Integrating an economic evaluation within a trial provides researchers the opportunity to collect patient-level data at multiple time points on both cost and effectiveness outcomes, and to measure the comparative impact of treatments on these outcomes while controlling for confounding factors, such as a participant’s medical history and socioeconomic characteristics.
However, bias associated with missing data is a common problem encountered when collecting data from trial participants, and sampling uncertainty must also be accounted for when determining the likelihood that a treatment is cost-effective. The nonparametric bootstrap method is widely recommended to account for sampling uncertainty in economic evaluations, but there are multiple valid approaches that researchers can choose from to address missing data biases. There are also multiple ways of combining these methods together in a single analysis.
In a new study published in the 2022 theme issue Epidemiological Reviews titled “Clinical Trials: Current Status and Future Directions,” Assistant Professor Ali Jalali from Weill Cornell Medicine; Associate Professor Sean M. Murphy from Weill Cornell Medicine; Professor Rulla Tamimi from Weill Cornell Medicine; and Professor Sterling M. McPherson from Elson S. Floyd College of Medicine at Washington State University, outline potential pros and cons of approaches to address these statistical issues. The authors argue that the method a researcher chooses to implement to correct for missing data bias and how this method is combined with the nonparametric bootstrap are not independent decisions. This poses unique statistical and practical challenges for researchers conducting an economic evaluation alongside a clinical trial.
Using prior-published systematic reviews of economic evaluations and a targeted literature search, the authors identified and critically reviewed the statistical methods employed by 29 studies to address missingness, identified whether the nonparametric bootstrap was included in each analysis, and evaluated how each study combined these methods in order to highlight the advantages and shortcomings of each approach. Given the scope and complexity of potential statistical issues that may arise for researchers interested in including an economic evaluation in a clinical trial, the authors recommend establishing close and early collaboration between economists and clinical scientists in the clinical trial design process.