Adjusting survival estimates for differential loss to follow-up by race-ethnicity: a SEER analysis.

Abstract

Apparent survival advantages for Hispanics and Non-Hispanic Asians and Pacific Islanders (NHAPIs), often labeled the "Hispanic Paradox" or "Healthy Immigrant Effect," have been reported for multiple cancers in comparison to Non-Hispanic Whites (NHWs) and Non-Hispanic Blacks (NHBs). However, higher proportions loss to follow-up (LFU) among Hispanics and NHAPIs overestimate survival due to informative censoring. We developed a sensitivity analysis-based method to address potential impacts of differential LFU. We estimated 5-year age-standardized cancer-specific survival for NHWs (reference), NHBs, NHAPIs, Hispanics, and Non-Hispanic American Indians/Alaska Natives (NHAIANs) diagnosed 2005 to 2019 in 22 Surveillance, Epidemiology, and End Results (SEER) registries. Cases alive with < 5 years or maximum possible follow-up were classified as LFU. For each cancer-stage-race/ethnicity combination, for LFU cases exceeding NHW levels, we modeled two plausible scenarios relative to NHW survival 1) "lower" survival and 2) "higher" survival. Across 1,000 simulations, we constructed a "sensitivity interval" spanning the average lower 95% confidence bound of "lower" survival to the average upper 95% confidence bound of "higher" survival. LFU was higher for NHAPIs and Hispanics across cancer sites (min-max 2.2 to 20.7%) than for NHWs (0.4 to 5.4%). Sensitivity intervals were consistently wider than standard 95% confidence intervals. After accounting for differential LFU, previously observed equivalent or higher survival disappeared for NHAPIs (eg, prostate) and Hispanics (eg, lung). Ignoring differential LFU overstates survival advantages among racial-ethnic minorities. Sensitivity analysis reveals some reported survival advantages are artifacts of higher LFU, underscoring the need to address this censoring bias in population-based survival studies.