OBJECTIVES: Data scarcity poses significant challenges in rare disease research, particularly in survival analysis. This study leverages established methodologies to estimate individual patient data from published Kaplan-Meier (KM) curves, aiming to enhance survival estimates by pooling data across studies. We propose modifications to a popular method to increase its accuracy, applying these modifications to a case study in spinal muscular atrophy (SMA).

METHODS: A systematic literature review was performed to collect studies featuring KM curves for all subtypes of SMA. We modified the method developed by Guyot et al. to estimate individual patient data (IPD) to estimate censoring rate more accurately. We assumed varying censoring rates (compared to constant rate in Guyot’s method), and we calculated the cumulative number of censored patients (rather than rounding to nearest integer). To validate our method, we used the reverse engineering approach using a publicly available survival dataset. Then, we calculated the squared residual of the censoring rate (the error in estimating the censoring rate, which is the lower, the better) using the modified and original methods. Applying this to SMA, we aggregated the IPD data from the included studies and parametrized the aggregated KM curves.

RESULTS: The systematic literature review yielded seven included studies. IPD from these studies was successfully estimated, and survival analysis was performed separately for SMA subtypes. Using the reverse engineering method to compare the modified to the original method, the modified method was proved more accurate in estimating the IPD. The median survival was 40.6 , 568.2 months for type I and II, respectively. For type III, no disease-specific mortality was observed across the studies.

CONCLUSIONS: The study represents an improvement to existing method for estimating IPD from published KM curves. Our results provide the most comprehensive survival data to date for SMA where data about patients’ survival is scarce.