Cleveland Clinic Journal of Medicine

Alzheimer disease (AD) is a degenerative disorder characterized by a gradual deterioration in memory. In its clinically overt stages, obvious signs of neural degeneration on magnetic resonance imaging (MRI) appear as global cerebral atrophy. Even in the earliest stages of the disease, regional cell loss can be observed, particularly in the mesial temporal lobe regions, specifically the hippocampus and entorhinal cortex.^1–3

MRI is used primarily as a diagnostic tool to rule out conditions other than AD. However, MRI may be useful in understanding whether the underlying processes that are associated with these cognitive changes can be attributed to general or specific effects of the disease process. Volumetric changes observed with MRI in the hippocampal region have been correlated with disease progression^4,5 and predict development of AD in individuals with isolated memory impairment,⁶ suggesting that neuroimaging quantification may serve as a useful measure of brain integrity in patients with AD.

New treatments for AD are emerging, and assessing their efficacy is of critical importance. The rationale for testing statin drugs as a therapy for AD was bolstered by ever-mounting preclinical animal and human data suggesting that elevated circulating cholesterol exacerbates AD-like pathology and that statin treatment, in part, reverses the effect of cholesterol. This article surveys current evidence on the association between cholesterol and AD as well as between statin use and AD risk. We conclude by focusing on results from the first clinical investigation of statin therapy in patients with AD and present new results of a substudy of this trial examining the morphologic effects of statin therapy in AD patients.

LINK BETWEEN CHOLESTEROL AND AD

Early epidemiologic surveys suggested an association between a high-fat/high-cholesterol diet and increased risk of AD,^7–10 and this suggestion has been supported by more recent investigations.^11,12 Cholesterol levels are increased in the blood of AD patients,^7,13–16 and increased cholesterol has been observed in the AD brain as a function of the apolipoprotein E allotype.^8,17

Numerous clinical studies suggest a link between elevated cholesterol and increased risk of AD,^17–23 with one study reporting a threefold increase in the risk of AD with elevated serum cholesterol, even after adjusting for age and presence of the apolipoprotein E4 allele.¹⁹ Another study indicates that persistently elevated cholesterol levels in midlife increase the risk of AD.²³ A retrospective analysis of the Framingham Study suggested, however, that there is no relationship between total cholesterol levels and risk of incident AD.²⁴ A more recent report indicated that language performance in elderly subjects without dementia declined faster among those individuals with higher cholesterol levels, but this effect did not remain significant after accounting for multiple comparisons.²⁵ In contrast, the Three-City Study, a population-based cohort investigation of 9,294 subjects in France, demonstrated a significant increase in the risk of dementia among subjects who had hyperlipidemia (odds ratio [OR] = 1.43; 95% confidence interval [CI], 1.03 to 1.99).¹²

STATIN USE AND RISK OF AD

The preponderance of clinical data suggests that statin therapy may reduce the risk of AD later in life. Since the initial epidemiologic investigation assessing the effect of statin use on later risk of AD in the elderly, there have been 13 additional studies; all but two of these studies have reported benefit with cholesterol-lowering therapy.

In the two earliest epidemiologic studies, Wolozin et al demonstrated benefit with the use of lovastatin and pravastatin, but not with simvastatin or non-statin therapy,²⁶ and Jick et al showed benefit associated with cholesterol-lowering therapy, but not specifically with statin use.²⁷ Five epidemiologic studies published in 2002 suggested that prior statin use reduced the risk of dementia or AD.^28–32 Meta-analysis of these first seven retrospective studies suggested a significant reduction in the risk of later cognitive impairment with statin use (relative risk = 0.43; 95% CI, 0.31 to 0.62), but the risk reduction with lipid-lowering agents collectively (not just statins) was not statistically significant.³³

In 2004, Zamrini et al reported a 39% reduction in the risk of AD in statin users compared with nonusers (OR = 0.61; 95% CI, 0.42 to 0.87).³⁴ That same year, Li et al suggested that there was no association between statin use and a reduced incidence of probable AD using a time-dependent proportional hazards model, but if the data were analyzed (inappropriately) as a case-control study, a significant protective effect was identified.³⁵

Data from the Cache County Study cohort demonstrated no significant reduction in the risk of AD with statin use but allowed for the possibility that some benefit could be provided with longer-term statin therapy.³⁶ In constrast, the Three-City Study of 9,294 individuals in France identified a significant reduction in the risk of AD with statin use (OR = 0.61; 95% CI, 0.41 to 0.91).¹² Rea et al reported that prior statin use did not decrease the risk of dementia or AD, but when they included in their analysis individuals currently using a statin, there was a significant reduction in the hazard ratios for AD and for all-cause dementia.³⁷ The two most recent epidemiologic studies both suggest that statin therapy slows cognitive decline in AD.^38,39

COGNITIVE PERFORMANCE AND STATIN USE

A retrospective cohort study that assessed intelligence and cognition at a young age and again when subjects were in their 80s indicated that statin use had a significant beneficial effect on cognitive ability.⁴⁰

In contrast, two very large prospective studies published in 2002 suggested that statins produce no positive effect on cognition in younger individuals at risk for heart disease.^41,42 The Prospective Study of Pravastatin in the Elderly at Risk (PROSPER) found that the mean Mini-Mental State Examination (MMSE) score, which was assessed only at subjects’ last on-treatment clinical visit, was comparable between the study’s pravastatin and placebo groups.⁴¹ Likewise, the Medical Research Council/British Heart Foundation (MRC/BHF) Heart Protection Study, which used the Telephone Interview for Cognitive Status questionnaire at the end of the investigation, reported that simvastatin had no positive effect on cognitive performance compared with placebo, but this finding was obtained in the absence of baseline data.⁴² Given the limited cognitive assessments performed in these two studies, no firm conclusions should be drawn.

A more recent prospective comparison of atorvastatin and placebo in younger subjects did include baseline and follow-up assessment of cognitive function, and it identified significantly superior performance in the statin-treated population on the MMSE and on tests of attention, psychomotor speed, mental flexibility, working memory, and memory retrieval.⁴³

The initial clinical investigation of statin therapy in patients with AD—the Alzheimer’s Disease Cholesterol-Lowering Treatment (ADCLT) trial—involved atorvastatin.⁴⁴ Patients with mild to moderate AD were randomized to either placebo or 80 mg/day of atorvastatin for a 1-year period. Evaluable data were available for 63 patients (32 in the atorvastatin group, 31 in the placebo group). End points included the change in performance on the following measures:

MMSE

• Alzheimer’s Disease Assessment Scale–cognitive subscale (ADAS-cog)
• Neuropsychiatric Inventory Caregiver Distress Scale (NPI)
• Clinical Global Impression of Change scale (CGIC)
• Alzheimer’s Disease Cooperative Study–Activities of Daily Living Inventory (ADCS-ADL)
• Geriatric Depression Scale (GDS).

Cognitive results

In the setting of continued cholinesterase inhibitor use, atorvastatin provided significant benefit on the ADAS-cog at 26 weeks compared with placebo (P = .003) and marginally significant benefit at 1 year (P = .055) while producing a trend for benefit on the CGIC and NPI and a statistically significant improvement on the GDS after 1 year of active treatment.⁴⁴ The observed benefit on the MMSE with atorvastatin versus placebo did not reach statistical significance, and no discernible difference was observed on the ADCS-ADL.⁴⁴ In contrast, a significant difference in the slope of deterioration on the MMSE and the GDS in the atorvastatin group versus the placebo group suggested disease modification.⁴⁵

Blood test results

Secondary analysis and initial morphometric substudy

Secondary assessment indicated that the subjects who garnered the greatest benefit from atorvastatin therapy in terms of their 6-month ADAS-cog score were those who had higher cholesterol levels at trial entry, those who harbored the apolipoprotein E4 allele, and those who were less affected by AD at trial entry (ie, with higher entry MMSE scores).⁴⁷

In an ADCLT substudy using new voxel-based morphometry techniques, we quantitatively assessed gray matter density in 15 ADCLT trial participants and compared it with density findings in 15 normal elderly controls.⁴⁸ Regional reductions in gray matter density were observed in the AD patients compared with the controls. Large differences in gray matter concentration were observed bilaterally in the temporal lobe. The anterior cingulate, right superior temporal, left superior frontal, and posterior cingulate regions also showed significantly decreased gray matter density in the AD patients compared with the controls. A significant relationship was observed between gray matter density and ADAS-cog error scores—ie, more severe levels of cognitive impairment correlated with reduced gray matter density.⁴⁸

PILOT SUBSTUDY OF ADCLT: ASSESSING MORPHOLOGIC CHANGES WITH STATIN THERAPY

Eleven of the 15 ADCLT trial participants from the above morphometric substudy returned for MRI assessment after 1 year of treatment with either atorvastatin or placebo. We report here the comparative effects of atorvastatin and placebo on hippocampal volume and the relationship with cognitive performance.

Participants

Subjects were participating in the ADCLT trial, an investigator-initiated, double-blind, placebo-controlled study. Neuroimaging was performed at the Barrow Neurological Institute, Phoenix, AZ, for a subset of the participants in the trial (n = 11) as a pilot study to examine neural changes associated with atorvastatin therapy.

Each patient underwent screening, assignment to either atorvastatin 80 mg/day or placebo, and medical and cognitive assessment at Sun Health Research Institute, Sun City, AZ, prior to imaging at Barrow. All  patients met Diagnostic and Statistical Manual of Mental Disorders, fourth edition, criteria for dementia as well as NINCDS-ADRDA criteria for probable AD. Each patient was free of significant psychiatric and neurological history and had a score of 4 or less on the Hachinski Modified Ischemia Scale. All MRIs were reviewed by a neuroradiologist to ensure that there was no evidence of stroke or cortical or lacunar infarcts.

Both sites’ institutional review boards approved this project, and all subjects gave written informed consent.

Cognitive assessment

A primary efficacy measure used in the parent study was the ADAS-cog,⁴⁹ and the MMSE⁵⁰ was a secondary measure. Change scores were determined by comparing values obtained at baseline, prior to randomization to treatment with either atorvastatin or placebo, and after 1 year of treatment. MMSE scores were obtained at the same session as the ADAS-cog scores. Cognitive assessments were obtained within 2 weeks prior to MRI.

Image acquisition

All participants underwent imaging on a single 1.5­tesla GE scanner at Barrow Neurological Institute. Imaging was conducted both prior to treatment randomization and again after 1 year of treatment. Images of the whole brain were collected using a coronal SPGR (spoiled gradient) T₁-weighted, three-dimensional acquisition with the following parameters:

• Number of acquisitions = 1
• Repetition time = 23 msec
• Echo time = 8 msec
• Flip angle = 35 degrees
• Bandwidth = 12.5 kHz
• Slice thickness = 1.5 mm or 1.9 mm
• 0 skip between slices
• In-plane resolution = 0.9375 x 0.9375.

Hippocampal volumetrics

All imaging analysis was performed within the Analysis of Functional Neuroimages (AFNI) package.⁵¹ We traced the outline of the hippocampus using the three-dimensional SPGR images. The hippocampi were visualized in all three planes, landmarked in the coronal and sagittal planes, and drawn in the coronal plane. We employed the guidelines of Insausti et al⁵² and Machulda et al⁵³ to define the hippocampal boundaries. First we defined the anterior boundary by observing the white matter band and/or the cerebrospinal fluid space between the amygdala and hippocampus in the sagittal plane. The posterior aspect of the posterior region was initially landmarked in the sagittal plane by locating the posterior edge of the hippocampus and then checking in the coronal plane to ensure that the fornices were completely visualized. Volumes were calculated by importing the extracted hippocampi into MATLAB to measure the volumes.

Mean differences between the atorvastatin and placebo groups were evaluated using two-tailed Student t tests. Correlation between changes in cognitive measures and changes in the hippocampal volume for the total population and for the treatment groups was determined using Pearson’s r coefficient. Significance was defined as a P value less than .05; a P value between .05 and .10 was deemed a trend.

Results

In contrast to other studies,^54–58 we found in this pilot study that right hippocampal volume was slightly less than left hippocampal volume (2,015 ± 141 mm³ vs 2,135 ± 183 mm³).

Mean changes in performance on the ADAS-cog and MMSE were less pronounced in the atorvastatin group than in the placebo group, but not significantly so (Table 1). However, there was a trend toward superiority in the atorvastatin group on performance on the free word-recall subscale of the ADAS-cog.

No significant correlations were found between change in cognitive performance and change in hip-pocampal volume.

DISCUSSION

The preponderance of evidence clearly indicates that hippocampal volume is reduced in patients with AD compared with individuals with normal cognitive ability for their age. There is also evidence indicating that as cognitive performance deteriorates in AD patients, there are concurrent further reductions in hippocampal volume.⁵⁴ Many studies reported that there was no significant volume difference between the right and left hippocampi, but most suggested that the left hip-pocampus was slightly smaller than the right.^54–58 We identified no significant difference in volume between the sides, but we did find that the right hippocampus was smaller than the left in a very limited population of subjects with mild to moderate AD.

