STUDY OBJECTIVES:To evaluate the effect of an employer-mandated obstructive sleep apnea (OSA) program on the risk of serious preventable truck crashes. METHODS:Data are from the first large-scale, employer-mandated program to screen, diagnose, and monitor OSA treatment adherence in the US trucking industry. A retrospective analysis of cohorts was constructed: polysomnogram-diagnosed drivers (OSA positive n = 1,613, OSA negative n = 403) were matched to control drivers unlikely to have OSA (n = 2,016) on two factors affecting crash risk, experience-at-hire and length of job tenure; tenure was matched on the date of each diagnosed driver's polysomnogram. Auto-adjusting positive airway pressure (APAP) treatment was provided to all cases (i.e. OSA positive drivers); treatment adherence was objectively monitored. Cases were grouped by treatment adherence: "Full Adherence" (n = 682), "Partial Adherence" (n = 571), or "No Adherence" (n = 360). Preventable Department-of-Transportation-reportable crashes/100,000 miles were compared across study subgroups. Robustness was assessed. RESULTS:After the matching date, "No Adherence" cases had a preventable Department of Transportation-reportable crash rate that was fivefold greater (incidence rate ratio = 4.97, 95% confidence interval: 2.09, 10.63) than that of matched controls (0.070 versus 0.014 per 100,000 miles). The crash rate of "Full Adherence" cases was statistically similar to controls (incidence rate ratio = 1.02, 95% confidence interval: 0.48, 2.04; 0.014 per 100,000 miles). CONCLUSIONS:Nontreatment-adherent OSA-positive drivers had a fivefold greater risk of serious preventable crashes, but were discharged or quit rapidly, being retained only one-third as long as other subjects. Thus, the mandated program removed risky nontreatment-adherent drivers and retained adherent drivers at the study firm. Current regulations allow nonadherent OSA cases to drive at another firm by keeping their diagnosis private. COMMENTARY:A commentary on this article appears in this issue on page 961.