Association between Self-reported Stressful Feeling by Sacl and Mortality in a Japanese Community

A cohort study consisting of 4,291 residents of a Japanese community has been conducted since 1987. The associations between stressful feelings measured by Stress Arousal Checklist (SACL) and mortality of all causes and cancer were examined. The relative risks (RRs) for mortality for 7 years for high stressful state (SACL score:7-17) and moderate stressful state (3-6) subjects compared with low stressful state subjects (0-2) were 1. for cancer respectively, after being adjusted for sex, age, smoking, drinking, exercise and medical care use. When each item of the SACL was examined independently, six out of 17 items of SACL which were "not comfortable", "not contented", "not cheerful", "dejected", "nervous", "not pleasant" showed elevated RRs for all causes of mortality with statistical significance, after being adjusted for the possible confounding factors stated above. Especially, the adjusted RRs of "not cheerful" and "dejected" for all causes was 1.7 (95% CI=1.20-2.33), 1.7 (95% CI=1.17-2.46), respectively. The results suggested that stressful feelings of "not cheerful" and "dejected" might increase mortality. It has been considered that a stressful condition is related to mortality 1"23). Most of these studies have been done under the conditions of unusual life experiences such as concentration camp, natural disaster, surgical operation, or have focused on the specific disease, age group, psychological factors such as depression, hopelessness and so on. However, there are a few studies 24-27) on the association between stressful conditions in daily life measured by a self-administered questionnaire and mortality in a community population. The association of stress measured by the personality-stress inventory with mortality of cancer, coronary heart disease and other causes, was reported by Eysenck et al 2426). In these papers, stress was defined in terms of a subjects report of strain. It was stated that mortality of all causes was 20% or higher in the stressed group than in the non-stressed one when smoking and drinking were adjusted in the study 26). Rosengren et al also reported that stress defined as feeling tense, irritable or filled with anxiety, or as having sleeping difficulties at work or at home, was significantly associated with mortality from cardiovascular disease and for all other causes but not associated with mortality from cancer 27), even after adjusting for age, systolic blood pressure, serum cholesterol, smoking, BMI, diabetes , family history of myocardial infarction, occupational class, marital status, leisure time physical activity and registration for alcohol abuse. No such study has been …

It has been considered that a stressful condition is related to mortality 1"23).Most of these studies have been done under the conditions of unusual life experiences such as concentration camp, natural disaster, surgical operation, or have focused on the specific disease, age group, psychological factors such as depression, hopelessness and so on.
However, there are a few studies [24][25][26][27] on the association between stressful conditions in daily life measured by a selfadministered questionnaire and mortality in a community population.
The association of stress measured by the personality-stress inventory with mortality of cancer, coronary heart disease and other causes, was reported by Eysenck et al 2426).In these papers, stress was defined in terms of a subjects report of strain.It was stated that mortality of all causes was 20% or higher in the stressed group than in the non-stressed one when smoking and drinking were adjusted in the study 26).Rosengren et al also reported that stress defined as feeling tense, irritable or filled with anxiety, or as having sleeping difficulties at work or at home, was significantly associated with mortality from cardiovascular disease and for all other causes but not associated with mortality from cancer 27), even after adjusting for age, systolic blood pressure, serum cholesterol, smoking, BMI, diabetes, family history of myocardial infarction, occupational class, marital status, leisure time physical activity and registration for alcohol abuse.
No such study has been reported up to the present in Japan where sociocultural situation is quite different from western countries.This study aims to determine whether self-reported stressful feelings measured by Stress Arousal Checklist 2&31) or each item of the checklist is associated with mortality among the people living in a Japanese community.

Study Populations
The study subjects were derived from a part of the Miyako study.The Miyako Study consists of three community cohort populations (one rural area, one commercial area and one industrial area) in Miyako county, Fukuoka, Japan.This study has been carried out since November 20, 1987.The Miyako study was established to clarify the relationship between lifestyle and its health effects.One commercial area -Y city among 3 cohort populations was used in the present study.Y city is located in the southeastern part of Fukuoka Prefecture, in Kyushu and the population in 1994 was approximately 69,000.
All of the residents (5,386 persons) born between 1907 and 1956 (30 to 79 years of age) and living in 39 randomly selected regions out of a total of 176 regions in the city were identified as a study population.The subjects were requested to answer the self-administered questionnaire consisting of 254 items regarding stressful feelings as well as other questions such as health habits, lifestyle, medical care use and so on.The response rate (No. of respondents / No. of residents) was 79.7% (4,293 / 5,386).After excluding the persons with an incomplete questionnaire, 3,298 persons were defined as cohort subjects in the present study.The study population was followed up until December 31, 1994.Twenty-seven persons were excluded due to the follow-up loss.As a result, 3,271 persons (1,480 male, 1,791 female) with complete answers for SACL were analyzed in the present study.