The major finding of this pilot study flies in the face of conventional wisdom in that there seems to be significant shrinkage of the right hippocampus with atorvastatin therapy compared with placebo in a randomized AD treatment trial that demonstrated clinical benefit with atorvastatin therapy.⁴⁴ A similar finding was reported from the beta-amyloid immunization (AN1792) treatment trial in AD.⁵⁹ In that study the active immunization was associated with significant clinical benefit, reduced beta-amyloid load, and reduced hippocampal volume.⁵⁹ The authors suggested that removal of beta-amyloid and/or other protein constituents from the tissue might have caused a “fluid shift” out of the tissue, resulting in shrinkage.

Based on our previous finding of reduced brain tissue density in AD patients compared with age-matched normal controls,⁴⁸ an alternative explanation can be proposed. Neuronal loss in the hippocampus may be accompanied by increased fluid balance (reduced density) in an attempt to retain the previous volume at the expense of function. Accordingly, as the hippocampus shrinks, it approaches a more normal density for the remaining neuronal complement, and cognitive function improves.

Alzheimer disease (AD) is a degenerative disorder characterized by a gradual deterioration in memory. In its clinically overt stages, obvious signs of neural degeneration on magnetic resonance imaging (MRI) appear as global cerebral atrophy. Even in the earliest stages of the disease, regional cell loss can be observed, particularly in the mesial temporal lobe regions, specifically the hippocampus and entorhinal cortex.^1–3

MRI is used primarily as a diagnostic tool to rule out conditions other than AD. However, MRI may be useful in understanding whether the underlying processes that are associated with these cognitive changes can be attributed to general or specific effects of the disease process. Volumetric changes observed with MRI in the hippocampal region have been correlated with disease progression^4,5 and predict development of AD in individuals with isolated memory impairment,⁶ suggesting that neuroimaging quantification may serve as a useful measure of brain integrity in patients with AD.

New treatments for AD are emerging, and assessing their efficacy is of critical importance. The rationale for testing statin drugs as a therapy for AD was bolstered by ever-mounting preclinical animal and human data suggesting that elevated circulating cholesterol exacerbates AD-like pathology and that statin treatment, in part, reverses the effect of cholesterol. This article surveys current evidence on the association between cholesterol and AD as well as between statin use and AD risk. We conclude by focusing on results from the first clinical investigation of statin therapy in patients with AD and present new results of a substudy of this trial examining the morphologic effects of statin therapy in AD patients.

LINK BETWEEN CHOLESTEROL AND AD

Early epidemiologic surveys suggested an association between a high-fat/high-cholesterol diet and increased risk of AD,^7–10 and this suggestion has been supported by more recent investigations.^11,12 Cholesterol levels are increased in the blood of AD patients,^7,13–16 and increased cholesterol has been observed in the AD brain as a function of the apolipoprotein E allotype.^8,17

Numerous clinical studies suggest a link between elevated cholesterol and increased risk of AD,^17–23 with one study reporting a threefold increase in the risk of AD with elevated serum cholesterol, even after adjusting for age and presence of the apolipoprotein E4 allele.¹⁹ Another study indicates that persistently elevated cholesterol levels in midlife increase the risk of AD.²³ A retrospective analysis of the Framingham Study suggested, however, that there is no relationship between total cholesterol levels and risk of incident AD.²⁴ A more recent report indicated that language performance in elderly subjects without dementia declined faster among those individuals with higher cholesterol levels, but this effect did not remain significant after accounting for multiple comparisons.²⁵ In contrast, the Three-City Study, a population-based cohort investigation of 9,294 subjects in France, demonstrated a significant increase in the risk of dementia among subjects who had hyperlipidemia (odds ratio [OR] = 1.43; 95% confidence interval [CI], 1.03 to 1.99).¹²

STATIN USE AND RISK OF AD

The preponderance of clinical data suggests that statin therapy may reduce the risk of AD later in life. Since the initial epidemiologic investigation assessing the effect of statin use on later risk of AD in the elderly, there have been 13 additional studies; all but two of these studies have reported benefit with cholesterol-lowering therapy.

In the two earliest epidemiologic studies, Wolozin et al demonstrated benefit with the use of lovastatin and pravastatin, but not with simvastatin or non-statin therapy,²⁶ and Jick et al showed benefit associated with cholesterol-lowering therapy, but not specifically with statin use.²⁷ Five epidemiologic studies published in 2002 suggested that prior statin use reduced the risk of dementia or AD.^28–32 Meta-analysis of these first seven retrospective studies suggested a significant reduction in the risk of later cognitive impairment with statin use (relative risk = 0.43; 95% CI, 0.31 to 0.62), but the risk reduction with lipid-lowering agents collectively (not just statins) was not statistically significant.³³

In 2004, Zamrini et al reported a 39% reduction in the risk of AD in statin users compared with nonusers (OR = 0.61; 95% CI, 0.42 to 0.87).³⁴ That same year, Li et al suggested that there was no association between statin use and a reduced incidence of probable AD using a time-dependent proportional hazards model, but if the data were analyzed (inappropriately) as a case-control study, a significant protective effect was identified.³⁵

Data from the Cache County Study cohort demonstrated no significant reduction in the risk of AD with statin use but allowed for the possibility that some benefit could be provided with longer-term statin therapy.³⁶ In constrast, the Three-City Study of 9,294 individuals in France identified a significant reduction in the risk of AD with statin use (OR = 0.61; 95% CI, 0.41 to 0.91).¹² Rea et al reported that prior statin use did not decrease the risk of dementia or AD, but when they included in their analysis individuals currently using a statin, there was a significant reduction in the hazard ratios for AD and for all-cause dementia.³⁷ The two most recent epidemiologic studies both suggest that statin therapy slows cognitive decline in AD.^38,39

COGNITIVE PERFORMANCE AND STATIN USE

A retrospective cohort study that assessed intelligence and cognition at a young age and again when subjects were in their 80s indicated that statin use had a significant beneficial effect on cognitive ability.⁴⁰

In contrast, two very large prospective studies published in 2002 suggested that statins produce no positive effect on cognition in younger individuals at risk for heart disease.^41,42 The Prospective Study of Pravastatin in the Elderly at Risk (PROSPER) found that the mean Mini-Mental State Examination (MMSE) score, which was assessed only at subjects’ last on-treatment clinical visit, was comparable between the study’s pravastatin and placebo groups.⁴¹ Likewise, the Medical Research Council/British Heart Foundation (MRC/BHF) Heart Protection Study, which used the Telephone Interview for Cognitive Status questionnaire at the end of the investigation, reported that simvastatin had no positive effect on cognitive performance compared with placebo, but this finding was obtained in the absence of baseline data.⁴² Given the limited cognitive assessments performed in these two studies, no firm conclusions should be drawn.

A more recent prospective comparison of atorvastatin and placebo in younger subjects did include baseline and follow-up assessment of cognitive function, and it identified significantly superior performance in the statin-treated population on the MMSE and on tests of attention, psychomotor speed, mental flexibility, working memory, and memory retrieval.⁴³

The initial clinical investigation of statin therapy in patients with AD—the Alzheimer’s Disease Cholesterol-Lowering Treatment (ADCLT) trial—involved atorvastatin.⁴⁴ Patients with mild to moderate AD were randomized to either placebo or 80 mg/day of atorvastatin for a 1-year period. Evaluable data were available for 63 patients (32 in the atorvastatin group, 31 in the placebo group). End points included the change in performance on the following measures:

MMSE

• Alzheimer’s Disease Assessment Scale–cognitive subscale (ADAS-cog)
• Neuropsychiatric Inventory Caregiver Distress Scale (NPI)
• Clinical Global Impression of Change scale (CGIC)
• Alzheimer’s Disease Cooperative Study–Activities of Daily Living Inventory (ADCS-ADL)
• Geriatric Depression Scale (GDS).

Cognitive results

In the setting of continued cholinesterase inhibitor use, atorvastatin provided significant benefit on the ADAS-cog at 26 weeks compared with placebo (P = .003) and marginally significant benefit at 1 year (P = .055) while producing a trend for benefit on the CGIC and NPI and a statistically significant improvement on the GDS after 1 year of active treatment.⁴⁴ The observed benefit on the MMSE with atorvastatin versus placebo did not reach statistical significance, and no discernible difference was observed on the ADCS-ADL.⁴⁴ In contrast, a significant difference in the slope of deterioration on the MMSE and the GDS in the atorvastatin group versus the placebo group suggested disease modification.⁴⁵

Blood test results

Secondary analysis and initial morphometric substudy

Secondary assessment indicated that the subjects who garnered the greatest benefit from atorvastatin therapy in terms of their 6-month ADAS-cog score were those who had higher cholesterol levels at trial entry, those who harbored the apolipoprotein E4 allele, and those who were less affected by AD at trial entry (ie, with higher entry MMSE scores).⁴⁷

In an ADCLT substudy using new voxel-based morphometry techniques, we quantitatively assessed gray matter density in 15 ADCLT trial participants and compared it with density findings in 15 normal elderly controls.⁴⁸ Regional reductions in gray matter density were observed in the AD patients compared with the controls. Large differences in gray matter concentration were observed bilaterally in the temporal lobe. The anterior cingulate, right superior temporal, left superior frontal, and posterior cingulate regions also showed significantly decreased gray matter density in the AD patients compared with the controls. A significant relationship was observed between gray matter density and ADAS-cog error scores—ie, more severe levels of cognitive impairment correlated with reduced gray matter density.⁴⁸

PILOT SUBSTUDY OF ADCLT: ASSESSING MORPHOLOGIC CHANGES WITH STATIN THERAPY

Eleven of the 15 ADCLT trial participants from the above morphometric substudy returned for MRI assessment after 1 year of treatment with either atorvastatin or placebo. We report here the comparative effects of atorvastatin and placebo on hippocampal volume and the relationship with cognitive performance.

Participants

Subjects were participating in the ADCLT trial, an investigator-initiated, double-blind, placebo-controlled study. Neuroimaging was performed at the Barrow Neurological Institute, Phoenix, AZ, for a subset of the participants in the trial (n = 11) as a pilot study to examine neural changes associated with atorvastatin therapy.

Each patient underwent screening, assignment to either atorvastatin 80 mg/day or placebo, and medical and cognitive assessment at Sun Health Research Institute, Sun City, AZ, prior to imaging at Barrow. All  patients met Diagnostic and Statistical Manual of Mental Disorders, fourth edition, criteria for dementia as well as NINCDS-ADRDA criteria for probable AD. Each patient was free of significant psychiatric and neurological history and had a score of 4 or less on the Hachinski Modified Ischemia Scale. All MRIs were reviewed by a neuroradiologist to ensure that there was no evidence of stroke or cortical or lacunar infarcts.

Both sites’ institutional review boards approved this project, and all subjects gave written informed consent.

Cognitive assessment

A primary efficacy measure used in the parent study was the ADAS-cog,⁴⁹ and the MMSE⁵⁰ was a secondary measure. Change scores were determined by comparing values obtained at baseline, prior to randomization to treatment with either atorvastatin or placebo, and after 1 year of treatment. MMSE scores were obtained at the same session as the ADAS-cog scores. Cognitive assessments were obtained within 2 weeks prior to MRI.

Image acquisition

All participants underwent imaging on a single 1.5­tesla GE scanner at Barrow Neurological Institute. Imaging was conducted both prior to treatment randomization and again after 1 year of treatment. Images of the whole brain were collected using a coronal SPGR (spoiled gradient) T₁-weighted, three-dimensional acquisition with the following parameters:

• Number of acquisitions = 1
• Repetition time = 23 msec
• Echo time = 8 msec
• Flip angle = 35 degrees
• Bandwidth = 12.5 kHz
• Slice thickness = 1.5 mm or 1.9 mm
• 0 skip between slices
• In-plane resolution = 0.9375 x 0.9375.

Hippocampal volumetrics

All imaging analysis was performed within the Analysis of Functional Neuroimages (AFNI) package.⁵¹ We traced the outline of the hippocampus using the three-dimensional SPGR images. The hippocampi were visualized in all three planes, landmarked in the coronal and sagittal planes, and drawn in the coronal plane. We employed the guidelines of Insausti et al⁵² and Machulda et al⁵³ to define the hippocampal boundaries. First we defined the anterior boundary by observing the white matter band and/or the cerebrospinal fluid space between the amygdala and hippocampus in the sagittal plane. The posterior aspect of the posterior region was initially landmarked in the sagittal plane by locating the posterior edge of the hippocampus and then checking in the coronal plane to ensure that the fornices were completely visualized. Volumes were calculated by importing the extracted hippocampi into MATLAB to measure the volumes.

Mean differences between the atorvastatin and placebo groups were evaluated using two-tailed Student t tests. Correlation between changes in cognitive measures and changes in the hippocampal volume for the total population and for the treatment groups was determined using Pearson’s r coefficient. Significance was defined as a P value less than .05; a P value between .05 and .10 was deemed a trend.

Results

In contrast to other studies,^54–58 we found in this pilot study that right hippocampal volume was slightly less than left hippocampal volume (2,015 ± 141 mm³ vs 2,135 ± 183 mm³).

Mean changes in performance on the ADAS-cog and MMSE were less pronounced in the atorvastatin group than in the placebo group, but not significantly so (Table 1). However, there was a trend toward superiority in the atorvastatin group on performance on the free word-recall subscale of the ADAS-cog.

No significant correlations were found between change in cognitive performance and change in hip-pocampal volume.

