Extroversion and Neuroticism and the Associated ...

American Journal of Epidemiology Copyright © 2001 by The Johns Hopkins University School of Hygiene and Public Health All rights reserved

Vol. 153, No. 8 Printed in U.S.A.

Extroversion, Neuroticism, and Risk of Cancer Schapiro et al.

Extroversion and Neuroticism and the Associated Risk of Cancer: A Danish Cohort Study

Ina Rytter Schapiro,1 Lone Ross-Petersen,1 Henrik Sælan,2,3 Karin Garde,4 Jørgen H. Olsen,1 and Christoffer Johansen1

extraversion (psychology); introversion (psychology); neoplasms; neurotic disorders; personality

In 1962, Kissen and Eysenck (1) reported that 116 patients with lung cancer showed a high degree of extroversion (an outgoing, uninhibited social disposition) and a low degree of neuroticism (general emotional lability, emotional overresponsiveness, and liability to neurotic breakdown under stress) when compared with 123 controls in an ageadjusted analysis. On the basis of these findings, the authors developed the hypothesis of a cancer-prone personality. The results of subsequent case-control studies did not confirm the proposed association between Eysenck’s personality dimensions (2) and the risk of cancer (3–6). A proportion of retrospective investigations based on miscellaneous personality assessments did, however, indicate that personality affects the risk of cancer (7–11). We used a prospective design to investigate the hypothesis of a cancer-prone personality as defined by Kissen and Eysenck (1) and estimated the incidence of cancer in a random sample of 1,052 individuals from the general popula-

tion aged 40 years at the date of entry and characterized according to the Eysenck Personality Inventory (2). The specific hypothesis of this study was that a high degree of extroversion and/or a low degree of neuroticism is associated with an increased risk of cancer at all sites combined. MATERIALS AND METHODS

Originating from the Copenhagen County, Denmark, population of persons born in 1936 (n 9,000), a random sample of 1,198 persons living in four municipalities on the outskirts of Copenhagen, Denmark, in 1976–1977 were invited to participate in an epidemiologic health survey at the Glostrup Population Studies. The survey included a social-psychiatric interview (12). Of these, 1,052 (88 percent), equally distributed by sex, accepted the invitation. In a study by Hollnagel et al. (12), the study population was found to be representative of the age group 40 years in Copenhagen County. We excluded from this study population 21 persons who had foreign citizenship, leaving 1,031 Danish citizens (494 men and 537 women) for analysis. A total of 1,023 persons had completed the answers constituting the extroversion scale. Similarly, a total of 1,024 persons had completed the answers constituting the neuroticism scale. The social-psychiatric interview lasted approximately 1 hour and was performed by a physician. The questionnaire included marital status, socioeconomic status, smoking habits, alcohol consumption, personality, and the presence of symptoms of a psychiatric disease as evaluated by the

Received for publication March 16, 2000, and accepted for publication August 8, 2000. Abbreviations: CI, confidence interval; EPI-Q, Eysenck Personality Inventory, version Q; RR, relative risk. 1 The Danish Cancer Society, Institute of Cancer Epidemiology, Copenhagen, Denmark. 2 Copenhagen County, Centre of Preventive Medicine, Glostrup University Hospital, Glostrup, Denmark. 3 Medical Office of Health, Copenhagen, Denmark. 4 Sct. Hans Hospital, Roskilde, Denmark. Reprint requests to Dr. Christoffer Johansen, Danish Cancer Society, Institute of Cancer Epidemiology, Strandboulevarden 49, DK-2100 Copenhagen, Denmark (e-mail: [email protected]).

