Abstract

Background: The role of personality in the causation of cancer has been controversial. We examined this question in a large, prospective study. Methods: From June through August 1990, 30 277 residents of Miyagi Prefecture in northern Japan completed a Japanese version of the short form of the Eysenck Personality Questionnaire-Revised and a questionnaire on various health habits. There were 671 prevalent cases of cancer at baseline, and 986 incident cases of cancer were identified during 7 years of follow-up, through December 1997. We used Cox proportional hazards regression to estimate the relative risk (RR) of incident cancer (total, stomach, colorectal, breast, and lung) according to four levels of each of four personality subscales (extraversion, neuroticism, psychoticism, and lie), with adjustment for sex, age, education, smoking, alcohol use, body mass index, and family history of cancer. Statistical tests were two-sided. Results: Multivariable RRs of total cancer for individuals in the highest level of each personality subscale as compared with those in the lowest were 0.9 for extraversion (95% confidence interval [CI] = 0.7 to 1.1; Ptrend = .32), 1.1 for psychoticism (95% CI = 0.9 to 1.3; Ptrend = .96), 0.9 for lie (95% CI = 0.7 to 1.0; Ptrend = .19), and 1.2 for neuroticism (95% CI = 1.0 to 1.4; Ptrend = .06). There were no associations between any personality subscale and risk of specific cancers. Neuroticism showed statistically significant positive, linear associations with prevalent cancer at baseline (Ptrend<.001) and with the 320 incident cancer cases diagnosed within the first 3 years of follow-up (Ptrend = .03); however, it showed no association with the 666 cases diagnosed during the fourth through the seventh years of follow-up (Ptrend = .43). Conclusion: Our data do not support the hypothesis that personality is a risk factor for cancer incidence. The association between neuroticism and prevalent cancer may be a consequence, rather than a cause, of cancer diagnosis or symptoms.

Personality has long been hypothesized to play a causative role in the development of cancer. Around 200 A.D., Galen observed that melancholic women were more prone to cancer than were women of a sanguine temperament (1). In 1962, Kissen and Eysenck (2) conducted one of the first modern studies of the association between personality and cancer and reported that patients with lung cancer were more likely to be extraverted and less likely to be neurotic as compared with hospital controls.

The findings of subsequent studies examining the association between personality and cancer incidence (1,311) or mortality (1214) have been inconsistent. Of six retrospective studies reported to date (16), four found statistically significant associations between higher scores for extraversion (2,3), lower scores for neuroticism (2,4,5), and lower scores for trait anxiety (4) and increased risk of cancer of the breast (4), lung (2,5), and all sites combined (3). The other two studies found no association between personality scales and breast (1) or gastric (6) cancer. Of 10 prospective cohort studies (714), five found statistically significant associations between “Type 1” personality [understimulation (12)] or hopelessness (8) and total cancer and between anti-emotionality (an absence of emotional behavior or a lack of trust in one’s own feelings) (7) and breast cancer. The other five prospective studies found no statistically significant associations between personality and breast cancer (10) or total cancer (9,11,13,14). The majority of these studies had methodological limitations, including the use of a retrospective design (16), a small number of cancer cases [ranging from 29 (12) to 249 (5)], and failure to control sufficiently for potentially confounding variables such as smoking and alcohol use (17,10,12).

To further examine the association between personality and the risk of cancer, we conducted a population-based prospective cohort study in rural Japan. This study used the Japanese version of the personality questionnaire developed by Eysenck and colleagues (15) and involved 986 incident and 671 prevalent cancer cases, the largest number of cases to date.

Subjects and Methods

Study Cohort

We have reported the design of this prospective cohort study in detail elsewhere (16,17). Briefly, we delivered two self-administered questionnaires to all 51 921 residents aged 40–64 years in 14 municipalities of Miyagi Prefecture in rural northern Japan from June through August 1990. The first questionnaire asked about various health habits, and the second was the Japanese version of the Eysenck Personality Questionnaire-Revised (EPQ-R) Short Form (15). The questionnaires were delivered to, and collected at, the subjects’ residences by members of health promotion committees appointed by the municipal governments. The response rate for the first questionnaire was 91.7% (n = 47 605), and that for the second questionnaire among the respondents to the first was 79.8% (n = 41 442). The study protocol was approved by the institutional review board of Tohoku University Graduate School of Medicine. We considered the return of self-administered questionnaires signed by the subjects to imply their consent to participate in the study.

Exposure Data

The first questionnaire asked about demographic variables, personal and family histories of cancer and other diseases, and health habits, including smoking, alcohol use, diet, use of health services, and self-reported height and weight. The second questionnaire was a Japanese translation of the original English version of the EPQ-R Short Form, one of a series of personality inventories developed by Eysenck and colleagues (18). The EPQ-R has 48 questions with dichotomized responses (yes or no); there are 12 questions for each of four personality subscales (extraversion, neuroticism, psychoticism, and lie). Scores on each subscale range from 0 to 12, with higher scores indicating a greater tendency to possess the personality trait represented by each subscale. Extraversion represents sociability, liveliness, and surgency; neuroticism represents emotional instability and anxiousness; psychoticism represents tough-mindedness, aggressiveness, coldness, and egocentricity; and lie represents unsophisticated dissimulation and social naivety or conformity (19).

Several of Eysenck’s personality questionnaires have been translated to Japanese (15,20,21) and have been applied in various research and clinical settings, including studies examining associations with tobacco and nicotine dependence in male smokers (22); health perception, life satisfaction, and other measures of quality of life in rural community residents (23); behaviors for coping with stress in female college students (24); and adjustment of high school students in an international exchange program (25). These studies indicate that Eysenck’s personality theory is as applicable to Japanese individuals as it is to Western individuals.

In previous work (15), we developed the Japanese version of the EPQ-R and examined its reproducibility and validity among 329 college students and 253 adults. Cronbach’s α coefficient, a measure of internal consistency, was greater than 0.70 for all subscales except psychoticism (0.42 for college students and 0.48 for adults). Test–retest reliability coefficients of the four subscales over a 6-month period ranged from 0.70 to 0.85, indicating substantial stability. Confirmatory factor analysis supported the original theoretical structure of the four scales proposed by Eysenck and colleagues (18). Scores on the four subscales were highly correlated with scores on similar subscales in the Japanese versions of the Sixteen Personality Factor Questionnaire (26) and the Maudsley Personality Inventory (20), indicating that the questionnaire had a high degree of concurrent validity.