DISCUSSION

The preponderance of evidence clearly indicates that hippocampal volume is reduced in patients with AD compared with individuals with normal cognitive ability for their age. There is also evidence indicating that as cognitive performance deteriorates in AD patients, there are concurrent further reductions in hippocampal volume.⁵⁴ Many studies reported that there was no significant volume difference between the right and left hippocampi, but most suggested that the left hip-pocampus was slightly smaller than the right.^54–58 We identified no significant difference in volume between the sides, but we did find that the right hippocampus was smaller than the left in a very limited population of subjects with mild to moderate AD.

The major finding of this pilot study flies in the face of conventional wisdom in that there seems to be significant shrinkage of the right hippocampus with atorvastatin therapy compared with placebo in a randomized AD treatment trial that demonstrated clinical benefit with atorvastatin therapy.⁴⁴ A similar finding was reported from the beta-amyloid immunization (AN1792) treatment trial in AD.⁵⁹ In that study the active immunization was associated with significant clinical benefit, reduced beta-amyloid load, and reduced hippocampal volume.⁵⁹ The authors suggested that removal of beta-amyloid and/or other protein constituents from the tissue might have caused a “fluid shift” out of the tissue, resulting in shrinkage.

Based on our previous finding of reduced brain tissue density in AD patients compared with age-matched normal controls,⁴⁸ an alternative explanation can be proposed. Neuronal loss in the hippocampus may be accompanied by increased fluid balance (reduced density) in an attempt to retain the previous volume at the expense of function. Accordingly, as the hippocampus shrinks, it approaches a more normal density for the remaining neuronal complement, and cognitive function improves.

Alzheimer disease (AD) is a degenerative disorder characterized by a gradual deterioration in memory. In its clinically overt stages, obvious signs of neural degeneration on magnetic resonance imaging (MRI) appear as global cerebral atrophy. Even in the earliest stages of the disease, regional cell loss can be observed, particularly in the mesial temporal lobe regions, specifically the hippocampus and entorhinal cortex.^1–3

MRI is used primarily as a diagnostic tool to rule out conditions other than AD. However, MRI may be useful in understanding whether the underlying processes that are associated with these cognitive changes can be attributed to general or specific effects of the disease process. Volumetric changes observed with MRI in the hippocampal region have been correlated with disease progression^4,5 and predict development of AD in individuals with isolated memory impairment,⁶ suggesting that neuroimaging quantification may serve as a useful measure of brain integrity in patients with AD.

New treatments for AD are emerging, and assessing their efficacy is of critical importance. The rationale for testing statin drugs as a therapy for AD was bolstered by ever-mounting preclinical animal and human data suggesting that elevated circulating cholesterol exacerbates AD-like pathology and that statin treatment, in part, reverses the effect of cholesterol. This article surveys current evidence on the association between cholesterol and AD as well as between statin use and AD risk. We conclude by focusing on results from the first clinical investigation of statin therapy in patients with AD and present new results of a substudy of this trial examining the morphologic effects of statin therapy in AD patients.

LINK BETWEEN CHOLESTEROL AND AD

Early epidemiologic surveys suggested an association between a high-fat/high-cholesterol diet and increased risk of AD,^7–10 and this suggestion has been supported by more recent investigations.^11,12 Cholesterol levels are increased in the blood of AD patients,^7,13–16 and increased cholesterol has been observed in the AD brain as a function of the apolipoprotein E allotype.^8,17

Numerous clinical studies suggest a link between elevated cholesterol and increased risk of AD,^17–23 with one study reporting a threefold increase in the risk of AD with elevated serum cholesterol, even after adjusting for age and presence of the apolipoprotein E4 allele.¹⁹ Another study indicates that persistently elevated cholesterol levels in midlife increase the risk of AD.²³ A retrospective analysis of the Framingham Study suggested, however, that there is no relationship between total cholesterol levels and risk of incident AD.²⁴ A more recent report indicated that language performance in elderly subjects without dementia declined faster among those individuals with higher cholesterol levels, but this effect did not remain significant after accounting for multiple comparisons.²⁵ In contrast, the Three-City Study, a population-based cohort investigation of 9,294 subjects in France, demonstrated a significant increase in the risk of dementia among subjects who had hyperlipidemia (odds ratio [OR] = 1.43; 95% confidence interval [CI], 1.03 to 1.99).¹²

STATIN USE AND RISK OF AD

The preponderance of clinical data suggests that statin therapy may reduce the risk of AD later in life. Since the initial epidemiologic investigation assessing the effect of statin use on later risk of AD in the elderly, there have been 13 additional studies; all but two of these studies have reported benefit with cholesterol-lowering therapy.

In the two earliest epidemiologic studies, Wolozin et al demonstrated benefit with the use of lovastatin and pravastatin, but not with simvastatin or non-statin therapy,²⁶ and Jick et al showed benefit associated with cholesterol-lowering therapy, but not specifically with statin use.²⁷ Five epidemiologic studies published in 2002 suggested that prior statin use reduced the risk of dementia or AD.^28–32 Meta-analysis of these first seven retrospective studies suggested a significant reduction in the risk of later cognitive impairment with statin use (relative risk = 0.43; 95% CI, 0.31 to 0.62), but the risk reduction with lipid-lowering agents collectively (not just statins) was not statistically significant.³³

In 2004, Zamrini et al reported a 39% reduction in the risk of AD in statin users compared with nonusers (OR = 0.61; 95% CI, 0.42 to 0.87).³⁴ That same year, Li et al suggested that there was no association between statin use and a reduced incidence of probable AD using a time-dependent proportional hazards model, but if the data were analyzed (inappropriately) as a case-control study, a significant protective effect was identified.³⁵

Data from the Cache County Study cohort demonstrated no significant reduction in the risk of AD with statin use but allowed for the possibility that some benefit could be provided with longer-term statin therapy.³⁶ In constrast, the Three-City Study of 9,294 individuals in France identified a significant reduction in the risk of AD with statin use (OR = 0.61; 95% CI, 0.41 to 0.91).¹² Rea et al reported that prior statin use did not decrease the risk of dementia or AD, but when they included in their analysis individuals currently using a statin, there was a significant reduction in the hazard ratios for AD and for all-cause dementia.³⁷ The two most recent epidemiologic studies both suggest that statin therapy slows cognitive decline in AD.^38,39

COGNITIVE PERFORMANCE AND STATIN USE

A retrospective cohort study that assessed intelligence and cognition at a young age and again when subjects were in their 80s indicated that statin use had a significant beneficial effect on cognitive ability.⁴⁰

In contrast, two very large prospective studies published in 2002 suggested that statins produce no positive effect on cognition in younger individuals at risk for heart disease.^41,42 The Prospective Study of Pravastatin in the Elderly at Risk (PROSPER) found that the mean Mini-Mental State Examination (MMSE) score, which was assessed only at subjects’ last on-treatment clinical visit, was comparable between the study’s pravastatin and placebo groups.⁴¹ Likewise, the Medical Research Council/British Heart Foundation (MRC/BHF) Heart Protection Study, which used the Telephone Interview for Cognitive Status questionnaire at the end of the investigation, reported that simvastatin had no positive effect on cognitive performance compared with placebo, but this finding was obtained in the absence of baseline data.⁴² Given the limited cognitive assessments performed in these two studies, no firm conclusions should be drawn.

A more recent prospective comparison of atorvastatin and placebo in younger subjects did include baseline and follow-up assessment of cognitive function, and it identified significantly superior performance in the statin-treated population on the MMSE and on tests of attention, psychomotor speed, mental flexibility, working memory, and memory retrieval.⁴³

The initial clinical investigation of statin therapy in patients with AD—the Alzheimer’s Disease Cholesterol-Lowering Treatment (ADCLT) trial—involved atorvastatin.⁴⁴ Patients with mild to moderate AD were randomized to either placebo or 80 mg/day of atorvastatin for a 1-year period. Evaluable data were available for 63 patients (32 in the atorvastatin group, 31 in the placebo group). End points included the change in performance on the following measures:

MMSE

• Alzheimer’s Disease Assessment Scale–cognitive subscale (ADAS-cog)
• Neuropsychiatric Inventory Caregiver Distress Scale (NPI)
• Clinical Global Impression of Change scale (CGIC)
• Alzheimer’s Disease Cooperative Study–Activities of Daily Living Inventory (ADCS-ADL)
• Geriatric Depression Scale (GDS).

Cognitive results

In the setting of continued cholinesterase inhibitor use, atorvastatin provided significant benefit on the ADAS-cog at 26 weeks compared with placebo (P = .003) and marginally significant benefit at 1 year (P = .055) while producing a trend for benefit on the CGIC and NPI and a statistically significant improvement on the GDS after 1 year of active treatment.⁴⁴ The observed benefit on the MMSE with atorvastatin versus placebo did not reach statistical significance, and no discernible difference was observed on the ADCS-ADL.⁴⁴ In contrast, a significant difference in the slope of deterioration on the MMSE and the GDS in the atorvastatin group versus the placebo group suggested disease modification.⁴⁵

Blood test results

Secondary analysis and initial morphometric substudy

Secondary assessment indicated that the subjects who garnered the greatest benefit from atorvastatin therapy in terms of their 6-month ADAS-cog score were those who had higher cholesterol levels at trial entry, those who harbored the apolipoprotein E4 allele, and those who were less affected by AD at trial entry (ie, with higher entry MMSE scores).⁴⁷

In an ADCLT substudy using new voxel-based morphometry techniques, we quantitatively assessed gray matter density in 15 ADCLT trial participants and compared it with density findings in 15 normal elderly controls.⁴⁸ Regional reductions in gray matter density were observed in the AD patients compared with the controls. Large differences in gray matter concentration were observed bilaterally in the temporal lobe. The anterior cingulate, right superior temporal, left superior frontal, and posterior cingulate regions also showed significantly decreased gray matter density in the AD patients compared with the controls. A significant relationship was observed between gray matter density and ADAS-cog error scores—ie, more severe levels of cognitive impairment correlated with reduced gray matter density.⁴⁸

PILOT SUBSTUDY OF ADCLT: ASSESSING MORPHOLOGIC CHANGES WITH STATIN THERAPY

Eleven of the 15 ADCLT trial participants from the above morphometric substudy returned for MRI assessment after 1 year of treatment with either atorvastatin or placebo. We report here the comparative effects of atorvastatin and placebo on hippocampal volume and the relationship with cognitive performance.

Participants

Subjects were participating in the ADCLT trial, an investigator-initiated, double-blind, placebo-controlled study. Neuroimaging was performed at the Barrow Neurological Institute, Phoenix, AZ, for a subset of the participants in the trial (n = 11) as a pilot study to examine neural changes associated with atorvastatin therapy.

Each patient underwent screening, assignment to either atorvastatin 80 mg/day or placebo, and medical and cognitive assessment at Sun Health Research Institute, Sun City, AZ, prior to imaging at Barrow. All  patients met Diagnostic and Statistical Manual of Mental Disorders, fourth edition, criteria for dementia as well as NINCDS-ADRDA criteria for probable AD. Each patient was free of significant psychiatric and neurological history and had a score of 4 or less on the Hachinski Modified Ischemia Scale. All MRIs were reviewed by a neuroradiologist to ensure that there was no evidence of stroke or cortical or lacunar infarcts.

Both sites’ institutional review boards approved this project, and all subjects gave written informed consent.

Cognitive assessment

A primary efficacy measure used in the parent study was the ADAS-cog,⁴⁹ and the MMSE⁵⁰ was a secondary measure. Change scores were determined by comparing values obtained at baseline, prior to randomization to treatment with either atorvastatin or placebo, and after 1 year of treatment. MMSE scores were obtained at the same session as the ADAS-cog scores. Cognitive assessments were obtained within 2 weeks prior to MRI.

Image acquisition

All participants underwent imaging on a single 1.5­tesla GE scanner at Barrow Neurological Institute. Imaging was conducted both prior to treatment randomization and again after 1 year of treatment. Images of the whole brain were collected using a coronal SPGR (spoiled gradient) T₁-weighted, three-dimensional acquisition with the following parameters:

• Number of acquisitions = 1
• Repetition time = 23 msec
• Echo time = 8 msec
• Flip angle = 35 degrees
• Bandwidth = 12.5 kHz
• Slice thickness = 1.5 mm or 1.9 mm
• 0 skip between slices
• In-plane resolution = 0.9375 x 0.9375.

Hippocampal volumetrics

All imaging analysis was performed within the Analysis of Functional Neuroimages (AFNI) package.⁵¹ We traced the outline of the hippocampus using the three-dimensional SPGR images. The hippocampi were visualized in all three planes, landmarked in the coronal and sagittal planes, and drawn in the coronal plane. We employed the guidelines of Insausti et al⁵² and Machulda et al⁵³ to define the hippocampal boundaries. First we defined the anterior boundary by observing the white matter band and/or the cerebrospinal fluid space between the amygdala and hippocampus in the sagittal plane. The posterior aspect of the posterior region was initially landmarked in the sagittal plane by locating the posterior edge of the hippocampus and then checking in the coronal plane to ensure that the fornices were completely visualized. Volumes were calculated by importing the extracted hippocampi into MATLAB to measure the volumes.

