Christian Nasr, MD

Whether subclinical hypothyroidism is clinically important and should be treated remains controversial. Studies have differed in their findings, and although most have found this condition to be associated with a variety of adverse outcomes, large randomized controlled trials are needed to clearly demonstrate its clinical impact in various age groups and the benefit of levothyroxine therapy.

Currently, the best practical approach is to base treatment decisions on the magnitude of elevation of thyroid-stimulating hormone (TSH) and whether the patient has thyroid autoantibodies and associated comorbid conditions.

HIGH TSH, NORMAL FREE T₄ LEVELS

Subclinical hypothyroidism is defined by elevated TSH along with a normal free thyroxine (T₄).¹

The hypothalamic-pituitary-thyroid axis is a balanced homeostatic system, and TSH and thyroid hormone levels have an inverse log-linear relation: if free T₄ and triiodothyronine (T₃) levels go down even a little, TSH levels go up a lot.²

TSH secretion is pulsatile and has a circadian rhythm: serum TSH levels are 50% higher at night and early in the morning than during the rest of the day. Thus, repeated measurements in the same patient can vary by as much as half of the reference range.³

WHAT IS THE UPPER LIMIT OF NORMAL FOR TSH?

The upper limit of normal for TSH, defined as the 97.5th percentile, is approximately 4 or 5 mIU/L depending on the laboratory and the population, but some experts believe it should be lower.³

In favor of a lower upper limit: the distribution of serum TSH levels in the healthy general population does not seem to be a typical bell-shaped Gaussian curve, but rather has a tail at the high end. Some argue that some of the individuals with values in the upper end of the normal range may actually have undiagnosed hypothyroidism and that the upper 97.5th percentile cutoff would be 2.5 mIU/L if these people were excluded.⁴ Also, TSH levels higher than 2.5 mIU/L have been associated with a higher prevalence of antithyroid antibodies and a higher risk of clinical hypothyroidism.⁵

On the other hand, lowering the upper limit of normal to 2.5 mIU/L would result in 4 times as many people receiving a diagnosis of subclinical hypothyroidism, or 22 to 28 million people in the United States.^4,6 Thus, lowering the cutoff may lead to unnecessary therapy and could even harm from overtreatment.

Another argument against lowering the upper limit of normal for TSH is that, with age, serum TSH levels shift higher.⁷ The third National Health and Nutrition Education Survey (NHANES III) found that the 97.5th percentile for serum TSH was 3.56 mIU/L for age group 20 to 29 but 7.49 mIU/L for octogenarians.^7,8

It has been suggested that the upper limit of normal for TSH be adjusted for age, race, sex, and iodine intake.³ Currently available TSH reference ranges are not adjusted for these variables, and there is not enough evidence to suggest age-appropriate ranges,⁹ although higher TSH cutoffs for treatment are advised in elderly patients.¹⁰ Interestingly, higher TSH in older people has been linked to lower mortality rates in some studies.¹¹

Authors of the NHANES III⁸ and Hanford Thyroid Disease study¹² have proposed a cutoff of 4.1 mIU/L for the upper limit of normal for serum TSH in patients with negative antithyroid antibodies and normal findings on thyroid ultrasonography.

SUBCLINICAL HYPOTHYROIDISM IS COMMON

In different studies, the prevalence of subclinical hypothyroidism has been as low as 4% and as high as 20%.^1,8,13 The prevalence is higher in women and increases with age.⁸ It is higher in iodine-sufficient areas, and it increases in iodine-deficient areas with iodine supplementation.¹⁴ Genetics also plays a role, as subclinical hypothyroidism is more common in white people than in African Americans.⁸

A difficulty in estimating the prevalence is the disagreement about the cutoff for TSH, which may differ from that in the general population in certain subgroups such as adolescents, the elderly, and pregnant women.^10,15

A VARIETY OF CAUSES

The most common cause of subclinical hypothyroidism, accounting for 60% to 80% of cases, is Hashimoto (autoimmune) thyroiditis,² in which thyroid peroxidase antibodies are usually present.^2,16

Also important to rule out are false-positive elevations due to substances that interfere with TSH assays (eg, heterophile antibodies, rheumatoid factor, biotin, macro-TSH); reversible causes such as the recovery phase of euthyroid sick syndrome; subacute, painless, or postpartum thyroiditis; central hypo- or hyperthyroidism; and thyroid hormone resistance.

SUBCLINICAL HYPOTHYROIDISM CAN RESOLVE OR PROGRESS 


“Subclinical” suggests that the disease is in its early stage, with changes in TSH already apparent but decreases in thyroid hormone levels yet to come.¹⁷ And indeed, subclinical hypothyroidism can progress to overt hypothyroidism,¹⁸ although it has been reported to resolve spontaneously in half of cases within 2 years,¹⁹ typically in patients with TSH values of 4 to 6 mIU/L.²⁰ The rate of progression to overt hypothyroidism is estimated to be 33% to 55% over 10 to 20 years of follow-up.¹⁸

Figure 1 shows the natural history of subclinical hypothyroidism.¹

GUIDELINES FOR SCREENING DIFFER

Guidelines differ on screening for thyroid disease in the general population, owing to lack of large-scale randomized controlled trials showing treatment benefit in otherwise-healthy people with mildly elevated TSH values.

Various professional societies have adopted different criteria for aggressive case-finding in patients at risk of thyroid disease. Risk factors include family history of thyroid disease, neck irradiation, partial thyroidectomy, dyslipidemia, atrial fibrillation, unexplained weight loss, hyperprolactinemia, autoimmune disorders, and use of medications affecting thyroid function.²³

The US Preventive Services Task Force in 2014 found insufficient evidence on the benefits and harms of screening.²⁴

The American Thyroid Association (ATA) recommends screening adults starting at age 35, with repeat testing every 5 years in patients who have no signs or symptoms of hypothyroidism, and more frequently in those who do.²⁵

The American Association of Clinical Endocrinologists recommends screening in women and older patients. Their guidelines and those of the ATA also suggest screening people at high risk of thyroid disease due to risk factors such as history of autoimmune diseases, neck irradiation, or medications affecting thyroid function.²⁶

The American Academy of Family Physicians recommends screening after age 60.¹⁸

The American College of Physicians recommends screening patients over age 50 who have symptoms.¹⁸

Our approach. Although evidence is lacking to recommend routine screening in adults, aggressive case-finding and treatment in patients at risk of thyroid disease can, we believe, offset the risks associated with subclinical hypothyroidism.²⁴

CLINICAL PRESENTATION

About 70% of patients with subclinical hypothyroidism have no symptoms.¹³

Tiredness was more common in subclinical hypothyroid patients with TSH levels lower than 10 mIU/L compared with euthyroid controls in 1 study, but other studies have been unable to replicate this finding.^27,28

Other frequently reported symptoms include dry skin, cognitive slowing, poor memory, muscle weakness, cold intolerance, constipation, puffy eyes, and hoarseness.¹³

The evidence in favor of levothyroxine therapy to improve symptoms in subclinical hypothyroidism has varied, with some studies showing an improvement in symptom scores compared with placebo, while others have not shown any benefit.^29–31

In one study, the average TSH value for patients whose symptoms did not improve with therapy was 4.6 mIU/L.³¹ An explanation for the lack of effect in this group may be that the TSH values for these patients were in the high-normal range. Also, because most subclinical hypothyroid patients have no symptoms, it is difficult to ascertain symptomatic improvement. Though it is possible to conclude that levothyroxine therapy has a limited role in this group, it is important to also consider the suggestive evidence that untreated subclinical hypothyroidism may lead to increased morbidity and mortality.

ADVERSE EFFECTS OF SUBCLINICAL HYPOTHYROIDISM, EFFECTS OF THERAPY

INDIVIDUALIZED MANAGEMENT AND SHARED DECISION-MAKING

The management of subclinical hypothyroidism should be individualized on the basis of extent of thyroid dysfunction, comorbid conditions, risk factors, and patient preference.¹¹⁸ Shared decision-making is key, weighing the risks and benefits of levothyroxine treatment and the patient’s goals.

The risks of treatment should be kept in mind and explained to the patient. Levothyroxine has a narrow therapeutic range, causing a possibility of overreplacement, and a half-life of 7 days that can cause dosing errors to have longer effect.^118,119

Adherence can be a challenge. The drug needs to be taken on an empty stomach because foods and supplements interfere with its absorption.^118,120 In addition, the cost of medication, frequent biochemical monitoring, and possible need for titration can add to financial burden.

When choosing the dose, one should consider the degree of hypothyroidism or TSH elevation and the patient’s weight, and adjust the dose gently.

If the TSH is high-normal

It is proposed that a TSH range of 3 to 5 mIU/L overlaps with normal thyroid function in a great segment of the population, and at this level it is probably not associated with clinically significant consequences. For these reasons, levothyroxine therapy is not thought to be beneficial for those with TSH in this range.

Pollock et al¹²¹ found that, in patients with symptoms suggesting hypothyroidism and TSH values in the upper end of the normal range, there was no improvement in cognitive function or psychological well-being after 12 weeks of levothyroxine therapy.

However, due to the concern for possible adverse maternal and fetal outcomes and low IQ in children of pregnant patients with subclinical hypothyroidism, levothyroxine therapy is advised in those who are pregnant or planning pregnancy who have TSH levels higher than 2.5 mIU/L, especially if they have thyroid peroxidase antibody. Levothyroxine therapy is not recommended for pregnant patients with negative thyroid peroxidase antibody and TSH within the pregnancy-specific range or less than 4 mIU/L if the reference ranges are unavailable.

Keep in mind that, even at these TSH values, there is risk of progression to overt hypothyroidism, especially in the presence of thyroid peroxidase antibody, so patients in this group should be monitored closely.

If TSH is mildly elevated

The evidence to support levothyroxine therapy in patients with subclinical hypothyroidism with TSH levels less than 10 mIU/L remains inconclusive, and the decision to treat should be based on clinical judgment.² The studies that have looked at the benefit of treating subclinical hypothyroidism in terms of cardiac, neuromuscular, cognitive, and neuropsychiatric outcomes have included patients with a wide range of TSH levels, and some of these studies were not stratified on the basis of degree of TSH elevation.

The risk that subclinical hypothyroidism will progress to overt hypothyroidism in patients with TSH higher than 8 mIU/L is high, and in 70% of these patients, the TSH level rises to more than 10 mIU/L within 4 years. Early treatment should be considered if the TSH is higher than 7 or 8 mIU/L.

If TSH is higher than 10 mIU/L

Figure 2 outlines an algorithmic approach to subclinical hypothyroidism in nonpregnant patients as suggested by Peeters.¹²²

Whether subclinical hypothyroidism is clinically important and should be treated remains controversial. Studies have differed in their findings, and although most have found this condition to be associated with a variety of adverse outcomes, large randomized controlled trials are needed to clearly demonstrate its clinical impact in various age groups and the benefit of levothyroxine therapy.

Currently, the best practical approach is to base treatment decisions on the magnitude of elevation of thyroid-stimulating hormone (TSH) and whether the patient has thyroid autoantibodies and associated comorbid conditions.

HIGH TSH, NORMAL FREE T₄ LEVELS

Subclinical hypothyroidism is defined by elevated TSH along with a normal free thyroxine (T₄).¹

The hypothalamic-pituitary-thyroid axis is a balanced homeostatic system, and TSH and thyroid hormone levels have an inverse log-linear relation: if free T₄ and triiodothyronine (T₃) levels go down even a little, TSH levels go up a lot.²

TSH secretion is pulsatile and has a circadian rhythm: serum TSH levels are 50% higher at night and early in the morning than during the rest of the day. Thus, repeated measurements in the same patient can vary by as much as half of the reference range.³

WHAT IS THE UPPER LIMIT OF NORMAL FOR TSH?

The upper limit of normal for TSH, defined as the 97.5th percentile, is approximately 4 or 5 mIU/L depending on the laboratory and the population, but some experts believe it should be lower.³

In favor of a lower upper limit: the distribution of serum TSH levels in the healthy general population does not seem to be a typical bell-shaped Gaussian curve, but rather has a tail at the high end. Some argue that some of the individuals with values in the upper end of the normal range may actually have undiagnosed hypothyroidism and that the upper 97.5th percentile cutoff would be 2.5 mIU/L if these people were excluded.⁴ Also, TSH levels higher than 2.5 mIU/L have been associated with a higher prevalence of antithyroid antibodies and a higher risk of clinical hypothyroidism.⁵

On the other hand, lowering the upper limit of normal to 2.5 mIU/L would result in 4 times as many people receiving a diagnosis of subclinical hypothyroidism, or 22 to 28 million people in the United States.^4,6 Thus, lowering the cutoff may lead to unnecessary therapy and could even harm from overtreatment.

Another argument against lowering the upper limit of normal for TSH is that, with age, serum TSH levels shift higher.⁷ The third National Health and Nutrition Education Survey (NHANES III) found that the 97.5th percentile for serum TSH was 3.56 mIU/L for age group 20 to 29 but 7.49 mIU/L for octogenarians.^7,8

It has been suggested that the upper limit of normal for TSH be adjusted for age, race, sex, and iodine intake.³ Currently available TSH reference ranges are not adjusted for these variables, and there is not enough evidence to suggest age-appropriate ranges,⁹ although higher TSH cutoffs for treatment are advised in elderly patients.¹⁰ Interestingly, higher TSH in older people has been linked to lower mortality rates in some studies.¹¹

Authors of the NHANES III⁸ and Hanford Thyroid Disease study¹² have proposed a cutoff of 4.1 mIU/L for the upper limit of normal for serum TSH in patients with negative antithyroid antibodies and normal findings on thyroid ultrasonography.

SUBCLINICAL HYPOTHYROIDISM IS COMMON

In different studies, the prevalence of subclinical hypothyroidism has been as low as 4% and as high as 20%.^1,8,13 The prevalence is higher in women and increases with age.⁸ It is higher in iodine-sufficient areas, and it increases in iodine-deficient areas with iodine supplementation.¹⁴ Genetics also plays a role, as subclinical hypothyroidism is more common in white people than in African Americans.⁸

A difficulty in estimating the prevalence is the disagreement about the cutoff for TSH, which may differ from that in the general population in certain subgroups such as adolescents, the elderly, and pregnant women.^10,15

A VARIETY OF CAUSES

The most common cause of subclinical hypothyroidism, accounting for 60% to 80% of cases, is Hashimoto (autoimmune) thyroiditis,² in which thyroid peroxidase antibodies are usually present.^2,16

Also important to rule out are false-positive elevations due to substances that interfere with TSH assays (eg, heterophile antibodies, rheumatoid factor, biotin, macro-TSH); reversible causes such as the recovery phase of euthyroid sick syndrome; subacute, painless, or postpartum thyroiditis; central hypo- or hyperthyroidism; and thyroid hormone resistance.

SUBCLINICAL HYPOTHYROIDISM CAN RESOLVE OR PROGRESS 


“Subclinical” suggests that the disease is in its early stage, with changes in TSH already apparent but decreases in thyroid hormone levels yet to come.¹⁷ And indeed, subclinical hypothyroidism can progress to overt hypothyroidism,¹⁸ although it has been reported to resolve spontaneously in half of cases within 2 years,¹⁹ typically in patients with TSH values of 4 to 6 mIU/L.²⁰ The rate of progression to overt hypothyroidism is estimated to be 33% to 55% over 10 to 20 years of follow-up.¹⁸

Figure 1 shows the natural history of subclinical hypothyroidism.¹

GUIDELINES FOR SCREENING DIFFER

Guidelines differ on screening for thyroid disease in the general population, owing to lack of large-scale randomized controlled trials showing treatment benefit in otherwise-healthy people with mildly elevated TSH values.

Various professional societies have adopted different criteria for aggressive case-finding in patients at risk of thyroid disease. Risk factors include family history of thyroid disease, neck irradiation, partial thyroidectomy, dyslipidemia, atrial fibrillation, unexplained weight loss, hyperprolactinemia, autoimmune disorders, and use of medications affecting thyroid function.²³

The US Preventive Services Task Force in 2014 found insufficient evidence on the benefits and harms of screening.²⁴

The American Thyroid Association (ATA) recommends screening adults starting at age 35, with repeat testing every 5 years in patients who have no signs or symptoms of hypothyroidism, and more frequently in those who do.²⁵

The American Association of Clinical Endocrinologists recommends screening in women and older patients. Their guidelines and those of the ATA also suggest screening people at high risk of thyroid disease due to risk factors such as history of autoimmune diseases, neck irradiation, or medications affecting thyroid function.²⁶

The American Academy of Family Physicians recommends screening after age 60.¹⁸

The American College of Physicians recommends screening patients over age 50 who have symptoms.¹⁸

Our approach. Although evidence is lacking to recommend routine screening in adults, aggressive case-finding and treatment in patients at risk of thyroid disease can, we believe, offset the risks associated with subclinical hypothyroidism.²⁴

CLINICAL PRESENTATION

About 70% of patients with subclinical hypothyroidism have no symptoms.¹³

Tiredness was more common in subclinical hypothyroid patients with TSH levels lower than 10 mIU/L compared with euthyroid controls in 1 study, but other studies have been unable to replicate this finding.^27,28

Other frequently reported symptoms include dry skin, cognitive slowing, poor memory, muscle weakness, cold intolerance, constipation, puffy eyes, and hoarseness.¹³

The evidence in favor of levothyroxine therapy to improve symptoms in subclinical hypothyroidism has varied, with some studies showing an improvement in symptom scores compared with placebo, while others have not shown any benefit.^29–31

In one study, the average TSH value for patients whose symptoms did not improve with therapy was 4.6 mIU/L.³¹ An explanation for the lack of effect in this group may be that the TSH values for these patients were in the high-normal range. Also, because most subclinical hypothyroid patients have no symptoms, it is difficult to ascertain symptomatic improvement. Though it is possible to conclude that levothyroxine therapy has a limited role in this group, it is important to also consider the suggestive evidence that untreated subclinical hypothyroidism may lead to increased morbidity and mortality.

ADVERSE EFFECTS OF SUBCLINICAL HYPOTHYROIDISM, EFFECTS OF THERAPY

INDIVIDUALIZED MANAGEMENT AND SHARED DECISION-MAKING

The management of subclinical hypothyroidism should be individualized on the basis of extent of thyroid dysfunction, comorbid conditions, risk factors, and patient preference.¹¹⁸ Shared decision-making is key, weighing the risks and benefits of levothyroxine treatment and the patient’s goals.

The risks of treatment should be kept in mind and explained to the patient. Levothyroxine has a narrow therapeutic range, causing a possibility of overreplacement, and a half-life of 7 days that can cause dosing errors to have longer effect.^118,119

Adherence can be a challenge. The drug needs to be taken on an empty stomach because foods and supplements interfere with its absorption.^118,120 In addition, the cost of medication, frequent biochemical monitoring, and possible need for titration can add to financial burden.

When choosing the dose, one should consider the degree of hypothyroidism or TSH elevation and the patient’s weight, and adjust the dose gently.

If the TSH is high-normal

It is proposed that a TSH range of 3 to 5 mIU/L overlaps with normal thyroid function in a great segment of the population, and at this level it is probably not associated with clinically significant consequences. For these reasons, levothyroxine therapy is not thought to be beneficial for those with TSH in this range.

Pollock et al¹²¹ found that, in patients with symptoms suggesting hypothyroidism and TSH values in the upper end of the normal range, there was no improvement in cognitive function or psychological well-being after 12 weeks of levothyroxine therapy.

However, due to the concern for possible adverse maternal and fetal outcomes and low IQ in children of pregnant patients with subclinical hypothyroidism, levothyroxine therapy is advised in those who are pregnant or planning pregnancy who have TSH levels higher than 2.5 mIU/L, especially if they have thyroid peroxidase antibody. Levothyroxine therapy is not recommended for pregnant patients with negative thyroid peroxidase antibody and TSH within the pregnancy-specific range or less than 4 mIU/L if the reference ranges are unavailable.

Keep in mind that, even at these TSH values, there is risk of progression to overt hypothyroidism, especially in the presence of thyroid peroxidase antibody, so patients in this group should be monitored closely.

