Association of rotating shift work with incident irritable bowel syndrome: a large population-based prospective cohort study

The study involved over 500,000 participants from the UK Biobank, a large-scale prospective cohort. Participants were recruited from 22 assessment centers across England, Wales, and Scotland. At these centers, participants completed a touchscreen questionnaire, underwent a face-to-face interview, and underwent a series of physical measurements (23).

We excluded participants with a diagnosis of IBS and those with missing data of job involves shiftwork at baseline. Participants with the most common benign digestive system diseases with symptoms similar to IBS were also excluded, including patients with Crohn’s disease, ulcerative colitis and celiac disease (Supplementary Table S1). Finally, 268,290 participants were included in this study, as shown in Figure 1.

Only participants who were employed at the time of recruitment were considered in shift work. UK Biobank defined shift work as “a schedule falling outside of 9 am to 5 pm.” By definition, participants were classified as “never/rarely shift workers” or “shift workers” based on their current shift work status. Among shift workers, participants were further categorized as “always/usually shift workers” and “sometimes shift workers.”

An index score for the sleep quality was created including 5 aspects of sleep behaviors (24) and in this paper, daytime dozing is used as a substitute for daytime sleepiness. Healthy sleep factors were defined as morning chronotype (“definitely a morning person” or “more a morning than evening person”); adequate sleep duration (adequate); not usually insomnia (never/rarely); no snoring and no daytime dozing (never/rarely or sometimes). Each sleep factor was coded 1 if meeting the healthy criterion and 0 if not. All component scores were summed to obtain a healthy sleep score ranging from 0 to 5, with higher scores indicating a better sleep quality (Supplementary Table S2).

The primary outcome of the study was the incidence of IBS. The diagnosis of IBS was determined using self-report or linkage to death register and/or primary care and/or hospital admission data, based on International Classification of Disease-10 (ICD-10) codes of K58 (). Supplementary Table S1

To reduce the effect of potential confounding, sociodemographic characteristics, lifestyle factors, comorbidities, and markers of inflammation were considered to be covariates based on previous literature (16, 25).

The sociodemographic factors include: age (≤ 50 or > 50 years), sex (female or male), ethnicity (white or not white), Townsend Deprivation Index (< median (−2.09) or ≥ median), and BMI (< 25 or ≥ 25 kg/m²). Lifestyle factors variables include smoking status (never or previous), alcohol drinking(never or previous), tea intake (continuous variable), healthy diet score (26) (poor or ideal), optimal physical activity (yes or no: defined as meeting the 2017 UK Physical activity guidelines of 150 min of moderate activity per week or 75 min of vigorous activity), sleep duration (categorized as <6 and ≥ 6 h/day), chronotype preference (definitely a “morning” person, more a “morning” than “evening” person, more an “evening” than a “morning” person, and definitely an “evening” person), daytime dozing (all of the time/often, never/rarely, and sometimes), Snoring (yes or no), sleeplessness insomnia (never, sometimes and usually), anxiety/depression (normal or abnormal) (27). Comorbidities variables include: diabetes diagnosed by doctor (yes or no), high blood pressure diagnosed by doctor (yes or no); markers of inflammation variables include: SII, PLR, NLR, LMR (all categorized as high or low). Calculations were as follows: SII = (neutrophils * platelets)/lymphocytes, NLR = neutrophils/lymphocytes, PLR = platelets/lymphocytes, and LMR = lymphocytes/monocytes (25).

For baseline characteristics, continuous variables were presented as means and standard deviations, and categorical variables were summarized as number and percentages. Univariate comparisons between incident IBS cases and non-IBS were conducted using Student’s t-test or χ² test, respectively. The UK Biobank data were available up to 31 October 2022, which was considered the end of the study period. Person-years of follow-up were calculated from the baseline assessment until the occurrence of IBS, death, loss to follow-up, or the end of the study, whichever came first.

In the primary analysis, a time-to-event analysis of IBS was performed using a Cox proportional hazards model. After collinearity diagnosis, it can be concluded that there is no significant multicollinearity issue in the model. Then, we constructed three models with different covariates to estimate the hazard ratio (HR) and its 95% confidence interval (CI). Crude model was adjusted for nothing; Model 1 was adjusted for age, Townsend Deprivation Index, sex, ethnicity; Model 2 was further adjusted for optimal physical activity, BMI, smoking status, tea intake, alcohol drinking, healthy diet score, chronotype, sleep duration, sleeplessness insomnia, snoring, daytime dozing, high blood pressure diagnosed by doctor, diabetes diagnosed by doctor, anxiety/depression, LMR, PLR, NLR, SII based on Model 1.

