Iranian Journal of

Background & Objectives: The case-cohort study is one of the youngest designs in epidemiology and some methodological aspects of it are still in debate. This study aimed at comparing the estimated hazard ratio, standard error, and interaction hazard ratio between the case-cohort and cohort studies for assessing the relationship between diabetes and cardiovascular diseases.

Methods: A total of 1701 men and 2253 women aged between 40 and 75 years were considered as the main cohort. Subcohort sampling was performed using simple random sampling with a sampling fraction of 0.3%. The hazard ratio of the cohort study was calculated using Cox regression model and the 3 methods of Prentice, Self-Prentice, and Barlow were used for calculating the hazard ratio of the case-cohort study. The mentioned regression models were used to assess the interactions.

Results: The results of the two studies were similar in populations with higher incidence (cohort of men) and lower incidence (the cohort of women) when frequency percent of exposure variable was greater than 10%. When the sample size of the initial cohort was less than 1250 subjects, discrepancies were observed between the results of the two studies. In addition, the standard error of the case-cohort study was higher than the cohort study. The results of both studies were similar in assessing the considered interactions.

Conclusion: The results are similar when the initial cohort sample sizes are sufficient. Meanwhile, unlike the percentage of exposure frequency, the outcome incidence has a negligible impact on the discrepancy between the results while the effect of the relative frequency of the exposure levels on the results discrepancy is noticeable.

Case-Cohort and nested case-control designs are the most well-known and cost-effective derivatives of prospective studies. Due to the popularity of these designs in recent years, the aim of this study is to provide a comprehensive and up-to-date description of them with respect to methodological issues such as sampling scheme, effect measure estimation, and analysis strategies. Overall, random sampling of controls in case-cohort and matching for time in nested design are considered their hallmarks. Eventually, decision on a better design depends on study objectives, sub-cohort selection, case and control determination, and method of data collection. It should be kept in mind that any particular study requires unique elaboration and delicate examination to choose from the designs.