Introduction. P Hakala 1 *, A Rissanen 2, M Koskenvuo 3, J Kaprio 4 and T RoÈnnemaa 3

Transcription

1 International Journal of Obesity (1999) 23, 746±753 ß 1999 Stockton Press All rights reserved 0307±0565/99 $ Environmental factors in the development of obesity in identical twins P Hakala 1 *, A Rissanen 2, M Koskenvuo 3, J Kaprio 4 and T RoÈnnemaa 3 1 Research and Development Centre, Social Insurance Institution, Turku, Finland; 2 Department of Psychiatry, University of Helsinki, Finland; 3 Departments of Public Health and Medicine, University of Turku, Finland; and 4 Department of Public Health, University of Helsinki, Finland OBJECTIVE: To study environmental factors promoting obesity when genetic factors are identical. DESIGN: Monozygotic (MZ) twins discordant for overweight were examined during a 3-day stay in an inpatient setting. SUBJECTS: The subjects were selected from the Finnish Twin Cohort. The study sample consisted of 23 healthy adult MZ twin pairs (14 female, nine male) with a difference of at least 3 kg=m 2 in BMI. The mean BMI was 29.5 kg=m 2 for the overweight twins and 22.9 kg=m 2 for their lean co-twins. MEASUREMENTS: Interviews and standardized questionnaires were used to obtain information about energy and nutrient intake, eating behaviour, physical activity history, smoking and other background factors. RESULTS: The overweight co-twins had higher disinhibition scores (P ˆ 0.007) and hunger scores (P ˆ 0.005) in the 3-Factor Eating Questionnaire than their lean co-twins. Among women the mean daily energy intake was higher in the overweight twins than in their lean co-twins (8.8 vs 7.4 MJ; P ˆ 0.045). In the twins discordant for smoking the mean BMI was higher in non-smokers than in smokers (BMI kg=m 2 vs kg=m 2 ; P ˆ 0.031). CONCLUSIONS: Dif culty in controlling eating in both sexes and high energy intake in women were related to overweight, independent of genetic background. Smoking explained the BMI difference among smoking discordant pairs. It is probable that individual twin pairs had different reasons behind the variation in weight gain resulting in non-signi cant intrapair differences in single obesity-promoting factors. Difference in living conditions (e.g. familyand work-related factors) may have promoted different living habits, especially eating behaviour, and may have led to different weight gain in identical twins. Keywords: identical twins; overweight; slimness; energy intake; eating behavior; physical activity history; sociodemographic factors Introduction It is well documented that obesity is a familial disorder with a strong genetic component. The heritability of body fat mass derived from underwaterweighing measurement of body density was observed to reach about 25% of the age- and gender adjusted variance. 1,2 Recent twin studies have shown that the amount of abdominal fat is in uenced by a genetic component, which attained 50 ± 60% of the variance. 3 However, it is dif cult to distinguish the genetic effects from the effects of a family lifestyle. Parents and other family members share the same environment, which may promote overeating or an inactive lifestyle. However, several twin studies indicate that family environment does not play a major role in the aetiology of obesity. Comparisons between twins reared together and twins reared apart have suggested that a shared family environment contributes only *Correspondence: Dr Paula Hakala, Research and Development Centre, Social Insurance Institution, Peltolantie 3, FIN Turku, Finland paula.hakala@kela.memonet. Received 7 October 1998; revised 2 February 1999; accepted 1 March 1999 little to similarity in body mass index (BMI). 4 Similarly, in the twin study of Stunkard et al 5 genetic variability between individuals accounted for 70% of the interindividual differences in BMI later in life, whereas childhood environment had little or no in uence. Adoption studies have also shown that the weight and lifestyle of adoptive parents has only a limited in uence on the weight of their adopted children. 6,7 Mechanisms by which genes confer susceptibility to weight gain are not well known. According to one theory, genetically predisposed individuals are particularly susceptible to the effects of fatty foods and to adoption of a lifestyle which promotes development of obesity. 8 On the other hand, the rapidly increasing prevalence of obesity in all westernized societies is unlikely to be due to changes in the gene pool, but rather to changes in society and the environment. In addition, because twins in adult monozygotic (MZ) twin pairs do not always share the same weight, it is obvious that environmental factors, taken in the broadest sense, must have an in uence. An ef cient way to determine the impact of environmental factors on the onset of obesity without confounding by genetic factors is to examine MZ twins discordant for obesity. Thus, the aim of the present study was to nd environmental factors pro-

