Secular increases in waist ± hip ratio among Swedish women

Transcription

1 International Journal of Obesity (1998) 22, 1116±1120 ß 1998 Stockton Press All rights reserved 0307±0565/98 $ Secular increases in waist ± hip ratio among Swedish women L Lissner *1,2, C BjoÈrkelund 2, BL Heitmann 3, L Lapidus 1, P BjoÈrntorp 4, C Bengtsson 2 1 Department of Internal Medicine, Sahlgrenska University Hospital, GoÈteborg University, GoÈteborg, Sweden; 2 Department of Primary Health Care, Vasa Hospital, GoÈteborg University, GoÈteborg, Sweden; 3 Danish Epidemiology Science Center at the Institute of Preventive Medicine, Municipal Hospital of Copenhagen, Copenhagen, Denmark, and 4 Heart Lung Institute, Sahlgrenska University Hospital, GoÈteborg University, GoÈteborg, Sweden INTRODUCTION: Secular increases in obesity have been documented in numerous populations. However, little is known about trends in fat distribution. Because men and women with elevated waist ± hip ratios (WHR) constitute a high cardiovascular risk group, it is relevant to document secular changes in WHR. This paper compares WHR in three cohorts of women, one cohort recruited in the late 1960s and the others after 12 y and 24 y intervals. SUBJECTS AND METHODS: In 1968 ± 1969, a randomly selected sample of women aged of 38 y and 50 y, was given anthropometric examinations (n ˆ 761, total). The same measurements were taken on representative cohorts aged 38 y and 50 y in 1980 ± 1981 (n ˆ 677) and 1992 ± 1993 (n ˆ 167). All analyses of trends in WHR as a function of time are age-speci c and body mass index (BMI)-adjusted. RESULTS: An interesting feature of this population is that BMI was stable from 1968 ± 1969 to 1992 ± However, WHR increased signi cantly in those aged 38 y and 50 y, independent of BMI (P ˆ 0.001, both ages). The source of these changes in WHR was a combination of increasing waist circumferences and decreasing hip circumferences. Skinfold measurements, taken only at the rst two examinations, also increased signi cantly. CONCLUSIONS: This female population appears to have experienced some changes in body shape and composition. However, we cannot explain the increasingly centralized fat patterning by changes in BMI, subcutaneous skinfold thickness or those obesity-related aspects of the modern lifestyle that we were able to measure. Keywords: secular trends; obesity; waist ± hip ratio; waist circumference; hip circumference; skinfolds; women; population studies Introduction As recently as the mid-1980s, Sweden had one of the lowest rates of obesity in the developed world, 1 although the prevalence has increased in Sweden, as it has elsewhere in Europe. 2 Secular (time-related) trends in fat distribution could, however, have as serious health implications for the population as trends in prevalence of overweight. Numerous population studies have found that waist ± hip ratio (WHR) is a powerful predictor of cardiovascular disease (CVD) and all-cause mortality, independent of body mass index (BMI). 3,4 Intra-abdominal or visceral obesity, has been proposed to represent a major explanation for the excess risk associated with an elevated WHR. More speci cally, mesenteric omental adipose tissue is associated with multiple central endocrine aberrations, which may clarify why obese individuals with large abdominal fat depots are particularly prone to develop CVD and non-insulin *Correspondence: L Lissner, SOS Secretariat, Sahlgrenska University Hospital, S GoÈ teborg, Sweden. Received 14 May 1998; revised 23 June 1998; accepted 23 July 1998 dependent diabetes. 5 Recently, increasing attention has been placed upon waist circumference rather than WHR, since it is the waist circumference that directly re ects the abdominal depots. 6 However, WHR continues to be used as an explanatory risk factor in much research on obesity and disease. Epidemiological studies involving both BMI and WHR measurements have demonstrated clearly that a strong positive correlation is to be expected between generalized and centralized fatness, 7 implying that a population that is getting fatter may also display increased central fat patterning. The purpose of the present paper is to examine whether Swedish women are becoming centrally obese to a greater extent than would be predicted by their relative body weights. An interesting feature of this particular population of middle-aged women is that they have not experienced signi cant increases in BMI during the observation period. Subjects and methods The women studied here were participants from selected age groups of the Prospective Population Study of Women in GoÈteborg, Sweden. In 1968, the

