“…Cox proportional hazards regression models were applied to assess adjusted risks of the outcome measures in anemic and nonanemic patients and for continuous haemoglobin levels at baseline. Multivariate models were adjusted for prespecified variables known to be associated with worse outcomes in coronary artery disease and PCI (Model 1): Age, gender, diabetes mellitus, dyslipidemia, hypertension, smoking, prior myocardial infarction, prior PCI, prior coronary artery bypass grafting (CABG), multivessel disease, moderate/severe coronary calcification, chronic kidney disease, estimated glomerular filtration rate (eGFR), C‐reactive protein, reduced left ventricular ejection fraction (<40%), number of stents implanted, procedural success, and use of bare metal stents. In order to exclude any potential bias introduced by overfitting, a second model was generated including the following variables only (Model 2): Age, diabetes mellitus, chronic kidney disease, and procedural success .…”