Diabetes Mellitus: A Fundamental and Clinical Text
3rd Edition

The paradigm for this mechanism in humans is the toxic effect of the rodenticide N-3 pyridyl methyl N8 p-nitrophenyl urea (PNU, Vacor) (5,6,7). Accidental or deliberate ingestion of this chemical causes rapid death of the insulin-producing islet β-cells. There are two phases to this process: first, insulin release from dying cells, which can be accompanied by hypoglycemia; and second, absolute insulin deficiency, which is accompanied by diabetes or diabetic ketoacidosis. This chemical appears to act by antagonizing the nucleotide niacinamide, which is an essential enzymatic cofactor. Early treatment with niacinamide is believed to be protective, although there is no case report that documents the efficacy of this therapy (5,7).

Tacrolimus is a macrolide that has become the cornerstone of many immunosuppressive regimens used for organ transplantation. Although its immunologic actions are similar to those of cyclosporine, it appears that tacrolimus induces diabetes by limiting β-cell insulin production, whereas cyclosporine acts at both insulin secretion and insulin sensitivity. Tacrolimus is commonly used in regimens with steroids, and it is this combination that is associated with the incidence of diabetes. In a recent analysis of pediatric renal transplant recipients, the risk for post-transplantation diabetes was ninefold higher in tacrolimus-based regimens than in those based on cyclosporine (8). The
overall incidence of new-onset diabetes in renal transplant recipients receiving tacrolimus is 10% to 20%. Likewise, the incidence of diabetes in heart transplant recipients taking tacrolimus was 19.6 % (9). In pediatric recipients of heart-lung transplants, and in adult liver transplant patients, the incidence of new-onset diabetes was over 40% (10,11). Although some studies show that a first-degree family history of type 2 diabetes is a major risk factor, most studies have not found features to predict the development of diabetes, and most patients who develop diabetes will require insulin therapy.

Pentamidine is an antiparasitic agent frequently used to treat infections with Pneumocystis carinii in patients with acquired immune deficiency syndrome (AIDS) who are intolerant of trimethoprim-sulfa regimens. The mesylate isethionate and aerosolized forms have been reported to induce a dose- and duration-dependent cytolytic injury to the β-cell (12,13,14,15,16,17,18). Up to 26% of patients treated with this agent experience hypoglycemia within 1 week due to acute cytolytic release of insulin. Diabetic ketoacidosis has been reported (18). Following one or several courses of pentamidine therapy, hyperglycemia develops within 2 to 6 months in up to 19% of patients, most of whom have already experienced the hypoglycemic phase. Diabetes has not been reported in patients with AIDS treated with other antiparasitic agents, such as cotrimazole. The mechanism of injury is not known, although the histologic picture in the islet cells resembles that following PNU (Vacor) ingestion (10). Although diabetes may be temporary after one course of pentamidine, permanent diabetes is more common, especially after repeated courses of therapy.

Didanosine (DDI) is a nucleoside analogue used in the treatment of human immunodeficiency virus (HIV) infection and AIDS. Pancreatitis is an infrequent but serious adverse reaction to this medication, and DDI-induced pancreatitis can be accompanied
by new diabetes due to β-cell injury (19). Another possible mechanism is inhibition of insulin release due to hypokalemia, which can be associated with this therapy. There are over 80 reported cases.

The antineoplastic agent l-asparaginase causes diabetes without injuring β-cells (20,21,22). L-asparaginase is an enzyme that metabolizes asparagine and limits the availability of this essential nutrient to leukemic cells. There are several case reports of diabetes and diabetic ketoacidosis in nondiabetic children treated with L-asparaginase. In these cases, there were very low levels of circulating insulin. L-asparaginase does not directly metabolize the insulin molecule (22). Instead, low levels of asparagine may inhibit insulin production, since the insulin molecule contains three asparagine residues. Consistent with this hypothesis, diabetes usually resolves when treatment is withdrawn or completed (20).

Antagonists to β-adrenergic receptors are commonly used medications known to impair insulin secretion, especially agents that are not selective for the β₁-receptor subtype. β-receptor blockade inhibits insulin secretion by pancreatic islets in response to glucagon, glucose, or argenine. Several studies have linked chronic use of β-blockers with an increased risk for the development of diabetes. In two studies of men treated for hypertension, a community-based health survey in England over a 9-year span (23) and a 12-year follow-up of a cohort of Swedish men (24), there was a relative risk of 6 to 6.1 compared with nonhypertensive controls. Most recently, a review of the use of antihypertensives in the Atherosclerosis Risk in Communities (ARIC) study in 3,804 subjects indicated that there was a relative risk of 2.43 for the development of diabetes in those with hypertension over a 6-year period, and a multivariate analysis indicated the the risk for diabetes was 28% greater in those using a β-blocker than those using other medications (25). The risk for diabetes reported in most studies of β-receptor antagonists exceeds the known twofold increase in the risk for diabetes found in hypertensive populations (26).