Method of Follow-Up
The vital status was confirmed by basic resident registers after receiving permission from the City Government.The cause of death for deceased cases was obtained from death forms at a local health center with permission of the Management and Coordination Agency of the Japanese Government.The successful follow-up rate was 99.2%.The underlying causes of death were decided and coded, according to the International Classification of Diseases, the 9th Revision (ICD 9).The causes of death used in this analysis were all causes, cancer (ICD 9: 140-208) and circulatory diseases combined with coronary heart disease (ICD 9: 393-398, 410-429) and cerebrovascular disease (ICD 9: 430-438).

Stress Arousal Checklist (SACL) for Stressful Feeling
Stress Arousal Checklist (SACL) was used as a part of the self administered questionnaire in the present cohort study.SACL was originally a mood adjective checklist consisting of 17 items on stress and 13 items on arousal 28).In this study, a Japanese translated version (Kumashiro M, personal communication) of the SACL consisting of 17 items for stress was used to evaluate the stressful feelings.
Self-reported stressful feelings were examined by asking questions such as, "Describe your feelings at this moment : comfortable, calm, distressed, relaxed, contented, tense, easy, up-tight, cheerful, apprehensive, peaceful, dejected, nervous, bothered, pleasant, worried and jittery".The response scales are "definitely feel", "feel slightly", "do not understand or cannot decide" and "definitely do not feel".If "definitely feel" or "feel slightly" had been circled for a positive adjective , it was assigned 1, otherwise 0. If "do not understand or cannot decide" or "definitely do not feel" had been circled for a negative adjective, it was assigned 1, otherwise 0, according to the original scoring method ').Thus the stress scale ranged from 0 to 17. Approximately each 30 percentile of the total score was used as the cut off point to categorize the total subjects into three groups.The total score of SACL was, then, categorized into 3 groups.These 3 groups were low stressful state (total SACL score: 0-2), moderate stressful state (3)(4)(5)(6) and high stressful state (7)(8)(9)(10)(11)(12)(13)(14)(15)(16)(17).

Possible Confounding Factors
Age, sex, smoking, drinking, exercise and medical care use were considered as possible confounding factors in our study.The definitions of each possible confounding factor are as follows; smoking : present smoker only, drinking : alcohol intake everyday or almost everyday during the past year, exercise getting exercise more than twice a week, medical care use under medical care for angina, gastric ulcer, hepatitis, liver cirrhosis, hypertension, myocardinal infarction, valvular disease of the heart, asthma, stroke, diabetes mellitus, cancer, hysteromyoma and arrhythmia.

Analytical Methods
For analysis, relative risks were calculated by the Cox proportional hazard model with a SAS PHREG procedure .
Using the total SACL score for each subject, crude and adjusted relative risks with 95% confidence intervals (CI) for both, all causes and cancer mortality, were obtained for the moderate stressful state and high stressful state compared to low stressful state.
Using each item of SACL independently, effects on mortality for each item were examined by crude and adjusted relative risks with 95% confidence interval calculated and analized as stated above.These relative risks on mortality calculated for "No" to "Yes" in positive feeling items to "Yes" , and for subjects with persons who answered "Yes" to "No" answers in negative feeling items along with the 95 percent confidence interval.

