Effect of accumulated night work during the working lifetime, on subjective health and sleep in monozygotic twins


Michael Ingre, National Institute for Psychosocial Medicine (IPM), PO Box 230, 171 77 Stockholm, Sweden. Tel.: +46 (0) 8 524 820 48; fax: +46 (0) 8 32 05 21; e-mail: michael.ingre@ipm.ki.se


The present study sought to investigate the effects of lifetime accumulated night work on subjective health and sleep in monozygotic twins discordant on the exposure to night work. A total of 169 twin pairs (83 males, 86 females) over 65 years of age were analyzed on variables that describe the presence or absence of complaints regarding subjective health, disturbed sleep, repeated awakenings and not being well rested. Significant (P < 0.05) risk ratios (RR) were found for subjective health (RR = 1.67) and disturbed sleep (RR = 2.83) indicating a higher risk of developing complaints to former night workers. There was neither a dose-dependent effect of night work nor an interaction effect with gender observed.


One of the major health problems in shift work is disturbed sleep (Åkerstedt 2003). A number of studies have shown that particularly the night shift is associated with reports of sleep problems (Aanonsen 1964; Graf et al. 1958; Pilcher and Coplen 2000; Ribet and Derriennic 1999; Thiis-Evensen 1958). Polysomnographical studies have shown that sleep is reduced by about 2 h after a night shift or before a morning shift but is of normal length after an evening shift (Åkerstedt et al. 1991; Foret and Benoit 1974; Tilley et al. 1982). Most of the loss seems to affect stage 2 and REM sleep. The problems of sleep (and wakefulness) in shift work have led authors to suggest the existence of a shift work maladaptation syndrome or a shift work sleep disorder (Guilleminault et al. 1982; Lavie et al. 1989).

It also seems that sleep problems may increase with accumulated exposure to shift work, as evidenced in the only study of long-term follow-up of sleep in shift workers (Verhaegen et al. 1986a). On the other hand, very little information is available on persistent/chronic effects of shift work on sleep. One such study of 15 former and 10 present shift workers found more sleep interruptions in the former group (Webb 1983). However, Cervinka et al. (1986) failed to find any polysomnographical differences between 10 night nurses and 40 university staff. Dumont et al. (1997) found that an insomnia index had worse values in those who had worked during the night for more than 5 months per year for 4–10 years, but nurses with more than 10 years of night work had a low insomnia index. A polysomnographical recording showed less slow wave sleep (stages 3 + 4) in former night workers. It was suggested that the lower insomnia in experienced shift workers might have been a selection effect.

One has to bear in mind that interpretation of studies of former shift workers who have transformed to day work are complicated because selection into (Knutsson and Åkerstedt 1992) and out of (Aanonsen 1964; Marquiéet al. 1999) shift work, may bias the results. Individuals with a poor tolerance for shift work are likely to avoid such hours and seek transfer to day work if recruited for shift work. Also, long-term studies almost invariably loose subjects over time, which often makes the comparisons between day and shift workers suffer from a healthy worker bias.

There is no study regarding the effect of lifetime exposure to night work on subjective health or any other health parameters not related to sleep, except mortality. The latter was studied by Taylor and Pocock (1972) but no effects of shift work were observed.

One approach to avoiding selection effects is to investigate a well-defined group after retirement. This was carried out by Michel-Briand et al. (1981) who did not find any effect on sleep, but the results were not complete. Taking this approach one step further, one might investigate retired monozygotic (MZ) twins discordant on shift work. Genetic and early life factors would then be highly controlled, thus minimizing selection effects. This approach was used in the present study of the Swedish Twin Registry (Lichtenstein et al. 2002).

In a recent updating effort of the Swedish Twin Registry we were able to include questions of accumulated time with night work, disturbed sleep, sleep duration, sleep behavior and subjective health. The present study sought to investigate the relation between lifetime accumulated night work and subjective health and sleep in MZ twins discordant on night work.


The overall design of the present study was that of a retrospective matched-pair cohort study in which MZ twins exposed to night work (at least one year) was matched with his/her co-twin (with no exposure to night work). Thus, the cohort consisted of paired MZ twins discordant on the exposure to night work. To study the effect across the working lifespan, the target population was MZ twins over 65 years of age (the formal retirement age in Sweden).

