This cross-sectional study included 55,727 evacuees, aged 40-89 years (mean age, 61 years), who experienced the Great East Japan Earthquake, and who had taken the Mental Health and Lifestyle Survey as a part of the Fukushima Health Management Survey. Detailed information about the survey is described elsewhere [17, 18]. Briefly, the Mental health and lifestyle Survey is one part of the Fukushima Health Management Survey. The target population were all the residents who lived in the evacuation zones, which is a government-designated area (20 km radius) around the nuclear power plant. The earthquake occurred on 11 March 2011, and the study period was from 18 January 2012 to 31 October 2012. We sent a self-administered questionnaire on various lifestyle and mental health factors to all the residents. The overall response rate of the Mental health and lifestyle Survey was 40.7% .
We defined an evacuee as a resident in evacuation areas who reported to have experienced the earthquake, including those who had experienced the tsunami and/or the indirect effects of the nuclear power plant accident. The exclusion criteria were as follows: not having experienced the earthquake (n=2,026); having any major disabilities in activities of daily living such as eating, dressing, going to the toilet, or shopping (n=3,351). We imposed this latter restriction because individuals with disabilities may represent a distinct lifestyle compared to individuals without disabilities, which may lead to misclassification due to their avoiding activity-related pain. We also excluded missing information or abnormal values for basic characteristics and medical history (n=6,256), musculoskeletal pain and lifestyle-related information (n=7,052), and psychological factors (n=2,113). Finally, we used 34,919 people (female, n=18,156; male, n=16,763) without major disabilities at about 1.0 ± 0.1 years (average ± standard deviation) after the disaster. The Ethics Committee of Fukushima Medical University approved this survey protocol. The participants were determined to have provided their consent to participate by returning the survey questionnaires.
We asked all participants to answer the multiple-choice question “In the past few days, have you experienced uncomfortable symptoms due to any diseases or injuries?”. If the participant responded “Yes” to low back pain or limb pain, we regarded that the participant had major musculoskeletal symptoms. The most common areas of musculoskeletal pains are the low back, knee, hip, ankle, foot, shoulder, elbow, wrist, and fingers (the latter eight are in the category of limb pain) . We further asked whether the pain was exacerbated after the disaster. The subjects were categorised into 3 groups: absent musculoskeletal pain (absent), prevalent plus exacerbated musculoskeletal pain (exacerbated), and prevalent musculoskeletal pain (prevalent).
We asked participants about lifestyle factors including the use of evacuation shelters or temporary housing at any point in time (shelter use: no/yes), having lost their job after the disaster (job loss: no/yes), decrease in income after the disaster (decreased income: no/yes), current smoking status (smoking status: never/past/current light smoking/current heavy smoking), current drinking status (drinking status: never/past/current light drinking/current heavy drinking), subjective lack of sleep (insomnia: no/yes), current regular exercise (regular exercise: almost none/1 time per week/2-4 times per week/almost daily), and participation in recreational or community activities (community activities: never or rarely/sometimes/often).
We assessed psychological factors that may affect musculoskeletal pain such as traumatic reactions (post-traumatic stress disorder check list: 44 points or more) [20, 21], psychological distress (Kessler psychological distress scale: 13 points or more) [22, 23], and somatic symptoms (none/one/two or more subjective symptoms except for musculoskeletal pain). Their subjective symptoms were categorised into 4 groups as follows based on a study by Zijlema et al. (2013): cardiopulmonary/autonomic (palpitations, breathing trouble), gastrointestinal (constipation/diarrhoea, loss of appetite, abdominal pain), general (headaches, dizziness), and others (difficulty urinating) .
Other confounding factors
We also gathered information regarding potential confounding factors such as age (40-64 years/65-74 years/75-89 years); sex (male/female); history of hypertension, diabetes, and dyslipidaemia (no/yes); education status (junior high school or less/high school/colleges/university or higher); having experienced the tsunami and/or indirect effects of the nuclear power plant accident (no/yes); house damage (none/partly destroyed/half or more destroyed); and loss of relatives or friends (no/yes).
A generalised linear model was used to compare basic characteristics, psychological factors, and lifestyle factors between the groups (absent versus prevalent, absent versus exacerbated, and prevalent versus exacerbated). Because even trivial differences may be statistically significant due to our large sample, we regarded differences of ±3% or more between the groups in addition to p-values less than 0.05 as statistically significant to highlight the significance of our findings. Multinomial logistic regression analysis was used to calculate odds ratios and 95% confidence intervals of each lifestyle factor related to prevalent musculoskeletal pain and exacerbated musculoskeletal pain. Two multivariable adjusted models were constructed. Model 1 was the adjustment for background information (age, sex, history of hypertension/diabetes/dyslipidaemia, and educational status) and disaster experience (tsunami/indirect nuclear power plant accident, house damage, loss of close person) and each lifestyle factor. This model was used to estimate odds ratios of what were considered unmodifiable characteristics. Model 2 was the further adjustment for psychological factors (traumatic reactions and psychological distress) and the number of uncomfortable symptoms (none, one, and two or more) to the model 1. This model aimed to estimate the non-psychometric effects (e.g. mechanical stress) from the psychometric effects.