Statistical Methods for Case-Cohort Study Designs

Case-cohort designs combine the efficiency of nested case-control studies with the broader inference of full cohort analyses by selecting all individuals who experience an event (cases) together with a randomly drawn subcohort, irrespective of outcome. Core statistical approaches rely on weighted estimators—most notably inverse probability weighting—to correct for the unequal sampling of cases and non-cases. Cox proportional hazards models are adapted via specialised weights (Prentice weights) to estimate hazard ratios, while extensions handle time-dependent covariates, non-proportional hazards and multiple outcomes. Stratified or outcome-dependent sampling schemes, including two-phase and goodness-of-fit designs, improve efficiency by oversampling informative subjects based on auxiliary variables or preliminary risk models. Variance estimation exploits robust sandwich estimators or bootstrap methods for valid confidence intervals. Missing covariate data are addressed through multiple imputation strategies tailored to the design, leveraging full cohort information where available. Recent advances also propose pseudo-Poisson and pseudo-normal linear regression for direct risk ratio and risk difference estimation in binary outcomes, avoiding rare-disease assumptions inherent in logistic regression. Overall, these methods expand the analytic toolkit for large-scale epidemiological studies by balancing cost, complexity and statistical validity.

Research from Nature Portfolio

No recent Nature Portfolio content available.

Research from all publishers

Recent methodological work has introduced a goodness-of-fit two-phase sampling design for time-to-event outcomes that uses an external survival model to oversample subjects with poor fit, yielding unbiased hazard estimates with improved efficiency compared to standard case-cohort sampling. Another development proposes pseudo-Poisson and pseudo-normal linear regression techniques for direct estimation of risk ratios and risk differences in case-cohort studies with binary outcomes; by applying inverse-probability weights and incorporating auxiliary covariate information, these methods provide interpretable effect measures without relying on the rare-disease assumption. Furthermore, multiple imputation approaches have been adapted for supersampled nested case-control and case-cohort designs: by imputing expensive or missing covariates using both the subsample and full cohort data, analysts achieve substantial efficiency gains while preserving computational feasibility in large cohorts.

Statistical Methods for Case-Cohort Study Designs publication trend

The graph below shows the total number of articles in statistical methods for case-cohort study designs across all publications each year (not limited to Nature Index journals).

Technical terms

Case-cohort study: A cohort-based sampling design in which all cases are analysed alongside a randomly selected subcohort of the original cohort, enabling efficient estimation of exposure-outcome associations.

Subcohort: A random sample of the full cohort drawn at baseline, irrespective of future event status, which serves as controls and provides the risk set for analysis.

Inverse probability weighting (IPW): A technique that assigns weights equal to the reciprocal of sampling probabilities, correcting for unequal selection of subjects.

Prentice weights: Case‐cohort specific weights applied in Cox regression to account for the sampling design and to yield consistent hazard ratio estimates.

Multiple imputation: A method to handle missing covariate data by creating several complete datasets through imputation models and combining results to reflect uncertainty.

References

1. Goodness-of-fit two-phase sampling designs for time-to-event outcomes: a simulation study based on New York University Women’s Health Study for breast cancer. BMC Medical Research Methodology (2023).
2. Multiple Imputation of Missing Data in Nested Case-Control and Case-Cohort Studies. Biometrics (2018).
3. Risk Ratio and Risk Difference Estimation in Case-cohort Studies. Journal of Epidemiology (2022).
4. Use of multiple imputation in supersampled nested case‐control and case‐cohort studies. Scandinavian Journal of Statistics (2022).