Follow-Up

We used population registries in the 14 municipalities to ascertain the vital and residential status of the subjects from June 1, 1990, through December 31, 1997. We identified incident cases of cancer by computerized record linkage with the Miyagi Prefectural Cancer Registry, which covers the study area (27).

Of the 41 442 subjects who responded to the two questionnaires, we excluded 8618 subjects for whom responses to any of the 48 items in the EPQ-R were missing and 54 subjects who responded “yes” or “no” to every one of the 48 items. We also excluded 2493 subjects who indicated that the two questionnaires had been completed with the aid of other family members because we considered that such aid might have affected the response patterns of the study subjects. For the analysis of incident cancer, we also excluded 671 subjects with prevalent cancer ascertained either by self-reporting in the health-habit questionnaire or from cancer registry records. Consequently, 29 606 subjects (14 473 men and 15 133 women) remained for the analysis of incident cancer. Over the follow-up period, 986 incident cases of cancer (588 in men and 398 in women) were identified.

We observed little difference between subjects who completed the two questionnaires (n = 41 442) and those who remained for the analysis of incident cancer (n = 29 606). Specifically, for men in the two groups, the mean ages were 51.7 and 51.1 years, the prevalence of current smokers was 61.1% and 60.8%, respectively, and the prevalence of family history of cancer was 25.8% in both groups. Characteristics of women in the two groups were also similar. Furthermore, the subjects who remained for the current analysis made up 57% of the total population of adults aged 40–64 years at baseline (June 1, 1990) in the study area. The two groups were also similar with respect to the distribution of sex (men made up 48.9% and 48.7% of the groups, respectively) and age classes (e.g., subjects aged 50–59 years made up 36.9% and 38.7% of the groups, respectively). Taken together, these findings indicate that the subjects we analyzed were representative of the total population in the study area.

Statistical Analysis

We counted person-years of follow-up for each subject from June 1, 1990, until the date of diagnosis of cancer, date of emigration outside the study districts (because of logistical limitations), date of death, or the end of the study period (December 31, 1997), whichever occurred first. A total of 215 810 person-years resulted. A total of 1088 subjects (3.7% of the analytic cohort) were lost to follow-up during the study period.

Each personality subscale was divided into four categories to divide the total subjects as closely as possible into even-sized quartiles. Relative risk (RR) was computed as the incidence among subjects in each category of a personality subscale divided by the incidence among subjects in the lowest category. We used Cox proportional hazards regression to adjust for sex, age, and other potentially confounding variables (28) using the SAS PHREG procedure in the SAS version 8.2 statistical software package (SAS Institute, Cary, NC). The assumption of proportional hazards was verified graphically. We repeated all analyses after excluding the 320 cancer cases diagnosed after baseline in the first 3 years of follow-up (194 men and 126 women) in case health issues related to as-yet-undiagnosed cancer affected personality. P values for testing statistical significance of linear trends were calculated by treating personality subscales as continuous variables. All P values were two-tailed. This study has approximately 85% statistical power, with a two-sided α error level of 5%, to detect a true hazard ratio of 1.3 for total cancer among the highest versus lowest categories of the personality subscales (29).

In addition to sex and age, we considered the following variables as potential confounders a priori: cigarette smoking (never smoked, smoked in the past, currently smoking 1–19 cigarettes per day, or currently smoking ⩾20 cigarettes per day), alcohol consumption (never drank alcohol, drank in the past, currently drinking 22.7 g of alcohol or less per day, or currently drinking 22.8 g of alcohol [corresponding to one small bottle of sake] or more per day), body mass index in kg/m2 (⩾18.4, 18.5 to 24.9, or ⩾25.0); education (in school until age 15 years or younger, age 16–18 years, or age 19 years or older), and family history of cancer (presence or absence in first-degree relatives). For the adjustment of smoking, we entered three types of smoking variables into separate multivariable models, including 1) categorical variables for smoking status, as mentioned above; 2) the number of pack-years of smoking as a continuous variable; or 3) the number of the cigarettes currently smoked per day as a continuous variable. Because we observed similar results irrespective of the choice of these variables, we presented the results with the categorical variables for smoking status.

For the cancer risk estimates, we analyzed both total incident cancer (986 cases) and the four most common cancer sites among the subjects: stomach (229 cases), colorectum (186 cases), lung (108 cases), and breast (87 cases, all women). We also used combined cancer endpoints according to whether they were associated with smoking; for smoking-associated cancers, we included cancers of the lung (108 cases), esophagus (34 cases), pancreas (32 cases), cervix (25 cases), bladder (21 cases), oral cavity (21 cases), and pharynx (16 cases) (30), for a total of 257 cases. We excluded from the analysis of smoking-associated cancers 12 cancer cases whose primary sites were not specified, thus leaving 717 cases designated as not smoking-associated.

We also conducted additional analyses to examine how study design (retrospective or prospective) and duration of follow-up (in prospective analyses) affected the associations between personality scales and the risk of total cancer. In the retrospective analysis, we used as an endpoint the 671 cases of cancer that were prevalent at baseline and had been ascertained from self-reports in the health-habit questionnaire or from cancer registry records. We estimated odds ratios of the presence of prevalent cancers for different levels of personality subscales by unconditional logistic regression (28). In the prospective analyses, we used two durations of follow-up: the first analysis involved the first 3 years of follow-up from the baseline, and the second involved 7 years of follow-up but excluded cancer cases diagnosed within the first 3 years.

Results

We first compared the characteristics of subjects in the highest and the lowest categories (i.e., approximate quartiles) of each personality subscale (Table 1). Both men and women in the highest category of extraversion were more likely than those in the lowest to be current smokers, current drinkers, and overweight (i.e., BMI of 25 kg/m2 or higher). Subjects in the highest category of neuroticism were less likely than those in the lowest category to be overweight, and those in the highest category of psychoticism were more likely than those in the lowest to be current smokers. Subjects in the highest category of the lie scale were older than those in the lowest category and less likely to be current smokers, current drinkers, overweight (men only), and to have attended school until age 19 or older (women only).

In multivariable Cox proportional hazards regression analyses, we found no statistically significant association between any of the four personality subscales and the RR of total cancer (Table 2). Neuroticism showed a borderline statistically significant linear trend association with increased RR of cancer (Ptrend = .06), but the association disappeared after the 320 cases of cancer diagnosed in the first 3 years of follow-up were excluded from the analysis.

