence of efficacy, these drugs have well-known side effects such as diarrhea, nausea, vomiting, anorexia, and rhinitis, as well as significant but lesser-known side effects such as syncope, bradycardia, gastrointestinal bleeding, and vivid dreams.²⁶

Nevertheless, some patients with MCI, particularly those at high risk with amnestic MCI, may still want to try these medications. In these cases, the risks and possible benefits (or lack of them) should be reviewed thoroughly with the patient and family, and the discussion should be documented before starting therapy. The lowest starting dose of acetylcholinesterase inhibitor should be used to determine tolerability; generally, the dose is increased after 4 weeks to a maintenance dosage, with particular consideration of side effects.

Other agents have also been evaluated for MCI but have shown no evidence of benefit. Nonsteroidal anti-inflammatory drugs have not been found to either improve symptoms or delay progression to dementia. Ginkgo biloba has shown unclear benefit in achieving important treatment goals for MCI,²⁷ and it increases the risk of bleeding in the elderly. Vitamin E was evaluated in one study and did not slow progression to dementia.²⁸

STAYING HEALTHY AND ACTIVE MAY HELP

We recommend optimizing vascular risk factors such as diabetes, blood pressure, smoking, and lipid levels in managing MCI, given that uncontrolled vascular risk factors may lead to progression to dementia. However, we can point to no research to support this recommendation.

Cognitive rehabilitation involves training in deficient domains and developing strategies to compensate for deficits. Different interventions are used, including computerized simulation exercises, memory aids, organizational techniques, personal digital assistants, crossword puzzles, mind games, and other mentally engaging activities.²⁹

Increasing physical activity is another aspect of treatment. Some studies have shown that it improves cognitive performance in MCI, at least in the short term.^30,31

Optimizing mood and emotions is also important. If present, depression should be identified and optimally treated. Social activity can be useful and leads to less emotional stress and to better coping mechanisms.

A multidisciplinary approach may help patients and may also help relieve the burden on the caregiver. Periodic reassessment of cognitive and functional symptoms may be warranted.

Maintaining disease-specific registries of patients who have MCI may be useful to longitudinally follow patients and ensure that they get the care they need.

PROGNOSIS VARIES

MCI is a heterogeneous condition that often does not predictably progress to dementia. Patients and families should be told that having MCI does not mean that the patient will necessarily get dementia.

Several studies have shown that the annual risk of progression to dementia for patients with MCI is 5% to 10% in community-dwelling populations and up to 15% in specialty-clinic patients.^24,32 In comparison, the incidence of dementia in the general elderly population is 1% to 3% per year.

On the other hand, a number of studies show that MCI improves significantly in up to 15% to 40% of patients and sometimes reverts to a normal cognitive state.^33,34 But prospective studies of patients with clinically diagnosed MCI usually find a low rate of reversion to a normal state.^35,36 Many are short-term follow-up studies of different populations, making generalizations difficult.¹⁴

Patients with impairment in instrumental activities of daily living may be more likely to have nonreversible MCI and may be at higher risk of progressing to dementia.³⁷

PATIENT AND FAMILY EDUCATION AND FOLLOW-UP CONSIDERATIONS

Caregiver education and stress management are important components of managing patients with MCI. Formally assessing caregiver stress is useful. Steps to prevent caregiver burnout include making use of respite care, counseling, education, and community resources such as adult day care and those offered by the Alzheimer’s Association.

Clinicians should follow patients with MCI closely to evaluate progression, address specific concerns, minimize risks, emphasize healthy habits, manage concurrent illnesses, and evaluate management.

Functional status, as demonstrated by activities of daily living, is the most important determinant of progression of MCI to dementia and should be evaluated at each visit. Repeat cognitive testing should be done on patients who have significant loss of functional status. Changes in work habits also warrant further attention.

Patients diagnosed with MCI or those who have persistent cognitive concerns should be considered for neuropsychological evaluation after 1 year to assess specific deficits and progression of cognitive impairment.

Finally, consideration should be given to current clinical research, and referrals should be made to research centers that focus on MCI management and treatment.

A 68-year-old woman presented to the emergency department with constipation and abdominal pain 13 days after left hip arthroplasty. Abdominal computed tomography (CT) revealed possible bowel obstruction, left renal infarction, and a thrombus in the abdominal aorta near the left renal artery (Figure 1).

Because of the aortic thrombus, anticoagulation with intravenous heparin and with warfarin (Coumadin) was started. Three days later, her platelet count decreased to 54 × 10⁹/L (reference range 150–400), her serum creatinine rose to 2.38 mg/dL (0.70–1.40), sodium was stable at 131 mmol/L (132–148), and potassium was 4.1 mmol/L (3.5–5.0).

