Shiftwork experience, age and cognitive performance

Abstract

Changes of alertness and cognitive efficiency has been suggested in people whose circadian rhythms are disrupted, e.g. night or shift-workers. Data from field and laboratory studies have demonstrated short-term cognitive disturbances related to circadian rhythm disruption. By contrast, little is known about the long-term consequences of chronic sleep deprivation, as can be observed with shift-work, on cognitive abilities. The present paper is aimed at evaluating, on a large cross-sectional sample of workers, the long-term influence of shift-work on verbal memory and speed performances. Participants were 3237 workers aged 32, 42, 52, and 62 years of various occupational statuses included in the VISAT (Aging, Health and Work) cohort. Data collected by questionnaires included items on working hours and shift-work and sleep disorders. Cognitive abilities were assessed using neuropsychological tests. Current male shift-workers had lower cognitive performance than never exposed workers. In the same population, memory performance tended to decrease with increasing shift-work duration. Among former shift-workers, the cognitive performance of the participant having stopped shiftwork more than 4 years ago seemed to be increased, suggesting a possible reversibility of effects. In conclusion, this study demonstrated that cognitive functioning tends to be impaired by a long-term exposure to SW. As found by other authors, neuropsychological performance tends to decrease with the increases in the duration of exposure to SW.

Keywords:

• Shiftwork
• Aging
• Neuropsychological tests
• Epidemiology
• Cohort

1. Introduction

Many studies have reported a variety of adverse biological, psychological and social effects of shiftwork and other atypical work schedules on the worker. For instance, effects on a wide set of health and well being components have been observed, including sleep, eating behaviour, gastrointestinal, neuro-psychic and cardiovascular functions, menstrual cycle, work accidents, absenteeism, family role (e.g., Costa 1996). Consequences for several aspects of performance have also been documented (e.g., Folkard 1996), which suggest changes in alertness and cognitive efficiency in people whose circadian rhythms are disrupted. The study of effects of night and shift work on alertness and cognitive performance, however, is complex. This is especially due to the fact that these performances vary both with circadian and homeostatic factors (e.g., see, Folkard & Akerstedt 1992), and can suffer from the effects of sleep deprivation and other consequences of desynchronization of the circadian system. There is some evidence, mainly from laboratory research, that sleep deprivation is associated with cognitive impairment. Adverse effects of sleep deprivation have been shown on executive functions (Harrison and Horne 1998), temporal memory (Harrison and Horne 2000), attentional processes (Kim et al. 2001), and working memory (Drummond et al. 2001). Data from field and laboratory studies also strongly suggest that disruption of circadian rhythm induces short-term effects on cognitive performance (e.g., see Folkard 1996; Folkard & Monk 1979; Rogers et al. 1989; Winget et al. 1984). Although much remains to be learnt concerning the mechanisms of the adverse effects on cognitive performance of working during abnormal work hours, there is little doubt about the existence of short-term harmful effects.

By contrast, very few studies have assessed long-term consequences of chronic sleep deprivation and repeated disruption of circadian rhythms on cognitive functioning, among workers having been exposed to day and/or night shift work over time, sometimes decades. One reason may be the great diversity of the working conditions involving abnormal working hours, but also the difficulty of distinguishing between the respective influences, in the long run, of work hours and of other time-related factors, such as ageing. There are good reasons to believe that such long-term effects may exist. One possible mechanism is the hypothesized chronic sleep disruption resulting from night and shift work exposure (e.g., Butat et al. 1993, Webb 1983) and its possible consequences on cognitive efficiency suggested by the above mentioned studies. Another mechanism may be the repeated stress induced by the desynchronization of the circadian system, as evidenced by the work of Cho et al. (2000) who studied the influence of chronic jet lag on cognitive efficiency. They compared two groups of 24-to-29 year-old women who were employees of international airline companies. One group had little or no circadian rhythm reset history (ground crew), while the other group consisted of airline flight attendants who had over 8 hours jet lag per week (cabin crew). The authors observed weaker cognitive performance for cabin crew with more than 3 years of service. At the same time, significantly higher salivary cortisol levels were found in cabin crew than in ground crew, and these higher levels correlated with lower cognitive performance. Cho (2001) also examined the relationship between cognitive performance, the cortisol level and the volume of the temporal lobe of two flight attendant groups with different jet lag recovery periods (n = 10 in each group). The short-recovery group (recovery interval: ⩽5 days) showed significantly smaller volume of the right temporal lobe than the long-recovery group (recovery interval >5 days). Moreover, a significant correlation was found between lower cognitive performance, higher salivary cortisol level and smaller volume of the right temporal lobe in the short-recovery crew. One important feature of these studies is that they suggest a cumulative effect of exposure to chronic disruption of circadian rhythms on cognitive functions and the underlying cerebral structures. Indeed, the cognitive deficit, as compared with ground crew, became significant only after four years of transmeridian flying experience.

