Estimating a population cumulative incidence under calendar time trends

Area of Science:

• Epidemiology
• Biostatistics

Background:

• Disease risk is often measured using age-specific cumulative incidence.
• Cohort studies with staggered entry and specific end dates commonly use Kaplan-Meier or Aalen-Johansen estimators.
• These methods are frequently applied to the total sample to describe disease risk.

Purpose of the Study:

• To highlight the limitations of standard estimators when calendar time influences disease risk.
• To propose alternative analytical methods for more accurate disease risk estimation.
• To provide useful measures of disease risk in observational studies with time trends.

Main Methods:

• Investigating the utility of proportional hazards models for extrapolating disease risk.
• Exploring age-specific cumulative incidence curves stratified by time of entry.
• Examining end-of-follow-up estimates across strata.
• Proposing a weighted average of end-of-follow-up estimates as a summary measure.

Main Results:

• Total sample Kaplan-Meier and Aalen-Johansen estimators may not accurately reflect general population risk if calendar time trends are present.
• Proportional hazards models can be used for extrapolation if the proportionality assumption holds.
• Stratified age-specific cumulative incidence curves or end-of-follow-up estimates offer more useful risk descriptions.
• A weighted average of end-of-follow-up estimates can serve as a valuable summary measure.

Conclusions:

• Calendar time trends diminish the usefulness of total sample estimators in observational studies with staggered entry and administrative censoring.
• Proportional hazards modeling or stratified analysis are recommended as superior alternatives.
• Accurate disease risk assessment requires methods that account for temporal influences.