Propranolol therapy induced small increases in fasting glucose levels in 687 men treated in a Veterans Administration study for 48 months, from 99.6 to 106 mg/dL (27). In a 10-year study in Sweden, patients treated with hydrochlorothiazide plus propranolol showed an increase of 0.56 mM (10 mg/dL) compared with 0.18 mM (3.24 mg/dL) in patients treated with hydrochlorothiazide alone (28). These data suggest that propranolol therapy may shift individuals from the glucose intolerance to the frank diabetes portion of the curve shown in Fig. 59.1. Because many factors, including age, weight, family history, and hypertension, play a role in the risk for diabetes, the true risk can only be assessed when these variables are controlled through a randomized study design. In a randomized trial comparing hydrochlorothiazide and propranolol (a nonselective β-blocker), oral glucose tolerance testing at 0,1, and 6 years did not show significant effect from β-blockade (29). However, more and larger randomized trials will be needed to assess the true attributable risk of diabetes from β-blockade.

Diphenylhydantoin is a commonly used anticonvulsant that is known to inhibit insulin secretion (30). In a randomized trial, 46 patients were treated with placebo or diphenylhydantoin for 3 years following myocardial infarction (31). At the end of the study, the two groups did not differ significantly from each other or from their respective baseline levels when compared using oral glucose tolerance tests. However, the diphenylhydantoin group showed a small decrease in insulin secretion during the test, suggesting a compensatory increase in insulin sensitivity. Because there are numerous case reports of diphenylhydantoin-induced diabetes, it is likely that individuals with other risk factors for diabetes such as glucose intolerance and a limited ability to increase insulin sensitivity may develop diabetes with this medication. Diabetes usually resolves once the medication is discontinued.

Diazoxide is a vasodilator commonly used in the past for control of malignant hypertension. Parenteral diazoxide therapy inhibits insulin secretion and has been used in the treatment of insulinoma. There are reports of diabetic ketoacidosis in nondiabetic patients treated with multiple intravenous doses of diazoxide (32,33). The mechanism of its action on insulin secretion is not known but is reversible upon discontinuing the drug.

Glucocorticoids such as hydrocortisone, dexamethasone, prednisone, and methylprednisolone may induce diabetes. These drugs are used in a wide variety of disorders and in a wide range of doses. The actual incidence of diabetes induced by these agents is unknown because of these variations and because the most powerful influences on the risk for steroid-induced diabetes are likely to be the underlying metabolic and nonmetabolic disorders of the patient (34). This is underscored in data presented in a study from 1954, in which Fajans and Conn (35) proposed a new combined cortisone glucose tolerance test to identify those at risk for diabetes. In the report, 50 to 62.5 mg of oral cortisone (12–15 mg of prednisone) were administered 8.5 and 2 hours before a glucose challenge. Of 37 individuals with normal glucose tolerance, 1 developed a diabetic degree of glucose intolerance after the cortisone, an incidence of 2.7%. However, when 75 individuals with normal glucose tolerance but also a family history of diabetes were tested, 24% had a diabetic response to the glucose load. Thus, the presence of an
asymptomatic underlying genetic risk or metabolic disorder (i.e., glucose intolerance) increased the risk for acute steroid-induced diabetes 10-fold. However, the use of corticosteroids as replacement therapy has not been shown to change the risk of the development of diabetes. In patients with hypopituitarism taking standard (30 mg/day of hydrocortisone) replacement corticosteroids, 96% had normal glucose tolerance to a 75-g oral glucose load (36).

Glucocorticoid-induced diabetes can be detected within hours of administration of the steroid (35). Usually, there is improvement in glucose tolerance with continued steroid use and a pathophysiologic state consistent with insulin resistance (increased endogenous insulin secretion) (37,38,39,40). The exact pathophysiology is controversial, because several mechanisms are operating simultaneously (34). Glucocorticoids encourage breakdown of stored protein and fat stores, which causes an increased stream of free fatty acids and branched amino acids to the liver (40). Steroids also induce increased cellular concentrations of gluconeogenic enzymes. The result of increased amounts of substrate for gluconeogenesis and increased amounts of the hepatic enzymatic machinery for gluconeogenesis is increased hepatic glucose output. Hepatic glucose output is usually regulated by insulin, but the effect of insulin is diminished in the presence of steroids. Glucose uptake by fat and muscle is reduced due to insulin resistance and direct steroid effects. Glucocorticoid therapy is a challenge to endogenous insulin secretion, and those who have limited reserves may become diabetic.

