Shift work and cognitive impairment in later life – results of a cross-sectional pilot study testing the feasibility of a large-scale epidemiologic investigation

With a global prevalence of about 24 million cases, neurodegenerative disorders such as Alzheimer’s disease (AD) are a major burden of illness in the modern world [1]. They not only cause severe suffering among patients and their families but also generate substantial costs for medical services, health insurances, and society as a whole. For instance, in the US alone the annual costs for health care and long-term care of AD patients are estimated at 259 billion US Dollars [2]. Given the ageing of the population especially in Western societies, this burden will further increase in the future [3]. However, despite the importance of this type of diseases and the urgency for action, it is so far not well understood which genetic, environmental, and occupational factors contribute to their aetiology [1, 4].

Recent studies indicate that the risk of developing neurodegenerative diseases might be associated with sleep disturbance and shift work, which is usually defined as work outside a regular daytime work schedule [5‐9]. The proposed mechanism for this effect is circadian misalignment, which means that the circadian rhythm, i.e., the internal biological cycle that regulates sleep and wakefulness, is disturbed by an external stimulus such as shift work [10, 11]. Potential consequences include cancer, cardiovascular disease, diabetes, mental impairment, and neurologic as well as neurodegenerative disorders [7, 12‐15]. With respect to potential underlying biological mechanisms, neuroinflammation due to sleep disturbances [16, 17], changes in the neuronal plasticity of the hippocampus [18], or impairment of melatonin production due to shift work could play a role – the latter as melatonin appears to contribute to the protection from neurodegenerative disease [19‐21]. The plausibility of these mechanisms was shown in animal studies [22‐24]. However, it has not yet been sufficiently investigated whether circadian misalignment has a causal influence on the development of neurodegenerative diseases such as AD [7].

The manifestation of AD includes different components, with cognitive symptoms being among the most important ones [3]. Even before a manifest diagnosis of AD can be made, most patients show mild forms of cognitive impairment as a precursor [2]. Thus, epidemiological studies investigating AD usually include subjects who show early forms of cognitive impairment and who are thus likely to develop a clinical diagnosis of AD in the future [25, 26]. So far, few studies examined the long-term effects of shift work on chronic impairment of cognition in later life yielding inconsistent results [6, 13, 14, 27]. In addition, from a methodological point of view, it is also interesting that studies including shift-workers or elderly individuals from the general population reported different experiences with respect to invitee’s willingness to take part in such studies. While prospective studies, for example the Nurses Health Study [27] or the French VISAT study [6], reported excellent response rates, other approaches such as cross-sectional studies in Australia and Norway could enrol only about a third of the invited individuals [25, 28]. When planning a study that aims to assess cognitive effects in current and former shift workers, it thus seems to be wise to first evaluate the feasibility of the planned design by means of a pilot study.

A major limitation of previous studies is the lack of adequate consideration of study subjects’ chronotype, i.e., their internal biological clock. Taking into account chronotype and sleeping habits, however, is crucial to determine circadian misalignment and potential health effects due to shift work [29‐31]. For instance, results from studies on chronic disease such as diabetes or breast cancer point towards effect modification by chronotype [15, 32].

Therefore, we planned to design a large-scale epidemiologic study to investigate the putative association between shift work and cognitive impairment in later life with adequate consideration of individual sleeping habits and chronotype. Before implementing further steps, we evaluated the feasibility of the planned project by assessing study subjects’ willingness to participate, indication of selective participation, and the viability of the study instruments by means of the pilot study presented here.

Of the 500 randomly selected persons, 75 could not be contacted (70 had an invalid address, three persons had died, one person lived in a nursing home, another one in a hospice), resulting in a net sample of 425 potential participants. Of these, 75 subjects (18%) completed the questionnaire, of which 47 (11% of the total sample) participated in the cognitive testing. The remaining 28 persons either agreed to take part in the cognitive testing in the informed consent form, but could not be reached for scheduling, or did not appear for the test. Except for four variables (country of birth, time of falling asleep, minutes until falling asleep, time of waking up), the proportion of missing values for all variables in the questionnaire was less than 10 %.

Our study aimed at examining the feasibility of a large-scale epidemiologic study to investigate the relationship between exposure to shift work and cognitive impairment in older age. The feasibility was evaluated based on invited subjects’ willingness to participate in the pilot study (response rate), indication of selective participation, and completeness of the collected data. While the willingness to participate was very low, especially among non-shift workers, the completeness of the data indicates that the survey instruments are very well applicable. Significant cognitive impairment was observed in very few subjects, but the test results indicated that around one third of the participants had at least slight difficulties with at least one cognitive test.

