Tobacco Documents Online

DR. C.R.E. COGGINS Pergamon 0191.8869(95)00123-9 Pdnted in Great Bdtain. All fights reserved 0191-8869/95 $9.50 + 0.00 SELF-REGULATION AND MORTALITY FROM CANCER, CORONARY HEART DISEASE, AND OTHER CAUSES: A PROSPECTIVE STUDY R. Grossarth-MaticekI and H. J. Eysenck2 1University for Peace, United Nations and 2Department of Psychology, Institute of Psychiatry, University of London, De Crespigny Park, Denmark Hill, London SE5 8A5, England (Received 3 July 1995) SummaryDThis article introduces a new personality inventory dealing with self-regulation. This is in some ways the opposite of neuroticism, and measures personal autonomy or independence, particularly as far as emotional dependence is concerned. Our concern was the relation between self-regulation and health, and large samples of healthy men and women were tested and followed up to demonstrate high predictability of mortality from cancer, coronary heart disease and other causes of death from scores on the questionnaire. It was also demonstrated that psychological risk factors were largely independent from physical risk factors, and could bc changed by behavioural--cognitive treatment, reducing mortality. INTRODUCTION The ancients had a motto for happiness: Mens sana in corpore sano. They believed, following Hippocrates, that the sound mind was related to the sound body, and that there were cancer-prone personalities predisposed to develop this disease more readily, and die of it more quickly, than others not so prone (Mettler & Mettler, 1947; Kowal, 1955; Greer, 1983; Rosch, 1979, 1980). In recent years, many studies have given support to the idea of a cancer-prone personality (Eysenck, 1991, 1994a; Temoshok & Dreher, 1992), as well as a coronary heart disease-prone personality (Friedman, 1991; .~ohnson, 1990; Turner, Sherwood & Light, 1992). The latter is often referred to as Type A, contrasted with the healthy Type B (Eysenck, 1990); the cancer-prone type is sometimes referred to as Type C. Disease-prone types share certain similarities, but can be differentiated successfully both experimentally (Kneier & Temoshok, 1984) and by interview/questionnaire (Eysenck, 1988). While most interest has been directed towards Types A and C, there has also been some interest in the study of the healthy type of person, Type B (Friedman & Rosenman, 1974) or Type 4 (Grossarth-Maticek, Eysenck & Vetter, 1988). This may be defined negatively in terms of the absence of traits characteristic of cancer-prone and CHD-prone personalities, or positively in terms of active health-promoting traits. The difference is of course of little practical importance; a given trait may be formulated positively, or negatively, and scored in the health-giving or disease-prone direction. Table 1 shows some of the concepts related to the disease-prone and the healthy personality (Friedman & Booth-Kewley, 1987), respectively. Obviously these varied conceptions have a great deal in common, and the present study reports an attempt to bring this consensus to a focus, and demonstrate its relevance to actual physical health. What seemed to us to be the defining feature of the healthy personality was autonomy, emotional independence, and self-regulation, i.e. the ability to actively regulate one' s own life, without a degree of emotional dependence on other people that acted in such a way as to thwart one's needs and aims. The concept of 'locus of control' has some similarity to self-regulation, but is rather narrower in its meaning. Both are clearly related to low neuroticism (Eysenck, 1994a). The term 'self-regulation' has been used in the past with similar but somewhat differing meaning (Schwartz, 1983; Leventhal, Nerenz & Strauss, 1980; Carver & Scheier, 1982) as an aspect of control theory. These authors review self-regulation in terms of coping with symptoms or medical treatments 781

