Article Text

Download PDFPDF

Original article
Night shift work and incidence of diabetes in the Danish Nurse Cohort
  1. Anne B Hansen1,
  2. Leslie Stayner2,
  3. Johnni Hansen3,
  4. Zorana J Andersen1
  1. 1Department of Public Health, Center for Epidemiology and Screening, University of Copenhagen, Copenhagen, Denmark
  2. 2Division of Epidemiology & Biostatistics, University of Illinois at Chicago School of Public Health, Chicago, Illinois, USA
  3. 3Danish Cancer Research Centre, Danish Cancer Society, Copenhagen, Denmark.
  1. Correspondence to Dr Zorana J Andersen, Department of Public Health, Center for Epidemiology and Screening, University of Copenhagen, Øster Farimagsgade 5, Copenhagen 1014, Denmark; zorana.andersen{at}


Objectives Night shift work has been associated with poor sleep, weight gain, metabolic syndrome, which are recognised risk factor for diabetes. However, only a few studies have examined the effect of shift work on diabetes risk. Here, we study the association between shift work and incidence of diabetes in Danish nurses.

Methods We used the Danish Nurse Cohort with 28 731 participating female nurses recruited in 1993 (19 898) or 1999 (8833), when self-reported baseline information on diabetes prevalence, lifestyle and working time were collected, and followed them in the Danish Diabetes Register for incidence of diabetes until 2013. Nurses reported whether they worked night, evening, rotating or day shifts. We analysed the association between working time and diabetes incidence using a Cox proportional hazards model adjusted for diabetes risk factors, separately with and without adjustment for body mass index (BMI) which might be an intermediate variable.

Results Of 19 873 nurses who worked and were diabetes-free at recruitment, 837 (4.4%) developed diabetes during 15 years of follow-up. The majority of nurses (62.4%) worked day shifts, 21.8% rotating shift, 10.1% evening and 5.5% night shifts. Compared with nurses who worked day shifts, we found statistically significantly increased risk of diabetes in nurses who worked night (1.58; 1.25 to 1.99) or evening shifts (1.29; 1.04 to 1.59) in the fully adjusted models including BMI.

Conclusions Danish nurses working night and evening shifts have increased risk for diabetes, with the highest risk associated with current night shift work.

  • Diabetes
  • night shift
  • nurse cohort
  • type 2 diabetes

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

What this paper adds

  • Night shift work has been associated in prior studies with poor sleep, weight gain and metabolic syndrome.

  • A few epidemiological studies suggest that shift work may increase the risk of diabetes, but shift work is still not a confirmed risk factor.

  • Existing studies used different definitions of shift work, and most often dichotomised as shift versus no shift work.

  • In our study, we were able to separate the effects of night, evening, rotating and day shift work with respect to diabetes risk.

  • We found a significantly increased risk for diabetes in nurses working night shift or evening shifts as compared with those working day shifts. The highest risk was observed in night shift workers.


Increasing demand for 24 h services in the modern world has contributed to the growing number of companies requiring employees to work in shift work schedules, including rotating shifts, evening or night work. It is estimated that about 20–23% of the European working population is employed in shift or night work.1 Working during night time, including exposure to artificial light, can adversely affect the social life and well-being as it may disrupt circadian rhythms, compromise cognitive capacity, induce tiredness, irregular and poor sleep, and digestive problems, which may lead to weight gain, obesity, breast cancer, cardiovascular disease, metabolic disorders and possibly type 2 diabetes.2

The global diabetes epidemic is considered one of the major public health challenges and one of the greatest contributors to the global burden of the disease, with a predicted 65% increase in patients with diabetes by 2025, to 380 million.3 ,4

Although a few epidemiological studies suggest that night shift work may increase the risk of diabetes,5 ,6 shift work is still not a confirmed risk factor, and more data are needed.7 The epidemiological evidence on shift work and incidence of type 2 diabetes consists of six prospective cohort studies: two from the USA,8 ,9 three from Japan10–12 and a single study from Europe (Sweden);13 a cohort study on diabetes mortality,14 and several cross-sectional studies on diabetes prevalence.15–18 In the largest study to date, Pan et al8 found that rotating night shift work is associated with diabetes incidence in 177 184 American nurses from Nurses’ Health Study I and II. Similarly, Vimalananda et al9 has in the Black Women's Health Study with 28 041 participants detected association between night shift work and diabetes incidence, and like Pan et al,8 detected a dose–response relationship of increasing diabetes incidence with increasing numbers of years working night shift. Three Japanese studies, each in several thousand male factory workers, found increased risk of diabetes in men working rotating shift as compared with day work.10–12 A study in 5432 Swedish men and women found increased risk of diabetes related to shift work, but only in women.13