2 moting overweight when genetic factors are controlled. Therefore, we examined a unique sample of 23 MZ twin pairs discordant for BMI with special emphasis on variables related to diet, physical activity and sociodemographic factors. Subjects and methods Subjects The subjects were selected from the older part of the Finnish Twin Cohort, which consists of like sexed twin pairs born before The structure and format of the Twin Cohort has been described in detail elsewhere 9 The BMIs were calculated using the selfreported height and weight values obtained from postal questionnaires mailed in 1975, 1981 and ,11 The height and weight was measured in a sample of twins who had self-reported height and weight in The correlation coef cient of BMI was 0.89 for men and 0.90 for women. 12 For this study we included only MZ twin pairs in which the overweight twin was consistently heavier than the co-twin in all three questionnaires over the 15-year period. Initially, the criterion for inclusion in the present study was a difference of at least 4 kg=m 2 in BMI between the co-twins in the 1990 survey. In addition, the BMI of the heavier twin had to be > 27 kg=m 2 and that of the lean co-twin < 25 kg=m 2. Subjects with chronic diseases (diabetes, thyroid disorders and other metabolic diseases or psychiatric diseases) were excluded from the study. All MZ twin pairs ful lling the inclusion criteria (n ˆ 56) were invited to take part in the present study. Twenty-six twin pairs (46% of all eligible pairs) volunteered to participate. The zygosity of the twin pairs was originally estimated on the bases of selfadministered questionnaire that was validated earlier by examination of 11 blood group markers. 13 To con rm the monozygosity of the pairs an expert did dermatoglyphic analysis of ngertip prints. 14,15 All but six pairs were con rmed to be monozygotic. Samples of DNA from these six pairs were typed for markers at six different polymorphic gene logi (DIS 80, APOB, DI7S30, COL2AI, VWA, and HUMTH). Four of the six pairs were found to be monozygotic; the other two were dizygotic and were excluded. Physical examinations showed that ve twin pairs did not meet the original BMI criteria; two pairs with a BMI difference < 3kg=m 2 were rejected but the other three pairs with a BMI difference between 3 and 4kg=m 2 were included in the study sample despite the initially more stringent criterion for BMI difference. In one pair, the overweight twin was found to have previously undiagnosed overt diabetes mellitus, so this pair was also excluded from the study. Thus, the nal study sample consisted of 23 healthy MZ twin pairs (14 female, 9 male) with an intrapair BMI difference of at least 3 kg=m 2. The mean age of the 46 twins was 45 y (range 35 ± 60 y). The twins had lived apart for an average of 25 y (range 1 ± 46 y). One of the twin pairs lived in the same household at the time of the study. The ethics committee of the Social Insurance Institution approved the study protocol, and all subjects gave their informed consent. Methods All measurements and interviews using standardized questionnaires were carried out during a three-day stay at the Research and Development Centre of the Social Insurance Institution (SII) in Turku in 1991 ± Anthropometric data An experienced technician using standardized methods determined the anthropometric measures. The BMI was calculated as weight (kg) divided by the square of height (m 2 ). The waist=hip ratio was calculated as the circumference in the middle between the lowest rib and the iliac crest at the midaxillary line divided by the circumference at the level of the major trochanters. The proportion of fat tissue was measured by a four-component method, which is validated. 16,17 This method is based on the division of body mass into four components, each with a different density: fat tissue (density g=cm 3 ), water (density g=cm 3 ), minerals (density g=cm 3 ) and proteins (density 1.34 g=cm 3 ). Body water mass was estimated by using the bioelectrical impedance method (BIA-101A=S, RJL System Inc., Clemens, MI). 18 Mineral mass was estimated by using dual-energy X-ray absorptiometry (Norland XR26, Norland Corp., Fort Atkinson, WS). The density of the whole body was estimated by underwater weighing and was corrected for information on body water and mineral mass. The proportion of fat tissue was calculated from the density of the whole body according to the formula of Siri. 19 Basal metabolic rate The basal metabolic rate (BMR) was measured by indirect calorimetry (DELTA-TRACK, Instrumentarium, Finland) at 7:30 a.m. before the subjects got out of bed. The mid-day and the evening meals were standardized during the previous day in the following way: The macronutrient composition of the meals was 50 energy% (E%) carbohydrates, 30 E% fat and 20 E% protein. The energy content of the meals was adjusted according to the body weight of each subject, using intervals of 10 kg. The requirement of total energy per day for each subject was calculated (30 kcal=kg=24 h). The proportion of mid-day meal was 30% and the proportion of evening meal was 15% from the total requirement of energy. The purpose of this standardisation was to ensure that diet-induced 747