If TSH is mildly elevated

The evidence to support levothyroxine therapy in patients with subclinical hypothyroidism with TSH levels less than 10 mIU/L remains inconclusive, and the decision to treat should be based on clinical judgment.² The studies that have looked at the benefit of treating subclinical hypothyroidism in terms of cardiac, neuromuscular, cognitive, and neuropsychiatric outcomes have included patients with a wide range of TSH levels, and some of these studies were not stratified on the basis of degree of TSH elevation.

The risk that subclinical hypothyroidism will progress to overt hypothyroidism in patients with TSH higher than 8 mIU/L is high, and in 70% of these patients, the TSH level rises to more than 10 mIU/L within 4 years. Early treatment should be considered if the TSH is higher than 7 or 8 mIU/L.

If TSH is higher than 10 mIU/L

Figure 2 outlines an algorithmic approach to subclinical hypothyroidism in nonpregnant patients as suggested by Peeters.¹²²

Whether subclinical hypothyroidism is clinically important and should be treated remains controversial. Studies have differed in their findings, and although most have found this condition to be associated with a variety of adverse outcomes, large randomized controlled trials are needed to clearly demonstrate its clinical impact in various age groups and the benefit of levothyroxine therapy.

Currently, the best practical approach is to base treatment decisions on the magnitude of elevation of thyroid-stimulating hormone (TSH) and whether the patient has thyroid autoantibodies and associated comorbid conditions.

HIGH TSH, NORMAL FREE T₄ LEVELS

Subclinical hypothyroidism is defined by elevated TSH along with a normal free thyroxine (T₄).¹

The hypothalamic-pituitary-thyroid axis is a balanced homeostatic system, and TSH and thyroid hormone levels have an inverse log-linear relation: if free T₄ and triiodothyronine (T₃) levels go down even a little, TSH levels go up a lot.²

TSH secretion is pulsatile and has a circadian rhythm: serum TSH levels are 50% higher at night and early in the morning than during the rest of the day. Thus, repeated measurements in the same patient can vary by as much as half of the reference range.³

WHAT IS THE UPPER LIMIT OF NORMAL FOR TSH?

The upper limit of normal for TSH, defined as the 97.5th percentile, is approximately 4 or 5 mIU/L depending on the laboratory and the population, but some experts believe it should be lower.³

In favor of a lower upper limit: the distribution of serum TSH levels in the healthy general population does not seem to be a typical bell-shaped Gaussian curve, but rather has a tail at the high end. Some argue that some of the individuals with values in the upper end of the normal range may actually have undiagnosed hypothyroidism and that the upper 97.5th percentile cutoff would be 2.5 mIU/L if these people were excluded.⁴ Also, TSH levels higher than 2.5 mIU/L have been associated with a higher prevalence of antithyroid antibodies and a higher risk of clinical hypothyroidism.⁵

On the other hand, lowering the upper limit of normal to 2.5 mIU/L would result in 4 times as many people receiving a diagnosis of subclinical hypothyroidism, or 22 to 28 million people in the United States.^4,6 Thus, lowering the cutoff may lead to unnecessary therapy and could even harm from overtreatment.

Another argument against lowering the upper limit of normal for TSH is that, with age, serum TSH levels shift higher.⁷ The third National Health and Nutrition Education Survey (NHANES III) found that the 97.5th percentile for serum TSH was 3.56 mIU/L for age group 20 to 29 but 7.49 mIU/L for octogenarians.^7,8

It has been suggested that the upper limit of normal for TSH be adjusted for age, race, sex, and iodine intake.³ Currently available TSH reference ranges are not adjusted for these variables, and there is not enough evidence to suggest age-appropriate ranges,⁹ although higher TSH cutoffs for treatment are advised in elderly patients.¹⁰ Interestingly, higher TSH in older people has been linked to lower mortality rates in some studies.¹¹

Authors of the NHANES III⁸ and Hanford Thyroid Disease study¹² have proposed a cutoff of 4.1 mIU/L for the upper limit of normal for serum TSH in patients with negative antithyroid antibodies and normal findings on thyroid ultrasonography.

SUBCLINICAL HYPOTHYROIDISM IS COMMON

In different studies, the prevalence of subclinical hypothyroidism has been as low as 4% and as high as 20%.^1,8,13 The prevalence is higher in women and increases with age.⁸ It is higher in iodine-sufficient areas, and it increases in iodine-deficient areas with iodine supplementation.¹⁴ Genetics also plays a role, as subclinical hypothyroidism is more common in white people than in African Americans.⁸

A difficulty in estimating the prevalence is the disagreement about the cutoff for TSH, which may differ from that in the general population in certain subgroups such as adolescents, the elderly, and pregnant women.^10,15

A VARIETY OF CAUSES

The most common cause of subclinical hypothyroidism, accounting for 60% to 80% of cases, is Hashimoto (autoimmune) thyroiditis,² in which thyroid peroxidase antibodies are usually present.^2,16

Also important to rule out are false-positive elevations due to substances that interfere with TSH assays (eg, heterophile antibodies, rheumatoid factor, biotin, macro-TSH); reversible causes such as the recovery phase of euthyroid sick syndrome; subacute, painless, or postpartum thyroiditis; central hypo- or hyperthyroidism; and thyroid hormone resistance.

SUBCLINICAL HYPOTHYROIDISM CAN RESOLVE OR PROGRESS 


“Subclinical” suggests that the disease is in its early stage, with changes in TSH already apparent but decreases in thyroid hormone levels yet to come.¹⁷ And indeed, subclinical hypothyroidism can progress to overt hypothyroidism,¹⁸ although it has been reported to resolve spontaneously in half of cases within 2 years,¹⁹ typically in patients with TSH values of 4 to 6 mIU/L.²⁰ The rate of progression to overt hypothyroidism is estimated to be 33% to 55% over 10 to 20 years of follow-up.¹⁸

Figure 1 shows the natural history of subclinical hypothyroidism.¹

GUIDELINES FOR SCREENING DIFFER

Guidelines differ on screening for thyroid disease in the general population, owing to lack of large-scale randomized controlled trials showing treatment benefit in otherwise-healthy people with mildly elevated TSH values.

Various professional societies have adopted different criteria for aggressive case-finding in patients at risk of thyroid disease. Risk factors include family history of thyroid disease, neck irradiation, partial thyroidectomy, dyslipidemia, atrial fibrillation, unexplained weight loss, hyperprolactinemia, autoimmune disorders, and use of medications affecting thyroid function.²³

The US Preventive Services Task Force in 2014 found insufficient evidence on the benefits and harms of screening.²⁴

The American Thyroid Association (ATA) recommends screening adults starting at age 35, with repeat testing every 5 years in patients who have no signs or symptoms of hypothyroidism, and more frequently in those who do.²⁵

The American Association of Clinical Endocrinologists recommends screening in women and older patients. Their guidelines and those of the ATA also suggest screening people at high risk of thyroid disease due to risk factors such as history of autoimmune diseases, neck irradiation, or medications affecting thyroid function.²⁶

The American Academy of Family Physicians recommends screening after age 60.¹⁸

The American College of Physicians recommends screening patients over age 50 who have symptoms.¹⁸

Our approach. Although evidence is lacking to recommend routine screening in adults, aggressive case-finding and treatment in patients at risk of thyroid disease can, we believe, offset the risks associated with subclinical hypothyroidism.²⁴

CLINICAL PRESENTATION

About 70% of patients with subclinical hypothyroidism have no symptoms.¹³

Tiredness was more common in subclinical hypothyroid patients with TSH levels lower than 10 mIU/L compared with euthyroid controls in 1 study, but other studies have been unable to replicate this finding.^27,28

Other frequently reported symptoms include dry skin, cognitive slowing, poor memory, muscle weakness, cold intolerance, constipation, puffy eyes, and hoarseness.¹³

The evidence in favor of levothyroxine therapy to improve symptoms in subclinical hypothyroidism has varied, with some studies showing an improvement in symptom scores compared with placebo, while others have not shown any benefit.^29–31

In one study, the average TSH value for patients whose symptoms did not improve with therapy was 4.6 mIU/L.³¹ An explanation for the lack of effect in this group may be that the TSH values for these patients were in the high-normal range. Also, because most subclinical hypothyroid patients have no symptoms, it is difficult to ascertain symptomatic improvement. Though it is possible to conclude that levothyroxine therapy has a limited role in this group, it is important to also consider the suggestive evidence that untreated subclinical hypothyroidism may lead to increased morbidity and mortality.

ADVERSE EFFECTS OF SUBCLINICAL HYPOTHYROIDISM, EFFECTS OF THERAPY

INDIVIDUALIZED MANAGEMENT AND SHARED DECISION-MAKING

The management of subclinical hypothyroidism should be individualized on the basis of extent of thyroid dysfunction, comorbid conditions, risk factors, and patient preference.¹¹⁸ Shared decision-making is key, weighing the risks and benefits of levothyroxine treatment and the patient’s goals.

The risks of treatment should be kept in mind and explained to the patient. Levothyroxine has a narrow therapeutic range, causing a possibility of overreplacement, and a half-life of 7 days that can cause dosing errors to have longer effect.^118,119

Adherence can be a challenge. The drug needs to be taken on an empty stomach because foods and supplements interfere with its absorption.^118,120 In addition, the cost of medication, frequent biochemical monitoring, and possible need for titration can add to financial burden.

When choosing the dose, one should consider the degree of hypothyroidism or TSH elevation and the patient’s weight, and adjust the dose gently.

If the TSH is high-normal

It is proposed that a TSH range of 3 to 5 mIU/L overlaps with normal thyroid function in a great segment of the population, and at this level it is probably not associated with clinically significant consequences. For these reasons, levothyroxine therapy is not thought to be beneficial for those with TSH in this range.

Pollock et al¹²¹ found that, in patients with symptoms suggesting hypothyroidism and TSH values in the upper end of the normal range, there was no improvement in cognitive function or psychological well-being after 12 weeks of levothyroxine therapy.

However, due to the concern for possible adverse maternal and fetal outcomes and low IQ in children of pregnant patients with subclinical hypothyroidism, levothyroxine therapy is advised in those who are pregnant or planning pregnancy who have TSH levels higher than 2.5 mIU/L, especially if they have thyroid peroxidase antibody. Levothyroxine therapy is not recommended for pregnant patients with negative thyroid peroxidase antibody and TSH within the pregnancy-specific range or less than 4 mIU/L if the reference ranges are unavailable.

Keep in mind that, even at these TSH values, there is risk of progression to overt hypothyroidism, especially in the presence of thyroid peroxidase antibody, so patients in this group should be monitored closely.

If TSH is mildly elevated

The evidence to support levothyroxine therapy in patients with subclinical hypothyroidism with TSH levels less than 10 mIU/L remains inconclusive, and the decision to treat should be based on clinical judgment.² The studies that have looked at the benefit of treating subclinical hypothyroidism in terms of cardiac, neuromuscular, cognitive, and neuropsychiatric outcomes have included patients with a wide range of TSH levels, and some of these studies were not stratified on the basis of degree of TSH elevation.

The risk that subclinical hypothyroidism will progress to overt hypothyroidism in patients with TSH higher than 8 mIU/L is high, and in 70% of these patients, the TSH level rises to more than 10 mIU/L within 4 years. Early treatment should be considered if the TSH is higher than 7 or 8 mIU/L.

If TSH is higher than 10 mIU/L

Figure 2 outlines an algorithmic approach to subclinical hypothyroidism in nonpregnant patients as suggested by Peeters.¹²²

A 28-year-old woman returns for follow-up of her hypothyroidism. She was diagnosed 4 years ago when she presented with fatigue, “foggy” thinking, poor concentration, cold intolerance, and constipation. Her thyroid-stimulating hormone (TSH) level at that time was elevated at 15 mIU/L (reference range 0.4–4). She was started on 50 µg of levothyroxine daily, which helped her symptoms, but she continued to complain of tiredness and the inability to lose weight. She has been on 100 µg of levothyroxine daily since her last visit 1 year ago.

On examination, she has a small, diffuse, and firm goiter; she has no Cushing-like features, visual field abnormalities, or signs of hypothyroidism.

Her TSH level today is 1.2 mIU/L. Based on this, you recommend no change in her daily levothyroxine dose. She expresses dissatisfaction that you had ordered only a TSH, and she asks you to order thyroxine (T₄) and triiodothyronine (T₃) measurements because she read on the Internet that those were needed to determine the appropriateness of the levothyroxine dose.

Should T₄ or T₃ be routinely measured when adjusting thyroid replacement therapy?

IN PRIMARY HYPOTHYROIDISM, TSH IS ENOUGH

In a patient with primary hypothyroidism and no suspicion of pituitary abnormality, a serum TSH is sufficient for monitoring thyroid status and adjusting the dose of thyroid hormone.

Hypothyroidism is one of the most common endocrine disorders, affecting about 4% of the adult US population.¹ In areas of iodine sufficiency, primary hypothyroidism is due predominantly to Hashimoto thyroiditis.

The role of the lack of thyroid hormone in the pathogenesis of myxedema was recognized in the late 19th century through the observation of a “cretinoid” state occurring in middle-aged women, associated with atrophy of the thyroid gland and a similar severe state noted after total thyroidectomy.²

In 1891, George R. Murray was able to “cure” myxedema in a patient by injecting sheep thyroid extract subcutaneously. Thyroid extracts continued to be the only treatment for hypothyroidism until 1950, when levothyroxine was introduced and later became the main treatment. Around that time, T₃ was discovered and was described as being the physiologically active thyroid hormone. Later, it was noted that 80% to 90% of circulating T₃ is generated through peripheral deiodination of T₄, the latter being considered a prohormone.²

The pituitary-thyroid axis is regulated through negative feedback. At concentrations of free T₄ below normal, plasma TSH rises rapidly with small decrements in T₄ levels.³ The opposite phenomenon occurs with free T₄ concentrations above normal. Since T₄ has a long disappearance half-time—about 7 days—a normal TSH tends to stay relatively stable in the same individual.⁴ The relationship between TSH and T₄ was long thought to be inverse log-linear, but Hadlow et al⁵ found that it is complex and nonlinear and differs by age and sex. TSH and T₄ concentrations have narrower within-individual variability than inter-individual variability. Although environmental factors may affect this hypothalamic-pituitary set-point, there is evidence that heritability is a major determinant of individual variability.⁶

GUIDELINES AND CHOOSING WISELY

In 2014, the American Thyroid Association published comprehensive, evidence-based guidelines for the treatment of hypothyroidism.⁷ The guidelines state that the goal of thyroid hormone replacement is to achieve clinical and biochemical euthyroidism.⁷ TSH continues to be the most reliable marker of adequacy of thyroid hormone replacement in primary hypothyroidism. The guidelines recommend aiming for a TSH in the normal range (generally 0.4–4 mIU/L).

Most studies of the risks associated with hypothyroidism or thyrotoxicosis have looked at TSH levels. Significantly increased risk of cardiovascular mortality and morbidity is seen in individuals with TSH levels higher than 10 mIU/L.⁸ On the other hand, excess thyroid hormone leading to a TSH level lower than 0.1 mIU/L has been associated with an increased risk of atrial fibrillation in older persons and osteoporosis in postmenopausal women.

The classic symptoms and signs of hypothyroidism correlate with biochemical hypothyroidism and usually improve with the restoration of euthyroidism. Some of these symptoms, however, lack sensitivity and specificity, especially with modest degrees of hypothyroidism.⁷ A randomized controlled trial showed that patients were unable to detect any difference in symptoms when the levothyroxine dose was changed by about 20%.⁹

HARMS ASSOCIATED WITH ORDERING T₄ AND T₃

Other than the financial burden to the patient and society, there is no major morbidity caused by obtaining T₄ or T₃ levels, or both. However, knowing the T₄ or T₃ level does not help with management beyond the information offered by the TSH value. Hypothyroid patients treated with levothyroxine to maintain a normal TSH generally have higher free T₄ levels and lower free T₃ levels than euthyroid patients with similar TSH values.¹⁰ Therefore, reacting to a high T₄ level or a low T₃ level in a treated hypothyroid patient with a normal TSH may lead to inappropriate dose adjustment. On the other hand, increasing the dose of thyroid hormone in a patient with a low TSH whose T₃ level is low-normal may lead to morbidity.

SPECIAL SCENARIO: PITUITARY COMPROMISE

We assume that the patient described above has primary hypothyroidism and that her pituitary-thyroid axis is intact. Primary hypothyroidism is diagnosed by a high TSH along with a low or low-normal T₄. In this typical case, TSH can be used to guide therapy without the need for other tests.

However, when there is pituitary compromise (hypopituitarism, congenital central hypothyroidism), the TSH will not be reliable to monitor the adequacy of thyroid hormone replacement therapy. The aim of levothyroxine management in these patients is to maintain a free T₄ concentration in the upper half of the normal range. If the free T₃ concentration is followed and is found to be elevated, the dose of levothyroxine should be reduced.¹¹

CLINICAL BOTTOM LINE

Since our patient’s dose of levothyroxine has been stable and her TSH is not elevated, measuring serum levels of T₄ and T₃ would not contribute to her management. For such a patient, if the TSH were less than 3 mIU/L, increasing the dose would be unlikely to offer clinical benefit.

On the other hand, if her TSH was higher than 4 mIU/L, then it would be legitimate to tweak the dose upward and reassess her thyroid state clinically and biochemically 6 to 8 weeks later. One would need to be careful not to induce thyrotoxicosis through such an intervention because of the potential morbidity.

The TSH level is typically monitored every 6 to 12 months when the patient is clinically stable. It should be measured sooner in circumstances that include the following:

• Symptoms of hypothyroidism or thyrotoxicosis
• Starting a new medication known to affect thyroid hormone levels
• Significant weight change
• Hospitalization
• Pregnancy.

A 28-year-old woman returns for follow-up of her hypothyroidism. She was diagnosed 4 years ago when she presented with fatigue, “foggy” thinking, poor concentration, cold intolerance, and constipation. Her thyroid-stimulating hormone (TSH) level at that time was elevated at 15 mIU/L (reference range 0.4–4). She was started on 50 µg of levothyroxine daily, which helped her symptoms, but she continued to complain of tiredness and the inability to lose weight. She has been on 100 µg of levothyroxine daily since her last visit 1 year ago.

On examination, she has a small, diffuse, and firm goiter; she has no Cushing-like features, visual field abnormalities, or signs of hypothyroidism.

Her TSH level today is 1.2 mIU/L. Based on this, you recommend no change in her daily levothyroxine dose. She expresses dissatisfaction that you had ordered only a TSH, and she asks you to order thyroxine (T₄) and triiodothyronine (T₃) measurements because she read on the Internet that those were needed to determine the appropriateness of the levothyroxine dose.

Should T₄ or T₃ be routinely measured when adjusting thyroid replacement therapy?

IN PRIMARY HYPOTHYROIDISM, TSH IS ENOUGH

In a patient with primary hypothyroidism and no suspicion of pituitary abnormality, a serum TSH is sufficient for monitoring thyroid status and adjusting the dose of thyroid hormone.

Hypothyroidism is one of the most common endocrine disorders, affecting about 4% of the adult US population.¹ In areas of iodine sufficiency, primary hypothyroidism is due predominantly to Hashimoto thyroiditis.

The role of the lack of thyroid hormone in the pathogenesis of myxedema was recognized in the late 19th century through the observation of a “cretinoid” state occurring in middle-aged women, associated with atrophy of the thyroid gland and a similar severe state noted after total thyroidectomy.²

In 1891, George R. Murray was able to “cure” myxedema in a patient by injecting sheep thyroid extract subcutaneously. Thyroid extracts continued to be the only treatment for hypothyroidism until 1950, when levothyroxine was introduced and later became the main treatment. Around that time, T₃ was discovered and was described as being the physiologically active thyroid hormone. Later, it was noted that 80% to 90% of circulating T₃ is generated through peripheral deiodination of T₄, the latter being considered a prohormone.²

The pituitary-thyroid axis is regulated through negative feedback. At concentrations of free T₄ below normal, plasma TSH rises rapidly with small decrements in T₄ levels.³ The opposite phenomenon occurs with free T₄ concentrations above normal. Since T₄ has a long disappearance half-time—about 7 days—a normal TSH tends to stay relatively stable in the same individual.⁴ The relationship between TSH and T₄ was long thought to be inverse log-linear, but Hadlow et al⁵ found that it is complex and nonlinear and differs by age and sex. TSH and T₄ concentrations have narrower within-individual variability than inter-individual variability. Although environmental factors may affect this hypothalamic-pituitary set-point, there is evidence that heritability is a major determinant of individual variability.⁶

GUIDELINES AND CHOOSING WISELY

In 2014, the American Thyroid Association published comprehensive, evidence-based guidelines for the treatment of hypothyroidism.⁷ The guidelines state that the goal of thyroid hormone replacement is to achieve clinical and biochemical euthyroidism.⁷ TSH continues to be the most reliable marker of adequacy of thyroid hormone replacement in primary hypothyroidism. The guidelines recommend aiming for a TSH in the normal range (generally 0.4–4 mIU/L).

