Observational Studies

Video Transcript

With an increased regulatory focus on incorporating patient perspectives in drug development, patient-reported outcomes, known as PROs, are playing an increasingly key role. Often an important step in helping stakeholders understand the relevance of PROs, is to demonstrate the relationship between a PRO and a more familiar clinician-reported outcome or biological marker. 

Most of the research in this area uses cross-sectional analyses that offer only an evaluation of the data at individual time points and does not account for the longitudinal design of most trials. Moreover, there is a risk that results from cross-sectional methods may be misinterpreted by examining the relationship across time-points.

Our goal for this study was to demonstrate the advantages of using longitudinal data methods – and particularly a joint mixed model for repeated measures – versus more commonly used cross-sectional methods based on an example data set consisting of a PRO and clinician-reported outcome. Our research showed that the joint MMRM approach modeled the relationship between both outcome variables simultaneously, allowing for a more thorough examination including an evaluation of the statistical significance of the correlations both within and between the outcomes over time.

We hope that this article helps researchers understand the value of using more robust analysis approaches, such as longitudinal joint MMRM, when examining relationships between different types of outcome assessments.

Longitudinal Modeling Approaches to Assess Two Clinical Outcome Assessments

Analyzing real-world, observational data requires specialized statistical skills. Our biostatisticians work closely with RTI-HS outcomes researchers to design and analyze data from a wide range of study types, including surveys, cross-sectional or longitudinal non-interventional studies, registries, chart abstractions, and electronic medical records (EMRs). 

Registry and Database Analyses

Our biostatisticians are knowledgeable in the statistical techniques necessary to analyze non-randomized studies. These studies require special consideration to address potential confounding, typically using methods such as propensity scores. Our researchers have the necessary expertise to define exposure and outcome(s), and to account for censoring and missing data. In addition, we assist researchers in the interpretation of the results and examination of potential biases so that you can be confident in using your results to make informed decisions.

Survey Analyses

Whether it is a simple survey or a complex multi-stage survey design, our statisticians have extensive experience in the planning and analysis of surveys including survey sampling and weighting.  We understand that analyses of complex sample surveys require the use of appropriate statistical software that accounts for the sample design so that the results accurately answer your research questions about the population of interest.