Mean differences between the atorvastatin and placebo groups were evaluated using two-tailed Student t tests. Correlation between changes in cognitive measures and changes in the hippocampal volume for the total population and for the treatment groups was determined using Pearson’s r coefficient. Significance was defined as a P value less than .05; a P value between .05 and .10 was deemed a trend.

Results

In contrast to other studies,^54–58 we found in this pilot study that right hippocampal volume was slightly less than left hippocampal volume (2,015 ± 141 mm³ vs 2,135 ± 183 mm³).

Mean changes in performance on the ADAS-cog and MMSE were less pronounced in the atorvastatin group than in the placebo group, but not significantly so (Table 1). However, there was a trend toward superiority in the atorvastatin group on performance on the free word-recall subscale of the ADAS-cog.

No significant correlations were found between change in cognitive performance and change in hip-pocampal volume.

DISCUSSION

The preponderance of evidence clearly indicates that hippocampal volume is reduced in patients with AD compared with individuals with normal cognitive ability for their age. There is also evidence indicating that as cognitive performance deteriorates in AD patients, there are concurrent further reductions in hippocampal volume.⁵⁴ Many studies reported that there was no significant volume difference between the right and left hippocampi, but most suggested that the left hip-pocampus was slightly smaller than the right.^54–58 We identified no significant difference in volume between the sides, but we did find that the right hippocampus was smaller than the left in a very limited population of subjects with mild to moderate AD.

The major finding of this pilot study flies in the face of conventional wisdom in that there seems to be significant shrinkage of the right hippocampus with atorvastatin therapy compared with placebo in a randomized AD treatment trial that demonstrated clinical benefit with atorvastatin therapy.⁴⁴ A similar finding was reported from the beta-amyloid immunization (AN1792) treatment trial in AD.⁵⁹ In that study the active immunization was associated with significant clinical benefit, reduced beta-amyloid load, and reduced hippocampal volume.⁵⁹ The authors suggested that removal of beta-amyloid and/or other protein constituents from the tissue might have caused a “fluid shift” out of the tissue, resulting in shrinkage.

Based on our previous finding of reduced brain tissue density in AD patients compared with age-matched normal controls,⁴⁸ an alternative explanation can be proposed. Neuronal loss in the hippocampus may be accompanied by increased fluid balance (reduced density) in an attempt to retain the previous volume at the expense of function. Accordingly, as the hippocampus shrinks, it approaches a more normal density for the remaining neuronal complement, and cognitive function improves.

The autonomic nervous system has an important role in the genesis, maintenance, and interruption of ventricular arrhythmias.¹ In most instances, sympathetic activation precipitates or enhances ventricular arrhythmias, whereas vagal tone suppresses their occurrence.^2,3 Therefore, modulating autonomic tone has been proposed as a method to potentially suppress ventricular arrhythmias.⁴

An important mechanism underlying the development of ventricular arrhythmias is electrophysiologic heterogeneity. Electrical heterogeneity predisposes to the development of reentrant arrhythmias and other types of arrhythmias.⁵

SYMPATHETIC AND PARASYMPATHETIC INNERVATION OF THE HEART

Sympathetic nerve fibers are located subepicardially and travel along the routes of the major coronary arteries. In contrast, the vagus nerve is subendocardial in its location after it crosses the atrioventricular groove. A lesion of the heart produced by infarct or fibrosis can result in denervation of otherwise normal myocardium by interruption of neural axons traveling through the lesion. A defect in sympathetic function following myocardial infarction (MI) has been demonstrated in both animals and humans as measured by iodine-123-metaiodobenzylguanidine (MIBG) and C-11 hydroxyephedrine.⁶

Reduced uptake of MIBG in the inferior wall has recently been observed in patients with idiopathic ventricular fibrillation as compared with controls. Although no difference in survival could be detected between the two groups, patients with reduced uptake of MIBG had an increased incidence of ventricular tachyarrhythmias compared with those who did not have such a defect.⁷

Similar observations of sympathetic dysfunction have been made in a variety of animal models and humans with heart failure, coronary disease, and ventricular tachycardia in the absence of structural heart disease. In such instances, the speculation is that sympathetic heterogeneity may produce electrical heterogeneity and spur the development of ventricular arrhythmias. The arrhythmic mechanism is probably more complex than this description, however, because the response to sympathetic inhibition using beta-blockers is not uniform.

Evidence of nerve sprouting

Using a growth-associated protein antibody that marks axonal growth, nerve sprouting has been demonstrated in mice in areas of denervation following MI.⁸ Similarly, using growth-associated protein 43 staining, researchers have demonstrated nerve sprouting in the right atrial free wall, right atrial isthmus, and right ventricle in dogs after radiofrequency catheter ablation.⁹

Neural component in ventricular arrhythmias

Sympathetic hypersensitivity has been shown in areas of denervation, which may be related in part to nerve sprouting. Other sympathetic and electrical phenomena following myocardial injury include an upregulation of nerve growth factor, a heterogeneous distribution of sympathetic innervation, and electrical heterogeneity with areas of denervation, hyperinnervation, and normal nerve density.

Two discoveries by Chen and colleagues are perhaps most noteworthy. One is that nerve growth factor infusion and stellate ganglion stimulation following MI increase nerve density and ventricular arrhythmias, with increased burst frequency discharge of the stellate ganglion prior to the onset of ventricular tachycardia/ventricular fibrillation (VT/VF) in dogs.⁸ More recently, they have shown that infusion of nerve growth factor into the stellate ganglion prolongs the QT interval and prolongs ventricular arrhythmias.¹⁰

A relationship has been established between the hyperinnervation that occurs following myocardial injury and ventricular arrhythmias. Using immunocytochemical staining in explanted native hearts of transplant recipients, Chen and colleagues demonstrated colocalization of Schwann cells, sympathetic nerves, and nerve axons, as well as regional cardiac hyperinnervation, with the most abundant nerve sprouting in the areas bordering myocardial injury and normal myocardium.⁸ In addition, they demonstrated positive tyrosine hydroxylase staining of cardiac nerves in areas around coronary arteries in patients with coronary disease and idiopathic dilated nonischemic cardiomyopathy. At the origin of ventricular tachycardia (prior to transplant), nerve sprouting was shown by staining for S100 protein and tyrosine hydroxylase. The authors hypothesized that nerve sprouting may give rise to ventricular arrhythmia and sudden cardiac death, in which MI results in nerve injury, followed by sympathetic nerve sprouting and regional myocardial hyperinnervation.¹⁰

A link with circadian variations in QT interval length?

The observation that nerve growth factor infused into the left stellate ganglion prolongs the QT interval and prolongs ventricular arrhythmias, resulting in an inordinate risk of sudden death, is fascinating in the context of recent findings of a circadian variation in duration of the QT interval. In measuring QT intervals in 3,700 men without ventricular arrhythmias, we found that the QT interval peaked in winter (between October and January), with a 6-msec difference between the longest and shortest QT intervals.¹¹ This increase in the QT interval in winter coincides with an increase in the incidence of sudden death, which occurs in many regions of the world regardless of climate. Whether or not this increase in sudden death in winter is related to a longer QT interval is supposition, but the potential interaction deserves further exploration. A similar surge in sudden death in winter was observed in patients who were eligible for an implantable cardioverter-defibrillator (ICD) but did not receive one, as opposed to those who did receive an ICD, which suggests that the mechanism responsible for the increase in sudden death in winter is a ventricular tachyarrhythmia that can be prevented by an ICD.¹²

How sympathetic hyperinnervation promotes cardiac arrhythmias is speculative, but increased density of sympathetic nerve endings could promote the release of sympathetic neurotransmitters during sympathetic excitation. The autonomic remodeling is associated with heterogeneous electrical remodeling of cardiomyocytes, resulting in prolongation of action potentials in hyperinnervated regions. Further, acute release of sympathetic neurotransmitters probably accentuates the heterogeneity of excitability and refractoriness, likely contributing to arrhythmia susceptibility.⁵

Inhibiting sympathetic activity pharmacologically reduces the incidence of sudden cardiac death in patients with heart failure. In the Eplerenone Post-Acute Myocardial Infarction Heart Failure Efficacy and Survival Study (EPHESUS), the aldosterone inhibitor eplerenone was associated with a clear reduction in sudden cardiac arrest in patients with acute MI complicated by left ventricular dysfunction.¹³ Beta-blockers and angiotensin-converting enzyme inhibitors have had the same effect. These findings indicate that adverse electrophysiologic consequences from sympathetic stimulation may contribute to the development of a pro-arrhythmic substrate, and that antagonizing sympathetic activation can reduce the extent of adverse electrical remodeling to reduce the risk of sudden cardiac death.

SPINAL CORD STIMULATION

Acute spinal cord stimulation

The possibility of using spinal cord stimulation to modulate cardiac arrhythmias is intriguing, as electrodes introduced paraspinally may activate nerves that could affect sympathetic function. In Europe, spinal cord stimulation is already approved for treatment of patients with intractable angina and end-stage coronary disease.

The mechanism responsible for elimination of angina pectoris via carotid sinus massage is presumably an increase in vagal activity to the heart. In a study of whether thoracic spinal cord stimulation eliminates angina pectoris via a vagal mechanism, Olgin et al confirmed that spinal cord stimulation at the T1-T2 segments enhanced parasympathetic activity and that this action is mediated via the vagus.¹⁴ These findings suggest that thoracic spinal cord stimulation may protect against ventricular arrhythmias through its effect on autonomic tone.

This suggestion led to the development of a canine model of spontaneous ventricular arrhythmias to investigate the mechanisms responsible for ventricular arrhythmias related to acute myocardial ischemia in the setting of healed MI.¹⁵ An infarct was produced via occlusion of the left anterior descending coronary artery and a permanent ventricular pacemaker was placed. After a 2-week recovery period, heart failure was induced by continuous rapid ventricular pacing for 2 to 3 weeks. Transient myocardial ischemia was induced by transient occlusion of the proximal left circumflex coronary artery. Seventy-two percent of dogs surviving the rapid pacing period developed VT/VF during acute left circumflex artery occlusion or within 1 to 2 minutes thereafter.

The effect of thoracic spinal cord stimulation applied at the dorsal T1-T2 segments was studied on the surviving dogs.₁₆ Spinal cord stimulation reduced the occurrence of VT/VF from 59% to 23%.

Another study examined the effect of intrathecal clonidine, an alpha-2 antagonist that reduces concentrations of catecholamines, on ventricular arrhythmias in the canine model.¹⁷ Ischemia-induced VT/VF occurred in 9 of 12 dogs before administration of intrathecal clonidine in contrast to only 3 of 12 dogs after clonidine administration, a degree of efficacy similar to that with spinal cord stimulation.

Chronic spinal cord stimulation

Studies of the effects of chronic thoracic spinal cord stimulation on ventricular function and ventricular arrhythmias in a canine postinfarction heart failure model have recently been completed, and the results will be published in the near future.¹⁸

CONCLUSIONS

Sudden cardiac death continues to be a major health problem in Western countries. Many approaches have been explored in attempting to reduce this modern-day plague.¹⁹ A better understanding of the risks, mechanisms, and treatments is required.²⁰ An animal model has demonstrated that acute modulation of autonomic tone with thoracic spinal cord stimulation or intrathe-cal clonidine reduces susceptibility to ischemic ventricular arrhythmias, presumably via a sympatholytic mechanism. Modulation of autonomic tone—sympatholytic, vagomimetic, or both—may play a significant role in protecting against spontaneous and ischemic ventricular tachyarrhythmias.

The autonomic nervous system has an important role in the genesis, maintenance, and interruption of ventricular arrhythmias.¹ In most instances, sympathetic activation precipitates or enhances ventricular arrhythmias, whereas vagal tone suppresses their occurrence.^2,3 Therefore, modulating autonomic tone has been proposed as a method to potentially suppress ventricular arrhythmias.⁴

An important mechanism underlying the development of ventricular arrhythmias is electrophysiologic heterogeneity. Electrical heterogeneity predisposes to the development of reentrant arrhythmias and other types of arrhythmias.⁵

SYMPATHETIC AND PARASYMPATHETIC INNERVATION OF THE HEART

Sympathetic nerve fibers are located subepicardially and travel along the routes of the major coronary arteries. In contrast, the vagus nerve is subendocardial in its location after it crosses the atrioventricular groove. A lesion of the heart produced by infarct or fibrosis can result in denervation of otherwise normal myocardium by interruption of neural axons traveling through the lesion. A defect in sympathetic function following myocardial infarction (MI) has been demonstrated in both animals and humans as measured by iodine-123-metaiodobenzylguanidine (MIBG) and C-11 hydroxyephedrine.⁶

Reduced uptake of MIBG in the inferior wall has recently been observed in patients with idiopathic ventricular fibrillation as compared with controls. Although no difference in survival could be detected between the two groups, patients with reduced uptake of MIBG had an increased incidence of ventricular tachyarrhythmias compared with those who did not have such a defect.⁷

Similar observations of sympathetic dysfunction have been made in a variety of animal models and humans with heart failure, coronary disease, and ventricular tachycardia in the absence of structural heart disease. In such instances, the speculation is that sympathetic heterogeneity may produce electrical heterogeneity and spur the development of ventricular arrhythmias. The arrhythmic mechanism is probably more complex than this description, however, because the response to sympathetic inhibition using beta-blockers is not uniform.

