The protocol for this systematic review and meta-analysis was registered to PROSPERO (CRD420251060086). This article followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines (23).

We comprehensively searched six electronic databases, including PubMed, Embase, Cochrane Library, Web of Science, CINAHL, and Scopus, conducted from inception to August 10, 2025, without language restrictions. The search strategy entailed a combination of exposure-related factors (“shift work,” “night work,” “Shift Work Schedule,” “night shift work,” “rotating shift”) and a mixture of Medical Subject Headings (MESH) terms and keywords pertinent to outcomes (“Cardiovascular Diseases,” “coronary heart disease,” “Myocardial Ischemia,” “ischemic heart disease,” “stroke,” “cardiovascular events,” “cardiovascular mortality”) (Detailed search strategy can be found in the Supplementary Table S1). Additionally, a manual search of the reference lists from the retrieved articles was conducted to identify any additional relevant studies.

Inclusion criteria (based on PICO framework): P (Population): The studies examined outcomes in populations engaged in night shift work (including shift work schedules involving night shifts); I (Intervention): Any type of night shift work as the exposure variable (rotating shift work, fixed night shift work); C (comparisons): Daytime work exclusively, with no history of night shift work; O (Outcome): The outcomes focused on cardiovascular events (including both incidence and mortality) and encompassed major CVD subtypes such as coronary heart disease (CHD), ischemic heart disease (IHD), and stroke, and reported the provision of the 95% confidence intervals (CI), relative risk (RR), odds ratios (OR), hazard ratios (HR). When multiple publications utilized the same cohort, we prioritized the study with the most comprehensive follow-up, complete outcome data, or superior methodological quality to avoid double-counting of participants.

Exclusion criteria included: (1) studies with self-reported cardiovascular outcomes or those reporting only cardiovascular markers; (2) non-human studies; (3) reviews, case reports, letters, or conference abstracts; (4) incomplete data or inability to access full-text articles; (5) duplicate publications; and (6) case reports, systematic reviews, or meta-analyses.

Literature management and screening were conducted using EndNote 20. The literature screening procedure was carried out in two phases. Initially, two authors (XJY and MWL) assessed the eligibility of studies by evaluating the titles and abstracts according to the predefined inclusion and exclusion criteria. The second phase entailed a thorough examination of the full texts to ascertain which studies qualified for inclusion in the meta-analysis. In cases of disagreement during the literature screening, the issues were deliberated with a third author (TYM) until a consensus was achieved.

Data regarding results and study characteristics were extracted to tables by the first author (XJY) and checked for accuracy by the second author (MWL). Microsoft Excel was used to extract the required data, including lead author, year of publication, location by country, sample source, study design, age, sex, sample size, follow-up duration, definition of night shift work, type of night shift work, occupation, Outcome, outcome assessment method, adjusted effect estimates (OR, RR, and HR) with 95% CI and a list of adjusted factors.

Two authors (XJY and MWL) independently assessed the quality of the studies using the nine-star Newcastle–Ottawa Scale (NOS) (24), which evaluates three domains: selection (0–4 points), comparability (0–2 points), and outcomes (0–3 points). Total scores were categorized into three groups: 0–3 (fair), 4–6 (moderate), and 7–9 (good). Inter-rater agreement in study quality assessment was quantified using Cohen’s kappa coefficient. According to Cohen’s kappa coefficient, inter-rater agreement was categorized into three levels: coefficients ≤0.60 indicated poor agreement, values between 0.61 and 0.80 represented moderate agreement, and coefficients ≥0.81 signified good agreement (25). Any disagreements were resolved through consensus discussion with the third author (TYM).