Self-regulation and mortality Table 2. Personality and other correlate~ of the six Grossarth-Maticek types (Schmitz. 1992. 1993) Types 1 2 3 4 5 6 N +++ +++ ++ + +++ E .... ++ - + P ~ + =, -- _-- ++ L = = -, = + - Autonomic anxiety + + + =* - + := Cognitive anxiety + + + + + - - + + + State anxiety + + + + -- - - + + Dogmatism + + + + + + + - - + + + Alienation + + + + + +'+ - + + + Can't describe feelings + + + + + + ,,~ - - = + Can't communicate feelings + + + + + - + + + Alexithymia + + = = - + = Task-oriented ,= - = + + + - Emotion-oriented + + + + + + + + - - = + + Coping Avoidance-oriented + + + + + = = = Distraction + + + + + + = = = Social diversion = = + = = = Psychosomatic complaints Physical exhaustion + + + + 4- - - + + Insom~tia + + + - - + + + + + Cardiovascular problems + + + + + - + + + + Depressive tendencies + + + + + + + -- - + + + + + Impulsiveness + + + + + + - - = + + + Abuse of drugs Psychopath. +. + + = - = + Alcohol + + + + = - + + + + + Drugs + + + + ffi -- -- = + Smoking + + + - -- = + + 783 interviewer-administration involving the establishment of trust, and the explanation of obscure or complex questions, has the greatest validity. The second problem is that response may depend on the circumstances leading to the establishment of a given sample; a sample consisting of people coming for psychotherapy is more likely to give truthful answers than a random sample uncertain about the relevance of the questions asked. These problems are closely linked to the hypothesis that a major aspect of the cancer-prone personality, for instance, is the suppression of feelings and emotional responses; such denial may lead to differential responding in different conditions of test Table 3. Personality of other correlates of the six Grossarth-Maticek types (Sandin et al., 1993a, b) Types 1 2 3 4 5 6 Immunological, Cardiovascular ' Respiratory Gastrointestinal Neurology: sensorial Skin Musculoskeletal Genito-urinary N p Alexithymia Coping Task-oriented Emotion-oriented Avoidance-oriented Social support Anger-state Anger-strait Angevin Anger-out Anger-control AnSer.~-x

~ett-regulat~on ant1 mortality Table 2, Personality and other correlates of the six Orossarth-Maticek types (Schmitz, 1992, 1993) Types ! 2 3 4 5 6 N +++ +++ ++ + +++ E .... ++ - + p = + = - __ ++ I. = -- =, = 4- - Autonomic anxiety + + + Cognitive anxiety + + + + + + + + State anxiety + + + + = - - + + Dogmatism + + + + + + + - - + + + Alienation + + + + + + + - + + + Can't describe feelings + + + + + + .... + Can't communicate feelings + + + + + - + + + Alexithymia + + = Task-oriented ffi - ffi + + + - Emotion-oriented + + + + + + + + - - = + + Coping Avoidance-oriented + + + + + Distraction + + + + + + -- Social diversion = = + = = Psychosomatic complaints Physical exhaustion + + + + ÷ - - + + Insomnia + + + - - + + + + + Cardiovascular problems + + + + + - + + + + Depressive tendencies + + + + + + + - - + + + + + Impulsiveness + + + + + + - - = + + + Abuse of drugs Psychopath. + + + = - = + Alcohol + + + + ffi - + + + + + Drugs + + + + .... + Smoking + + + - - = + + 783 interviewer-administration involving the establishment of trust, and the explanation of obscure or complex questions, has the greatest validity. The second problem is that response may depend on the circumstances leading to the establishment of a given sample; a sample consisting of people coming for psychotherapy is more likely to give truthful answers than a random sample uncertain about the relevance of the questions asked. These problems are closely linked to the hypothesis that a major aspect of the cancer-prone personality, for instance, is the suppression of feelings and emotional responses; such denial may lead to differential responding in different conditions of test Table 3. Personality of other correlates of the six Grossarth-Maticek types (Sandin et al., 1993a, b) Types 1 2 3 4 5 6 Immunological + + + - = + Cardiovascular + + + + + - = + Respiratory + + + + =~ - = + Gastrointestinal + + + + - = + Neurology: sensorial + + + + + + + - - + + Skin + + + + = = + Musculoskeletal . + + + + + + - = + . Genito-urinary + + + + - = + N ++ ++ ++ - = + E .... 4- = = p = ++ - ffi = + Alexithymia + + ..... Coping Task-oriented - - - + + = ffi Emotion-oriented + + + - = + + Avoidance-oriented + + + + - = + Social support - - ffi + + ffi = STAXI Anger-state = + =, m = + Anger-strait + + + + + -' = + + Anger-in + + + + + + = + + Anger-out =, + + + + = = + + Anger-control = - - + = -- Anger..ex + + + + + + = ffi +