Further research is needed on this topic, with larger samples, especially field research since this is the only way to get a sufficient amount of shiftwork experience to really examine long-term effects of any disruption of the circadian system. Moreover, age is a crucial variable that needs to be controlled in such studies, as ageing is known to be associated with changes in tolerance to shiftwork, greater occurrence of rhythm disturbance, increased sleep troubles (e.g., Härmä 1993, Härmä & Ilmarinen 1999) and cognitive decline (Craik and Salthouse 2000). Finally, knowing whether, if they exist, such long-term effects are permanent or reversible, is also an important issue. This requires the possibility of examining the effect of various time intervals since the former shiftworker has returned to normal work times.

The aim of the present study was to examine the relationship between shiftwork and cognitive efficiency, and to assess the hypothetical mediating role of sleep quality in this relationship. The study was based on data from the VISAT study (Aging, Health and Work), a large cohort stratified on age, of workers currently employed or retired (Marquie et al. 2002). Because they include a wide variety of information about present and past working conditions, an assessment of sleep quality and of some memory, speed and attention abilities, these data offer the opportunity to examine some aspects of the relationship between atypical work times, sleep, and cognitive efficiency, especially the effect of shiftwork duration and recency. The research questions were: (i) Do current shiftworkers show lower cognitive functioning as compared with workers who never worked on shift and former shiftworkers? (ii) Do the effects of shiftwork depend on exposure duration and worker's age? (iii) Are these hypothetical cognitive decrements reversible; in other words, is cognitive performance influenced by the length of the time since former shiftworkers have returned to a normal schedule?

2 Sample and methods

2.1 Participants

The data were taken from the cross-sectional phase of the VISAT study (for further details on the methodology and aims of this study more, see Marquié et al., 2002). The sample was composed of 3237 present and former salaried subjects born in 1964, 1954, 1944 or 1934. They were exactly 32, 42, 52 and 62 years old at the time of the data collection (1996). Only the group born in 1934 included retirees (83%). Participants were randomly drawn from the patient list of 94 occupational physicians in three southern regions of France, who volunteered for VISAT. The participation rate was 76%. Data were collected during the annual medical examination by an occupational physician to which any salaried worker is entitled in France. Retired workers, who were no longer followed-up by the occupational physicians, were especially invited for the purpose of the study.

All participant physicians took a specific training course. It was designed to achieve maximal standardization of data collection. This was especially important for the part of the protocol which was less familiar to the physicians, like cognitive tests. Two questionnaires concerning working conditions and non-occupational factors were self-administered but validated in a face to face interview with the physician. A health questionnaire was further filled in by the physician.

2.2 Shiftwork and other working hours

The work conditions questionnaire contained questions on whether the job (1) involved shiftwork with changing schedules, (2) often involved going to bed after midnight (more than 50 days per year), (3) often involved rising before 5 AM (more than 50 days per year), (4) and often involved not sleeping during the night (more than 50 days per year). Three answers were possible: ‘no never’, ‘yes now’, ‘not now, but yes in the past’. If the answer was ‘now’ or ‘in the past’, participants were asked for how many years they had been working on that schedule. For ‘in the past’ answers, participants were further asked how many years had elapsed since they had stopped working on that schedule (recency).

