The Application of Quantitative Methods to Merge Clinical Data between Studies

Abstract

Randomized controlled trials (RCTs) are a standard for assessing the benefit of competing healthcare interventions. However, obtaining RCT data is time-consuming and expensive. Observational data is more readily available, however it is not oftentimes clear how comparable the findings are between observational and RCT data. This work assesses that question directly. Three different clinical studies are evaluated. The first is a double-blinded RCT comparing the difference in lowering intraocular pressure between the anti-glaucoma medications Xalatan and Xalacom. The second study is a single-blinded RCT comparing Xalatan to a placebo. The third study is an observational study comparing Xalacom to a placebo. Three quantitative techniques are used to merge the results of studies 2 and 3 in order to determine which technique best approximates the gold standard double-blinded RCT results. The three quantitative techniques are: linear regression, propensity score matching, and simple subtraction. Study 1 calculated the difference in intraocular pressure between Xalatan and Xalacom as 0.97 mmHg. The approximation results were: 0.86 mmHg (linear regression); -1.56 mmHg (nearest neighbor propensity scoring); 0.79 mmHg (kernel matching propensity scoring); 0.77 mmHg (radius matching propensity scoring); 1.03 mmHg (stratification propensity scoring); and -1.10 mmHg (simple subtraction). This work finds that stratification-based propensity score matching best approximated the gold standard results. It provides preliminary evidence that quantitative techniques may be used to merge studies and approximate double-blinded RCT data.