Dr Tsukasa Sasaki, Department of Psychiatry, Health Service Center, University of Tokyo, 7-3-1 Hongo, Bunkyo, Tokyo 113, Japan. Email: email@example.com
Abstract Environmental factors, in addition to genetic factors, may be related to the development of panic attack (PA) and panic disorder (PD). Previous studies suggested that there may be seasonal variation in the onset of PA/PD and possibly a higher prevalence of PA/PD in colder areas. Also observed were lactate-induced PA and elevated serum cholesterol in PD patients. These suggest that living environment and lifestyle, such as weather conditions, preference of food and physical exercise, might play a role in the occurrence of PA and PD. The present study explored the association of such candidate factors with the development of PA and PD in 4000 Japanese subjects, using a questionnaire. The subjects were recruited from the general population of Japan, using stratified random sampling. Logistic regression with stepwise selection of variables was employed for statistical analysis. Variables including ‘dislike of physical exercise’, mostly in female subjects, and ‘living in areas with longer winter’, in male subjects, were suggested for associations with PA and PD among the candidate factors. The result is preliminary but indicates that lifestyle such as like/dislike of physical exercise and environmental factors including weather conditions could play a partial role in the development of PA and PD. Further investigations are required before firm conclusions can be reached.
The development of panic attack (PA) and panic disorder (PD) may be associated with a number of etiological factors. Many of the factors might be related to genetic components, but environmental components may also be involved.1 The present study attempted to explore factors that might have a correlation with environmental components in the development of PA/PD. Weather conditions may be one such possible factor. A greater rate of the first PA has been observed during warmer seasons in PD subjects, suggesting a role of weather in the development of PA/PD.2 When data from an international survey were analyzed, the prevalence of PA/PD appeared lower in warmer areas than areas with colder winter.3 According to the survey, in the Latin American region, for example, PA and PD were less frequent in warm Rio de Janeiro (1.4% and 0.0%, respectively) than in colder Santiago (5.5% and 0.6%, respectively).3 In the East Asian area, prevalence of PD was estimated to be low in warm Taiwan (0.4%),4 while the prevalence was moderate in Korea (1.5%), which has a much colder climate.5 The observations suggest that environmental factors such as weather condition (more specifically, living in a colder area) might affect the development of PA and PD, although other factors including genetic and sociocultural ones could have an effect on the observations.
We also focused on daily lifestyle factors that may play a role in PA and PD. Lifestyles may be closely correlated with environmental components, although they could also have a correlation with genetic factors. Elevated serum cholesterol has been consistently observed in PD.6–8 This might indicate a role of dietary habits and/or habits of daily exercise in the development of the disorder. In addition, an induction of PA by lactate or hard physical exercises has been observed.9–11 This also suggests a possible relation between exercise and PA/PD. Thus, the present study explored the role of factors related to lifestyle and living environment in the development of PA and PD. Four thousand subjects from the general population of Japan participated in the study by completing a questionnaire. The possible role of weather and lifestyle, such as dietary habits and daily exercise, were the focus of interest. The study is the first survey of PA and PD in the general population of Japan to be undertaken on this relatively large scale.
Subjects were recruited, using stratified random sampling, from a list of households nationwide maintained by Japan Statistics and Research, a market research company. The company maintains a panel of over 200 000 households with more than 800 000 persons for marketing and survey purposes. The panel was initially constructed from basic residence registers in Japan in 1984. A proportion of the sample is continuously renewed. Established panel usage policies are employed to guard against potential overuse and bias.
A cross-sectional survey was conducted from 1 to 13 November 2000. A questionnaire was mailed to a total of 5800 subjects. Relevant information pertaining to the purpose and conditions of survey (e.g. that the addressee was to respond without including his or her signature) was presented on the cover page of the questionnaire. A book coupon for 500 yen was enclosed, to encourage participation.
The stratified random sampling was employed to prevent deviation of sampling, with stratification by living area, age, and sex. Regarding the living area, the whole of Japan (or the 47 prefectures) was divided into 10 districts according to geographic location. As for age, subjects were stratified into five age groups covering 10-year intervals from 20 to 69 years. The stratification by living area, age, and sex therefore formed 100 cells. Each cell was to contain almost equal numbers of subjects to ensure a similar accuracy of data analysis across all cells. To achieve this balance, anonymous replies from 4000 subjects (69.0%) were retained for use in the statistical analysis. Thus, the subjects consisted of 2000 men (age: 44.3 ± 14.5 years) and 2000 women (age: 44.3 ± 14.2 years), consisting of 400 subjects (200 men and 200 women) from each of the 10 living areas and 800 subjects (400 men and 400 women) in each of the age groups (20s, 30s, 40s, 50s and 60s).
Items on the questionnaire assessed the presence of panic attack (single or repetitive), relevant psycho-physiological symptoms, expectation anxiety and agoraphobia, in addition to age, sex, current marital status, living area (prefecture) and questions related to lifestyle and habits. Panic attack and PD were defined according to Diagnostic and Statistical Manual of Mental Disorders (4th edn; DSM-IV) criteria. Attacks with four or more psycho-physiological symptoms were defined as ‘panic attack’. Subjects who experienced two or more PA and expectation anxiety were defined as suffering from PD.
