Chinese Medical Journal 2012;125(23):4301-4306
Risk of cardiovascular disease and all-cause mortality among diabetic patients prescribed rosiglitazone or pioglitazone: a meta-analysis of retrospective cohort studies

CHEN Xin,  YANG Li,  ZHAI Suo-di

CHEN Xin (Department of Pharmacy, Peking University Third Hospital, Beijing 100191, China)

YANG Li (Department of Pharmacy, Peking University Third Hospital, Beijing 100191, China)

ZHAI Suo-di (Department of Pharmacy, Peking University Third Hospital, Beijing 100191, China)

Correspondence to:ZHAI Suo-di,Department of Pharmacy, Peking University Third Hospital, Beijing 100191, China (Tel: 86-10-82266686. Fax:.
cardiovascular risk; meta-analysis; pioglitazone; rosiglitazone; cohort study
Background  The difference of cardiovascular effects between rosiglitazone and pioglitazone treatment for diabetic patients has not been thoroughly studied. We performed a meta-analysis to compare the risk of cardiovascular adverse effects in patients with type 2 diabetes treated with rosiglitazone compared to pioglitazone.
Methods  The Cochrane Library, PubMed, and Embase were searched to identify retrospective cohort studies assessing cardiovascular outcomes with rosiglitazone and pioglitazone. Meta-analysis of retrospective cohort studies was conducted using RevMan 5.0 software to calculate risk ratios.
Results  Of the 74 references identified, eight studies involving 945 286 patients fit the inclusion criteria for the analysis. The results of meta-analyses showed that, compared with pioglitazone, rosiglitazone therapy significantly increased the risk of myocardial infarction (risk ratios (RR) 1.17, 95% confidence interval (CI) 1.04–1.32; P=0.01), the risk of heart failure (RR 1.18, 95% CI 1.02–1.36; P=0.03), and total mortality (RR 1.13, 95% CI 1.08–1.20; P <0.000 01).
Conclusion  Compared with pioglitazone, rosiglitazone was associated with an increased risk of myocardial infarction, heart failure, and all-cause mortality in diabetic patients.
Rosiglitazone and pioglitazone, the only marketed thiazolidinediones (TZDs), are oral hypoglycaemic agents that have been shown to improve glycaemic control and may act to slow the progression of beta cell failure. In May 2007, a meta-analysis of data from 42 randomized controlled clinical trials found an increased risk of myocardial infarction (MI) and death from cardiovascular causes in relation to the use of rosiglitazone.1 However, several meta-analyses did not come to the same conclusion. Moreover, the results of the Rosiglitazone Evaluated for Cardiac Outcomes and Regulation of Glycemia in Diabetes (RECORD) trial, which was once considered to be the strongest evidence evaluating the cardiovascular safety of rosiglitazone, showed no significant increase in these events with rosiglitazone compared to metformin or sulfonylurea in 2009.2 On the other hand, a meta-analysis of 19 randomised controlled trials with pioglitazone found a statistically significant reduction in the composite outcome of nonfatal MI stroke, and all-cause mortality.3 Thus, the cardiovascular risk data for rosiglitazone and benefits of pioglitazone are less than conclusive and an intriguing disparity between the drugs has emerged, suggesting an intraclass variation in TZD effects. The debate about the TZDs effects on cardiovascular health of has gone on for years; especially in the case of rosiglitazone.4-8 This debate is set against a background of a lack of sufficient data and uncertainty about the cardiovascular safety differences between the two types of TZDs, mainly due to limitations in the randomised, controlled trials to date, such as short duration and small sample size.9
In contrast, surveillance data routinely collected through general practice are able to capture information on drugs and events in a wide range of patients as they present for clinical care. That is to say, cohort studies provide data on real-life use, rare outcomes, and long-term effects that are undetectable in randomised controlled trials.10 With large healthcare utilisation databases being used more and more frequently, active surveillance in the form of retrospective cohort studies can provide evidence of risk that is not necessarily provided by randomised controlled trials.11 Therefore, the comparison of cardiovascular events in diabetic patients treated with rosiglitazone and pioglitazone in the same cohort study can illustrate the differential cardiovascular outcomes of these two medications, which may provide strong evidence in the debate on TZD safety.
