Purpose: Cohort participants usually have lower mortality rates than nonparticipants, but it is unclear if this survival advantage decreases or increases as cohort studies age.
Methods: We used a 1975 private census of Washington County, Maryland, to compare mortality among cohort participants to nonparticipants for three cohorts, Campaign Against Cancer and Stroke (CLUE I), Campaign Against Cancer and Heart Disease (CLUE II), and Atherosclerosis Risk In Communities (ARIC) initiated in 1974, 1989, and 1986, respectively. We analyzed mortality risk using time-truncated Cox regression models.
Results: Participants had lower mortality risk in the first 10 years of follow-up compared with nonparticipants (fully adjusted average hazard ratio [95% confidence intervals] were 0.72 [0.68, 0.77] in CLUE I, 0.69 [0.65, 0.73] in CLUE II, and 0.74 [0.63, 0.86] in ARIC), which persisted over 20 years of follow-up (0.81 [0.78, 0.84] in CLUE I, 0.87 [0.84, 0.91] in CLUE II, and 0.90 [0.83, 0.97] in ARIC). This lower average hazard for mortality among participants compared with nonparticipants attenuated with longer follow-up (0.99 [0.96, 1.01] after 30+ years in CLUE I, 1.02 [0.99, 1.05] after 30 years in CLUE II, and 0.95 [0.89, 1.00] after 30+ years in ARIC). In ARIC, participants who did not attend visits had higher mortality, but those who did attend visits had similar mortality to the community.
Conclusions: Our results suggest the volunteer selection for mortality in long-standing epidemiologic cohort studies often diminishes as the cohort ages.
Keywords: Cohort study; Mortality; Selection bias; Survival advantage.
Copyright © 2020 Elsevier Inc. All rights reserved.