3 748 thermogenesis should be similar to that in the normal life of each subject. Nutrient intake Food consumption data were collected by a careful one-hour retrospective dietary history interview, which dealt with the subject's normal diet during the past year. Quantities were assessed by means of food models, cups and household measures. One experienced dietitian conducted all interviews. A nutrient data bank and a computer system (Nutrica) designed for nutrition research at the SII was used for calculations of total energy, micronutrients and dietary bre. 20 A study comparing calculated values, based on the Nutrica data base, with analysed values revealed reasonably good agreement between the results concerning energy, micronutrients and dietary bre (r ˆ 0.60 ± 0.75). 21 In 1975 and 1981 the intensity of activity was assessed by three questions. The rst question estimated the level of physical activity compared with walking and running (score 0 ± 3). The second question ranked sweating (score 0 ± 3) and the third question breathlessness (score 0 ± 3) during exercise. Summing up the scores (range 0 ± 9) formed the intensity score. The duration of each session was strati ed in ve categories. Another question asked the number of such sessions per month strati ed in ve categories. The frequency and duration of sessions were multiplied to obtain the time in minutes spent in physical activity per month. The intensity and duration (expressed as hours per week) of physical activity were multiplied to obtain the total score of physical activity during leisure time (range 0 ± 90) according to Kaprio et al. 24 Eating behaviour Two standardised psychometric questionnaires were used to evaluate the subjects' eating behaviour. The Three-Factor Eating Questionnaire (3-FEQ) including 51 items 22 yields three dimensions of eating behaviour (cognitive restraint of eating, disinhibition of control and susceptibility to hunger). The Eating Attitudes Test (EAT) including 40 items 23 reveals symptoms of eating disorders. Scores higher than 30 are suggestive of anorexia nervosa. Physical activity Physical activity during leisure time was assessed during the present study in In addition, physical activity data assessed in 1975 and 1981 were available for this study. In 1992, physical activity history was evaluated by a retrospective interview. The subjects were asked to recollect the frequency, duration and intensity of various types of physical exercises in 5-y periods after 15 years of age. A physical activity index for regular physical exercising was based on three questions. The rst question asked whether the time used for exercising lasted for less than 15 min=week (score zero), 15 ± 60 min=week (score 1) or more than 60 min=week (score 2). The second and third questions asked whether the exercising caused sweating (positive answer scored 1) and breathlessness (positive answer scored 1). These intensity scores were summed up and multiplied by the time score. The index for exercising in connection with other leisure activities was based on two questions. Time was asked as in the case of regular physical exercise. Intensity of exercise was scored as light (0), moderate (1) or heavy (2). Time and intensity scores were multiplied. Only the answers concerning two latest 5-y periods were taken into consideration. The mean value of the scores of these two periods formed the nal indices (range 0 ± 4). Background information Socioeconomic characteristics and smoking were assessed by a standardized interview. It included questions concerning the level of basic and vocational education (classi ed as primary, secondary or advanced level), marital status, parity (for women) and working situation (employment status, basis of salary, work rhythm including day, shift and night work, irregular work, number of unemployment months during lifetime and self-assessed working capacity). Smoking was evaluated according to the number of smoking years, number of cigarettes smoked daily and cessation of smoking. The present height and weight of spouses of the twins were obtained by a questionnaire. Data analysis Statistical analyses were carried out using SAS statistical software. Intrapair differences between the overweight and lean co-twins were tested with the paired t-test (for continuous variables) or McNemar test (for categorical variables). Chi-square test was used for the comparison of the means between overweight and lean co-twins. In addition, Mean intrapair differences were compared in BMI, body weight and body fatness between concordant and discordant twins for smoking, marital status, educational level, working situation and physical activity (Wilcoxon 2-Sample Test). Student's t-test was used to compare the differences in BMI, weight, and body fatness within twins discordant for smoking, marital status, other socioeconomic factors and physical activity. Furthermore, the intrapair difference in intake of energy and macronutrients was correlated against the intrapair difference in BMI and body fatness. P values less than 0.05 were considered statistically signi cant.

