Thiazolidinediones: the Forgotten Diabetes Medications

For years there was an intense debate as to whether the primary defect in type 2 diabetes (T2D) was insulin resistance or insulin secretory deficiency. This conflict has been resolved in that hyperglycemia only occurs when insulin secretion is insufficient to overcome barriers to insulin action [1•]. Thus, when insulin action is normal, hyperglycemia occurs when absolute insulin secretion is deficient. In contrast, when insulin action is impaired (insulin resistance), hyperglycemia occurs when insulin secretion is inadequate to overcome the insulin resistance. While a few investigators suggest that hyperinsulinemia itself may be a primary cause of insulin resistance, the overwhelming data indicate that insulin resistance is a primary defect due to abnormalities in the distribution of adipose tissue products in hepatic and extrahepatic tissues [2••, 3••, 4]. These ectopic increases in lipids define the metabolic syndrome. The metabolic syndrome results in increases in atherosclerotic diseases, hepatic steatosis and steato-hepatitis, and hypertension [5, 6••]. Insulin resistance increases the prevalence of T2D sixfold as marginal insulin secretion becomes inadequate insulin secretion [7, 8].

Patients with T2D may have hyperglycemia and no metabolic syndrome (up to 15% of patients with T2D) or hyperglycemia with the metabolic syndrome (85% of patients with T2D) [9, 10••]. Data from NHANES III participants 50 years or older showed that patients with T2D and no metabolic syndrome had a prevalence of coronary heart disease of 7.5% which did not differ from a non-diabetic population with no metabolic syndrome (prevalence 8.7%) [10••]. Patients with no diabetes and the metabolic syndrome had a 13.9% prevalence of coronary heart disease and those with diabetes and the metabolic syndrome had a 19.2% prevalence of coronary heart disease [10••]. These and other data suggest that the metabolic syndrome is the driving factor in the development of cardiovascular disease and that hyperglycemia is a factor that exaggerates cardiovascular disease in the metabolic syndrome setting.

This formulation of insulin-resistant T2D as two interacting pathophysiologic entities (Fig. 1) suggests that appropriate treatment should target both the insulin resistance and the inadequate insulin secretion. Therapeutic agents such as insulin, sulfonylureas, and meglitinides target primarily insulin availability [11]. Glucagon-like peptide 1 (GLP-1) receptor agonists and dipeptidyl dipeptidase 4 (DDP-4) inhibitors increase insulin secretion and decrease hyperglucagonemia, thereby lowering hyperglycemia [12‐14]. GLP-1 receptor agonists have modest effects in decreasing body weight and lowering systolic blood pressure [12, 13]. The sodium-glucose transport 2 (SGLT-2) inhibitors lower blood glucose through an increase in renal excretion of glucose with a secondary benefit of decreasing glucose toxicity [15]. They increase hepatic glucose production, glucagon secretion, ketogenesis, and lipid oxidation [15, 16]. A significant side effect of their treatment is a modest weight loss, a decrease in blood pressure, and a modest reduction in insulin resistance [15, 16]. Weight loss itself decreases hepatic triglycerides, peripheral and visceral adipose tissue mass, and plasma triglycerides [17•]. GLP-1 receptor agonists and SGLT-2 inhibitors secondarily decrease insulin resistance in patients in proportion to their effect in promoting weight loss. Thiazolidinediones (TZDs) decrease insulin resistance directly through activation of PPARγ receptors which facilitate differentiation of mesenchymal stem cells into adipocytes, promote lipogenesis in peripheral adipocytes, decrease hepatic and peripheral triglycerides, decrease activity of visceral adipocytes, and increase adiponectin [18••, 19, 20]. These primary effects of TZDs markedly ameliorate insulin resistance and the metabolic syndrome and decrease insulin requirements [18••, 19, 20]. While metformin decreases hepatic insulin resistance, it has little or no primary effect on peripheral insulin resistance [21, 22]. Metformin’s effects on hepatic insulin resistance are likely to be due to its inhibition of mitochondrial complex 1 resulting in defective cyclic AMP signaling [22].

