Margulis AV, Varas-Lorenzo C, Bui CL, McQuay LJ, Ziemiecki R, Reynolds M, Rebordosa C, Pladevall-Vila M, Fortuny J, Rivero-Ferrer E, Plana E, Atsma WJ, Appenteng K, Franks B, de Vogel S, D'Silva M, Perez-Gutthann S, Arana A. Validation of cardiovascular events and covariates in CPRD GOLD using questionnaires to general practitioners (GPS). Presented at the 32nd ICPE International Conference on Pharmacoepidemiology & Therapeutic Risk Management; August 27, 2016. Dublin, Ireland. [abstract] Pharmacoepidemiol Drug Saf. 2016 Aug; 25(Suppl 3):255-6.

BACKGROUND: Reliable assessment of events and covariates is needed for valid study results.

OBJECTIVES: To evaluate a process to identify and verify cases of cardiovascular (CV) events and patient characteristics within a study of CV safety in users of overactive bladder (OAB) drugs in the CPRD.

METHODS: We identified a cohort of new OAB-drug users aged greater than or equal to 18 years in 2004-2012. Epidemiological definitions for acute myocardial infarction (AMI) and stroke were used. Events and patient characteristics in cases were compared, in a sample, against information from questionnaires sent to GPs.

RESULTS: A total of 5,593 CV events were identified in CPRD GOLD via algorithms. GPs'  response rate was 81%. Positive predictive values (PPVs) and 95% confidence intervals: definite AMI, 98% (94%-100%); probable AMI, 92% (86%-96%); possible AMI, 92% (73%-99%); definite stroke, 92% (85%-96%); probable stroke, 79% (69%-87%); and possible stroke, 84% (70%-93%). The negative predictive value (95% CI) was 99% (95%-100%) for noncases that were alive at the end of follow-up and 84% (77%-90%) for noncases who died during the study. In CPRD GOLD, smoking information was available for 1728 of 1731 patients with questionnaire information on smoking; on the closest day before the event, 17% of patients were current smokers, 41%, former smokers, and 42%, never smokers. Of patients identified as current smokers in GOLD, 84% were also current smokers according to GP questionnaires; likewise, 77% of former smokers and 97% of never smokers were confirmed. In CPRD GOLD, information on obesity (BMI greater than or equal to 30 kg/m²) was present in 74% of the 1713 patients with questionnaire information on obesity. Of the patients classified as obese in CPRD GOLD, 82% were confirmed through GP questionnaires; of patients classified as not obese, this was confirmed for 92%. History of AMI was confirmed in 70% of the cases, and history of stroke in 44%. Of patients assumed to be premenopausal, 12% were confirmed and 77% were reported as having gone through menopause by their GPs.

CONCLUSIONS: AMIs, definite strokes, smoking status and obesity identified via electronic algorithms in CPRD GOLD could be confirmed in most cases.