Most studies of the risks associated with hypothyroidism or thyrotoxicosis have looked at TSH levels. Significantly increased risk of cardiovascular mortality and morbidity is seen in individuals with TSH levels higher than 10 mIU/L.⁸ On the other hand, excess thyroid hormone leading to a TSH level lower than 0.1 mIU/L has been associated with an increased risk of atrial fibrillation in older persons and osteoporosis in postmenopausal women.

The classic symptoms and signs of hypothyroidism correlate with biochemical hypothyroidism and usually improve with the restoration of euthyroidism. Some of these symptoms, however, lack sensitivity and specificity, especially with modest degrees of hypothyroidism.⁷ A randomized controlled trial showed that patients were unable to detect any difference in symptoms when the levothyroxine dose was changed by about 20%.⁹

HARMS ASSOCIATED WITH ORDERING T₄ AND T₃

Other than the financial burden to the patient and society, there is no major morbidity caused by obtaining T₄ or T₃ levels, or both. However, knowing the T₄ or T₃ level does not help with management beyond the information offered by the TSH value. Hypothyroid patients treated with levothyroxine to maintain a normal TSH generally have higher free T₄ levels and lower free T₃ levels than euthyroid patients with similar TSH values.¹⁰ Therefore, reacting to a high T₄ level or a low T₃ level in a treated hypothyroid patient with a normal TSH may lead to inappropriate dose adjustment. On the other hand, increasing the dose of thyroid hormone in a patient with a low TSH whose T₃ level is low-normal may lead to morbidity.

SPECIAL SCENARIO: PITUITARY COMPROMISE

We assume that the patient described above has primary hypothyroidism and that her pituitary-thyroid axis is intact. Primary hypothyroidism is diagnosed by a high TSH along with a low or low-normal T₄. In this typical case, TSH can be used to guide therapy without the need for other tests.

However, when there is pituitary compromise (hypopituitarism, congenital central hypothyroidism), the TSH will not be reliable to monitor the adequacy of thyroid hormone replacement therapy. The aim of levothyroxine management in these patients is to maintain a free T₄ concentration in the upper half of the normal range. If the free T₃ concentration is followed and is found to be elevated, the dose of levothyroxine should be reduced.¹¹

CLINICAL BOTTOM LINE

Since our patient’s dose of levothyroxine has been stable and her TSH is not elevated, measuring serum levels of T₄ and T₃ would not contribute to her management. For such a patient, if the TSH were less than 3 mIU/L, increasing the dose would be unlikely to offer clinical benefit.

On the other hand, if her TSH was higher than 4 mIU/L, then it would be legitimate to tweak the dose upward and reassess her thyroid state clinically and biochemically 6 to 8 weeks later. One would need to be careful not to induce thyrotoxicosis through such an intervention because of the potential morbidity.

The TSH level is typically monitored every 6 to 12 months when the patient is clinically stable. It should be measured sooner in circumstances that include the following:

• Symptoms of hypothyroidism or thyrotoxicosis
• Starting a new medication known to affect thyroid hormone levels
• Significant weight change
• Hospitalization
• Pregnancy.

A 28-year-old woman returns for follow-up of her hypothyroidism. She was diagnosed 4 years ago when she presented with fatigue, “foggy” thinking, poor concentration, cold intolerance, and constipation. Her thyroid-stimulating hormone (TSH) level at that time was elevated at 15 mIU/L (reference range 0.4–4). She was started on 50 µg of levothyroxine daily, which helped her symptoms, but she continued to complain of tiredness and the inability to lose weight. She has been on 100 µg of levothyroxine daily since her last visit 1 year ago.

On examination, she has a small, diffuse, and firm goiter; she has no Cushing-like features, visual field abnormalities, or signs of hypothyroidism.

Her TSH level today is 1.2 mIU/L. Based on this, you recommend no change in her daily levothyroxine dose. She expresses dissatisfaction that you had ordered only a TSH, and she asks you to order thyroxine (T₄) and triiodothyronine (T₃) measurements because she read on the Internet that those were needed to determine the appropriateness of the levothyroxine dose.

Should T₄ or T₃ be routinely measured when adjusting thyroid replacement therapy?

IN PRIMARY HYPOTHYROIDISM, TSH IS ENOUGH

In a patient with primary hypothyroidism and no suspicion of pituitary abnormality, a serum TSH is sufficient for monitoring thyroid status and adjusting the dose of thyroid hormone.

Hypothyroidism is one of the most common endocrine disorders, affecting about 4% of the adult US population.¹ In areas of iodine sufficiency, primary hypothyroidism is due predominantly to Hashimoto thyroiditis.

The role of the lack of thyroid hormone in the pathogenesis of myxedema was recognized in the late 19th century through the observation of a “cretinoid” state occurring in middle-aged women, associated with atrophy of the thyroid gland and a similar severe state noted after total thyroidectomy.²

In 1891, George R. Murray was able to “cure” myxedema in a patient by injecting sheep thyroid extract subcutaneously. Thyroid extracts continued to be the only treatment for hypothyroidism until 1950, when levothyroxine was introduced and later became the main treatment. Around that time, T₃ was discovered and was described as being the physiologically active thyroid hormone. Later, it was noted that 80% to 90% of circulating T₃ is generated through peripheral deiodination of T₄, the latter being considered a prohormone.²

The pituitary-thyroid axis is regulated through negative feedback. At concentrations of free T₄ below normal, plasma TSH rises rapidly with small decrements in T₄ levels.³ The opposite phenomenon occurs with free T₄ concentrations above normal. Since T₄ has a long disappearance half-time—about 7 days—a normal TSH tends to stay relatively stable in the same individual.⁴ The relationship between TSH and T₄ was long thought to be inverse log-linear, but Hadlow et al⁵ found that it is complex and nonlinear and differs by age and sex. TSH and T₄ concentrations have narrower within-individual variability than inter-individual variability. Although environmental factors may affect this hypothalamic-pituitary set-point, there is evidence that heritability is a major determinant of individual variability.⁶

GUIDELINES AND CHOOSING WISELY

In 2014, the American Thyroid Association published comprehensive, evidence-based guidelines for the treatment of hypothyroidism.⁷ The guidelines state that the goal of thyroid hormone replacement is to achieve clinical and biochemical euthyroidism.⁷ TSH continues to be the most reliable marker of adequacy of thyroid hormone replacement in primary hypothyroidism. The guidelines recommend aiming for a TSH in the normal range (generally 0.4–4 mIU/L).

Most studies of the risks associated with hypothyroidism or thyrotoxicosis have looked at TSH levels. Significantly increased risk of cardiovascular mortality and morbidity is seen in individuals with TSH levels higher than 10 mIU/L.⁸ On the other hand, excess thyroid hormone leading to a TSH level lower than 0.1 mIU/L has been associated with an increased risk of atrial fibrillation in older persons and osteoporosis in postmenopausal women.

The classic symptoms and signs of hypothyroidism correlate with biochemical hypothyroidism and usually improve with the restoration of euthyroidism. Some of these symptoms, however, lack sensitivity and specificity, especially with modest degrees of hypothyroidism.⁷ A randomized controlled trial showed that patients were unable to detect any difference in symptoms when the levothyroxine dose was changed by about 20%.⁹

HARMS ASSOCIATED WITH ORDERING T₄ AND T₃

Other than the financial burden to the patient and society, there is no major morbidity caused by obtaining T₄ or T₃ levels, or both. However, knowing the T₄ or T₃ level does not help with management beyond the information offered by the TSH value. Hypothyroid patients treated with levothyroxine to maintain a normal TSH generally have higher free T₄ levels and lower free T₃ levels than euthyroid patients with similar TSH values.¹⁰ Therefore, reacting to a high T₄ level or a low T₃ level in a treated hypothyroid patient with a normal TSH may lead to inappropriate dose adjustment. On the other hand, increasing the dose of thyroid hormone in a patient with a low TSH whose T₃ level is low-normal may lead to morbidity.

SPECIAL SCENARIO: PITUITARY COMPROMISE

We assume that the patient described above has primary hypothyroidism and that her pituitary-thyroid axis is intact. Primary hypothyroidism is diagnosed by a high TSH along with a low or low-normal T₄. In this typical case, TSH can be used to guide therapy without the need for other tests.

However, when there is pituitary compromise (hypopituitarism, congenital central hypothyroidism), the TSH will not be reliable to monitor the adequacy of thyroid hormone replacement therapy. The aim of levothyroxine management in these patients is to maintain a free T₄ concentration in the upper half of the normal range. If the free T₃ concentration is followed and is found to be elevated, the dose of levothyroxine should be reduced.¹¹

CLINICAL BOTTOM LINE

Since our patient’s dose of levothyroxine has been stable and her TSH is not elevated, measuring serum levels of T₄ and T₃ would not contribute to her management. For such a patient, if the TSH were less than 3 mIU/L, increasing the dose would be unlikely to offer clinical benefit.

On the other hand, if her TSH was higher than 4 mIU/L, then it would be legitimate to tweak the dose upward and reassess her thyroid state clinically and biochemically 6 to 8 weeks later. One would need to be careful not to induce thyrotoxicosis through such an intervention because of the potential morbidity.

The TSH level is typically monitored every 6 to 12 months when the patient is clinically stable. It should be measured sooner in circumstances that include the following:

• Symptoms of hypothyroidism or thyrotoxicosis
• Starting a new medication known to affect thyroid hormone levels
• Significant weight change
• Hospitalization
• Pregnancy.

In this issue of the Journal, Millstein et al provide an elegant, practical, and up-to-date review of insulin pump therapy (also known as continuous subcutaneous insulin infusion), emphasizing its benefits and comparing it with multiple daily insulin injections.¹

See related article

NOT FOR EVERYONE

While insulin pumps make the lives of many patients much easier, we should be careful when generalizing their indications. These devices have been with us for 4 decades, during which they have progressively been made more precise and more intelligent—and smaller. The technology may be attractive to some patients but undesirable to others (Table 1).

Many healthcare providers are unfamiliar with pump technology, and some are intimidated by it because it involves a dynamic device-user interface that is more complex than that of other concealed programmed devices such as pacemakers. Inadequate glycemic management is complex and may result from factors such as fear of hypoglycemia, difficulty with insulin dose adjustment, and poor math skills.²

Unfortunately, some patients are given a pump without proper screening and education, and they tend to call the pump manufacturer’s help line or their provider often for help with technical problems. Selecting the right patient for this technology is more important than the converse.

Indications for an insulin pump vary by country. In some countries, a pump is started as soon as type 1 diabetes is diagnosed. In the United States, the indications are very rigorous and restrictive, especially for patients with type 2 diabetes, in whom a lack of endogenous insulin production must first be proved.

There is no question that a pump should be offered to every patient with type 1 diabetes who demonstrates good motivation to improve his or her glucose control, but only after a rigorous education program. This option is too costly to be tried just to see if the patient likes it.

ADVERSE EVENTS WITH INSULIN PUMPS: MORE DATA NEEDED

A worrisome aspect of continuous subcutaneous insulin infusion at a population level is a lack of information on the root causes of adverse events (diabetic ketoacidosis or severe hypoglycemia) in patients who use it. These events may be serious and sometimes even fatal.

Outside of a controlled environment, it is difficult to ascertain whether an adverse event represents device error or user error, since pumps contain different components (electronic, mechanical, and pharmacologic) that interface with the human user.³ How adverse events are tracked or categorized is unclear, and given the risks associated with this technology, better postmarketing evaluation is needed. Furthermore, we do not know if the precision of insulin delivery decreases over the life of a pump.

While most pump manufacturers have good customer service and make every effort to provide the patient with a replacement pump in case of failure, we do not know if anyone maintains a database of such failures or adverse events, and if those failures can be analyzed to improve safety.³

INTERFACES ARE NOT STANDARD

When one buys a new car, little time is needed to learn how to operate it because most cars use the same basic features.

The situation is different with insulin pumps. To compete with each other, pump manufacturers create different looks, different insulin delivery methods, and different ways of administering a bolus. Switching from one pump to another is difficult without detailed education on the “bells and whistles” of the new pump.

Most patients use just a few features of the pump. They look at it as more of a convenience. They sometimes forget they are wearing it, and even forget to take a bolus before a meal.

PATIENT SATISFACTION DEPENDS ON THE PATIENT

For years, we thought insulin pumps were better at improving hypoglycemia awareness. But in a prospective study, multiple daily injections with frequent self-monitoring of blood glucose provided identical outcomes without worsening hemoglobin A_1c compared with continuous infusion with real-time continuous glucose monitoring, although satisfaction with treatment was better in the latter group.⁴

Patients’ satisfaction with continuous subcutaneous insulin infusion depends on their baseline hemoglobin A_1c level. Patients with relatively low hemoglobin A_1c tend to take an active approach to self-care, describe the pump as a tool for meeting glycemic goals, and say the pump makes them feel more normal. Patients with high hemoglobin A_1c tend to have a more passive approach to their self-care and have more negative experiences with the pump. Women are more concerned than men with the effect of the pump on body image and social acceptance.⁵

DOLLARS AND CENTS

According to 2012 estimates, 29 million Americans had diabetes mellitus, of whom 1.25 million had type 1. The direct medical costs of diabetes are estimated at $176 billion, of which 12% covers overall pharmacy costs.⁶ About 31% of adults with diabetes use insulin.⁷

For a device that costs $6,000, has a life span of only 4 years, and requires supplies that cost $300 per month, rigorous interpretation of superiority data would be needed to confirm that this technology would have a positive impact on public health if every insulin-using patient with diabetes were to say yes to it. It is true that switching from multiple daily injections to a pump leads to a significant reduction in insulin expenditures in patients with type 2 diabetes, according to a retrospective analysis of claims data.⁸

However, not all studies comparing pumps and multiple daily injections in type 2 diabetes have shown an advantage of one over the other in terms of a reduction in fasting glucose, hemoglobin A_1c, or incidence of hypoglycemia.⁹ A meta-analysis¹⁰ found that the two therapies had similar effects on glycemic control and hypoglycemia. Continuous infusion had a more favorable effect in adults with type 1 diabetes.¹⁰

Neither continuous infusion nor multiple daily injections can mimic physiologic endogenous insulin secretion. Endogenous insulin is secreted into the portal system, and its main site of action is the liver. As a result, there is more hepatic glucose uptake and thus a lower peripheral plasma insulin concentration with endogenous secretion than with systemic administration. Endogenous insulin secretion also suppresses hepatic glucose production and reduces the risk of hypoglycemia.¹¹

PROGRESS, BUT NOT PERFECTION

Diabetes mellitus constitutes a big burden on patients and on society. The discovery of insulin was a giant leap forward; the insulin pump was another great advance. We are getting closer to an integrated bionic pancreas. We are far from achieving a perfect system, but we are much better off than we were 50 or 80 years ago. And although insulin pump technology is sophisticated and precise, it still interfaces with a human user, and the human user still must press its buttons.

In this issue of the Journal, Millstein et al provide an elegant, practical, and up-to-date review of insulin pump therapy (also known as continuous subcutaneous insulin infusion), emphasizing its benefits and comparing it with multiple daily insulin injections.¹

See related article

NOT FOR EVERYONE

While insulin pumps make the lives of many patients much easier, we should be careful when generalizing their indications. These devices have been with us for 4 decades, during which they have progressively been made more precise and more intelligent—and smaller. The technology may be attractive to some patients but undesirable to others (Table 1).

Many healthcare providers are unfamiliar with pump technology, and some are intimidated by it because it involves a dynamic device-user interface that is more complex than that of other concealed programmed devices such as pacemakers. Inadequate glycemic management is complex and may result from factors such as fear of hypoglycemia, difficulty with insulin dose adjustment, and poor math skills.²

Unfortunately, some patients are given a pump without proper screening and education, and they tend to call the pump manufacturer’s help line or their provider often for help with technical problems. Selecting the right patient for this technology is more important than the converse.

Indications for an insulin pump vary by country. In some countries, a pump is started as soon as type 1 diabetes is diagnosed. In the United States, the indications are very rigorous and restrictive, especially for patients with type 2 diabetes, in whom a lack of endogenous insulin production must first be proved.

There is no question that a pump should be offered to every patient with type 1 diabetes who demonstrates good motivation to improve his or her glucose control, but only after a rigorous education program. This option is too costly to be tried just to see if the patient likes it.

ADVERSE EVENTS WITH INSULIN PUMPS: MORE DATA NEEDED

A worrisome aspect of continuous subcutaneous insulin infusion at a population level is a lack of information on the root causes of adverse events (diabetic ketoacidosis or severe hypoglycemia) in patients who use it. These events may be serious and sometimes even fatal.

Outside of a controlled environment, it is difficult to ascertain whether an adverse event represents device error or user error, since pumps contain different components (electronic, mechanical, and pharmacologic) that interface with the human user.³ How adverse events are tracked or categorized is unclear, and given the risks associated with this technology, better postmarketing evaluation is needed. Furthermore, we do not know if the precision of insulin delivery decreases over the life of a pump.

While most pump manufacturers have good customer service and make every effort to provide the patient with a replacement pump in case of failure, we do not know if anyone maintains a database of such failures or adverse events, and if those failures can be analyzed to improve safety.³

INTERFACES ARE NOT STANDARD

When one buys a new car, little time is needed to learn how to operate it because most cars use the same basic features.

The situation is different with insulin pumps. To compete with each other, pump manufacturers create different looks, different insulin delivery methods, and different ways of administering a bolus. Switching from one pump to another is difficult without detailed education on the “bells and whistles” of the new pump.

Most patients use just a few features of the pump. They look at it as more of a convenience. They sometimes forget they are wearing it, and even forget to take a bolus before a meal.

PATIENT SATISFACTION DEPENDS ON THE PATIENT

For years, we thought insulin pumps were better at improving hypoglycemia awareness. But in a prospective study, multiple daily injections with frequent self-monitoring of blood glucose provided identical outcomes without worsening hemoglobin A_1c compared with continuous infusion with real-time continuous glucose monitoring, although satisfaction with treatment was better in the latter group.⁴

Patients’ satisfaction with continuous subcutaneous insulin infusion depends on their baseline hemoglobin A_1c level. Patients with relatively low hemoglobin A_1c tend to take an active approach to self-care, describe the pump as a tool for meeting glycemic goals, and say the pump makes them feel more normal. Patients with high hemoglobin A_1c tend to have a more passive approach to their self-care and have more negative experiences with the pump. Women are more concerned than men with the effect of the pump on body image and social acceptance.⁵

DOLLARS AND CENTS

According to 2012 estimates, 29 million Americans had diabetes mellitus, of whom 1.25 million had type 1. The direct medical costs of diabetes are estimated at $176 billion, of which 12% covers overall pharmacy costs.⁶ About 31% of adults with diabetes use insulin.⁷

For a device that costs $6,000, has a life span of only 4 years, and requires supplies that cost $300 per month, rigorous interpretation of superiority data would be needed to confirm that this technology would have a positive impact on public health if every insulin-using patient with diabetes were to say yes to it. It is true that switching from multiple daily injections to a pump leads to a significant reduction in insulin expenditures in patients with type 2 diabetes, according to a retrospective analysis of claims data.⁸

However, not all studies comparing pumps and multiple daily injections in type 2 diabetes have shown an advantage of one over the other in terms of a reduction in fasting glucose, hemoglobin A_1c, or incidence of hypoglycemia.⁹ A meta-analysis¹⁰ found that the two therapies had similar effects on glycemic control and hypoglycemia. Continuous infusion had a more favorable effect in adults with type 1 diabetes.¹⁰

Neither continuous infusion nor multiple daily injections can mimic physiologic endogenous insulin secretion. Endogenous insulin is secreted into the portal system, and its main site of action is the liver. As a result, there is more hepatic glucose uptake and thus a lower peripheral plasma insulin concentration with endogenous secretion than with systemic administration. Endogenous insulin secretion also suppresses hepatic glucose production and reduces the risk of hypoglycemia.¹¹

PROGRESS, BUT NOT PERFECTION

Diabetes mellitus constitutes a big burden on patients and on society. The discovery of insulin was a giant leap forward; the insulin pump was another great advance. We are getting closer to an integrated bionic pancreas. We are far from achieving a perfect system, but we are much better off than we were 50 or 80 years ago. And although insulin pump technology is sophisticated and precise, it still interfaces with a human user, and the human user still must press its buttons.