Physical examination revealed a temperature of 100.2°F (37.9°C), blood pressure 110/59 mm Hg, pulse 100 bpm, and no abdominal pain on palpation. Renal ultrasonography revealed a mass (2.4 × 2.1 × 3.6 cm) above the right kidney. Abdominal CT without contrast showed bilateral high-density adrenal masses (Figure 2).

Q: Which is the most likely diagnosis?

• Metastasis to the adrenal glands
• Adrenal adenoma
• Pheochromocytoma
• Adrenal cortical carcinoma
• Adrenal gland hemorrhage

A: The correct diagnosis is adrenal gland hemorrhage.

Acute or subacute hemorrhage typically results in an oval hyperdense mass with an attenuation of 50 to 90 Hounsfield units (H) on noncontrast CT (Figure 2),^1,2 and this attenuation does not increase with the use of contrast.²

In contrast, adrenal cortical carcinoma typically appears as a large heterogeneous mass, with some lesions demonstrating central necrosis or calcification. Pheochromocytoma is well defined, with intense enhancement after contrast is given. Adenoma is usually homogenous, with well-defined margins. Many adenomas have increased intracytoplasmic lipid content and, therefore, will have an attenuation of less than 10 H or will demonstrate rapid washout of contrast. Metastasis to the adrenal gland may not have a characteristic radiographic appearance but typically has a slower contrast washout rate than adenoma.¹

This patient’s initial abdominal image showed normal-appearing adrenal glands, thus making adenoma, adrenal metastasis, pheochromocytoma, and adrenal cortical carcinoma unlikely.

The patient’s baseline cortisol level, a random afternoon reading, was 0.4 μg/dL (reference range 3.4–26.9), and a 1-hour cortrosyn-stimulated cortisol was 0 μg/dL, which is diagnostic of primary adrenal insufficiency in the context of this clinical setting. She received hydrocortisone and fludrocortisone, and 8 am cortisol measurements 3 weeks and 5 months later, after the patient was off hydrocortisone for 24 hours, remained undetectable.

The diagnosis of adrenal hemorrhage can be difficult because the symptoms can be nonspecific and attributable to other clinical factors. In a review of 141 patients with adrenal hemorrhage,³ only 19% of patients with bilateral adrenal hemorrhage developed hypotension with a systolic blood pressure less than 90 mm Hg, only 15% developed hyponatremia (sodium < 130 mmol/L), and only 24% developed hyperkalemia (potassium > 5 mmol/L).³

If unrecognized, adrenal insufficiency from adrenal hemorrhage is fatal. Abdominal CT or magnetic resonance imaging can diagnose adrenal hemorrhage. Adrenal function may recover (although it did not in this patient), and a morning cortisol level should be obtained to reevaluate adrenal function.³

Risk factors for adrenal hemorrhage include anticoagulation therapy, sepsis, surgery, hypotension, and coagulopathy as seen in heparininduced thrombocytopenia and disseminated intravascular coagulation. This patient had coagulopathy, as evidenced by her abdominal aortic thrombus, for which she was placed on anticoagulation. Patients without these risk factors for adrenal gland hemorrhage should be investigated for an underlying adrenal neoplasm or cyst.²

A 68-year-old woman presented to the emergency department with constipation and abdominal pain 13 days after left hip arthroplasty. Abdominal computed tomography (CT) revealed possible bowel obstruction, left renal infarction, and a thrombus in the abdominal aorta near the left renal artery (Figure 1).

Because of the aortic thrombus, anticoagulation with intravenous heparin and with warfarin (Coumadin) was started. Three days later, her platelet count decreased to 54 × 10⁹/L (reference range 150–400), her serum creatinine rose to 2.38 mg/dL (0.70–1.40), sodium was stable at 131 mmol/L (132–148), and potassium was 4.1 mmol/L (3.5–5.0).

Physical examination revealed a temperature of 100.2°F (37.9°C), blood pressure 110/59 mm Hg, pulse 100 bpm, and no abdominal pain on palpation. Renal ultrasonography revealed a mass (2.4 × 2.1 × 3.6 cm) above the right kidney. Abdominal CT without contrast showed bilateral high-density adrenal masses (Figure 2).

Q: Which is the most likely diagnosis?

• Metastasis to the adrenal glands
• Adrenal adenoma
• Pheochromocytoma
• Adrenal cortical carcinoma
• Adrenal gland hemorrhage

A: The correct diagnosis is adrenal gland hemorrhage.

Acute or subacute hemorrhage typically results in an oval hyperdense mass with an attenuation of 50 to 90 Hounsfield units (H) on noncontrast CT (Figure 2),^1,2 and this attenuation does not increase with the use of contrast.²