[50] Zheng H, Patel M, Hryniewicz K, Katz SD. Association of extended work shifts, vascular function, and inflammatory markers in internal medicine residents: a randomized crossover trial. JAMA. 2006 Sep 6;296(9):1049–1050. doi: 10.1001/jama.296.9.1049. PMID: 16954481. RefID-595 PMID-16954481
Study Setting	Clinical research laboratory, Yale University, New Haven, CT
Eligible Participants	Internal Medicine Residents who did not consume caffeine and other unspecified medications
Study Aims	The study aims to test the hypothesis that sleep loss during extended work shifts is associated with evidence of vascular inflammation and dysfunction.
Study Design & Participants	A prospective single-blind, crossover design during an intensive care unit rotation. Participants were assigned at random to two study sessions in random order. One of these sessions was conducted at 1 p.m. after completion of a 30 h extended work shift (from 7 a.m. until 1 p.m. the following day). The other session was conducted at 1 p.m. after completion of a 6 h nonextended work shift (from 7 a.m. until 1 p.m. on the same day). Participants fasted from 7 a.m. to 1 p.m. on the study day.
Intervention	Participants completed a 30 h extended work shift and a 6 h non-extended work shift. After each session, sleep hours from the past night were recorded with a written diary. Flow-mediated dilation of the brachial artery was obtained non-invasively using ultrasonography.
Comparison(s)	FMD measurements taken after the 30 h shift were compared to those taken after the 6 h shift.
Outcome Measure(s)	Sleep duration was self-reported with a written diary. FMD was measured using high-resolution ultrasound imaging.
Key Finding(s)	Sleep duration was found to differ significantly between the extended work group and the non-extended work group, with a median of 0.3 h and 6.5 h, respectively. The p-value for this comparison was <0.001. FMD was also found to differ significantly between the extended work group and the non-extended work group, with a median of 3.2% and 7.9%, respectively. The p-value for this comparison was also <0.001.