Abstract:
|
Surprisingly, the duration of survival from the time of a lung cancer diagnosis has been found to be longer for prior cancer survivors than for those with no prior cancer. A possible explanation is lead-time bias, which confers an extension of the survival on those with prior cancer diagnoses. We propose a discrete semiparametric model to jointly describe survival in both no-prior-cancer group and prior-cancer group. We model the lead time with a negative binomial distribution and the post-lead-time survival with a linear spline on the logit hazard scale, which allows for survival to differ between prior-cancer groups even in the absence of bias. We fit the model to data from the SEER-Medicare linked data set, conducting a sensitivity analysis to assess the potential effects of lead-time bias on estimates of the survival difference between groups. With lung-cancer death as the endpoint, mean lead time is estimated as roughly 9 months for stage I&II patients. For patients with higher-stage lung cancers, the lead-time bias is on the order of one month or less. Even accounting for lead-time bias, there are modest but statistically significant survival differences between the groups.
|