The metabolic consequences of insulin resistance are an increase in the incidence of T2D and an increase in cardiovascular diseases [7, 8]. Insulin resistance decreases pancreatic beta-cell function by causing compensatory hyperinsulinemia, beta-cell lipotoxicity, and increasing islet inflammation [22, 23]. These factors result in increased beta-cell apoptosis and progressive decreases in insulin secretory capacity. Insulin resistance is associated with increases in multiple cardiovascular risk factors [2••, 3••, 4, 5, 6•]. Insulin resistance creates a dyslipidemia consisting of increases in plasma-free fatty acids and triglycerides, decreases in HDL-cholesterol, decreases in adiponectin, and increases in small dense LDL particles [2••, 6•, 18••, 24••]. Insulin resistance is a major cause of hepatic steatosis and steato-hepatitis (NASH) [24••]. Insulin resistance creates a procoagulant state with increases in fibrinogen and plasminogen activator inhibitor-1 (PAI-1) [5, 6•]. Insulin resistance increases inflammation and is associated with an increase in C-reactive protein (CRP), IL-6, and TNFα [25, 26]. Insulin resistance is associated with endothelial dysfunction as measured by decreased vasodilatation, increase in adhesion molecules, and an increase in cellular proliferation [5, 27]. An increase in hepatic and visceral fat leads to ectopic deposits of lipids in tissues such as skeletal muscle, myocardial muscle, and endothelial cells, causing a decrease response to insulin [2••, 3••, 24••]. These ectopic deposits of intracellular lipids within muscle and endothelial cells increase serine and threonine phosphorylation and decrease tyrosine phosphorylation of insulin receptor substrates and phospho-inositol 3-kinase, thereby decreasing intracellular insulin signaling [2••, 24••]. An increase in beta-cell lipids decreases its insulin secretory function [23, 28].

Significant weight loss in obese insulin-resistant individuals will secondarily reduce insulin resistance and the components of the insulin resistance syndrome [17, 29]. TZDs through their activation of PPARγ have primary effects on adipose tissue and decrease insulin resistance by reducing hepatic triglycerides, decreasing visceral fat mass and activity and increasing subcutaneous fat mass [2••, 18••, 19, 20]. This primary action of TZDs in reducing insulin resistance accounts for their beneficial effects in ameliorating the detrimental effects of insulin resistance. Table 1 describes the beneficial effects of TZDs in the treatment of patients with T2D [2••, 18••, 20, 30, 33]. TZDs decrease both fasting and postprandial hyperglycemia by decreasing insulin resistance and allowing endogenous insulin to be more effective. Where measured, TZDs have a more durable effect in reducing hyperglycemia than other antihyperglycemic agents [34]. The multiple effects in decreasing dyslipidemia, endothelial dysfunction, inflammation, and the procoagulant state and increasing adiponectin are thought to provide some potential cardiovascular benefits. The effects of TZDs in reducing hepatic steatosis suggest potential therapeutic benefits in reducing NASH. The reduction in insulin resistance in patients with prediabetes decreases their rate of progression to T2D, presumably by preserving beta-cell function [31•, 32•]. Between 2000 and 2008, pioglitazone and rosiglitazone were among the most widely prescribed antidiabetic medications. However, there were increasing concerns about TZDs’ side effects: increased fluid retention; increased incidence of heart failure; weight gain; and increased peripheral fractures [35•, 36••]. Regulatory concerns about cardiovascular safety issues with rosiglitazone and increases in bladder cancer with pioglitazone markedly reduced the availability of these drugs [37‐39]. Since the regulatory issues appear to have been greatly exaggerated and TZDs are the only agents that primarily target insulin resistance, the use of these relatively inexpensive drugs justifies a re-evaluation of their clinical use.

Since TZD treatment improves insulin resistance and decreases many cardiovascular risk factors, it had been proposed that treatment of patients at high risk for cardiovascular events with TZDs might have a beneficial effect in reducing long-term cardiovascular events. Clinical studies have examined the effects of pioglitazone and rosiglitazone on cardiovascular outcomes, and several meta-analyses and large database analyses have compared their relative cardiovascular outcomes.