Evidence of nerve sprouting

Using a growth-associated protein antibody that marks axonal growth, nerve sprouting has been demonstrated in mice in areas of denervation following MI.⁸ Similarly, using growth-associated protein 43 staining, researchers have demonstrated nerve sprouting in the right atrial free wall, right atrial isthmus, and right ventricle in dogs after radiofrequency catheter ablation.⁹

Neural component in ventricular arrhythmias

Sympathetic hypersensitivity has been shown in areas of denervation, which may be related in part to nerve sprouting. Other sympathetic and electrical phenomena following myocardial injury include an upregulation of nerve growth factor, a heterogeneous distribution of sympathetic innervation, and electrical heterogeneity with areas of denervation, hyperinnervation, and normal nerve density.

Two discoveries by Chen and colleagues are perhaps most noteworthy. One is that nerve growth factor infusion and stellate ganglion stimulation following MI increase nerve density and ventricular arrhythmias, with increased burst frequency discharge of the stellate ganglion prior to the onset of ventricular tachycardia/ventricular fibrillation (VT/VF) in dogs.⁸ More recently, they have shown that infusion of nerve growth factor into the stellate ganglion prolongs the QT interval and prolongs ventricular arrhythmias.¹⁰

A relationship has been established between the hyperinnervation that occurs following myocardial injury and ventricular arrhythmias. Using immunocytochemical staining in explanted native hearts of transplant recipients, Chen and colleagues demonstrated colocalization of Schwann cells, sympathetic nerves, and nerve axons, as well as regional cardiac hyperinnervation, with the most abundant nerve sprouting in the areas bordering myocardial injury and normal myocardium.⁸ In addition, they demonstrated positive tyrosine hydroxylase staining of cardiac nerves in areas around coronary arteries in patients with coronary disease and idiopathic dilated nonischemic cardiomyopathy. At the origin of ventricular tachycardia (prior to transplant), nerve sprouting was shown by staining for S100 protein and tyrosine hydroxylase. The authors hypothesized that nerve sprouting may give rise to ventricular arrhythmia and sudden cardiac death, in which MI results in nerve injury, followed by sympathetic nerve sprouting and regional myocardial hyperinnervation.¹⁰

A link with circadian variations in QT interval length?

The observation that nerve growth factor infused into the left stellate ganglion prolongs the QT interval and prolongs ventricular arrhythmias, resulting in an inordinate risk of sudden death, is fascinating in the context of recent findings of a circadian variation in duration of the QT interval. In measuring QT intervals in 3,700 men without ventricular arrhythmias, we found that the QT interval peaked in winter (between October and January), with a 6-msec difference between the longest and shortest QT intervals.¹¹ This increase in the QT interval in winter coincides with an increase in the incidence of sudden death, which occurs in many regions of the world regardless of climate. Whether or not this increase in sudden death in winter is related to a longer QT interval is supposition, but the potential interaction deserves further exploration. A similar surge in sudden death in winter was observed in patients who were eligible for an implantable cardioverter-defibrillator (ICD) but did not receive one, as opposed to those who did receive an ICD, which suggests that the mechanism responsible for the increase in sudden death in winter is a ventricular tachyarrhythmia that can be prevented by an ICD.¹²

How sympathetic hyperinnervation promotes cardiac arrhythmias is speculative, but increased density of sympathetic nerve endings could promote the release of sympathetic neurotransmitters during sympathetic excitation. The autonomic remodeling is associated with heterogeneous electrical remodeling of cardiomyocytes, resulting in prolongation of action potentials in hyperinnervated regions. Further, acute release of sympathetic neurotransmitters probably accentuates the heterogeneity of excitability and refractoriness, likely contributing to arrhythmia susceptibility.⁵

Inhibiting sympathetic activity pharmacologically reduces the incidence of sudden cardiac death in patients with heart failure. In the Eplerenone Post-Acute Myocardial Infarction Heart Failure Efficacy and Survival Study (EPHESUS), the aldosterone inhibitor eplerenone was associated with a clear reduction in sudden cardiac arrest in patients with acute MI complicated by left ventricular dysfunction.¹³ Beta-blockers and angiotensin-converting enzyme inhibitors have had the same effect. These findings indicate that adverse electrophysiologic consequences from sympathetic stimulation may contribute to the development of a pro-arrhythmic substrate, and that antagonizing sympathetic activation can reduce the extent of adverse electrical remodeling to reduce the risk of sudden cardiac death.

SPINAL CORD STIMULATION

Acute spinal cord stimulation

The possibility of using spinal cord stimulation to modulate cardiac arrhythmias is intriguing, as electrodes introduced paraspinally may activate nerves that could affect sympathetic function. In Europe, spinal cord stimulation is already approved for treatment of patients with intractable angina and end-stage coronary disease.

The mechanism responsible for elimination of angina pectoris via carotid sinus massage is presumably an increase in vagal activity to the heart. In a study of whether thoracic spinal cord stimulation eliminates angina pectoris via a vagal mechanism, Olgin et al confirmed that spinal cord stimulation at the T1-T2 segments enhanced parasympathetic activity and that this action is mediated via the vagus.¹⁴ These findings suggest that thoracic spinal cord stimulation may protect against ventricular arrhythmias through its effect on autonomic tone.

This suggestion led to the development of a canine model of spontaneous ventricular arrhythmias to investigate the mechanisms responsible for ventricular arrhythmias related to acute myocardial ischemia in the setting of healed MI.¹⁵ An infarct was produced via occlusion of the left anterior descending coronary artery and a permanent ventricular pacemaker was placed. After a 2-week recovery period, heart failure was induced by continuous rapid ventricular pacing for 2 to 3 weeks. Transient myocardial ischemia was induced by transient occlusion of the proximal left circumflex coronary artery. Seventy-two percent of dogs surviving the rapid pacing period developed VT/VF during acute left circumflex artery occlusion or within 1 to 2 minutes thereafter.

The effect of thoracic spinal cord stimulation applied at the dorsal T1-T2 segments was studied on the surviving dogs.₁₆ Spinal cord stimulation reduced the occurrence of VT/VF from 59% to 23%.

Another study examined the effect of intrathecal clonidine, an alpha-2 antagonist that reduces concentrations of catecholamines, on ventricular arrhythmias in the canine model.¹⁷ Ischemia-induced VT/VF occurred in 9 of 12 dogs before administration of intrathecal clonidine in contrast to only 3 of 12 dogs after clonidine administration, a degree of efficacy similar to that with spinal cord stimulation.

Chronic spinal cord stimulation

Studies of the effects of chronic thoracic spinal cord stimulation on ventricular function and ventricular arrhythmias in a canine postinfarction heart failure model have recently been completed, and the results will be published in the near future.¹⁸

CONCLUSIONS

Sudden cardiac death continues to be a major health problem in Western countries. Many approaches have been explored in attempting to reduce this modern-day plague.¹⁹ A better understanding of the risks, mechanisms, and treatments is required.²⁰ An animal model has demonstrated that acute modulation of autonomic tone with thoracic spinal cord stimulation or intrathe-cal clonidine reduces susceptibility to ischemic ventricular arrhythmias, presumably via a sympatholytic mechanism. Modulation of autonomic tone—sympatholytic, vagomimetic, or both—may play a significant role in protecting against spontaneous and ischemic ventricular tachyarrhythmias.

The autonomic nervous system has an important role in the genesis, maintenance, and interruption of ventricular arrhythmias.¹ In most instances, sympathetic activation precipitates or enhances ventricular arrhythmias, whereas vagal tone suppresses their occurrence.^2,3 Therefore, modulating autonomic tone has been proposed as a method to potentially suppress ventricular arrhythmias.⁴

An important mechanism underlying the development of ventricular arrhythmias is electrophysiologic heterogeneity. Electrical heterogeneity predisposes to the development of reentrant arrhythmias and other types of arrhythmias.⁵

SYMPATHETIC AND PARASYMPATHETIC INNERVATION OF THE HEART

Sympathetic nerve fibers are located subepicardially and travel along the routes of the major coronary arteries. In contrast, the vagus nerve is subendocardial in its location after it crosses the atrioventricular groove. A lesion of the heart produced by infarct or fibrosis can result in denervation of otherwise normal myocardium by interruption of neural axons traveling through the lesion. A defect in sympathetic function following myocardial infarction (MI) has been demonstrated in both animals and humans as measured by iodine-123-metaiodobenzylguanidine (MIBG) and C-11 hydroxyephedrine.⁶

Reduced uptake of MIBG in the inferior wall has recently been observed in patients with idiopathic ventricular fibrillation as compared with controls. Although no difference in survival could be detected between the two groups, patients with reduced uptake of MIBG had an increased incidence of ventricular tachyarrhythmias compared with those who did not have such a defect.⁷

Similar observations of sympathetic dysfunction have been made in a variety of animal models and humans with heart failure, coronary disease, and ventricular tachycardia in the absence of structural heart disease. In such instances, the speculation is that sympathetic heterogeneity may produce electrical heterogeneity and spur the development of ventricular arrhythmias. The arrhythmic mechanism is probably more complex than this description, however, because the response to sympathetic inhibition using beta-blockers is not uniform.

Evidence of nerve sprouting

Using a growth-associated protein antibody that marks axonal growth, nerve sprouting has been demonstrated in mice in areas of denervation following MI.⁸ Similarly, using growth-associated protein 43 staining, researchers have demonstrated nerve sprouting in the right atrial free wall, right atrial isthmus, and right ventricle in dogs after radiofrequency catheter ablation.⁹

Neural component in ventricular arrhythmias

Sympathetic hypersensitivity has been shown in areas of denervation, which may be related in part to nerve sprouting. Other sympathetic and electrical phenomena following myocardial injury include an upregulation of nerve growth factor, a heterogeneous distribution of sympathetic innervation, and electrical heterogeneity with areas of denervation, hyperinnervation, and normal nerve density.

Two discoveries by Chen and colleagues are perhaps most noteworthy. One is that nerve growth factor infusion and stellate ganglion stimulation following MI increase nerve density and ventricular arrhythmias, with increased burst frequency discharge of the stellate ganglion prior to the onset of ventricular tachycardia/ventricular fibrillation (VT/VF) in dogs.⁸ More recently, they have shown that infusion of nerve growth factor into the stellate ganglion prolongs the QT interval and prolongs ventricular arrhythmias.¹⁰

A relationship has been established between the hyperinnervation that occurs following myocardial injury and ventricular arrhythmias. Using immunocytochemical staining in explanted native hearts of transplant recipients, Chen and colleagues demonstrated colocalization of Schwann cells, sympathetic nerves, and nerve axons, as well as regional cardiac hyperinnervation, with the most abundant nerve sprouting in the areas bordering myocardial injury and normal myocardium.⁸ In addition, they demonstrated positive tyrosine hydroxylase staining of cardiac nerves in areas around coronary arteries in patients with coronary disease and idiopathic dilated nonischemic cardiomyopathy. At the origin of ventricular tachycardia (prior to transplant), nerve sprouting was shown by staining for S100 protein and tyrosine hydroxylase. The authors hypothesized that nerve sprouting may give rise to ventricular arrhythmia and sudden cardiac death, in which MI results in nerve injury, followed by sympathetic nerve sprouting and regional myocardial hyperinnervation.¹⁰

A link with circadian variations in QT interval length?

The observation that nerve growth factor infused into the left stellate ganglion prolongs the QT interval and prolongs ventricular arrhythmias, resulting in an inordinate risk of sudden death, is fascinating in the context of recent findings of a circadian variation in duration of the QT interval. In measuring QT intervals in 3,700 men without ventricular arrhythmias, we found that the QT interval peaked in winter (between October and January), with a 6-msec difference between the longest and shortest QT intervals.¹¹ This increase in the QT interval in winter coincides with an increase in the incidence of sudden death, which occurs in many regions of the world regardless of climate. Whether or not this increase in sudden death in winter is related to a longer QT interval is supposition, but the potential interaction deserves further exploration. A similar surge in sudden death in winter was observed in patients who were eligible for an implantable cardioverter-defibrillator (ICD) but did not receive one, as opposed to those who did receive an ICD, which suggests that the mechanism responsible for the increase in sudden death in winter is a ventricular tachyarrhythmia that can be prevented by an ICD.¹²

How sympathetic hyperinnervation promotes cardiac arrhythmias is speculative, but increased density of sympathetic nerve endings could promote the release of sympathetic neurotransmitters during sympathetic excitation. The autonomic remodeling is associated with heterogeneous electrical remodeling of cardiomyocytes, resulting in prolongation of action potentials in hyperinnervated regions. Further, acute release of sympathetic neurotransmitters probably accentuates the heterogeneity of excitability and refractoriness, likely contributing to arrhythmia susceptibility.⁵

Inhibiting sympathetic activity pharmacologically reduces the incidence of sudden cardiac death in patients with heart failure. In the Eplerenone Post-Acute Myocardial Infarction Heart Failure Efficacy and Survival Study (EPHESUS), the aldosterone inhibitor eplerenone was associated with a clear reduction in sudden cardiac arrest in patients with acute MI complicated by left ventricular dysfunction.¹³ Beta-blockers and angiotensin-converting enzyme inhibitors have had the same effect. These findings indicate that adverse electrophysiologic consequences from sympathetic stimulation may contribute to the development of a pro-arrhythmic substrate, and that antagonizing sympathetic activation can reduce the extent of adverse electrical remodeling to reduce the risk of sudden cardiac death.

