Development of rare disease meta-analysis, a meta-analytical method incorporating case studies
P Patel, E Shaeffer, B Willie, S Komarova
The Canadian Institutes of Health Research estimates that 1 in 15 children is born with a rare disease. Despite their prevalence, there is a limited representation of rare diseases in academic literature, which limits the progress of basic and translational research. In our systematic review of collagen pathophysiology in osteogenesis imperfecta, we have noted a large portion of the literature presents
case studies between 1 to 5 patients. Meta-analysis is a robust and powerful knowledge translational tool to assess the difference between disease and a healthy state quantitatively. However, traditional
meta-analysis cannot be applied to rare disease datasets due to the nature of case studies. We developed the 鈥淩are-Disease Meta-Analysis鈥� framework, which handles the inability to calculate effect sizes for single-patient publications and the significantly low variance in publications with 2 to 5 patients compared to the global variance. Employing the bootstrap approach which iteratively resamples the dataset to obtain an estimate of population variance distribution. The variance distribution is then applied to the individual data points. The validation of this method was performed using simulated data
and working datasets. When the random effects model was chosen, the overall effect size and the heterogeneity were not affected by the choice of the inter-study variance estimator tau. Simulation results revealed minimal heterogeneity in the data, with the overall effect converging to a stable value with maximum likelihood and restricted maximum likelihood estimator. This work establishes that applying Bootstrap on rare disease datasets provides an alternative to overcome issues with variance. Rare disease meta-analysis will be an essential tool for researchers to navigate data scarcity in the field of rare diseases.