Thank you for a much needed review addressing the gaps in literature regarding the risks and benefits of antiplatelets in the chronic kidney disease (CKD) population. We thought the literature search was very thorough and well done. However, we came up with a few questions upon reading this review and felt that the stated conclusion "antiplatelets reduce myocardial infarction...including those with early stages of CKD who do not have clinically-evident occlusive cardiovascular disease" may not be accurately reflected by the presented data.
Looking at the first primary outcome - fatal and non-fatal myocardial infarction (MI), it was unclear whether the population studied was addressing primary prevention, secondary prevention or acute treatment of MI as the included populations had different cardiovascular histories. Of the two studies that were given the most weight in the analysis (HOT Study 2010 and PURSUIT Study 1998), one investigated primary prevention of MI using ASA versus placebo, while the other investigated acute treatment of MI using eptifibatide + ASA + heparin compared to ASA + heparin. In the non-CKD population, efficacy of antiplatelets is dependent on the indication (ie. primary or secondary prophylaxis or treatment). Different antiplatelets also have different places in therapy.
We therefore feel that it may be inappropriate to pool these trials together as they were investigating different populations.
In this same analysis, there were also multiple interventions such as single antiplatelets versus placebo (HOT study 2010, Ell study 1982, Creek 1990, Dember 2008, STOP study 1995, UK-HARP-I study 2005, ETDRS 1992), dual antiplatelets versus placebo (Kaufman 2003), as well as dual antiplatelets versus single antiplatelet agents (CREDO study 2008, CHARISMA study 2009, EPILOG study 1997, EPIC study 1994, EPISTENT study 1998, Dixon study 2009, RAPPORT study 1998, PURSUIT study 1998, and IMPACT II 1997). With both placebo and antiplatelet in the "control" arms of one meta-analysis, comparison groups and treatment groups are not clearly delineated from one another. As this was unclear, readers may be misled into believing that the effect is driven purely from antiplatelet compared to placebo, when this is not the case. Even pooling the data on the seven placebo-controlled trials may be inappropriate as they were studied in different patient
populations and indications (e.g. primary prevention, non-cardiovascular outcomes). Similarly, the "treatment arms" of the meta-analysis contained one or more antiplatelet agents, which may have
biased the result towards the treatment arm over single agent or placebo "control". This can also make it difficult to isolate the beneficial agent in the dual antiplatelet studies. Due to the differences in treatment arms and patient populations, we feel it would valuable to investigate the outcomes of these factors in separate analyses.
It should also be mentioned that the patients included in this review were derived as subgroups from larger studies with different baseline cardiovascular risk factors (e.g. diabetes, coronary artery disease, hypertension, etc). As a result, one cannot conclude that patients with only CKD, and no additional cardiovascular risk factors, would benefit from antiplatelet use to decrease cardiovascular outcomes such as fatal and non-fatal MIs. Dixon 2009 and Dember 2008 were two
studies enrolling hemodialysis patients with a primary outcome of AV graft patency or thrombosis; fatal and non-fatal MIs were only reported as an adverse effect and could have been under-reported in the study.
We commend the authors for assessing bias in the included trials and for performing a sensitivity analysis to explore the impact of the bias. We feel that with the relatively high percentage of unclear or high risk of biases that exist in the trials, it would have been beneficial for the authors to report on the results of their sensitivity analyses to clarify the role of the bias and to substantiate the reported results.
We feel that the author's conclusion "antiplatelet agents reduce myocardial infarction" may be too broad of a conclusion to be drawn based on the analysis that was performed looking at fatal and non-fatal MI. As well, their specific reference to "patients with early stages of CKD who do not have a clinically-evident occlusive cardiovascular disease" suggests this effect is shown in the CKD
population when using antiplatelets for primary prevention; however, this aspect was not separated out in their analysis. We feel that the pooling of studies with varying patient populations and treatments is not appropriate in helping clinicians determine whether antiplatelets provide any benefit for MI in patients with CKD. While we did not explore the other identified primary outcomes in this review, we wonder if similar concerns exist for not only the efficacy but also the safety outcomes. We would appreciate an investigation into single antiplatelet therapy versus placebo for various cardiovascular indications. We hope the authors will provide clarification and address these concerns in their future updates.