We conducted stratified analyses of relative risk of cancer according to sex, 5-year age class, cigarette smoking, alcohol consumption, body mass index, education, and family history of cancer (data not shown). However, the associations between personality subscales and the risk of cancer were not markedly modified by any of these variables.

We also found no statistically significant associations between any of the personality subscales and the risk of the four cancers that were most common among the subjects (stomach, colorectum, lung, and breast) and cancer at sites defined according to whether or not they are associated with smoking (Table 3). Across all levels of extraversion, from lowest to highest, there was some evidence of a decreased risk of breast cancer (Ptrend = .046), but the association disappeared when the 32 cases of breast cancer diagnosed in the first 3 years of follow-up were excluded.

In the analyses that examined the effect of study design (i.e., retrospective or prospective) and duration of follow-up (in the prospective study) on associations between personality subscales and the risk of total cancer, we found that the association with neuroticism was different depending on which analysis was done (Fig. 1). In the retrospective analysis, neuroticism showed a statistically significant linear, positive association with the presence of cancer at baseline (Ptrend<.001). In the prospective analysis with only 3 years of follow-up, neuroticism also showed a statistically significant linear, positive association with the RR of incident cancer (Ptrend = .03). However, in the second prospective analysis, in which subjects were followed for 7 years but in which cancer cases diagnosed in the first 3 years of follow-up were excluded, neuroticism showed no association with the incident risk of cancer (Ptrend = .43). The other three subscales of personality showed no association with the risk of total cancer, regardless of study design or duration of follow-up (data not shown).

Discussion

In this population-based prospective cohort study in rural northern Japan, we found no association between any of four personality subscales measured by the Eysenck questionnaire (extroversion, neuroticism, psychoticism, and lie) and the risk of either total cancer or cancers of the stomach, lung, colorectum, and breast. Our study had several methodologic advantages compared with previous studies of personality and cancer, including a prospective design, recruitment of subjects from the general population, and extensive control for potentially confounding variables such as smoking, alcohol use, body mass index, family history of cancer, and education (a measure of socioeconomic status).

Our findings in the analyses according to study design and the duration of follow-up (Fig. 1) may provide clues for interpreting the discrepancies among previous studies on personality and cancer. That is, we found a statistically significant positive association between neuroticism and prevalent cases of cancer at baseline and with cancer cases diagnosed in the first 3 years of follow-up. A neurotic tendency among subjects with prevalent cases of cancer may be a consequence, rather than a cause, of having been diagnosed with cancer. Similarly, a neurotic tendency of subjects with cancer diagnosed in the early years of follow-up may be a consequence of subclinical symptoms caused by cancers that were present but undiagnosed at baseline. In contrast, we found no association between neuroticism and the risk of being diagnosed with cancer during the later years of follow-up. The majority of these cancer cases would have been asymptomatic at baseline, when personality was assessed. This analysis would be least susceptible to distortion when examining the temporal association between personality (putative cause) and the development of cancer (effect).

Although the questionnaire that we used, EPQ–R, measures general personality dimensions rather than personality dimensions more specifically hypothesized to be associated with cancer, previous studies finding statistically significant associations with cancer risk also use questionnaires developed by Eysenck (25) and another questionnaire developed for general use (7). It is therefore unlikely that the observed lack of association is due mainly to the use of the general personality questionnaire.

Several investigators have developed measures of personality dimensions that are more specifically hypothesized to be associated with cancer and have examined the associations between these measures and cancer risk. The first line of work was initiated by Temoshok et al. (31), who observed that tumor thickness in patients with malignant melanoma was positively associated with “Type C” personality, which they described as cooperative, unassertive, patient, suppressive of negative emotions, and accepting/compliant with external authorities. Subsequent studies have focused mainly on the association between components or related factors of “Type C” personality (such as emotional suppression and helplessness/hopelessness) and cancer recurrence or survival rather than incidence, but there is little consistent evidence that these factors play an important role in cancer survival or recurrence (32). Other studies have examined the association between cancer incidence and anti-emotionality (7,33), repression (10), and hopelessness (8). Although some studies (7,8,33) observed statistically significant positive associations, they had methodologic limitations, including a small number of cancer cases [ranging from 73 (8) to 166 (33)] and poor control for potentially confounding variables (33).

The second line of work was conducted by Grossarth-Maticek and colleagues (12). They classified personality into four subtypes: “Type 1” (understimulation), “Type 2” (overarousal), “Type 3” (ambivalence), and “Type 4” (personal autonomy) and examined the association between the personality type and disease in three prospective studies among a population in Yugoslavia, a population in Heidelberg, and a stressed group in Heidelberg. They found “Type 1” personalities to be prone to cancer, “Type 2” to be prone to coronary heart disease, and “Type 3” and “Type 4” to be healthy (12). They observed that the proportion of subjects who died from cancer during the follow-up was markedly higher in “Type 1” individuals than in “Type 4” individuals (e.g., 46.2% versus 0.6% in the Yugoslav study). However, various criticisms have been raised about these studies, including vague descriptions of inclusion and exclusion criteria for the study subjects (9,34,35), insufficient information on establishment and confirmation of cancer diagnosis (9,3537), and no information on validity of personality measurement (9,34,37,38).

In this study, we examined in detail potential confounding and effect modification by smoking and other covariates on the associations between personality scales and cancer risk. Various studies have reported that smoking is associated with extraversion or other personality characteristics (39,40), and some investigators have considered these findings as evidence that the association between smoking and lung cancer is confounded by personality variables (41). We found that current smokers tended to score high on extraversion, low on neuroticism, high on psychoticism, and low on the lie scale compared with never smokers (Table 1). We also observed that other risk factors for cancer, such as alcohol consumption and being overweight, were associated with these personality variables. Nevertheless, we observed little difference in the point estimates of RRs whether or not multivariable adjustments were made for smoking and other variables (Table 2). We also observed no material variations in the results of analyses stratified by these covariates. These findings indicate that the observed lack of association between personality factors and incident risk of total cancer in our study was not substantially confounded or modified by smoking, using alcohol, or being overweight.

Although this study focused on the association between personality and cancer incidence, we have observed an association between high alcohol consumption and all-cause mortality among men in this cohort (17). We have also observed in this cohort positive associations between smoking and lung cancer, between alcohol consumption and several cancers, and between family history of cancer and risk of total cancers (Tsubono Y: unpublished observations). Therefore, our subjects should be considered a typical series of persons with regard to the associations between risk factors other than personality and cancer incidence or general health.