As additional exploratory analyses, possible modifications were assessed for variables including age, sex, ethnicity, Townsend deprivation index, BMI, smoking status, alcohol drinking, diabetes diagnosed by doctor, high blood pressure diagnosed by doctor, healthy diet score, optimal physical activity, sleep duration, chronotype preference, sleeplessness insomnia, snoring, daytime dozing, anxiety/depression, LMR, PLR, NLR, SII. Potential modifying effects were assessed by modeling the cross-product term of the stratifying variable with shiftwork variable. The joint association between sleep quality and shiftwork with the risk of IBS were examined using Cox proportional hazard models, which was performed after full adjustment with the combination of never/sometimes insomnia-never/rarely shiftwork or adequate sleep duration-never/rarely shiftwork as the reference. Furthermore, mediation analysis was performed to investigate the mediating effect of anxiety/depression and sleep quality on the shiftwork-IBS association.

Multiple imputation was used for missing data to obtain the final dataset. All p values were reported as two-sided tests with significance defined as p < 0.05. Statistical analyses were performed in the R software (Version 4.3.2, R Core Team).1

Table 1 presents the baseline characteristics of the study participants who being in paid employment or self-employment at baseline. Among the 268,290 participants (mean age 52.77 years, 50.9% female), 26,609 (9.92%) participants had always/usually shiftwork behaviors, 221,709 (82.64%) had never/rarely shiftwork behavior, 19,972 (7.44%) participants had sometimes shiftwork behaviors.

Compared with participants without incident IBS, those with incident IBS were more likely to be younger, female, to have less exercise, work more shifts, have shorter sleep duration, more likely to be an “evening person,” usually suffer from insomnia, do not snore, always be sleepy during the day, and had higher PHQ-4 and SII and LMR scores.

During a mean follow-up of 13.7 years, a total of 5,359 (1.997%) participants developed IBS. Shiftwork status was classified as always/usually, sometimes and never/rarely. Compared to participants with never/rarely rotating shift work, those adherence to always/usually rotating shift work had a significantly higher risk of developing IBS, (HR: 1.12, 95% CI: 1.03–1.23; Table 2), but no significant associations were found with other intestinal diseases (for Crohn, (HR: 1.20, 95% CI: 0.93–1.60); for intestinal vascular disease, (HR: 1.20, 95% CI: 0.96–1.40); for ulcerative colitis, (HR: 1.10, 95% CI: 0.94–1.30); and for colorectal cancer, (HR: 1.01, 95% CI: 0.89–1.15), respectively; Supplementary Table S3). Although for paralytic intestinal obstruction the association is significant, (HR: 1.20, 95% CI: 1.10–1.30), the statistical power is not sufficient to indicate an association between them.

In subgroup analysis, we found significant association between rotating shift work and the risk of IBS among participants with sometimes or usually sleeplessness insomnia and both sexes (for never/rarely sleeplessness insomnia, (HR: 0.91, 95% CI: 0.82–1.01); for sometimes sleeplessness insomnia, (HR: 1.07, 95% CI: 1.00–1.13); for usually sleeplessness insomnia, (HR: 1.09, 95% CI: 1.02–1.17); for males, (HR: 1.10, 95% CI: 1.02–1.17); and for females, (HR: 1.08, 95% CI: 1.03–1.14), respectively), shiftwork interacted with sleeplessness insomnia on the risk of IBS (P for interaction = 0.011). Significant modification effects by high blood pressure diagnosed by doctor and level of PLR were also detected when using per SD change (P for interaction = 0.022 for High blood pressure diagnosed by doctor and P for interaction = 0.034 for level of PLR). Additionally, shiftwork interacted with BMI on the risk of IBS (P for interaction = 0.028), with the risks generally observed among participants whose BMI ≥ 25 kg/m² but not BMI < 25 kg/m² (Figure 2).

Moreover, the increased IBS risk (not significant) associated with shiftwork was generally observed across TDI, sex, age, BMI, optimal physical activity, smoking status, alcohol drinking, healthy diet score, sleep duration, snoring, daytime dozing, high blood pressure diagnosed by doctor, diabetes diagnosed by doctor, anxiety/depression, LMR, PLR, NLR, SII. The risk of IBS did not differ based on chronotype, except for definitely a ‘morning’ person subgroup (Figure 2).