757

Downloaded from http://aje.oxfordjournals.org/ by guest on June 11, 2013

The authors have investigated the effect of personality, as measured with the Eysenck Personality Inventory, on the incidence of cancer among 1,031 persons participating in a Danish health survey in 1976–1977 and followed up for 20 years. They thereby accrued a total of 19,993 person-years. The expected number of cancer cases was estimated on the basis of age-, sex-, and site-specific incidence rates in Copenhagen County, Denmark. Overall, 113 malignancies were observed among the cohort members between the date of interview and December 31, 1996. Since 114.3 were expected from county incidence rates, the standardized incidence ratio was 0.99 (95% confidence interval: 0.81, 1.19). No statistically significant deviation of the relative risk from unity was seen for any measure of personality, and no excess risk was seen for any particular type of cancer. A regression model, in which adjustment was made for age, sex, calendar period, alcohol consumption, tobacco smoking, psychiatric illness as rated by the interviewing doctor, marital status, and social class, showed no excess risk of cancer among persons considered to be in medium- or high-risk groups according to the Eysenck Personality Inventory. The authors’ data provide no support for the hypothesis of an association between personality and the risk of cancer. Am J Epidemiol 2001;153:757–63.

758

Schapiro et al.

g/day), psychiatric group (normal, neurotic, deviant/ psychotic, other/unknown), social class (white-collar, bluecollar), and marital status (married, not married). In this analysis, a third categorization of personality scores was chosen according to the hypothesized risk of cancer, with risk being defined as background (low extroversion, high neuroticism or low extroversion and high neuroticism combined), medium (low extroversion and low neuroticism combined; high extroversion and high neuroticism combined), and high (high extroversion, low neuroticism or high extroversion and low neuroticism combined). The number of cancers within each stratum was assumed to be Poisson distributed, with a mean proportional to the follow-up period. The cancer incidence rate was assumed to be constant in each stratum. Because of missing values in each of the personality dimensions, part of the multivariate analysis included only 1,020 persons. RESULTS

The 1,031 members of the study population accrued approximately 19,982 person-years of follow-up, with an average of 19.4 years (range, 1–20.7 years). The male-tofemale ratio was 0.9. The distribution of sociodemographic variables, smoking habits, and alcohol consumption by personality score is shown in table 1. There was a significant difference in social class, tobacco smoking, and psychiatric group between persons who scored high and those who scored low on both personality dimensions. Likewise, marital status and alcohol intake differed significantly between the low and the high groups of the extroversion scale and the neuroticism scale, respectively. Overall, a total of 113 cancers were observed, whereas 114.3 were expected, yielding a standardized incidence ratio of 0.99 (95 percent confidence interval (CI): 0.81, 1.19) (table 2). A stratified analysis by scores on the extroversion and neuroticism scales according to our hypothesis did not reveal a link between the personality dimensions under study and cancer, and no trend in the risk pattern for either scale was observed. Because the distribution of personality scores differed among men and women, the analyses were further stratified on sex, but this did not change the results. Table 3 shows the standardized incidence ratios for sitespecific cancers or groups of cancers that are thought to be influenced by psychosomatic, endocrine, or immunologic mechanisms. We observed no increased risk for any subtype of cancer in any of the analyses. Furthermore, we did not observe a pattern or trend in risk according to our hypothesis. In the multivariate analyses, we adjusted for age, sex, calendar period, marital status, social class, tobacco smoking, alcohol consumption, and psychiatric group, which we thought might confound any association between personality scores and the risk of cancer. The risk of cancer was not significantly increased for any of the categories of personality (table 4). In contrast to our expectation, the smallest increase in risk of cancer compared with the background risk was observed among cohort members assigned to the high-risk group, and the extroverted persons in the highrisk-group even had a decreased risk of developing cancer. Am J Epidemiol

