Association Between Night-Shift Work and Cancer Risk: Updated Systematic Review and Meta-Analysis

Background: Nightshift work introduces light at night and causes circadian rhythm among night workers, who are considered to be at increased risk of cancer. However, in the last 2 years, nine population-based studies reported insignificant associations between night-shift work and cancer risks. We aimed to conduct an updated systematic review and meta-analysis to ascertain the effect of night-shift work on the incidence of cancers. Methods: Our protocol was registered in PROSPERO and complied with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA). Embase, PubMed, and Web of Science databases were used to comprehensively search studies published up to May 31, 2019. The random-effect model (Der Simonian-Laird method) was carried out to combine the risk estimates of night-shift work for cancers. The dose-response meta-analysis was performed to verify whether the association was in a dose-dependent manner. Results: Our literature searching retrieved 1,660 publications. Included in the meta-analyses were 57 eligible studies with 8,477,849 participants (mean age 55 years; 2,560,886 men, 4,220,154 women, and 1,696,809 not mentioned). The pooled results showed that night-shift work was not associated with the risk of breast cancer (OR = 1.009, 95% CI = 0.984–1.033), prostate cancer (OR = 1.027, 95% CI = 0.982–1.071), ovarian cancer (OR = 1.027, 95% CI = 0.942–1.113), pancreatic cancer (OR = 1.007, 95% CI = 0.910–1.104), colorectal cancer (OR = 1.016, 95% CI = 0.964–1.068), non-Hodgkin's lymph (OR = 1.046, 95% CI = 0.994–1.098), and stomach cancer (OR = 1.064, 95% CI = 0.971–1.157), while night-shift work was associated with a reduction of lung cancer (OR = 0.949, 95% CI = 0.903–0.996), and skin cancer (OR = 0.916, 95% CI = 0.879–0.953). The dose-response meta-analysis found that cancer risk was not significantly elevated with the increased light exposure of night- shift work. Conclusion: This systematic review of 57 observational studies did not find an overall association between ever-exposure to night-shift work and the risk of breast, prostate ovarian, pancreatic, colorectal, non-Hodgkin's lymph, and stomach cancers.


INTRODUCTION
Night-shift work is increasingly frequent among both full-time and part-time workers worldwide. Night-shift workers have to face the biological challenges of work shifts, light at night and altered circadian rhythm cycles. These challenges, as well as alterations in daily life and activity may introduce potential harms to night workers. In different employment sectors today, the number of people working overtime or on a night shift has been increasing, especially in transportation, health care, and manufacturing (1). Surveys of Americans, Europeans and Australians have shown that 15-30% of adults were engaged in shift work experience, and more than 30% of them fell asleep at work at least once a week (2). Apart from an increased risk of work-related injury, night-shift workers have a greater chance of having long-term disorders. Currently, epidemiological evidences indicated that night-shift work is recognized to be associated with increased susceptibilities to cancer (3,4).
The International Agency for Research on Cancer (IARC) has defined that night-shift work is probably carcinogenic to humans (IARC Group 2A) (5). Further studies have proposed the followings as a potential mechanism of carcinogenicity of night-shift work as: (1) circadian rhythm disruption, (2) melatonin suppression due to exposure to light at night, (3) physiological changes, (4) lifestyle disturbances, and (5) decreased vitamin D levels (resulting from lack of sunlight) (6). However, studies focusing on the association between night-shift work and cancer risks have reached contradictory conclusions. Even though several systematic reviews and meta-analyses have been conducted, they presented inconsistent findings (7). Nine reviews reported that night work may be positively associated with breast cancer, skin cancer, prostate cancer, colorectal cancer, and lung cancer (2,(8)(9)(10)(11)(12)(13)(14)(15), while four reviews reported slightly elevated but statistically insignificant results, among which the review published in 2017 included nine studies and 2,570,790 participants (2), the review in 2016 included ten studies and 4,660 breast cancer patients (16), the review in 2013 included 16 studies and 1,444,881 participants (17), and another review in 2013 included 15 studies and 1,422,189 participants (18).
In the last 2 years, nine original population-based studies reported that night-shift work was not associated with cancer development, which have not been included in previously published reviews (19)(20)(21)(22)(23)(24)(25)(26). We conducted this study to systematically summarize the evidence regarding the associations between night-shift work and cancer risks. We expect to facilitate recognition of the health-related problems among night-shift workers.