1~. ~JrossarLil-~la~lcoK an~ t~. J. l~ysencK administration. Intelligence, too, may play a part; complex questions embodying complicated theories may not be easily understood by persons with below-average IQs. If we may take the results of the studies summarized in Tables 2 and 3 as suggesting the nature of the 'healthy' (Type 4) and 'unhealthy' (Types I and 2) personality, we see that the 'healthy personality' is low in psychopathology (neuroticism and psychoticism), extraverted, task-oriented rather than emotion-oriented, and controlled in his anger. We may compare these characteristics with those noted in an early but still valuable study that played a pioneering role in this field (Hinkle & Wolff, 1957). They studied three rather homogeneous groups, composed of over 4000 men and women, looking at their history of major and minor illnesses, as well as their circumstances, personalities, and stresses and stress reactions. Their first finding was that the distribution of illnesses was not Gaussian, but negative binomial, a sort of distribution that occurs in groups when the members of the group have different 'risks' of becoming ill. In other words, people differ in their predisposition to become ill. In addition, those so predisposed showed an increased susceptibility to illness in general; they developed many different types of minor or major illness, not just one or two. (Number of major illnesses correlated 0.40 with number of minor illnesses.) There was a clear correlation between number of illnesses and stress experienced, in terms of objective events like divorces, separations, conflicts with family members, uncongenial living and working arrangements, etc. Further, clusters of illness often occurred during periods of significant stress. Constitutional differences predisposing to disease have not been found to differentiate the 'healthy' from the 'diseased'. The subjectivity of the 'stresses' involved becomes apparent in the conclusion drawn by the authors "that illness often occurs when a person perceives his life situation as peculiarly threatening to him, even though this life situation may not appear to be threatening to an outside observer, and that people who maintain good health in a setting of what are 'objectively' difficult life situations do not usually perceive these situations as difficult." The study closely targeted psychological factors similar to those found in Tables 2 and 3 as related to illness predisposition. "Those people who had the greater number of bodily illnesses, regardless of their nature and regardless of their etiology, were the ones who experienced the greater number of disturbances of mood, thought, and behaviour. For example, not uncommonly, persons were seen with recurrent episodes of anxiety, depression, chronic obsessive and compulsive symptoms, or character disturbances; symptoms of this type, with exacerbations and remissions, might predominate in their illness pattern throughout life. But such people, as a group, also had more bodily illnesses of all types than were found among those who had few or no disturbances of mood, thought, or behaviour. This can be put in other terms by saying that ... there was a parallelism between the occurrence of psychoneuroses and psychoses and the occurrence of bodily illness." (p. 446; italics not in original). 2063633598 THE SELF-REGULATION INVENTORY(SRI) To investigate the hypothetical relationship between personality and illness, a self-regulation inventory wasconstructed using questions based on those that had in past research proved useful in predicting good health or poor health respectively, reversing the scoring for the latter so that a high score indicated good health, a low score poor health. Likert-scale scoring on a six-point scale was used. Scores can vary between 105 and 630. The Cronbach ~ reliability for various groups centred on 0.80. For purposes of presentation scores were grouped into six groups, from 1 (low self-regulation) to 6 (high self-regulation). The six steps are coded in multiples of 105. Thus a score of 1 is obtained when the total point score is between 105 and 209; a score of 2 is obtained when the total point score is between 210 and 314, etc. The number of men and women with each score is given in Table 4. A detailed statistical analysis of the questionnaire will be given in a later publication; here we shall be concerned with the validity of the questionnaire as regards predictive accuracy of mortality. Ss were tested by trained interviewers in 1973, and mortality established in 1988; thus the study reports a 15-year follow-up. Data were collected by 116 trained students in all. Ss were randomly selected on the basis of lists of inhabitants in Heidelberg, Germany, at the time. (Copies of the questionnaire can be obtained from H. J. Eysenck.) Table 4 shows the degree of self-regulation for the men and women who took part in the study.

