Adult-onset diabetics have about twice the mortality rate as nondiabetics. A trial was mounted to determine whether screening for the presence of diabetes would lead to treatment that would reduce this difference. Among 19,226 residents of Great Britain not known to have diabetes, 15,089 were assigned at random to be invited for screening for blood glucose and HbA1c (a marker of longer-term glucose levels). Potential cases were evaluated by means of an oral glucose tolerance test, and those confirmed as being diabetic were entered into an aggressive treatment program. The remaining 4137 patients did not undergo screening.

During an average follow-up of 9.6 years, 3% of patients in the group assigned to be screened were diagnosed with diabetes. The primary outcome measure employed by the investigators, the all-cause mortality rate, was no lower in the intervention group (10.5 per 1000 person-years) than among patients in the control arm of the trial (9.9 per 1000 person-years, 95% CI of the relative mortality = 0.90–1.25).

a. One factor that limited the ability of the trial to document a health benefit from screening was the substantial fraction—27%—of persons in the intervention arm who did not respond to the invitation to be screened. What was (likely) an even more important factor? Explain.

b. The authors of the article reporting the results of this study acknowledged the failure of the intervention to lead to a decrease in all cause mortality in the screened population, and suggested that the benefits of screening may be restricted to the diabetics themselves.

From the data gathered in the study, it would be straightforward to compare all-cause mortality in persons after a diagnosis of diabetes in the intervention and control arms (p.348) of the trial.

However, even if there truly were no mortality reduction resulting from screening, such an analysis almost certainly would observe a lower all-cause mortality rate among intervention-arm diabetics than among the control-arm diabetics. Why?