, Dilip R. Karnad2 and Snehal Kothari3
(1)
Cardiac Safety Services Quintiles, Durham, North Carolina, USA
(2)
Research Team, Cardiac Safety Services Quintiles, Mumbai, India
(3)
Cardiac Safety Services Global Head, Cardiac Safety Center of Excellence Quintiles, Mumbai, India
A meta-analysis is always considered less convincing than a large prospective trial designed to assess the outcome of interest (Nissen and Wolski 2007).
12.1 Introduction
The noncommunicable disease of diabetes mellitus has reached epidemic proportions, and it knows no geographic boundaries (Turner 2014). Diabetes, a chronic disorder characterized by elevations in both basal and postprandial (post-meal) glucose levels, is associated with a two- to fourfold increased risk of cardiovascular disease and a threefold increased risk of mortality (Meigs 2003). There are two major forms of diabetes, type 1 and type 2. The typical age of onset of type 1 diabetes is in childhood. This form is also known as insulin-dependent diabetes since patients do not produce enough insulin for healthy function: multiple daily injections of insulin (i.e., exogenous insulin) are required to maintain life, and strict dietary rules, planned physical activity, and daily home glucose tests are also necessary.
The characterization of type 2 diabetes has changed over recent years. Previously used terms such as non-insulin-dependent diabetes and adult-onset diabetes are no longer appropriate: some individuals with type 2 diabetes do require insulin, and children now have this form of the disease. It is characterized by decreased insulin secretion and insulin resistance, a condition in which insulin is not able to carry out its functions effectively, i.e., it cannot decrease plasma glucose levels via suppression of hepatic glucose and stimulation of glucose use in skeletal muscle and adipose tissue.
As one example from a country in the Western world, the US National Diabetes Statistics Report, 2014, observed that in 2012, 29.1 million Americans (9.3 %) had diabetes and that 86 million individuals aged 20 years and older had prediabetes: the respective figures for just 2 years earlier were 25.8 million (8.3 %) and 79 million (CDC 2014). Type 1 diabetes has long been a disease observed in pediatric populations, but, as noted a few moments ago, the same is now true for type 2 diabetes. The prevalence of pediatric type 1 diabetes in the USA increased from 1.48 to 1.93 per 1000 from 2001 to 2009: when adjusted for completeness of ascertainment, these figures represent a 21.1 % increase (Dabelea et al. 2014). Although it is true that the absolute number of pediatric patients with type 2 diabetes is less than for type 1 diabetes, the rate of increase is greater: the estimated US prevalence increased from 0.34 to 0.46 per 1000 across the same time span, representing a 30.5 % increase (Dabelea et al. 2014).
Now consider a regional example outside the Western world. Close to 20 % of all adults in the world with diabetes live in the Southeast Asia region. As the worldwide figure is estimated to approach 600 million by 2035, the number of people with diabetes in Southeast Asia is estimated to increase to 123 million (International Diabetes Federation 2013). It is extremely concerning from a global public health perspective that most people with diabetes live in low- and middle-income countries; these countries will see the greatest increase over the next two decades (Whiting et al. 2011). The life expectancy of a patient with type 2 diabetes is likely to be reduced by up to 10 years as a result of this condition (Diabetes in the UK Report 2010), a dramatic statistic driven to a large extent by increased risk of heart disease, renal disease, and stroke. Quality as well as quantity of life can also be seriously affected given the additional burdens of nervous system damage, blindness, and lower-limb amputation.
When first-line interventions addressing a patient’s dietary and exercise habits have failed to prevent progression to diabetes, initiation of pharmaceutical regimens becomes necessary (an appropriate diet and appropriate levels of exercise should certainly be maintained as adjunctive therapy). At the time of writing this chapter, there are 12 classes of antidiabetic drugs approved in the USA for adults with type 2 diabetes: insulins, biguanides, second-generation sulfonylureas, glinides, thiazolidinediones, α-glucosidase inhibitors, DPP-4 inhibitors, GLP-1 agonists, SGLT2 inhibitors, bile acid sequestrants, dopamine agonists, and amylin analogues. While it may initially appear that this array of currently approved drugs comprises a robust list, the continuing medical need for additional pharmaceutical agents is well captured by the European Medicines Agency’s (EMA’s) guideline addressing clinical investigation of medicinal products in the treatment or prevention of diabetes, which comments as follows (EMA 2012):
That is, a patient’s drug regimen becomes insufficient over several years, meaning that an additional drug will be added. When this happens, the drugs the patient is already taking are often kept as they still have a beneficial effect, but their effect alone is not great enough to counteract disease progression. A constant provision of new drugs is therefore needed. Accordingly, it is encouraging to report that new drugs within various classes continue to be developed and that there are multiple new classes of drugs being developed, including PPAR agonists/modulators (PPAR-α agonists, PPAR-δ agonists, PPAR-α/δ agonists, PPAR-δ/γ agonists, PPAR-α/γ co-agonists, and PPAR-α/δ/γ pan-agonists), glucokinase activators, C-C chemokine receptor type 2 antagonists, IL-1 modulators, G-protein-coupled receptor agonists, apical sodium-dependent bile acid transporter inhibitors, and 11-beta-HSD1 inhibitors (see Mittermayer et al. 2015).
