The Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) 2020 statement ⁹ was followed in the conduct and reporting of this systematic review. Supplementary Table 1 contains the PRISMA 2020 checklist. This review was prospectively registered with PROSPERO prior to the search (CRD420261378070), and the specified process was followed exactly.

As mechanistic proxies for circadian disturbance linked to night shift scheduling, sleep-related variables (sleep duration and quality) were incorporated as a pre-specified secondary exposure category. They are clearly marked as such throughout and were examined independently from direct shift-work exposure studies. Several publications from the same cohort were not considered distinct populations, but rather connected datasets reflecting consecutive follow-up analyses.

The following broad research question served as the basis for this systematic review: Among working-age adults employed in healthcare, healthcare-adjacent, or industrial settings, does exposure to night shift work or its mechanistic proxy of sleep disturbance — independently increase the risk of subclinical vascular dysfunction and structural arterial changes as measured by objective, instrument-based vascular parameters?

Studies were eligible if they: (i) used a cross-sectional, cohort, or case-control observational design; (ii) enrolled adults (≥18 years) working in healthcare, healthcare-adjacent, or industrial settings with verifiable night shift exposure; (iii) defined night shift work as at least three hours of work between 22:00 and 06:00 on a regular or rotating basis; and (iv) reported at least one objective, instrument-measured vascular parameters, such as CIMT, PWV (any method), AIx, RHI, carotid plaque presence or total plaque area, or cerebrovascular imaging metric. Only English-language research papers were considered. Case reports, reviews, and editorials were not used as primary sources of evidence.

Since sleep disturbance is the most direct and quantifiable physiological effect of working nights, the pre-specified inclusion of sleep-proxy studies was warranted. Due to the fundamental difference in population type, one study ¹⁰ that examined sleep architecture in clinical ICA stenosis patients was only included as indirect mechanistic evidence; it was not included in the primary evidence synthesis and is not eligible for JBI appraisal.

An electronic search was carried out in PubMed/MEDLINE, Embase, Web of Science, Scopus, and CINAHL. For every database, the following Boolean search string was used and modified:

Filters applied: English language; humans. Reference lists of all included studies and relevant reviews were hand-searched.

A full-text evaluation of possibly eligible records was conducted after titles and abstracts were evaluated by two independent, blind reviewers. A third reviewer or consensus was used to settle disagreements. The research selection procedure is summarized in the PRISMA 2020 flow diagram ( Figure 1). First author and year, country, study design, population characteristics (occupation, n, age, sex), exposure definition and metric, vascular outcome(s) and measurement tool, adjusted effect estimate with 95% confidence interval (CI) when available, confounders adjusted for, and JBI appraisal score were all recorded on a standardized extraction form.

The JBI Critical Appraisal Checklist for Analytical Cross-Sectional research (8 criteria) and the JBI Critical Appraisal Checklist for Cohort Studies (11 criteria), if applicable, were used to evaluate the risk of bias in individual research. There were four possible ratings for each criterion: yes (Y), no (N), uncertain (U), and not applicable (NA). Studies were categorized as poor quality (<50%), moderate quality (50–74%), or high quality (≥75%). Independent appraisals were conducted by two reviewers; differences were settled by consensus.

In order to give a comparator quality measure (0–9 stars) that allowed for domain-level dissection across the Selection, Comparability, and Outcome/Exposure domains, the Newcastle-Ottawa Scale (NOS) was also used. Studies with a score of at least six were considered to be of excellent quality. Self-reported shift work exposure (information bias), healthy worker survivor bias (selection bias), cross-sectional design that precludes causal inference, and diverse CIMT assessment techniques were among the major concerns found throughout the evidence base.

Database searches (n = 2,253) and additional sources (n = 4) yielded a total of 2,257 records, as shown in Figure 1. 1,626 records remained after duplicates (n = 631) were eliminated, and they were filtered using the abstract and title.1,616 records were eliminated during the screening phase for the following reasons: lack of vascular outcomes (n = 673); unrelated to night shift work or exposure mismatch (n = 312); review articles, editorials, or conference abstracts (n = 238); animal or in vitro studies (n = 174); non-English publications (n = 121); irrelevant population (n = 62); and inadequate or incomplete data (n = 36). Ten full-text publications were then evaluated for eligibility; no additional exclusions were made.