The present study was part of a larger effort of the Swedish Twin Registry to collect information on health and behavior in all twins in Sweden born in the year 1958 or earlier. Data were collected between March 1998 and March 2002 by means of a computer-assisted telephone interview. Efforts were made to interview both twins of a pair within a month of each other to avoid possible effects of aging. A more detailed presentation of the methods used has been presented by Lichtenstein et al. (2002).

A total of 48 216 twins responded to the questionnaire (74% response rate) of which 9168 were MZ and 1814 were MZ over 65 years of age. Of the latter group, 94 twins were excluded because their twin-mate was missing, which resulted in a total of 1720 twins (860 pairs). Finally, 169 pairs (83 males and 86 females, mean age 69.7 years, SD = 2.9 years) were found to be discordant on the exposure to night work and selected for further analyses.

It was not feasible to include detailed questions on cardiovascular, gastrointestinal and other diseases that are normally related to shift work because of insufficient space (Knutsson 2003). Subjective health was chosen as an integrative measure of health effects. It has been estimated in many of the health-oriented studies of shift work, but perhaps most relevant are the approaches related to duration of exposure, showing negative effects (Meers et al. 1978; Verhaegen et al. 1986b).

The variables used in the present study describe years of exposure to night work, years of formal education, weight, body mass index, diurnal type (1 = extreme morningness, 4 = extreme eveningness), habitual rise time, habitual bed time, time in bed, self-rated health (1 = very poor, 5 = excellent) and complaints (1 = always, 5 = never) regarding disturbed sleep, early awakenings, and being well rested (Åkerstedt et al. 2002).

Subjective health and complaint variables were recoded into binary form to reflect the presence or absence of complaints (categories 1 and 2 = poor/very poor subjective health or complaints mostly/always).

Continuous data were analyzed with t-tests for dependent measures between exposed and unexposed twins. McNemar's chi-squared test for dependent measures was used to test for differences between exposed and unexposed twins on binary data (complaints). In addition, risk ratios (RR) and 95% confidence intervals (CI) were calculated. An alpha level of 0.05 was used to test for significance. All statistical analyses were performed with Stata 8.1 for the Macintosh.


The mean exposure to night work during the lifespan was 12.4 years (range 1–51) in the exposed ‘night’ twins and no exposure was present in the unexposed ‘day’ twins. There were no differences between night twins and day twins with respect to formal education, weight, body mass index, diurnal type, habitual rise time, habitual bed time or time in bed (Table 1).

Table 1.  Descriptive data and t-test for exposed (night) and unexposed (day) twins
VariableNumberMean ± SEt-test
  1. Number, number of pairs; P-value, level of significance.

Exposure to night work (years)16912.37 ± 0.960.00 ± 0.0012.90.000
Formal education (years)1578.80 ± 0.258.69 ± 0.230.490.622
Weight (kg)16773.81 ± 1.0473.57 ± 0.960.360.717
Body mass index16425.76 ± 0.3425.58 ± 0.300.830.405
Diurnal type (1–4 evening)1512.08 ± 0.102.12 ± 0.100.380.702
Habitual rise time1627.03 ± 0.087.07 ± 0.080.540.593
Habitual bed time16222.51 ± 0.0622.61 ± 0.061.530.128
Time in bed (h)1628.52 ± 0.098.46 ± 0.080.700.487

There was a difference between the night and day twins on self-rated health and disturbed sleep as indicated by a significant chi-square (Table 2). Regression analyses showed that twins exposed to night work had a significantly higher risk than day twins of reporting poor subjective health (17.8% versus 10.7%) and disturbed sleep (10.2% versus 3.6%).

Table 2.  Prevalence of complaints and estimation of risk ratios for exposed (night) twins
ComplaintsNo. of twin pairsMcNemarRegression
DNdNDndnχ2P-valueRR95% CI
  1. Upper case letters indicate the presence of a complaint in the night (N) or day (D) twin, lower case letters indicate the absence of complaints; P, level of significance for McNemar's χ2; RR, risk ratio; CI, confidence interval.

