Article Text

Meta-analyses of incomplete trial datasets: unreliable and potentially misleading
  1. Victor Serebruany1,
  2. Nikita Lomakin2,
  3. Thomas Marciniak3
  1. 1Division of Neurology, Johns Hopkins University School of Medicine, Baltimore, Maryland, USA
  2. 2Department of Intensive Cardiology, Central Clinical Hospital of the Administrative Affairs of the President of the Russian Federation, Moscow, Russian
  3. 3Physician, Bethany Beach, Delaware, USA
  1. Correspondence to Dr Victor Serebruany, Johns Hopkins University School of Medicine, Division of Neurology, Stroke Unit, 7600 Osler Drive, Suite 307, Towson 21204, Maryland, USA; heartdrug{at}

Statistics from

While clopidogrel remains the dominant oral antiplatelet agent (OAA), newer P2Y12 platelet inhibitors such as prasugrel or ticagrelor are expanding for postacute coronary syndrome indications. However, reliable randomised data among these three OAA are lacking, and thus, the comparative outcomes remain unclear. Even less is known regarding special patient subsets dichotomised by age, gender, comorbidities and concomitant medications. Therefore, data from carefully designed and properly implemented meta-analyses may provide critical, but yet missing evidence. Moreover, yielding specific advantages or risks of certain OAA for various clinical scenarios represent an urgent, unmet and unsolved need with regard to the choice of agents, duration of treatment and potential dose/regimen adjustment. The lack of large, randomised clinical trials that are adequately powered to compare the various OAA and designed to specifically assess the impact of gender, the absence of standardised efficacy and safety data and the overly optimistic publications that have been based on post hoc subanalyses of the primary mega-trials have tended to exaggerate the clinical benefits and minimise risks of the newer agents.

Lately, both prasugrel and ticagrelor suffer from array of secondary subset publications claiming superiority although the quality of such data are usually poor, while their interpretation is somewhat biased. Finally, triaging patients by any variable in general, and gender in particular is problematic due to massive missing data.

Missed and incomplete evidence frequently encountered problems in clinical research that challenge the validity of the overall results.1 In fact, a senior food and drug administration (FDA) official has identified missing data as the major problem with recent outcome trials. The FDA itself commissioned the US Institute of Medicine to prepare a special report to address optimal handling of the missing data in clinical trials.2 We recently published the analysis of 23 trials which randomised 276 967 patients for a median duration of follow-up of 19 months.3 Ironically, the ‘officially’ published rates of lost to follow-up (median 0.3%) were substantially less than actual total incomplete follow-up rates (median 13%) and there was no correlation between the two sets of rates. Most strikingly, the median difference in primary endpoint rates was about 1.0%, usually greater than the published lost to follow-up rates, but much less than the true incomplete rates challenging the entire credibility of the clinical trial concept. Moreover, the documented drug discontinuation rates were very high (median 23.7%). These data are relevant for our critical comprehension of the clinical trials since published follow-up statistics heavily under-represent the true completeness of follow-up in antithrombotic trials. With these reservations in mind, we should be upmost sceptical in the comprehension of any subanalyses of published trials, and especially with the attempts to pool these incomplete datasets together for meta-analyses. Not only separate trials may have different primary outcomes, definitely not identical bleeding definitions, and various patient selections, the published data do not adjust the results to massive double-digit missing or/and incomplete patients, which in turn differ from trial to trial as well. Our inability to assess and handle these issues adequately has been well recognised.1–3

In their Heart paper, Lee and colleagues4 analysed pooled datasets of five earlier clopidogrel trials (CURE, COMMIT, CLARITY CHANCE and SPS3), two trials with prasugrel (TRITON and TRILOGY) and two trials with ticagrelor (PLATO and SOCRATES). The authors dichotomised the entire cohort by gender and confirmed few important issues. First, women (n=17 842) are heavily underrepresented than men (n=27 818) in randomised trials with OAA. Second, and upmost important, outcomes in women are far inferior with mortality being higher for women than men for all-cause (5708, 32.0% vs 5891, 21.2%), cardiovascular (4032, 22.6% vs 4117, 14.8%) and bleeding (193, 1.1% vs 228, 0.8%) deaths. Finally, the absolute risk reduction for mortality at 1 year was similar for women and men for all-cause (2.30% (95% CI −0.92 to 5.22) vs 2.47% (95% CI 0.62 to 4.10)), cardiovascular (2.70% (95% CI −0.63 to 5.74)) vs 2.72% (95% CI 0.92 to 4.35)) and bleeding (−0.27% (95% CI −1.06 to 0.30) vs −0.18% (95% CI −0.71 to 0.24)) deaths.4 In fact, the FDA scrutinised five out of nine trials pooled in the discussed paper. These data are presented in table 1.

The complete data from other four studies (COMMIT, CLARITY, CHANCE and SOCRATES) are unavailable, since they were never independently generated or verified by the FDA reviewers. Aside from unquestionable quality of the discussed meta-analysis and advanced statistical techniques, there are few important considerations to be yielded from these latest elegant data. More women should be enrolled in future randomised trials to diminish disparity. This is especially true since females exhibited worsened cardiovascular outcomes when compared with males. The index data are in full agreement and confirm the existing evidence that despite huge differences among the trials with regard to exclusions, baseline characteristics, randomisation or enrolment patterns and length of follow-up, women are in consistently higher risks to experience primary vascular endpoint event.5 6 Also, women experience much more frequent bleeding events, especially those with low weight, and advanced age.7 8 These disturbing findings raise obvious concerns that we may consider OAA dose/regimen downgrades in females, the strategy which is currently not recommended by the regulatory agencies.

In summary, all OAA mega trials suffer from massive double-digit incomplete follow-up rates and/or drug discontinuations challenging the quality of the analysed datasets and limiting the practical value of further subanalyses mining. Overall, the current knowledge suggests that gender differences should be considered when picking the particular OAA, length of therapy and careful selection of concomitant medications. Whether gender per se is responsible for such differences is still unclear. Further comparative randomised studies of different OAA in patients of both sexes are urgently needed. Finally, from the practical standpoint, we are currently at the edge to seriously consider dose and/or regiment modification for antiplatelet agents. Indeed, the industry will viciously resist this trend, however, the escalating body of evidence unquestionably suggest patient benefit from such alternative strategies. Since no randomised data are available, and most likely there will be lack of solid evidence in the future, common sense and experience of prescribing physicians will rule. Capturing and reporting of follow-up needs to be improved to ensure the validity and interpretability of trial results, publications, meta-analyses and drug labelling.

Table 1

Follow-up and drug discontinuation rates in antiplatelet trials


View Abstract


  • Contributors Contributors VS and NL initially outlined a sketch of the text. TM provided the final draft. All authors contributed to critical review, revisions and agreed to the final version.

  • Competing interests None declared.

  • Provenance and peer review Commissioned; internally peer reviewed.

Request permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

Linked Articles