In this issue of the Journal, Millstein et al provide an elegant, practical, and up-to-date review of insulin pump therapy (also known as continuous subcutaneous insulin infusion), emphasizing its benefits and comparing it with multiple daily insulin injections.¹

See related article

NOT FOR EVERYONE

While insulin pumps make the lives of many patients much easier, we should be careful when generalizing their indications. These devices have been with us for 4 decades, during which they have progressively been made more precise and more intelligent—and smaller. The technology may be attractive to some patients but undesirable to others (Table 1).

Many healthcare providers are unfamiliar with pump technology, and some are intimidated by it because it involves a dynamic device-user interface that is more complex than that of other concealed programmed devices such as pacemakers. Inadequate glycemic management is complex and may result from factors such as fear of hypoglycemia, difficulty with insulin dose adjustment, and poor math skills.²

Unfortunately, some patients are given a pump without proper screening and education, and they tend to call the pump manufacturer’s help line or their provider often for help with technical problems. Selecting the right patient for this technology is more important than the converse.

Indications for an insulin pump vary by country. In some countries, a pump is started as soon as type 1 diabetes is diagnosed. In the United States, the indications are very rigorous and restrictive, especially for patients with type 2 diabetes, in whom a lack of endogenous insulin production must first be proved.

There is no question that a pump should be offered to every patient with type 1 diabetes who demonstrates good motivation to improve his or her glucose control, but only after a rigorous education program. This option is too costly to be tried just to see if the patient likes it.

ADVERSE EVENTS WITH INSULIN PUMPS: MORE DATA NEEDED

A worrisome aspect of continuous subcutaneous insulin infusion at a population level is a lack of information on the root causes of adverse events (diabetic ketoacidosis or severe hypoglycemia) in patients who use it. These events may be serious and sometimes even fatal.

Outside of a controlled environment, it is difficult to ascertain whether an adverse event represents device error or user error, since pumps contain different components (electronic, mechanical, and pharmacologic) that interface with the human user.³ How adverse events are tracked or categorized is unclear, and given the risks associated with this technology, better postmarketing evaluation is needed. Furthermore, we do not know if the precision of insulin delivery decreases over the life of a pump.

While most pump manufacturers have good customer service and make every effort to provide the patient with a replacement pump in case of failure, we do not know if anyone maintains a database of such failures or adverse events, and if those failures can be analyzed to improve safety.³

INTERFACES ARE NOT STANDARD

When one buys a new car, little time is needed to learn how to operate it because most cars use the same basic features.

The situation is different with insulin pumps. To compete with each other, pump manufacturers create different looks, different insulin delivery methods, and different ways of administering a bolus. Switching from one pump to another is difficult without detailed education on the “bells and whistles” of the new pump.

Most patients use just a few features of the pump. They look at it as more of a convenience. They sometimes forget they are wearing it, and even forget to take a bolus before a meal.

PATIENT SATISFACTION DEPENDS ON THE PATIENT

For years, we thought insulin pumps were better at improving hypoglycemia awareness. But in a prospective study, multiple daily injections with frequent self-monitoring of blood glucose provided identical outcomes without worsening hemoglobin A_1c compared with continuous infusion with real-time continuous glucose monitoring, although satisfaction with treatment was better in the latter group.⁴

Patients’ satisfaction with continuous subcutaneous insulin infusion depends on their baseline hemoglobin A_1c level. Patients with relatively low hemoglobin A_1c tend to take an active approach to self-care, describe the pump as a tool for meeting glycemic goals, and say the pump makes them feel more normal. Patients with high hemoglobin A_1c tend to have a more passive approach to their self-care and have more negative experiences with the pump. Women are more concerned than men with the effect of the pump on body image and social acceptance.⁵

DOLLARS AND CENTS

According to 2012 estimates, 29 million Americans had diabetes mellitus, of whom 1.25 million had type 1. The direct medical costs of diabetes are estimated at $176 billion, of which 12% covers overall pharmacy costs.⁶ About 31% of adults with diabetes use insulin.⁷

For a device that costs $6,000, has a life span of only 4 years, and requires supplies that cost $300 per month, rigorous interpretation of superiority data would be needed to confirm that this technology would have a positive impact on public health if every insulin-using patient with diabetes were to say yes to it. It is true that switching from multiple daily injections to a pump leads to a significant reduction in insulin expenditures in patients with type 2 diabetes, according to a retrospective analysis of claims data.⁸

However, not all studies comparing pumps and multiple daily injections in type 2 diabetes have shown an advantage of one over the other in terms of a reduction in fasting glucose, hemoglobin A_1c, or incidence of hypoglycemia.⁹ A meta-analysis¹⁰ found that the two therapies had similar effects on glycemic control and hypoglycemia. Continuous infusion had a more favorable effect in adults with type 1 diabetes.¹⁰

Neither continuous infusion nor multiple daily injections can mimic physiologic endogenous insulin secretion. Endogenous insulin is secreted into the portal system, and its main site of action is the liver. As a result, there is more hepatic glucose uptake and thus a lower peripheral plasma insulin concentration with endogenous secretion than with systemic administration. Endogenous insulin secretion also suppresses hepatic glucose production and reduces the risk of hypoglycemia.¹¹

PROGRESS, BUT NOT PERFECTION

Diabetes mellitus constitutes a big burden on patients and on society. The discovery of insulin was a giant leap forward; the insulin pump was another great advance. We are getting closer to an integrated bionic pancreas. We are far from achieving a perfect system, but we are much better off than we were 50 or 80 years ago. And although insulin pump technology is sophisticated and precise, it still interfaces with a human user, and the human user still must press its buttons.

The question is complicated, as there are different types of thyroid cancer, and a causal relationship is hard to prove.

Glucagon-like peptide 1 (GLP-1) receptor agonists can be safely used in all patients with thyroid cancers that are derived from the thyroid follicular epithelium (papillary and follicular thyroid cancer). However, they are currently contraindicated in patients with medullary thyroid cancer and in patients with multiple endocrine neoplasia 2 (MEN-2), which is not a form of thyroid cancer but is relevant to our discussion. We probably should be cautious about using them in patients with familial thyroid cancer and those with a genetic predisposition for papillary or follicular thyroid cancer.

GLP-1 DRUGS ARE WIDELY USED

The glucagon-like peptide 1 (GLP-1) receptor agonists are widely used to treat type 2 diabetes mellitus. The currently available drugs of this class—exenatide (Byetta), liraglutide (Victoza), albiglutide (Tanzeum), dulaglutide (Trulicity), and extended-release exenatide (Bydureon)—are popular because they lower glucose levels, pose a low risk of hypoglycemia, can induce weight loss,¹ and, in the case of extended-release exenatide and albiglutide, are given once weekly. They are currently recommended as add-on therapy to metformin. These drugs mimic the action of GLP-1, an endogenous hormone released by the intestine in response to food. They bind to receptors on beta cells, stimulating insulin production.¹

FOUR TYPES OF THYROID CANCER

There are four types of thyroid cancer: medullary (a contraindication to GLP-1 agonists), papillary, follicular, and anaplastic.

Medullary thyroid cancer is extremely rare in humans, with 976 cases diagnosed from 1992 to 2006 in the United States, compared with 36,583 cases of papillary and 4,560 cases of follicular cancer. Anaplastic cancer is also rare (556 cases).² The highest incidence rates of medullary thyroid cancer are in people of Hispanic descent (0.21 per 100,000 woman-years and 0.18 per 100,000 man-years).²

EXPERIMENTAL EVIDENCE

Pancreatic beta cells are not the only cells in the body that can express GLP-1 receptors. Notably, the parafollicular cells (also called C cells) of the thyroid, which secrete calcitonin and which are the cells involved in medullary thyroid cancer, also sometimes express these receptors if cancer develops.

GLP-1 receptor agonists are contraindicated in patients with medullary thyroid cancer or multiple endocrine neoplasia 2

In experiments in mice and rats, the incidence of thyroid C-cell tumors was higher in animals given GLP-1 analogues. Liraglutide, exenatide, taspoglutide, and lixisenatide potently activated GLP-1 receptors in thyroid C cells, increasing calcitonin gene expression and stimulating calcitonin release in a dose-dependent manner.³ Moreover, sustained activation of these receptors caused C-cell hyperplasia and resulted in medullary thyroid cancer. However medullary thyroid cancer also occurred in rodents receiving placebo.

C cells in monkeys and humans express fewer GLP-1 receptors than those in rodents; in fact, healthy human C cells do not express them at all.^3,4 In rats with C-cell hyperplasia or medullary thyroid cancer, GLP-1 receptors are present in 100% of cases (and in increased density), compared with 27% of human medullary thyroid cancers.⁴

In addition to medullary thyroid cancer, various other human tumors have been shown to express GLP-1 receptors.⁵ Based on limited data, KÖrner et al⁵ found that these receptors are also present in various other human tumors, eg:

• Pheochromocytoma (60%)
• Paraganglioma (28%)
• Meningioma (35%)
• Astrocytoma (25%)
• Glioblastoma (9%)
• Ependymoma (16%)
• Medulloblastoma (25%)
• Nephroblastoma (22%)
• Neuroblastoma (18%)
• Ovarian adenocarcinoma (16%)
• Prostate carcinoma (5%).

Madsen et al⁶ reported that liraglutide binding to the GLP-1 receptor on murine thyroid C cells led to C-cell hyperplasia. However, prolonged administration of liraglutide at very high doses did not produce C-cell proliferation in monkeys.³

Gier et al⁷ looked at GLP-1 receptor expression in normal human C cells, hyperplastic C cells, and medullary thyroid cancer cells, as well as in papillary thyroid cancer cells, which do not arise from C cells. They demonstrated concurrent calcitonin and GLP-1 receptor immunostaining, not only in those with C-cell hyperplasia (9 of 9 cases) and medullary thyroid cancer (11 of 12 cases), but also in 3 (18%) of 17 patients with papillary thyroid cancer and 5 (33%) of 15 with normal thyroid follicular cells. However, the choice of polyclonal anti­bodies and radioligands used and concerns about methodology have led investigators to interpret these results cautiously.^8–10

STUDIES OF GLP-1 AGONISTS IN HUMANS

Several prospective clinical studies showed no increase in calcitonin levels during therapy with GLP-1 receptor agonists in patients with type 2 diabetes.^3,11 Long-term use of liraglutide in high doses (up to 3 mg per day) did not lead to elevations in serum calcitonin levels.¹¹

In a retrospective Adverse Event Reporting System database review, the incidence rate of thyroid cancer in patients treated with exenatide was higher—with an odds ratio of 4.7 (30 events)—than with a panel of control drugs (3 events).¹² However, this study did not differentiate between types of thyroid cancer, and the inherent limitations of retrospective databases complicate its interpretation. Such a high odds ratio would imply a significant increase in the incidence of medullary thyroid cancer, but this does not seem to be true.

These studies are hypothesis-generating and do not prove that GLP-1 receptor agonists cause medullary thyroid cancer

Alves et al¹³ performed a meta-analysis of randomized controlled trials and long-term observational studies. None of the studies evaluating exenatide reported cases of thyroid cancer, whereas five of the studies evaluating liraglutide did. In total, nine patients treated with liraglutide were diagnosed with thyroid cancer, compared with one patient on glimepiride. The odds ratio for thyroid cancer occurrence associated with liraglutide treatment was 1.54, but that was not statistically significant (95% confidence interval 0.40–6.02, P = .53, I² = 0%).

These studies are hypothesis-generating and do not prove that GLP-1 receptor agonists cause medullary thyroid cancer. Given the extremely low incidence of medullary thyroid cancer, to prove or disprove a causal relationship would require an enormous number of patients, who would need to be followed for several years.

OFFICIAL RECOMMENDATIONS

Considerable differences in the biology of the rodent vs human thyroid GLP-1 receptor systems have led regulatory authorities to conclude that the risk for development of medullary thyroid cancer with GLP-1 therapy in humans is difficult to quantify, but low.¹⁴ Consequently, the US Food and Drug Administration recommends neither monitoring of calcitonin levels nor ultrasound imaging as a screening tool in patients taking GLP-1 agonists.¹⁴

BENEFITS OUTWEIGH RISKS

At present, the benefits of using GLP-1 receptor agonists to treat type 2 diabetes mellitus outweigh the risks, and there seems to be little reason to withhold this effective therapy except in patients who have a personal or family history of medullary thyroid cancer or MEN-2. Until the effects of GLP-1 agonists are systematically studied in follicular-cell-derived thyroid cancer, we also recommend caution when considering their use in patients with familial thyroid cancer and those with a genetic predisposition for papillary and follicular thyroid cancer—eg, patients with familial adenomatous polyposis, phosphate and tensin homolog hamartoma tumor syndrome, Carney complex type 1, Werner syndrome, or familial papillary thyroid cancer.

Methodologically superior studies and careful long-term monitoring of patients treated with GLP-1 agonists are required to clarify the risk vs benefit of these therapies.

The question is complicated, as there are different types of thyroid cancer, and a causal relationship is hard to prove.

Glucagon-like peptide 1 (GLP-1) receptor agonists can be safely used in all patients with thyroid cancers that are derived from the thyroid follicular epithelium (papillary and follicular thyroid cancer). However, they are currently contraindicated in patients with medullary thyroid cancer and in patients with multiple endocrine neoplasia 2 (MEN-2), which is not a form of thyroid cancer but is relevant to our discussion. We probably should be cautious about using them in patients with familial thyroid cancer and those with a genetic predisposition for papillary or follicular thyroid cancer.

GLP-1 DRUGS ARE WIDELY USED

The glucagon-like peptide 1 (GLP-1) receptor agonists are widely used to treat type 2 diabetes mellitus. The currently available drugs of this class—exenatide (Byetta), liraglutide (Victoza), albiglutide (Tanzeum), dulaglutide (Trulicity), and extended-release exenatide (Bydureon)—are popular because they lower glucose levels, pose a low risk of hypoglycemia, can induce weight loss,¹ and, in the case of extended-release exenatide and albiglutide, are given once weekly. They are currently recommended as add-on therapy to metformin. These drugs mimic the action of GLP-1, an endogenous hormone released by the intestine in response to food. They bind to receptors on beta cells, stimulating insulin production.¹

FOUR TYPES OF THYROID CANCER

There are four types of thyroid cancer: medullary (a contraindication to GLP-1 agonists), papillary, follicular, and anaplastic.

Medullary thyroid cancer is extremely rare in humans, with 976 cases diagnosed from 1992 to 2006 in the United States, compared with 36,583 cases of papillary and 4,560 cases of follicular cancer. Anaplastic cancer is also rare (556 cases).² The highest incidence rates of medullary thyroid cancer are in people of Hispanic descent (0.21 per 100,000 woman-years and 0.18 per 100,000 man-years).²

EXPERIMENTAL EVIDENCE

Pancreatic beta cells are not the only cells in the body that can express GLP-1 receptors. Notably, the parafollicular cells (also called C cells) of the thyroid, which secrete calcitonin and which are the cells involved in medullary thyroid cancer, also sometimes express these receptors if cancer develops.

GLP-1 receptor agonists are contraindicated in patients with medullary thyroid cancer or multiple endocrine neoplasia 2

In experiments in mice and rats, the incidence of thyroid C-cell tumors was higher in animals given GLP-1 analogues. Liraglutide, exenatide, taspoglutide, and lixisenatide potently activated GLP-1 receptors in thyroid C cells, increasing calcitonin gene expression and stimulating calcitonin release in a dose-dependent manner.³ Moreover, sustained activation of these receptors caused C-cell hyperplasia and resulted in medullary thyroid cancer. However medullary thyroid cancer also occurred in rodents receiving placebo.

C cells in monkeys and humans express fewer GLP-1 receptors than those in rodents; in fact, healthy human C cells do not express them at all.^3,4 In rats with C-cell hyperplasia or medullary thyroid cancer, GLP-1 receptors are present in 100% of cases (and in increased density), compared with 27% of human medullary thyroid cancers.⁴

In addition to medullary thyroid cancer, various other human tumors have been shown to express GLP-1 receptors.⁵ Based on limited data, KÖrner et al⁵ found that these receptors are also present in various other human tumors, eg:

• Pheochromocytoma (60%)
• Paraganglioma (28%)
• Meningioma (35%)
• Astrocytoma (25%)
• Glioblastoma (9%)
• Ependymoma (16%)
• Medulloblastoma (25%)
• Nephroblastoma (22%)
• Neuroblastoma (18%)
• Ovarian adenocarcinoma (16%)
• Prostate carcinoma (5%).

Madsen et al⁶ reported that liraglutide binding to the GLP-1 receptor on murine thyroid C cells led to C-cell hyperplasia. However, prolonged administration of liraglutide at very high doses did not produce C-cell proliferation in monkeys.³

Gier et al⁷ looked at GLP-1 receptor expression in normal human C cells, hyperplastic C cells, and medullary thyroid cancer cells, as well as in papillary thyroid cancer cells, which do not arise from C cells. They demonstrated concurrent calcitonin and GLP-1 receptor immunostaining, not only in those with C-cell hyperplasia (9 of 9 cases) and medullary thyroid cancer (11 of 12 cases), but also in 3 (18%) of 17 patients with papillary thyroid cancer and 5 (33%) of 15 with normal thyroid follicular cells. However, the choice of polyclonal anti­bodies and radioligands used and concerns about methodology have led investigators to interpret these results cautiously.^8–10

STUDIES OF GLP-1 AGONISTS IN HUMANS

Several prospective clinical studies showed no increase in calcitonin levels during therapy with GLP-1 receptor agonists in patients with type 2 diabetes.^3,11 Long-term use of liraglutide in high doses (up to 3 mg per day) did not lead to elevations in serum calcitonin levels.¹¹

In a retrospective Adverse Event Reporting System database review, the incidence rate of thyroid cancer in patients treated with exenatide was higher—with an odds ratio of 4.7 (30 events)—than with a panel of control drugs (3 events).¹² However, this study did not differentiate between types of thyroid cancer, and the inherent limitations of retrospective databases complicate its interpretation. Such a high odds ratio would imply a significant increase in the incidence of medullary thyroid cancer, but this does not seem to be true.

These studies are hypothesis-generating and do not prove that GLP-1 receptor agonists cause medullary thyroid cancer

Alves et al¹³ performed a meta-analysis of randomized controlled trials and long-term observational studies. None of the studies evaluating exenatide reported cases of thyroid cancer, whereas five of the studies evaluating liraglutide did. In total, nine patients treated with liraglutide were diagnosed with thyroid cancer, compared with one patient on glimepiride. The odds ratio for thyroid cancer occurrence associated with liraglutide treatment was 1.54, but that was not statistically significant (95% confidence interval 0.40–6.02, P = .53, I² = 0%).

These studies are hypothesis-generating and do not prove that GLP-1 receptor agonists cause medullary thyroid cancer. Given the extremely low incidence of medullary thyroid cancer, to prove or disprove a causal relationship would require an enormous number of patients, who would need to be followed for several years.

OFFICIAL RECOMMENDATIONS

Considerable differences in the biology of the rodent vs human thyroid GLP-1 receptor systems have led regulatory authorities to conclude that the risk for development of medullary thyroid cancer with GLP-1 therapy in humans is difficult to quantify, but low.¹⁴ Consequently, the US Food and Drug Administration recommends neither monitoring of calcitonin levels nor ultrasound imaging as a screening tool in patients taking GLP-1 agonists.¹⁴

BENEFITS OUTWEIGH RISKS

At present, the benefits of using GLP-1 receptor agonists to treat type 2 diabetes mellitus outweigh the risks, and there seems to be little reason to withhold this effective therapy except in patients who have a personal or family history of medullary thyroid cancer or MEN-2. Until the effects of GLP-1 agonists are systematically studied in follicular-cell-derived thyroid cancer, we also recommend caution when considering their use in patients with familial thyroid cancer and those with a genetic predisposition for papillary and follicular thyroid cancer—eg, patients with familial adenomatous polyposis, phosphate and tensin homolog hamartoma tumor syndrome, Carney complex type 1, Werner syndrome, or familial papillary thyroid cancer.