Selt~regulauon an~l ~0rtality Tabl~ 4. Degre~ of self-regulation and mortality in women and men Group 1 2 3 4 5 6 Total Women 150 316 535 912 502 193 2608 5.7% 12.1% 20.5% 34.9% 19.2% 7.4% Men 154 509 1221 813 308 I03 3108 4.9% 16.3% 39.2% 26.1% 9.9% 3.3% Total 304 825 1756 1725 810 296 5716 5.3% 14.4% 30.7% 30.1% 30.1% 5.1% 785 Table 5. Degree of self-regulation and mortality in women Group Score Score Score Score Score Score 1 2 3 4 5 6 150 316 535 912 502 193 N 5.7% 12.1% 20.5% 34.9% 19.2% 7.4% Cancer 25 43 58 35 15 4 16.6% 13.6% 10.9% 3.8% 2.9% 2.0% CHD 45 60 96 51 14 5 30.0% 18.9% 17.9% 5.5% 2.7% 2.5% Other causes 52 79 147 130 37 7 of death 34.6% 25.0% 27.4% 14.2% 7.3% 3.6% Still alive 28 134 234 696 436 177 18.6% 42.4% 43.7% 76.3% 86.8% 91.7% Total 122 182 301 216 66 18 mortality 8 ! .3% 57.5% 56.2% 23.6% 13.1% 8.2% Average age (1973) 55.7 56.1 57.8 58.3 56.9 58.8 Table 6. Degree of self-regulation and mortality in men Group Score Score Score Score Score Score I 2 3 4 5 6 154 509 1221 813 308 103 N 4.9% 16.3% 39.2% 26.1% " 9.9% 3.3% Cancer 22 63 126 29 8 2 14.2% 12.3% 10.3% 3.5% 2.5% 1.9% 49 121 251 48 10 2 31.15% 23.7% 20.5% 5.9% 3.5% 1.9% 51 128 349 92 15 5 33.1% 25.1% 28.5% 11.3% 4.8% 4.8% 32 197 495 644 275 94 20.7% 38.7% 40.5% 79.2% 89.3% 91.2% 122 312 726 169 33 9 79.2% 61.2% 59.4% 20.7% 10.7% 8.7% CHD O~her causes of death Still alive Total mortality Average age (1973) 57.8 56.5 55.9 57.2 58.9 58.4 It is clear that women are significantly higher on the S-R scale (P < 0.001 by Mann-Whitney U-test). This agrees well with the universal tendency of women to live longer than men. Tables 5 and 6 show, separately for women (Table 5) and men (Table 6) the interaction between degree of self-regulation and mortality from Cancer, CHD, and other causes of death. Also given are number still living and total mortality, g2 values were calculated for total mortality vs still living, cancer vs still living, CHD vs still living, and other causes of death vs still living; all were significant at P < 0001 for the sexes separately. Also given are the average ages of the S-R groups. (Ages ranged from 45 to 68 yr in 1973.) Thus for all causes of death (cancer, CHD, other) there is a very significant correlation between S-R and mortality. Figures 1 and 2 show the results diagramatically. Table 7 shows the relationship between S-R scores and a number of risk factors in a small group of 571 persons where more detailed investigation was possible. Clearly those low on self-regulation have higher blood pressure, suffer more from diabetes, are more overweight and lacking in exercise,

786 R. Grossarth-Maticek and H. J. Eysenck Prospective 1973 - 1988 study: females (N -- 2608) 35 - • Cancer " / / 30- . CHD 25- " Other o 20 -- o 15 -- / / 5 4 3 2 1 High Self regulation Low Fig. 1. Mo~lity and de~ee of self-~gulafion; 2608 women. smoke more, drink more, have more accidents, have a poorer diet, are more often ill, spend more time in hospital, and report more symptoms leading to medical treatment. All these are at high levels of significance, with P < 0.001. Table 8 lists smokers in relation to self-regulation for men only. There are two groups, those still alive, and those who had died. (There were too few women smokers in 1973 to make results meaningful.) Among the former, smoking is positively related to higher degrees in self-regulation. In those who died, smoking was more frequent in those with low self-regulation, and they smoked Pros ~ective 1973 - 1988 study: males (N = 3108) 35 30 "" 25 o 20 O 0 6 High 5 4 .3 2 I Self regulation Low 0 O~ o~ O~ O O • . Fig. 2. Mortality and degree of self-regulation; 3108 men.

