The FDA acknowledges that Real World Data (RWD) “are playing an increasing role in health care decisions.” However, key information is often missing and thus determining causal inference of treatment effect and RWD internal validity is challenging. Methods that increase our confidence in the correspondence between conclusions from RWD and randomized clinical trial (RCT) data could allow clinicians and policymakers to make enlightened decisions using either source of data. To address these challenges, we describe a modeling approach applied to RCT data that explicitly considers the absence of predictive data to identify estimates of treatment efficacy that more closely match RWD. We use data from 6 RCTs investigating the effect of secukinumab on chronic plaque-type psoriasis as our sources of data. We describe a random censoring approach of the RCT data to create a data set that is missing key information and more closely aligns with RWD. Using the censored data, we show how a longitudinal analytic technique (latent growth modeling [LGM]) models individual-level change across all time points to derive precise estimates of treatment effects at 4-week intervals. By conducting LGMs in each RCT, we validate the model by testing invariance (i.e., whether they are measuring the same thing) and differences in mean scores. The validated model estimates treatment effect in the absence of key information and can be compared to results from RWD sources. Comparable results can yield confidence in making causal inferences of treatment effect from RWD sources. These methods are a flexible and efficient way to estimate treatment efficacy that more closely match those observed within the real world and thus can be used to inform the treatment decision making process.