Methodologically superior studies and careful long-term monitoring of patients treated with GLP-1 agonists are required to clarify the risk vs benefit of these therapies.

The question is complicated, as there are different types of thyroid cancer, and a causal relationship is hard to prove.

Glucagon-like peptide 1 (GLP-1) receptor agonists can be safely used in all patients with thyroid cancers that are derived from the thyroid follicular epithelium (papillary and follicular thyroid cancer). However, they are currently contraindicated in patients with medullary thyroid cancer and in patients with multiple endocrine neoplasia 2 (MEN-2), which is not a form of thyroid cancer but is relevant to our discussion. We probably should be cautious about using them in patients with familial thyroid cancer and those with a genetic predisposition for papillary or follicular thyroid cancer.

GLP-1 DRUGS ARE WIDELY USED

The glucagon-like peptide 1 (GLP-1) receptor agonists are widely used to treat type 2 diabetes mellitus. The currently available drugs of this class—exenatide (Byetta), liraglutide (Victoza), albiglutide (Tanzeum), dulaglutide (Trulicity), and extended-release exenatide (Bydureon)—are popular because they lower glucose levels, pose a low risk of hypoglycemia, can induce weight loss,¹ and, in the case of extended-release exenatide and albiglutide, are given once weekly. They are currently recommended as add-on therapy to metformin. These drugs mimic the action of GLP-1, an endogenous hormone released by the intestine in response to food. They bind to receptors on beta cells, stimulating insulin production.¹

FOUR TYPES OF THYROID CANCER

There are four types of thyroid cancer: medullary (a contraindication to GLP-1 agonists), papillary, follicular, and anaplastic.

Medullary thyroid cancer is extremely rare in humans, with 976 cases diagnosed from 1992 to 2006 in the United States, compared with 36,583 cases of papillary and 4,560 cases of follicular cancer. Anaplastic cancer is also rare (556 cases).² The highest incidence rates of medullary thyroid cancer are in people of Hispanic descent (0.21 per 100,000 woman-years and 0.18 per 100,000 man-years).²

EXPERIMENTAL EVIDENCE

Pancreatic beta cells are not the only cells in the body that can express GLP-1 receptors. Notably, the parafollicular cells (also called C cells) of the thyroid, which secrete calcitonin and which are the cells involved in medullary thyroid cancer, also sometimes express these receptors if cancer develops.

GLP-1 receptor agonists are contraindicated in patients with medullary thyroid cancer or multiple endocrine neoplasia 2

In experiments in mice and rats, the incidence of thyroid C-cell tumors was higher in animals given GLP-1 analogues. Liraglutide, exenatide, taspoglutide, and lixisenatide potently activated GLP-1 receptors in thyroid C cells, increasing calcitonin gene expression and stimulating calcitonin release in a dose-dependent manner.³ Moreover, sustained activation of these receptors caused C-cell hyperplasia and resulted in medullary thyroid cancer. However medullary thyroid cancer also occurred in rodents receiving placebo.

C cells in monkeys and humans express fewer GLP-1 receptors than those in rodents; in fact, healthy human C cells do not express them at all.^3,4 In rats with C-cell hyperplasia or medullary thyroid cancer, GLP-1 receptors are present in 100% of cases (and in increased density), compared with 27% of human medullary thyroid cancers.⁴

In addition to medullary thyroid cancer, various other human tumors have been shown to express GLP-1 receptors.⁵ Based on limited data, KÖrner et al⁵ found that these receptors are also present in various other human tumors, eg:

• Pheochromocytoma (60%)
• Paraganglioma (28%)
• Meningioma (35%)
• Astrocytoma (25%)
• Glioblastoma (9%)
• Ependymoma (16%)
• Medulloblastoma (25%)
• Nephroblastoma (22%)
• Neuroblastoma (18%)
• Ovarian adenocarcinoma (16%)
• Prostate carcinoma (5%).

Madsen et al⁶ reported that liraglutide binding to the GLP-1 receptor on murine thyroid C cells led to C-cell hyperplasia. However, prolonged administration of liraglutide at very high doses did not produce C-cell proliferation in monkeys.³

Gier et al⁷ looked at GLP-1 receptor expression in normal human C cells, hyperplastic C cells, and medullary thyroid cancer cells, as well as in papillary thyroid cancer cells, which do not arise from C cells. They demonstrated concurrent calcitonin and GLP-1 receptor immunostaining, not only in those with C-cell hyperplasia (9 of 9 cases) and medullary thyroid cancer (11 of 12 cases), but also in 3 (18%) of 17 patients with papillary thyroid cancer and 5 (33%) of 15 with normal thyroid follicular cells. However, the choice of polyclonal anti­bodies and radioligands used and concerns about methodology have led investigators to interpret these results cautiously.^8–10

STUDIES OF GLP-1 AGONISTS IN HUMANS

Several prospective clinical studies showed no increase in calcitonin levels during therapy with GLP-1 receptor agonists in patients with type 2 diabetes.^3,11 Long-term use of liraglutide in high doses (up to 3 mg per day) did not lead to elevations in serum calcitonin levels.¹¹

In a retrospective Adverse Event Reporting System database review, the incidence rate of thyroid cancer in patients treated with exenatide was higher—with an odds ratio of 4.7 (30 events)—than with a panel of control drugs (3 events).¹² However, this study did not differentiate between types of thyroid cancer, and the inherent limitations of retrospective databases complicate its interpretation. Such a high odds ratio would imply a significant increase in the incidence of medullary thyroid cancer, but this does not seem to be true.

These studies are hypothesis-generating and do not prove that GLP-1 receptor agonists cause medullary thyroid cancer

Alves et al¹³ performed a meta-analysis of randomized controlled trials and long-term observational studies. None of the studies evaluating exenatide reported cases of thyroid cancer, whereas five of the studies evaluating liraglutide did. In total, nine patients treated with liraglutide were diagnosed with thyroid cancer, compared with one patient on glimepiride. The odds ratio for thyroid cancer occurrence associated with liraglutide treatment was 1.54, but that was not statistically significant (95% confidence interval 0.40–6.02, P = .53, I² = 0%).

These studies are hypothesis-generating and do not prove that GLP-1 receptor agonists cause medullary thyroid cancer. Given the extremely low incidence of medullary thyroid cancer, to prove or disprove a causal relationship would require an enormous number of patients, who would need to be followed for several years.

OFFICIAL RECOMMENDATIONS

Considerable differences in the biology of the rodent vs human thyroid GLP-1 receptor systems have led regulatory authorities to conclude that the risk for development of medullary thyroid cancer with GLP-1 therapy in humans is difficult to quantify, but low.¹⁴ Consequently, the US Food and Drug Administration recommends neither monitoring of calcitonin levels nor ultrasound imaging as a screening tool in patients taking GLP-1 agonists.¹⁴

BENEFITS OUTWEIGH RISKS

At present, the benefits of using GLP-1 receptor agonists to treat type 2 diabetes mellitus outweigh the risks, and there seems to be little reason to withhold this effective therapy except in patients who have a personal or family history of medullary thyroid cancer or MEN-2. Until the effects of GLP-1 agonists are systematically studied in follicular-cell-derived thyroid cancer, we also recommend caution when considering their use in patients with familial thyroid cancer and those with a genetic predisposition for papillary and follicular thyroid cancer—eg, patients with familial adenomatous polyposis, phosphate and tensin homolog hamartoma tumor syndrome, Carney complex type 1, Werner syndrome, or familial papillary thyroid cancer.

Methodologically superior studies and careful long-term monitoring of patients treated with GLP-1 agonists are required to clarify the risk vs benefit of these therapies.

In Reply: Dr. Fountas et al highlight further data on insulin therapy and cancer risk, specifically in regard to insulin detemir and insulin degludec. Detemir first gained US Food and Drug Administration (FDA) approval in 2005 as a basal insulin, dosed once or twice daily.¹ Compared with regular human insulin, detemir has demonstrated proliferative and antiapoptotic activities in vitro in various cancer cell lines—eg, HCT-116 (colorectal cancer), PC-3 (prostate cancer), and MCF-7 (breast adenocarcinoma).² But clinically, detemir has not demonstrated increased cancer risk compared with other basal insulins in randomized controlled trials or cohort studies.^3–5

Degludec (U-200 insulin) is equal to twice the concentration of the usual U-100 insulin therapies presently available. In February 2013, the drug application for insulin degludec failed to obtain FDA approval, and the FDA requested additional data on cardiovascular safety. Thus, degludec is not currently available in the United States.⁶

Besides ameliorating nocturnal hypoglycemia,⁷ U-200 insulin may mitigate potential mitogenic effects.⁸ However, there are still very few data on degludec compared with the amount of data on insulin glargine. Insulin analogues with a decreased dissociation rate from the insulin receptor are associated with higher mitogenic potency than metabolic potency compared with human insulin.^9,10 Degludec, like detemir, has an elevated dissociation rate from the insulin receptor, a low affinity for IGF-1 receptors, and a low mitogenic activity in vitro.⁸

At this juncture, neither detemir nor degludec has been associated with higher cancer risk, but these therapies are relatively new. And as Dr. Fountas et al indicated, their safety, particularly in regard to cancer risk in diabetes patients, should continue to be assessed.

In Reply: Dr. Fountas et al highlight further data on insulin therapy and cancer risk, specifically in regard to insulin detemir and insulin degludec. Detemir first gained US Food and Drug Administration (FDA) approval in 2005 as a basal insulin, dosed once or twice daily.¹ Compared with regular human insulin, detemir has demonstrated proliferative and antiapoptotic activities in vitro in various cancer cell lines—eg, HCT-116 (colorectal cancer), PC-3 (prostate cancer), and MCF-7 (breast adenocarcinoma).² But clinically, detemir has not demonstrated increased cancer risk compared with other basal insulins in randomized controlled trials or cohort studies.^3–5

Degludec (U-200 insulin) is equal to twice the concentration of the usual U-100 insulin therapies presently available. In February 2013, the drug application for insulin degludec failed to obtain FDA approval, and the FDA requested additional data on cardiovascular safety. Thus, degludec is not currently available in the United States.⁶

Besides ameliorating nocturnal hypoglycemia,⁷ U-200 insulin may mitigate potential mitogenic effects.⁸ However, there are still very few data on degludec compared with the amount of data on insulin glargine. Insulin analogues with a decreased dissociation rate from the insulin receptor are associated with higher mitogenic potency than metabolic potency compared with human insulin.^9,10 Degludec, like detemir, has an elevated dissociation rate from the insulin receptor, a low affinity for IGF-1 receptors, and a low mitogenic activity in vitro.⁸

At this juncture, neither detemir nor degludec has been associated with higher cancer risk, but these therapies are relatively new. And as Dr. Fountas et al indicated, their safety, particularly in regard to cancer risk in diabetes patients, should continue to be assessed.

In Reply: Dr. Fountas et al highlight further data on insulin therapy and cancer risk, specifically in regard to insulin detemir and insulin degludec. Detemir first gained US Food and Drug Administration (FDA) approval in 2005 as a basal insulin, dosed once or twice daily.¹ Compared with regular human insulin, detemir has demonstrated proliferative and antiapoptotic activities in vitro in various cancer cell lines—eg, HCT-116 (colorectal cancer), PC-3 (prostate cancer), and MCF-7 (breast adenocarcinoma).² But clinically, detemir has not demonstrated increased cancer risk compared with other basal insulins in randomized controlled trials or cohort studies.^3–5

Degludec (U-200 insulin) is equal to twice the concentration of the usual U-100 insulin therapies presently available. In February 2013, the drug application for insulin degludec failed to obtain FDA approval, and the FDA requested additional data on cardiovascular safety. Thus, degludec is not currently available in the United States.⁶

Besides ameliorating nocturnal hypoglycemia,⁷ U-200 insulin may mitigate potential mitogenic effects.⁸ However, there are still very few data on degludec compared with the amount of data on insulin glargine. Insulin analogues with a decreased dissociation rate from the insulin receptor are associated with higher mitogenic potency than metabolic potency compared with human insulin.^9,10 Degludec, like detemir, has an elevated dissociation rate from the insulin receptor, a low affinity for IGF-1 receptors, and a low mitogenic activity in vitro.⁸

At this juncture, neither detemir nor degludec has been associated with higher cancer risk, but these therapies are relatively new. And as Dr. Fountas et al indicated, their safety, particularly in regard to cancer risk in diabetes patients, should continue to be assessed.

In Reply: In regard to Dr. Weiss’s first point, the Kaiser Permanente Northern California diabetes registry study aimed to assess the association between bladder cancer and pioglitazone in 193,099 patients. In their 2011 interim 5-year analysis, Lewis et al reported a modest but statistically significant increased risk of bladder cancer in patients with type 2 diabetes mellitus who used pioglitazone for 2 or more years.¹

We appreciate Dr. Weiss’s comment on the 10-year study conclusion data. As Dr. Weiss has indicated, the recent Takeda news release² showed that the primary analysis found no association between pioglitazone use and bladder cancer risk. Furthermore, no association was found between bladder cancer risk and duration of use, higher cumulative doses, or time since initiation of pioglitazone.²

Regarding Dr. Weiss’s second point, we agree that at this time the cumulative data are not supportive of pancreatitis as per Egan et al.³ Recent publication of the SAVOR-TIMI trial⁴ of saxagliptin documented no increased risk of pancreatitis or pancreatic cancer over 2.1 years of follow-up in more than 16,000 patients over the age of 40 with type 2 diabetes. However, since amylase and lipase levels were not routinely checked in study participants, subclinical and asymptomatic cases may not have been recognized.⁴ Therefore, we stand by our statement that pancreatitis is a potential side effect.

It is important to recognize that although the observational data reviewed by both agencies (the US Food and Drug Administration and European Medicine Agency) in the publication by Egan et al³ are reassuring, we cannot yet say with absolute certainty that there is no associated risk. In fact, the concluding statements of the publication are as follows: “Although the totality of the data that have been reviewed provides reassurance, pancreatitis will continue to be considered a risk associated with these drugs until more data are available; both agencies continue to investigate this safety signal.”³

On September 18, 2014, the newest approved GLP-1 receptor agonist, dulaglutide, was approved with a boxed warning that it causes thyroid C-cell tumors in rats, that whether it causes thyroid C-cell tumors including medullary thyroid carcinoma (MTC) in humans is unknown, and that since relevance to humans could not be determined from clinical or nonclinical studies, dulaglutide is contraindicated in patients with a personal or family history of MTC, as well as in patients with multiple endocrine neoplasia syndrome type 2.⁵

It is important to recognize that despite these controversies, which have not been well-supported to date, incretin-based therapies have numerous metabolic benefits, including favorable glycemic and weight effects.

In regard to Dr. Weiss’s last point, we would like to point out the study by Gier et al⁶ in which GLP-1 receptor expression was found in 3 of 17 cases of human papillary thyroid cancer. The implication is that abnormal thyroid tissue does not behave the same way as normal tissue.

Furthermore, Dr. Weiss brings up the point that patients with thyroid cancer, if it is adequately treated, should have no remnant thyroid tissue. Certainly, adequate treatment would be an easy call to make if a stimulated thyroglobulin level is below the assay’s detection limit and there is no imaging evidence of residual thyroid cancer. For example, in someone with a history of thyroid cancer diagnosed more than 10 years ago without biochemical or imaging evidence of disease, any potential concerns of GLP-1 receptor agonist use in regards to thyroid cancer would be nominal. But not everyone with thyroid cancer falls into this category.

We do not suggest that these potential risks preclude the use of these agents in all patients, but rather that a discussion should occur between physician and patient. Diabetes therapy, as in treatment of other medical conditions, should be tailored to the individual patient, and all potential risk and benefits should be disclosed and considered.

In Reply: In regard to Dr. Weiss’s first point, the Kaiser Permanente Northern California diabetes registry study aimed to assess the association between bladder cancer and pioglitazone in 193,099 patients. In their 2011 interim 5-year analysis, Lewis et al reported a modest but statistically significant increased risk of bladder cancer in patients with type 2 diabetes mellitus who used pioglitazone for 2 or more years.¹

We appreciate Dr. Weiss’s comment on the 10-year study conclusion data. As Dr. Weiss has indicated, the recent Takeda news release² showed that the primary analysis found no association between pioglitazone use and bladder cancer risk. Furthermore, no association was found between bladder cancer risk and duration of use, higher cumulative doses, or time since initiation of pioglitazone.²

Regarding Dr. Weiss’s second point, we agree that at this time the cumulative data are not supportive of pancreatitis as per Egan et al.³ Recent publication of the SAVOR-TIMI trial⁴ of saxagliptin documented no increased risk of pancreatitis or pancreatic cancer over 2.1 years of follow-up in more than 16,000 patients over the age of 40 with type 2 diabetes. However, since amylase and lipase levels were not routinely checked in study participants, subclinical and asymptomatic cases may not have been recognized.⁴ Therefore, we stand by our statement that pancreatitis is a potential side effect.

It is important to recognize that although the observational data reviewed by both agencies (the US Food and Drug Administration and European Medicine Agency) in the publication by Egan et al³ are reassuring, we cannot yet say with absolute certainty that there is no associated risk. In fact, the concluding statements of the publication are as follows: “Although the totality of the data that have been reviewed provides reassurance, pancreatitis will continue to be considered a risk associated with these drugs until more data are available; both agencies continue to investigate this safety signal.”³

On September 18, 2014, the newest approved GLP-1 receptor agonist, dulaglutide, was approved with a boxed warning that it causes thyroid C-cell tumors in rats, that whether it causes thyroid C-cell tumors including medullary thyroid carcinoma (MTC) in humans is unknown, and that since relevance to humans could not be determined from clinical or nonclinical studies, dulaglutide is contraindicated in patients with a personal or family history of MTC, as well as in patients with multiple endocrine neoplasia syndrome type 2.⁵

It is important to recognize that despite these controversies, which have not been well-supported to date, incretin-based therapies have numerous metabolic benefits, including favorable glycemic and weight effects.

In regard to Dr. Weiss’s last point, we would like to point out the study by Gier et al⁶ in which GLP-1 receptor expression was found in 3 of 17 cases of human papillary thyroid cancer. The implication is that abnormal thyroid tissue does not behave the same way as normal tissue.

Furthermore, Dr. Weiss brings up the point that patients with thyroid cancer, if it is adequately treated, should have no remnant thyroid tissue. Certainly, adequate treatment would be an easy call to make if a stimulated thyroglobulin level is below the assay’s detection limit and there is no imaging evidence of residual thyroid cancer. For example, in someone with a history of thyroid cancer diagnosed more than 10 years ago without biochemical or imaging evidence of disease, any potential concerns of GLP-1 receptor agonist use in regards to thyroid cancer would be nominal. But not everyone with thyroid cancer falls into this category.

We do not suggest that these potential risks preclude the use of these agents in all patients, but rather that a discussion should occur between physician and patient. Diabetes therapy, as in treatment of other medical conditions, should be tailored to the individual patient, and all potential risk and benefits should be disclosed and considered.

In Reply: In regard to Dr. Weiss’s first point, the Kaiser Permanente Northern California diabetes registry study aimed to assess the association between bladder cancer and pioglitazone in 193,099 patients. In their 2011 interim 5-year analysis, Lewis et al reported a modest but statistically significant increased risk of bladder cancer in patients with type 2 diabetes mellitus who used pioglitazone for 2 or more years.¹

We appreciate Dr. Weiss’s comment on the 10-year study conclusion data. As Dr. Weiss has indicated, the recent Takeda news release² showed that the primary analysis found no association between pioglitazone use and bladder cancer risk. Furthermore, no association was found between bladder cancer risk and duration of use, higher cumulative doses, or time since initiation of pioglitazone.²

Regarding Dr. Weiss’s second point, we agree that at this time the cumulative data are not supportive of pancreatitis as per Egan et al.³ Recent publication of the SAVOR-TIMI trial⁴ of saxagliptin documented no increased risk of pancreatitis or pancreatic cancer over 2.1 years of follow-up in more than 16,000 patients over the age of 40 with type 2 diabetes. However, since amylase and lipase levels were not routinely checked in study participants, subclinical and asymptomatic cases may not have been recognized.⁴ Therefore, we stand by our statement that pancreatitis is a potential side effect.