4 Results Anthropometric data and BMR The selection criteria of the twin pairs discordant for obesity resulted in statistically signi cant intrapair differences in anthropometric data. The BMR=day was higher and the BMR=weight was lower in the overweight that in the lean twins. The BMR=LBM (lean body mass) in the overweight twins was not different from that in the lean twins, as was expected. There were no intrapair differences in the BMIs of the spouses of the overweight and lean twins (Table 1). Socioeconomic factors and smoking There were no statistically signi cant differences between the overweight and lean twins in the background variables, including socioeconomic status, education, marital status, working situation and smoking (Table 2). No differences were found between the overweight and non-obese female twins in parity ( Table 1 Anthropometric data and basal metabolic rate (BMR) in overweight and lean twins and the BMI of their spouses 1. The results are presented for men and women together, because there were no differences between the sexes Overweight Lean Intrapair (n ˆ 23) (n ˆ 23) difference P-value 2 BMI, kg=m < (26 ± 38) (18 ± 25) Weight, kg < (72 ± 109) (47 ± 94) Percentage of body fat, % < (20.3 ± 51.6) (12.8 ± 36.2) Waist=hip ratio < (0.68 ± 1.04) (0.71 ± 0.99) BMR, MJ=day < (5.7 ± 10.1) (5.1 ± 9.3) BMR=weight, kj=kg < (72 ± 103) (80 ± 124) BMR=LBM, kj=kg ( ) (99 174) BMI of the spouse, kg=m (20 ± 32) (20 ± 30) 1 mean SD; range in parentheses 2 paired t-test 3 BMR=weight ˆ basal metabolic rate (kj) divided by body weight (kg) 4 BMR=LBM ˆ basal metabolic rate (kj) divided by lean body mass (kg) 5 n ˆ 22; 6 n ˆ 19; 7 n ˆ 14. Table 2 Sociodemographic variables, smoking and physical activity indices in overweight and lean twins Female Male Overweigt Lean Overweight Lean (n ˆ14) (n ˆ14) P-value 1 (n ˆ 9) (n ˆ 9) P-value 1 Marital status married, % divorced, % ± single, % 7 21 ± ± Vocational education more than vocational school, % Unemployed, % Smoking nonsmoker, % previous smoker, % ± 0.34 current smoker, % Physical activity indices 2 regular exercising during previous 6 ± 10 yrs assessed in exercising in connection with other leisure activities during previous 6 ± 10 yrs assesses in total physical activity score (mean of yrs 1975 and 1981) paired t-test, McNemar test or Chi-square test 2 mean SD; 3 n ˆ 12; 4 n ˆ 8; 5 n ˆ 13; 6 n ˆ 12

5 750 compared with 1.9) or in menopausal status (33% vs 27% postmenopausal women). There were also no statistically signi cant intrapair differences in adiposity between pairs concordant and discordant for socioeconomic factors and within pairs discordant for socioeconomic factors. Sixteen twin pairs were concordant and seven pairs discordant for smoking. Among the discordant pairs, in six pairs (three males, three females) the lean twin was a smoker and the overweight co-twin was nonsmoker and in only one pair (female) the situation was opposite. In the twins discordant for smoking the mean BMI and body weight were clearly higher in non-smokers than in smokers (BMI kg=m 2 vs kg=m 2 ; P ˆ and weight 82.7 kg 14.3 vs kg; P ˆ 0.031). The corresponding difference in body fatness was not statistically signi cant ( % vs %; P ˆ 0.31). Physical activity The index for regular exercising (assessed in 1992) and the total score of physical activity (mean of years 1975 and 1981) tended to be slightly but not statistically signi cantly lower in the overweight, compared with the lean twins. This tendency was greater in female than in male twin pairs (Table 2). Statistically signi cant intrapair differences in adiposity were not found between pairs concordant and discordant for physical activity or within pairs discordant for physical activity. In the seven female twin pairs discordant for physical activity indices (regular exercising assessed in 1992) the adiposity tended, however, to be higher among physically inactive than among active (for example BMI kg=m 2 vs kg=m 2 ; P ˆ 0.30 and body fatness % vs ; P ˆ 0.09). A similar trend was shown in the female twins discordant for physical activity measured in 1975 and 1981 (for example BMI kg=m 2 vs kg=m 2 ; P ˆ 0.25 and body fatness % vs %; P ˆ 0.29). than in their lean co-twins (Table 3). The energy intake of overweight males tended to be lower than that of lean co-twins, but this difference was not statistically signi cant. The percentages of energy derived from fat, protein, carbohydrates