MATERIALS AND METHODS
This systematic review and meta-analysis were conducted in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) ( Table S1). The study protocol was registered in the online database of PROSPERO (CRD42019138215). This systematic review aimed to answer the medical question of the association between night-shift work and cancer risks by reference to PICOS: (1) study population were night-shift workers; (2) compared to population without nightshift work; (3) the exposure was defined as night-shift work; (4) the outcome of cancer risk was evaluated; (5) observational studies on this topic were included.

Search Strategy
We used Embase, PubMed, and Web of Science databases to systematically search English language publications issued up to May 31, 2019. The search terms were "carcinoma" or "tumor" or "cancer" or "neoplasm, " and "night-shift work" or "night work" or "shift work" or "work schedule tolerance" or "rotatingshift work." The detailed literature search strategy was shown in Supplementary Box 1. Two investigators independently searched and then screened the retrieved studies. In addition, we manually screened the reference lists of included studies to collect additional literature.

Inclusion and Exclusion Criteria
Literature was included based the following criteria: (1) nightshift work was reported. Night-shift work was defined by questionnaire interview or occupational history of those who have ever exposed to shift system (rotating or fixed, forward or backward rotation). The durations that participants have ever engaged in night-shift work were collected by retrospective investigation or follow-up interview. (2) Cancer risk was investigated. (3) Cohort studies, case-control studies, or nested case-control studies. (4) The risk was estimated by odds ratio (OR), risk ratio (RR), or hazard ratio (HR), with 95% confidence interval (CI). (5) For studies reporting overlapping data, the studies newly published or with a larger sample size were included. (6) Publications in English language. Exclusion criteria were (1) studies without sufficient data; (2) studies referring to recurrent cancer.

Quality Assessment
We assessed the bias risk as low, high, or unclear by verifying the checklist for measuring bias in risk factor studies to counter 10 important sources (domains) of bias (17,27). The following are domains of bias risk assessment: (1) exposure definition, (2) exposure assessments, (3) reliability of assessments, (4) analysis methods in research (research-specific bias), (5) confusion, (6) attrition, (7) blinding of assessors, (8) selective reporting, (9) funding, and (10) conflict of interest. We then rated the studylevel risk of bias as: low (low risk in all major domains and ≥2 of the minor domains), moderate (low risk of bias in ≥4 major and 2 minor domains), or high risk of bias (low risk of bias in <4 major domains). The detailed information is available in Supplementary Box 2.

Statistical Analysis
All statistical analyses were done with the Stata14.0 software (Stata Corp, College Station, TX, USA). We preferentially measured the association between night-shift work and all cancer risks via the pooled estimates (i.e., OR, RR, and HR) and 95% CI. The Q test along with I 2 statistic was used to identify whether heterogeneity was significant between eligible studies. When P < 0.10 or I 2 > 50% heterogeneity was considered significant, therefore the random-effect model (Der Simonian-Laird method) meta-analysis was applied, otherwise, the fixedeffect model was used. In addition, subgroup analyses were conducted to stratify the results on specific study design, occupation, night-shift work status, cancer type, and sex. A doseresponse meta-analysis was performed to evaluate the risk for cancers per year increase of night-shift work. We established the curve of dose-response relationship using a method proposed by Greenland and Longnecker (28). In this process, we combined the data from studies that reported the estimate effects of cancer risks with night-shift work at ≥3 quantitative categories. To assess the stability of the results, sensitivity analysis was conducted by sequential removal of each original study. Potential publication bias was assessed by the Begg's regression asymmetry test and funnel plot.

Literature Search and Study Selection
Details of the literature search and screening are shown in Figure 1. A total of 1660 publications were initially retrieved from Embase, PubMed and Web of Science. Among them, 538 duplicate publications were removed. After review of abstracts 1,037 studies were excluded for the following reasons: not human studies (n = 53), not studies on cancer and night-shift work (n = 713), reviews/editorials/letters (n = 271). By full-text review eight    studies with overlapping data and 20 studies on sleep patterns were removed. Altogether, this meta-analysis included 57 articles.