To our knowledge, this is the largest prospective cohort study of personality and cancer to date, involving 986 incident cases of cancer at all sites combined. However, the number of cases of individual cancer sites was still only modest at best, even for the most common cancer sites observed among our subjects. Therefore, our study may not have had sufficient statistical power for detecting small increases or decreases in the risk of cancer at individual sites associated with personality scales. This is particularly the case for hormone-related cancers, such as breast, endometrial, and prostate cancers, for which the incidence rates are lower in Japan than in Western countries (27). Nevertheless, our findings support the hypothesis that personality does not play a substantial role in the causation of cancer in general.

Table 1.

Characteristics of study subjects according to the highest and lowest of four categories of each of four personality subscales

 Personality subscale score* 
 Extraversion Neuroticism Psychoticism Lie 
Characteristics ⩽3 ⩾9 ⩽3 ⩾9 ⩽2 ⩾5 ⩽5 ⩾10 
*Each personality subscale (scored on a scale of 0–12) was divided into four categories approximately equal in size, on the basis of the score in the population as a whole. Consequently, different personality subscales have different cutoff scores. 
Men 
    No. of subjects 4103 3249 4505 2682 3356 4717 4391 2609 
    Age, y, mean ± SD 50.8 ± 7.6 51.2 ± 7.6 51.4 ± 7.7 50.6 ± 7.6 52.6 ± 7.7 49.7 ± 7.4 48.3 ± 7.1 54.5 ± 7.4 
    Current smoker, % 56.4 61.8 61.1 57.3 52.7 64.6 64.4 53.4 
    Current drinker, % 71.4 81.1 76.7 76.9 75.7 76.9 79.6 71.4 
    Body mass index ⩾25.0 kg/m2, % 23.0 32.3 29.0 25.4 25.2 28.0 28.4 24.8 
    Education, in school until age 19 years or older, % 13.9 16.7 15.9 14.1 15.8 14.1 15.7 14.4 
    Family history of cancer in first-degree relatives, % 26.1 25.9 25.3 25.1 27.6 25.5 26.7 26.0 
Women 
    No. of subjects 4625 2941 4038 3118 6246 2189 2812 3415 
    Age, y, mean ± SD 50.8 ± 7.4 51.1 ± 7.3 51.6 ± 7.5 50.6 ± 7.4 51.6 ± 7.5 50.0 ± 7.1 47.5 ± 6.6 54.3 ± 6.8 
    Current smoker, % 5.6 10.1 6.3 7.8 5.1 11.0 9.9 4.7 
    Current drinker, % 17.6 31.6 22.9 24.2 20.2 26.5 32.2 15.8 
    Body mass index ⩾25.0 kg/m2, % 25.6 33.7 34.5 25.3 28.2 30.2 26.6 29.3 
    Education, in school until age 19 years or older, % 13.8 15.5 14.5 13.9 13.5 14.1 18.4 11.5 
    Family history of cancer in first-degree relatives, % 28.8 28.0 28.6 30.3 29.2 26.3 28.5 28.6 
 Personality subscale score* 
 Extraversion Neuroticism Psychoticism Lie 
Characteristics ⩽3 ⩾9 ⩽3 ⩾9 ⩽2 ⩾5 ⩽5 ⩾10 
*Each personality subscale (scored on a scale of 0–12) was divided into four categories approximately equal in size, on the basis of the score in the population as a whole. Consequently, different personality subscales have different cutoff scores. 
Men 
    No. of subjects 4103 3249 4505 2682 3356 4717 4391 2609 
    Age, y, mean ± SD 50.8 ± 7.6 51.2 ± 7.6 51.4 ± 7.7 50.6 ± 7.6 52.6 ± 7.7 49.7 ± 7.4 48.3 ± 7.1 54.5 ± 7.4 
    Current smoker, % 56.4 61.8 61.1 57.3 52.7 64.6 64.4 53.4 
    Current drinker, % 71.4 81.1 76.7 76.9 75.7 76.9 79.6 71.4 
    Body mass index ⩾25.0 kg/m2, % 23.0 32.3 29.0 25.4 25.2 28.0 28.4 24.8 
    Education, in school until age 19 years or older, % 13.9 16.7 15.9 14.1 15.8 14.1 15.7 14.4 
    Family history of cancer in first-degree relatives, % 26.1 25.9 25.3 25.1 27.6 25.5 26.7 26.0 
Women 
    No. of subjects 4625 2941 4038 3118 6246 2189 2812 3415 
    Age, y, mean ± SD 50.8 ± 7.4 51.1 ± 7.3 51.6 ± 7.5 50.6 ± 7.4 51.6 ± 7.5 50.0 ± 7.1 47.5 ± 6.6 54.3 ± 6.8 
    Current smoker, % 5.6 10.1 6.3 7.8 5.1 11.0 9.9 4.7 
    Current drinker, % 17.6 31.6 22.9 24.2 20.2 26.5 32.2 15.8 
    Body mass index ⩾25.0 kg/m2, % 25.6 33.7 34.5 25.3 28.2 30.2 26.6 29.3 
    Education, in school until age 19 years or older, % 13.8 15.5 14.5 13.9 13.5 14.1 18.4 11.5 
    Family history of cancer in first-degree relatives, % 28.8 28.0 28.6 30.3 29.2 26.3 28.5 28.6 
Table 2.