and alcohol and the intake of dietary bre in the overweight twins were not different from those of the lean twins. Correlation analysis concerning differences in intake of energy or macronutrients and differences in BMI or body fatness did not give any valuable additional information. Eating behaviour The scores of the 3-FEQ were for disinhibition and hunger statistically signi cantly higher in the overweight twins than in their lean co-twins (Fig. 1). There was a clear tendency among the lean twins to avoidance of weight gain. Seventy percent of the non-obese twins agreed with a statement `I consciously hold back at meals in order not to gain weight', while the corresponding percentage among the overweight Nutrient intake The mean daily intake of total energy during the preceding year was higher in the overweight females Figure 1 The mean scores of the 3-Factor Eating Questionnaire in overweight and lean twins. The results are presented for men and women together, because there were no differences between the sexes. Table 3 Intake of total daily energy, energy yielding nutrients and dietary bre, according to interviewed dietary history over one year, and energy intake=bmr ratio in female and male overweight and lean twins 1 Female Male Overweight Lean Intrapair Overweight lean Intrapair (n ˆ14) (n ˆ14) difference P-value 2 (n ˆ 9) (n ˆ 9) difference P-value 2 Energy, MJ Fat (E%) Protein (E%) Carbohydrates (E%) Alcohol (E%) Dietary bre, g Energy intake=bmr mean SD; 2 paired t-test

6 Table 4 Positive answers to the question: `Do you have feelings which make you eat or drink although you are not hungry?' among the overweight and lean twins 1. The results are presented for men and women together, because there were no differences between the sexes Overweight Lean Emotional factors (n ˆ 23) (n ˆ 23) P-value 2 Unsatisfactory human relations Stress Irascibility, ill temper Displeasure (failure etc.) Strain Feeling of loneliness Good mood (success) percentage; 2 McNemar test twins was 30% (P ˆ 0.005). Overweight twins also had a tendency to emotional eating and binge eating, as shown by the responses to statements like `When I feel anxious, I nd myself eating' (agreed by 71% of the overweight compared with 29% of the non-obese; P ˆ 0.034), `When I feel blue, I often overeat' (agreed by 73% compared with 27%; P ˆ 0.020) and `Sometimes when I start eating, I just can't seem to stop' (agreed by 65% compared with 35%; P ˆ 0.025). The mean score of the EAT was slightly higher among the overweight ( ) than among their lean co-twins ( ) but the intrapair difference was not statistically signi cant. None of the twins suffered from anorexia nervosa. The overweight female twins agreed more often than their lean cotwins with statements showing a tendency to preoccupation with food and eating, e.g., `I feel that food controls my life' and `I use too much time and thinking on food' (intrapair difference of the scores ; P ˆ and ; P ˆ 0.015, respectively). Additional questions concerning emotional factors indicated that the overweight twins tended more often respond to negative feelings by eating or drinking than their lean co-twins. The difference reached statistical signi cance only for the feeling of irascibility (Table 4). Discussion We observed that the scores for disinhibition and hunger were distinctly higher among the overweight twins, indicating more problems of control over eating. This nding is in agreement with the results of Lindroos et al 25 who showed that disinhibition was the strongest 3-FEQ factor independently differentiating a sample of the obese Swedish women from the non-obese population-based controls. In addition, many of the obese twins showed a tendency to emotional eating and admitted that they have dif culties in controlling their food consumption. Our psychometric measurement (3-FEQ) showed that dietary restraint was slightly, but not signi cantly greater among the non-obese than among the overweight subjects. Recently, links between obesity and binge eating have been observed in many studies. 