Diabetes Drug Therapy -- First Do No Harm

I n the November/December 2006 issue of JMCP, Watkins et al. reported the estimated value of exenatide in treating obese patients with type 2 diabetes over a 30-year time horizon, as calculated using the Center for Outcomes Research Diabetic Model, a validated Markov model that predicts clinical and economic end points over time horizons up to 100 years in duration. 1 Adopting the premise that " sophisticated disease-based economic models can be substituted " when " long-term outcomes data are lacking, " the authors sought to provide guidance to managed care organizations making formulary decisions about new products. In fact, development of the model, a collaborative effort between exenatide's manufacturer and a 1.6 million member health plan with an active formulary review process, began several months before approval of the product by the U.S. Food and Drug Administration (FDA). The model's findings for exenatide, including incremental cost-effectiveness ratios of $13,000 per quality-adjusted life-year compared with insulin and $32,000 compared with generic glyburide, were used as a basis for the health plan's decision to cover exenatide after a step-therapy trial of metformin. Two key assumptions used in this pharmacoeconomic analysis are noteworthy. First was an assumed 8.5% (about 15-20 pound) weight loss associated with exenatide treatment in a hypothetical patient population with mean body mass index (ratio of weight in kilograms to height in meters squared [kg/m 2 ]) of 35 kg per m 2 , assumed to result in mean reductions of 10 mm Hg in sys-tolic blood pressure, 20 mg per dL in low-density lipoprotein cholesterol, 59 mg per dL in triglycerides, and corresponding reductions in medical costs to treat cardiovascular events. These assumptions were extrapolated liberally from the approximate 3% weight loss observed in the 30-week clinical trials that were cited in the product label and the approximate 5% weight loss observed in the open-label follow-up for 314 (57.0%) of 551 subjects randomized to exenatide who completed 82 weeks of therapy. Second and more important, the study by Watkins et al. failed to include the cost to treat adverse drug events in the pharmacoeconomic model, in effect assuming that these were $0 or equal across treatment arms, but did acknowledge that more patients discontinued therapy with exenatide compared with placebo. As often happens with pharmacoeconomic modeling performed early in a product's history, the actual data from real-world use ultimately do not support and potentially could even contradict the …

I n the November/December 2006 issue of JMCP, Watkins et al. reported the estimated value of exenatide in treating obese patients with type 2 diabetes over a 30-year time horizon, as calculated using the Center for Outcomes Research Diabetic Model, a validated Markov model that predicts clinical and economic end points over time horizons up to 100 years in duration. 1 Adopting the premise that "sophisticated disease-based economic models can be substituted" when "long-term outcomes data are lacking," the authors sought to provide guidance to managed care organizations making formulary decisions about new products. In fact, development of the model, a collaborative effort between exenatide's manufacturer and a 1.6 million member health plan with an active formulary review process, began several months before approval of the product by the U.S. Food and Drug Administration (FDA). The model's findings for exenatide, including incremental cost-effectiveness ratios of $13,000 per quality-adjusted life-year compared with insulin and $32,000 compared with generic glyburide, were used as a basis for the health plan's decision to cover exenatide after a step-therapy trial of metformin.
Two key assumptions used in this pharmacoeconomic analysis are noteworthy. First was an assumed 8.5% (about 15-20 pound) weight loss associated with exenatide treatment in a hypothetical patient population with mean body mass index (ratio of weight in kilograms to height in meters squared [kg/m 2 ]) of 35 kg per m 2 , assumed to result in mean reductions of 10 mm Hg in systolic blood pressure, 20 mg per dL in low-density lipoprotein cholesterol, 59 mg per dL in triglycerides, and corresponding reductions in medical costs to treat cardiovascular events. These assumptions were extrapolated liberally from the approximate 3% weight loss observed in the 30-week clinical trials that were cited in the product label and the approximate 5% weight loss observed in the open-label follow-up for 314 (57.0%) of 551 subjects randomized to exenatide who completed 82 weeks of therapy. Second and more important, the study by Watkins et al. failed to include the cost to treat adverse drug events in the pharmacoeconomic model, in effect assuming that these were $0 or equal across treatment arms, but did acknowledge that more patients discontinued therapy with exenatide compared with placebo.
As often happens with pharmacoeconomic modeling performed early in a product's history, the actual data from realworld use ultimately do not support and potentially could even contradict the model assumptions. 2 Absent evidence to support the assumption of sustained weight loss with exenatide or what this weight loss might mean in actual cardiovascular event risk reduction, validation of this key assumption is currently an empirical question. Notably, the model's assumption regarding the unimportance of adverse events was quickly found to be fundamentally flawed. Less than a year after the study by Watkins et al. was published, the FDA announced that it had reviewed 30 cases of acute pancreatitis associated with the use of exenatide 3 and announced that the manufacturer had agreed to "update Byetta's label to discuss pancreatitis." 4 Ten months later on August 18, 2008, the FDA updated its warning regarding acute pancreatitis with a notice to health care professionals that 6 cases involved hemorrhagic or necrotizing pancreatitis, all 6 cases required hospitalization, and 2 patients died. 5 The upgraded warning includes a contraindication for the use of exenatide in patients with a history of pancreatitis. This example highlights one of the shortcomings of pharmacoeconomic modeling for new drugs -that efficacy outcomes are more likely to be known than are adverse effects for drugs that are new to market.