SPINAL CORD STIMULATION

Acute spinal cord stimulation

The possibility of using spinal cord stimulation to modulate cardiac arrhythmias is intriguing, as electrodes introduced paraspinally may activate nerves that could affect sympathetic function. In Europe, spinal cord stimulation is already approved for treatment of patients with intractable angina and end-stage coronary disease.

The mechanism responsible for elimination of angina pectoris via carotid sinus massage is presumably an increase in vagal activity to the heart. In a study of whether thoracic spinal cord stimulation eliminates angina pectoris via a vagal mechanism, Olgin et al confirmed that spinal cord stimulation at the T1-T2 segments enhanced parasympathetic activity and that this action is mediated via the vagus.¹⁴ These findings suggest that thoracic spinal cord stimulation may protect against ventricular arrhythmias through its effect on autonomic tone.

This suggestion led to the development of a canine model of spontaneous ventricular arrhythmias to investigate the mechanisms responsible for ventricular arrhythmias related to acute myocardial ischemia in the setting of healed MI.¹⁵ An infarct was produced via occlusion of the left anterior descending coronary artery and a permanent ventricular pacemaker was placed. After a 2-week recovery period, heart failure was induced by continuous rapid ventricular pacing for 2 to 3 weeks. Transient myocardial ischemia was induced by transient occlusion of the proximal left circumflex coronary artery. Seventy-two percent of dogs surviving the rapid pacing period developed VT/VF during acute left circumflex artery occlusion or within 1 to 2 minutes thereafter.

The effect of thoracic spinal cord stimulation applied at the dorsal T1-T2 segments was studied on the surviving dogs.₁₆ Spinal cord stimulation reduced the occurrence of VT/VF from 59% to 23%.

Another study examined the effect of intrathecal clonidine, an alpha-2 antagonist that reduces concentrations of catecholamines, on ventricular arrhythmias in the canine model.¹⁷ Ischemia-induced VT/VF occurred in 9 of 12 dogs before administration of intrathecal clonidine in contrast to only 3 of 12 dogs after clonidine administration, a degree of efficacy similar to that with spinal cord stimulation.

Chronic spinal cord stimulation

Studies of the effects of chronic thoracic spinal cord stimulation on ventricular function and ventricular arrhythmias in a canine postinfarction heart failure model have recently been completed, and the results will be published in the near future.¹⁸

CONCLUSIONS

Sudden cardiac death continues to be a major health problem in Western countries. Many approaches have been explored in attempting to reduce this modern-day plague.¹⁹ A better understanding of the risks, mechanisms, and treatments is required.²⁰ An animal model has demonstrated that acute modulation of autonomic tone with thoracic spinal cord stimulation or intrathe-cal clonidine reduces susceptibility to ischemic ventricular arrhythmias, presumably via a sympatholytic mechanism. Modulation of autonomic tone—sympatholytic, vagomimetic, or both—may play a significant role in protecting against spontaneous and ischemic ventricular tachyarrhythmias.

Only a few brain structures have been implicated in the autonomic control of blood pressure and heart rate. Among them are heteromodal association areas in the cortex, especially the insular cortex. Insular infarction has been associated with both cardiac arrhythmias and mortality. However, stroke may not be the only insular pathology with the potential to disrupt autonomic function. Alzheimer disease (AD) is associated with both insular pathology and autonomic dysfunction.

This article presents the hypothesis that autonomic dysfunction reflects subclinical stages of AD pathology affecting the insular cortex and discusses the resulting clinical implications.

AUTONOMIC DYSFUNCTION AS A PRODUCT OF SUBCLINICAL ALZHEIMER DISEASE

Braak and Braak have demonstrated a hierarchical progression of AD pathology that includes the insular cortex.¹ This may explain why AD has effects on blood pressure and central autonomic cardioregulatory functions. However, AD reaches the insular cortex at a “preclinical” stage in the Braak and Braak sequence (before “dementia” can be diagnosed). Thus, AD pathology should also be considered as a possible explanation for autonomic morbidity and mortality in nondemented elderly persons.²

Suggestive evidence

The following observations support this possibility:

• Clinical AD is associated with a wide range of dysautonomic phenomena. These can already be demonstrated at the initial diagnosis, which suggests a preclinical onset.
• Only a limited set of brain regions are capable of affecting autonomic control. The insulae are affected at a preclinical stage in the sequence of Braak and Braak (ie, stage III of VI).
• Neurofibrillary tangle (NFT) counts inside the insulae moderate the association between the heart rate–corrected QT interval (QTc) and survival. This has been demonstrated by my colleagues and I in collaboration with the Honolulu-Asia Aging Study, which is examining the association between insular pathology at autopsy and the slope of premorbid change in the QTc.

Implications of AD-mediated autonomic dysfunction

AD-mediated autonomic dysfunction could have important clinical implications:

• The prevalence of preclinical AD is likely to be higher than the number of demented cases. Many apparently well elderly persons may be affected solely on the basis of subclinical AD pathology.
• Autonomic functions have widespread effects; many cardiac and noncardiac “age-related” changes may actually be related to AD.
• Pharmacologic therapies for AD are known to delay the progression of symptoms and to reduce mortality; these medications may also impact AD-related autonomic problems.
• Conversely, the association between other medications and cardiac arrhythmias/sudden death may be mediated via effects on insular function.

ALZHEIMER DISEASE DISRUPTS AUTONOMIC CONTROL

AD has been associated with a wide variety of dys-autonomic phenomena, including increased pupillary dilation, altered skin conductivity, blunted autonomic response to noxious stimuli, diminished heart rate variability, depressed baroreflex sensitivity, and orthostasis. Autonomic instability has yet to be sought in mild cognitive impairment or even earlier preclinical stages of AD. However, nondemented subjects with mild cognitive impairment and early AD do experience more frequent falls and more gait and balance problems than do age-matched controls.

INSULAR CORTEX: A LIKELY TARGET

The insulae have been specifically implicated in the cortical control of autonomic function.³ The vulnerability of the insular cortex to AD is easy to understand. NFTs appear to spread retrogradely along cortico-cortical and cortico-subcortical connections.⁴ The insulae are mesiotemporal structures with direct connections to the hippocampus and entorhinal cortex.

Insular lesions result in changes in cardiovascular and autonomic control that are readily detectable by a variety of measures and procedures, including blood pressure, tilt table, balance platform, and electrocardiogram. The electrocardiographic effects of insular pathology include diminished heart rate variability, determined in either the time domain or the frequency domain. Diminished heart rate variability has been associated with increased mortality in cardiovascular disease and type 2 diabetes. It is important to note, however, that the effects of diminished heart rate variability are statistically independent of disease severity in these disorders, and that they can be demonstrated in the absence of clinically significant cardiovascular disease.⁵

Autopsy studies suggest that as many as 40% of non-demented septuagenarians and octogenarians may have AD pathology that is sufficiently advanced to affect the insular cortex.¹ This might explain the high prevalence of supraventricular arrhythmias and longitudinal decreases in heart rate variability among well elderly persons who are free of cardiovascular disease. In fact, unexplained supraventricular arrhythmias are quite common among such individuals. Both tachy-arrhythmias and bradyarrhythmias are common on 24-hour Holter monitor recordings among subjects older than 80 years, and most are unexplained. In a study of the causes of syncope in a large (N = 711) sample of octogenarians, Lipsitz et al confirmed a cardiac etiology in only 21% of cases, whereas 31% of cases were unexplained.⁶

MORTALITY IN ALZHEIMER DISEASE IS ASSOCIATED WITH RIGHT HEMISPHERE DYSFUNCTION

AD pathology is widely thought to be symmetrically distributed. However, this may not be true of the pre-clinical “limbic” stages of AD.⁷ Since insular effects on autonomic function are highly lateralized, the side of the brain affected by NFTs may be relevant to effects on cardiac rhythm and, hence, mortality risk.

Interestingly, mortality in AD is specifically associated with right hemisphere metabolic changes by electroencephalography, single-photon emission computed tomography, and positron emission tomography. Mortality can also be specifically associated with tests of constructional praxis. Claus and colleagues found that only the praxis subscore of the Cambridge Cognitive Examination (CAMCOG) was significantly related to survival in patients with early AD (P < .001).⁸ Its predictive power was based on only two items: copying ability for a spiral and for a three-dimensional house. The effect was independent of age, sex, education, dementia severity, total CAMCOG score, and symptom duration. Similarly, Swan et al found a significant association between performance on the digit symbol substitution test and 5-year mortality among 1,118 subjects (with a mean age of 70.6 years) in the Western Collaborative Group Study.⁹ In Cox regression analyses, the relative risk for all-cause mortality was 1.44 (95% confidence interval, 1.12 to 1.86) after adjustment for age, education, blood pressure, cancer, cardiovascular/cerebrovascular disease, and smoking.⁹

RIGHT HEMISPHERE DYSFUNCTION AFFECTS MORTALITY IN OTHER CONDITIONS

The effect of right hemisphere dysfunction on mortality is not limited to AD; it can also be demonstrated in other disorders, including epilepsy, head injury, and stroke.

We have been studying the cognitive correlates of mortality among well elderly septuagenarians and octogenarians living in a single comprehensive care retirement community (CCRC). Once again, visuo-spatial measures have been found to be selectively associated with mortality.¹⁰ Clock drawing appears to be the cognitive predictor most strongly correlated with mortality,¹⁰ a finding that has been independently replicated in a second CCRC cohort.¹¹ This effect is independent of other cognitive domains, notably executive function.¹⁰

SUMMARY

Right hemisphere dysfunction is associated with mortality in AD and other conditions. These associations may be mediated by insular pathology. AD affects the insulae at a preclinical stage, and insular AD pathology may affect as many as 40% of nondemented septuagenarians and octogenarians. This pathology can be shown to affect in vivo cardiac conduction, and may dispose elderly persons to cardiac arrhythmias and sudden death. If so, then AD must be considered a potential cause of cardiac arrhythmia, sudden death, and other autonomic disturbances in nondemented older adults.

Acknowledgments

The author wishes to acknowledge the important cooperation and support received from the Air Force Villages and the Honolulu-Asia Aging Study. This work has been supported by a grant from the National Institute for Neurological Disorders and Stroke (NS45121-01A1).

Only a few brain structures have been implicated in the autonomic control of blood pressure and heart rate. Among them are heteromodal association areas in the cortex, especially the insular cortex. Insular infarction has been associated with both cardiac arrhythmias and mortality. However, stroke may not be the only insular pathology with the potential to disrupt autonomic function. Alzheimer disease (AD) is associated with both insular pathology and autonomic dysfunction.

This article presents the hypothesis that autonomic dysfunction reflects subclinical stages of AD pathology affecting the insular cortex and discusses the resulting clinical implications.

AUTONOMIC DYSFUNCTION AS A PRODUCT OF SUBCLINICAL ALZHEIMER DISEASE

Braak and Braak have demonstrated a hierarchical progression of AD pathology that includes the insular cortex.¹ This may explain why AD has effects on blood pressure and central autonomic cardioregulatory functions. However, AD reaches the insular cortex at a “preclinical” stage in the Braak and Braak sequence (before “dementia” can be diagnosed). Thus, AD pathology should also be considered as a possible explanation for autonomic morbidity and mortality in nondemented elderly persons.²

Suggestive evidence

The following observations support this possibility:

• Clinical AD is associated with a wide range of dysautonomic phenomena. These can already be demonstrated at the initial diagnosis, which suggests a preclinical onset.
• Only a limited set of brain regions are capable of affecting autonomic control. The insulae are affected at a preclinical stage in the sequence of Braak and Braak (ie, stage III of VI).
• Neurofibrillary tangle (NFT) counts inside the insulae moderate the association between the heart rate–corrected QT interval (QTc) and survival. This has been demonstrated by my colleagues and I in collaboration with the Honolulu-Asia Aging Study, which is examining the association between insular pathology at autopsy and the slope of premorbid change in the QTc.

Implications of AD-mediated autonomic dysfunction

AD-mediated autonomic dysfunction could have important clinical implications:

• The prevalence of preclinical AD is likely to be higher than the number of demented cases. Many apparently well elderly persons may be affected solely on the basis of subclinical AD pathology.
• Autonomic functions have widespread effects; many cardiac and noncardiac “age-related” changes may actually be related to AD.
• Pharmacologic therapies for AD are known to delay the progression of symptoms and to reduce mortality; these medications may also impact AD-related autonomic problems.
• Conversely, the association between other medications and cardiac arrhythmias/sudden death may be mediated via effects on insular function.

ALZHEIMER DISEASE DISRUPTS AUTONOMIC CONTROL

AD has been associated with a wide variety of dys-autonomic phenomena, including increased pupillary dilation, altered skin conductivity, blunted autonomic response to noxious stimuli, diminished heart rate variability, depressed baroreflex sensitivity, and orthostasis. Autonomic instability has yet to be sought in mild cognitive impairment or even earlier preclinical stages of AD. However, nondemented subjects with mild cognitive impairment and early AD do experience more frequent falls and more gait and balance problems than do age-matched controls.