26,27 The 3-FEQ included some questions concerning binge eating, and the answers indicated that this symptom was more prevalent among the overweight than among the lean twins. The main result of the EAT questionnaire was that in general the lean twins did not show anorectic behaviour. Our nding that energy intake was higher among the overweight female twins than in the lean females is consistent with the study of Rissanen et al 28 in which the high energy intake was found to predict weight gain in women, but not in men. This nding is noteworthy despite the fact that basal metabolic rate is higher in obese than in normal weight subjects, as it is well documented that obese individuals tend to underestimate substantially their food intake. It is 29 ± 31 unclear if this applies similarly to both genders. According to Goldberg et al, 32 energy intake=basal metabolic rate ratio (EI=BMR) of less than 1.2 usually indicates an energy intake too low for the maintenance of bodyweight. In our sample, the EI=BMR ratio was < 1.2 in 15 subjects. Nine of them were overweight ( ve females, four males) and six were non-obese females. Our ndings support earlier ndings that both obese and normal weight individual may underreport their food intake. 31,33 The epidemiological studies suggest that a high-fat diet facilitates the development of obesity although the positive relationship has been rather weak. 34 In our study the interview on dietary history revealed no differences in the proportion of fat and other macronutrients between the overweight and lean subjects. According to Heitmann and Lissner, 35 obese persons tend to underreport fatty foods or foods rich in carbohydrates, or both, rather than underreport their total dietary intake. Thus, the E% of fat and carbohydrates may be biased due to selective underreporting by the overweight twins. It has been suggested that one of the most important reasons for the increased prevalence of obesity in the af uent societies is a low level of physical activity. 36 ± 38 It is suggested, too, that a sedentary lifestyle may have an obesity-promoting effect in men with a genetic predisposition. 39 Our results did not support this nding, because no statistically signi cant differences were shown in the physical activity indices between the overweight and lean twins. It has been suspected that measurement errors are likely to occur in self-reported physical activity. 40 In the study of Lichtman et al 30 many obese subjects highly overestimated their level of physical activity. It may be that our indices for physical activity were not sensitive enough to detect some of the possible differences between the groups. The results may also have been biased because of possible overreporting of physical activity by the obese co-twins. On the other hand, among the discordant female twin pairs physical activity may possibly explain at least partly the adiposity. 751

7 752 Statistically signi cant differences in background factors (for example level of education and cessation of smoking), which have earlier been suggested to explain the BMI difference, 28,41,42 were not found in the present study. One reason for this may be that because our study sample was small, only very powerful factors promoting obesity can reach statistical signi cance. It is also worth pointing out that majority of the overweight co-twins (70%) were only mildly overweight (BMI 30). It may also be that the minimum BMI difference used as an inclusion criterion in this study was too small to reveal weaker factors in uencing BMI difference. It is also probable that individual twin pairs had different reasons behind the variation in weight gain, resulting in non-signi cant intrapair differences in single obesity-promoting factors. For example, although the prevalence of smoking did not differ among overweight and lean twins, in the seven pairs discordant for smoking, the leaner co-twin was a smoker in six and the overweight only in one of the pairs. This can be interpreted so that although smoking does not explain the BMI difference in the whole population of 23 twin pairs, smoking is an important factor explaining the BMI difference among smoking discordant pairs. On the other hand, it is possible that both individuals within the twin-pairs share a genetic predisposition to obesity. It may be that the overweight cotwins have merely ful lled their genetic predisposition in the current environment, but the lean cotwins, on the contrary, may have maintained a lower weight through dietary restriction, smoking or increased physical activity. This assumption is supported by the nding that there was a tendency among the lean twins to avoid weight gain. Intrapair discordance for the variables of the present study may also be a consequence, not