Glucose control in type 2 diabetes deteriorates progressively over time, and, after failure of diet and exercise alone, needs on average a new intervention with glucose-lowering agents every 3–4 years in order to obtain/retain good control.
12.1.1 Prospective Exclusion of Unacceptable Cardiovascular Risk for New Antidiabetic Drugs for Type 2 Diabetes
There are multiple aspects of bringing a new antidiabetic drug for type 2 diabetes to market, including safety and efficacy investigations. Of specific interest in this and the following chapter is one component of the overall safety evaluation required by regulators, namely, the prospective exclusion of unacceptable cardiovascular risk associated with the drug. This chapter provides an overview of the genesis of the regulatory landscapes in the USA (FDA 2008) and Europe (EMA 2012) addressing this issue, which were formalized in 2008 and 2012, respectively. Chapter 13 then reviews the requirements of these regulatory landscapes and provides examples of the exoneration of drugs from an unacceptable cardiovascular risk.
12.2 Publication of a Meta-analysis Involving Rosiglitazone
Rosiglitazone is a member of the thiazolidinedione drug class and was approved for marketing by the FDA in 1999 and the EMA in 2000. On May 21, 2007, the New England Journal of Medicine e-published a paper entitled “Effect of rosiglitazone on the risk of myocardial infarction and death from cardiovascular causes,” which was published in the journal in print format on June 14, 2007 (Nissen and Wolski 2007). One result presented in the paper was an odds ratio for myocardial infarction in the rosiglitazone group compared with the control group of 1.43 (95 % CI, 1.03–1.98, p = 0.03). This result can be interpreted as follows:
The fact that the two-sided 95 % confidence interval excludes zero informs us that the result is statistically significant at the 5 % level, and the authors appropriately provided the exact p-value as well. This result received considerable attention not only in clinical circles but also in government and media settings.
The result of this meta-analysis is compatible with an increase in risk of myocardial infarction of as little as 3 % and as great at 98 %, and the best estimate is an increase of 43 %.
12.2.1 Details and Critique of the Meta-analysis
In preparing for their meta-analysis, the meta-analysts screened 116 trials, deeming that 48 met their predefined inclusion criteria of employing a randomized comparator group, a similar duration of treatment in all groups, and more than 24 weeks of drug exposure. Six of the 48 trials did not report myocardial infarctions (or deaths from cardiovascular causes), and therefore their results could not be included in the form of analysis chosen by the authors. Hence, summary statistics from 42 trials were included in this study-level meta-analysis: the combined number of participants was almost 28,000.
Given that the risks of myocardial infarction are low, the odds ratio can be interpreted as a relative risk, which facilitates the following statement: compared with the control treatment group, treatment with rosiglitazone was associated with 1.03–1.98 times the risk of a myocardial infarction. Leaving aside discussion of the analytical methodology chosen for a moment, it is informative to examine the absolute number of events (myocardial infarctions) that went into the analysis. In the rosiglitazone group (totaling approximately 15,600 participants), there were 86 events, and in the comparator group (totaling approximately 12,300 participants), there were 72 events. Making the reasonable assumption that the results for the control group reflect the general background incidence of myocardial infarction in individuals with type 2 diabetes not treated with rosiglitazone, the increase in absolute number of events, 14, is small (Turner and Durham 2009).