The cardiovascular risks of rosiglitazone and pioglitazone have been compared with one another in several observational studies. A meta-analysis of these cohort studies in type 2 diabetic patients was conducted to compare the risk of myocardial infarction, heart failure, and all-cause mortality between rosiglitazone and pioglitazone.
Search strategy
The search strategy was designed to identify observational cohort studies conducted in any country, in people of any age, and aimed to compare cardiovascular outcomes of both rosiglitazone and pioglitazone at the same time. A computer-based literature search was conducted up to July 2010 using the Cochrane library, PubMed, and Embase. MeSH terms used to identify articles included “rosiglitazone”, “pioglitazone”, “cohort study”, and “cardiovascular risk”, as well as “myocardial infraction” and “heart failure”. Additionally, we screened the reference lists of selected researches and wrote to authors for further detailed data. Although no language restrictions were applied, the search results produced studies published only in English. The literature search was conducted by Chen and Zhai independently.
Inclusion criteria
All identified articles were reviewed according to inclusion criteria before proceeding with further analysis. The inclusion criteria were: (1) retrospective cohort study in patients with diabetes mellitus; (2) studies that compared rosiglitazone to pioglitazone with or without other classes of oral antidiabetes drugs; (3) parameters used to estimate cardiovascular risks, such as MI, heart failure, or mortality. Studies were excluded if they were not original research, were not available as full papers, did not meet the three parameters mentioned above, or were case-controlled studies.
Data extraction and quality assessment
Each potentially eligible study was independently assessed by two qualified reviewers (Chen and Zhai) to determine whether it met the inclusion criteria and to assess its methodological quality. Conflict between the reviewers was resolved through discussion, and if consensus could not be reached, a third reviewer (Yang) determined the outcome. The two reviewers independently extracted data from each included study using a standardised form. The characteristics of each study were identified and extracted, including: methodology, database, number of patients, intervention, follow-up period, population characteristics, diagnostic criteria of diabetes and outcomes. The outcome data extracted were: number of patients treated with rosiglitazone as well as those treated with pioglitazone, number of events of each group.
The most important quality criteria for studies included: prior cardiovascular outcome researches, reliability of database researches used, comparability of population characteristics between rosiglitazone groups and pioglitazone groups, a follow-up period of at least one year, and a clear definition of diabetes and outcomes.
Statistical analysis
RevMan version 5.0 (Cochrane Collaboration, Copenhagen, Denmark) was used to combine results from more than two separate trials. Statistical heterogeneity among the studies was identified using the χ2 test (P=0.05). Random-effects models were applied when heterogeneity was identified in a group of studies (P ≤0.10); otherwise, fixed-effects models were used. Since the outcomes were dichotomous, risk ratios (RRs) were calculated, and the range of the RRs was expressed as a 95% confidence interval (CI). P <0.05 was considered statistically significant. Subgroup analysis was performed to answer specific questions about particular patient groups or types of interventions.
Study inclusion
The search of PubMed and Embase revealed 18 and 54 researches, respectively, of which 15 were relevant for our study, while there was no relevant research from Cochrane Library. Two studies were excluded from the analysis based on the criterion for case control cohort studies.12,13 One study were excluded because the data of rosiglitazone and pioglitazone respectively were not available.  The other studies were excluded because the original data relating to numbers of events were unpublished. After attempting to contact corresponding authors, three authors provided further data for analysis. Finally, eight studies remained eligible for inclusion (Table 1).14-21 Figure 1 shows search process and reasons for excluding studies.