In order to assess the robustness of our findings, we conducted several sensitivity analyses. We did not find significant changes of the impact of shift work on IBS risk after excluding subjects with follow-up time < 1 year from baseline (HR: 1.14, 95% CI: 1.04–1.24) or the model of the dataset excluding non-White participants (HR: 1.15, 95% CI: 1.05–1.26). The results showed no substantial change of the impact of shift work on IBS (,). Supplementary Tables S4 S5

After including sleep duration and shift work interaction terms in the cox proportional hazard models, we found evidence of a joint association of sleep duration and rotating shift work with the incidence of IBS. Compared to those with both adequate sleep duration and never/rarely shift work, participants with both adequate sleep duration and sometimes shift work was positively associated with risk of IBS (HR: 1.02, 95% CI: 0.91–1.14), although the statistical significance is not significant. Participants with both adequate sleep duration and always/usually shift work had a 12% increased risk of IBS (HR: 1.12, 95% CI: 1.02–1.23), 35% increased risk (HR: 1.35, 95% CI: 1.19–1.52) for those with both inadequate sleep duration and never/rarely shift work, and 45% increased risk (HR: 1.45, 95% CI: 1.07–1.97) for those with both inadequate sleep duration and sometimes shift work. Of note, those with both inadequate sleep duration and always/usually shift work had a 54% increased risk of IBS (HR: 1.54, 95% CI: 1.35–1.82), which is the highest risk of all the combinations (Figure 3a).

Similarly, Figure 3b illustrated the HR for each combination of sleeplessness insomnia and rotating shift work interaction terms. Compared to those with never/sometimes insomnia and never/rarely shift work, participants with both never/sometimes insomnia and always/usually shift work was positively associated with risk of IBS (HR: 1.10, 95% CI: 0.98–1.22), although the statistical significance is not significant. Participants with both usually insomnia and never/rarely shift work had a 41% increased risk (HR: 1.41, 95% CI: 1.32–1.50), and 59% increased risk (HR: 1.59, 95% CI: 1.32–1.88) for those with both usually insomnia and sometimes shift work. It is worth noting that those with both usually insomnia and always/usually shift work had the highest increased risk of IBS, which is 65% (HR: 1.65, 95% CI: 1.43–1.90) higher than those with never/sometimes insomnia and never/rarely shift work (Figure 3b).

We observed significant mediating effects by healthy sleep and anxiety/depression of the association between job involves shift work and IBS incidence (Table 3). The mediation analysis showed that the association between shift work and IBS was partially mediated by sleep health (Mediation proportion = 16.10%, p < 0.001) with pure natural indirect effects of 0.011 (0.009–0.014) and anxiety/depression (Mediation proportion = 3.60%, p = 0.029) with pure natural indirect effects of 0.002 (0.001–0.003).

First, a large number of participants provided a detailed employment history, allowing us to categorize the frequency of shifts, thus overcoming limitations of many previous studies. Second, individuals in the UK Biobank were recruited entirely independent of employment status, ensuring that the sample was representative of the entire UK workforce. This approach effectively mitigated the potential selection bias that could arise from studying a single occupation. Third, large-scale prospective cohort studies have explored the joint effect of sleep and shift work, and found that there is a synergistic effect between them, which has enriched the research on related mechanisms. Meanwhile, combined with inflammatory markers, the potential mechanism of physiological circadian rhythm and IBS pathogenesis was explored. Furthermore, we were able to explore the potential mediation pathway by sleep health and anxiety/depression in this association.

Firstly, despite the substantial sample size, our analysis is confined to adults aged 40 to 69 years, as the UK Biobank does not provide data on younger individuals and is predominantly composed of a White population. Consequently, the findings may necessitate validation across other age groups and racial or ethnic backgrounds. Secondly, only the baseline shift schedules were counted, but in reality, participants’ work status might change over time during the follow-up while people tend to stop doing shift or night shift work at an older age, which might bias our results toward the null hypothesis, resulting in an underestimation of the effect size. Thirdly, IBS might be misdiagnosed or underdiagnosed, hence some IBS cases might not be captured by EHRs. Fourthly, the Rome criteria, which define disorders of gut-brain interaction, are extensively applied in epidemiologic research, pathophysiologic studies, treatment trials, and clinical practice (38). However, diagnoses in the UK biobank are coded using ICD-10, and we did not have access to information on diseases diagnosed using the ROME criteria. Thus, this is one of the limitations of our study.

The datasets presented in this study can be found in online repositories. The names of the repository/repositories and accession number(s) can be found in the article/. Supplementary material