Analysis of statistics performed with STATA 18.0. The RR was considered as the common measure of the association between night shift work and CVD. The HRs were deemed equivalent to RRs because HRs were considered to approximate RRs (26, 27). The ORs were transformed into RRs using the formula RR=OR(1−P0)+(P0×OR), where P₀ indicates the incidence of the outcome of interest in the non-exposed group, and then synthesized with the RRs and the HRs into pooled RRs (28). For each included study, we extracted covariate-adjusted outcome measures. When multiple effect estimates were reported within a single study, these were first pooled at the study level prior to the overall meta-analysis. Heterogeneity across studies was evaluated using Cochran’s Q statistic test and I² statistic (29, 30). A random-effects model was applied if significant heterogeneity was detected (p-value <0.05 or I² > 50%); otherwise, a fixed-effects model was utilized (31). Sensitivity analysis was conducted by sequentially excluding individual studies to assess the robustness of the pooled results. At least three quantitative categories (involving the duration of night shift work) were reported in the included studies and were incorporated into the dose-response meta-analysis. The dose-response relationship model proposed by Greenland and Longnecker (32) was adopted to calculate the trend based on the estimated values of logRR in different shift work categories. According to this method, the distribution of cases and participants, as well as RR and 95% confidence intervals, were extracted for each shift work category. Additionally, we used restricted cubic splines with three knots set at the 10th, 50th, and 90th percentiles of the distribution to assess the potential curvilinear association between years of night shift work and the risk of CVD incidence and mortality. In addition, subgroup analyses were conducted to stratify the results on cardiovascular outcome type, sex, type of night shift work, and geographic region. Publication bias was examined through visual inspection of funnel plots and quantified using Egger’s test, if there was a large publication bias, correction is performed using trim-and-fill method (33, 34).

We assessed the strength of evidence for each outcome using the Grading of Recommendations Assessment, Development, and Evaluation (GRADE) framework. The evaluation included the following domains: risk of bias, inconsistency, indirectness, imprecision, and other considerations. Evidence quality was categorized into four levels: Very Low, Low, Moderate, and High.

The initial database search identified 5,510 articles: 860 from PubMed, 1,107 from Web of Science, 1835 from Embase, 232 from Cochrane Library, 217 from CINAHL, and 1,259 from Scopus. After removing 2,237 duplicates, 3,273 unique records underwent title and abstract screening. Of these, 154 articles were deemed eligible for full-text review. Following detailed evaluation, 131 studies were excluded based on predefined criteria: (1) non-human studies; (2) cross-sectional design; (3) irrelevant outcomes; (4) duplicate data from overlapping cohorts; (6) insufficient data availability. Ultimately, 23 articles met all inclusion criteria and were included in the meta-analysis (selection process detailed in Figure 1).

This meta-analysis incorporated 23 observational studies comprising 20 prospective cohort studies (13, 14, 19–22, 35–48), 3 retrospective cohort studies (49–51), with a total population of 3,340,377 participants. Geographically, 16 studies were conducted in Europe (13, 14, 19–22, 36, 37, 41–43, 45–47, 50, 51), 3 in North America (35, 38, 40), 1 in Oceania (48), and 3 in Asia (39, 44, 49). Regarding occupational distribution, 7 studies specifically focused on healthcare workers (19, 20, 22, 35, 38, 40, 42), while the remaining studies encompassed diverse occupations, predominantly involving industrial workers.

The included studies investigated associations between night shift work and CVD incidence/mortality risks, primarily targeting the following outcomes: CVD, CHD, IHD, and stroke. Case ascertainment was primarily based on local hospital admission records and death certificates. Definitions of night shift work varied across studies: most involved rotating night shifts, whereas others assessed fixed night shifts. Study characteristics are summarized in Table 1.

Supplementary Table S2 shows the quality assessment of the included studies. All 23 studies were rated as high-quality based on the NOS, with scores ≥7. High methodological quality was primarily manifested in cohort selection, ascertainment of night shift work exposure, and assessment of CVD outcomes. Some potential biases were noted, primarily arising from inadequate follow-up and comparability limitations. Inter-rater reliability was assessed using Cohen’s kappa coefficient. The analysis yielded a kappa value of 0.857 (p < 0.001), indicating good agreement between the two assessors in quality evaluation.

Fifteen studies reported the association between night shift work and the risk of total CVD events, with a total of 2,891,280 participants (13, 14, 19–22, 35–38, 44–46, 48, 49). Due to the absence of significant heterogeneity among the studies (I² = 37.17%, p = 0.07), a fixed-effect model was selected for analysis. The summary RR (95% CI) was 1.13 (1.10–1.16) (Figure 2), suggesting that night shift work was associated with a higher risk of adverse cardiovascular events compared to daytime work.