It is important to recognize that although the observational data reviewed by both agencies (the US Food and Drug Administration and European Medicine Agency) in the publication by Egan et al³ are reassuring, we cannot yet say with absolute certainty that there is no associated risk. In fact, the concluding statements of the publication are as follows: “Although the totality of the data that have been reviewed provides reassurance, pancreatitis will continue to be considered a risk associated with these drugs until more data are available; both agencies continue to investigate this safety signal.”³

On September 18, 2014, the newest approved GLP-1 receptor agonist, dulaglutide, was approved with a boxed warning that it causes thyroid C-cell tumors in rats, that whether it causes thyroid C-cell tumors including medullary thyroid carcinoma (MTC) in humans is unknown, and that since relevance to humans could not be determined from clinical or nonclinical studies, dulaglutide is contraindicated in patients with a personal or family history of MTC, as well as in patients with multiple endocrine neoplasia syndrome type 2.⁵

It is important to recognize that despite these controversies, which have not been well-supported to date, incretin-based therapies have numerous metabolic benefits, including favorable glycemic and weight effects.

In regard to Dr. Weiss’s last point, we would like to point out the study by Gier et al⁶ in which GLP-1 receptor expression was found in 3 of 17 cases of human papillary thyroid cancer. The implication is that abnormal thyroid tissue does not behave the same way as normal tissue.

Furthermore, Dr. Weiss brings up the point that patients with thyroid cancer, if it is adequately treated, should have no remnant thyroid tissue. Certainly, adequate treatment would be an easy call to make if a stimulated thyroglobulin level is below the assay’s detection limit and there is no imaging evidence of residual thyroid cancer. For example, in someone with a history of thyroid cancer diagnosed more than 10 years ago without biochemical or imaging evidence of disease, any potential concerns of GLP-1 receptor agonist use in regards to thyroid cancer would be nominal. But not everyone with thyroid cancer falls into this category.

We do not suggest that these potential risks preclude the use of these agents in all patients, but rather that a discussion should occur between physician and patient. Diabetes therapy, as in treatment of other medical conditions, should be tailored to the individual patient, and all potential risk and benefits should be disclosed and considered.

In the last quarter century, many new drugs have become available for treating type 2 diabetes mellitus. The American Association of Clinical Endocrinologists incorporated these new agents in its updated glycemic control algorithm in 2013.¹ Because diabetes affects 25.8 million Americans and can lead to blindness, renal failure, cardiovascular disease, and amputation, agents that help us treat it more effectively are valuable.²

One of the barriers to effective treatment is the side effects of the agents. Because some of these drugs have been in use for only a short time, concerns of potential adverse effects have arisen. Cancer is one such concern, especially since type 2 diabetes mellitus by itself increases the risk of cancer by 20% to 50% compared with no diabetes.³

Type 2 diabetes has been linked to risk of cancers of the pancreas,⁴ colorectum,^5,6 liver,⁷ kidney,^8,9 breast,¹⁰ bladder,¹¹ and endometri-um,¹² as well as to hematologic malignancies such as non-Hodgkin lymphoma.¹³ The risk of bladder cancer appears to depend on how long the patient has had type 2 diabetes. Newton et al,¹⁴ in a prospective cohort study, found that those who had diabetes for more than 15 years and used insulin had the highest risk of bladder cancer. On the other hand, the risk of prostate cancer is actually lower in people with diabetes,¹⁵ particularly in those who have had diabetes for longer than 4 years.¹⁶

Cancer and type 2 diabetes share many risk factors and underlying pathophysiologic mechanisms. Nonmodifiable risk factors for both diseases include advanced age, male sex, ethnicity (African American men appear to be most vulnerable to both cancer and diabetes),^17,18 and family history. Modifiable risk factors include lower socioeconomic status, obesity, and alcohol consumption. These common risk factors lead to hyperinsulinemia and insulin resistance, changes in mitochondrial function, low-grade inflammation, and oxidative stress,³ which promote both diabetes and cancer. Diabetes therapy may influence several of these processes.

Several classes of diabetes drugs, including exogenous insulin,^19–22 insulin secretagogues,^23–25 and incretin-based therapies,^26–28 have been under scrutiny because of their potential influences on cancer development in a population already at risk (Table 1).

INSULIN ANALOGUES: MIXED EVIDENCE

Insulin promotes cell division by binding to insulin receptor isoform A and insulin-like growth factor 1 receptors.²⁹ Because endogenous hyperinsulinemia has been linked to cancer risk, growth, and proliferation, some speculate that exogenous insulin may also increase cancer risk.

In 2009, a retrospective study by Hemkens et al linked the long-acting insulin analogue glargine to risk of cancer.¹⁹ This finding set off a tumult of controversy within the medical community and concern among patients. Several limitations of the study were brought to light, including a short duration of follow-up, and several other studies have refuted the study’s findings.^20,21

More recently, the Outcome Reduction With Initial Glargine Intervention (ORIGIN) trial²² found no higher cancer risk with glargine use than with placebo. This study enrolled 12,537 participants from 573 sites in 40 countries. Specifically, risks with glargine use were as follows:

• Any cancer—hazard ratio 1.00, 95% confidence interval (CI) 0.88–1.13, P = .97
• Cancer death—hazard ratio 0.94, 95% CI 0.77–1.15, P = .52.

However, the study was designed to assess cardiovascular outcomes, not cancer risk. Furthermore, the participants were not typical of patients seen in clinical practice: their insulin doses were lower (the median insulin dose was 0.4 units/kg/day by year 6, whereas in clinical practice, those with type 2 diabetes mellitus often use more than 1 unit/kg/day, depending on duration of diabetes, diet, and exercise regimen), and their baseline median hemoglobin A_1c level was only 6.4%. And one may argue that the median follow-up of 6.2 years was too short for cancer to develop.²²

In vitro studies indicate that long-acting analogue insulin therapy may promote cancer cell growth more than endogenous insulin,³⁰ but epidemiologic data have not unequivocally substantiated this.^20–22 There is no clear evidence that analogue insulin therapy raises cancer risk above that of human recombinant insulin, and starting insulin therapy should not be delayed because of concerns about cancer risk, particularly in uncontrolled diabetes.

INSULIN SECRETAGOGUES

Sulfonylureas: Higher risk

Before 1995, only two classes of diabetes drugs were approved by the US Food and Drug Administration (FDA)—insulin and sulfonylureas.

Sulfonylureas lower blood sugar levels by binding to sulfonylurea receptors and inhibiting adenosine triphosphate-dependent potassium channels. The resulting change in resting potential causes an influx of calcium, ultimately leading to insulin secretion.

Sulfonylureas are effective, and because of their low cost, physicians often pick them as a second-line agent after metformin.

The main disadvantage of sulfonylureas is the risk of hypoglycemia, particularly in patients with renal failure, the elderly, and diabetic patients who are unaware of when they are hypoglycemic. Other potential drawbacks are that they impair cardiac ischemic preconditioning³¹ and possibly increase cancer risk.^21,32 (Ischemic preconditioning is the process in which transient episodes of ischemia “condition” the myocardium so that it better withstands future episodes with minimal anginal pain and tissue injury.³³) Of the sulfonylureas, glyburide has been most implicated in cardiovascular risk.³²

In a retrospective cohort study of 62,809 patients from a general-practice database in the United Kingdom, Currie et al²¹ found that sulfonylurea monotherapy was associated with a 36% higher risk of cancer (95% CI 1.19–1.54, P < .001) than metformin monotherapy. Prescribing bias may have influenced the results: practitioners are more likely to prescribe sulfonylureas to leaner patients, who have a greater likelihood of occult cancer. However, other studies also found that the cancer death rate is higher in those who take a sulfonylurea alone than in those who use metformin alone.^23,24

Some evidence indicates that long-acting sulfonylurea formulations (eg, glyburide) likely hold the most danger, certainly in regard to hypoglycemia, but it is less clear if this translates to cancer concerns.³¹

Meglitinides: Limited evidence

Meglitinides, the other class of insulin secretagogues, are less commonly used but are similar to sulfonylureas in the way they increase endogenous insulin levels. The data are limited regarding cancer risk and meglitinide therapy, but the magnitude of the association is similar to that with sulfonylurea therapy.²⁵

INSULIN SENSITIZERS

There are currently two classes of insulin sensitizers: biguanides and thiazolidinediones (TZDs, also known as glitazones). These drugs show less risk of both cancer incidence and cancer death than insulin secretagogues such as sulfonylureas.^21,23,24 In fact, they may decrease cancer potential by alteration of signaling via the AKT/mTOR (v-akt murine thymoma viral oncogene homolog 1/mammalian target of rapamycin) pathway.³⁴

Metformin, a biguanide, is the oral drug of choice

Metformin is the only biguanide currently available in the United States. It was approved by the FDA in 1995, although it had been in clinical use since the 1950s. Inexpensive and familiar, it is the oral antihyperglycemic of choice if there are no contraindications to it, such as renal dysfunction (creatinine ≥ 1.4 mg/dL in women and ≥ 1.5 mg/dL in men), acute decompensated heart failure, or pulmonary or hepatic insufficiency, all of which may lead to an increased risk of lactic acidosis.¹

Metformin lowers blood sugar levels primarily by inhibiting hepatic glucose production (gluconeogenesis) and by improving peripheral insulin sensitivity. It directly activates AMP-activated protein kinase (AMPK), which affects insulin signaling and glucose and fat metabolism.³⁵ It may exert further beneficial effects by acutely increasing glucagon-like peptide-1 (GLP-1) levels and inducing islet incretin-receptor gene expression.³⁶ Although the exact mechanisms have not been fully elucidated, metformin’s insulin-sensitizing properties are likely from favorable effects on insulin receptor expression, tyrosine kinase activity, and influences on the incretin pathway.^36,37 These effects also mitigate carcinogenesis, both directly (via AMPK and liver kinase B1, a tumor-suppressor gene) and indirectly (via reduction of hyperinsulinemia).³⁵

Overall, biguanide therapy is associated with a lower cancer incidence or, at worst, no effect on cancer incidence. In vitro studies demonstrate that metformin both suppresses cancer cell growth and induces apoptosis, resulting in fewer live cancer cells.³⁴ Several retrospective studies found lower cancer risk in metformin users than in patients receiving antidiabetes drugs other than insulin-sensitizing agents,^{21,23,25,38–40} while others have shown no effect.⁴¹ Use of metformin was specifically associated with lower risk of cancers of the liver, colon and rectum, and lung.⁴² Further, metformin users have a lower cancer mortality rate than nonusers.^24,43

Thiazolidinediones

TZDs, such as pioglitazone, work by binding to peroxisome proliferator-activated gamma receptors in the cell nucleus, altering gene transcription.⁴⁴ They reduce insulin resistance and levels of endogenous insulin levels and free fatty acids.⁴⁴

Concern over bladder cancer risk with TZD use, particularly with pioglitazone, has increased in the last few years, as various cohort studies found a statistically significant increased risk with this agent.⁴⁴ The risk appears to rise with cumulative dose.^45,46

Randomized controlled trials also found an increased risk of bladder cancer with TZD therapy, although the difference was not statistically significant.^47–49 In a mean follow-up of 8.7 years, the Prospective Pioglitazone Clinical Trial in Macrovascular Events reported 23 cases of bladder cancer in the pioglitazone group vs 22 cases in the placebo group, for rates of 0.9% vs 0.8% (relative risk [RR] 1.06, 95% CI 0.59–1.89).⁴⁹

On the other hand, the risk of cancer of the breast, colon, and lung has been found to be lower with TZD use.⁴⁷ In vitro studies support the clinical data, showing that TZDs inhibit growth of human cancer cells derived from cancers of the lung, colon, breast, stomach, ovary, and prostate.^50–53

Home et al⁵⁴ compared rosiglitazone against a sulfonylurea in patients already taking metformin in the Rosiglitazone Evaluated for Cardiovascular Outcomes in Oral Agent Combination Therapy for Type 2 Diabetes (RECORD) trial. Malignancies developed in 6.7% of the sulfonylurea group compared with 5.1% of the rosiglitazone group, for a hazard ratio of 1.33 (95% CI 0.94–1.88).

Both ADOPT (A Diabetes Outcome Progression Trial) and the RECORD trial found rosiglitazone comparable to metformin in terms of cancer risk.⁵⁴

Colmers et al⁴⁷ pooled data from four randomized controlled trials, seven cohort studies, and nine case-control studies to assess the risk of cancer with TZD use in type 2 diabetes. Both the randomized and observational data showed neutral overall cancer risk with TZDs. However, pooled data from observational studies showed significantly lower risk with TZD use in terms of:

• Colorectal cancer RR 0.93, 95% CI 0.87–1.00
• Lung cancer RR 0.91, 95% CI 0.84–0.98
• Breast cancer RR 0.89, 95% CI 0.81–0.98.

INCRETIN-BASED THERAPIES

Incretins are hormones released from the gut in response to food ingestion, triggering release of insulin before blood glucose levels rise. Their action explains why insulin secretion increases more after an oral glucose load than after an intravenous glucose load, a phenomenon called the incretin effect.⁵⁵

There are two incretin hormones: glucose-dependent insulinotropic polypeptide (GIP) and glucagon-like peptide 1 (GLP-1). They have short a half-life because they are rapidly degraded by dipeptidyl peptidase-IV (DPP-IV).⁵⁵ Available incretin-based therapies are GLP-1 receptor agonists and DPP-IV inhibitors.

When used as monotherapy, incretin-based therapies do not cause hypoglycemia because their effect is glucose-dependent.⁵⁵ GLP-1 receptor antagonists have the added benefit of inducing weight loss, but DPP-IV inhibitors are considered to be weight-neutral.

GLP-1 receptor agonists

Exenatide, the first of the GLP-1 receptor agonists, was approved in 2005. The original formulation (Byetta) is taken by injection twice daily, and timing in conjunction with food intake is important: it should be taken within 60 minutes before the morning and evening meals. Extended-release exenatide (Bydureon) is a once-weekly formulation taken without regard to timing of food intake. Exenatide (either twice-daily Byetta or once-weekly Bydureon) should not be used in those with creatinine clearance less than 30 mL/min or end-stage renal disease and should be used with caution in patients with renal transplantation.

Liraglutide (Victoza), a once-daily formulation, can be injected irrespective of food intake. The dose does not have to be adjusted for renal function, although it should be used with caution in those with renal impairment, including end-stage renal disease. Approval for a 3-mg formulation is pending with the FDA as a weight-loss drug on the basis of promising results in a randomized phase 3 trial.⁵⁶

Albiglutide (Tanzeum), a once-weekly GLP-1 receptor antagonist, was recently approved by the FDA.

DPP-IV inhibitors

Whereas GLP-1 receptor agonists are injected, the DPP-IV inhibitors have the advantage of being oral agents.

Sitagliptin (Januvia), the first DPP-IV inhibitor, became available in the United States in 2006. Since then, three more have become available: saxagliptin (Onglyza), linagliptin (Tradjenta), and alogliptin (Nesina).

Concerns about thyroid cancer with incretin drugs

Concerns of increased risk of cancer, particularly of the thyroid and pancreas, have been raised since GLP-1 receptor agonists and DPP-IV inhibitors became available.

Studies in rodents have shown C-cell hyperplasia, sometimes resulting in increased incidence of thyroid carcinoma, and dose-dependent rises in serum calcitonin, particularly with liraglutide.²⁶ This has raised concern about an increased risk of medullary thyroid carcinoma in humans. However, the density of C cells in rodents is up to 45 times greater than in humans, and C cells also express functional GLP-1 receptors.²⁶

Gier et al²⁷ assessed the expression of calcitonin and human GLP-1 receptors in normal C cells, C cell hyperplasia, and medullary cancer. In this study, calcitonin and GLP-1 receptor were co-expressed in medullary thyroid cancer (10 of 12 cases) and C-cell hyperplasia (9 of 9 cases) more commonly than in normal C cells (5 of 15 cases). Further, GLP-1 receptor was expressed in 3 of 17 cases of papillary thyroid cancer.

Calcitonin, a polypeptide hormone produced by thyroid C cells and used as a medullary thyroid cancer biomarker, was increased in a slightly higher percentage of patients treated with liraglutide than in controls, without an increase above the normal range.⁵⁷

A meta-analysis by Alves et al⁵⁸ of 25 studies found that neither exenatide (no cases reported) nor liraglutide (odds ratio 1.54, 95% CI 0.40–6.02) was associated with increased thyroid cancer risk.

MacConell et al⁵⁹ pooled the results of 19 placebo-controlled trials of twice-daily exenatide and found a thyroid cancer incidence rate of 0.3 per 100 patient-years (< 0.1%) vs 0 per 100 patient-years in pooled comparators.

Concerns about pancreatic cancer with incretin drugs

Increased risk of acute pancreatitis is a potential side effect of both DPP-IV inhibitors and GLP-1 receptor agonists and has led to speculation that this translates to an increased risk of pancreatic cancer.

In a point-counterpoint debate, Butler et al²⁸ argued that incretin-based medications have questionable safety, with increased rates of pancreatitis possibly leading to pancreatic cancer. In counterpoint, Nauck⁶⁰ argued that the risk of pancreatitis or cancer is extremely low, and clinical cases are unsubstantiated.

Bailey⁶¹ outlined the complexities and difficulties in drawing firm conclusions from individual clinical trials regarding possible adverse effects of diabetes drugs. The trials are typically designed to assess hemoglobin A_1c reduction at varying doses and are typically restricted in patient selection, patient numbers, and drug-exposure duration, which may introduce allocation and ascertainment biases. The attempt to draw firm conclusions from such trials can be problematic and can lead to increased alarm, warranted or not.

Type 2 diabetes mellitus itself is associated with an increased incidence of pancreatic cancer, and whether incretin therapy enhances this risk is still controversial. Whether more episodes of acute pancreatitis without chronic pancreatitis can be extrapolated to an increased incidence of pancreatic cancer is doubtful. A normal pancreatic duct cell may take up to 12 years to become a tumor cell from which pancreatic carcinoma develops, another 7 years to develop metastatic capacity, and another 3 years before a diagnosis is made from clinical symptoms (which are usually accompanied by metastases).⁶²

The risks and benefits of incretin therapies remain a contentious issue, and there are no clear prospective data at this time on increased pancreatic cancer incidence. Long-term prospective studies designed to analyze these specific outcomes (pancreatitis, pancreatic cancer, and medullary thyroid cancer) need to be undertaken.⁶³

OTHER DIABETES THERAPIES

Alpha glucosidase inhibitors

Oral glucosidase inhibitors ameliorate hyperglycemia by inhibiting alpha glucosidase enzymes in the brush border of the small intestines, preventing conversion of polysaccharides to monosaccharides.⁶⁴ This slows digestion of carbohydrates and glucose release into the bloodstream and blunts the postprandial hyperglycemic excursion.

The two alpha glucosidase inhibitors currently available in the United States are acarbose and miglitol, and although data are limited, they do not appear to increase the risk of cancer.^65,66

Sodium-glucose-linked cotransporter 2 inhibitors

The newest class of oral diabetes agents to be approved are the sodium-glucose-linked cotransporter 2 (SGLT2) inhibitors canagliflozin (Invokana) and dapagliflozin (Farxiga).

SGLT2 is a protein in the S1 segment of the proximal renal tubules responsible for over 90% of renal glucose reabsorption. SGLT2 inhibitors lower serum glucose levels by promoting glycosuria and have also been shown to have favorable effects on blood pressure and weight.^67,68

Canagliflozin was the first of its class to gain FDA approval in the United States. It has not been found to be associated with increased cancer risk.⁶⁸

Dapagliflozin, originally approved in Europe, was approved in the United States on January 8, 2014. Because of a possible increased incidence of breast and bladder malignancies, the FDA advisory committee initially recommended against approval and required further data. In those who were treated, nine cases of bladder cancer and nine cases of breast cancer were reported, compared with one case of bladder cancer and no cases of breast cancer in the control group; however, the difference was not statistically significant.⁶⁸

Since SGLT2 inhibitors are still new, data on long-term outcomes are lacking. Early clinical data do not show a significant increase in cancer risk.

