Curriculum in Cardiology
Challenges in meta-analysis of randomized clinical trials for rare harmful cardiovascular events: The case of rosiglitazone

https://doi.org/10.1016/j.ahj.2008.03.002Get rights and content

Rare cardiovascular events of commonly used drugs are important to document and investigate, but single trials are notoriously underpowered to provide conclusive evidence. Recently, meta-analyses have been used to improve on the power. A recent rosiglitazone meta-analysis heightened the debate about the usefulness and limitations of meta-analysis in this setting. In this review, we examined the methods used in previous published meta-analyses for harmful cardiovascular events, with special attention to the rosiglitazone meta-analyses, and give suggestions for the improvement of methods and interpretation of such meta-analyses. The conduct of meta-analysis in this context is particularly difficult and requires timely investigation, availability of high-quality data on harms, and statistical expertise. There are important decisions that need to be made about selecting the appropriate analytical methods and performing sensitivity analyses to evaluate whether the results are robust to different analytical choices.

Section snippets

Methods

We retrieved and examined English-language meta-analyses of RCTs exploring adverse cardiovascular events of drugs. Keywords used were “meta-analysis” or “meta-analy”; “randomized clinical trials” or “randomized controlled trials” or “randomized trials”; “adverse cardiovascular event” or “cardiovascular event” or “cardiovascular risk” or “cardiovascular safety” or “cardiovascular harm”; “rare event”; and “method.” If 2 or more similar meta-analyses were performed by the same research group,

Meta-analyses of rare cardiovascular events: interventions other than rosiglitazone

We retrieved 8 major meta-analyses of RCTs for rare harmful cardiovascular events due to drugs other than rosiglitazone. Each meta-analysis included data on >3,000 patients (range 3,089-145,373) and summarized between 4 and 121 RCTs (Table I). A total of 7 meta-analyses15, 17, 18, 19, 20, 21, 22 were based on published results, and 1 was based on individual-patient data16; 6 meta-analyses16, 17, 19, 20, 21, 22 focused on cyclooxygenase-2 inhibitors (coxibs), and 5 of them15, 16, 18, 19, 21

Rosiglitazone meta-analyses

Seven rosiglitazone meta-analyses6, 9, 10, 11, 12, 13, 14 were conducted in a short span of time after the publication of the original meta-analysis.1 Nissen and Wolski1 reported that in >27,000 patients with type 2 diabetes, rosiglitazone was associated with a statistically significant increase in the odds of MI (OR 1.43, 95% CI 1.03-1.98, P = .03) with a trend for increase in the odds of cardiovascular death (OR 1.64, 95% CI 0.98-2.74, P = .06) when compared with a control group (active

Inherent problems in meta-analyses of rare cardiovascular events

Most of the problems of meta-analyses of rare events, as highlighted in the rosiglitazone situation, are intrinsic of the original trials. Short-term scope, small sample size, or the definition of cardiovascular events as secondary outcomes are examples of shortcomings of individual trials. Meta-analysis allows a systematic examination and discussion of these shortcomings, and this is a clear advantage over the fragmented examination of single trials. Some of these shortcomings can be improved

What is desirable in a meta-analysis for cardiovascular harms?

A clearer description of the absolute risks per treatment arm is advised. For example, it would be desirable, given its direct clinical usefulness,26 to express the results using NNH (calculated from the summary OR or RR) with the caveats of heterogeneity of absolute risks. For instance, Singh et al13 incorporated NNHs in their meta-analysis of cardiovascular effects of long-term use of rosiglitazone. They used baseline risks from several RCTs or cohorts and incorporated the RR found in their

Conclusions

Rare harmful cardiovascular events associated to commonly used drugs are very important from clinical and public health perspectives because they can have a heavy toll on morbidity and mortality. However, the evaluation of these rare events present several challenges. Because of the sparcity of harms, meta-analyses are powerful tools in this endeavor. The conduct of meta-analysis in this context is particularly difficult and requires timely investigation, availability of high-quality data on

References (42)

  • The Lancet

    Rosiglitazone: seeking a balanced perspective

    Lancet

    (2007)
  • P.D. Home et al.