view in a new window
Table 1. Basic characteristics of included studies

view in a new window

Figure 1. Search process showing reasons for excluding studies.

Study characteristics
Table 2 shows the main characteristics of the studies. Most cohort studies were performed in the USA, one in the UK, one in Canada, and one in Taiwan, China. Three of the studies involved patients older than 65 years.14-16 Another study by Gerrits included patients older than 45 years,17 while the patients of Tzoulaki’s study were between 35 and 90 years old.18 Only one study specifically narrowed its population to newly diagnosed diabetic patients.19 Unpublished data from Pantalone’s study were added to our analysis.20 In all studies, the rosiglitazone arm was compared with the pioglitazone arm. Four studies were head-to-head cohorts of these two TZD drugs.14-17 Four studies excluded patients that received a prescription for insulin during TZD treatment.14,16-18 Moreover, two studies compared rosiglitazone and pioglitazone as monotherapy or in combination, respectively.18,19

view in a new window
Table 2. Patient characteristics of included studies
Data synthesis and sensitivity analysis
Myocardial infraction
A total of seven studies reported numbers of myocardial infraction, and they showed statistical heterogeneity across studies (Q-test P values=0.002); hence, a random effects model was used for analysis. As showed in Figure 2, compared with patients given pioglitazone, patients given rosiglitazone had an increased risk of MI (RR 1.17; 95% CI 1.04–1.32, P=0.010). In the head-to-head studies, the RR was 1.08 (95% CI 1.01–1.15, P=0.030). Considering that the follow-up period of Gerrits’ study was shorter than others, exclusion of this study from the analysis gave an RR of 1.15 (95% CI 1.01–1.32, P=0.040). For the studies that excluded patients treated with insulin, the RR was 1.09 (95% CI 0.98–1.20, P=0.11).

view in a new window
Figure 2. Comparison of myocardial infarction.
Heart failure
Since there was significant heterogeneity among the five studies (P <0.0001, I2=84%) that reported the number of heart failure events, the analysis used a random effects model. The risk of heart failure was increased in the rosiglitazone group compared to the pioglitazone group (RR 1.18, 95% CI 1.02–1.36; P=0.03) (Figure 3). When unpublished data for Pantalone’s study were included, among the six studies, the difference between rosiglitazone and pioglitazone was not significant (RR 1.14, 95% CI 0.98–1.32; P=0.08). Analysis of the four studies that excluded patients on insulin combination therapy found no difference between the two TZDs (RR 1.00, 95% CI 0.77–1.31; P=0.99)

view in a new window
Figure 3. Comparison of risk of heart failure.
All-cause mortality
All-cause mortality was reported in four studies. The RR was 1.13 (95% CI 1.08–1.20; P <0.00001) (Figure 4). After including data provided by the author of one study, the risk of mortality remained significant in the rosiglitazone group for the whole analysis (RR 1.13, 95% CI 1.00–1.19; P <0.00001). Sensitivity analysis of studies that did not involve insulin treatment showed more risk with rosiglitazone (RR 1.16, 95% CI 1.08–1.24; P <0.00001).