With respect to the response rate, only a very small proportion of the invited persons (11%) agreed both to participate in the cognitive testing and to complete the questionnaire. In addition, some invitees completed the questionnaire without participating in the testing. This clearly contradicts the feasibility of the proposed approach, since results that are not prone to considerable levels of selection bias require substantially greater willingness to participate. The fact that 80% of those who participated in the pilot study reported that they had worked in a shift system before, indicated that exposed individuals were more probable to take part in the study than non-exposed invitees were. Hence, there is a high likelihood of selection bias in our study population.

An explanation for the low response could be that it was too much effort for the invited persons to come to the examination centre and undergo a test procedure lasting one hour, especially without any financial incentive. The incentive to participate in a free screening of their cognitive abilities apparently did not suffice to convince the majority of the invitees to participate. At the same time, the examination of their cognitive performance could have led to a feeling of fear or shame in the participants, especially among those who felt to suffer some form of cognitive decline and among those who were still working. Beyond that, individuals with subjective decline even may have been concerned that the cognitive testing could have revealed indication for more severe impairment such as an early diagnosis of dementia which is sometimes associated with fear of discrimination or stigmatisation [41, 42]. This effect may have especially occurred as the aim of investigating cognitive impairment was explicitly mentioned and not encapsulated in a more general description. In that regard, our study was similar to an Australian pilot study with a related research question, which also suffered from a low response rate – although it was still higher than in our study [25]. The largest difference between our and the Australian study, however, was that they recruited a population-based sample while we conducted our study in an occupational setting. The fact that this setting was the one in which also the researchers are employed may have negatively influenced the invitees’ willingness to take part. There is also a tendency towards selective participation by sex as almost 90% of all participants were female (among all invitees, about 75% were females). Moreover, in another study using a test battery similar to ours and inviting individuals aged 35–74 years, willingness to participate was lowest in the oldest age group [43]. This suggests that it is generally difficult to recruit elderly people for psychometric testing. Apart from the potentially deterrent effect of the cognitive test battery, also additional reasons for the low response can be debated. One of them may be study design since, as already indicated in the introduction, shift work studies achieving high response rates seem to be predominantly prospective approaches.

In contrast to the low willingness to participate, the quality of the collected data can be considered as high. All cognitive tests were completed and the proportion of missing information in the questionnaire was below 10 % for all but four variables. Thus, the procedure that participants first receive the questionnaire, fill it out at home and then bring it to the study centre where it is checked for missing or ambiguous information, can be regarded as highly feasible.

In terms of cognitive performance, in all three tests that were used to calculate a standardised performance score, at least 80% of all subjects tested had an average to above-average score. The Trail Making Test had the highest difficulty, with 17% of respondents performing substandard while the corresponding value for RBANS was 6 % and in the Vocabulary Test no participant had such a low test score. Accordingly, the clinical evaluation of the test results from all tests provided clear evidence of reduced cognitive performance in only two subjects. However, minor difficulties in at least one subscale were observed in about a third of the sample. This corresponds to the proportion of people with the syndrome CIND in the aforementioned Australian pilot study [25].

For the further planning of the project, selection and recruitment of the sample need to be improved considerably. First, it needs to be easier for subjects to participate in the cognitive testing. One way to do this within a retrospective cohort study could be to cooperate with the Occupational Health Service of a company that incorporates the testing into the regular occupational health check-ups. An alternative option could be a cross-sectional study recruiting participants via cooperating GPs that ask their patients to undergo the testing when they are in the GP’s surgery for medical examination. Another option could be a case-control study, which recruits cases with diagnosed mild cognitive impairment and a matched control group retrospectively assessing occupational history and chronotype of both groups. The largest potential limitation of a case-control design, however, is the retrospective exposure assessment, which in the case of our research question would be very prone to recall bias, especially among cases. Another means of possibly increasing the response rate might be to diminish the size of the cognitive test battery in order to reduce the expenditure of time for participation in the study. This would also reduce the time-wise and hence financial effort needed for field work on the part of the study team. The big drawback of shrinking the test battery, however, would be that less cognitive domains could be tested so that the assessment of participant’s cognition would be less comprehensive and thus potentially less reliable.

Regardless of the study design, subjects in the main study could be offered not only to receive their individual test results, but also other incentives that have been shown to be effective such as financial rewards or participation in a raffle [44, 45].

In summary, the present pilot study only partially supported the feasibility of the planned large-scale study. The further planning in particular needs to focus on improving the recruitment of subjects to avoid selection bias. The study questionnaire and the cognitive test battery have been proven to be suitable.