    Rosiglitazone evaluated for cardiovascular outcomes—an interim analysis

    N Engl J Med

    (2007)
  • D.M. Nathan

    Rosiglitazone and cardiotoxicity—weighing the evidence

    N Engl J Med

    (2007)
  • B.M. Psaty et al.

    The record of rosiglitazone and the risk of myocardial infarction

    N Engl J Med

    (2007)
  • V. Fuster et al.

    Faster publication isn't always better

    Nat Clin Pract Cardiovasc Med

    (2007)
  • GlaxoSmithKline

    Study no. ZM2005/00181/01; Avandia Cardiovascular Event Modeling Project

  • G.A. Diamond et al.

    Uncertain effects of rosiglitazone on the risk of myocardial infarction and cardiovascular death

    Ann Intern Med

    (2007)
  • M.B. Bracken

    Rosiglitazone and cardiovascular risk

    N Engl J Med

    (2007)
  • J.J. Shuster et al.

    Fixed vs. random effects meta-analysis in rare event studies: the rosiglitazone link with myocardial infarction and cardiac death

    Stat Med

    (2007)
  • S. Singh et al.

    Long-term risk of cardiovascular events with rosiglitazone. A meta-analysis

    JAMA

    (2007)
  • S.D. Ross et al.

    Risk of serious adverse events in hypertensive patients receiving isradipine: a meta-analysis

    J Hum Hypertens

    (1997)
  • Cited by (35)

    • Harms in Systematic Reviews Paper 1: An introduction to research on harms

      2022, Journal of Clinical Epidemiology
      Citation Excerpt :

      Harms data that have been collected in studies may be insufficient for analyses if they do not include relevant dimensions such as date of onset, duration and severity of the event, and concomitant medications. Additionally, the lack of use of standardized terminology when collecting harms, as described in section 2, can lead to incompleteness and inconsistency in the data which further threatens the validity of harms analyses.[55–60] Interviews with trialists have revealed many reasons for not analyzing and reporting harms, including: an unawareness of the importance of “negative” results, data being considered “uninteresting”, having “too few events worth reporting”, and a perception of limited space in publications”.[55]

    • Effect of diastolic dysfunction on postoperative outcomes after cardiovascular surgery: A systematic review and meta-analysis

      2016, Journal of Thoracic and Cardiovascular Surgery
      Citation Excerpt :

      Secondary study outcomes were postoperative complications, which include major adverse cardiac events (MACE), prolonged mechanical ventilation, atrial fibrillation (AF), and myocardial infarction (MI). We performed meta-analysis with RevMan 5.3 (Cochrane Collaboration, Copenhagen, Denmark) and R 3.0.1 (www.r-project.org) using the Mantel-Haenszel fixed-effects model, given the scarcity of events.7 Random effects meta-analyses via the inverse variance method were performed otherwise.

    • Aggregated adverse-events outcomes in oncology phase III reports: A systematic review

      2016, European Journal of Cancer
      Citation Excerpt :

      Furthermore, a clear description and definition of A-AEs outcomes would also facilitate future meta-analyses on AEs. Meta-analyses are one such use and can provide additional insights about the tolerance and the relative risks of anticancer therapies [20,21]. It is important to note that even the use of tools such as the MedDRA may not entirely overcome the problems of using A-AEs without appropriate definition.

    • Reply

      2014, American Journal of Cardiology
    • Reporting of Harm in Randomized Controlled Trials Published in the Urological Literature

      2010, Journal of Urology
      Citation Excerpt :

      However, we believe it is important to provide these data since systematic review of all available studies may allow detection of important differences.12 Unfortunately if harm outcomes are not presented in published articles, meta-analyses of harm data are limited or impossible.13–16 In our review even in instances in which harm information was reported, 53% lacked enough detail to be used in a meta-analysis (ie the number of patients in each group who provided adverse event information).

    View all citing articles on Scopus
    View full text