Objective: This study was designed to measure the impact of variation in patient follow-up on survival statistics.
Methods: surveillance, epidemiology and end results (SEER) data were used to construct four additional datasets. These datasets simulated scenarios of complete, incomplete, and no follow-up of live patients; and complete and incomplete death ascertainment. Sixty-month observed survival proportions were calculated using the actual SEER data and the four additional datasets.
Results: The 60-month observed survival proportion increased from 54.44% under the original SEER dataset to 54.62% under complete ascertainment of deaths with no follow-up among live patients. Under complete death ascertainment, randomly imputing loss to follow-up among 20% of live cases resulted in a 1%-2% decrease in 60-month observed survival for 71 of the 102 SEER site categories. With follow-up limited to ascertainment of deaths, randomly missing 6% of deaths resulted in a 1% or greater increase in 60-month observed survival for 99 SEER site categories.
Conclusions: This study provides evidence to support the importance of complete death ascertainment for producing accurate cancer survival statistics, and that ascertainment of deaths only should generally be sufficient for survival analysis.