2.3 Sleep questionnaire

The participant had to rate on a 4-point scale (never, seldom, sometimes, often) the frequency in the last month of five sleep symptoms: (1) difficulty initiating sleep, (2) difficulty staying asleep, (3) difficulty getting back to sleep, (4) early morning awakening, (5) hypnotic medication use.

2.4 Cognitive tests

Three types of cognitive tests were administered to the participants: (1) a memory test adapted from the Rey Verbal Learning Test, followed by two memory retrieval tests. The participant was read a list of 16 words, and then immediately asked to recall the words. Three consecutive learning/recall trials were given. After a delay of 15 minutes, filled with other tests, the participant was submitted to a delayed free recall test, followed by a delayed recognition test. In the latter, the participant was required to locate the 16 previously learned words which were randomly mixed in with 32 new words. Five verbal memory measures were thus recorded: 3 immediate free recalls, 1 delayed free recall and 1 delayed recognition measure; (2) the Digit-Symbol Substitution subtest of the Wechsler Adult Intelligence Scale (Wechsler 1955), a test mainly reflecting processing speed (Salthouse 1992). The test consists of assigning the correct symbol to digits ranging from 1 to 9 according to a code table displaying pairs of digits and symbols. The participant had to copy as many symbols as possible in a time period of 90 s; (3) a selective attention test derived from the Sternberg test (Sternberg 1975) was composed of two subtests. The first was a task consisting of looking as quickly as possible through a line of 58 alphabetic characters (57 of which were distractors) to find a target letter shown in the margin, and then crossing it out. This task was repeated six times, on six lines with a different target each time. The location of the target letter among the distractors was random. The second subtest also had six lines of 58 alphabetic characters, but this time the memory load was greater because the target to locate was one of the four letters shown in the margin.

2.5 Statistical analyses

As sleep quality was assessed with five different items, Pearson's correlations between these items were computed, and then a synthetic variable was created. Likewise, as cognitive efficiency was assessed through eight memory and speed measures, a factor analysis was performed to define new synthetic variables. A Promax rotation was used (no assumption of independency between factors, unlike Varimax rotation).

Then, the relationships between shiftwork and sleep, between sleep and cognitive performance, and between the latter and shiftwork were analysed using multiple linear regressions. The influence of shiftwork duration and of shiftwork recency on cognitive scores were also examined with the same statistical method. As former results showed greater effects of shiftwork on sleep quality for women than for men in the long run (Marquié and Foret 1999), men and women were studied separately in the present study.

The analyses were performed using the Statistical Analysis System version 8 (SAS Institute 1999).

3 Results

3.1 Descriptive results

The socio-demographic characteristics of the participants are displayed in Table 1. The distribution of men and women according to age and educational level was relatively similar.

Table 1. Participants' characteristics.

	Men (n = 1660) N (%)	Women (n = 1577) N (%)
Age
32 years	431 (26.0)	465 (29.5)
42 years	488 (29.4)	484 (30.6)
52 years	461 (27.7)	438 (27.8)
62 years	280 (16.9)	190 (12.1)
Years of education
<8 years	379 (22.8)	264 (17.0)
8 to 9 years	425 (25.6)	380 (24.1)
10 to 11 years	222 (13.4)	241 (15.3)
12 to 13 years	225 (13.6)	280 (18.0)
> 13 years	408 (25.0)	409 (26.0)
Shiftwork
Current	265 (15.9)	321 (20.4)
Former	346 (20.8)	242 (15.3)
Never	1049 (63.2)	1014 (64.3)
Work implying to rise before 5 AM
Current	269 (16.2)	106 (6.7)
Former	342 (20.6)	161 (10.2)
Never	1049 (63.2)	1310 (83.1)
Work implying to go to bed after midnight
Current	181 (10.9)	74 (4.7)
Former	279 (16.8)	132 (8.4)
Never	1200 (72.3)	1371 (86.9)
Work implying to stay awake during all the night
Current	135 (8.1)	92 (5.8)
Former	251 (15.1)	125 (7.9)
Never	1274 (76.7)	1360 (86.2)