Ten papers were ultimately included in the study as they satisfied all inclusion requirements.

Five vascular outcome domains were covered by the ten included studies: arterial stiffness/PWV (three studies; two from the same cohort), endothelial function/RHI (one study), CIMT (six studies; overlap with sleep proxy studies), carotid plaque/total plaque area (one study), and cerebrovascular white matter integrity (one study). Only the indirect mechanistic evidence from one other research ¹⁰ was kept. Table 1 summarizes the features of the study. Interestingly, two of the included studies ^{11 , 12} are not regarded as independent replications because they come from the same Norwegian industrial cohort.

Detailed study characteristics and risk of bias assessments are provided in Table 2 and Table 3. Overall, the included studies were of moderate-to-high methodological quality based on JBI and NOS assessments, although common limitations included cross-sectional design, self-reported exposure, and potential residual confounding.

The included studies demonstrated generally consistent relationships between night shift exposure, or its sleep-disturbance proxy, and adverse subclinical vascular indicators across five vascular outcome domains: arterial stiffness, endothelial function, carotid intima-media thickness, carotid plaque burden, and cerebrovascular white matter integrity. Although they are suggestive of potential cumulative vascular effects, the majority of data come from cross-sectional designs, which record group-level changes rather than longitudinal development and cannot demonstrate causal directionality.

Different studies employed different measurement instruments. Pulse Wave Velocity (PWV) and Augmentation Index (AIx) are complementary measures of arterial stiffness — PWV quantifies the speed of pressure wave propagation along the arterial wall, while AIx reflects the contribution of reflected waves to central aortic pressure; higher values of both indicate stiffer vessels. The Stiffness Index (SI), obtained from finger photoplethysmography, is an easier method for estimating large artery stiffness that can be used in large-scale population studies. The Reactive Hyperemia Index (RHI), obtained using peripheral arterial tonometry (EndoPAT), measures endothelial function in microvessels, while low values suggest reduced vasodilation. The carotid intima-media thickness (CIMT) measurement with B-mode ultrasonography is a well-known structural measure of atherosclerotic disease, although there is inconsistency across studies due to differences in anatomical localization, number of segments measured, and instrumentation. Total Plaque Area (TPA) takes this further by calculating the total cross-sectional area of all carotid plaques. Peak Skeletonized Mean Diffusivity (PSMD) is an advanced structural imaging parameter using diffusion tensor MRI.

The results on arterial stiffness were the most reliable. In the GHS study (n = 8,065), a dose-dependent effect was shown, where the stiffness index increased from 0.33 m/s after more than 660 night shifts (p < 0.05). On the longitudinal level, Skogstad et al. (2023a) reported an increase in PWV of 1.29 m/s in shift workers compared to 0.11 m/s in day workers over three years, with the follow-up studies showing that lowering shift work frequency slows down this process (Skogstad et al., 2023b – same sample size). The only result on RHI was observed in the GHS study, which showed worse endothelial function (−0.054 points for lifetime years of night work; p < 0.05).

CIMT findings were the most heterogeneous. Wang et al. ¹³ found elevated CIMT in shift workers that was attenuated after full adjustment, while sleep-proxy studies showed more consistent signals — worse subjective sleep quality independently predicted CIMT in resident doctors (Agudelo et al. ¹⁴ ), objective short sleep inversely correlated with CIMT in men (Sands et al. ¹⁵ ), and a J-shaped relationship emerged between sleep duration and CIMT in the SHIP cohort (Wolff et al. ¹⁶ ), with both short (<6 h) and long (>8 h) sleep associated with higher values. Agudelo et al. similarly identified a J-shaped pattern for TPA and plaque presence. Finally, Rimmele et al. ¹⁷ found greater PSMD in former shift workers after adjustment, with the persistence of this signal beyond exposure cessation suggesting that cerebrovascular changes may not fully reverse following shift-work discontinuation.