view in a new window
Figure 4. Comparison of risk of all-cause mortality
This meta-analysis compared the cardiovascular risks of rosiglitazone and pioglitazone based on a total of eight retrospective cohort studies. The risk of all three outcomes—MI, heart failure and all cause mortality—were higher in the rosiglitazone group than in the pioglitazone group. Moreover, the results of the main analysis involving the whole population was significant, and the results did not change with inclusion of the unpublished data from one study or sensitivity analysis of studies that did not involve insulin treatment, though part of the analysis did not show significant differences. Our findings suggest clear evidence for the cardiovascular harm of rosiglitazone, which may have very important consequences in clinical practice. Recently, the FDA restricted rosiglitazone to a narrowed population, while the EMA has suspended it. The results of our study provide further evidence to support the decision. As far as pioglitazone is concerned, the cardiovascular safety issue has not been fully researched. Our study suggests that pioglitazone does not present as high a risk as rosiglitazone; however, it should also be used with caution.
Though randomized controlled trials (RCTs) are the golden standard for efficiency assessment of drugs, they are not necessarily so for drug safety because of their inadequate power to detect either multiple or rare adverse events. Furthermore, the conditions under which drugs are approved for market use are often different from the settings in actual use. Take rosiglitazone for example, the population of rosiglitazone’s RCTs are often younger than the patients who are treated with it. Thus observational studies are increasing in uptake because they reflect the real-life utility of drugs, which can be far more useful to drug safety assessment activities than generally acknowledged. In recent years both FDA and EMA consider observational researches as more appropriate vehicle of exploring drug safety issues. Our study assessed the difference in cardiovascular outcomes between rosiglitazone and pioglitazone through a meta-analysis of retrospective cohort studies. One of the strengths of this study was that we combined the results of cohort studies that reflected the differential effects of cardiovascular outcomes between rosiglitazone and pioglitazone during real-life use. To guard against heterogeneity, we had strict inclusion criteria, including that all the researches should be designed to study the cardiovascular outcomes of patients treated with anti-diabetic drugs, especially rosiglitazone or pioglitazone as first-line treatment. We excluded case-control studies because these will bring more heterogeneity to the analysis. Moreover, of all of the corresponding authors of the included studies, three authors out of eight studies responded to our request, one author provided further unpublished data, allowing a more detailed analysis. By the addition of data from Pantalone’s study, the risk of all outcomes increased for rosiglitazone compared with pioglitazone; probably because the number of patients included in this study was relatively small, so the result did not have much effect in the analysis. As the Tzoulaki’s study implies that the risk of heart failure favours rosiglitazone, and the risk of MI and all cause mortality did not show significant difference between the two TZDs, it is noticeable that the design of this study was to compare those two drugs as one category with metformin or SU. Thus the sample size of TZDs was relatively small, especially for pioglitazone which combined mono therapy and multi therapy as one group.
The weaknesses of this meta-analysis included the limitations of observational studies which brought together biases from all of the cohort studies. The differences of research characteristic reminded us that cohort studies were less homogeneous without control mechanism, such as the treatment intervention, patients’ characteristic, sample size, follow-up period and control group are all varied from each other. So the results of this meta analysis were less compelling as analysis of RCT’s. To guard against this, the inclusion criteria were strengthened and further sensitive analysis and subgroup analysis were performed. Since the follow-up time of the included studies varied, a sensitivity analysis was performed after excluding one study with only a 1-year follow-up. Finally, as we did not have access to the patient characteristic data, further analysis of the risk factors which contributed to the adverse outcomes could not be performed.
Since diabetes mellitus is complicated by cardiovascular diseases, the adverse effects of hypoglycaemic agents cannot be sensitively detected through comparisons with placebos or healthy volunteers. Only in the most natural circumstances can a rare adverse effect of one drug be detected in a large population, without any interference due to the specific purpose of any clinical trial. On the other hand, the comparison of rosiglitazone to any other treatment, including insulin, oral hypoglycaemic agents, or even placebo, resulted in great heterogeneity among the studies involved in the previous meta-analysis. Therefore, we chose retrospective cohort studies of diabetic patients who were treated with the two TZDs in order to provide direct evidence in otherwise comparable patients. The studies we included were all cohort studies that compared these two drugs directly, since there has been no such previous randomised controlled trial. These retrospective cohort studies provide the best evidence to evaluate the safety profiles of these two drugs.