INSULAR CORTEX: A LIKELY TARGET

The insulae have been specifically implicated in the cortical control of autonomic function.³ The vulnerability of the insular cortex to AD is easy to understand. NFTs appear to spread retrogradely along cortico-cortical and cortico-subcortical connections.⁴ The insulae are mesiotemporal structures with direct connections to the hippocampus and entorhinal cortex.

Insular lesions result in changes in cardiovascular and autonomic control that are readily detectable by a variety of measures and procedures, including blood pressure, tilt table, balance platform, and electrocardiogram. The electrocardiographic effects of insular pathology include diminished heart rate variability, determined in either the time domain or the frequency domain. Diminished heart rate variability has been associated with increased mortality in cardiovascular disease and type 2 diabetes. It is important to note, however, that the effects of diminished heart rate variability are statistically independent of disease severity in these disorders, and that they can be demonstrated in the absence of clinically significant cardiovascular disease.⁵

Autopsy studies suggest that as many as 40% of non-demented septuagenarians and octogenarians may have AD pathology that is sufficiently advanced to affect the insular cortex.¹ This might explain the high prevalence of supraventricular arrhythmias and longitudinal decreases in heart rate variability among well elderly persons who are free of cardiovascular disease. In fact, unexplained supraventricular arrhythmias are quite common among such individuals. Both tachy-arrhythmias and bradyarrhythmias are common on 24-hour Holter monitor recordings among subjects older than 80 years, and most are unexplained. In a study of the causes of syncope in a large (N = 711) sample of octogenarians, Lipsitz et al confirmed a cardiac etiology in only 21% of cases, whereas 31% of cases were unexplained.⁶

MORTALITY IN ALZHEIMER DISEASE IS ASSOCIATED WITH RIGHT HEMISPHERE DYSFUNCTION

AD pathology is widely thought to be symmetrically distributed. However, this may not be true of the pre-clinical “limbic” stages of AD.⁷ Since insular effects on autonomic function are highly lateralized, the side of the brain affected by NFTs may be relevant to effects on cardiac rhythm and, hence, mortality risk.

Interestingly, mortality in AD is specifically associated with right hemisphere metabolic changes by electroencephalography, single-photon emission computed tomography, and positron emission tomography. Mortality can also be specifically associated with tests of constructional praxis. Claus and colleagues found that only the praxis subscore of the Cambridge Cognitive Examination (CAMCOG) was significantly related to survival in patients with early AD (P < .001).⁸ Its predictive power was based on only two items: copying ability for a spiral and for a three-dimensional house. The effect was independent of age, sex, education, dementia severity, total CAMCOG score, and symptom duration. Similarly, Swan et al found a significant association between performance on the digit symbol substitution test and 5-year mortality among 1,118 subjects (with a mean age of 70.6 years) in the Western Collaborative Group Study.⁹ In Cox regression analyses, the relative risk for all-cause mortality was 1.44 (95% confidence interval, 1.12 to 1.86) after adjustment for age, education, blood pressure, cancer, cardiovascular/cerebrovascular disease, and smoking.⁹

RIGHT HEMISPHERE DYSFUNCTION AFFECTS MORTALITY IN OTHER CONDITIONS

The effect of right hemisphere dysfunction on mortality is not limited to AD; it can also be demonstrated in other disorders, including epilepsy, head injury, and stroke.

We have been studying the cognitive correlates of mortality among well elderly septuagenarians and octogenarians living in a single comprehensive care retirement community (CCRC). Once again, visuo-spatial measures have been found to be selectively associated with mortality.¹⁰ Clock drawing appears to be the cognitive predictor most strongly correlated with mortality,¹⁰ a finding that has been independently replicated in a second CCRC cohort.¹¹ This effect is independent of other cognitive domains, notably executive function.¹⁰

SUMMARY

Right hemisphere dysfunction is associated with mortality in AD and other conditions. These associations may be mediated by insular pathology. AD affects the insulae at a preclinical stage, and insular AD pathology may affect as many as 40% of nondemented septuagenarians and octogenarians. This pathology can be shown to affect in vivo cardiac conduction, and may dispose elderly persons to cardiac arrhythmias and sudden death. If so, then AD must be considered a potential cause of cardiac arrhythmia, sudden death, and other autonomic disturbances in nondemented older adults.

Acknowledgments

The author wishes to acknowledge the important cooperation and support received from the Air Force Villages and the Honolulu-Asia Aging Study. This work has been supported by a grant from the National Institute for Neurological Disorders and Stroke (NS45121-01A1).

Only a few brain structures have been implicated in the autonomic control of blood pressure and heart rate. Among them are heteromodal association areas in the cortex, especially the insular cortex. Insular infarction has been associated with both cardiac arrhythmias and mortality. However, stroke may not be the only insular pathology with the potential to disrupt autonomic function. Alzheimer disease (AD) is associated with both insular pathology and autonomic dysfunction.

This article presents the hypothesis that autonomic dysfunction reflects subclinical stages of AD pathology affecting the insular cortex and discusses the resulting clinical implications.

AUTONOMIC DYSFUNCTION AS A PRODUCT OF SUBCLINICAL ALZHEIMER DISEASE

Braak and Braak have demonstrated a hierarchical progression of AD pathology that includes the insular cortex.¹ This may explain why AD has effects on blood pressure and central autonomic cardioregulatory functions. However, AD reaches the insular cortex at a “preclinical” stage in the Braak and Braak sequence (before “dementia” can be diagnosed). Thus, AD pathology should also be considered as a possible explanation for autonomic morbidity and mortality in nondemented elderly persons.²

Suggestive evidence

The following observations support this possibility:

• Clinical AD is associated with a wide range of dysautonomic phenomena. These can already be demonstrated at the initial diagnosis, which suggests a preclinical onset.
• Only a limited set of brain regions are capable of affecting autonomic control. The insulae are affected at a preclinical stage in the sequence of Braak and Braak (ie, stage III of VI).
• Neurofibrillary tangle (NFT) counts inside the insulae moderate the association between the heart rate–corrected QT interval (QTc) and survival. This has been demonstrated by my colleagues and I in collaboration with the Honolulu-Asia Aging Study, which is examining the association between insular pathology at autopsy and the slope of premorbid change in the QTc.

Implications of AD-mediated autonomic dysfunction

AD-mediated autonomic dysfunction could have important clinical implications:

• The prevalence of preclinical AD is likely to be higher than the number of demented cases. Many apparently well elderly persons may be affected solely on the basis of subclinical AD pathology.
• Autonomic functions have widespread effects; many cardiac and noncardiac “age-related” changes may actually be related to AD.
• Pharmacologic therapies for AD are known to delay the progression of symptoms and to reduce mortality; these medications may also impact AD-related autonomic problems.
• Conversely, the association between other medications and cardiac arrhythmias/sudden death may be mediated via effects on insular function.

ALZHEIMER DISEASE DISRUPTS AUTONOMIC CONTROL

AD has been associated with a wide variety of dys-autonomic phenomena, including increased pupillary dilation, altered skin conductivity, blunted autonomic response to noxious stimuli, diminished heart rate variability, depressed baroreflex sensitivity, and orthostasis. Autonomic instability has yet to be sought in mild cognitive impairment or even earlier preclinical stages of AD. However, nondemented subjects with mild cognitive impairment and early AD do experience more frequent falls and more gait and balance problems than do age-matched controls.

INSULAR CORTEX: A LIKELY TARGET

The insulae have been specifically implicated in the cortical control of autonomic function.³ The vulnerability of the insular cortex to AD is easy to understand. NFTs appear to spread retrogradely along cortico-cortical and cortico-subcortical connections.⁴ The insulae are mesiotemporal structures with direct connections to the hippocampus and entorhinal cortex.

Insular lesions result in changes in cardiovascular and autonomic control that are readily detectable by a variety of measures and procedures, including blood pressure, tilt table, balance platform, and electrocardiogram. The electrocardiographic effects of insular pathology include diminished heart rate variability, determined in either the time domain or the frequency domain. Diminished heart rate variability has been associated with increased mortality in cardiovascular disease and type 2 diabetes. It is important to note, however, that the effects of diminished heart rate variability are statistically independent of disease severity in these disorders, and that they can be demonstrated in the absence of clinically significant cardiovascular disease.⁵

Autopsy studies suggest that as many as 40% of non-demented septuagenarians and octogenarians may have AD pathology that is sufficiently advanced to affect the insular cortex.¹ This might explain the high prevalence of supraventricular arrhythmias and longitudinal decreases in heart rate variability among well elderly persons who are free of cardiovascular disease. In fact, unexplained supraventricular arrhythmias are quite common among such individuals. Both tachy-arrhythmias and bradyarrhythmias are common on 24-hour Holter monitor recordings among subjects older than 80 years, and most are unexplained. In a study of the causes of syncope in a large (N = 711) sample of octogenarians, Lipsitz et al confirmed a cardiac etiology in only 21% of cases, whereas 31% of cases were unexplained.⁶

MORTALITY IN ALZHEIMER DISEASE IS ASSOCIATED WITH RIGHT HEMISPHERE DYSFUNCTION

AD pathology is widely thought to be symmetrically distributed. However, this may not be true of the pre-clinical “limbic” stages of AD.⁷ Since insular effects on autonomic function are highly lateralized, the side of the brain affected by NFTs may be relevant to effects on cardiac rhythm and, hence, mortality risk.

Interestingly, mortality in AD is specifically associated with right hemisphere metabolic changes by electroencephalography, single-photon emission computed tomography, and positron emission tomography. Mortality can also be specifically associated with tests of constructional praxis. Claus and colleagues found that only the praxis subscore of the Cambridge Cognitive Examination (CAMCOG) was significantly related to survival in patients with early AD (P < .001).⁸ Its predictive power was based on only two items: copying ability for a spiral and for a three-dimensional house. The effect was independent of age, sex, education, dementia severity, total CAMCOG score, and symptom duration. Similarly, Swan et al found a significant association between performance on the digit symbol substitution test and 5-year mortality among 1,118 subjects (with a mean age of 70.6 years) in the Western Collaborative Group Study.⁹ In Cox regression analyses, the relative risk for all-cause mortality was 1.44 (95% confidence interval, 1.12 to 1.86) after adjustment for age, education, blood pressure, cancer, cardiovascular/cerebrovascular disease, and smoking.⁹

RIGHT HEMISPHERE DYSFUNCTION AFFECTS MORTALITY IN OTHER CONDITIONS

The effect of right hemisphere dysfunction on mortality is not limited to AD; it can also be demonstrated in other disorders, including epilepsy, head injury, and stroke.

We have been studying the cognitive correlates of mortality among well elderly septuagenarians and octogenarians living in a single comprehensive care retirement community (CCRC). Once again, visuo-spatial measures have been found to be selectively associated with mortality.¹⁰ Clock drawing appears to be the cognitive predictor most strongly correlated with mortality,¹⁰ a finding that has been independently replicated in a second CCRC cohort.¹¹ This effect is independent of other cognitive domains, notably executive function.¹⁰

SUMMARY

Right hemisphere dysfunction is associated with mortality in AD and other conditions. These associations may be mediated by insular pathology. AD affects the insulae at a preclinical stage, and insular AD pathology may affect as many as 40% of nondemented septuagenarians and octogenarians. This pathology can be shown to affect in vivo cardiac conduction, and may dispose elderly persons to cardiac arrhythmias and sudden death. If so, then AD must be considered a potential cause of cardiac arrhythmia, sudden death, and other autonomic disturbances in nondemented older adults.

Acknowledgments

The author wishes to acknowledge the important cooperation and support received from the Air Force Villages and the Honolulu-Asia Aging Study. This work has been supported by a grant from the National Institute for Neurological Disorders and Stroke (NS45121-01A1).

To the Editor: In their otherwise excellent review, “Masquerade: Medical causes of back pain” (Cleve Clin J Med 2007; 74:905–913), Dr. Klineberg et al seem to confuse two distinct pathologic processes—aortic dissection and rupture of an aortic aneurysm. Parts of their description seem to fit the pathology of abdominal aortic aneurysm, with a pulsatile abdominal mass, sentinel bleeding, and rupture risk with a size over 6 cm, whereas other parts seem to correspond to aortic dissection, with severe, ripping pain and an association with Marfan syndrome. They also use the terminology “dissecting aortic aneurysm,” which again implies a single entity, when in fact the two conditions rarely occur together. The authors are not alone in their use of this misnomer: a review of the Web sites of renowned universities reveals use of the same terminology. The readers would have been better served if the authors had discussed “acute aortic dissection” and “ruptured aortic aneurysm” as two separate causes of back pain, with a note that in rare cases an aortic aneurysm can develop a dissection.

To the Editor: In their otherwise excellent review, “Masquerade: Medical causes of back pain” (Cleve Clin J Med 2007; 74:905–913), Dr. Klineberg et al seem to confuse two distinct pathologic processes—aortic dissection and rupture of an aortic aneurysm. Parts of their description seem to fit the pathology of abdominal aortic aneurysm, with a pulsatile abdominal mass, sentinel bleeding, and rupture risk with a size over 6 cm, whereas other parts seem to correspond to aortic dissection, with severe, ripping pain and an association with Marfan syndrome. They also use the terminology “dissecting aortic aneurysm,” which again implies a single entity, when in fact the two conditions rarely occur together. The authors are not alone in their use of this misnomer: a review of the Web sites of renowned universities reveals use of the same terminology. The readers would have been better served if the authors had discussed “acute aortic dissection” and “ruptured aortic aneurysm” as two separate causes of back pain, with a note that in rare cases an aortic aneurysm can develop a dissection.