Relative risk (RR) of cancer according to personality subscales*

 Category†  
Personality subscale 1 (referent) Ptrend‡ 
*Multivariable RRs are adjusted for sex, age (in years), cigarette smoking (never smoked, smoked in the past, currently smoking 1–19 cigarettes per day, or currently smoking ⩾20 cigarettes per day), alcohol consumption (never drank alcohol, drank in the past, currently drinking ⩽22.7 g of alcohol per day, or currently drinking ⩾22.8 g of alcohol per day), body mass index in kg/m2 (⩽18.4, 18.5–24.9, or ⩾25.0), education (in school until up to age 15 years, until age 16–18 years, or until age 19 years or older), and family history of cancer (presence or absence in first-degree relatives). RR1 denotes the RR with all cases of cancer included in the multivariable analysis; RR2 denotes the RR with cases diagnosed in the first 3 years of follow-up excluded from the analysis. All RRs are given with 95% confidence intervals (CIs) in parentheses. 
†Each personality subscale (scored on a scale of 0–12) was divided into four categories as equal in size as possible, on the basis of scores in the total population; different subscales therefore have different category cutpoints. 
Ptrend values were calculated by treating personality subscales as continuous variables. 
Extraversion ⩽3 4–5 6–8 ⩾9  
    No. of cases of cancer 305 200 282 199  
    Person-years of follow-up 63 551 47 385 59 647 45 227  
    Sex- and age-adjusted RR (95% CI) 1.0 0.8 (0.7 to 1.0) 0.9 (0.8 to 1.1) 0.9 (0.7 to 1.0) .24 
    Multivariable RR1 (95% CI) 1.0 0.8 (0.7 to 1.0) 0.9 (0.8 to 1.1) 0.9 (0.7 to 1.1) .32 
    Multivariable RR2 (95% CI) 1.0 0.8 (0.7 to 1.0) 0.9 (0.7 to 1.1) 0.9 (0.7 to 1.1) .45 
Neuroticism ⩽3 4–5 6–8 ⩾9  
    No. of cases of cancer 285 200 289 212  
    Person-years of follow-up 62 475 46 776 64 449 42 110  
    Sex- and age-adjusted RR (95% CI) 1.0 1.0 (0.8 to 1.2) 1.0 (0.9 to 1.2) 1.2 (1.0 to 1.5) .03 
    Multivariable RR1 (95% CI) 1.0 1.0 (0.8 to 1.2) 1.0 (0.9 to 1.2) 1.2 (1.0 to 1.4) .06 
    Multivariable RR2 (95% CI) 1.0 0.9 (0.7 to 1.1) 1.0 (0.8 to 1.2) 1.1 (0.9 to 1.4) .43 
Psychoticism ⩽2 ⩾5  
    No. of cases of cancer 334 247 156 249  
    Person-years of follow-up 70 034 52 623 42 935 50 218  
    Sex- and age-adjusted RR (95% CI) 1.0 1.0 (0.8 to 1.2) 0.8 (0.6 to 0.9) 1.1 (0.9 to 1.3) .87 
    Multivariable RR1 (95% CI) 1.0 1.0 (0.8 to 1.2) 0.8 (0.6 to 0.9) 1.1 (0.9 to 1.3) .96 
    Multivariable RR2 (95% CI) 1.0 1.0 (0.8 to 1.2) 0.7 (0.6 to 0.9) 1.1 (0.9 to 1.4) .87 
Lie ⩽5 6–7 8–9 ⩾10  
    No. of cases of cancer 214 224 316 232  
    Person-years of follow-up 52 348 54 137 65 311 44 014  
    Sex- and age-adjusted RR (95% CI) 1.0 0.9 (0.7 to 1.1) 0.9 (0.8 to 1.1) 0.8 (0.7 to 1.0) .12 
    Multivariable RR1 (95% CI) 1.0 0.9 (0.7 to 1.1) 0.9 (0.8 to 1.1) 0.9 (0.7 to 1.0) .19 
    Multivariable RR2 (95% CI) 1.0 0.9 (0.7 to 1.1) 0.9 (0.7 to 1.1) 0.9 (0.7 to 1.1) .37 
 Category†  
Personality subscale 1 (referent) Ptrend‡ 
*Multivariable RRs are adjusted for sex, age (in years), cigarette smoking (never smoked, smoked in the past, currently smoking 1–19 cigarettes per day, or currently smoking ⩾20 cigarettes per day), alcohol consumption (never drank alcohol, drank in the past, currently drinking ⩽22.7 g of alcohol per day, or currently drinking ⩾22.8 g of alcohol per day), body mass index in kg/m2 (⩽18.4, 18.5–24.9, or ⩾25.0), education (in school until up to age 15 years, until age 16–18 years, or until age 19 years or older), and family history of cancer (presence or absence in first-degree relatives). RR1 denotes the RR with all cases of cancer included in the multivariable analysis; RR2 denotes the RR with cases diagnosed in the first 3 years of follow-up excluded from the analysis. All RRs are given with 95% confidence intervals (CIs) in parentheses. 
†Each personality subscale (scored on a scale of 0–12) was divided into four categories as equal in size as possible, on the basis of scores in the total population; different subscales therefore have different category cutpoints. 
Ptrend values were calculated by treating personality subscales as continuous variables. 
Extraversion ⩽3 4–5 6–8 ⩾9  
    No. of cases of cancer 305 200 282 199  
    Person-years of follow-up 63 551 47 385 59 647 45 227  
    Sex- and age-adjusted RR (95% CI) 1.0 0.8 (0.7 to 1.0) 0.9 (0.8 to 1.1) 0.9 (0.7 to 1.0) .24 
    Multivariable RR1 (95% CI) 1.0 0.8 (0.7 to 1.0) 0.9 (0.8 to 1.1) 0.9 (0.7 to 1.1) .32 
    Multivariable RR2 (95% CI) 1.0 0.8 (0.7 to 1.0) 0.9 (0.7 to 1.1) 0.9 (0.7 to 1.1) .45 
Neuroticism ⩽3 4–5 6–8 ⩾9  
    No. of cases of cancer 285 200 289 212  
    Person-years of follow-up 62 475 46 776 64 449 42 110  
    Sex- and age-adjusted RR (95% CI) 1.0 1.0 (0.8 to 1.2) 1.0 (0.9 to 1.2) 1.2 (1.0 to 1.5) .03 
    Multivariable RR1 (95% CI) 1.0 1.0 (0.8 to 1.2) 1.0 (0.9 to 1.2) 1.2 (1.0 to 1.4) .06 
    Multivariable RR2 (95% CI) 1.0 0.9 (0.7 to 1.1) 1.0 (0.8 to 1.2) 1.1 (0.9 to 1.4) .43 
Psychoticism ⩽2 ⩾5  
    No. of cases of cancer 334 247 156 249  
    Person-years of follow-up 70 034 52 623 42 935 50 218  
    Sex- and age-adjusted RR (95% CI) 1.0 1.0 (0.8 to 1.2) 0.8 (0.6 to 0.9) 1.1 (0.9 to 1.3) .87 
    Multivariable RR1 (95% CI) 1.0 1.0 (0.8 to 1.2) 0.8 (0.6 to 0.9) 1.1 (0.9 to 1.3) .96 
    Multivariable RR2 (95% CI) 1.0 1.0 (0.8 to 1.2) 0.7 (0.6 to 0.9) 1.1 (0.9 to 1.4) .87 
Lie ⩽5 6–7 8–9 ⩾10  
    No. of cases of cancer 214 224 316 232  
    Person-years of follow-up 52 348 54 137 65 311 44 014  
    Sex- and age-adjusted RR (95% CI) 1.0 0.9 (0.7 to 1.1) 0.9 (0.8 to 1.1) 0.8 (0.7 to 1.0) .12 
    Multivariable RR1 (95% CI) 1.0 0.9 (0.7 to 1.1) 0.9 (0.8 to 1.1) 0.9 (0.7 to 1.0) .19 
    Multivariable RR2 (95% CI) 1.0 0.9 (0.7 to 1.1) 0.9 (0.7 to 1.1) 0.9 (0.7 to 1.1) .37 
Table 3.