S~lf-regulation and mortality 787 Table 7. Self-regulation as related to various physical risk factors Type Type Type T);pe Type Type 1 2 3 4 5 6 304 825 175 172 810 296 Blood pressure 168/93 155/90 144/86 135/75 123/71 121/70 Diabetes 39 68 69 11 2 1 12.6% 8.2% 3.9% 0.6% 0.2% 0.3% Overweight 183 478 80 I" 159 40 13 60.0% 57.9% 45.6% 9.2% 4.9% 4.3% Lack of exercise 194 536 961 201 62 10 63.8% 64.9% 54.7% 11.6% 7.6% 3.3% Number of cigarettes smoked per day Alcohol consumed daily (g) Number of accidents per year treated individ. (1970-1973) Poor nutrition Days ill per year (1970-1973) Days in hospital per year (1970-1973) Needing medicare care over 1 yr Number of symptoms leading to medical treatment (1970-1973) 40.2 35.1 30.6 15.1 11.2 7.7 83.6 80.2 64.9 19.8 I 1.6 I0 84 155 167 90 I0 I 27.5% 18.7% 9.5% 5.2% 1.2% 0.3% 265 557 718 401 89 19 87.1% 67.5% 40.9% 23.2% I0.9% 6.4% 64.7 57.2 31.5 16 18 15 22.8 20.6 10.6 4.3 2.5 1.1 71 125 216 99 36 8 23.3% 15.1% 12.3% 5.7% 4.4% 2.7% 14.3 12.8 11.4 4.7 2.3 1.2 more per day. These results for the relation between smoking and self-regulation may at first seem contradictory, but both are highly significant by X2 (P < 0.001). The results are in good agreement with previous studies (e.g. Friedman, Firman, Petitti, Siegelaub, Ury & Klatsky, 1983; Howard, Curmingham & Rechnitzer, 1985) which demonstrated that personality acts as a moderator of the effects of cigarette smoking on coronary risk, in the sense that smoking was having deleterious effects on heal.th only for people with CHD-prone personality, but not on those with psychologically healthy personalities. Eysenck (1994b) has shown that this effect occurs equally for cancer, and the results in Table 8 are clearly in line with this general rule. Data for alcohol consumption are given in Table 9. Among those alive in 1988, the relation between drinking and degree of self-regulation is reasonably linear, with low S-R scorers drinking less than Table 8. Self-regulation ahd smoking--in live and dead pmbands Type Type Type Type Type Type I 2 3 4 5 6 Total 154 509 1221 813 308 103 3108 Group 1" 32 197 495 644. 275 94 1737 55.9% 9 57 164 303 108 49 690 28.1% 28.9% 33.1% 47.0% 39.3% 52.1% 39.7% 15.3 15.6 14.7 24.6 21.7 22.0 122 312 726 169 33 8 1370 44.1% N Still alive Smokers (n;%) Cigarettes per day Total mortality No longer living Smokers (n;%) Cigarettes per day Total smokers Group 122 312 726 169 33 8 91 224 415 89 7 1 74.5% 71.7% 57.1% 40.8% 21.2% 12.5% 26.9% 25.6% 24.3% 23.9% 21.3% 21.3% 100 281 579 372 115 50 64.9% 55.2% 47.4% 45.7% 37.3% 48.5% *~. (linear) ffi 20.63, d.f. = I, P = 0.0000. ~'~ (linear) = 70.59, d.f. = I, P ffi 0.0000. 1370 44.1% 807 59.0% 1497 48.2;c 0

788 R. Grossarth-Maticck and H. J. Eysenck Table 9. Self-regulation and drinking---in live and dead probands Type Type Type Type Type Type 1 2 3 4 5 6 Total N 154 509 1221 813 308 • 103 3108 Group !* Still alive32 197 495 644 275 94 1737 55.9% Alcohol consumed (n;%) 4 48 51 353 210 50 718 12.5% 24.3% 10.3% 54.8% 76.3% 53.1% 23.1% Daily intake (g) 21.6 23.6 39.8 48.7 42.6 44.6 Group llt No longer living 122 312 726 169 33 8.0 1370 44.1% Alcohol consumed (n;%) 85 197 617 17 6 + I 923 69.6% 63.1% 84.9% 10.0% 18.1% 12.5% 67.3% Daily intake (g) 75.8 79.4 69.6 28.3 24.2 25.3 Total alcohol consumed (n;%) 89 245 668 370 216 51 1639 57.7% 48.1% 54.7% 45.5% 70.1% 49.5% 52.7% 0 O~ O~ C.O 0 *.