Vol. 153, No. 8, 2001


interviewing doctor (13). Personality dimensions were measured against a short form of the Eysenck Personality Inventory (EPI-Q) (2) consisting of 18 items (14). This abbreviated version, the EPI-Q (appendix 1) allows measurement of two dimensions: degree of extroversion and degree of neuroticism. In our study, subjects were grouped according to their scores on each of the two dimensions, first by a dichotomization of each scale and subsequently by defining four categories of risk whereby the extroversion scores were grouped according to the hypothesized risk of cancer into “background risk” (0–3 points), “low risk” (4–5 points), “medium risk” (6–7 points), and “high risk” (8–9 points). Likewise, the neuroticism scores were grouped as background risk (6–9 points), low risk (4–5 points), medium risk (2–3 points), and high risk (0–1 points). Data on all members of the study population were linked to the Central Population Register for verification of the personal identification number and for information on vital statistics and migration. The Central Population Register was established in Denmark on April 1, 1968, and all Danish residents are assigned a 10-digit personal identification number, which incorporates sex and date of birth and permits accurate linkage of information among registries. Subsequently, the study cohort was linked to the Danish Cancer Registry, which began reporting cancer incidence on a nationwide scale in 1943. Each record includes the personal identification number, date of diagnosis of the tumor, and information on the tumor. Tumors are coded according to a modified Danish version of the International Classification of Diseases, Seventh Revision (15). Members of the cohort were followed up for cancer occurrence from the day of interview to the date of emigration, a cancer diagnosis, death, or December 31, 1996, whichever came first. The observed numbers of cancers were compared with those expected on the basis of Copenhagen County incidence rates, which are divided into groups according to sex and 5year age and calendar periods. Multiplication of the personyears of observation by the incidence rate yields the number of cancers that would be expected had the cohort members experienced the same risk of cancer as that prevailing in the population of Copenhagen County. Tests of significance and 95 percent confidence intervals for the standardized incidence ratio, taken as the ratio of the observed to the expected numbers of cancers, were calculated. The Miettinen exact confidence limits were used when the observed number of endpoints was small; otherwise, an accurate asymptotic approximation was used (16). Because of missing values in each of the personality dimensions, the calculations stratified on extroversion and neuroticism included 1,023 persons and 1,024 persons, respectively. For the regression analysis, multiplicative Poisson regression models were estimated through the use of PROC GENMOD in the SAS 6.12 statistical package (SAS, Inc., Cary, North Carolina). Strata were defined by classifying the observation period by age (0–54, ≥55), sex (male, female), 5-year calendar period (1973–1977, 1978–1982, 1983–1987, 1988–1992, 1993–1996), alcohol (0, 1–6, and ≥7 units/week), smoking habit (nonsmoker, former smoker, 1–14 g/day, ≥15

Extroversion, Neuroticism, and Risk of Cancer 759 TABLE 1. Sociodemographic variables, smoking habits, and alcohol consumption based on total numbers and personality scores (extraversion and neuroticism), respectively, among 1,031 Danish citizens participating in the Glostrup Population Studies, Denmark, 1976 Extraversion†

Neuroticism†

Total

Variable

High

High

Low

Low

%

No.

%

No.

%

No.

%

No.

%

Marital status Married

890

86.3

545

86.2

339

86.7

677

207

80.5

Not married

141

13.7

87

13.8

52

13.3

90

88.3 * 11.7

50

19.5

Social class White-collar

514

49.9

349

162

41.4

401

42.4

517

50.1

283

229

55.2

366

52.3 * 47.7

109

Blue-collar

55.2 * 44.8

148

57.6

332 133

32.2 12.9

184 76

146 56

37.3 14.3

263 97

26.9 13.2

22.7 32.2

154 218

78 111

20.0 28.4

172 235

34.3 12.7 * 22.4 30.6

69 34

234 332

29.1 12.0 * 24.4 34.5

59 95

23.0 37.0

206

20.0

104

101

49.3

154

20.1

51

19.8

398 427

38.6 41.4

230 298

16.5 * 36.4 47.2

164 126

41.6 29.7

295 318

38.5 41.5

99 107

38.5 41.6

901 83

87.4 8.1

573 34

90.7 5.4

322 49

82.4 12.5

714 24

181 59

93.1 3.1

20 27

1.9 2.6

6 19

1.0 3.0

12 8

3.1 2.1

8 6

70.4 23.0 * 4.3 2.3

Tobacco smoking Nonsmokers Former smokers Current smokers (g/day) 1–14 ≥15 Alcohol consumption (units/week) Abstainers 1–6 ≥7 Psychiatric group‡ Normal Neurotic Deviant/psychotic Other/unknown

*

11 21

1.0 2.7

* Intragroup variation reaching significance; p < 0.05. † After dichotomization of the Eysenek Personality Inventory, version Q scale: 0–4 = low, 5–9 = high. ‡ As evaluated by the physician who conducted the social-psychiatric interview.