Multivariable relative risk (RR) of particular cancers for individuals in the highest as compared with the lowest of four categories of each of four personality subscales*

 Extraversion Neuroticism Psychoticism Lie 
Cancer site RR (95% CI) Ptrend† RR (95% CI) Ptrend RR (95% CI) Ptrend RR (95% CI) Ptrend 
*Cox proportional hazards regression was used to adjust multivariable RRs for sex, age (in years), cigarette smoking (never smoked, smoked in the past, currently smoking 1–19 cigarettes per day, or currently smoking ⩾20 cigarettes per day), alcohol consumption (never drank alcohol, drank in the past, currently drinking ⩽22.7 g of alcohol per day, or currently drinking ⩾22.8 g of alcohol per day), body mass index in kg/m2 (⩽18.4, 18.5–24.9, or ⩾25.0), education (in school until up to age 15 years, until age 16–18 years, or until age 19 years or older), and family history of cancer (presence or absence in first-degree relatives). RR1 denotes the relative risk with all cases of cancer included in the multivariable analysis; RR2 denotes the relative risk with cases diagnosed in the first 3 years of follow-up excluded from the analysis. CI = confidence interval. 
Ptrend values were calculated by treating personality subscales as continuous variables. 
Stomach (n = 229) 
    RR1 1.0 (0.7 to 1.4) .94 1.1 (0.8 to 1.6) .86 1.1 (0.8 to 1.6) .60 0.9 (0.6 to 1.3) .66 
    RR2 0.8 (0.5 to 1.3) .41 1.3 (0.8 to 2.0) .54 1.2 (0.7 to 1.8) .58 1.0 (0.6 to 1.7) .93 
Colorectum (n = 186) 
    RR1 0.8 (0.5 to 1.3) .26 1.2 (0.8 to 1.8) .16 1.3 (0.8 to 1.9) .61 0.8 (0.5 to 1.2) .66 
    RR2 0.9 (0.5 to 1.4) .24 1.2 (0.7 to 2.1) .22 1.0 (0.6 to 1.7) .56 1.0 (0.6 to 1.6) .75 
Lung (n = 108) 
    RR1 0.6 (0.3 to 1.0) .11 1.2 (0.7 to 2.1) .44 0.9 (0.5 to 1.5) .63 1.1 (0.6 to 1.9) .59 
    RR2 0.7 (0.3 to 1.2) .27 1.0 (0.5 to 1.8) .71 1.3 (0.7 to 2.4) .39 1.3 (0.7 to 2.4) .32 
Breast (n = 87) 
    RR1 0.6 (0.3 to 1.1) .046 1.5 (0.9 to 2.7) .11 0.7 (0.4 to 1.5) .17 0.6 (0.3 to 1.1) .056 
    RR2 0.9 (0.4 to 1.7) .42 1.3 (0.6 to 2.5) .75 1.0 (0.4 to 2.2) .75 0.7 (0.3 to 1.6) .14 
Cancer sites associated with smoking (n = 257) 
    RR1 0.7 (0.5 to 1.1) .50 1.2 (0.9 to 1.7) .28 0.8 (0.6 to 1.2) .30 0.9 (0.6 to 1.3) .70 
    RR2 0.8 (0.5 to 1.3) .95 0.9 (0.6 to 1.4) .70 0.9 (0.6 to 1.4) .73 0.8 (0.5 to 1.2) .34 
Cancer sites not associated with smoking (n = 717) 
    RR1 0.9 (0.8 to 1.1) .38 1.2 (1.0 to 1.5) .10 1.2 (1.0 to 1.4) .61 0.8 (0.7 to 1.1) .18 
    RR2 0.9 (0.7 to 1.2) .29 1.2 (0.9 to 1.5) .23 1.2 (0.9 to 1.5) .79 0.9 (0.7 to 1.2) .57 
 Extraversion Neuroticism Psychoticism Lie 
Cancer site RR (95% CI) Ptrend† RR (95% CI) Ptrend RR (95% CI) Ptrend RR (95% CI) Ptrend 
*Cox proportional hazards regression was used to adjust multivariable RRs for sex, age (in years), cigarette smoking (never smoked, smoked in the past, currently smoking 1–19 cigarettes per day, or currently smoking ⩾20 cigarettes per day), alcohol consumption (never drank alcohol, drank in the past, currently drinking ⩽22.7 g of alcohol per day, or currently drinking ⩾22.8 g of alcohol per day), body mass index in kg/m2 (⩽18.4, 18.5–24.9, or ⩾25.0), education (in school until up to age 15 years, until age 16–18 years, or until age 19 years or older), and family history of cancer (presence or absence in first-degree relatives). RR1 denotes the relative risk with all cases of cancer included in the multivariable analysis; RR2 denotes the relative risk with cases diagnosed in the first 3 years of follow-up excluded from the analysis. CI = confidence interval. 
Ptrend values were calculated by treating personality subscales as continuous variables. 
Stomach (n = 229) 
    RR1 1.0 (0.7 to 1.4) .94 1.1 (0.8 to 1.6) .86 1.1 (0.8 to 1.6) .60 0.9 (0.6 to 1.3) .66 
    RR2 0.8 (0.5 to 1.3) .41 1.3 (0.8 to 2.0) .54 1.2 (0.7 to 1.8) .58 1.0 (0.6 to 1.7) .93 
Colorectum (n = 186) 
    RR1 0.8 (0.5 to 1.3) .26 1.2 (0.8 to 1.8) .16 1.3 (0.8 to 1.9) .61 0.8 (0.5 to 1.2) .66 
    RR2 0.9 (0.5 to 1.4) .24 1.2 (0.7 to 2.1) .22 1.0 (0.6 to 1.7) .56 1.0 (0.6 to 1.6) .75 
Lung (n = 108) 
    RR1 0.6 (0.3 to 1.0) .11 1.2 (0.7 to 2.1) .44 0.9 (0.5 to 1.5) .63 1.1 (0.6 to 1.9) .59 
    RR2 0.7 (0.3 to 1.2) .27 1.0 (0.5 to 1.8) .71 1.3 (0.7 to 2.4) .39 1.3 (0.7 to 2.4) .32 
Breast (n = 87) 
    RR1 0.6 (0.3 to 1.1) .046 1.5 (0.9 to 2.7) .11 0.7 (0.4 to 1.5) .17 0.6 (0.3 to 1.1) .056 
    RR2 0.9 (0.4 to 1.7) .42 1.3 (0.6 to 2.5) .75 1.0 (0.4 to 2.2) .75 0.7 (0.3 to 1.6) .14 
Cancer sites associated with smoking (n = 257) 
    RR1 0.7 (0.5 to 1.1) .50 1.2 (0.9 to 1.7) .28 0.8 (0.6 to 1.2) .30 0.9 (0.6 to 1.3) .70 
    RR2 0.8 (0.5 to 1.3) .95 0.9 (0.6 to 1.4) .70 0.9 (0.6 to 1.4) .73 0.8 (0.5 to 1.2) .34 
Cancer sites not associated with smoking (n = 717) 
    RR1 0.9 (0.8 to 1.1) .38 1.2 (1.0 to 1.5) .10 1.2 (1.0 to 1.4) .61 0.8 (0.7 to 1.1) .18 
    RR2 0.9 (0.7 to 1.2) .29 1.2 (0.9 to 1.5) .23 1.2 (0.9 to 1.5) .79 0.9 (0.7 to 1.2) .57 
Fig. 1.