Quantitative Meta-Analyses
As shown in Table 2, the risks of breast cancer (pooled OR = 1.009, 95% CI = 0.984-1.033), prostate cancer (pooled OR = 1.027, 95% CI = 0.982-1.071), ovarian cancer (pooled OR = 1.027, 95% CI = 0.942-1.113), pancreatic cancer (pooled OR = 1.007, 95% CI = 0.910-1.104), colorectal cancer (pooled OR = 1.016, 95% CI = 0.964-1.068) were not significantly associated with night-shift work. Besides, the pooled results, which were from small number of original studies, showed that stomach cancer, esophageal cancer, leukemia, oral cancer, uterine cancer, laryngeal cancer, testicular cancer, and non-Hodgkin lymphoma were not associated with night-shift work. Combination of three original studies showed that night-shift work increased the risk of urinary cancer. However, decreased risks of lung cancer and skin cancer were observed on the basis of pooled results of six and four studies, respectively.

Subgroup Analysis
As subgroup analyses on the association between night-shift work and combined risk for cancers ( Table 3), no significant associations were observed in cohort studies (pooled OR = 0.996, 95% CI = 0.982-1.011) and nest case-control studies (pooled OR = 0.960, 95% CI = 0.887-1.032). However, the pooled result from case-control studies was statistically significant (pooled  For types of night-shift work, neither rotating nor fixed nightshift work was associated with an increased risk of cancer, with pooled ORs 0.982 (95% CI = 0.960-1.004), and 0.880 (95% CI = 0.746-1.014), respectively. The association between night-shift work and cancer risks were also analyzed among occupational groups. Night-shift work was associated with a decreased risk of cancers among textile workers, while no significant association was found for nurses.
We included 13 original studies from Asia, 26 from Europe, 16 from North America, and 2 from Oceania. Subgroup analyses indicated that night-shift work was associated with an increased risk for cancers in Europe, and a decreased risk in Asia, while no significant associations were observed in America or Oceania.

Dose Response Meta-Analysis
For studies that reported more than one category of durations of night-shift work, we assessed whether the risk of cancer increased in a dose-response manner per year night-shift work. As shown in Figure 3, a dose-response curve was established, in which the solid line is a curve model established by the dose response meta-analysis, while the dotted line represents a reference of linear model. The result indicated that, for every 1 year increase of night-shift work, there is no increased risk for cancer (χ 2 = 3.34, P = 0.067).
In addition, we validated this finding with comparisons of cancer risks among individuals with different classifications of night work duration (0-5, 6-10, 11-15, 16-20, 21-25, and ≥26 years). Taking all eligible studies together, night-shift work did not increase the risk of cancer in any group of night workers (Figure 4 and Table S3).

Heterogeneity Analysis
The Q-test and I 2 statistics were used to assess heterogeneity across included studies. No obvious heterogeneity was observed in the overall analyses, while heterogeneity was found in four subgroup analyses (i.e., men, North America area, prostate cancer, lung cancer, and colorectal cancer).

Sensitivity Analysis and Publication Bias
We performed the sensitivity analysis by omitting one study at a time, and recalculated the pooled OR of the remaining studies. As shown in Figure S1, no significant alteration was observed by