Table 1 also shows that the proportion of shiftworkers was higher in women. However, considering some particular characteristics of worktime helps us to get a more precise view of gender differences in this respect: work involved rising before 5 AM, going to bed after midnight, and staying awake in the night were more frequent for men than for women. This, along with further analyses based on the crossing of “current” answers for shiftwork and these other work time characteristics, showed that shiftwork corresponded to less atypical work times for women than for men. For instance, 64.2% vs. 22.4% of present shiftworkers, for men and women respectively, reported activity resulting in rising time before 5 hours; 44.5% vs. 14.9% reported not being able to go to bed before midnight; and 37.3% vs 19.9%, also reported missing night sleep. The analysis of occupations showed that 65% of female shiftworkers were employed in hospitals (by contrast, only 20% of the male shiftworkers worked in a hospital).

The proportion of current shift-workers decreased with age, but the relationship was weaker in women than in men. At the same time, the proportion of shiftworkers decreased with increasing education level for men only: for women it was not related to education. These gender differences also justify our decision to consider men and women separately in subequent analyses.

3.2 Sleep variables

The correlations between the different sleep items ranged from 0.11 to 0.44 and were all statistically significant at the 5% level. A global score of sleep disorders was created by summing the ratings of the 5 sleep items. As the ratings ranged from 1 to 4, the global score ranged from 4 to 20, where 20 represented the most disturbed sleep.

3.3 Cognitive variables

Cognitive performance was assessed through 8 variables: 3 immediate free recalls (IR1, IR2, IR3), 1 delayed free recall (DFR) and 1 recognition test (RECOG), 1 speed measure for the Digit Symbol Substitution Test (DSST) and 2 speed measures for selective attention tests (SA1, SA2). A logarithmic transformation was done for speed variables in order to improve distribution normality. The factor analysis made with cognitive tests resulted in a 3-factor solution, which was similar for men and women. The first factor, called Immediate Free Recall, included the first three immediate free recall tests, with the first two having high loadings on the Factor 1 only (IR1: 0.52, IR2: 0.93), and IR3 (0.45) also loading on Factor 3 (0.49). Factor 3, called Delayed Retrieval, had high loadings from DFR (0.72) and RECOG (0.59). Factor 2, called Speed, had high loadings from speed variables (DSST: 0.62, SA1: 0.63 and SA2: 0.67). Three cognitive scores were then created corresponding to each factor: (1) Immediate Free Recall = (IR1 + IR2 + IR3)/3; (2) Delayed Retrieval  = (DFR + RECOG)/2; (3) for speed variables, z scores were computed, resulting in Speed = [zscore(DSST) + zscore(SA1) + zscore(SA2)]/3. This latter variable was coded so that higher scores reflected higher speeds.

Summary statistics of these new synthetic cognitive variables according to age and educational level are given in Table 2. Not surprisingly, cognitive scores decreased with age and increased with educational level. It can be noted that women had in general better scores than men but that the difference between the extreme categories of age and educational level was similar among the two genders. Moreover contrary to the speed score, the 2 memory scores varied more with educational levels than with age.

Table 2. Cognitive scores for men and women according to age and educational level (1 : ≤ 7 years of education; 2 : 8 – 9 years; 3 : 10 – 11 years; 4 : 12 – 13 years; 5 : ≥ 14 years). P-values based on two-way analysis of variance including both age and educational level.