Another point that can informatively be made here is the usefulness of presenting statements (best estimates) of absolute risk along with statements of relative risk. Consider the “worst-case scenario” value, in this case provided by the upper limit of the confidence interval presented, i.e., 1.98. Rounding this value up to 2.00 for present purposes, it is meaningful to consider several hypothetical sets of numbers where, in all cases, the risk of some specified event doubles. A given risk could increase as follows: from 1 in 10 to 2 in 10; from 1 in 100 to 2 in 100; from 1 in 1,000 to 2 in 1,000; from 1 in 100,000 to 2 in 1000,000; 1 in a million to 2 in a million; and so on. While the increase in relative risk is mathematically identical in all scenarios, addition of information concerning absolute risk dramatically influences one’s perception of the acceptability of a doubling in risk when using that information to make a decision. It is therefore appropriate to communicate a risk by using both relative and absolute statements.
Returning now to the meta-analysis of interest, limitations have been provided by various authors. The meta-analysts themselves noted the following points (Nissen and Wolski 2007):
Moreover, they also noted that “a meta-analysis is always considered less convincing than a large prospective trial designed to assess the outcome of interest.”Pignone’s (2007) commentary on this meta-analysis indicated some of the limitations noted by the authors themselves and also discussed other limitations, including the following:
The meta-analysts did not have access to the original source data for any of the trials included in the analysis (i.e., they did not have access to participant-level data).
The meta-analysis pooled the results of a group of trials that were not originally intended to explore cardiovascular outcomes.
Many of the included trials were small and short term, resulting in few adverse cardiovascular events or deaths. Accordingly, since the results of the analysis were based on a relatively small number of events, the odds ratio “could be affected by small changes in the classification of events.”
Directly related to the previous point, most of the included trials did not utilize centralized adjudication of cardiovascular outcomes (recall discussions in Sect. 6.5), and the definitions of myocardial infarction were not available.
The confidence interval is wide, “resulting in considerable uncertainty about the magnitude of the observed hazard.”
The lack of participant-level data limits assessment of the role of important covariates such as age and sex, as well as precluding time-to-event analysis.
A fixed-effects model was used to combine information from the studies included in the analysis.
With regard to the last bullet point, Pignone provided additional commentary (Pignone 2007):
Diamond and colleagues (2007) expressed this view:
A fixed-effects model assumes that all of the trials included drew their participants from the same underlying patient pool, an assumption that is difficult to support. It would have been more appropriate to use a random-effects model, which accounts for both within- and between-study variability. Fixed-effects models usually produce narrower confidence intervals for their summary estimates, which leads to overestimation of the precision of the data.
The meta-analysis was not based on a comprehensive search for all studies that might yield evidence about rosiglitazone’s cardiovascular effects. Studies were combined on the basis of a lack of statistical heterogeneity, despite substantial variability in study design and outcome assessment. The meta-analytic approach that was used required the exclusion of studies with zero events in the treatment and control groups. Alternative meta-analytic approaches that use continuity corrections show lower odds ratios that are not statistically significant. We conclude that the risk for diabetic patients taking rosiglitazone is uncertain: Neither increased nor decreased risk is established.
12.3 FDA Advisory Committee Meetings Following the Publication of the Meta-analysis and Their Consequences
Reactions to the publication of the meta-analysis led to the FDA convening a joint meeting of its Endocrinologic and Metabolic Drugs Advisory Committee and its Drug Safety and Risk Management Committee on July 30, 2007, to discuss the cardiovascular ischemic and thrombotic risk of the thiazolidinediones, with a focus on rosiglitazone. (One of this book’s authors [JRT] was an invited speaker at the Open Public Hearing session at this meeting: his testimony was that the meta-analysis had been so poorly conducted that no decision-making weight should be afforded to it.) Representatives from rosiglitazone’s sponsor and the FDA testified at the meeting. Representatives from the sponsor of another thiazolidinedione, pioglitazone, were invited to attend the meeting, but they did not actively participate.
Two votes were taken by members of the advisory committees. First, they voted 20–3 that rosiglitazone increased the cardiac risk in patients with type 2 diabetes, although many members “made statements accompanying their votes that drew a distinction between the risk as compared with placebo and the risk as compared with other antidiabetic drugs” (Krall 2007). Second, they voted 22–1 that rosiglitazone should not be removed from the market and hence should remain available to patients.
![](https://freepngimg.com/download/social_media/63059-media-icons-telegram-twitter-blog-computer-social.png)
Stay updated, free articles. Join our Telegram channel
![](https://clinicalpub.com/wp-content/uploads/2023/09/256.png)
Full access? Get Clinical Tree
![](https://videdental.com/wp-content/uploads/2023/09/appstore.png)
![](https://videdental.com/wp-content/uploads/2023/09/google-play.png)