In this issue of JMCP, Asche and Nelson comment on a previous review by VanDeKoppel et al. of the place in therapy of 3 new agents for diabetes: exenatide, pramlintide, and sitagliptin. 6 Asche and Nelson call for research with administrative claims and other post-marketing surveillance to help better define the safety and effectiveness of these new diabetes drugs. 7 However, there is already quite a bit known about the adverse effects associated with these drugs, particularly sitagliptin and exenatide.
Valdez in her commentary in this issue of JMCP expresses hope for the superiority of newer diabetes agents in providing some advantage over existing agents, 8 certainly a need given the explosion in the number of cases of type 2 diabetes in the United States and worldwide. 9 However, there is only limited evidence, primarily from animal models, of some potential value of the dipeptidyl peptidase-4 (DPP-4) inhibitors in possibly preserving beta-cell function, the foundation for the hope expressed by inhibitors, 52 were evaluated in detail and 25 of these were rejected as reviews or overviews. Of the 25 studies (reported in 27 publications) that met the inclusion criteria, 11 RCTs involved a total of 6,743 patients randomized to sitagliptin and 14 RCTs involved 6,121 patients randomized to vildagliptin, a drug not yet approved by the FDA. This 134-page systematic review showed that sitagliptin and vildagliptin in meta-analysis of the combined clinical trial results had modest effects on A1c (-0.7% for sitagliptin and -0.6% for vildagliptin). Adverse event outcomes showed a significantly higher incidence of all-cause infections with sitagliptin and a possibly higher incidence with vildagliptin, but comparison with placebo was not statistically significant for vildagliptin. Since the DPP-4 inhibitors may affect immune function, the 2008 Cochrane Review called for additional data on the long-term safety of these drugs.
Letters posted on the FDA website last year mentioned "hypersensitivity post-marketing adverse reaction data," and the label for Januvia (sitagliptin) was revised in October 2007 to include in the "Warnings and Precautions" section the following additional wording, "There have been postmarketing reports of serious allergic and hypersensitivity reactions in patients treated with Januvia such as anaphylaxis, angioedema, and exfoliative skin conditions including Stevens-Johnson syndrome." 11 Dermatologic reactions may be characteristic of the class of DPP-4 drugs since the use of vildagliptin (Galvus) in monkeys has been associated with dermatologic reactions that have reportedly kept vildagliptin off the U.S. market, and in November 2007, dose-related elevations in liver enzymes were discovered with vildagliptin 100 mg once daily, making entry to the U.S. market more unlikely. 12 The Cochrane Review published in April 2008 was preceded by a systematic review and meta-analysis (Amori et al.) in July 2007 of the safety and efficacy of incretin therapy, 13 which encompasses glucagonlike peptide 1 (GLP-1) drugs such as exenatide that mimic incretin, and the DPP-4 inhibitors such as sitagliptin that enhance the effect of incretin. This earlier review included many of the RCTs of the DPP-4 inhibitors that were evaluated in the Cochrane Review that followed in 2008. However, Amori et al. noted an unexplained increased risk of headache with DPP-4 inhibitors and put some context around the concern about the possible adverse effect on immune function -patients with diabetes are more likely to develop urinary tract infections (UTIs) and are at a higher risk of complications, including death from urosepsis. Amori et al. used the example of a seemingly small relative risk of 1.52 for UTIs, which in their estimate potentially increases the number of UTIs by 1 million new cases per year in a population of 20 million persons with type 2 diabetes.
For the GLP-1 drugs, Amori et al. reported that nausea was the most common side effect, and 4% of clinical trial patients who received exenatide withdrew because of nausea and vomiting. Zinman et al. (2007) found in a 16-week trial of add-on therapy (added to thiazolinedione [TZD] with or without metformin) that 16% of the patients who received exenatide versus 2% who received placebo did not complete the clinical trial because of adverse events; 40% of exenatide patients versus 15% of placebo patients experienced nausea, and 13% of exenatide patients versus 1% of placebo patients experienced vomiting. 14 Also, despite only 16 weeks duration, 29% of the exenatide patients and 14% of placebo patients did not complete the study.
Nathan et al. in the 2008 Consensus Algorithm from the American Diabetes Association and the European Association for the Study of Diabetes find no place in therapy for pramlintide, exenatide or sitagliptin in patients with type 2 diabetes. 15 Separately, Nathan criticized the FDA for approving sitagliptin despite the "paucity of published data from long-term clinical trials." 16 The evidence of the risk of adverse events with new GLP-1 and DPP-4 antidiabetic drug therapies combined with 5 risk-warning events in 2007 for the 2 TZDs (rosiglitazone and pioglitazone) 17 and the absence of conclusive evidence of reduction in macrovascular risk with any of the FDA-approved drugs for type 2 diabetes, 18 provide more context for the recommendation by the FDA outside expert panel in July this year to require more evidence than A1c reduction for FDA approval of new drug applications for antidiabetic therapies. 19,20 More evidence of the safety of new drugs for type 2 diabetes is sorely needed.