Weight gain has been associated with both rosiglitazone and pioglitazone treatments [35•, 40••]. The weight gain appears to be dose-related (1 or 1.5 kg at low doses) and is greater during combination therapy with insulin secretagogues (2 to 3 kg) and remarkably so when TZDs are combined with insulin therapy (3.5 to 6 kg). The weight gain is attributable to several factors: increase in subcutaneous fat mass with either no change or a small decrease in visceral fat mass, fluid retention, and positive calorie balance because of improved glycemic control. Though the increase in subcutaneous fat mass is distressing to patients with T2D who are trying to lose weight, the TZD-mediated changes in fat mass distribution is related to the improvement in insulin resistance and glycemic control.

During the pre-marketing clinical trials of pioglitazone and rosiglitazone, it was noted that an increase in peripheral edema and the development of congestive heart failure occurred in a few patients. Patients with NYHA functional class III and IV heart failure had been excluded from the clinical studies. Shortly after rosiglitazone and pioglitazone were marketed for clinical use, data appeared which indicated that edema and congestive heart failure were significant complications of TZD treatment of patients with diabetes mellitus. The magnitude of the problem was sufficiently important that the American Diabetes Association and the American Heart Association held a Consensus Conference to discuss and publish a Consensus Statement on “Thiazolidinedione Use, Fluid Retention, and Congestive Heart Failure” [35•].

The consensus derived from meta-analyses and large databases was that monotherapy with either pioglitazone or rosiglitazone is associated with a 3 to 5% incidence of peripheral edema which increases to 7.5 to 8% when they are combined with other antidiabetic agents such as sulfonylureas [35•]. When TZDs are combined with insulin therapy, the incidence of peripheral edema is approximately 15 to 16% as compared with insulin alone which is 5 to 7% [35•].

Initially, it was thought that activation of PPARγ receptors in the renal collecting duct epithelium’s sodium channel (ENaC) was responsible for TZD-related peripheral edema, hemodilution, and even macular edema [62]. It now appears that sodium transporters in the proximal tubule including NHE3 also contribute to sodium reabsorption [63]. Increased vascular permeability through increased vascular endothelial growth factor secretion and decreased systemic vascular resistance are non-kidney factors that contribute to the edema [63].

Table 2 presents data on TZD-mediated edema from randomized, controlled clinical trials and a large French Pharmacovigilance data base. The incidence of TZD-induced edema depends on drug dose, the stage of glucose intolerance, and the magnitude of the underlying cardiovascular disease. The lowest incidence is in prediabetes (DREAM study), with intermediate incidence in T2D (ADOPT and French PharmacoVigilance studies), and the highest in those with advanced cardiovascular disease (PROactive and IRIS studies) [31•, 34, 64••, 65••, 66•]. The extent to which TZD use exacerbates fluid retention in congestive heart failure is unclear [67].

Heart failure is the most significant side effect of treatment with TZDs. Many meta-analyses and database analyses have been published and have concluded that both pioglitazone and rosiglitazone increase the incidence of heart failure in patients with T2D [68‐71]. However, they cannot provide data on the severity, progression, and clinical outcome of the patients. Table 3 presents the available data on TZD treatment and the development of heart failure in randomized, controlled clinical trials. Data are available in patients with insulin-resistant and no diabetes, prediabetes, ordinary T2D, and T2D with previous cardiovascular events. Patients with prediabetes treated with TZDs have a very low incidence of heart failure (< 1%) [31•]. Patients with T2D, similarly, had a very low incidence of heart failure (< 1%) [34]. Patients without diabetes but with insulin resistance and a previous ischemic cerebrovascular event had an increase in the incidence of heart failure and hospitalization for heart failure, but these were not related to pioglitazone treatment [65]. VA ambulatory patients with heart failure and diabetes had an increase in heart failure hospitalizations and death over a 2-year follow-up period and they were unrelated to TZD treatment [72••]. Three studies (PROactive, RECORD, and A GSK-sponsored ECHOCARDIOGRAPHIC study) showed significant increases in heart failure hospitalizations in patients treated with a TZD [43••, 61, 64••]. The populations in these three studies had in common a long duration of diabetes and a high prevalence of preceding cardiovascular events. While death from heart failure was increased by rosiglitazone in the RECORD study, death was not increased by pioglitazone in PROactive. It appears that TZDs increase edema and heart failure requiring additional treatment or hospitalization in patients with T2D with significant preceding cardiovascular disease. The available data indicate that pioglitazone treatment does not increase death from heart failure.