The questions on lifestyles and habits included those regarding drinking habits (frequency of drinking per month), smoking habits (daily smoker or not) and self-evaluation of coping skills to stressors (designated as poor stress-coping), in addition to questions on dietary habit (preference for oily food including meat, designated as food preference) and like/dislike of physical exercise (designated as dislike of exercise). The questions on poor stress-coping, food preference and dislike of exercise were on a five-point scale that ranged from 1, strongly agree, to 5, strongly disagree. In considering weather conditions, ambient temperature was used as the criterion because we hypothesized, based on a review of previous studies, that PA/PD might be more prevalent in colder areas.3–5 More specifically, the number of days with the highest temperature of <0°C (designated as ‘winter days’)/year at the city (or cities) with the largest population in each living prefecture was analyzed.
Statistical analyses were conducted using logistic regression with stepwise selection of variables, employing the sas package (SAS Institute Inc, Cary, NC, USA). The independent variables employed in the logistic regression analyses consisted of age, sex, marital status (married or not), issues regarding lifestyle and the number of winter days/year in the living areas. The dependent variable was the diagnosis of PA or PD (present or absent). The diagnosis of PD with agoraphobia was not employed as the dependent variable because the number of the subjects with the diagnosis was too small for multivariant statistical analysis.
The prevalence of subjects whose response met the criteria for PA, PD and PD with agoraphobia was 6.6% and 5.0% and 3.3% in the women and 3.1%, 1.7% and 1.0% in the men, respectively. The age-distribution of PD in the female subjects was bimodal, with the bottom in the 40s and wide peaks in the 20–30s and the 50–60s, while the prevalence in the male subjects appeared to decrease with age after the 30s (Fig. 1).
Results of the logistic regression analyses are summarized in Table 1. In the analyses of all subjects, gender had the largest effect on the development of PA and PD (P < 0.0001 and P < 0.0001; odds ratio (female : male) = 2.11 and 3.03 for PA and PD, respectively). The dislike of exercise factor had a statistical trend for associations with PA and PD (P < 0.02 and P < 0.006; odds ratios = 2.16 and 2.73, respectively, when compared between the score 1 and the score 5 of the five-point question). The poor stress-coping factor also had a statistical trend for associations with PA and PD (P < 0.02 and P < 0.03; odds ratios = 1.97 and 2.05, respectively, when compared between the scores 1 and 5). Current marital status (not married) appeared to be associated with PA (P < 0.04; odds ratio = 1.49).
Table 1. Variables suggested for an association with the development of PA and PD by logistic regression
Odds ratio (95%CI)
Odds ratio (95%CI)
CI, confidence interval.
Odds ratios are for an increase of 10 ‘winter days’ (or days with the highest temperature <0°C).
Odds ratios are for comparisons between scores 1 and 5 on the five-point scale.
The logistic regression analyses were then conducted by sex, considering its marked effect. In female subjects, the poor stress-coping factor was indicated for associations with PA and PD (P < 0.0001 and P < 0.0001; odds ratios = 4.53 and 6.26, respectively, when compared between the scores 1 and 5). Dislike of exercise (P < 0.06; odds ratio = 1.89 when compared between the scores 1 and 5) and current marital status (or not currently married, P < 0.03; odds ratio = 1.69) had statistical trends for associations with PA in women. In men, instead, living in areas with longer winter (or more winter days/year), which are defined as days with the highest temperature <0°C) seemed to be associated with PA and PD (P < 0.02 and P = 0.0002; odds ratio = 1.08 and 1.16 for an increase of 10 winter days, respectively). No other variables were suggested for the association in male subjects. Frequencies of PA and PD by number of winter days and answers to the questions on like/dislike of physical exercise and self-assessment of stress-coping are summarized in Tables 2–4.
Table 2. Self-assessment of having methods to cope with stressors and prevalence of PA/PD
Strongly agree n (%)
Agree n (%)
Neutral n (%)
Disagree n (%)
Strongly disagree n (%)
PA, panic attack; PD, panic disorder.
Table 3. ‘Liking/dislike of doing physical exercises’ and prevalence of PA/PD
Strongly agree n (%)
Agree n (%)
Neutral n (%)
Disagree n (%)
Strongly disagree n (%)
PA, panic attack; PD, panic disorder.
Table 4. Length of ‘winter days’ (or days with the highest temperature <0°C) and prevalence of PA/PD
0–19 n (%)
20–79 n (%)
80 n (%)
PA, panic attack; PD, panic disorder.