~ (Iinear)= 216.44, d.f. = 1, P = 0.0000. ~.Ztk-~ = 160.29, d.f. = 1, P = 0.0000. high scorers. For those who died, low S-R scores clearly drank more than high S-R scorers. We again see a paradox, and again this finds an explanation in previous research that showed clearly that the effects of alcohol are dependent on personality factors; Grossarth-Maticek & Eysenck (1991 a) found that alcohol consumption had a negative valence for health if drunk to drown one's sorrows, but not if drunk for pleasure, celebration, etc. This is an interesting feature common to smoking and drinking, showing that leaving out of account psychological factors may lead to serious misinterpretations of epidemiological data concerning the effects of cigarette and alcohol consumption. (The X2 results for our conclusions show P < 0.001 levels.) SELF-REGULATION AND GROSSARTH-MATICEK TYPOLOGY It is of interest to see to what extent the Grossarth-Maticek Typology (Grossarth-Maticek & Eysenck, 1990), with its six types, interacts with the self-regulation typology. It has often been objected that the Grossarth-Maticek methodology of assigning a person to one or other of the six types is faulty because: (1) it uses only a small portion of the available data, (2) it does not correct scores on one type by drawing on information regarding another type. Thus a Type 1 person with a high score on Type 4 might be expected to do better health-wise than a Type 1 person with a low Type 4 score. Profile scoring might be a better method of analysis ~han simply assigning a person to a given type just because he happened to score highest for that type, but from the beginning Grossarth-Maticek has used the simple typology concept, rather like Friedman and Rosenman used the Type A concept because to a medical audience this method of analysis might seem more natural and easier to follow. The fact that this simple typological approach has been very successful (Eysenck, 1991) does not mean that better methods should not be tried; it might be hoped that their use would improve predictive accuracy. A sub-group of 3240 men and women was selected on a random basis and administered the Personality Stress Inventory (Grossarth-Maticek & Eysenck, 1990), in order to cross-validate the two inventories. Table 10 shows the major findings. Results are given separately for bad and for good self-regulation (scores of 1, 2 or 3 vs 4, 5 or 6), subdivided by subjects according to Type (1, 2, 3, 4, 5 or 6). For each of the 12 sub-divisions (2 × 6) are given the number and percentage of deaths from cancer, CHD (infarct) and other causes. Clearly SR is vitally important, as the percentage of mortality figures for the High and low S-R scores show. This of course merely mirrors the data in Figs 1 and 2. Within the low S-R group, clearly Type 1 has the highest cancer mortality, Type 4 the least, while for Type 2 CHD has the highest mortality, with all the other types roughly on a par. For