Four cross-sectional studies, in 1111 retired participants from the USA,15 in 26 463 retired workers from China,16 in 1601 Japanese workers,17 and in 14 427 men and women from ELSA-Brazil cohort,18 reported positive associations between shift work and diabetes prevalence. In a recent meta-analysis, Gan et al6 reported that diabetes risk was higher for those working rotating night shifts than those working night, mixed, evening or unspecific shifts, and called for more studies, especially on night shift work to address the limited power of existing studies, and the heterogeneity between the working time definitions used in the studies.6

In this prospective cohort study, we investigated the association between shift work and incidence of diabetes over 13 years among Danish female nurses who were members of the Danish Nurse Cohort. We present novel findings on the effects of different types of shift work on diabetes risk. Although the importance of shift work for diabetes risk has been suggested,6 this study will contribute with data on which shift workers (night, evening or rotating shift) are most susceptible with respect to diabetes risk, which may be useful in tailoring of prevention strategies and better understanding of mechanisms.


The Danish Nurse Cohort

The Danish Nurse Cohort was inspired by the US Nurse Health Study and was designed to investigate the health effects of hormone replacement therapy (HRT) in an European population.19 The cohort was initiated in 1993 by sending a questionnaire to 23 170 female Danish nurses (44 years and older), members of the Danish Nursing Organization, which includes 95% of all nurses in Denmark, both retired and working nurses. In total, 19 898 (86%) nurses replied, and follow-up of the cohort was updated in 1999 when an additional 10 534 nurses (who turned 44 in the meantime) were included, and in 2009, but without inclusion of new nurses. A self-administered questionnaire provided information on socioeconomic and working conditions, parents’ occupations, weight and height including weight at birth, lifestyle (diet, smoking, alcohol consumption and leisure time physical activity), self-reported health, family history of cardiovascular and cancer disease, parity, age at first birth, age of menarche and menopause, use of oral contraceptives and HRT, removal of uterus and ovaries. In this study, we utilised the earliest available information from the 1993 (19 898) or 1999 (8833) questionnaire for 28 731 female nurses. Nurses who were working at the time of recruitment were asked to answer following question about shift work: ‘Do you normally work in: (a) day, (b) evening, (c) night or (d) rotating shifts?’ The cohort was linked to the Danish Central Population Register (CPR)20 to obtain the vital status information as of 31 December 2012 (date of death, emigration or disappearance).

Danish National Diabetes Register

The Danish National Diabetes Register (NDR)21 was established to describe and monitor the prevalence and incidence of diabetes in Denmark since 1995. The NDR was made by linking existing Danish registries: the National Patient Register (NPR),22 containing hospital discharge diagnosis since 1977; the National Health Service Register (NHSR),23 with information on all services provided by general and specialist practitioners since 1990; and the Danish National Prescription Registry (DNPR), containing all prescriptions dispensed at Danish pharmacies since 1995.24 To be registered in the NDR, a person has to fulfil a minimum of one of the following criteria: (1) has been hospitalised with diabetes as discharge diagnosis (International Classification of Diseases (ICD)-10 code E10-14, DH36.0, DO24) in the NPR; (2) has used chiropody as a diabetes patient; (3) had five blood glucose measures within 1 year; or (4) two blood glucose measures per year in five consecutive years, all (2–4) registered in the NHSR; (5) made a second purchase of insulin or oral antidiabetic drugs within 6 months, registered in DNPR.19 Majority (50–60%) of people registered in NDR had multiple inclusion criteria, in which case the date of the first inclusion criteria is considered the incidence of diabetes. Since results of blood glucose measurements (criteria 3 or 4) are not available in the NHSR, people who had blood glucose measurements as the single inclusion criteria in the NDR are not considered diabetic in this study, as they cannot be confirmed as diabetic patients. All the participants are followed in the national Danish registries with 100% coverage until record of diabetes in Diabetes Register, or date of death or emigration in CPR register, and thus we had no loss to follow-up.