WHAT THIS MEANS IN PRACTICE

Many studies have found associations between diabetes, obesity, hyperinsulinemia, and cancer risk. In the last decade, concerns implicating antihyperglycemic agents in cancer development have arisen but have not been well substantiated. At this time, there are no definitive prospective data indicating that the currently available type 2 diabetes therapies increase the incidence of cancer beyond the inherent increased risk in this population. What, then, is one to do?

Educate. Lifestyle modification, including weight management, should continue to be emphasized in diabetes education, as no therapy is completely effective without adjunct modifications in diet and physical activity. Epidemiologic studies have shown the benefits of lifestyle modifications, which ameliorate many of the adverse metabolic conditions that coexist in type 2 diabetes and cancer.

Screen for cancer. Given the associations between diabetes and malignancy, cancer screening is especially important in this high-risk population.

Customize therapy to individual patients. Those with a personal history of bladder cancer should avoid pioglitazone, and those who have had pancreatic cancer should avoid sitagliptin until definitive clinical data become available.

Moreover, patients with a personal or family history of medullary thyroid cancer should not receive GLP-1 receptor agonists. These agents should also probably be avoided in patients with a personal history of differentiated thyroid carcinoma or a history of familial nonmedullary thyroid carcinoma. Until we have further elucidating data, it is not possible to say whether a family history of any of the other types of cancer should represent a contraindication to the use of any of these agents.

Discuss. The multitude of diabetes therapies warrants physician-patient discussions that carefully weigh the risks and benefits of additional agents to optimize glycemic control and metabolic factors in individual patients.

In the last quarter century, many new drugs have become available for treating type 2 diabetes mellitus. The American Association of Clinical Endocrinologists incorporated these new agents in its updated glycemic control algorithm in 2013.¹ Because diabetes affects 25.8 million Americans and can lead to blindness, renal failure, cardiovascular disease, and amputation, agents that help us treat it more effectively are valuable.²

One of the barriers to effective treatment is the side effects of the agents. Because some of these drugs have been in use for only a short time, concerns of potential adverse effects have arisen. Cancer is one such concern, especially since type 2 diabetes mellitus by itself increases the risk of cancer by 20% to 50% compared with no diabetes.³

Type 2 diabetes has been linked to risk of cancers of the pancreas,⁴ colorectum,^5,6 liver,⁷ kidney,^8,9 breast,¹⁰ bladder,¹¹ and endometri-um,¹² as well as to hematologic malignancies such as non-Hodgkin lymphoma.¹³ The risk of bladder cancer appears to depend on how long the patient has had type 2 diabetes. Newton et al,¹⁴ in a prospective cohort study, found that those who had diabetes for more than 15 years and used insulin had the highest risk of bladder cancer. On the other hand, the risk of prostate cancer is actually lower in people with diabetes,¹⁵ particularly in those who have had diabetes for longer than 4 years.¹⁶

Cancer and type 2 diabetes share many risk factors and underlying pathophysiologic mechanisms. Nonmodifiable risk factors for both diseases include advanced age, male sex, ethnicity (African American men appear to be most vulnerable to both cancer and diabetes),^17,18 and family history. Modifiable risk factors include lower socioeconomic status, obesity, and alcohol consumption. These common risk factors lead to hyperinsulinemia and insulin resistance, changes in mitochondrial function, low-grade inflammation, and oxidative stress,³ which promote both diabetes and cancer. Diabetes therapy may influence several of these processes.

Several classes of diabetes drugs, including exogenous insulin,^19–22 insulin secretagogues,^23–25 and incretin-based therapies,^26–28 have been under scrutiny because of their potential influences on cancer development in a population already at risk (Table 1).

INSULIN ANALOGUES: MIXED EVIDENCE

Insulin promotes cell division by binding to insulin receptor isoform A and insulin-like growth factor 1 receptors.²⁹ Because endogenous hyperinsulinemia has been linked to cancer risk, growth, and proliferation, some speculate that exogenous insulin may also increase cancer risk.

In 2009, a retrospective study by Hemkens et al linked the long-acting insulin analogue glargine to risk of cancer.¹⁹ This finding set off a tumult of controversy within the medical community and concern among patients. Several limitations of the study were brought to light, including a short duration of follow-up, and several other studies have refuted the study’s findings.^20,21

More recently, the Outcome Reduction With Initial Glargine Intervention (ORIGIN) trial²² found no higher cancer risk with glargine use than with placebo. This study enrolled 12,537 participants from 573 sites in 40 countries. Specifically, risks with glargine use were as follows:

• Any cancer—hazard ratio 1.00, 95% confidence interval (CI) 0.88–1.13, P = .97
• Cancer death—hazard ratio 0.94, 95% CI 0.77–1.15, P = .52.

However, the study was designed to assess cardiovascular outcomes, not cancer risk. Furthermore, the participants were not typical of patients seen in clinical practice: their insulin doses were lower (the median insulin dose was 0.4 units/kg/day by year 6, whereas in clinical practice, those with type 2 diabetes mellitus often use more than 1 unit/kg/day, depending on duration of diabetes, diet, and exercise regimen), and their baseline median hemoglobin A_1c level was only 6.4%. And one may argue that the median follow-up of 6.2 years was too short for cancer to develop.²²

In vitro studies indicate that long-acting analogue insulin therapy may promote cancer cell growth more than endogenous insulin,³⁰ but epidemiologic data have not unequivocally substantiated this.^20–22 There is no clear evidence that analogue insulin therapy raises cancer risk above that of human recombinant insulin, and starting insulin therapy should not be delayed because of concerns about cancer risk, particularly in uncontrolled diabetes.

INSULIN SECRETAGOGUES

Sulfonylureas: Higher risk

Before 1995, only two classes of diabetes drugs were approved by the US Food and Drug Administration (FDA)—insulin and sulfonylureas.

Sulfonylureas lower blood sugar levels by binding to sulfonylurea receptors and inhibiting adenosine triphosphate-dependent potassium channels. The resulting change in resting potential causes an influx of calcium, ultimately leading to insulin secretion.

Sulfonylureas are effective, and because of their low cost, physicians often pick them as a second-line agent after metformin.

The main disadvantage of sulfonylureas is the risk of hypoglycemia, particularly in patients with renal failure, the elderly, and diabetic patients who are unaware of when they are hypoglycemic. Other potential drawbacks are that they impair cardiac ischemic preconditioning³¹ and possibly increase cancer risk.^21,32 (Ischemic preconditioning is the process in which transient episodes of ischemia “condition” the myocardium so that it better withstands future episodes with minimal anginal pain and tissue injury.³³) Of the sulfonylureas, glyburide has been most implicated in cardiovascular risk.³²

In a retrospective cohort study of 62,809 patients from a general-practice database in the United Kingdom, Currie et al²¹ found that sulfonylurea monotherapy was associated with a 36% higher risk of cancer (95% CI 1.19–1.54, P < .001) than metformin monotherapy. Prescribing bias may have influenced the results: practitioners are more likely to prescribe sulfonylureas to leaner patients, who have a greater likelihood of occult cancer. However, other studies also found that the cancer death rate is higher in those who take a sulfonylurea alone than in those who use metformin alone.^23,24

Some evidence indicates that long-acting sulfonylurea formulations (eg, glyburide) likely hold the most danger, certainly in regard to hypoglycemia, but it is less clear if this translates to cancer concerns.³¹

Meglitinides: Limited evidence

Meglitinides, the other class of insulin secretagogues, are less commonly used but are similar to sulfonylureas in the way they increase endogenous insulin levels. The data are limited regarding cancer risk and meglitinide therapy, but the magnitude of the association is similar to that with sulfonylurea therapy.²⁵

INSULIN SENSITIZERS

There are currently two classes of insulin sensitizers: biguanides and thiazolidinediones (TZDs, also known as glitazones). These drugs show less risk of both cancer incidence and cancer death than insulin secretagogues such as sulfonylureas.^21,23,24 In fact, they may decrease cancer potential by alteration of signaling via the AKT/mTOR (v-akt murine thymoma viral oncogene homolog 1/mammalian target of rapamycin) pathway.³⁴

Metformin, a biguanide, is the oral drug of choice

Metformin is the only biguanide currently available in the United States. It was approved by the FDA in 1995, although it had been in clinical use since the 1950s. Inexpensive and familiar, it is the oral antihyperglycemic of choice if there are no contraindications to it, such as renal dysfunction (creatinine ≥ 1.4 mg/dL in women and ≥ 1.5 mg/dL in men), acute decompensated heart failure, or pulmonary or hepatic insufficiency, all of which may lead to an increased risk of lactic acidosis.¹

Metformin lowers blood sugar levels primarily by inhibiting hepatic glucose production (gluconeogenesis) and by improving peripheral insulin sensitivity. It directly activates AMP-activated protein kinase (AMPK), which affects insulin signaling and glucose and fat metabolism.³⁵ It may exert further beneficial effects by acutely increasing glucagon-like peptide-1 (GLP-1) levels and inducing islet incretin-receptor gene expression.³⁶ Although the exact mechanisms have not been fully elucidated, metformin’s insulin-sensitizing properties are likely from favorable effects on insulin receptor expression, tyrosine kinase activity, and influences on the incretin pathway.^36,37 These effects also mitigate carcinogenesis, both directly (via AMPK and liver kinase B1, a tumor-suppressor gene) and indirectly (via reduction of hyperinsulinemia).³⁵

Overall, biguanide therapy is associated with a lower cancer incidence or, at worst, no effect on cancer incidence. In vitro studies demonstrate that metformin both suppresses cancer cell growth and induces apoptosis, resulting in fewer live cancer cells.³⁴ Several retrospective studies found lower cancer risk in metformin users than in patients receiving antidiabetes drugs other than insulin-sensitizing agents,^{21,23,25,38–40} while others have shown no effect.⁴¹ Use of metformin was specifically associated with lower risk of cancers of the liver, colon and rectum, and lung.⁴² Further, metformin users have a lower cancer mortality rate than nonusers.^24,43

Thiazolidinediones

TZDs, such as pioglitazone, work by binding to peroxisome proliferator-activated gamma receptors in the cell nucleus, altering gene transcription.⁴⁴ They reduce insulin resistance and levels of endogenous insulin levels and free fatty acids.⁴⁴

Concern over bladder cancer risk with TZD use, particularly with pioglitazone, has increased in the last few years, as various cohort studies found a statistically significant increased risk with this agent.⁴⁴ The risk appears to rise with cumulative dose.^45,46

Randomized controlled trials also found an increased risk of bladder cancer with TZD therapy, although the difference was not statistically significant.^47–49 In a mean follow-up of 8.7 years, the Prospective Pioglitazone Clinical Trial in Macrovascular Events reported 23 cases of bladder cancer in the pioglitazone group vs 22 cases in the placebo group, for rates of 0.9% vs 0.8% (relative risk [RR] 1.06, 95% CI 0.59–1.89).⁴⁹

On the other hand, the risk of cancer of the breast, colon, and lung has been found to be lower with TZD use.⁴⁷ In vitro studies support the clinical data, showing that TZDs inhibit growth of human cancer cells derived from cancers of the lung, colon, breast, stomach, ovary, and prostate.^50–53

Home et al⁵⁴ compared rosiglitazone against a sulfonylurea in patients already taking metformin in the Rosiglitazone Evaluated for Cardiovascular Outcomes in Oral Agent Combination Therapy for Type 2 Diabetes (RECORD) trial. Malignancies developed in 6.7% of the sulfonylurea group compared with 5.1% of the rosiglitazone group, for a hazard ratio of 1.33 (95% CI 0.94–1.88).

Both ADOPT (A Diabetes Outcome Progression Trial) and the RECORD trial found rosiglitazone comparable to metformin in terms of cancer risk.⁵⁴

Colmers et al⁴⁷ pooled data from four randomized controlled trials, seven cohort studies, and nine case-control studies to assess the risk of cancer with TZD use in type 2 diabetes. Both the randomized and observational data showed neutral overall cancer risk with TZDs. However, pooled data from observational studies showed significantly lower risk with TZD use in terms of:

• Colorectal cancer RR 0.93, 95% CI 0.87–1.00
• Lung cancer RR 0.91, 95% CI 0.84–0.98
• Breast cancer RR 0.89, 95% CI 0.81–0.98.

INCRETIN-BASED THERAPIES

Incretins are hormones released from the gut in response to food ingestion, triggering release of insulin before blood glucose levels rise. Their action explains why insulin secretion increases more after an oral glucose load than after an intravenous glucose load, a phenomenon called the incretin effect.⁵⁵

There are two incretin hormones: glucose-dependent insulinotropic polypeptide (GIP) and glucagon-like peptide 1 (GLP-1). They have short a half-life because they are rapidly degraded by dipeptidyl peptidase-IV (DPP-IV).⁵⁵ Available incretin-based therapies are GLP-1 receptor agonists and DPP-IV inhibitors.

When used as monotherapy, incretin-based therapies do not cause hypoglycemia because their effect is glucose-dependent.⁵⁵ GLP-1 receptor antagonists have the added benefit of inducing weight loss, but DPP-IV inhibitors are considered to be weight-neutral.

GLP-1 receptor agonists

Exenatide, the first of the GLP-1 receptor agonists, was approved in 2005. The original formulation (Byetta) is taken by injection twice daily, and timing in conjunction with food intake is important: it should be taken within 60 minutes before the morning and evening meals. Extended-release exenatide (Bydureon) is a once-weekly formulation taken without regard to timing of food intake. Exenatide (either twice-daily Byetta or once-weekly Bydureon) should not be used in those with creatinine clearance less than 30 mL/min or end-stage renal disease and should be used with caution in patients with renal transplantation.

Liraglutide (Victoza), a once-daily formulation, can be injected irrespective of food intake. The dose does not have to be adjusted for renal function, although it should be used with caution in those with renal impairment, including end-stage renal disease. Approval for a 3-mg formulation is pending with the FDA as a weight-loss drug on the basis of promising results in a randomized phase 3 trial.⁵⁶

Albiglutide (Tanzeum), a once-weekly GLP-1 receptor antagonist, was recently approved by the FDA.

DPP-IV inhibitors

Whereas GLP-1 receptor agonists are injected, the DPP-IV inhibitors have the advantage of being oral agents.

Sitagliptin (Januvia), the first DPP-IV inhibitor, became available in the United States in 2006. Since then, three more have become available: saxagliptin (Onglyza), linagliptin (Tradjenta), and alogliptin (Nesina).

Concerns about thyroid cancer with incretin drugs

Concerns of increased risk of cancer, particularly of the thyroid and pancreas, have been raised since GLP-1 receptor agonists and DPP-IV inhibitors became available.

Studies in rodents have shown C-cell hyperplasia, sometimes resulting in increased incidence of thyroid carcinoma, and dose-dependent rises in serum calcitonin, particularly with liraglutide.²⁶ This has raised concern about an increased risk of medullary thyroid carcinoma in humans. However, the density of C cells in rodents is up to 45 times greater than in humans, and C cells also express functional GLP-1 receptors.²⁶

Gier et al²⁷ assessed the expression of calcitonin and human GLP-1 receptors in normal C cells, C cell hyperplasia, and medullary cancer. In this study, calcitonin and GLP-1 receptor were co-expressed in medullary thyroid cancer (10 of 12 cases) and C-cell hyperplasia (9 of 9 cases) more commonly than in normal C cells (5 of 15 cases). Further, GLP-1 receptor was expressed in 3 of 17 cases of papillary thyroid cancer.

Calcitonin, a polypeptide hormone produced by thyroid C cells and used as a medullary thyroid cancer biomarker, was increased in a slightly higher percentage of patients treated with liraglutide than in controls, without an increase above the normal range.⁵⁷

A meta-analysis by Alves et al⁵⁸ of 25 studies found that neither exenatide (no cases reported) nor liraglutide (odds ratio 1.54, 95% CI 0.40–6.02) was associated with increased thyroid cancer risk.

MacConell et al⁵⁹ pooled the results of 19 placebo-controlled trials of twice-daily exenatide and found a thyroid cancer incidence rate of 0.3 per 100 patient-years (< 0.1%) vs 0 per 100 patient-years in pooled comparators.

Concerns about pancreatic cancer with incretin drugs

Increased risk of acute pancreatitis is a potential side effect of both DPP-IV inhibitors and GLP-1 receptor agonists and has led to speculation that this translates to an increased risk of pancreatic cancer.

In a point-counterpoint debate, Butler et al²⁸ argued that incretin-based medications have questionable safety, with increased rates of pancreatitis possibly leading to pancreatic cancer. In counterpoint, Nauck⁶⁰ argued that the risk of pancreatitis or cancer is extremely low, and clinical cases are unsubstantiated.

Bailey⁶¹ outlined the complexities and difficulties in drawing firm conclusions from individual clinical trials regarding possible adverse effects of diabetes drugs. The trials are typically designed to assess hemoglobin A_1c reduction at varying doses and are typically restricted in patient selection, patient numbers, and drug-exposure duration, which may introduce allocation and ascertainment biases. The attempt to draw firm conclusions from such trials can be problematic and can lead to increased alarm, warranted or not.

Type 2 diabetes mellitus itself is associated with an increased incidence of pancreatic cancer, and whether incretin therapy enhances this risk is still controversial. Whether more episodes of acute pancreatitis without chronic pancreatitis can be extrapolated to an increased incidence of pancreatic cancer is doubtful. A normal pancreatic duct cell may take up to 12 years to become a tumor cell from which pancreatic carcinoma develops, another 7 years to develop metastatic capacity, and another 3 years before a diagnosis is made from clinical symptoms (which are usually accompanied by metastases).⁶²

The risks and benefits of incretin therapies remain a contentious issue, and there are no clear prospective data at this time on increased pancreatic cancer incidence. Long-term prospective studies designed to analyze these specific outcomes (pancreatitis, pancreatic cancer, and medullary thyroid cancer) need to be undertaken.⁶³

OTHER DIABETES THERAPIES

Alpha glucosidase inhibitors

Oral glucosidase inhibitors ameliorate hyperglycemia by inhibiting alpha glucosidase enzymes in the brush border of the small intestines, preventing conversion of polysaccharides to monosaccharides.⁶⁴ This slows digestion of carbohydrates and glucose release into the bloodstream and blunts the postprandial hyperglycemic excursion.

The two alpha glucosidase inhibitors currently available in the United States are acarbose and miglitol, and although data are limited, they do not appear to increase the risk of cancer.^65,66

Sodium-glucose-linked cotransporter 2 inhibitors

The newest class of oral diabetes agents to be approved are the sodium-glucose-linked cotransporter 2 (SGLT2) inhibitors canagliflozin (Invokana) and dapagliflozin (Farxiga).

SGLT2 is a protein in the S1 segment of the proximal renal tubules responsible for over 90% of renal glucose reabsorption. SGLT2 inhibitors lower serum glucose levels by promoting glycosuria and have also been shown to have favorable effects on blood pressure and weight.^67,68

Canagliflozin was the first of its class to gain FDA approval in the United States. It has not been found to be associated with increased cancer risk.⁶⁸

Dapagliflozin, originally approved in Europe, was approved in the United States on January 8, 2014. Because of a possible increased incidence of breast and bladder malignancies, the FDA advisory committee initially recommended against approval and required further data. In those who were treated, nine cases of bladder cancer and nine cases of breast cancer were reported, compared with one case of bladder cancer and no cases of breast cancer in the control group; however, the difference was not statistically significant.⁶⁸

Since SGLT2 inhibitors are still new, data on long-term outcomes are lacking. Early clinical data do not show a significant increase in cancer risk.

WHAT THIS MEANS IN PRACTICE

Many studies have found associations between diabetes, obesity, hyperinsulinemia, and cancer risk. In the last decade, concerns implicating antihyperglycemic agents in cancer development have arisen but have not been well substantiated. At this time, there are no definitive prospective data indicating that the currently available type 2 diabetes therapies increase the incidence of cancer beyond the inherent increased risk in this population. What, then, is one to do?

Educate. Lifestyle modification, including weight management, should continue to be emphasized in diabetes education, as no therapy is completely effective without adjunct modifications in diet and physical activity. Epidemiologic studies have shown the benefits of lifestyle modifications, which ameliorate many of the adverse metabolic conditions that coexist in type 2 diabetes and cancer.