Megestrol acetate is a progestin steroid used to stimulate appetite and weight gain in cachexia related to cancer and AIDS. There are two case reports of new-onset diabetes in patients with AIDS who were taking 80 mg of megestrol four times a day (41,42). In one report, diabetes resolved when megestrol was discontinued but recurred upon rechallenge (42). The mechanism has not been studied but is probably a combination of steroid-induced decreased sensitivity to insulin and increased caloric intake.

Oral contraceptives are steroid combinations that are known to increase average glucose concentrations in patients with and without diabetes by decreasing insulin sensitivity. However, in large epidemiologic studies, there is little evidence to link the use of modern low-dose estrogen or triphasic oral contraceptives and diabetes. In the Nurses Health Study of 121,700 women over more than 15 years, current oral contraceptive use did not increase the relative risk (RR) for diabetes (RR = 0.86), and past use conferred a small increase (RR = 1.12) that was not related to dose or duration of exposure (43). Kjos et al. (44) screened 349 women over the age of 35 for metabolic abnormalities associated with oral contraceptives. In the older group, 183 women who used contraceptives did not demonstrate a greater risk for diabetes than the 63 nonusers. Gabal et al. (45) analyzed data from 848 women 50 to 70 years of age who were taking postmenopausal estrogen replacement therapy. The researchers did not find an increased risk for diabetes in users compared with controls. It is prudent to monitor women who have already manifested a tendency for diabetes in the presence of increased sex steroids, such as those with a history of gestational diabetes.

Ironically, both β-antagonists and β-agonists have been implicated in causing diabetes, albeit through different mechanisms. β-agonists mimic the effects of adrenergic members of the counterregulatory hormone system response to hypoglycemia. They cause insulin resistance, diminished glucose utilization, and increased glucose production. Stimulation of β-adrenergic receptors will increase hepatic glucose output and diminish insulin sensitivity. These agents do not confer an increased risk for diabetes when used topically for pulmonary disease. However, in the setting of pregnancy, in which insulin sensitivity is already reduced, the use of terbutaline to halt premature labor has been reported to result in an incidence of gestational diabetes of 12% to 33% (46,47,48,49). Increased concentrations of blood glucose and insulin have been demonstrated after 5 to 7 days of oral or subcutaneous therapy, suggesting insulin resistance (46). In a study of 91 women with preterm labor who were known to have normal glucose tolerance before terbutaline therapy, 11% of those treated with 30 mg/day of oral terbutaline developed gestational diabetes (47). A control cohort of 634 similar women had an incidence of 6%. Interestingly, women treated with subcutaneous terbutaline (<3 mg/day) had an incidence of gestational diabetes of 5%. These data suggest a dose-response relationship. A study by Fisher et al. (48) suggested that the combination of corticosteroids and β-adrenergic agents may be particularly diabetogenic.

Another member of the counterregulatory system with pharmacologic uses is growth hormone (GH), which causes insulin resistance at a cellular site after the binding of insulin to its receptor. Since recombinant GH has become available, there has been increased use of the peptide to treat GH-deficient children and exploration of widening its uses for other causes of short stature, cardiac remodeling, and reduced muscle mass in the elderly. In the natural example of GH excess, acromegaly, diabetes is a prominent problem. In a study of over 23,000 children treated with GH therapy, 11 had developed type 1 diabetes, 18 type 2 diabetes, and 14 had impaired glucose tolerance (50). In this analysis, it was found that there was a sixfold excess in the incidence of type 2 diabetes, but no increase in the incidence of type 1 diabetes for this age population (50). Termination of GH therapy did not result in resolution of type 2 diabetes. As GH therapy is expanded to the elderly and adult populations, it is likely to unmask latent diabetes in these populations as well. It is likely that GH therapy will cause diabetes in susceptible individuals, and the number of reported cases will increase if GH therapy is expanded to other patient groups.