a cause of intrapair discordance for obesity, e.g., because of different recollection of events by lean and overweight twins. Further, new understanding of molecular genetics 43,44 suggests that, `environmental' factors also cover random events, such as somatic mutations, which are not transmitted from one generation to the next but which, however, involve changes in DNA. Similarly, such phenomena as inactivation of gene expression by DNA methylation may occur randomly, resulting in phenotypical differences between individuals with an identical genotype. Among humans, MZ twins replicate uniquely the complex pattern of gene by gene interactions existing in the whole organism, while they differ in the outcome of gene by environment interactions. Conclusion In conclusion, differences suggestive of dif culty in controlling eating as well as differences in energy intake in women were related to overweight, independent of genetic background. Many of the overweight twins expressed behaviour exposing them to weight gain. These habits were associated with disinhibition and increased feeling of hunger in the overweight twins. The reasons behind the observed differences in habits are assumed to be acquired rather than inherited. Our ndings suggest that difference in living conditions (e.g. family- and work related factors) may have promoted different living habits and may have led to different weight gain in identical twins. Studies in larger MZ twin cohorts with a greater number of obesity-discordant pairs, if available, might identify more speci c differences in living conditions favouring weight gain. References 1 Bouchard C. Current understanding of the etiology of obesity: genetic and nongenetic factors. Am J Clin Nutr 1991; 53: S1561 ± S Bouchard C. The genetics of obesity: Promising advances and failures. Int J Obes 1996; 20 (Suppl. 4): 11 ± Bouchard C et al. The genetics of human obesity. In: Bray GA, Bouchard C, James WPT (eds) Handbook of obesity. New York, Marcel Dekker, 1998: 157 ± Allison DB, Kaprio J, Korkeila M, Koskenvuo M, Neale MC, Hayakawa K. The heritability of body mass index among an international sample of monozygotic twins reared apart. Int J Obes 1996; 20: 501 ± Stunkard AJ, Harris JR, Pedersen NL, McClearn GE. The body-mass index of twins who have been reared apart. N Engl J Med 1990; 322: 1483 ± Sùrensen TIA, Price RA, Stunkard AJ, Schulsinger F. Genetics of obesity in adult adopters and their biological siblings. Br Med J 1989; 298: 87 ± Vogler GP, Sùrensen TIA, Stunkard AJ, Srinivasan MR, Rao DC. In uences of genes and shared family environment on adult body mass index assessed in an adoption study by a comprehensive path model. Int J Obes 1995; 19: 40 ± Heitmann BL, Lissner L, Sùrensen TIA, Bengtson C. Dietary fat intake and weight gain in women genetically predisposed for obesity. Am J Clin Nutr 1995: 61: 1213 ± Kaprio J, Sarna S, Koskenvuo M, Rantasalo I. The Finnish twin registry: formation and compilation, questionnaire study, zygosity determination procedures and research program. Prog Clin Biol Res 1978; 24B: 179 ± Korkeila M, Kaprio J, Rissanen A, Koskenvuo M. Consistency and change of body mass index and weight. A study on 5967 Finnish twin pairs. Int J Obes 1995; 19: 310 ± Hublin C, Kaprio J, Partinen M, Koskenvuo M, HeikkilaÈ K, Koskimies S, Guilleminault C. The prevalence of narcolepsy: An epidemiologic study of the Finnish Twin Cohort. Annals of Neurology 1994; 35: 709 ± Korkeila M, Kaprio J, Rissanen A, Koskenvuo M, Sùrensen TIA. Predictors of major weight gain in adult Finns: Stress, life satisfaction and personality traits. Int J Obes 1998; 22: 949 ± Sarna S, Kaprio J, Sistonen P, Koskenvuo M. Diagnosis of twin zygosity by mailed questionnaire. Hum Hered 1978; 28: 241 ± Smith SM, Penrose LS. Monozygotic and dizygotic twin analysis. Ann Human Gen 1955: 19: 273 ± Reed T. Blind assessment of zygosity using dermatoglyphics from the NHLBI twin study. Am Dermat Assoc Newsletter 1986; 5: 6 ± 11.