Association between neuroticism and the risk of total cancer, according to study design (retrospective or prospective) and duration of follow-up (in the prospective analyses). Cut points for the neuroticism scores were chosen to divide the population into four groups as similar in size as possible. For all analyses, the referent group was the group in the lowest category of neuroticism score. OR denotes odds ratio estimated from the retrospective analysis using as the endpoint 671 prevalent cases of cancer at the baseline (Ptrend<.001). RR1 denotes relative risk estimated from the prospective analysis using as the endpoint 320 incident cases of cancer diagnosed in the first 3 years of follow-up (Ptrend = .03). RR2 denotes relative risk estimated from the prospective analysis using as the endpoint 666 incident cases of cancer diagnosed in the 4th through the 7th years of follow-up, that is, excluding cancer cases diagnosed in the first 3 years of follow-up (Ptrend = 0.43). All ORs and RRs have been adjusted for sex, age, cigarette smoking (never smoked, smoked in the past, currently smoking 1–19 cigarettes per day, or currently smoking ⩾20 cigarettes per day), alcohol consumption (never drank alcohol, drank in the past, currently drinking ⩽22.7 g of alcohol per day, or currently drinking ⩾22.8 g of alcohol per day), body mass index in kg/m2 (⩽18.4, 18.5–24.9, or ⩾25.0); education (in school until up to age 15 years, age 16–18 years, or age 19 years or older), and family history of cancer (presence or absence in first-degree relatives).

Fig. 1.

Association between neuroticism and the risk of total cancer, according to study design (retrospective or prospective) and duration of follow-up (in the prospective analyses). Cut points for the neuroticism scores were chosen to divide the population into four groups as similar in size as possible. For all analyses, the referent group was the group in the lowest category of neuroticism score. OR denotes odds ratio estimated from the retrospective analysis using as the endpoint 671 prevalent cases of cancer at the baseline (Ptrend<.001). RR1 denotes relative risk estimated from the prospective analysis using as the endpoint 320 incident cases of cancer diagnosed in the first 3 years of follow-up (Ptrend = .03). RR2 denotes relative risk estimated from the prospective analysis using as the endpoint 666 incident cases of cancer diagnosed in the 4th through the 7th years of follow-up, that is, excluding cancer cases diagnosed in the first 3 years of follow-up (Ptrend = 0.43). All ORs and RRs have been adjusted for sex, age, cigarette smoking (never smoked, smoked in the past, currently smoking 1–19 cigarettes per day, or currently smoking ⩾20 cigarettes per day), alcohol consumption (never drank alcohol, drank in the past, currently drinking ⩽22.7 g of alcohol per day, or currently drinking ⩾22.8 g of alcohol per day), body mass index in kg/m2 (⩽18.4, 18.5–24.9, or ⩾25.0); education (in school until up to age 15 years, age 16–18 years, or age 19 years or older), and family history of cancer (presence or absence in first-degree relatives).

We thank Dr. Seiichiro Yamamoto of the Cancer Information and Epidemiology Division, National Cancer Center Research Institute, Tokyo, Japan, for statistical advice.

References

1
Greer S, Morris T. Psychological attributes of women who develop breast cancer: a controlled study.
J Psychosom Res
 
1975
;
19
:
147
–53.
2
Kissen DM, Eysenck HJ. Personality in male lung cancer patients.
J Psychosom Res
 
1962
;
6
:
123
–7.
3
Coppen A, Metcalfe M. Cancer and extraversion.
BMJ
 
1963
;
2
:
18
–9.
4
Morris T, Greer S, Pettingale KW, Watson M. Patterns of expressing anger and their psychological correlates in women with breast cancer.
J Psychosom Res
 
1981
;
25
:
111
–7.
5
Kissen DM, Brown RI, Kissen M. A further report on personality and psychosocial factors in lung cancer.
Ann N Y Acad Sci
 
1969
;
164
:
535
–45.
6
Yamaoka K. Shigehisa T, Ogoshi K, Haruyama K, Watanabe M, Hayashi F, et al. Health-related quality of life varies with personality types: a comparison among cancer patients, non-cancer patients and healthy individuals in a Japanese population.
Qual Life Res
 
1998
;
7
:
535
–44.
7
Bleiker EM, van der Ploeg HM, Hendriks JH, Ader HJ. Personality factors and breast cancer development: a prospective longitudinal study.
J Natl Cancer Inst
 
1996
;
88
:
1478
–82.
8
Everson SA, Goldberg DE, Kaplan GA, Cohen RD, Pukkala E, Tuomilehto J, et al. Hopelessness and risk of mortality and incidence of myocardial infarction and cancer.
Psychosom Med
 
1996
;
58
:
113
–21.
9
Shapiro IR, Ross-Petersen L, Saelan H, Garde K, Olsen JH, Johansen C. Extraversion and neuroticism and the associated risk of cancer: A Danish cohort study.
Am J Epidemiol
 