RESULTS
The characteristics of the analyzed 3,271 subjects are presented in Table 1.There were more females (55%) than males among the study subjects.More than a quarter of the subjects were in their 30s and there was a decrease in number as the age increased.Age of the subjects was equally distributed in both sexes.Distribution of SACL scores was not statistically differ-Table 1. Basic characteristics of the subjects in this study.
smoker: present smoker only drinker: alcohol intake everyday or almost everyday during the past year exercise: getting exercise more than twice a week medical care user: under medical care for angina, gastric ulcer, hepatitis, liver cirrhosis, hypertension, myocardinal infarction, valvular disease of the heart, asthma, stroke, diabetes mellitus, cancer, hysteromyoma and arrhythmia ent between male and female.The proportion of daily habits such as smoking, drinking, exercise and medical care use are also shown in Table 1.
Table 2 shows the difference in risk factors among each stressful state in order to consider possible confounding factors.No differences in SACL scores between males and females were observed.The proportion of subjects with high stressful state decreased as the age increased.The proportion of smokers and exercisers are higher among the high stressful state group than those among the low group.It was found that subjects with a higher stressful state tended to be smokers and more active in doing exercise.
Among these subjects, the number of deaths was 175 as shown in Table 3. Forty-three percent (n=76) had died of cancer, which was the leading cause of death in this group.Twenty -three percent (n=41) had died of a stroke and heart disease.The remaining 33 percent (n=58) had died of other causes.
For all causes of death, crude and adjusted relative risks with 95 percent confidence intervals for 7 years and 4 years are presented in Table 4.
In order to reduce the possible effect on mortality due to unreported or unrecognized health conditions at the starting point, deaths during the first 3 years was excluded for analysis.items.These six items were "not comfortable", "not contented", "not cheerful" , "dejected", "nervous" and "not pleasant" with relative risks of 1.4, 1.4, 1.7, 1.7, 1.5; 1.5, respectively.
After excluding the first 3 years observation, the adjusted RRs for the 6 items of SACL were almost of the same magnitude or slightly smaller than those in the former analysis.However, the item of "not cheerful" showed even a higher RR of 2.1 (95% CI=1.38-3.12)after excluding the first 3 years.
For cancer mortality, crude and adjusted relative risks with 95 percent confidence intervals were shown in Table 5.During the 7 years follow up, after being adjusted for the possible confounding factors, the RRs of moderate and high stressful state to low stressful state tended to be slightly increased as 1.5 (95% CI=0.80-2.99)and 1.3 ( 95% CI=0.67-2.61),even though they were not statistically significant.After excluding the first 3 years observation, the adjusted RRs of moderate and high stressful state to low stressful state were 1.5 (95% CI=0.67-3.39)and 1.2 (95% CI=0.51-2.74).No definite association was observed, even though slightly elevated risks were obtained.There were no dose-response relationship observed.
Among the analysis of each item of SACL after adjusting for sex, age, smoking, drinking, exercise and medical care use, "dejected" showed elevated relative risk of 2 .0 with statistical significance at the 5% level over the 7-year observation period.The two items of "not comfortable" and "not cheerful" revealed also increased relative risks of 1.6 each, although they are not statistically significant.After calculating the adjusted RRs, excluding the first 3 years, the relative risk of "dejected" decreased to some extent (RR=1.7,95% CI=0.88-3.40).However, adjusted RRs for "not cheerful" increased up to 2.0 (95% CI=1.11-3.66).