Nine studies reported the association between night shift work and the risk of total CVD mortality, with a total of 638,255 participants (38–43, 47, 50, 51). Given the absence of significant heterogeneity across studies (I² = 14.88%, p = 0.31), a fixed-effect model was selected for analysis. The summary RR (95% CI) was 1.27 (1.18–1.36) (Figure 3), suggesting that night shift work was associated with a higher risk of adverse cardiovascular mortality outcomes compared to daytime work.

Studies with ≥3 categories of shift work duration were included in the dose-response analysis. Results demonstrated a linear dose-response relationship between night shift work and the risk of CVD incidence and mortality. For each 5-year increase in shift work duration, the risk of CVD incidence increased by 7% (RR = 1.07, 95% CI: 1.04–1.09), and the risk of CVD mortality increased by 4% (RR = 1.05, 95% CI: 1.03–1.06) (Figure 4a). Using restricted cubic splines, we did not detect a significant nonlinear dose-response association between shift work duration and the risk of CVD incidence (p = 0.068) or mortality (p = 0.096) (Figure 4b).

In subgroup analyses examining the association between night shift work and CVD events, night shift workers exhibited significantly higher risks of CHD (RR = 1.22, 95% CI = 1.16–1.28) and IHD (RR = 1.09, 95% CI = 1.05–1.14) compared to daytime workers. However, no statistically significant associations were observed for CVD (RR = 1.24, 95% CI = 0.95–1.60) or stroke (RR = 1.06, 95% CI = 0.95–1.18). In contrast, subgroup analyses focusing on CVD mortality revealed elevated risks among night shift workers for CHD (RR = 1.22, 95% CI = 1.10–1.36), CVD (RR = 1.35, 95% CI = 1.10–1.66), IHD (RR = 1.39, 95% CI = 1.06–1.84), and stroke (RR = 1.49, 95% CI = 1.04–2.12) (Figures 5, 6).

Additionally, night shift work patterns were significantly associated with both cardiovascular event incidence and mortality: fixed night work (RR = 1.34, 95% CI = 1.04–1.72) and rotating work (RR = 1.15, 95% CI = 1.10–1.21) were positively correlated with CVD incidence, while fixed night work (RR = 1.77, 95% CI = 1.28–2.46) and rotating work (RR = 1.25, 95% CI = 1.16–1.35) were associated with increased CVD mortality. Sex-specific analyses indicated a 18% higher risk of CVD events in female (RR = 1.18, 95% CI = 1.11–1.26) and a 20% higher risk in male (RR = 1.20 95% CI = 1.07–1.36) among night shift workers. For CVD mortality, both male (RR = 1.27, 95% CI = 1.07–1.45) and female (RR = 1.27, 95% CI = 1.17–1.38) showed statistically significant increases in risk. Geographic region subgroup analyses further demonstrated that night shift workers from Asia, Europe, North America, and Oceania are all significantly associated with the incidence and mortality risk of CVD (Tables 2, 3).

In this study, a sensitivity analysis was conducted on the association between night shift work and the risk of CVD incidence (Supplementary Figure S1) and mortality (Supplementary Figure S3). The application of sensitivity analysis confirmed the reliability of our meta-analysis results.

In this study, we used the Egger test and funnel plot to assess publication bias in the included studies. The analysis showed that the data distribution was slightly skewed in the association analysis between night shift work and overall CVD incidence risk (Supplementary Figure S2). The Egger test result (p = 0.025 < 0.05) indicated potential publication bias. However, the trim and fill method indicated that the impact of publication bias on the results was minimal. Imputation of six hypothetical studies showed no substantial alteration in the effect estimate (Supplementary Figure S2). This confirms the robustness of the primary conclusions. In the analysis of the association between night shift work and overall CVD mortality risk, the data distribution was generally symmetrical and relatively concentrated (Supplementary Figure S4). The Egger test results (p = 0.375 > 0.05) indicated no significant bias.

GRADE evaluation indicated predominantly low to very low certainty evidence (Supplementary Table S3). Given the observational cohort designs, initial certainty was low. Evidence for CVD incidence was downgraded to very low due to publication bias, while CVD mortality maintained low certainty absent downgrading factors. All other outcomes were rated very low, primarily because of imprecision and publication bias risks from limited studies.