2001
;
153
:
757
–63.
10
Hahn RC, Petitti DB. Minnesota Multiphasic Personality Inventory-rated depression and the incidence of breast cancer.
Cancer
 
1988
;
61
:
845
–8.
11
Persky VW, Kimpthorne-Rawson J, Shekelle RB. Personality and risk of cancer: 20-year follow-up of the Western Electric Study.
Psychosom Med
 
1987
;
49
:
435
–49.
12
Grossarth-Maticek R, Eysenck HJ, Vetter H. Personality type, smoking habit and their interaction as predictors of cancer and coronary heart disease.
Pers Individ Dif
 
1988
;
9
:
479
–95.
13
Shekelle RB, Raynor WJ, Ostfeld AM, Garron DC, Bieliauskas LA, Liu SC, et al. Psychological depression and 17-year risk of death from cancer.
Psychosom Med
 
1981
;
43
:
117
–25.
14
Almada SJ, Zonderman AB, Shekelle RB, Dyer AR, Daviglus ML, Costa PT Jr, et al. Neuroticism and cynicism and risk of death in middle-aged men: the Western Electric Study.
Psychosom Med
 
1991
;
53
:
165
–75.
15
Hosokawa T, Ohyama M. Reliability and validity of the Japanese version of the short form Eysenck Personality Questionnaire-Revised.
Psychol Rep
 
1993
;
72
:
823
–32.
16
Fukao A, Tsubono Y, Komatsu S, Tsuji I, Minami Y, Hisamichi S, et al. Cohort study on the relation of lifestyle, personality and biologic markers to cancer in Miyagi, Japan: study design, response rate and profiles of the cohort subjects.
J Epidemiol
 
1995
;
5
:
153
–7.
17
Tsubono Y, Yamada S, Nishino Y, Tsuji I, Hisamichi S. Choice of comparison group in assessing the health effects of moderate alcohol consumption.
JAMA
 
2001
;
286
:
1177
–8.
18
Eysenck HJ, Eysenck SB. Manual of the Eysenck Personality Questionnaire (adult and junior). London (UK): Hodder & Stoughton;
1975
.
19
Eysenck HJ, Eysenck SB. Manual of the Eysenck Personality Scales. (EPS Adult). London (UK): Hodder & Stoughton;
1991
.
20
MPI Kenkyukai. Shin Seikaku Kensahou [Manual of the Maudsley Personality Inventory, Japanese version]. Tokyo (Japan): Seishin Shobo;
1969
.
21
Shigehisa T, Ikeda S, Koike S. Rationality-antiemotionality, harmony-seeking and related variables: an analysis of premorbid personality.
J Tokyo Kasei Gakuin Univ
 
1995
;
35
:
421
–36.
22
Kawakami N, Takai A, Takatsuka N, Shimizu H. Eysenck’s personality and tobacco/nicotine dependence in male ever-smokers in Japan.
Addict Behav
 
2000
;
25
:
585
–91.
23
Kitamura T, Kawakami N, Sakamoto S, Tanigawa T, Ono Y, Fujihara S. Quality of life and its correlates in a community population in a Japanese rural area.
Psychiatry Clin Neurosci
 
2002
;
56
:
431
–41.
24
Nakano K. Role of personality characteristics in coping behaviors.
Psychol Rep
 
1992
;
71
:
687
–90.
25
Furukawa T, Shibayama T. Factors influencing adjustment of high school students in an international exchange program.
J Nerv Ment Dis
 
1994
;
182
:
709
–14.
26
Izawa S, Yamaguchi K, Tatsuoka M, Motigi M, Uchiyama T, Ueno K. Manual of the Sixteen Personality Factor Questionnaire (Japanese version). Tokyo (Japan): Nihon Bunka Kagakusha;
1982
.
27
Takano A, Okuno Y. Japan, Miyagi Prefecture. In: Parkin DM, Whelan SL, Ferlay J, Raymond L, Young J, editors. Cancer incidence in five continents. Vol. 7. IARC Sci Publ
1997
;(143):386–9.
28
Rothman KJ, Greenland S. Modern epidemiology. 2nd ed. Philadelphia (PA): Lippincott-Raven;
1998
. p. 359–99.
29
Schoenfeld DA, Richter JR. Nomograms for calculating the number of patients needed for a clinical trial with survival as an endpoint.
Biometrics
 
1982
;
38
:
163
–70.
30
World Cancer Research Fund in association with American Institute for Cancer Research. Food, nutrition and the prevention of cancer: a global perspective. Washington (DC):
1997
. p. 506–7.
31
Temoshok L, Heller B, Sagebiel RW, Blois MS, Sweet DM, DiClemente RJ, et al. The relationship of psychosocial factors to prognostic indicators in cutaneous malignant melanoma.
J Psychosom Res
 
1985
;
29
:
139
–53.
32
Petticrew M, Bell R, Hunter D. Influence of psychological coping on survival and recurrence in people with cancer: systematic review.
BMJ
 
2002
;
325
:
1066
.
33
Grossarth-Maticek R, Bastiaans J, Kanazir DT. Psychosocial factors as strong predictors of mortality from cancer, ischaemic heart disease and stroke: the Yugoslav prospective study.
J Psychosom Res
 
1985
;
29
:
167
–76.
34
Pelosi AJ, Appleby L. Psychological influences on cancer and ischaemic heart disease.
BMJ
 
1992
;
304
:
1295
–8.
35
Amelang M. Tales from Crvenka and Heidelberg: what about the empirical basis?
Psychol Inq
 
1991
;
2
:
233
–6.
36
Derogatis LR. Personality, stress, disease, and bias in epidemiologic research.
Psychol Inq
 
1991
;
2
:
238
–42.
37
Suinn RM. Prediction of cancer from psychological measures.
Psychol Inq
 
1991
;
2
:
268
–76.
38
Kiecolt-Glaser JK, Chee M. Personality, stress, and cancer: a re-examination.
Psychol Inq
 
1991
;
2
:
249
–51.
39
McManus IC, Weeks SJ. Smoking, personality and reasons for smoking.
Psychol Med
 
1982
;
12
:
349
–56.
40
Vollrath M, Torgersen S. Who takes health risks? A probe into eight personality types.
Pers Individ Dif
 
2002
;
32
:
1185
–97.
41
Eysenck HJ. Smoking, personality and psychosomatic disorders.
J Psychosom Res
 
1963
;
7
:
107
–30.