Protease inhibitors are a class of highly effective antiretrovirus agents used in the treatment of human immunodeficiency virus (HIV) infection and AIDS. They may be given in combination
with other diabetes-inducing drugs such as megestrol acetate or pentamidine. Metabolic side effects include centripetal obesity, hypertriglyceridemia, and diabetes (51). There have been many reports of new-onset diabetes in HIV-infected patients treated with these agents (52,53,54). The incidence of new diabetes is estimated to be 1% to 7% (54). Although there are reports of presentation with ketoacidosis that suggested a mechanism of acquired insulin deficiency (52,53), most studies have indicated that these agents induce an insulin resistance state and may even exacerbate an insulin resistance state associated with the viral infection (54). In most studies, C-peptide secretion is elevated as a marker of increased endogenous insulin secretion (55). One group has suggested that protease inhibitors may reduce the conversion of proinsulin to insulin as a novel mechanism to cause insulin deficiency, since proinsulin is much less potent than insulin (56). Acidosis may occur independently of diabetes, due to a propensity toward lactic acidosis in patients treated with antiretroviral therapy, and this propensity has limited the use of metformin in the treatment of protease inhibitor–related diabetes.

Thiazide diuretics, a commonly prescribed class of agents for control of hypertension, are often cited as causes of drug-induced diabetes. Many small uncontrolled trials show an increased incidence of glucose intolerance in patients with hypertension treated with thiazides; in two studies, up to 22% of patients treated for 6 years had diabetes (57,58). Bengtsson et al. (24) studied a population-based cohort of 1,462 women over a 10-year period. The relative risk for an abnormal glucose tolerance test at the end of the study for individuals with normal glucose tolerance at the start was 3.4 to 4.6 in those who were taking thiazide diuretics. Others have not replicated these results; in a case control study, Gurwitz et al. (59) found that the risk for developing diabetes was not increased by treatment with thiazides alone but was increased by treatment with more than one antihypertensive agent. The ARIC study (25) analysis also failed to show an association of the use of thiazide agents and the increased risk for diabetes in patients with hypertension. However, the strongest evidence would be that arising from a randomized intervention trial, since hypertension itself increases the risk for developing diabetes by twofold or more (26). Helgeland et al. (28) randomly assigned 785 men to either 50 mg of hydrochlorothiazide or placebo for 5 years, and did not demonstrate a difference in fasting glucose values between the two groups at the beginning or end of the study. The European Working Group on Hypertension in the Elderly randomly assigned 348 patients to 25 mg hydrochlorothiazide and 50 mg triampterene or placebo for up to 3 years (60). There was a large dropout rate in the study group that may have biased the results. Although the blood glucose levels before and after challenge with glucose were increased by 13.2 and 30.2 mg/dL, respectively, there were no new cases of diabetes noted in the treated group. Finally, two studies that compared thiazide therapy to propranolol therapy did not show an increased risk for diabetes after 6 years or 48 weeks of therapy in one regimen over the other (26,28).

Acute administration of thiazide diuretics has been shown to cause a 27% decrease in endogenous insulin response to a hyperglycemic clamp protocol (61). Gorden (62) had noted glucose intolerance and diminished serum insulin in association with potassium-depleted states, and Heldren et al. (63) repeated the hyperglycemic clamp studies to show that the majority of the defect could be corrected by careful potassium repletion. However, other data have shown diminished insulin sensitivity after thiazide treatment and suggest that thiazide effects on carbohydrate metabolism may be more complex. In two studies, one a randomized double-blind study, thiazide therapy increased plasma insulin and decreased index of insulin sensitivity over a 12-week treatment period (61,64).

In summary, thiazide diuretics are epidemiologically linked to an increased incidence of diabetes in hypertensive patients, but these data have not been supported by randomized controlled trials. Potassium depletion due to thiazide diuretics exacerbates the problem. It is likely that new diabetes induced by thiazides is uncommon.

Cyclosporine is a fungal metabolite used as an immunosuppressant to prevent the rejection of transplanted organs and to interfere with autoimmune processes. Cyclosporine interferes with the activation and proliferation of T cells by preventing the transcription of several genes, including those for interleukin-2 (IL-2) and c-myc. The incidence of diabetes in previously nondiabetic renal transplant recipients who receive cyclosporine has been reported to be from 2% to 46%. However, cyclosporine is usually given in concert with glucocorticoids, which are diabetogenic of themselves and may have an additive or synergistic effect (65).