	Cognitive Scores
	Immediate Free Recall (SD)		Delayed Retrieval (SD)		Speed Z score (SD)
	Men	Women	Men	Women	Men	Women
Age
32 years	8.4 (2.1)	9.2 (2.0)	10.5 (2.4)	11.6 (2.2)	0.29 (0.72)	0.41 (0.70)
42 years	7.8 (1.8)	8.7 (2.0)	10.0 (2.3)	11.2 (2.2)	0.06 (0.71)	0.21 (0.66)
52 years	7.3 (2.0)	7.7 (1.9)	9.3 (2.4)	10.0 (2.2)	−0.24 (0.78)	−0.17 (0.73)
62 years	6.7 (1.8)	7.4 (1.8)	8.7 (2.4)	9.6 (2.3)	−0.57 (0.86)	−0.49 (0.74)
	p < 0.0001	p < 0.0001	p < 0.0001	p < 0001	p < 0001	p < 0.0001
Educational Level
1	6.3 (1.7)	6.9 (1.8)	8.2 (2.2)	9.2 (2.3)	−0.68 (0.81)	−0.60 (0.78)
2	7.2 (1.8)	7.7 (1.9)	9.2 (2.1)	10.0 (2.2)	−0.24 (0.69)	−0.12 (0.65)
3	7.9 (1.7)	8.6 (1.8)	9.8 (2.2)	10.8 (2.1)	0.10 (0.70)	0.16 (0.61)
4	8.1 (1.8)	8.9 (1.6)	10.3 (2.2)	11.6 (1.9)	0.18 (0.67)	0.39 (0.66)
5	9.0 (1.9)	9.6 (1.8)	11.2 (2.3)	12.1 (1.9)	0.44 (0.62)	0.43 (0.64)
	p < 0.0001	p < 0.0001	p < 0.0001	p < 0.0001	p < 0.0001	p < 0.0001

3.4 Shiftwork and sleep

The relationship between exposure to shiftwork and sleep was assessed using linear regressions adjusted on age. For current shiftworkers, the sleep disturbances score was raised both for the two gender groups for both current and former shiftworkers although this increase was significant only for women (p = 0.01 for current and p = 0.001 for former shiftworkers).

3.5 Sleep and cognitive efficiency

No significant association was observed between reported sleep disorders during the last month and either Immediate Free Recall, Delayed Retrieval or Speed performance.

3.6 Shiftwork experience and cognitive efficiency

As shown in Table 3, current male shiftworkers tended to show lower cognitive performance than never exposed workers. The relationship was significant for speed, and approached significance for immediate free recall. This contrasted with the higher cognitive performance among former shiftworkers. On the other hand, no clear relationship was seen for women. As poor performance might occur for current shift workers if testing were made immediately after night shifts, the sleep duration of the night preceding the cognitive tests was tested in the models as an independent variable (data not shown): cognitive performances were not influenced by this variable. No significant interaction was found between shiftwork and age in their relationship with cognitive efficiency.

Table 3. Differences in cognitive efficiency scores between shiftwork categories adjusted for age and education in multiple linear regression analyses for men and women.

	Men (n = 1645)			Women (n = 1555)
	β	SE	p	β	SE	p
Immediate Free Recall
Shiftwork
Never	–			0.00
Now	−0.23	0.13	0.07	−0.06	0.12	0.60
Past	0.23	0.11	0.04	−0.03	0.13	0.82
Delayed Retrieval
Shiftwork
Never	0.00			0.00
Now	−0.19	0.16	0.23	−0.16	0.13	0.23
Past	0.17	0.14	0.22	0.05	0.15	0.74
Speed
Shiftwork
Never	0.00			0.000
Now	−0.10	0.05	0.04	0.001	0.042	0.98
Past	0.07	0.04	0.11	−0.042	0.047	0.36

3.7 Shiftwork duration and cognitive efficiency

Table 4 shows the relationship between shiftwork duration and cognitive scores among current shiftworkers. For men, memory performance tended to decrease as shiftwork duration increased. The difference between the 10 – 20 years group as compared with the lowest exposure duration (1 – 4 years) was statistically significant for Immediate Free Recall. No relationship was observed for women. Further adjustment on the sleep duration of the night preceding the cognitive tests did not influence these results.

Table 4. Differences in cognitive efficiency scores between categories of shiftwork duration adjusted for age and education in multiple linear regression analyses among current shiftworkers for men and women.