Screen for cancer. Given the associations between diabetes and malignancy, cancer screening is especially important in this high-risk population.

Customize therapy to individual patients. Those with a personal history of bladder cancer should avoid pioglitazone, and those who have had pancreatic cancer should avoid sitagliptin until definitive clinical data become available.

Moreover, patients with a personal or family history of medullary thyroid cancer should not receive GLP-1 receptor agonists. These agents should also probably be avoided in patients with a personal history of differentiated thyroid carcinoma or a history of familial nonmedullary thyroid carcinoma. Until we have further elucidating data, it is not possible to say whether a family history of any of the other types of cancer should represent a contraindication to the use of any of these agents.

Discuss. The multitude of diabetes therapies warrants physician-patient discussions that carefully weigh the risks and benefits of additional agents to optimize glycemic control and metabolic factors in individual patients.

In the last quarter century, many new drugs have become available for treating type 2 diabetes mellitus. The American Association of Clinical Endocrinologists incorporated these new agents in its updated glycemic control algorithm in 2013.¹ Because diabetes affects 25.8 million Americans and can lead to blindness, renal failure, cardiovascular disease, and amputation, agents that help us treat it more effectively are valuable.²

One of the barriers to effective treatment is the side effects of the agents. Because some of these drugs have been in use for only a short time, concerns of potential adverse effects have arisen. Cancer is one such concern, especially since type 2 diabetes mellitus by itself increases the risk of cancer by 20% to 50% compared with no diabetes.³

Type 2 diabetes has been linked to risk of cancers of the pancreas,⁴ colorectum,^5,6 liver,⁷ kidney,^8,9 breast,¹⁰ bladder,¹¹ and endometri-um,¹² as well as to hematologic malignancies such as non-Hodgkin lymphoma.¹³ The risk of bladder cancer appears to depend on how long the patient has had type 2 diabetes. Newton et al,¹⁴ in a prospective cohort study, found that those who had diabetes for more than 15 years and used insulin had the highest risk of bladder cancer. On the other hand, the risk of prostate cancer is actually lower in people with diabetes,¹⁵ particularly in those who have had diabetes for longer than 4 years.¹⁶

Cancer and type 2 diabetes share many risk factors and underlying pathophysiologic mechanisms. Nonmodifiable risk factors for both diseases include advanced age, male sex, ethnicity (African American men appear to be most vulnerable to both cancer and diabetes),^17,18 and family history. Modifiable risk factors include lower socioeconomic status, obesity, and alcohol consumption. These common risk factors lead to hyperinsulinemia and insulin resistance, changes in mitochondrial function, low-grade inflammation, and oxidative stress,³ which promote both diabetes and cancer. Diabetes therapy may influence several of these processes.

Several classes of diabetes drugs, including exogenous insulin,^19–22 insulin secretagogues,^23–25 and incretin-based therapies,^26–28 have been under scrutiny because of their potential influences on cancer development in a population already at risk (Table 1).

INSULIN ANALOGUES: MIXED EVIDENCE

Insulin promotes cell division by binding to insulin receptor isoform A and insulin-like growth factor 1 receptors.²⁹ Because endogenous hyperinsulinemia has been linked to cancer risk, growth, and proliferation, some speculate that exogenous insulin may also increase cancer risk.

In 2009, a retrospective study by Hemkens et al linked the long-acting insulin analogue glargine to risk of cancer.¹⁹ This finding set off a tumult of controversy within the medical community and concern among patients. Several limitations of the study were brought to light, including a short duration of follow-up, and several other studies have refuted the study’s findings.^20,21

More recently, the Outcome Reduction With Initial Glargine Intervention (ORIGIN) trial²² found no higher cancer risk with glargine use than with placebo. This study enrolled 12,537 participants from 573 sites in 40 countries. Specifically, risks with glargine use were as follows:

• Any cancer—hazard ratio 1.00, 95% confidence interval (CI) 0.88–1.13, P = .97
• Cancer death—hazard ratio 0.94, 95% CI 0.77–1.15, P = .52.

However, the study was designed to assess cardiovascular outcomes, not cancer risk. Furthermore, the participants were not typical of patients seen in clinical practice: their insulin doses were lower (the median insulin dose was 0.4 units/kg/day by year 6, whereas in clinical practice, those with type 2 diabetes mellitus often use more than 1 unit/kg/day, depending on duration of diabetes, diet, and exercise regimen), and their baseline median hemoglobin A_1c level was only 6.4%. And one may argue that the median follow-up of 6.2 years was too short for cancer to develop.²²

In vitro studies indicate that long-acting analogue insulin therapy may promote cancer cell growth more than endogenous insulin,³⁰ but epidemiologic data have not unequivocally substantiated this.^20–22 There is no clear evidence that analogue insulin therapy raises cancer risk above that of human recombinant insulin, and starting insulin therapy should not be delayed because of concerns about cancer risk, particularly in uncontrolled diabetes.

INSULIN SECRETAGOGUES

Sulfonylureas: Higher risk

Before 1995, only two classes of diabetes drugs were approved by the US Food and Drug Administration (FDA)—insulin and sulfonylureas.

Sulfonylureas lower blood sugar levels by binding to sulfonylurea receptors and inhibiting adenosine triphosphate-dependent potassium channels. The resulting change in resting potential causes an influx of calcium, ultimately leading to insulin secretion.

Sulfonylureas are effective, and because of their low cost, physicians often pick them as a second-line agent after metformin.

The main disadvantage of sulfonylureas is the risk of hypoglycemia, particularly in patients with renal failure, the elderly, and diabetic patients who are unaware of when they are hypoglycemic. Other potential drawbacks are that they impair cardiac ischemic preconditioning³¹ and possibly increase cancer risk.^21,32 (Ischemic preconditioning is the process in which transient episodes of ischemia “condition” the myocardium so that it better withstands future episodes with minimal anginal pain and tissue injury.³³) Of the sulfonylureas, glyburide has been most implicated in cardiovascular risk.³²

In a retrospective cohort study of 62,809 patients from a general-practice database in the United Kingdom, Currie et al²¹ found that sulfonylurea monotherapy was associated with a 36% higher risk of cancer (95% CI 1.19–1.54, P < .001) than metformin monotherapy. Prescribing bias may have influenced the results: practitioners are more likely to prescribe sulfonylureas to leaner patients, who have a greater likelihood of occult cancer. However, other studies also found that the cancer death rate is higher in those who take a sulfonylurea alone than in those who use metformin alone.^23,24

Some evidence indicates that long-acting sulfonylurea formulations (eg, glyburide) likely hold the most danger, certainly in regard to hypoglycemia, but it is less clear if this translates to cancer concerns.³¹

Meglitinides: Limited evidence

Meglitinides, the other class of insulin secretagogues, are less commonly used but are similar to sulfonylureas in the way they increase endogenous insulin levels. The data are limited regarding cancer risk and meglitinide therapy, but the magnitude of the association is similar to that with sulfonylurea therapy.²⁵

INSULIN SENSITIZERS

There are currently two classes of insulin sensitizers: biguanides and thiazolidinediones (TZDs, also known as glitazones). These drugs show less risk of both cancer incidence and cancer death than insulin secretagogues such as sulfonylureas.^21,23,24 In fact, they may decrease cancer potential by alteration of signaling via the AKT/mTOR (v-akt murine thymoma viral oncogene homolog 1/mammalian target of rapamycin) pathway.³⁴

Metformin, a biguanide, is the oral drug of choice

Metformin is the only biguanide currently available in the United States. It was approved by the FDA in 1995, although it had been in clinical use since the 1950s. Inexpensive and familiar, it is the oral antihyperglycemic of choice if there are no contraindications to it, such as renal dysfunction (creatinine ≥ 1.4 mg/dL in women and ≥ 1.5 mg/dL in men), acute decompensated heart failure, or pulmonary or hepatic insufficiency, all of which may lead to an increased risk of lactic acidosis.¹

Metformin lowers blood sugar levels primarily by inhibiting hepatic glucose production (gluconeogenesis) and by improving peripheral insulin sensitivity. It directly activates AMP-activated protein kinase (AMPK), which affects insulin signaling and glucose and fat metabolism.³⁵ It may exert further beneficial effects by acutely increasing glucagon-like peptide-1 (GLP-1) levels and inducing islet incretin-receptor gene expression.³⁶ Although the exact mechanisms have not been fully elucidated, metformin’s insulin-sensitizing properties are likely from favorable effects on insulin receptor expression, tyrosine kinase activity, and influences on the incretin pathway.^36,37 These effects also mitigate carcinogenesis, both directly (via AMPK and liver kinase B1, a tumor-suppressor gene) and indirectly (via reduction of hyperinsulinemia).³⁵

Overall, biguanide therapy is associated with a lower cancer incidence or, at worst, no effect on cancer incidence. In vitro studies demonstrate that metformin both suppresses cancer cell growth and induces apoptosis, resulting in fewer live cancer cells.³⁴ Several retrospective studies found lower cancer risk in metformin users than in patients receiving antidiabetes drugs other than insulin-sensitizing agents,^{21,23,25,38–40} while others have shown no effect.⁴¹ Use of metformin was specifically associated with lower risk of cancers of the liver, colon and rectum, and lung.⁴² Further, metformin users have a lower cancer mortality rate than nonusers.^24,43

Thiazolidinediones

TZDs, such as pioglitazone, work by binding to peroxisome proliferator-activated gamma receptors in the cell nucleus, altering gene transcription.⁴⁴ They reduce insulin resistance and levels of endogenous insulin levels and free fatty acids.⁴⁴

Concern over bladder cancer risk with TZD use, particularly with pioglitazone, has increased in the last few years, as various cohort studies found a statistically significant increased risk with this agent.⁴⁴ The risk appears to rise with cumulative dose.^45,46

Randomized controlled trials also found an increased risk of bladder cancer with TZD therapy, although the difference was not statistically significant.^47–49 In a mean follow-up of 8.7 years, the Prospective Pioglitazone Clinical Trial in Macrovascular Events reported 23 cases of bladder cancer in the pioglitazone group vs 22 cases in the placebo group, for rates of 0.9% vs 0.8% (relative risk [RR] 1.06, 95% CI 0.59–1.89).⁴⁹

On the other hand, the risk of cancer of the breast, colon, and lung has been found to be lower with TZD use.⁴⁷ In vitro studies support the clinical data, showing that TZDs inhibit growth of human cancer cells derived from cancers of the lung, colon, breast, stomach, ovary, and prostate.^50–53

Home et al⁵⁴ compared rosiglitazone against a sulfonylurea in patients already taking metformin in the Rosiglitazone Evaluated for Cardiovascular Outcomes in Oral Agent Combination Therapy for Type 2 Diabetes (RECORD) trial. Malignancies developed in 6.7% of the sulfonylurea group compared with 5.1% of the rosiglitazone group, for a hazard ratio of 1.33 (95% CI 0.94–1.88).

Both ADOPT (A Diabetes Outcome Progression Trial) and the RECORD trial found rosiglitazone comparable to metformin in terms of cancer risk.⁵⁴

Colmers et al⁴⁷ pooled data from four randomized controlled trials, seven cohort studies, and nine case-control studies to assess the risk of cancer with TZD use in type 2 diabetes. Both the randomized and observational data showed neutral overall cancer risk with TZDs. However, pooled data from observational studies showed significantly lower risk with TZD use in terms of:

• Colorectal cancer RR 0.93, 95% CI 0.87–1.00
• Lung cancer RR 0.91, 95% CI 0.84–0.98
• Breast cancer RR 0.89, 95% CI 0.81–0.98.

INCRETIN-BASED THERAPIES

Incretins are hormones released from the gut in response to food ingestion, triggering release of insulin before blood glucose levels rise. Their action explains why insulin secretion increases more after an oral glucose load than after an intravenous glucose load, a phenomenon called the incretin effect.⁵⁵

There are two incretin hormones: glucose-dependent insulinotropic polypeptide (GIP) and glucagon-like peptide 1 (GLP-1). They have short a half-life because they are rapidly degraded by dipeptidyl peptidase-IV (DPP-IV).⁵⁵ Available incretin-based therapies are GLP-1 receptor agonists and DPP-IV inhibitors.

When used as monotherapy, incretin-based therapies do not cause hypoglycemia because their effect is glucose-dependent.⁵⁵ GLP-1 receptor antagonists have the added benefit of inducing weight loss, but DPP-IV inhibitors are considered to be weight-neutral.

GLP-1 receptor agonists

Exenatide, the first of the GLP-1 receptor agonists, was approved in 2005. The original formulation (Byetta) is taken by injection twice daily, and timing in conjunction with food intake is important: it should be taken within 60 minutes before the morning and evening meals. Extended-release exenatide (Bydureon) is a once-weekly formulation taken without regard to timing of food intake. Exenatide (either twice-daily Byetta or once-weekly Bydureon) should not be used in those with creatinine clearance less than 30 mL/min or end-stage renal disease and should be used with caution in patients with renal transplantation.

Liraglutide (Victoza), a once-daily formulation, can be injected irrespective of food intake. The dose does not have to be adjusted for renal function, although it should be used with caution in those with renal impairment, including end-stage renal disease. Approval for a 3-mg formulation is pending with the FDA as a weight-loss drug on the basis of promising results in a randomized phase 3 trial.⁵⁶

Albiglutide (Tanzeum), a once-weekly GLP-1 receptor antagonist, was recently approved by the FDA.

DPP-IV inhibitors

Whereas GLP-1 receptor agonists are injected, the DPP-IV inhibitors have the advantage of being oral agents.

Sitagliptin (Januvia), the first DPP-IV inhibitor, became available in the United States in 2006. Since then, three more have become available: saxagliptin (Onglyza), linagliptin (Tradjenta), and alogliptin (Nesina).

Concerns about thyroid cancer with incretin drugs

Concerns of increased risk of cancer, particularly of the thyroid and pancreas, have been raised since GLP-1 receptor agonists and DPP-IV inhibitors became available.

Studies in rodents have shown C-cell hyperplasia, sometimes resulting in increased incidence of thyroid carcinoma, and dose-dependent rises in serum calcitonin, particularly with liraglutide.²⁶ This has raised concern about an increased risk of medullary thyroid carcinoma in humans. However, the density of C cells in rodents is up to 45 times greater than in humans, and C cells also express functional GLP-1 receptors.²⁶

Gier et al²⁷ assessed the expression of calcitonin and human GLP-1 receptors in normal C cells, C cell hyperplasia, and medullary cancer. In this study, calcitonin and GLP-1 receptor were co-expressed in medullary thyroid cancer (10 of 12 cases) and C-cell hyperplasia (9 of 9 cases) more commonly than in normal C cells (5 of 15 cases). Further, GLP-1 receptor was expressed in 3 of 17 cases of papillary thyroid cancer.

Calcitonin, a polypeptide hormone produced by thyroid C cells and used as a medullary thyroid cancer biomarker, was increased in a slightly higher percentage of patients treated with liraglutide than in controls, without an increase above the normal range.⁵⁷

A meta-analysis by Alves et al⁵⁸ of 25 studies found that neither exenatide (no cases reported) nor liraglutide (odds ratio 1.54, 95% CI 0.40–6.02) was associated with increased thyroid cancer risk.

MacConell et al⁵⁹ pooled the results of 19 placebo-controlled trials of twice-daily exenatide and found a thyroid cancer incidence rate of 0.3 per 100 patient-years (< 0.1%) vs 0 per 100 patient-years in pooled comparators.

Concerns about pancreatic cancer with incretin drugs

Increased risk of acute pancreatitis is a potential side effect of both DPP-IV inhibitors and GLP-1 receptor agonists and has led to speculation that this translates to an increased risk of pancreatic cancer.

In a point-counterpoint debate, Butler et al²⁸ argued that incretin-based medications have questionable safety, with increased rates of pancreatitis possibly leading to pancreatic cancer. In counterpoint, Nauck⁶⁰ argued that the risk of pancreatitis or cancer is extremely low, and clinical cases are unsubstantiated.

Bailey⁶¹ outlined the complexities and difficulties in drawing firm conclusions from individual clinical trials regarding possible adverse effects of diabetes drugs. The trials are typically designed to assess hemoglobin A_1c reduction at varying doses and are typically restricted in patient selection, patient numbers, and drug-exposure duration, which may introduce allocation and ascertainment biases. The attempt to draw firm conclusions from such trials can be problematic and can lead to increased alarm, warranted or not.

Type 2 diabetes mellitus itself is associated with an increased incidence of pancreatic cancer, and whether incretin therapy enhances this risk is still controversial. Whether more episodes of acute pancreatitis without chronic pancreatitis can be extrapolated to an increased incidence of pancreatic cancer is doubtful. A normal pancreatic duct cell may take up to 12 years to become a tumor cell from which pancreatic carcinoma develops, another 7 years to develop metastatic capacity, and another 3 years before a diagnosis is made from clinical symptoms (which are usually accompanied by metastases).⁶²

The risks and benefits of incretin therapies remain a contentious issue, and there are no clear prospective data at this time on increased pancreatic cancer incidence. Long-term prospective studies designed to analyze these specific outcomes (pancreatitis, pancreatic cancer, and medullary thyroid cancer) need to be undertaken.⁶³

OTHER DIABETES THERAPIES

Alpha glucosidase inhibitors

Oral glucosidase inhibitors ameliorate hyperglycemia by inhibiting alpha glucosidase enzymes in the brush border of the small intestines, preventing conversion of polysaccharides to monosaccharides.⁶⁴ This slows digestion of carbohydrates and glucose release into the bloodstream and blunts the postprandial hyperglycemic excursion.

The two alpha glucosidase inhibitors currently available in the United States are acarbose and miglitol, and although data are limited, they do not appear to increase the risk of cancer.^65,66

Sodium-glucose-linked cotransporter 2 inhibitors

The newest class of oral diabetes agents to be approved are the sodium-glucose-linked cotransporter 2 (SGLT2) inhibitors canagliflozin (Invokana) and dapagliflozin (Farxiga).

SGLT2 is a protein in the S1 segment of the proximal renal tubules responsible for over 90% of renal glucose reabsorption. SGLT2 inhibitors lower serum glucose levels by promoting glycosuria and have also been shown to have favorable effects on blood pressure and weight.^67,68

Canagliflozin was the first of its class to gain FDA approval in the United States. It has not been found to be associated with increased cancer risk.⁶⁸

Dapagliflozin, originally approved in Europe, was approved in the United States on January 8, 2014. Because of a possible increased incidence of breast and bladder malignancies, the FDA advisory committee initially recommended against approval and required further data. In those who were treated, nine cases of bladder cancer and nine cases of breast cancer were reported, compared with one case of bladder cancer and no cases of breast cancer in the control group; however, the difference was not statistically significant.⁶⁸

Since SGLT2 inhibitors are still new, data on long-term outcomes are lacking. Early clinical data do not show a significant increase in cancer risk.

WHAT THIS MEANS IN PRACTICE

Many studies have found associations between diabetes, obesity, hyperinsulinemia, and cancer risk. In the last decade, concerns implicating antihyperglycemic agents in cancer development have arisen but have not been well substantiated. At this time, there are no definitive prospective data indicating that the currently available type 2 diabetes therapies increase the incidence of cancer beyond the inherent increased risk in this population. What, then, is one to do?

Educate. Lifestyle modification, including weight management, should continue to be emphasized in diabetes education, as no therapy is completely effective without adjunct modifications in diet and physical activity. Epidemiologic studies have shown the benefits of lifestyle modifications, which ameliorate many of the adverse metabolic conditions that coexist in type 2 diabetes and cancer.

Screen for cancer. Given the associations between diabetes and malignancy, cancer screening is especially important in this high-risk population.

Customize therapy to individual patients. Those with a personal history of bladder cancer should avoid pioglitazone, and those who have had pancreatic cancer should avoid sitagliptin until definitive clinical data become available.

Moreover, patients with a personal or family history of medullary thyroid cancer should not receive GLP-1 receptor agonists. These agents should also probably be avoided in patients with a personal history of differentiated thyroid carcinoma or a history of familial nonmedullary thyroid carcinoma. Until we have further elucidating data, it is not possible to say whether a family history of any of the other types of cancer should represent a contraindication to the use of any of these agents.

Discuss. The multitude of diabetes therapies warrants physician-patient discussions that carefully weigh the risks and benefits of additional agents to optimize glycemic control and metabolic factors in individual patients.