DISCUSSION
The present study is the first report from Japan to examine the association between stressful feelings measured by SACL and mortality, using a population based cohort study.
In stress studies, it is quite difficult to define stress, because the term "stress" has been used in various ways.Kasl summarized the definition of stress as follows : 1) as an environmental condition ; 2) as the appraisal of an environmental situation ; 3) as a response to the environmental condition or to its appraisal ; 4) as an interactive term indicating the relationship between environmental demands and the persons capacity to meet these demands 33).
To measure psychological reactions to the stress defined as a response to the environmental condition or to its appraisal, there are several test batteries for measurement, such as Beck Depression Inventory 34), Center for Epidemiologic Studies Depression Scale 35), Self-rating Depression Scale 36), State-Trait Anxiety Inventory 37, Symptom Checklist-90-Revised 38) etc.None of these measurements, however, have been considered to be definite test batteries to measure stressful state, because no golden standard for stress level exists 29).
Stress in SACL was defined as an internal response to the perceived favorability of the external environment 28).
The first version of SACL with 34 mood describing adjectives was developed in 1978 by Mackay et al 28), after reanalysis of Mood Adjective Checklist (MACL) with 45 adjectives originated by Thayer in 1967 38).
The present version of SACL was proposed with a 30-item self-report list of adjectives to describe an individual's psychological state, as related to his perception of external environment and emotional reaction to it 31).SACL was translated by Kumashiro into Japanese for use to determine Japanese expressions.In order to verify usefulness of SACL inventory, several studies have been conducted as shown below.For example, Burrows and his colleagues reported that an intensive salestraining course produced significant changes in the stress score in SACL and that a significant correlation (r=-0.80,p=0.05) between stress score and blood glucose level was observed 40).
Cox found a significant increase in self-reported 'stress' after a prolonged and monotonous repetitive tasks which are supposed to be stressful 41).
Duckro compared SACL score with State-Trait Anxiety Inventory-State which is a well known measure of distress.It was found that scores on SACL were significantly correlated with those on STAI (r=0.85,p<0.01) 42).
In Japan, Kumashiro found that the average stress score of the software workers who were involved in the daily operation of a computer was significantly higher than that of people involved in other lines of work such as petrochemical or machine shop (p<0.05), which was as expected 43).The reliability coefficients (Cronbach's alpha) for 17 items of SACL was 0.87, which would be considered to have internal consistency among each item of SACL.
These, therefore, support that SACL could be one of the measurements for stressful conditions.
Vital status and cause of death were examined for all subjects without knowing the SACL score.Therefore, no information bias was possible.As no difference was found on the follow-up rates among 3 groups (low stressful state : 98.9%, moderate stressful state : 99.3%, high stressful state : 99.3%), it is considered that no selection bias existed.
Mortality in the cohort could be influenced by various factors, such as age, sex, smoking, drinking, exercise and disease status etc.In order to reduce the effect of these possible confounding factors on the relationship between SACL score and mortality, such factors as stated above have been adjusted for calculation of relative risks, using Cox proportional hazard model 44).
Although disease status reported at the baseline has been adjusted as a confounding factor by the Cox proportional hazard model, mortality might be influenced by unreported or unrecognized health conditions.In order to avoid any possible influence, the relative risks were calculated after the excluding of deaths during the first 3 years of follow up.Through this analysis, no substantial changes in the relative risks were obtained, although statistical power was reduced due to the small number of observed deaths.
Therefore, adjusted relative risks obtained here would not be biased nor confounded in order to see the association between SACL and mortality.
No significant difference was found in the association between SACL score and mortality in the present study.Possible explanations would be: 1) SACL may not be sensitive enough to measure stress in daily life, 2) Japanese may perceive or respond toward stress differently from western people, 3) It may be lack of statistical power.There is, however, no clear explanation for the results obtained here at this point.
Items of "cheerful" and "dejected" showed moderately elevated risks.The finding of elevated risks of "not cheerful" and "dejected" has been never reported so far .However, because no analysis or reports have been done on each item of SACL independently, there is no way to make a comparison of our results with others.
There might be a possibility that the result was obtained by chance.However, the items "dejected" and "cheerful" were found to be related to mortality in our preliminary analysis using a different population.It could be, these associations obtained in the present study would not be able to be explained by chance alone.
Although no psycho-biological mechanism on the effect of stress on mortality is known.Therefore, further studies should be required to verify these results obtained here.and the public health staff at Miyako Health Center for their cooperation.
We also would like to thank Ms.Yuko Uemura and Yoko Wada for their help during data analysis.

Table 2 .
Risk factors with stressful state.

Table 4 .
Relative Risk by Stressful state on the death of all causes.
# : adjusted for sex, age, smoking, drinking, exercise and medical care use Crude relative risks of moderate and high stressful state to low stressful state for all causes for the last 4 years were 0.8 (95% CI=0.48-1.26)and0.8(95%CI=0.49-1.26),respectively.Adjusted risks for moderate and high stressful state were 1.0 (95% CI=0.60-1.76)and1.2(95%CI=0.72-2.03),respectively.*:relativeriskNo clear association was observed.Using the response to each item of SACL independently, the association between stressful state of each item was examined by relative risks as shown in

Table 4 .
Adjusted relative risks were observed with statistical significance for 6 out of 17

Table 5 .
Relative Risk by Stressful state on the death of cancer.
# : adjusted for sex, age, smoking, drinking, exercise and medical care use * : relative risk