Shiftwork duration	Men (n = 265)			Women (n = 321)
	β	SE	p	β	SE	p
Immediate Free
Recall
1 to 4 years	0.00			0.00
5 to 9 years	−0.39	0.46	0.16	−0.16	0.39	0.68
10 to 20 years	−0.80	0.38	0.04	−0.01	0.32	0.97
>20 years	−0.66	0.45	0.40	0.33	0.37	0.38
Delayed Retrieval
1 to 4 years	0.00			0.00
5 to 9 years	−0.35	0.57	0.55	−0.55	0.43	0.21
10 to 20 years	−0.78	0.48	0.11	−0.18	0.35	0.60
>20 years	−0.70	0.58	0.23	−0.32	0.41	0.44
Speed
1 to 4 years	0.00			0.00
5 to 9 years	−0.04	0.17	0.79	0.12	0.13	0.35
10 to 20 years	−0.06	0.14	0.66	0.19	0.11	0.07
>20 years	0.02	0.17	0.92	0.081	0.12	0.54

3.8 Shiftwork recency and cognitive efficiency

The influence of shiftwork recency was assessed for former shiftworkers ( Table 5). As can be seen, there was no clear relationship between shiftwork recency and cognitive efficiency. All memory scores were higher for men and women who had stopped working on shift more than 4 years before, as compared with those who had stopped shiftwork in the last 4 years. These differences were statistically significant only for women and Immediate Free Recall. The cut-off of 4 years is the one that showed the largest and most significant differences in cognitive performance between recent and more ancient former shiftworkers.

Table 5. Differences in cognitive efficiency scores between 2 categories of shiftwork recency adjusted for age and education in multiple linear regression analyses for men and women among former shiftworkers.

Shiftwork Recency	Men (n = 331)			Women (n = 221)
	β	SE	p	β	SE	p
Immediate Free Recall
1 to 4 years	0.00			0.00
>4 years	0.40	0.23	0.08	0.52	0.25	0.04
Delayed Retrieval
1 to 4 years	0.00			0.00
>4 years	0.50	0.28	0.08	0.52	0.31	0.10
Speed
1 to 4 years	0.00			0.00
>4 years	0.04	0.09	0.69	0.08	0.10	0.41

4 Discussion

Short-term effects of shiftwork on cognitive efficiency have been assessed, but very little is known, so far, about long-term effects. The present study aimed at evaluating, on a large population of workers, the influence of shiftwork on verbal memory and speed performance, taking into account the possible role of sleep in the relation. The impact of shiftwork duration and recency were also assessed.

Self reported sleep quality by itself was not associated with cognitive performances. Presumably short-term sleep deprivation is more likely to have effects on cognitive efficiency rather than sleep disorders over a fairly long period (one month), as was assessed in the present study. On the other hand, current male shift-workers showed lower cognitive scores than workers who had never worked on shift. This finding was significant for speed (processing speed and selective attention) and approached significance for memory. The results thus suggest that it is not sleep disorders, but rather other consequences of the desynchronization of the circadian rhythms associated with shiftwork that account for the lower cognitive functioning observed in men. The fact that current shiftwork experience was associated with lower cognitive scores only for men, may be explained by gender difference in actual work hours (shiftwork corresponded to less atypical work times for women than for men, as shown in Table 1), and by other gender-related job characteristics. As already mentioned, 65% of the women shiftworkers in the present sample were employed in hospitals, compared to only 20% of the men. As older workers have been sometimes reported to experience more difficulties than the young to cope with shiftwork (e.g., Quéinnec et al. 1998), it was reasonable to examine whether greater effects could be found with age. But the results did not show any interaction between age and shiftwork, thus indicating that shiftwork had the same effect on cognitive efficiency at every age.

If shiftwork had long-lasting effects on cognitive efficiency, even after return to normal work hours, as it has sometimes been hypothesized for sleep (Butat et al. 1993, Webb 1983), former shiftworkers should have been found to show significant lower cognitive scores than their “never” counterparts. Little evidence was obtained in favour of this hypothesis: only speed scores were lower in women who were former shiftworkers as compared with women who never experienced shif work; and the opposite trend was found in men for immediate free recall (p = .04), which suggests that men who were former shiftworkers tended to better recall the words previously learned, a result that cannot be explained with the available data.