Only during the calendar-year period 1988–1992 did we observe a significant increase in cancer risk compared with the years 1973–1977 (relative risk (RR) 7.81, 95 percent CI: 1.06, 57.86) (data not shown). We observed an interaction between social class and extroversion when study subjects characterized as white-collar and highly extroverted were compared with those who had scores indicating low extroversion (RR 2.13, 95 percent CI: 1.07, 4.26) (data not shown). Cancer cases identified as blue-collar and highly extroverted were at a significantly decreased risk compared with the low-risk group (RR 0.54, 95 percent CI: 0.32, 0.92). We did not observe any interaction effects between the variables included in the model and the neuroticism score alone or in combination with the extroversion score. DISCUSSION

This study, based on data from a prospective, populationbased cohort study and on EPI-Q, does not support a relation between personality and cancer incidence. The positive association found in previous studies probably reflects methodological problems that will be discussed further. The Am J Epidemiol Vol. 153, No. 8, 2001

negative findings of our study and of investigations of better design (17–20) argue against an effect of personality on the risk of cancer. Our results are therefore consistent with the observation that with improved methods in this area of research, studies of the association between personality and the risk of cancer show no association. By determining the standardized incidence ratios, we were able to compare different scorings on two dimensions of the EPI-Q with respect to the risk of developing cancer. We were not able to confirm the previous findings by Kissen and Eysenck (1) because none of our results differed significantly from unity. It could be argued that a risk analysis that applied population-based rates on a representative sample from the population under study should result in estimates (standardized incidence ratio values) close to unity. However, we did not know the actual distribution of the personality dimensions of the general population. We observed a significant difference in the distribution of socioeconomic status and lifestyle factors between different groups in each of the two personality categories (table 1) and included this information in a multivariate analysis that did not confirm a cancer-prone personality according to the theory proposed


No.

760

Personality score

Observed Expected

Men (n = 494) PY* SIR* at risk

95% CI*

Observed Expected

Women (n = 537) PY at risk

SIR

95% CI

Observed Expected

Both sexes (n = 1,031) PY SIR at risk

95% CI

113

114.30

19,982

0.99

0.81, 1.19

1.85 1.44 1.67 1.89

34 26 39 14

29.49 30.95 34.84 18.13

5,013 5,352 6,140 3,316

1.15 0.84 1.12 0.77

0.80, 0.55, 0.80, 0.42,

1.61 1.23 1.53 1.30

1.66 1.83 1.56 1.70

15 32 40 25

18.96 24.64 39.45 30.43

3,265 4,195 6,873 5,510

0.79 1.30 1.01 0.82

0.44, 0.89, 0.72, 0.53,

1.30 1.83 1.38 1.21

Total cohort

40

45.94

9,484

0.87

0.62, 1.19

73

68.37

10,498

1.07

0.84, 1.34

Extraversion score† Background risk (0–3) Low risk (4–5) Medium risk (6–7) High risk (8–9)

8 8 17 7

8.92 11.17 14.86 10.49

1,851 2,322 3,073 2,137

0.90 1.72 1.14 0.67

0.39, 0.31, 0.67, 0.27,

1.77 1.41 1.83 1.38

26 18 22 7

20.57 19.78 19.98 7.64

3,162 3,030 3,067 1,179

1.26 0.91 1.10 0.92

0.83, 0.54, 0.69, 0.37,

Neuroticism score† Background risk (6–9) Low risk (4–5) Medium risk (2–3) High risk (0–1)

3 12 15 10

6.33 7.73 15.72 15.86

1,321 1,611 3.225 3,267

0.47 1.55 0.95 0.63

0.10, 0.80, 0.53, 0.30,

1.39 2.71 1.57 1.16

12 20 25 15

12.63 16.91 23.73 14.57

1,944 2,584 3,648 2,243

0.95 1.18 1.05 1.03

0.49, 0.72, 0.68, 0.58,

* PY, person-years; SIR, standardized incidence ratio; CI, confidence interval. † Grouped according to the hypothesis (see text).