Our results are consistent with a previously suggested, relatively adverse cardiovascular safety profile for rosiglitazone. Although our study was not able to provide evidence for the absolute risk of any outcome for either rosiglitazone or pioglitazone, as there was no reference group treated with medications other than those two drugs, our results indeed showed that, compared to pioglitazone, rosiglitazone was associated with an increased risk of all outcomes for the whole population. We also analyzed the risk of non-insulin treated patients, and the differences in rates of myocardial infarction and heart failure were not significant, suggesting that patients treated with insulin and these two TZDs are at higher risk of cardiovascular diseases. Moreover, previous studies that reported an insignificantly increased risk of MI with rosiglitazone were conducted in younger populations (mean age, 54–65 years). There is a possibility that the difference in cardiovascular outcomes for the two TZDs changes with increasing age. The progression of diabetes increases sudden cardiac death, along with aging. Thus, the influence of the drug was comparable when the events were more common in both groups.
In conclusion, our meta-analysis of eight retrospective cohort studies found that, compared with pioglitazone, rosiglitazone was associated with an increased risk of MI, heart failure, and all-cause mortality in diabetic patients. Further analyses would be recommended to compare the absolute risks of rosiglitazone and pioglitazone, as well as a comparison of these two drugs with other antidiabetic drugs of a different category, such as metformin or glimepiride.
1.  Nissen SE, Wolski K. Effect of rosiglitazone on the risk of myocardial infarction and death from cardiovascular causes. N Engl J Med 2007; 356: 2457-2471.
2.  Home PD, Pocock SJ, Beck-Nielsen H, Curtis PS, Gomis R, Hanefeld M, et al. Rosiglitazone evaluated for cardiovascular outcomes in oral agent combination therapy for type 2 diabetes (RECORD): a multicentre, randomised, open-label trial. Lancet 2009; 373: 2125-2135.
3.  Lincoff AM, Wolski K, Nicholls SJ, Nissen SE. Pioglitazone and risk of cardiovascular events in patients with type 2 diabetes mellitus: a meta-analysis of randomized trials. JAMA 2007; 298: 1180-1188.
4.  Dahabreh IJ, Economopoulos K. Meta-analysis of rare events: an update and sensitivity analysis of cardiovascular events in randomized trials of rosiglitazone. Clin Trials 2008; 5: 116-120.
5.  Khanderia U, Pop-Busui R, Eagle KA. Thiazolidinediones in type 2 diabetes: a cardiology perspective. Ann Pharmacother 2008; 42: 1466-1474.
6.  Rohatgi A, McGuire DK. Effects of the thiazolidinedione medications on micro- and macrovascular complications in patients with diabetes--update 2008. Cardiovasc Drugs Ther 2008; 22: 233-240.
7.  Singh S, Loke YK, Furberg CD. Long-term risk of cardiovascular events with rosiglitazone: a meta-analysis. JAMA 2007; 298: 1189-1195.
8.  Gerstein HC, Yusuf S, Bosch J, Pogue J, Sheridan P, Dinccag N, et al. Effect of rosiglitazone on the frequency of diabetes in patients with impaired glucose tolerance or impaired fasting glucose: a randomised controlled trial. Lancet 2006; 368: 1096-1105.
9.  Silverman SL. From randomized controlled trials to observational studies. Am J Med 2009;122: 114-120.
10.  Suissa S, Garbe E. Primer: administrative health databases in observational studies of drug effects--advantages and disadvantages. Nat Clin Pract Rheumatol 2007; 3: 725-732.
11.  Yang W, Zilov A, Soewondo P, Bech OM, Sekkal F, Home PD. Observational studies: going beyond the boundaries of randomized controlled trials. Diabetes Res Clin Pract 2010; 88 (Suppl 1): S3-S9.
12.  Loebstein R, Dushinat M, Vesterman-Landes J, Silverman B, Friedman N, Katzir I, et al. Database evaluation of the effects of long-term rosiglitazone treatment on cardiovascular outcomes in patients with type 2 diabetes. J Clin Pharmacol 2010. doi: 10.1177/0091270010368281.
13.  Koro CE, Fu Q, Stender M. An assessment of the effect of thiazolidinedione exposure on the risk of myocardial infarction in type 2 diabetic patients. Pharmacoepidemiol Drug Saf 2008; 17: 989-996.
14.  Juurlink DN, Gomes T, Lipscombe LL, Austin PC, Hux JE, Mamdani MM. Adverse cardiovascular events during treatment with pioglitazone and rosiglitazone: Population based cohort study. BMJ 2009; 339: b2942.
15.  Graham DJ, Ouellet-Hellstrom R, Macurdy TE, Ali F, Sholley C, Worrall C, et al. Risk of acute myocardial infarction, stroke, heart failure, and death in elderly medicare patients treated with rosiglitazone or pioglitazone. JAMA 2010; 304: 411-418.
16.  Winkelmayer WC, Setoguchi S, Levin R, Solomon DH. Comparison of cardiovascular outcomes in elderly patients with diabetes who initiated rosiglitazone vs pioglitazone therapy. Arch Intern Med 2008; 168: 2368-2375.
17.  Gerrits CM, Bhattacharya M, Manthena S, Baran R, Perez A, Kupfer S. A comparison of pioglitazone and rosiglitazone for hospitalization for acute myocardial infarction in type 2 diabetes. Pharmacoepidemiol Drug Saf 2007; 16: 1065-1071
18.  Tzoulaki I, Molokhia M, Curcin V, Little MP, Millett CJ, Ng A, et al. Risk of cardiovascular disease and all cause mortality among patients with type 2 diabetes prescribed oral antidiabetes drugs: retrospective cohort study using UK general practice research database. BMJ 2009; 339: b4731
19.  Hsiao FY, Huang WF, Wen YW, Chen PF, Kuo KN, Tsai YW. Thiazolidinediones and cardiovascular events in patients with type 2 diabetes mellitus: a retrospective cohort study of over 473 000 patients using the National Health Insurance database in Taiwan. Drug Saf 2009; 32: 675-690.
21.  Brownstein, JS, Murphy SN, Goldfine AB, Grant RW, Sordo M, Gainer V, et al. Rapid identification of myocardial infarction risk associated with diabetes medications using electronic medical records. Diabetes Care 2010; 33: 526-531.