On the pooling and subgrouping of data from percutaneous coronary intervention versus coronary artery bypass grafting trials: a call to circumspection