Statistical analysis

We used the Cox proportional hazards regression model to study the association between diabetes incidence and shift work (between day, evening, rotating and night) with age as the underlying time scale. Start of follow-up was the age at the date of response to questionnaire (1 April 1993 or 1 April 1999) or start of registration of diabetes (1 January 1995), whichever came latest, and the end of follow-up was age at the date of diabetes onset, date of death, emigration or 31 of December 2012, whichever came first. The effect of shift work was evaluated in several a priori defined adjustment steps: model (1) crude model, adjusted only for age; model (2) main, fully adjusted model, additionally adjusted for smoking (never, current, previously), smoking intensity (g/day), alcohol consumption (g/week), physical activity (low, medium, high), the consumption of fatty meat (yes, no), the consumption of fruit and vegetables regularly (yes/no), myocardial infarction (yes, no) and hypertension (yes, no), and marital status (married, separated, divorced, unmarried, widow); model (3) additionally adjusted for body mass index (BMI) (underweight, normal, overweight, obese), as BMI is also affected by working time and likely on one of the biological pathways from shift work to diabetes or in other words an intermediate variable. We examined the potential effect modification of the association between shift work and diabetes by BMI by testing the significance of the inclusion of an interaction term in the model using the Wald test. Finally, in a subset of the cohort (19 492) with available information, we investigated if adjusting for the self-reported intake of sleeping medicine at recruitment would influence the effect of the risk estimate for diabetes. We present results as HRs and 95% CIs. Analyses were performed using Stata V.13.1.


Of 28 731 nurses in the Danish Nurse Cohort, we excluded 192 who died between cohort entry (1 April 1993) and the start of NDR and start of follow-up in our study (1 January 1995), 588 who reported having diabetes at baseline at 1993 or 1999, and 31 who had diabetes record before start of follow-up in the NDR without self-reporting diabetes, 6493 who did not provide information about shift work (nurses not working at the recruitment), and 1554 who had missing information on one or more covariates. Of 19 873 women in the final analyses, 873 (4.4%) developed diabetes during the mean follow-up of 15.1 years, or 308 078 person-years, with diabetes incidence rate of 2.7 per 1000 person-years.

At recruitment, the majority of nurses (62.4%) worked day shifts, 21.8% worked rotating shift, 10.1% evening and 5.5% worked night shifts (table 1). Mean age at baseline was 51.4 years, and was higher (53.6 years) for nurses who later developed diabetes. Nurses who developed diabetes during the follow-up worked more often night shifts, had higher BMI, smoked more tobacco, were less physically active, consumed more fatty meat, and had more often hypertension than nurses without diabetes. Mean age for day shift workers at recruitment was 51.3 years, 53.0 years for evening shift workers, 54.3 years for night shift works and 50.2 years for rotating shift workers. Nurses working night shift were more overweight, smoked more tobacco, were more physically active, consumed more fat meat and were less likely to be married, compared with the other shifts type (table 2).

Table 1

Characteristics of 19 837 nurses at recruitment (1993/1999) by diabetes status at follow-up until 2013

Table 2

Characteristics of 19 837 nurses at recruitment (1993/1999) according to the different working shifts

We found statistically significantly increased risk for diabetes in nurses working night shift when compared with those working day shifts, both in age (HR=1.84; 95% CI 1.46 to 2.31) and fully adjusted models (1.73; 1.37 to 2.19; table 3). The association attenuated, but remained statistically significant after adjustment for BMI (1.58; 1.25 to 1.99). We also found statistically significantly increased risk of diabetes in nurses working evening shifts, in fully adjusted model without (1.21; 0.99 to 1.52) and with (1.29; 1.04 to 1.59) adjustment for BMI. Nurses working rotating shift did not have an increased risk of diabetes when compared with nurses working day shift, with HR of 1.06 (0.75 to 1.07) and 1.08 (0.91 to 1.28), respectively, in fully adjusted models without and with BMI (table 3).

Table 3

Association between shift work and diabetes among 19 873 Danish female nurses

We found no evidence of an interaction between shift work and obesity (BMI >30 kg/m2) or overweight (BMI >25 kg/m2; table 4), nor with BMI as continuous variable (results not shown). Results were also robust to adjustment for the intake of sleeping medicine (results not shown).

Table 4

Effect modification of an association* between shift work and diabetes incidence among 19 873 Danish female nurses by BMI


We found that Danish nurses working evening or night shifts had an increase in the risk of diabetes compared with nurse working day shifts. The risk was highest for night shift work, followed by evening work, while there was no significantly increased risk related to rotating shift work.