We look forward to hearing your response to our comments.
Gloria Su, BSc. Pharm
Wan-Yun Polinna Tsai, BSc. Pharm
Megan Harbin, BSc. Pharm
Asal Taheri, BSc. Pharm
Aaron M Tejani, BSc. Pharm, PharmD
Thanks for the constructive comments.
1. Primary versus secondary prevention versus acute treatment
We combined treatment estimates for all available studies comparing antiplatelet therapy (with or without standard therapy) versus placebo/no treatment (with or without standard therapy alone) to examine treatment effects, which is a standard starting point for meta-analyses. For the outcome of fatal or nonfatal myocardial infarction, there was little or no heterogeneity in the treatment effects observed in all the available trials, suggesting that treatment estimates could be appropriately summarised into a single effect size.
While not necessary in the absence of significant heterogeneity, we explored for pre-specified trial-level variables that might have modified the treatment estimates that we observed. We specifically wished to know whether treatment effects differed for patients with existing cardiovascular disease compared to those without cardiovascular disease but this was not feasible due to as we found insufficient numbers of studies that were clearly primary prevention or secondary prevention studies. However, the lack of heterogeneity in the overall summary estimate suggests that antiplatelet agents have similar effects irrespective of the presence or absence of cardiovascular disease.
2. Multiple interventions:
Unlike the relative lack of primary versus secondary prevention trials, there was sufficient studies to explore any differences in treatment effects based on the class of antiplatelet used. While there were numerous different strategies for antiplatelet treatment in contributing trials, all the treatment interventions could be characterised by an antiplatelet agent in addition to standard care versus no treatment/placebo in addition to standard care. We have called this antiplatelet therapy versus control to acknowledge the heterogeneity of the intervention strategies used (rather than antiplatelet treatment versus placebo).
We used stratified analyses according to overall class of antiplatelet drug where possible but there was lack of power from available studies to understand fully all the various treatment effects for each individual antiplatelet regimen. An individual patient meta-analysis would be needed to give a more fine-grained understanding of the different interventions and their combinations in the CKD population.
3. Deriving patients from subgroups of larger studies:
Patients with CKD were evaluated in post-hoc analyses of larger trials in broader populations. These included trials in populations with acute coronary syndromes requiring percutaneous coronary artery procedures, patients with hypertension and those with diabetes mellitus. Trials of treatment tended to use different interventions (glycoprotein IIb/IIIa inhibitors with or without clopidogrel) whereas trials of primary or secondary prevention did not use these agents, preventing useful stratified analyses for either class of agent or cardiovascular prevention in these trials. We have concluded that the lack of a priori assessment of glycoprotein IIb/IIIa inhibitors in people with CKD is an important limitation of the current evidence.
4. Potential under-reporting of clinical outcomes
We agree that many trials were not designed to evaluate mortality and cardiovascular outcomes and that these events were reported in an ad hoc fashion (not prespecified) which may have underestimated their frequency. We include evaluation of this aspect of trials when considering whether they are at risk of bias due to selective reporting of expected outcomes.
5. Risks of bias
We did not specify risk of bias items as sources of heterogeneity we would explore in stratified analyses. In further updates of this review and if deemed appropriate and feasible, we will explore attrition bias and allocation concealment as potential sources of heterogeneity in subgroup or sensitivity analyses.
In conclusion, we thank Dr Su and others for constructive comments to this review. We agree that the review cannot provide high quality information about antiplatelet agents as primary prevention for cardiovascular disease in people with CKD. We acknowledge the limitations of studies in which adults with CKD were studied post hoc and which are heterogeneous for presence of cardiovascular disease and antiplatelet agent studied. We agree that clinical events may be under-reported in available studies and will explore in future versions of this review the effects of risk of bias on the estimated treatment effects of antiplatelet treatment in CKD.