TABLE 3. Observed numbers of cancers and standardized incidence ratios at selected sites based on total numbers and personality scores (extraversion and neuroticism), respectively, among 1,031 Danish citizens participating in the Glostrup Population Studies, Denmark, 1976–1996

Vol. 153, No. 8, 2001

Observed

SIR*

95% CI*

Observed

SIR

95% CI

Observed

SIR

95% CI

Observed

SIR

95% CI

Observed

SIR

95% CI

27

0.88

0.58, 1.28

16

0.93

0.53, 1.51

11

0.83

0.41, 1.48

20

0.94

0.57, 1.45

7

0.81

0.33, 1.66

38

1.30

0.92, 1.78

23

1.29

0.82, 1.90

15

1.34

0.75, 2.21

30

1.38

0.93, 1.96

8

1.10

0.47, 2.16

16 15 17

1.05 1.03 0.69

0.60, 1.71 0.58, 1.70 0.40, 1.10

11 5 13

1.17 0.55 0.84

0.58, 2.08 0.18, 1.28 0.45, 1.44

5 10 4

0.89 1.89 0.44

0.29, 2.08 0.91, 3.48 0.12, 1.13

11 10 14

0.96 0.91 0.75

0.48, 1.72 0.44, 1.67 0.41, 1.26

5 5 2

1.39 1.47 0.34

0.45, 3.23 0.48, 3.43 0.04, 1.24

* PY, person-years; SIR, standardized incidence ratio; CI, confidence interval. † After dichotomization of Eysenek Personality Inventory, version Q, scale. ‡ Including cancers of the breast, corpus uteri, ovary, and prostate. § Including nonmelanoma skin cancer, non-Hodgkin’s lymphoma, Hodgkin’s disease, leukemia, liver cancer, and cancer of the cervix uteri.


Am J Epidemiol

Low risk (5–9) (PY at risk = 4,921)

High risk (0–4) (PY at risk = 14,921)

Low risk (0–4) (PY at risk = 7,582)

High risk (5–9) (PY at risk = 12,238)

Site

Hormone-related organs‡ Virus and immune-related cancers§ Digestive organs (excluding liver) Respiratory organs Other sites

Neuroticism†

Extraversion†

Total (PY* at risk = 19,982)

Schapiro et al.

TABLE 2. Observed and expected numbers of all cancers and standardized incidence ratios stratified on personality scores (extraversion and neuroticism) among 1,031 Danish citizens participating in the Glostrup Population Studies, Denmark, 1976–1996

Extroversion, Neuroticism, and Risk of Cancer 761 TABLE 4. Risk for cancer among 1,024* Danish citizens, adjusted for age, sex, calendar period, personality score, alcohol consumption, tobacco smoking, psychiatric group, marital status, and social class in the Glostrup Population Studies, Denmark, 1976

Personality measurement†

No. of cancer cases

Personyears at risk

RR‡ adjusted for sex, age, and calendar year

95% CI‡

Multivariateadjusted RR

95% CI

Background risk Low extraversion High neuroticism Low extraversion, high neuroticism

45 27 13

7,581 4,920 2,726

1.00 1.00 1.00

Medium risk Low extraversion, low neuroticism High extraversion, high neuroticism