Our results generally agree with existing evidence from other cohort studies8–13 suggesting that night shift work increases risk of diabetes, although estimates are not directly comparable, due to different definitions of shift work. The shift work definitions are different in existing studies, as seen in studies of shift work and cancer.25 Several studies have simply dichotomised shift work.10 ,11 ,13 For example, Eriksson et al13 has dichotomised information on current shift work as any shift work (night, evening or rotating vs day work), assessed by question: ‘Do you do shift work?’13 while Kawakami et al10 and Suwazono et al11 both classified work in Japanese male factory workers as day only versus rotating shift. Thus, the effect of night shift work was not explicitly estimated in these studies.10 ,11 ,13 In the largest existing cohort study, US nurses were asked how many years in total they had worked rotating night shifts (defined as at least three nights/month in addition to having worked days or evenings in that month), with prespecified response categories: never, 1–2, 3–5, 6–9, 10–14, 15–19, >20 years.8 In Vimalananda et al,9 women were asked if they had ever worked a night shift (graveyard shift, from 00:00 to 8:00), and for how many years. Silva-Costa et al18 require a minimum of four night shift per month to be classified as a night shift worker. Finally, Morikawa et al12 compared day work only to two shift work (day and evening shifts) and three shift work (day, evening and night shift work). Most studies used primarily day time work as reference category,10 ,11 ,12 ,13 others have evening and day shift in the reference category,9 and in some cases, the reference group included evening and rotating shift workers, including night shift less than three nights/month.8 Thus, our study is the only which separates the effect of night, evening, rotating and day shift work, allowing, for the first time, the estimation of separate effects of each on diabetes incidence. A limitation is, however, our lack on information on duration and intensity, that is, the frequency during a month of such non-day time work. Furthermore, the vast majority of Danish nurses have had night work, especially during young ages.26 If night work in young age is associated with diabetes later in life, this misclassification of exposure to evening and night shift work may have attenuated our relative risk estimates towards the null.

Comparison of our findings with other studies is also hampered by differences in the characteristics of the study populations. For example, three Japanese studies include only male workers.10–12 Our study in Danish female nurses and night shift work is most comparable to the American Nurse Health Study in American female nurses by Pan et al8 who, in contrast to ours, benefit from having data on number of years working rotating night shift. We have, however, detected considerably stronger associations with 50% (1.50; 1.20 to 1.89) increase in diabetes risk (adjusted for BMI) in women working night shifts compared with those not working night shifts, in comparison to Pan et al8 who, in BMI-adjusted model, found a 24% (1.24; 1.13 to 1.37) increase in diabetes incidence in nurses working rotating night shifts more than 20 years, and 10%, 6% and 3% in nurses working rotating night shifts 10–19, 3–9, and 1–2 years, respectively, compared with nurses not working night shifts.8 Vimalananda et al9 report similar estimate to Pan et al,8 with BMI-adjusted estimate of 23% (1.23; 1.03 to 1.47) increase in diabetes incidence in African-American women working more than 10 years night shift (11% and 9%, for those working 3–9 and 1–2 years, respectively) compared with women never working night shift.9 A stronger association was detected among female civil servants in Brazil who had a 42% (1.42; 1.39 to 1.45) higher risk of diabetes after 20 years of night shift work, compared with those who had never worked night shift work.18 Our findings are consistent with the meta-analyses by Gan et al6 who found that associations between shift work and diabetes are stronger in Europe (1.36; 1.05 to 1.73), than the USA (1.09; 1.03 to 1.14) and in Asia (1.07; 1.03 to 1.11), although this may be due to differences in study design as discussed above.8 Our estimate is comparable to a smaller Swedish cohort study by Eriksson et al13 who detected a very strong association between shift work and diabetes incidence in women (OR=1.9; 95% CI 0.8 to 4.4 in fully adjusted model and BMI-adjusted model), and none in men, although shift was defined as any shift work, not specifically as night shift.13 Finally, Gan et al6 in his meta-analyses reported an OR of 1.09 (1.04 to 1.14) for shift work and diabetes for women, a much lower risk estimate than ours, but this was based on a combination of cohort and cross-sectional studies. For men, Gan et al6 found OR of 1.37 (1.20 to 1.56), which is still lower than any of our adjusted results for night shift work.