Self-regulation and mortality Table 10. Degree of stir-regulation and six Grossarth-Maticek types as related to mortality Type Type Type Type Type Type 1 2 3 4 5 6 Total N 392 Cancer 117 29.8% Infarct 51 13.0% Other causes of death 101 25.7% Average age (yr) 57.6 Mean S-R score 2.4 N 26O Cancer 4 1.5% lnfaret 4 1.5% Other causes of death 21 8.0% Average age (yr) 56.2 Mean S-R score 3.8 N 652 Cancer 121 18.5% Infarct 55 8%4% Other causes 122 of death 18.7% Poor Self-regulation (1, 2 or 3 points) 403 102 52 507 64 1520 50 17 10 81 12 287 12.4% 16.6% 19.2% 15.9% 18.7% 18.8% 119 19 10 69 10 278 29.5% 18.6% 19.2% 13.6% 15.8% 18.3% 99 24 13 105 17 359 24.5% 23.5% 25.0% 20.7% 26.5% 23.6% 57.4 57.3 58.2 58.4 58.1 2.3 2.5 3.0 2.1 2.4 Good Self-regulatlon (4, 5 or 6 points) 204 351 477 358 70 1720 4 3 2 3 + I 17 1.9% 0.8% 0.4% 0.8% 1.4% 1.0% 5 7 2 2 1 21 2.4% 1.9% 0.4% 0.5% 1.4% 1,2% 27 29 34 38 15 164 13.2% 8.2% 7.2% 10.6% 21.4% 9.5% 56.9 57.1 56.2 56.4 55.7 3.9 4.1 4.7 3.8 3.9 Total Degree of Self-regulation 607 453 529 865 134 3240 54 20 12 84 13 304 8.9% 4.4% 2.2% 9.7% 9.7% 9.4% 204 26 12 71 11 379 33.6% 5.7% 2.3% 8.2% 8.2% 11.7% 126 53 47 143 32 523 20.8% 11.7% 8.9% 16.5% 23.9% 16.1% 789 'Other causes', there is little to choose between Types. For the good S-R scores, Type 4 does best overall, but the other Types have mortality too low to produce marked differences. It is interesting to look at the ratios of good/bad SRI scores for each of the typologies. Going from 1 to 6, these are: 0.66; 0.51; 3.34; 9.17; 0.71; 1.09. Not unexpectedly, the 'healthy' Type 4 has much the highest ratio, followed by the fairly healthy Type 3; while the cancer-prone and CHD-prone Types 1 and 2 have much the lowest. It is apparent that the SRI measures much the same traits as does the Grossarth-Maticek Typology Type 4. Analyses by generalized linear model shows the main effects (Typology and Self-regulation) as well as their interaction are all significant at the P < 0.001 level. One important consequence of these findings would seem to be that questionnaires using a positive wording are as useful, if not better, at indicating psychological disposition to good health, as questionnaires using a negative wording are in indicating psychological disposition to bad health. Most people are apparently more likely to respond truthfully to positive than to negative questions, although this point would have to be established by a specially designed experiment. Physical risk factors for disease To study the relationship of physical risk factors to mortality, a score was based on a specially designed questionnaire, based on known risk factors which could be obtained relatively easily. Table 11 gives the items involved and the points given for the various items. The scale has a minimum of 0 points (no positive factors, high risk), and a maximum of 24 points (many positive factors, low risk). The scale takes into account genetic factors, exercise, nutrition, alcohol, smoking and direct estimates of poor fitness---overweight, high blood pressure, high cholesterol, etc. Different numbers of points can be obtained for different items, thus blood pressure is more important than smoking or drinking. The various items were of course specified in considerable detail for the interviewers. Table 12 shows the relationship between physical risk factor scores and (a) mortality and (b) SRI Scores. There is clearly a close relation between physical risk factors and mortality; the greater the number of positive factors, the greater the chance of survival, and the lower the risk of mortality. Conversely,