The present study investigated factors that might be associated with the development of panic attack (PA) and panic disorder (PD) in the general population of Japan. The results of the multivariate logistic regression analysis indicated that the largest factor in the development of PA/PD was gender (or female sex). The prevalence of PA and PD was twice as high in women as in men. This is as expected from, and is consistent with, several previous studies.1,3,12,13 Among living environment, lifestyle and other factors, dislike of physical exercise and having poor measures to cope with stressors (poor stress-coping) had statistical trends for associations with PA and PD. When analyzed by sex, poor stress-coping was significantly associated with PA and PD in female subjects. The dislike of physical exercise and marital status (or being not married currently) factors had statistical trends for an association with PA in women. In the male subjects, however, these factors were not associated with either PA or PD, while living in areas with longer winter (or more specifically, more winter days) appeared to be associated with PA and PD.
Prior to the investigation, we hypothesized that several of these factors would be associated with PA and PD. An association with the dislike of exercise factor was expected for several reasons, including consistent observations of high serum cholesterol in PD,6–8 as well as lactate-induced PA in subjects with high susceptibility.9 We had also expected an association with preference for oily food for the former reason, but this was not found in the present analysis. Previous studies observed a reduced maximum intake of oxygen in PD patients,11 which might be related to an increase of lactate accumulation and susceptibility to panic attacks. Chronic avoidance of physical exercise may cause reduced intake of oxygen and accumulation of lactate, which could increase susceptibility to PA and PD. However, the situation might be that patients with PA/PD come to dislike and hence avoid physical exercise, which then contributes to accumulation of lactate and leads to further panic attacks, after the development of the attacks and disorder. Thus, whether dislike of physical exercise is a result or a cause of PA/PD remains unclear. This is also the case with poor stress-coping. Subjects with poorer measures to cope with stressors may be more likely to develop PA/PD, while patients with PA/PD, who often feel helpless due to persistent PA, might be likely to develop the self-evaluation of having poor ability to cope with stressors.
The reason why dislike of physical exercise and poor stress-coping were associated with PA/PD in female subjects but not in male subjects, is unclear. A possible explanation is that it could be due to the limited statistical power in the male sample; the number of male subjects with PA or PD was relatively small in the present study. A previous study observed elevated serum cholesterol, which could be related to dislike of physical exercise, in female PD subjects, but not in male subjects.6 This might be related to the observed associations between dislike of physical exercise and PA/PD in female subjects but not in male subjects, if avoidance of physical exercise has some correlation with elevated serum cholesterol. However, another study found elevation of serum cholesterol in PD regardless of gender.8 Thus, the reason remains to be further examined in future studies.
The associations between longer winter and PA/PD might be consistent with a trend toward higher prevalence in countries with a colder winter, which was found in a comparison of data from several areas of the world.3,4,12 For unknown reasons, however, the associations in the present study were confined to male subjects. This was also the case when the difference in average ambient temperatures between the coldest month and warmest month of year was employed as a variable, in place of number of winter days. The logistic regression suggested associations between temperature difference between the coldest and warmest months and PA/PD with P < 0.0001/P < 0.0001, respectively, in male subjects, but not in female subjects (data not shown in the Results or Table 1). Thus, the present data might suggest an association with weather conditions, especially seasonal change of ambient temperature, in male subjects. Our clinical impression is that weather conditions may affect the status of the disease in a substantial portion of PD patients. Panic attacks may be worsened during a hot–humid summer in a portion of patients, while in others expectation anxiety might be more frequent under colder and cloudy weather. But the effect does not appear to be confined to male patients. Thus, while the association of weather conditions in living areas might be partly consistent with the clinical impression, it is unclear why the association was confined to men. It is possible that the observed association was purely due to chance. The mechanism of the association also remains to be clarified, if it does exist. A difference in personality between people living in warm and cold areas is a candidate mechanism, in addition to the direct effect of weather on psychological and physical status.
The prevalence of PA and PD (at the time-point of the survey) was estimated at 6.6% and 5.0% in female subjects and 3.1% and 1.7% in male subjects (or 4.7% and 3.3% in all subjects). In previous studies, in comparison, lifetime prevalence of PD was estimated in women at 5.1% and 2.3% in North America12,13 and at 2.9% in South Korea;5 and in men, at 1.9% and 1.0% in North America12,13 and 0.5% in South Korea.5 Also, lifetime prevalence of PA was estimated to be 11.3% and 5.9% in North America, across men and women.12,13 Although the prevalence at the time-point of the survey may not be directly comparable with the estimates of lifetime prevalence, the prevalence of PD in the present study might be little higher than expected from the previous estimates, while the prevalence of PA could be comparable. This might be related to a limitation of the present study in that the investigation was conducted using a questionnaire and subjects were not directly interviewed. The prevalence, in addition to the results in the associated factors, should therefore be carefully considered. Thus far, however, no other large-scale investigation of PA or PD has been available in the general population of Japan. Further Japanese studies are required to give greater insight to these findings. The bimodal age-distribution in the prevalence of PD in women, however, might be consistent, to a degree, with a bimodal distribution of onset of panic disorder identified in a previous North American study.12
In conclusion, a significant effect of gender consistent with previous studies,1,3–5,12 a bimodal age distribution of the prevalence in women and possible associations between PA/PD and several factors were observed in the present study. These observations, while interesting, are mostly preliminary and further studies are required.
We thank Dr Mark Rogers for his assistance with English usage and grammar.