An effect of shiftwork duration on cognitive efficiency was observed only in men and for memory performances, especially immediate recall. As compared with people who had less than 5 years of shiftwork exposure, immediate recall performance decreased as shiftwork experience increased, and significantly so between 10 and 20 years. Beyond 20 years, the difference was no longer significant. This may be due to a healthy worker effect. Indeed, it can be hypothesized that with increasing age, workers who do not succeed any longer in coping with shiftwork tend to return to normal work times, thus increasing the probability of finding on shift only workers who are best fitted to this workschedule (Quéinnec et al. 1998). This finding is in agreement with the trends already observed for other shiftwork-related health risks (e.g., Knutsson et al. 1986; Scott et al. 1997).

Finally, more precise information was found concerning possible long-lasting effects of shiftwork on cognitive efficiency in former shiftworkers, by analysing the effect of the time elapsed since shiftwork experience had stopped. A positive effect on cognitive efficiency of getting back to normal worktimes was found after a 4-year delay mainly for the memory scores, this effect reaching statistical significance for one cognitive score among women. This result suggests that cognitive deficits that may be associated, under some conditions (type of cognitive performance, gender, or gender-related variables), to shiftwork might be reversible. Only longitudinal analyses, however, could provide a sound test of this hypothesis.

To the authors' knowledge, no earlier epidemiological survey has ever evaluated the chronic consequences of shiftwork on cognitive functioning. Although the available data in the present study provided only limited information about the magnitude of the current shiftworker's desynchronization (no information on the work-rest schedule in the few last days), several features of the study (e.g., large population, data on sleep quality, on exposure duration, and on shiftwork recency for former shiftworkers) offered an interesting opportunity to study this issue. On the whole, this first work provides some evidence of adverse effects of shiftwork experience on cognitive functioning, especially for men who are currently on shift and have been practising shiftwork for 10 to 20 years. It also provides some evidence that the effect can be reversed after shiftwork exposure cessation. In spite of differences between the present study and the one by Cho et al. (2000) and Cho (2001) (e.g., method, age, gender and number of participants, type of cognitive tests), some form of desynchronization of circadian rhythm is common to both studies, as are some findings suggesting that chronic exposure to such conditions may have disrupting consequences on the cognitive functioning in the long term.

Although there are findings in the present study that support this idea, the evidence is partial, and some limitations of the study must indicate caution in drawing conclusions. Indeed, significant effects were mainly obtained for men, and these effects differed according to type of cognitive measure. Future work should be able to shed light on whether the gender differences observed in this study reflected basic differences or were due to job or demographic characteristics that were not controlled. Moreover, in men, for whom significant effects of shiftwork were observed, the various cognitive measures were not equally sensitive. As compared with the “never” participants, “current” shiftworkers showed a lower speed score, regardless of exposure duration. By contrast, the lower immediate memory score of current shiftworkers as compared to the “never” was only marginally significant when exposure duration was not taken into account; this is likely because the effect of shiftwork is only reliable between 10 to 20 years of exposure. This difference between speed and verbal memory performance suggests that speed processes are rapidly affected by shiftwork while verbal memory processes would be impaired later, after several years of exposure. Again, this is compatible with the Cho (2001) finding of memory deficits associated with temporal lobe atrophy and high levels of cortisol after several years of jet-lag exposure. This also emphasizes the potential benefits of diversifying the types of cognitive measures. Finally, it must be remembered that the analyses were based on cross-sectional data, and thus do not allow the inference of any causality from the observed relationships. Further analyses on longitudinal data will enable this issue to be better addressed.

Acknowledgements

We thank the occupational physicians and researchers of the VISAT group for their help in managing the research program. We are especially grateful to B. Baracat, D. Blaise, F. Blanc, H. Cadeac, F. Casaux, C. Dalm, C. Duolé, Y. Esquirol, H. Fonds, P. Jansou, N. Lizano, J. Mandrette, C. Martinaud, M. Niezborala, and J.B. Ruidavets. This research was supported by grants from the Centre National de la Recherche Scientifique (CNRS), Conseil Régional Midi-Pyrénées, Ministère de l'Enseignement Supérieur et de la Recherche, Ministère du Travail, and the Institut National de la Santé et de la Recherche Médicale (INSERM) to Jean-Claude Marquié.