Night-shift Work and Risk of Prostate Cancer

Results From a Canadian Case-Control Study, the Prostate Cancer and Environment Study

Christine Barul; Hugues Richard; Marie-Elise Parent


Am J Epidemiol. 2019;188(10):1801-1811. 

In This Article


Study Design and Population

The present work was based on data from the Prostate Cancer and Environment Study (PROtEuS), a large population-based case-control study on prostate cancer conducted in Montreal, Quebec, Canada, in 2005–2012. PROtEuS was primarily conceived to study the role of occupational exposures in prostate cancer. The study design has been described previously.[9–11] In brief, eligible cases were patients aged ≤75 years who were diagnosed with a histologically confirmed primary tumor of the prostate in one of the 7 largest French-language hospitals (out of 9) in Montreal during 2005–2009. These patients represented more than 80% of all cases in the study base, according to the tumor registry. Concomitantly, population controls were randomly selected from the electoral list of French-speaking men residing in Montreal, which is continually updated. Eligible controls with a history of prostate cancer were excluded. Cases and controls were frequency-matched by age (±5 years). Among eligible subjects, 79% of cases (n = 1,937) and 56% of controls (n = 1,994) participated in the study. Refusal (86%) and untraceability (11%) were the main reasons for nonparticipation. Overall, 1,904 cases (1,472 low-grade prostate cancers and 432 high-grade prostate cancers) and 1,965 controls contributed to the analyses.

Data Collection

Subjects were interviewed face-to-face by trained interviewers. Data on sociodemographic characteristics, lifestyle habits, medical history, and anthropometric variables and a detailed occupational history were collected. For each job held for at least 2 years, information on work schedules, along with information on tasks, workplace characteristics, equipment used, and protective measures, was elicited. For complex occupations (industrial mechanics, firefighting, etc.), specialized questionnaires (n = 32) were also used. Occupations and industries were coded according to Canadian classifications.[12,13]

The PROtEuS protocol was approved by the ethics boards of all participating institutions. All participants provided written informed consent.

Assessment of Night-shift Work and Early-morning Shifts

Work schedules and schedule changes (hours, duration, etc.) within each of the 15,724 jobs held were recorded. On the basis of recommendations from the IARC Working Group,[7] we defined night-shift work as having ever worked for at least 3 hours between midnight and 5:00 AM. We then restricted our sample of night workers to men who had ever worked at night for at least 1 year with a minimum frequency of 3 nights per month, on average, over the course of their night-shift jobs. Subjects who had never worked at night constituted the reference category in all analyses in which night-shift work was considered.

We assessed exposure to night-shift work through several metrics: 1) ever engaging in night-shift work; 2) engaging in night-shift work with rotation, defined as having ever worked in night shifts involving a rotation with at least 1 other shift; 3) the number of night shifts worked in rotation, categorized as no night-shift work, no rotation, 2 rotations, or 3 rotations; 4) the direction of night-shift work rotation—that is, always forward, always backward, or both; 5) the rate of night-shift work rotation, based on the rate performed the longest over the course of the worker's lifetime—categorized as no night-shift work, daily or 2–4 days/week, weekly, or more than weekly; 6) the cumulative number of days of night-shift work, expressed as the sum of duration times intensity over the course of the worker's career; 7) the total duration of night-shift work, corresponding to the number of years of having worked at least 3 nights/month over the course of the worker's career; 8) the average intensity of night-shift work over the worker's career, expressed as the sum of the product of the number of days per year and the number of years of each job period in night-shift work, divided by the total number of years in night-shift work; and 9) work in night shifts only, without rotation (i.e., permanent night shifts).

Finally, we investigated the role of working in early-morning shifts—that is, starting work after 2:00 AM but before 6:00 AM—at least 3 times per month for at least 1 year. Consistently with the night-shift work metrics, we examined ever exposure, total duration, intensity, and cumulative exposure to early-morning shifts. For the latter analyses, subjects who had never worked in early-morning shifts and night shifts constituted the reference category.

In the main analyses, continuous variables were categorized according to approximate quartiles of the distributions among exposed controls.

Confounding Factors

We identified potential confounders using a directed acyclic graph (see Web Figure 1, available at based on the current knowledge and assumptions about the causal structure of the associations under investigation. Accordingly, our main models included covariates for age at diagnosis (cases) or interview (controls), expressed as <65 years or ≥65 years, ancestry (sub-Saharan African, Asian, French, other European, greater Middle Eastern, Latino, or other), and educational level (primary school or less, high school, college (2–3 years post–high school), university degree, or other). We had information on several lifestyle and occupational variables, but these were not retained for adjustment based on the directed acyclic graph.

Web Figure 1.

CausalDirected Acyclic Graph for the Association Between Nightshift Work and Prostate Cancer Riska, PROtEuS, Montreal, Canada, 2005–2012
Abbreviation: PROtEuS, Prostate Cancer and Environment Study
aRed nodes represent confounders and blue nodes represent other exposures.

Statistical Analysis

Multivariable unconditional logistic regression was used to estimate the association between the different schedule variables and the risk of prostate cancer and to calculate odds ratios and 95% confidence intervals. Assuming that missing data on night-shift work and early-morning shifts (approximately 8% of jobs) were missing at random, and including occupational codes as predictors, we performed multiple imputation by chained equations[14] using 15 data sets. Distributions in the latter were similar to those in observed data. Dose-response relationships were tested by modeling each category as a continuous variable. Polytomous logistic regression models were used to investigate associations with prostate cancer aggressiveness according to the Gleason score at diagnostic biopsy. Gleason scores of ≤6 or 7 (with 3 as the primary score and 4 as the secondary score) defined low-grade tumors (referred to as less aggressive cancers), while scores of ≥8 or 7 (with 4 as the primary score and 3 as the secondary score) defined high-grade (aggressive) cancers.[15] The Wald test was used to detect heterogeneity in odds ratios between the two groups.

Because prostate cancer is generally asymptomatic in its early stage, we conducted sensitivity analyses excluding controls who had not been screened in the 2-year period before the interview (n = 473), thereby reducing the likelihood of latent cancers in the control series. Moreover, because the cutoff of ≥3 nights/month, which was previously used in other studies of rotating night-shift work and cancer,[16,17] is arbitrary and represents a low frequency of night-shift work, we conducted sensitivity analyses with a higher cutoff (≥7 nights/month for at least 1 year) to capture night-shift workers with higher exposures.

We also explored the timing of night-shift work over the course of the participants' careers—that is, whether the last job entailing night-shift work had been held within the 20 years prior to the index date or further in the past.

All analyses were performed using SAS (version 9.4; SAS Institute, Inc., Cary, North Carolina). Statistical tests were 2-sided.