Rosiglitazone treatment of patients with T2D and NYHA functional class I or II heart failure for 52 weeks did not affect LVEF even though the patients experienced new or worsening edema and increased congestive heart failure medications [61].

The second major complication of TZD therapy is an increase in the risk of bone fractures. This was first reported in the results of the ADOPT study which was a long-term study to determine the durability of rosiglitazone as a treatment for glycemic control as compared with metformin or glyburide. In this study involving 2511 men and 1840 women, there was an increase in peripheral fractures in women but not men [36••]. After a median treatment of 4.0 years, 60 women (9.3% or 2.74/100 patient years) had at least one fracture. The cumulative incidence of fractures in women at 5 years of treatment was rosiglitazone 15.1%, metformin 7.3%, and glyburide 7.7%. The HR for rosiglitazone-induced fractures was 1.81 and 2.11 relative to metformin or glyburide [36••]. The increase in fractures started after 1 year of treatment. A retrospective analysis of the data from the PROactive study found similar results with an increase in fracture risk in women (5.1% vs. 2.5% or 1.0 vs. 0.5/100 person-years) but not in men (0.6 vs. 0.7/100 person-years) compared with the control population [73].

Two other large randomized, controlled clinical trials reported an increase in fracture risk in TZD-treated patients with T2D. In the RECORD trial which randomized 4447 participants to rosiglitazone or combination metformin and sulfonylurea for a mean of 5.5 years, fracture risk for women was increased by rosiglitazone exposure (2.1/100 person-years compared with the active controls (1.1/100 person-years) but not in men (1.0/100 person-years vs. active controls 0.8/100 person-years) [43••]. In the IRIS trial of 3876 insulin-resistant subjects without diabetes, the 5-year pioglitazone fracture risk was 13.6% compared with the placebo control of 8.8% with a HR of 1.53 [74•]. The fractures were due to a fall in 80%, and 45% were serious and required hospitalization or surgery. The fracture risk from pioglitazone exposure was increased in both men (9.4% vs. active control 5.2%, HR = 1.83) and women (14.9% vs. active control 11.6%, HR = 1.32, p = 0.13). The actual fracture rates were 2.3 vs. 1.3/100 person-years for men and 3.7 vs. 2.8/100 person-years for women. The rates in these IRIS participants were higher than in people with diabetes because of the high incidence of falls as these were subjects who had had a previous cerebral ischemic lesion [74•].

In the recently published data from the TOSCA-IT study comparing pioglitazone treatment to an active sulfonylurea therapy in 3028 participants with T2D patients for 57.3 months, there was no significant increases in fractures in pioglitazone exposed female or male participants [42].

Other data on TZD fracture risk come primarily from large medical databases or subanalyses from randomized controlled studies designed for other purposes. In the ACCORD study of intensive vs. ordinary glycemic control, 74% of participants received rosiglitazone and 17% received pioglitazone. During a mean follow-up of 4.8 years, 262 men and 287 women had at least one non-spinal fracture [75•]. There was no increase in fracture risk in men receiving a TZD. Fracture rate in women with 1 to 2 years or > 2 years TZD exposure had HR of 2.32 and 2.01, respectively. Discontinuing TZD use for 1 to 2 years or > 2 years reduced the HR to 0.57 and 0.42 compared with current TZD users [75•].

Database studies confirm the increased risk of fractures in both pre- and postmenopausal TZD-exposed women. Fractures occur starting at 1 year of exposure are usually peripheral and are likely dose- and time-related. Though a few studies find some increase fracture risk in men exposed to TZDs, most do not.

The specter that pioglitazone increases bladder cancer has been an issue for regulators and clinicians for more than a decade. A mismatch in the number of cases of bladder cancer was noted in the PROactive study (14 in the pioglitazone arm and 6 in the placebo arm) [40••]. An epidemiologic study of 1.5 million persons with diabetes followed for 4 years (2002–2006) by the French National Health Insurance Agency reported that pioglitazone-treated males had a statistically significant increase in bladder cancer (HR = 1.22, 95% CI = 1.03–1.43) compared with patients treated with other antidiabetic agents [39]. The risk was increased with a total cumulative dose > 28,000 mg (HR = 1.75) and for exposures > 1 year (HR = 1.34). The French Agency for the Safety of Health Products withdrew pioglitazone from the French market on June 9, 2011, and Germany’s Federal Institute for Drugs and Medical Devices on June 10, 2011, advised physicians not to prescribe pioglitazone until the issue of pioglitazone and bladder cancer was resolved.

