OBJECTIVES: NDRS rates are often utilized in survival extrapolations to avoid clinically implausible outcomes and to account for the possibility of cure. This study compared medium- (10-year) and long-term (20-year) NDRS estimated from local lifetables for the patient population in EORTC-30994 study to reported NDRS from the trial and its extrapolations, respectively.

METHODS: For each participating country in the trial, age- and sex-adjusted NDRS distributions were derived from the published lifetable data by World Health Organization and weighted with countries’ enrolment shares in the trial to generate an aggregate NDRS distribution. Trial median baseline age and sex distribution were assumed to be identical across countries. Parametric survival distributions and spline-based models were used to extrapolate reconstructed NDRS data pooled across both arms of the study. Weekly NDRS rates and restricted mean survival times (RMSTs) predicted from the lifetable-based NDRS distributions were compared versus their counterparts obtained from reported and extrapolated NDRS from the trial.

RESULTS: Log-normal distribution provided the best statistical and visual fits to the reported NDRS data for extrapolations. Lifetable-based NDRS overestimated trial-reported NDRS for the first 10 years during which it laid within the 95% CI of the reported NDRS 81% of the time. Estimated 10-year RMSTs from the reported NDRS and lifetable-based NDRS were 9.09 and 9.34 years, respectively. Estimated 20-year RMSTs from the extrapolated NDRS and lifetable-based NDRS were 17.12 and 16.39 years, respectively. Lifetable-based NDRS had an average rate gap of 0.025 versus the reported NDRS over 10-years and 0.062 versus extrapolated NDRS over 20-years.

CONCLUSIONS: For MIUC patients receiving immediate or no adjuvant chemotherapy after radical cystectomy, compared to NDRS estimated from trial data, lifetable-based NDRS can be slightly optimistic in medium-term and conservative in long-term. Generalization of results to broader adjuvant MIUC trial settings requires further analyses with varying baseline patient characteristics.