DISCUSSION
This updated systematic review included 57 publications with 8,477,849 participants. Our meta-analysis found an insignificant association between night-shift work and cancer risks. No increased risk for cancer was identified among female night-shift workers as well. Neither rotating night-shift workers nor fixed night-shift workers had an increased risk for cancer. However, analysis on geographical distribution showed an increased risk for cancer among night-shift workers in Europe.
As a common concern in case-control studies, recall bias might have been introduced into our study during the measurement of night work. This bias represents a major threat to the validity when the participants were investigated with self-reported questionnaires. In order to eliminate potential recall bias resulting from previous case-control studies on the association between night-shift work and cancer risks, we synthesized the data from cohort studies in which recall bias can be effectively controlled. Consequently, an insignificant association was noted again.
Researchers have proposed several underlying mechanism of cancer risks induced by night-shift work. Night-shift workers usually experience unnatural light at night, which reduces the release of melatonin (76). As a kind of methoxy indole compound secreted by the pineal gland, melatonin shows a variety of anti-tumor effects, such as anti-oxidant, anti-apoptosis, anti-angiogenesis, as well as modulation of hormones and immunity (77,78). It has been demonstrated that melatonin plays critical roles in breast, ovarian, endometrial, prostate, lung, and gastrointestinal cancers (79)(80)(81)(82)(83)(84)(85)(86). Decreased melatonin leads to an imbalance of inflammatory cytokine secretions, mutagenesis, and oxidative damage, which likely results in the progression of various cancers (87). Suppression of melatonin also induces the aberrant secretion of testosterone and estrogen which increases the risks of prostate, endometrial, ovarian, uterine, and breast cancers (88).
In addition, tumor suppression is a clock-controlled process. Night-shift workers are exposed to dysfunction of circadian genes that is understood to play a role in DNA repair and carcinogen metabolism (89)(90)(91). The disruption of the circadian time organization contributes to cancer development. The "clock" genes are known to be directly involved in the regulation of prostate tumorigenesis.
The intensity and duration, as well as the type of nightshift work may influence the effect on cancer risk. As other published systematic reviews, our study included all eligible studies on night-shift work (i.e., fixed and rotating night-shift work) in retrospective and prospective studies. Our subgroup analyses showed that neither fixed nor rotating night-shift work is associated with cancer risk. In addition, night-shift work has little association with cancer risk in spite of the variation of night work duration.
Surprisingly, our subgroup analysis demonstrated that nightshift work is associated with a reduction of cancer risk in Asians. It has been explained that Asian workers have different lifestyles and genotypes compared with Europeans and Americans (2). This finding, as well as the negative association between night-shift work and lung and skin cancers might result from publication bias or relatively small number of included studies.
The present study has more strengths than previous systematic reviews and meta-analyses on the same topic. As a newly released update, this study included many more eligible articles, among them nine studies were that were included in a meta-analysis for the first time. The larger populations enrolled in these studies could produce more accurate effect size at a higher statistical power. Furthermore, our study was conducted on the basis of strict inclusion and exclusion criteria. We rigorously included original data on night-shift work and excluded ineligible studies included in previous meta-analyses. These studies were conducted on work classifications, duration of sleep, sleep disturbance, and light at night. We checked all the database of original studies on night-shift work, and removed three studies on colorectal cancer (58,92,93), one study on lung cancer (94), one study on ovarian cancer (40), and five studies on breast cancer (3,58,(95)(96)(97) because these studies reported overlapping data from the Nurses' Health Study (NHS) and/or NHS2. As a preferred solution, the newly published studies were included in our meta-analysis (20,58).

Limitations
Our study has some limitations which might sometimes exist in common systematic reviews. First, even though we searched three most commonly used databases, there is potential studies that were missed, especially published in local languages. A slightly different search by a reviewer can lead to very different initial results, which should also cause some caution. Second, we observed moderate heterogeneity in the subgroup analyses of the cohort study group, case-control study group, women, European region, Asian region, breast cancer, lung cancer, and endocrine cancers. The data from case-control studies might be biased by different methods of night-shift work measurement. Third, due to the lack of information on occupations of participants and measurement of night-shift work, these variables were not taken into account in the adjustment model. Moreover, publication bias was statistically positive, which could hinder the quality of this study. We use a trim and fill approach, and no substantial differences were obtained. Our combination of the results on all type of cancers may lead to the neglect of cancerspecific differences. As is known, cancers with stronger hormone components appear to be substantially different from those with less hormone control. In addition, we reported the results for ever vs. never night-shift work. There are many other indicators in night-shift work studies that track exposure in a more variable way. It is possible that the use of more differentiated exposure metrics, such as frequency or intensity of night-shift work, might lead to other results.
In conclusion, this systematic review of 57 observational studies did not find an overall association between ever-exposure to night-shift work and the risk of breast, prostate ovarian, pancreatic, colorectal, non-Hodgkin's lymph, and stomach cancers. With regard to sex, night-shift work was only associated with increased risk of cancer among men.

AUTHOR CONTRIBUTIONS
HH and YW designed this study. AD, XZ, and XG contributed to literature search, review, and data extraction. XZ and TW conduced statistical analyses. XZ, XJ, and HH contributed to manuscript drafting. AD and YW contributed to manuscript revision. All authors have reviewed and approved the final version of this manuscript.