The generalizability of observational data to elderly patients was dependent on the research question in a systematic...

Despite the increasing use of data derived from randomized controlled trials (RCTs) to perform observational studies, little is known about the validity of this approach. We compared inferences from studies that were performed using Global Utilization of Streptokinase and t-PA for Occluded Coronary Arteries (GUSTO) RCT data with those derived from studies using data from the population-based Cooperative Cardiovascular Project (CCP).

We performed a systematic review. Articles were included if similar study questions were addressed by at least one manuscript that used GUSTO data and one that used CCP data.

GUSTO findings were disparate from CCP data regarding absolute rates of specific process or outcome measures, such as thrombolysis-associated intracranial hemorrhage (ICH) (0.65% versus 1.43%, respectively), atrial fibrillation (10% versus 21%, respectively), or use of beta-blockers (58% versus 37%, respectively). However, many important relations noted in GUSTO were corroborated by studies using CCP data. Both data sets identified similar predictors of ICH and presentation delay. The degree of variability in beta-blocker use (across geographic region) and angiography use (between genders) was remarkably similar when studied using CCP or GUSTO data.

Inferences derived from GUSTO about treatment variations and risk factors for outcomes were generalizable to community patients.