Self-rated health624121274.000.0461.6671.005–2.765
Disturbed sleep11651445.760.0162.8331.164–6.895
Early awakening521171240.420.5161.1820.713–1.959
Well rested512161330.570.4500.8100.468–1.402

To test if there was a dose-dependent response on poor subjective health and disturbed sleep, exposure to night work was introduced as a continuous predictor in two separate regression analyses for the exposed twins. The risk ratios for (every year of) exposure to night work was not significant neither for poor subjective health (RR = 1.008; 0.984–1.031; P = 0.532) nor for disturbed sleep (RR = 0.990; 0.952–1.030; P = 0.623).

The effect of gender was investigated by introducing gender as a between groups factor in the regression analyses. No significant interaction was found between gender and exposure to night work on any of the studied variables.


Exposure to night work was associated with 67% increased risk of reporting poor subjective health and 183% increased risk in reporting disturbed sleep. Other indicators of sleep quality, sleep length or timing of sleep did not show any differences.

The results are in agreement with a number of cross-sectional or short-term studies showing disturbed sleep in shift work, but represent the first study of accumulated affects with a reasonable control of dropouts and background factors.

Apparently, the reduction of sleep quality was not reflected in a reduced duration of sleep, nor in bedtimes or times of rising. Thus, much of the essential sleep pattern was left undisturbed. This is not in contradiction with the finding of reduced sleep quality – it is quite possible to have a long but disturbed sleep, even if the two are often correlated.

As with sleep, the results on subjective health in relation to shift work confirm previous observations (Meers et al. 1978; Verhaegen et al. 1986b), but adds to this the exposure to night work occurring for at least 1 year during an individual's working life. Apparently, the retired shift working twin more often perceives him/herself to have a poor health than his day working sibling.

An important question is whether the effects are large enough to be of concern. The present data do not, however, lend themselves to conclusions in this respect. One would need more data on health care consumption and perhaps mortality to draw such conclusions, and a longitudinal design would probably be necessary. Thus, the present results are merely suggestive of an effect of accumulated exposure to night work.

There were no dose-dependent effects of exposure. The result is in accordance with a selection effect out of night work (Aanonsen 1964; Marquiéet al. 1999). Individuals who develop sleep problems transfer to day work thus, counteracting an accumulation of problems with increasing duration of exposure – a healthy worker effect (Knutsson and Åkerstedt 1992). However, the results indicate that the exposure obtained was sufficient to establish an increased risk of developing chronic sleep and health problems.

The use of discordant MZ twins means that any genetic (or shared environment) influence on the outcome has been eliminated. It should also have eliminated much of the effects of the selection mechanism into shift work (Knutsson and Åkerstedt 1992), which otherwise may have made it more likely for individuals with a certain psychological and/or physiological make up to enter shift work. What were not controlled for in this study, however, were other aspects of the work situation that could have had an effect on sleep and health. This was unfortunately not possible in the present study although it would clearly have been an improvement to the design. However, one must also recognize the magnitude of problems if one would try to control for all relevant aspects in the environment across a whole lifespan. Even if the information were available, it would require numerous covariates in the analyses and a much larger sample size. The analyses would still suffer from potential confounders of heredity if not performed in discordant MZ twins. In the present study at least, the influence of heredity on sleep, diurnal type and health has been controlled.

Finally, it is important to bear in mind that in this study it was only possible to analyze twins still alive, and healthy enough to respond to the questionnaire. If exposure to night work had a negative effect on health (as the present study implies), then there could be a bias in the dropout pattern that would have strengthened the effects reported in this study (i.e. the shiftworking twin had died). Also, response recall bias may have affected the reported exposure to night work. It is possible that some of the ‘day’ twins in this study have forgotten that they have worked night thus, weakening the reported effect of exposure to night work.

In summary, the present study has demonstrated in discordant twins that night shift work is related to a higher risk of reporting poor subjective health and poor sleep quality after retirement. It should be emphasized that the term ‘night work’ in this study includes permanent night work, and traditional three-shift work as well as roster systems with more irregular forms of night work.


The authors are very much obliged to the Swedish Research Council for Working Life and Society (FAS) for their help in conducting the study. The Swedish Twin Registry is supported by grants from the Swedish Research Council and the National Institute of Health, USA, grant AG 08724.