31 14

4,835 2,174

1.39 1.41

0.73, 2.66 0.66, 3.01

1.63 1.33

0.83, 3.18 0.62, 2.84

High risk High extraversion Low neuroticism High extraversion, low neuroticism

68 85 54

12,237 14,919 10,023

1.00 1.09 1.24

0.68, 1.46 0.71, 1.69 0.67, 2.28

0.96 1.28 1.36

0.65, 1.42 0.79, 2.05 0.72, 2.59

1.00 1.00 1.00

by Kissen and Eysenck (1). The observed interaction effect between social class, extroversion score, and the risk of cancer may be explained by our definition of the social classes. Both groups were heterogeneously defined, e.g., the whitecollar group covered a range from highly educated persons to clerks without any formal education, and the blue-collar group included even unemployed people. In addition, we were not able to confirm this equivocal interaction between social class, personality, and the risk of cancer when we combined the two personality dimensions. In summary, we found no evidence to support the hypothesis that a high degree of extroversion and a low degree of neuroticism increase the risk of cancer. Our finding is in line with those of other studies in which the Eysenck Personality Inventory was used to assess personality (3–5). Morris et al. (21) did, however, find a significant association between low scores on the neuroticism scale and breast cancer risk among 75 women but did not find increased risk associated with scores on the extroversion scale. Our study has important advantages over previous investigations. It was of a prospective design with complete follow-up. The internal comparisons decreased the probability of confounding, and we were able to adjust for wellknown risk factors of cancer. Our study also has limitations. It might be argued that we used an inappropriate tool to measure personality, since only two major dimensions of personality are evaluated in the short form of the Eysenck Personality Inventory (2, 14). That might be considered too broad a measure, especially since each dimension is determined by only nine questions. During the 1970s, when our data were collected, however, this inventory was commonly used to assess personality in this area of research (1, 3–6, 21). Although our cohort was Am J Epidemiol Vol. 153, No. 8, 2001

restricted to 1,031 members, the follow-up period was of a reasonable length. We reviewed previous studies to see whether the methods might have accounted for observations of an association between personality and risk of cancer. In most, the study design may have led to observation bias. Thus, using a common case-control design as well as a case-control design in which patients are assessed before diagnosis of cancer but after they have developed sufficient symptoms to seek medical attention leads to the possibility that the recall of events by cases might be different from that of controls, so-called recall bias (7–11). When interviews are used, the case-control design also involves a risk of interviewer bias, whereby the interviewers’ awareness of clinical signs of diseases might have influenced the soliciting, recording, or interpretation of information from the participants (7, 8, 11). The resulting observation of a difference in personality between cancer patients and controls might be due to such observation bias. Apart from inappropriate study designs, the studies had other common problems. Selection bias may have occurred, since most of the investigations did not provide full information on the criteria for inclusion and/or exclusion for the study population or on nonparticipants (7–9, 22, 23). In one study, the control group was chosen after data inspection (10). Some investigations did not provide information on the establishment and confirmation of diagnoses, thereby calling into question the case or control status of the study participants (8, 22, 23). Conclusions based on small study populations are questionable. The populations used to show an association between personality and cancer ranged from 42 to 200 persons (7–11, 22, 24), but even investigations in which larger


* A total of 1,024 persons filled in the neuroticism scale and 1,023 persons filled in the extraversion scale, respectively. † Arranged by increasing risk according to the hypothesis (see text). ‡ RR, relative risk; CI, confidence interval.

762

Schapiro et al.

ACKNOWLEDGMENTS

Supported by the Psychosocial Research Committee, Danish Cancer Society. The authors thank Andrea Meersohn as well as Svend Larsen and Abdi Hersi from the Centre of Preventive Medicine, Glostrup, for skillful computer assistance and Dr. Ole Raaschou-Nielsen and Lars H. Thomassen for helpful assistance during the data processing.

REFERENCES 1. Kissen DM, Eysenck HJ. Personality in male lung cancer patients. J Psychosom Res 1962;6:123–7. 2. Eysenck HJ, Eysenck SBG. The Eysenck Personality Inventory. London, England: University Press of London, 1965.