Ost et al. used intravenous glucose tolerance testing before and after transplantation to find an incidence of new abnormal glucose tolerance in 53% of recipients taking cyclosporine and steroids compared with an incidence of 7% in recipients treated with azathioprine and steroids (65). Koselj et al. (66) reported a retrospective series of 158 renal transplant recipients and found that 13.9% of those treated with cyclosporine and steroids had developed diabetes. Other groups have reported an incidence of up to 30% for recipients on this immunosuppressive combination (67).

The mechanism of cyclosporine-associated diabetes is a combination of inhibition of islet cell function and increased insulin resistance. In an early case report, a renal transplant recipient receiving cyclosporine developed diabetes and a decrease in C-peptide secretion simultaneously (58). When cyclosporine treatment was decreased, C-peptide levels increased. However, Ost et al. (65) reported that there was no change in C-peptide concentrations with cyclosporine treatment in patients who developed diabetes despite a decline of intravenous glucose tolerance,
arguing for insulin resistance. In dog models, evidence for both mechanisms operating simultaneously has been demonstrated (68).

There has been a large pharmacologic revolution in psychiatric practice. The newer antipsychotic agents—clozapine, olanzapine, resperidone, and quetiapine—have many advantages over the conventional phenothiazines. However, their use is associated with significant weight gain, diabetes, and ketoacidosis. There have been several large retrospective reviews. In the United Kingdom, Koro et al. (69) found 451 incident cases of diabetes in 19,637 patients diagnosed and treated for schizophrenia from 1987 to 2000. The use of atypical antipsychotics (olanzapine and resperidone) were associated with a 1.6-fold (resperidone) to 4.2-fold (olanzapine) increased risk for diabetes over patients treated with conventional agents. In the United States, a similar review of 38,632 patients with schizophrenia in the Veterans Administration health system showed a significant increased risk for diabetes associated with the use of clozapine, quetiapine, olanzapine, and resperidone compared with conventional agents (70). In a review of the adverse event reports to the U.S. Food and Drug Administration from 1990 to 2001, there were 384 reports of hyperglycemia associated with clozapine. Within these were 80 cases of ketoacidosis (71). Although weight gain leading to increased insulin resistance has been postulated to be the primary mechanism for inducing diabetes with the use of these drugs, the development of ketoacidosis and occasionally weight loss makes it likely that inhibition of insulin secretion is also occurring in some patients (72).

Nicotinic acid is an effective therapy for dyslipidemias of low-density lipoproteins and very-low-density lipoproteins. Nicotinic acid therapy is associated with increased levels of blood glucose in both diabetic and nondiabetic patients, and uncontrolled hyperglycemia is a frequent reason to discontinue therapy. Henkin et al. (73) reviewed 82 patients treated with nicotinic acid, including 17 heart transplant recipients. In the transplant recipients who had not previously had diabetes, there was a 33% incidence of new diabetes while on nicotinic acid. In the nontransplanted patients, the incidence of new-onset diabetes was 15%. The two groups differed in that the transplant patients were taking additional diabetogenic agents, such as steroids and cyclosporine, and the mean dosage of nicotinic acid in the transplant patients (2.5 ± 0.4 g/day) was nearly twice that of the nontransplant patients.

The mechanism for nicotinic acid–induced hyperglycemia is an increase in hepatic glucose output due to enhanced gluconeogenesis (73). Actually, acute nicotinic acid administration results in a diminished flow of free fatty acids (FFAs) to the liver and diminished gluconeogenesis. However, the effects of nicotinic acid are short lived, and a rebound increase in FFAs, by 50% to 100% over baseline, occurs (74). This increased FFA flux to the liver results in increased oxidation of FFA by the liver and consumption of available cellular nicotinamide adenine dinucleotide (NAD). Ultimately, this decreased hepatocyte capacity for oxidation shifts metabolic activity to reductive pathways, such as glucose production. Kahn et al. (75) demonstrated a decline in both the responsiveness to insulin and the insulin sensitivity index in 11 patients treated for 2 weeks with nicotinic acid. Serum insulin levels increased, indicating an insulin-resistant state. Because of this mechanism, the use of long-acting nicotinic acid derivatives may avoid this hyperglycemic adverse effect.

Total parenteral nutrition in the intensive care and nonintensive care setting is frequently associated with significant elevations in blood glucose concentrations. There are often inflammatory conditions or drugs that impair carbohydrate metabolism, such as steroids, in effect concurrently. There are no large studies examining the frequency of diabetes induced by this therapy, which involves administration of high concentrations of glucose and FFAs, but the physician must monitor these patients for the development of diabetes. These patients strongly suggest interference with insulin secretion.