To the Editor: In their otherwise excellent review, “Masquerade: Medical causes of back pain” (Cleve Clin J Med 2007; 74:905–913), Dr. Klineberg et al seem to confuse two distinct pathologic processes—aortic dissection and rupture of an aortic aneurysm. Parts of their description seem to fit the pathology of abdominal aortic aneurysm, with a pulsatile abdominal mass, sentinel bleeding, and rupture risk with a size over 6 cm, whereas other parts seem to correspond to aortic dissection, with severe, ripping pain and an association with Marfan syndrome. They also use the terminology “dissecting aortic aneurysm,” which again implies a single entity, when in fact the two conditions rarely occur together. The authors are not alone in their use of this misnomer: a review of the Web sites of renowned universities reveals use of the same terminology. The readers would have been better served if the authors had discussed “acute aortic dissection” and “ruptured aortic aneurysm” as two separate causes of back pain, with a note that in rare cases an aortic aneurysm can develop a dissection.

In Reply: We appreciate Dr. Hirsch’s comments and are pleased to expand the discussion of this important point.

He is correct in his assertion that dissection and aneurysm are distinct processes. But the goal of this review was to remind practitioners to consider the aorta as a possible source of pain when it occurs acutely or in an atypical manner.

A number of aortic processes can cause back pain, and aneurysm and dissection are two of them, aneurysm being more common than aortic dissection. But the pain can also be from aortic ulceration, aortitis, contained rupture of an aneurysm, and other more esoteric problems.

Aortic dissection often presents as a tearing, severe, thoracic back pain. Pain from a progressive abdominal aneurysm is more commonly referred to the lower back or flank and can be severe and unrelenting. It is rarely described as a tearing pain like that of dissection.

It is difficult on initial physical examination to distinguish aneurysm from dissection. The key to diagnosing aneurysm is to detect the pulsatile abdominal mass. A pulsatile, tender abdominal mass with hypotension and back pain is classically associated with rupture of an abdominal aortic aneurysm. The combination of back pain, a deficit in peripheral pulses, and hypertension is more often associated with dissection.

Without imaging and appropriate consultation, it is difficult for even an experienced provider to definitively diagnose these disorders. Without a bit of suspicion, even with a careful physical examination either disorder might be overlooked entirely, with disastrous effect. The purpose of our review was to remind the reader that these conditions, while uncommon or even rare, do occur and should be sought out in patients presenting with acute, atypical lumbar and thoracic back pain. As with each of the conditions discussed in this review, the decision to linger a bit over the patient’s history and then perform a basic, focused physical examination can be life-saving.

In Reply: We appreciate Dr. Hirsch’s comments and are pleased to expand the discussion of this important point.

He is correct in his assertion that dissection and aneurysm are distinct processes. But the goal of this review was to remind practitioners to consider the aorta as a possible source of pain when it occurs acutely or in an atypical manner.

A number of aortic processes can cause back pain, and aneurysm and dissection are two of them, aneurysm being more common than aortic dissection. But the pain can also be from aortic ulceration, aortitis, contained rupture of an aneurysm, and other more esoteric problems.

Aortic dissection often presents as a tearing, severe, thoracic back pain. Pain from a progressive abdominal aneurysm is more commonly referred to the lower back or flank and can be severe and unrelenting. It is rarely described as a tearing pain like that of dissection.

It is difficult on initial physical examination to distinguish aneurysm from dissection. The key to diagnosing aneurysm is to detect the pulsatile abdominal mass. A pulsatile, tender abdominal mass with hypotension and back pain is classically associated with rupture of an abdominal aortic aneurysm. The combination of back pain, a deficit in peripheral pulses, and hypertension is more often associated with dissection.

Without imaging and appropriate consultation, it is difficult for even an experienced provider to definitively diagnose these disorders. Without a bit of suspicion, even with a careful physical examination either disorder might be overlooked entirely, with disastrous effect. The purpose of our review was to remind the reader that these conditions, while uncommon or even rare, do occur and should be sought out in patients presenting with acute, atypical lumbar and thoracic back pain. As with each of the conditions discussed in this review, the decision to linger a bit over the patient’s history and then perform a basic, focused physical examination can be life-saving.

In Reply: We appreciate Dr. Hirsch’s comments and are pleased to expand the discussion of this important point.

He is correct in his assertion that dissection and aneurysm are distinct processes. But the goal of this review was to remind practitioners to consider the aorta as a possible source of pain when it occurs acutely or in an atypical manner.

A number of aortic processes can cause back pain, and aneurysm and dissection are two of them, aneurysm being more common than aortic dissection. But the pain can also be from aortic ulceration, aortitis, contained rupture of an aneurysm, and other more esoteric problems.

Aortic dissection often presents as a tearing, severe, thoracic back pain. Pain from a progressive abdominal aneurysm is more commonly referred to the lower back or flank and can be severe and unrelenting. It is rarely described as a tearing pain like that of dissection.

It is difficult on initial physical examination to distinguish aneurysm from dissection. The key to diagnosing aneurysm is to detect the pulsatile abdominal mass. A pulsatile, tender abdominal mass with hypotension and back pain is classically associated with rupture of an abdominal aortic aneurysm. The combination of back pain, a deficit in peripheral pulses, and hypertension is more often associated with dissection.

Without imaging and appropriate consultation, it is difficult for even an experienced provider to definitively diagnose these disorders. Without a bit of suspicion, even with a careful physical examination either disorder might be overlooked entirely, with disastrous effect. The purpose of our review was to remind the reader that these conditions, while uncommon or even rare, do occur and should be sought out in patients presenting with acute, atypical lumbar and thoracic back pain. As with each of the conditions discussed in this review, the decision to linger a bit over the patient’s history and then perform a basic, focused physical examination can be life-saving.

To the Editor: I read with interest the exchange of letters between Drs. Norenberg and Graves in the December 2007 issue,^1,2 which followed Dr. Graves’ article in the October 2007 issue.³ Dr. Norenberg suggests that it is not always prudent to try to push systolic pressures below 140 mm Hg in the elderly, and Dr. Graves takes the position that physicians like Dr. Norenberg have been “too slow to adapt to evidence-based guidelines for quality of care.” I would like to focus on Dr. Graves’ reference to evidence-based guidelines for the treatment of systolic hypertension in the elderly.

Although there have been multiple published studies of the treatment of this disorder, none has achieved an average systolic blood pressure lower than 140. The Systolic Hypertension in the Elderly Program (SHEP)⁴ came closest with a final systolic blood pressure of 144. No study has ever documented the efficacy and safety of achieving systolic blood pressures less than 140 in a cohort of elderly patients, and there is substantial evidence that excessive lowering of diastolic blood pressure can be harmful.^5,6

Many elderly patients can achieve the target referenced by Dr. Graves, and it is reasonable to expect physicians to continue to strive for that goal, but it would be unwise to push all seniors below 140 systolic. Consider the elderly patient with systolic hypertension who is on a robust three-drug regimen including a diuretic, with a blood pressure of 144/60 and with persistent but tolerable drug side effects. I am aware of no clinical trials that demonstrate that further lowering of this patient’s blood pressure would provide incremental benefit to outweigh the potential risks and costs of additional medications.

We need to be careful not to confuse evidence-based medicine with high-placed opinions, which can result in rigid approaches to treatment that are not in the best interest of our patients.

To the Editor: I read with interest the exchange of letters between Drs. Norenberg and Graves in the December 2007 issue,^1,2 which followed Dr. Graves’ article in the October 2007 issue.³ Dr. Norenberg suggests that it is not always prudent to try to push systolic pressures below 140 mm Hg in the elderly, and Dr. Graves takes the position that physicians like Dr. Norenberg have been “too slow to adapt to evidence-based guidelines for quality of care.” I would like to focus on Dr. Graves’ reference to evidence-based guidelines for the treatment of systolic hypertension in the elderly.

Although there have been multiple published studies of the treatment of this disorder, none has achieved an average systolic blood pressure lower than 140. The Systolic Hypertension in the Elderly Program (SHEP)⁴ came closest with a final systolic blood pressure of 144. No study has ever documented the efficacy and safety of achieving systolic blood pressures less than 140 in a cohort of elderly patients, and there is substantial evidence that excessive lowering of diastolic blood pressure can be harmful.^5,6

Many elderly patients can achieve the target referenced by Dr. Graves, and it is reasonable to expect physicians to continue to strive for that goal, but it would be unwise to push all seniors below 140 systolic. Consider the elderly patient with systolic hypertension who is on a robust three-drug regimen including a diuretic, with a blood pressure of 144/60 and with persistent but tolerable drug side effects. I am aware of no clinical trials that demonstrate that further lowering of this patient’s blood pressure would provide incremental benefit to outweigh the potential risks and costs of additional medications.

We need to be careful not to confuse evidence-based medicine with high-placed opinions, which can result in rigid approaches to treatment that are not in the best interest of our patients.

To the Editor: I read with interest the exchange of letters between Drs. Norenberg and Graves in the December 2007 issue,^1,2 which followed Dr. Graves’ article in the October 2007 issue.³ Dr. Norenberg suggests that it is not always prudent to try to push systolic pressures below 140 mm Hg in the elderly, and Dr. Graves takes the position that physicians like Dr. Norenberg have been “too slow to adapt to evidence-based guidelines for quality of care.” I would like to focus on Dr. Graves’ reference to evidence-based guidelines for the treatment of systolic hypertension in the elderly.

Although there have been multiple published studies of the treatment of this disorder, none has achieved an average systolic blood pressure lower than 140. The Systolic Hypertension in the Elderly Program (SHEP)⁴ came closest with a final systolic blood pressure of 144. No study has ever documented the efficacy and safety of achieving systolic blood pressures less than 140 in a cohort of elderly patients, and there is substantial evidence that excessive lowering of diastolic blood pressure can be harmful.^5,6

Many elderly patients can achieve the target referenced by Dr. Graves, and it is reasonable to expect physicians to continue to strive for that goal, but it would be unwise to push all seniors below 140 systolic. Consider the elderly patient with systolic hypertension who is on a robust three-drug regimen including a diuretic, with a blood pressure of 144/60 and with persistent but tolerable drug side effects. I am aware of no clinical trials that demonstrate that further lowering of this patient’s blood pressure would provide incremental benefit to outweigh the potential risks and costs of additional medications.

We need to be careful not to confuse evidence-based medicine with high-placed opinions, which can result in rigid approaches to treatment that are not in the best interest of our patients.

In Reply: First, I am gratified by the tremendous interest in the care of the hypertensive patient that my article has generated. Dr. Norenberg and Dr. Kelleher are insightful clinicians, as evidenced by the issues that their letters raise. Secondly, as I am now 54 years old, SHEP’s definition of “elderly” as 60 years old and older appears less accurate to me! However, I think we might all agree that to date there has not been a trial with people 65 years old and younger that has not shown benefit to treatment of the blood pressure to less than 140/90 mm Hg.

I believe that Dr. Kelleher’s quest for more “evidence-based” data refers to treatment data in patients above that age. Hopefully, this quest will be answered by the results of the Hypertension in the Very Elderly Trial (HYVET).¹ In this trial, 3,845 patients older than 80 years were treated to less than 140/90 mm Hg. On July 12, 2007, the trial was stopped by the data safety and monitoring board, with the expectation of published results at the European Society of Hypertension and International Society of Hypertension joint meeting in Berlin in 2008.

Third, I must remind the reader that in practicing evidence-based medicine, we clinicians always must interpret the results of double-blind placebo-controlled trials, which tell us the mean effect of a treatment, but apply this information to the individual patient seated in front of us. A recent study² of individual blood pressure response to four forms of monotherapy showed that, in some patients, the blood pressure rose with hydrochlorothiazide instead of falling!

Fourth, Dr. Kelleher implies, correctly, that not all patients can reach the target of less than 140/90. In this regard I think the Antihypertensive and Lipid-Lowering Treatment to Prevent Heart Attack Trial (ALLHAT)³ is very instructive. ALLHAT is the first trial ever to show improvement in the percent of people reaching goal blood pressure, rising from 52% to 63% during the 5-year study. ALLHAT shows us how good we can be and that we should not accept the failure to reach goal blood pressure in at least two-thirds of our patients.

The final and most important point is that the time for arguing the guideline recommendations^4–6 based on our own opinion is past. Third-party payers and patients are demanding we meet those guidelines until new information suggests that they need to be altered. HYVET may force such an alteration, but until then Dr. Norenberg, Dr. Kelleher, and I must attempt to reach the target of less than 140/90 in the majority of our patients.

The final and most important point is that the time for arguing the guideline recommendations^4–6 based on our own opinion is past. Third-party payers and patients are demanding we meet those guidelines until new information suggests that they need to be altered. HYVET may force such an alteration, but until then Dr. Norenberg, Dr. Kelleher, and I must attempt to reach the target of less than 140/90 in the majority of our patients.