Response bias in the Honolulu Heart Program.

The 14-year incidence rates (1969-1982) for coronary heart disease, cerebrovascular disease (stroke), total mortality, and cause-specific mortality were compared between 8,006 examined and 3,130 nonexamined men of the Honolulu Heart Program using identical surveillance procedures. There was a significant decrease in examination participation with increasing age. Examined men smoked less, weighed more, had a higher level of education, and had a lower percentage of never-married status than did nonexamined men. Total mortality rates, cancer mortality rates, and coronary heart disease incidence rates were higher in nonexamined men, while there were no differences in stroke rates. The average annual response error for total mortality and coronary heart disease rates was underestimated at 8.7% and 5.4%, respectively. The differences in rates were greatest during the first half of the follow-up period and converged during the second half. By the end of 10 years, there were no differences between nonexamined and examined men for any of the endpoints studied. The pattern of convergence of rates suggests a diminishing healthy participant advantage over time. In conclusion, a response bias did occur in this study, but the effect was small and did not alter any of the earlier findings concerning the relative incidence of cardiovascular disease. Because the degree of response bias can vary widely depending on when during follow-up a particular analysis is undertaken, it is recommended that prospective studies monitor, insofar as possible, a sample of nonparticipants in order to ensure valid results.