Our findings that night shift workers have the highest risk of diabetes (1.58; 1.25 to 1.99) and that evening shift work (1.29; 1.04 to 1.59) increases diabetes risk as compared with women working day shift are novel. We found no increase in diabetes risk in women working rotating shifts (1.08; 0.91 to 1.28) when compared with women working day shifts. This may be explained by the fact that Danish nurses in rotating shifts both have day, evening and night shifts or only day and evening shifts, and generally have very few night shifts in a sequence, which implies none or only minor circadian disturbance. Furthermore, rotating shift workers possibly have more time to recover from night shifts, which may reduce the adverse effects related to night shift work. The most dominant schedules for Danish nurses who work day shifts are from 7:00 to 15:00, evening shift from 15:00 to 23:00, and night shifts from 23:00 to 7:00.26 This result is in contrast to Gan et al,6 who in meta-analyses reported the highest risk of diabetes for rotating shift work (1.42; 1.19 to 1.69) and notably weaker findings with night shift work schedules (1.09; 1.04 to 2.33), but again this analysis mixed results from both cohort and cross-sectional studies as well as results based on different definitions of rotating shifts.6 More studies on different shift work schedules are needed to elucidate which shift work regimen is most and less harmful.

Observed findings on increased risk of diabetes in nurses working night or evening shift are likely explained by several mechanisms. First, exposure to light at night leads to a decrease in pineal release of melatonin which is a strong antioxidant, and circadian rhythm disruption, but normally with at least several night shifts in a sequence.27–31 Persistent circadian stress, common in night shift workers, may cause excessive secretion of cortisol and interleukins, which together with increased insulin concentrations can lead to abdominal fat build up, lipid disorders and insulin resistance.27–31 Night and shift work are often accompanied by changes in lifestyle, such as changing mealtimes and thereby changes in timing of insulin response.29 ,32 Weight gain was more common among night than day shift Brazilian working nurses, as well in Japanese male night shift workers,33 ,34 although a review concluded that evidence between shift work and weight change is inconsistent.35 An Italian study has documented significant changes in BMI, smoking and alcohol intake in male night workers.36 Furthermore, stress related to shift work may lead to increased appetite, weight gain and glucose intolerance, may contribute to the development of diabetes in shift workers.2 ,4 Finally, short sleep duration and poor sleep quality, often found in shift workers,37 have been associated with larger psychological distress, higher levels of fasting insulin and inflammatory biomarkers, and decrease in insulin sensitivity, all contributing to increased diabetes risk.38–40

We found no evidence of an interaction between night shift work and BMI, suggesting that risk of diabetes from shift work is pertinent to all women. This is in agreement to Pan et al8 who found no interaction between rotating night shift and BMI at baseline in American nurses. Furthermore, our estimates of risk were only slightly changed when BMI was controlled for, suggesting that BMI is neither a strong confounder nor mediator in our study.

The strengths of this study include the large prospective cohort of Danish nurses, with well-defined information on diabetes risk factors and objective, prospective assessment of diabetes incidence from a nationwide registry, with minimal possibility of recall or information bias. Furthermore, we are the only cohort study to date that has information on different types of shift schedules, and could report estimates separately for night, evening and rotating shift work. The main weakness of this study is the lack of information on duration and intensity of working different shifts, precluding analyses of dose–response association with diabetes risk. Another weakness is that we lacked data on sleep patterns, including length and quality, which was also lacking in prior studies.8–14 As the association between night work and diabetes may be mediated by short sleep length, the observed associations in current and other studies may attenuate after adjustment for sleep length. We did not have detailed information on socioeconomic status (SES), such as household income or length of education, which may be a confounder, as nurse with lower SES may be more likely to work night shifts. However, we have a cohort of female nurses with generally similar education and uniform SES, and have furthermore adjusted for lifestyle an indirect proxy for SES, and marital status, as married nurses may have higher household income than unmarried. Furthermore, we were not able to distinguish between type 1 and type 2 diabetes, but we can assume with the age of the nurses above 44 years at baseline, that the vast majority of new diabetes cases are type 2 diabetes, because the other type of (juvenile) diabetes is normally diagnosed at young ages. The cohort has been compared with the general population of Danish women, and similar health status has been found, suggesting that external validity of the cohort is high.19

In conclusion, in a large Danish cohort of female nurses, we found that evening and night shift work is associated with an increased risk of diabetes. These findings are consistent with previous studies, and provide support for these associations. Further research is needed to improve our understanding of the mechanism by which these effects occur, so that effective prevention strategies may be developed.


View Abstract


  • Contributors ABH contributed with the idea for the study, performed statistical analyses and drafted the manuscript. ZJA contributed to the idea for the study, prepared data for analyses, and supervised ABH in performing statistical analyses and manuscript preparation. LS and JH contributed to the manuscript preparation. ABH and ZJA take full responsibility of the contents of this article.

  • Competing interests None declared.

  • Ethics approval Danish Data Protection Agency.

  • Provenance and peer review Not commissioned; externally peer reviewed.