790 R. Grossarth-Maticek and H.. J. Eysenck Table I I. Point scale for physical risk factors Points 1. A close member of fl~e family (parents, grandparents) has reached an age of 75 yr. Add one point for each such family member. Points 0-6, respectively 0-6 2. Regular exercise 2 3. Dally activity in fresh air, irrespective of the weather 1 4. Healthy nourishment 2 5. Sufficient amount of fluid intake 1 6. Normal body weight 1 7. Little alcohol 1 8. Non-smoker I 9. Normal blood pressure 2 10. Normal blood sugar 2 l I. Normal total cholesterol 2 12. Low consumption of coffee, black tea and Coca-Cola® 1 13. No stimulant or depressant psychopharmaca. I 14. Normal sensitivity for pain (not overly sensitive) 1 the smaller the number of positive factors, the greater the risk of mortality. Those with the most positive factors, i.e. 24 points, show a survival rate seven times greater than those with a score of 0 points. For those who died, probands with a score of 0 died 15 times more frequently than those with a score of 24. The relationship is significant by Mann-Whitney U-test, with P<0.00001. SRI also independently predicted mortality, with P < 0.00001 by Mann-Whitney U-test. The regression of S-R on physical risk factors appears linear for the dead group but curvilinear for the living; only a replication can show whether this is an accidental finding, of no importance. But clearly the S-R scale measures causes of death largely independent of physical causes. It will be obvious from Table 12 that physical risk factors, as expected, correlate separately with mortality, r (bis) between total mortality and physical risk factors is 0.36, P< 0.001. Using a Kruskal-Wallis ANOVA by ranks for the relationship between S-R and mortality, we obtain H= 3520.83, which with d.f. = 2 gives P < 0.00001. Carrying out the same type of analysis for physical risk factors, H = 1118.26, which with d.f. = 2 gives P < 0.00001. There is little correlation Table 12. Mortality as related to physical risk factors and self-regulation Still living Mortality Positive physical S-R S-R factors n % (%) n % (%) 0 20 0.6 4.8 315 13.9 3.1 1 21 0.6 4.9 206 9.1 3.0 2 34 1.0 4.7 170 7.5 2.8 3 47 1.4 4.6 ~ 153 6.7 3.9 4 96 2.8 4.3 104 4.6 3.1 5 78 2.3 3.9 107 4.7 3.3 6 103 3.0 3.6 I01 4.4 3.3 7 124 3.6 3.7 162 7.1 3.4 8 113 3.3 3.5 103 4.5 3.5 9 272 7.9 3.8 102 4.4 3.3 10 271 7.9 3.6 100 4.4 3.4 11 294 8.5 3.7 84 3.7 3.4 12 231 6.7 3.8 75 3.3 3.3 13 186 5.4 3.6 62 2.7 3.5 14 144 4.2 3.9 51 2.2 3.6 15 169 4.9 3.7 45 1.9 3.4 16 124 3.6 3.6 37 1.6 3.3 17 116 3.4 3.8 40 1.6 3.1 18 127 3.7 3.9 35 1.5 2.9 19 131 3.8 3.4 49 2.2 2.7 20 155 4.5 3.9 51 2.2 2.5 21 163 4.7 4.0 42 1.8 2.6 22 143 4.2 4.1 31 1.4 2.1 23 136 4.0 4.5 28 1.2 2.3 24 144 4.1 4.9 21 0.9 2.4 Total 3422 2274 0 0