A 10-year follow-up study of pioglitazone use and bladder cancer was agreed upon by the Kaiser Permanente Group of California and the US FDA. The study involved 193,099 persons aged 40 years and older in 1997–2002 who were diagnosed with diabetes and managed by the Kaiser health care system and were followed until December 2012 [76••]. Of the cohort, 34,181 (18%) received pioglitazone (mean 2.8 (0.2 to 13.2) years) and 1261 had incident bladder cancer. Ever use of pioglitazone was not associated with the development of bladder cancer (HR = 1.06; 95% CI = 0.89–1.26). In a comparison of 464 bladder cancer patients with 464 matched controls, the adjusted OR for pioglitazone use vs. no use was 1.18 (95% CI = 0.78–1.80) [76••]. Several additional studies support the findings that pioglitazone does not increase bladder cancer [77‐79].

There have been no randomized, controlled clinical trials comparing the efficacy and safety of rosiglitazone and pioglitazone. The TIDE study which was designed to obtain this information was discontinued after 162 days because of the FDA’s concern about the cardiovascular safety of rosiglitazone and the ensuing public apprehension which would have made recruitment difficult and biased [56]. The only comparative data available are database analyses and meta-analyses.

There are basic studies which do raise the possibility that there may be some differences in the clinical effects of rosiglitazone and pioglitazone. PPARs represent a complicated family of nuclear receptors: PPARα, PPARβδ, and PPARγ [80, 81]. PPARα primarily facilitates fatty acid oxidation and lipid utilization, while PPARγ promotes differentiation of stem cells into adipocytes and facilitates peripheral adipose tissue cells to store rather than release fatty acids. Furthermore, there are three distinct isoforms of PPARγ:PPARγ1 (expressed in almost all tissues), PPARγ2 (expressed mainly in adipose tissue), and PPARγ3 (expressed in macrophages, colon and white adipose tissue). Pioglitazone is a selective human PPARγ1 and a weak human PPARα activator [82]. Chemical proteomics–based analysis profiles show that while there are many common products for pioglitazone and rosiglitazone; there are also many unique products for each [83]. One proven difference between the two TZDs is their effect on human lipid profiles. Rosiglitazone has a less beneficial effect on plasma lipids than pioglitazone: greater increase in VLDL particles; increased LDL particle concentration vs. decrease in LDL particle concentration with pioglitazone; decreased HDL particle concentration and size while pioglitazone increased them [84].

Therefore, it would not be surprising to see differences in the clinical effects of these two TZDs. Unfortunately, as noted above, there are no randomized controlled trials comparing the two. There are database analyses and meta-analyses that conclude that pioglitazone has a greater benefit on cardiovascular outcomes and a lesser side effect profile than rosiglitazone. All these analyses are questionable because they are influenced by selection bias, and the studies that were carried out with rosiglitazone as compared with pioglitazone were suboptimal. For example, PROactive provided much more meaningful data than RECORD. The controversy surrounding rosiglitazone starting in 2007 would be expected to have influenced patient antidiabetic treatment selection and maintenance of treatment in all database analyses. Major unknowns with relevance to side effects and efficacy include the adjustments made in TZD doses and the effects of other drugs, such as diuretics, used to treat fluid retention and heart failure. The development of edema and heart failure appears to be influenced by the underlying degree of vascular disease. In Tables 2 and 3, patients with insulin resistance and prediabetes had very low incidence of edema and heart failure followed by the highest incidence in those who had very significant cardiovascular disease at entry into studies.

The TZD class of drugs uniquely treats insulin resistance which is the major cause of the increase in T2D and a large component of the increase in atherosclerotic cardiovascular disease. The positive benefits accrued from TZD treatment are summarized in Table 1. The use of these previously extensively prescribed and effective treatments for T2D has been remarkably diminished because of concerns about safety and side effects such as heart failure and increased risk of fractures. Data have been presented to dispel the safety concerns and put the complications into a rational perspective. These drugs are generic, cost is not overly excessive, and they are the most effective pharmacologic agents for treating insulin resistance.