8 16 Heyms eld SB, Lichtman S, Baumgartner RN, Wang J, Kamen Y, Aliprantis A, Pierson RN Jr. Body composition of human comparison of two improved four-compartment models that differ in expense, technical complexity, and radiation exposure. Am J Clin Nutr 1990; 52: 52 ± Friedl KE, Deluca JP, Marchitelli LJ, Vogel JA. Reliability of body fat estimations from a four-compartment model by using density, body water, and bone mineral measurements. Am J Clin Nutr 1992; 55: 764 ± Baumgartner RN, Chumlea WC, Roche AF. Bioelectric impedance for body composition. Exerc Sport Sci Rev 1990; 18: 193 ± Siri WE. The gross composition of the body. In: Advances in Biological and Medical Physics. New York: Academic Pr; 1956; 4: 239 ± Knuts L-R, Rastas M, Haapala P, SeppaÈnen R, Karvetti R-L, Aro A. Nutrica ± a computer program for calculation of food and nutrient intake ``abstr''. The third meeting of Eurofoods, Warsaw, 24 ± 27 May 1987; Hakala P, Marniemi J, Knuts L-R, Kumpulainen J, Tahvonen R, Plaami S. Calculated vs analysed nutrient composition of weight reduction diets. Food Chemistry 1996; 57: 71 ± Stunkard AJ, Messick S. The three-factor eating questionnaire to measure dietary restraint, disinhibition and hunger. J Psychosom Res 1985; 29: 71 ± Garner DM, Gar nkel P. The eating attitudes test: an index of the symptoms of anorexia nervosa. Psycholog Med 1979; 9: 273 ± Kaprio J, Sarna S, Koskenvuo M, Rantasalo I. Baseline characteristics of the Finnish Twin Registry: Section II: History of Symptoms and illnesses, use of drugs, physical characteristics, smoking, alcohol and physical activity. Department of Public Health. Publications. M37, Helsinki Lindroos A-K, Lissner L, Mathiassen M,, Karlsson J, Sullivan M, Bengtsson C, SjoÈstroÈm L. Dietary intake in relation to restrained eating, disinhibition and hunger in obese and nonobese Swedish women. Obes Res 1997; 5: 175 ± Bruce B, Agras WS. Binge eating in females: a populationbased investigation. Int J Eating Disord 1992; 12: 365 ± Wilson GT, Nonas CA, Rosenblum GD. Assessment of binge eating in obese patients. Int J Eating Disord 1993; 13: 25 ± Rissanen AM, HelioÈvaara M, Knekt P, Reunanen A, Aromaa A. Determinants of weight gain and overweight in adult Finns. Eur J Clin Nutr 1991; 45: 419 ± Black AE, Jebb SA, Bingham SA. Validation of energy and protein intakes assessed by diet history and weighed records against energy expenditure and 24 hr urinary nitrogen excretion. Proc Nutr Soc 1991; 50: 108A. 30 Lichtman SW, Pisarska K, Berman ER, Pestone M, Dowling H, Offenbacher E, Weisel H, Heshka S, Matthews DE, Heyms eld SB. Discrepancy between self-reported and actual caloric intake and exercise in obese subjects. N Engl J Med 1992; 327: 1893 ± Pryer JA, Vrijheid M, Nichols R, Kiggins M, Elliot P. Who are the `low energy reporters' in the Dietary and Nutritional survey of British Adults? Int J Epidemiol 1997; 26: 146± Goldberg GR, Black AE, Jebb SA, Cole TJ, Murgatroyd PR, Coward WA, Prentice AM. Critical evaluation of energy intake data using fundamental principles of energy physiology. 1. Derivation of cut-off limies to identify underrecording. Eur J Clin Nutr 1991; 45: 569 ± Poppitt SD, Swann D, Black AE, Prentice AM. Assessment of selective under-reporting of food intake by both obese and non-obese women in a metabolic facility. Int J Obes 1998; 22: 303 ± Lissner L, Heitmann BL. Dietary fat and obesity: evidence from epidemiology. Eur J Clin Nutr 1995; 49: 79 ± Heitmann BL, Lissner L. Dietary underreporting by obese individuals ± is it speci c or non-speci c? Br Med J 1995; 311: 986 ± Williamson DF, Madans J, Anda R, Kleinman JC, Kahn HS, Byers T. Recreational physical activity and 10-year weight change in a U.S. national cohort. Int J Obes 1993; 17: 279 ± Davies PSW, Gregory J, White A. Physical activity and body fatness in pre-school children. Int J Obes 1995; 19: 6 ± Haapanen N, Miilunpalo S, Pasanen M, Oja P, Vuori I. Association between leisure time physical activity and 10- year body mass change among working-aged men and women. Int J Obes 1997; 21: Heitmann BL, Kaprio J, Harris JR, Rissanen A, Korkeila M, Koskenvuo M. Are genetic determinants of weight gain modi ed by leisure-time physical activity? A prospective study of Finnish twins. Am J Clin Nutr 1997; 66: 672 ± DiPietro L. Physical activity, body weight, and adiposity: an epidemiologic perspective. Exerc Sport Sci Rev 1995; 23: 275 ± Albanes D, Jones DY, Micozzi MS, Mattson M. Associations between smoking and body weight in the U.S. population: analysis of NHANES II. Am J Public Health 1987; 77: 439 ± Flegal KM, Troiano RP, Pamuk ER, Kuczmanski RJ, Campbell SM. The in uence of smoking cessation on the prevalence of overweight in the united states. New Eng J Med 1995; 333: 1165 ± Morton NE. The future of genetic epidemiology. Ann Med 1992; 24: 557 ± Rennie J. DNA's new twists. Sci Am 1993; 268: 88 ±