3. Greer S, Morris T. Psychological attributes of women who develop breast cancer: a controlled study. J Psychosom Res 1975;19:147–53. 4. Berndt H, Günther H, Rothe G. Eysenck’s personality dimensions (MPI) in breast and lung cancer patients and their relationship to patients’ delay. (In German). Arch Geschwulstforsch 1980;50:359–68. 5. Priestman TJ, Priestman SG, Bradshaw C. Stress and breast cancer. Br J Cancer 1985;51:493–8. 6. Forsén A. Psychosocial aspects of breast cancer. Psychiatr Fenn 1990;21:189–200. 7. Huggan RE. Neuroticism and anxiety among women with cancer. J Psychosom Res 1968;12:215–22. 8. Abse DW, Wilkins MM, van de Castle RL, et al. Personality and behavioral characteristics of lung cancer patients. J Psychosom Res 1974;18:101–13. 9. Kirkcaldy BD, Kobylinska E. Psychological characteristics of breast cancer patients. Psychother Psychosom 1987;48:32–43. 10. Scherg H. Psychosocial factors and disease bias in breast cancer patients. Psychosom Med 1987;49:302–12. 11. Quander-Blaznik J. Personality as a predictor of lung cancer: a replication. Pers Individ Diff 1991;12:125–30. 12. Hollnagel H, Sælan H, Garde K, et al. The concept of representativity in epidemiological surveys. Examples from the population study of 40 years old in the Glostrup area, Denmark. (In Danish). Ugeskr Laeger 1981;143:1781–8. 13. Sælan H, Garde K. Social-psychiatric factors in a group of the Danish population aged 40 years. Description of the population and study of the methods (In Danish). Ugeskr Laeger 1979;141:2409–15. 14. Floderus B. Psychosocial factors in relation to coronary heart disease and associated risk factors. Nord Hyg Tidskr 1974;6 (suppl):7–148. 15. Jensen OM, Storm HH, Jensen H. Cancer registration in Denmark and the study of multiple primary cancers, 1943–1980. Natl Cancer Inst Monogr 1985;68:245–51. 16. Rothman KJ, Boice JD Jr. Epidemiologic analysis with a programable calculator. Washington, DC: US GPO,1979. 17. Hahn RC, Petitti DB. Minnesota Multiphasic Personality Inventory-rated depression and the incidence of breast cancer. Cancer 1988;61:845–8. 18. Edwards JR, Cooper CL, Pearl SG, et al. The relationship between psychosocial factors and breast cancer: some unexpected results. Behav Med 1990;16:5–14. 19. Ragland DR, Brand RJ, Fox BH. Type A/B behavior and cancer mortality: the confounding/mediating effect of covariates. Psychooncology 1992;1:25–33. 20. Bleiker EMA, van der Ploeg HM, Hendriks JHCL, et al. Personality factors and breast cancer development: a prospective longitudinal study. J Natl Cancer Inst 1996;88:1478–82. 21. Morris T, Greer S, Pettingale KW, et al. Patterns of expression of anger and their psychological correlates in women with breast cancer. J Psychosom Res 1981;25:111–17. 22. Dattore PJ, Schontz FC, Coyne L. Premorbid personality differentiation of cancer and non-cancer groups: a test of the hypothesis of cancer proneness. J Consult Clin Psychol 1980; 48:388–94. 23. Grossarth-Maticek R, Eysenck HJ, Vetter H. Personality type, smoking habit and their interaction as predictors of cancer and coronary heart disease. Pers Individ Diff 1988;9:479–95. 24. Hagnell O. The premorbid personality of persons who develop cancer in a total population investigated in 1947 and 1957. Ann N Y Acad Sci 1966;25:846–55.

[Appendix appears on the following page]

Am J Epidemiol

Vol. 153, No. 8, 2001


populations were used did not generally confirm the findings (17–20). In view of the multifactorial origin of cancer, another drawback of a number of studies is that they did not include confounding variables in their analyses or included only one variable, such as age (7, 9, 10, 24). It is therefore doubtful whether the positive association between personality and cancer observed in these studies is actually due to personality and not to the effect of confounders. The concept of personality differs in the earlier studies, as reflected in the use of different tools for personality assessment. It is difficult to conclude whether the questionnaires used were appropriate due to the inappropriate testing of the validity and reliability of a number of personality tests (8, 10, 11, 23). It is also possible that certain questionnaires are inappropriate for cancer patients, although they have been widely used (9, 22). The finding of an effect of personality on the risk of cancer might be due to the use of several personality questionnaires and scales, since positive results are more likely to be found in multiple comparisons. In conclusion, in this prospective cohort study, we found no evidence of an association between extroversion and neuroticism as personality dimensions and the risk of cancer. The positive findings in previous studies are likely to be explained by the design of the studies and the small study populations included. Likewise, these studies did not include potential confounders in the analyses of data. Further studies with the same and similar designs must be encouraged in order to investigate the association between personality and the risk of cancer.

Extroversion, Neuroticism, and Risk of Cancer 763 APPENDIX 1


Am J Epidemiol Vol. 153, No. 8, 2001