An EM algorithm for nonparametric estimation of the cumulative incidence function from repeated imperfect test...

Area of Science:

• Biostatistics
• Epidemiology
• Medical Screening

Background:

• Event times in screening and surveillance are often interval censored.
• Imperfect screening tests introduce uncertainty into event time intervals.

Purpose of the Study:

• To develop a nonparametric maximum likelihood estimator for cumulative incidence functions using screening test data.
• To address challenges posed by interval censoring and imperfect screening tests.

Main Methods:

• An expectation-maximization algorithm was developed for estimating cumulative incidence.
• The algorithm features a closed-form solution for combined expectation and maximization steps.
• The method is computationally efficient and undemanding.

Main Results:

• A simulation study showed estimator bias approaches zero with large sample sizes.
• The proposed estimator demonstrated lower mean squared error compared to methods assuming perfect screening.
• The algorithm was successfully applied to cervical precancer follow-up data.

Conclusions:

• The developed expectation-maximization algorithm provides a robust method for analyzing interval-censored event time data with imperfect screening.
• This approach offers improved accuracy in estimating cumulative incidence functions in real-world screening scenarios.
• The findings have implications for surveillance and follow-up studies in various medical fields.