Survival analysis without survival data: connecting length-biased and case-control data.

Area of Science:

• Biostatistics
• Epidemiology
• Survival Analysis

Background:

• Traditional survival analysis requires prospective follow-up for data collection.
• Estimating survival parameters often necessitates complete survival time data.

Purpose of the Study:

• To develop a method for estimating survival model parameters using only covariate data.
• To demonstrate that survival analysis is possible without prospective follow-up.
• To explore estimation of interaction and exposure effect parameters.

Main Methods:

• Utilizing covariate data from incident and prevalent samples.
• Employing an induced semiparametric density ratio model for covariates.
• Applying likelihood inference analogous to case-control studies.

Main Results:

• Relative mean survival parameters can be estimated without survival data.
• Interaction parameters can be estimated from prevalent samples alone (case-only analysis).
• Propensity score and conditional exposure effects on survival are estimable.

Conclusions:

• Covariate data from incident and prevalent samples are sufficient for survival model parameter estimation.
• This method offers a powerful alternative when traditional survival data is unavailable.
• The approach extends to estimating interaction and exposure effects in survival models.