2 sample was randomly selected from the local population registry 8 and has been studied prospectively during a 24-y follow-up period. 9 The present study is based on sub-groups of the sample, speci cally, all women who were aged 38 y or 50 y at the time of their 1968=69, 1980=81 or 1992=93 examinations. With this type of data, it is possible to compare anthropometric measurements in different cohorts of similaraged women examined at three distinct points in time. Data from an additional examination conducted in 1974=75 are not included here, because no subjects were aged 38 y or 50 y at that time. The methods used for anthropometric measurements have been documented 8 and these are assumed to have remained standardized according to the protocol at all examinations. Weight in underclothes (kg), standing height (cm), waist circumference at the midpoint between the lower rib margin and the iliac crest (cm) and hip circumference at the largest location around the buttocks (cm) were measured. Triceps skinfold (mm) and subscapular skinfold (mm) were measured in 1968=69 and 1980=81 only, by the methods of Keys. 10 At the baseline examination, all anthropometric measures were taken by one person, whereas at each of the subsequent examinations, different examiners were involved. Two composite anthropometric indicators were calculated: BMI (wt=ht 2 ) and WHR. The former variable is an indicator of general adiposity, while the latter variable is used as a summary index of regional adipose tissue distribution. To describe secular trends, separate analyses are conducted for subjects aged 38 y and 50 y. It should be noted that the data are not analysed longitudinally in individuals; the terms `increases' and `changes' thus refer to time trends between cohorts. Since most of the subjects aged 38 y at the rst or second examinations re-appeared aged 50 y at the second or Secular increases in waist± hip ratio third examinations, the age-speci c description of trends ensures that no subject appears twice in any analysis. For the main analysis, a general linear model is used, with time (expressed as examination 1, 2 or 3) as the main independent variable, BMI as a covariate and WHR as the dependent variable. To illustrate this result, least squares means and standard errors of WHR are plotted over time, after adjusting for BMI. In secondary analyses involving skinfold thickness and circumferences as the time dependent variables, changes are also adjusted for concomitant trends in other anthropometric measures. In addition to an overall test for trend, dummy variables are used to determine signi cance levels for WHR changes occurring speci cally during the rst and second 12 y observation periods. Additional lifestyle-related covariates are added to this model as necessary, to test for confounding, using only data from the rst 12 y. Similarity of secular changes in WHR by sub-group, strati ed on these factors, is illustrated with groupspeci c WHR means and mean changes. Results Trends in anthropometric measures, 1968=69 ± 1992=93 Table 1 gives the number of subjects per age group who participated at each of the three examinations together with mean anthropometric values at each point in time. As shown in Table 1, BMI did not vary signi cantly across the 24 y observation period in either age group (also illustrated by broken line in Figure 1). However, WHR increased signi cantly in both the 38 y and the 50 y age groups (test of trend: P < , both). At the three examinations, the mean WHR of those 1117 Table 1 Mean anthropometric measures (with s.d.) of participants aged 38 y and 50 y at the three examinations. Participation rates (%) are given below, where a indicates percent of those sampled from population registry and b indicates percent of those participating 12 y previously, gures that do not take into account new recruits. Sample sizes may vary slightly Age 1968= = =1993 Year 38 y n ˆ % a n ˆ % a n ˆ 69 72% b P-level for time trend Weight (kg) 63.4 (11.2) 63.2 (12.0) 64.8 (10.2) 0.5 Height (cm) (5.7) (6.0) (5.7) BMI (kg=m 2 ) 23.4 (3.8) 22.9 (4.2) 23.6 (3.3) 0.9 Waist (cm) 71.6 (8.2) 76.2 (10.8) 76.4 (9.2) Hip (cm) 98.9 (7.5) 96.6 (8.9) 94.8 (8.5) WHR (0.50) (0.61) (0.05) Triceps (mm) 16.7 (5.9) 21.7 (8.5) not taken Subscap (mm) 15.3 (7.2) 20.4 (11.2) not taken y n ˆ % a n ˆ % a n ˆ 98 76% b P-level for time trend Weight (kg) 66.2 (11.1) 66.3 (11.6) 69.6 (12.5) 0.04 Height (cm) (5.5) (5.8) (6.4) BMI (kg=m 2 ) 24.8 (3.80) 24.7 (4.04) 25.0 (4.03) 0.7 Waist (cm) 74.7 (8.6) 80.1 (11.1) 80.3 (12.3) Hip (cm) (8.0) 99.2 (8.4) 98.5 (9.2) 0.04 WHR (0.51) (0.74) (0.67) Triceps (mm) 18.2 (5.3) 23.8 (7.8) not taken Subscap (mm) 18.9 (8.2) 22.9 (10.4) not taken BMI ˆ body mass index; WHR ˆ waist ± hip ratio; subscap ˆ subscapular.

3 1118 Secular increases in waist ± hip ratio subjects aged 38 y was 0.725, and 0.806, respectively. In those aged 50 y, the WHR increased from to to The trends were essentially the same after adjusting for BMI, and these BMI-adjusted WHR mean values are plotted in Figure 1. Signi cance for the WHR trend was strongest between 1968=1969 and 1980=1981 (P < , both ages). The second 12 y change was only signi cant in those aged 38 y (P ˆ 0.03). The source of these changes in WHR was a combination of increasing waist circumferences and decreasing hip circumference, as shown in Table 1. The changes in waist circumference occurred between the rst two examinations, with little further change in 1992=1993. The trends in hip circumference were more continuous over 24 y, as well as being stronger in those aged 38 y than in those aged 50 y. Further adjustment for BMI did not have any major effects on the trends in waist and hip circumference measurements (not shown). Trends in direct measures of adiposity 1968=1969 ± 1980=1981 One hypothetical explanation for the elevated WHR, is that it could represent a better measure of total adiposity than BMI per se, which contains no information on body composition. Therefore, we examined secular changes in subscapular and triceps skinfold thicknesses, both of which are direct measures of subcutaneous adiposity, with the following results. First, both skinfold thicknesses were increasing, Figure 2 Trends in subscapular and triceps skinfold thicknesses between 1968=1969 ± 1980=1981. All changes were signi cant at P < , before and after adjustment for waist ± hip ratio (WHR) and body mass index (BMI). Unadjusted values shown here. independent of WHR and BMI, suggesting that the population was getting fatter without increasing in relative weight. These unadjusted trends are shown in Table 1 and Figure 2. However, neither of these skinfold measures, when tested as covariates, could statistically account for the secular increase in WHR or waist circumference, which remained similar in magnitude and signi cance after correction for the two skinfolds (not shown). This observation is consistent with the possibility that visceral adiposity, as distinct from subcutaneous adipose tissue, may be increasing in Swedish women. Figure 1 Secular trends in waist ± hip ratio (WHR) (with s.e.m. of least squares mean), after adjustment for body mass index (BMI), in women aged 38 y and 50 y. Overall trends in WHR and trends speci c to the rst 12 y were all highly signi cant (P < , all age groups). During the second 12 y, WHR differed signi cantly for the cohorts aged 38 y (P ˆ 0.03), but not those aged 50 y (P ˆ 0.6). BMI was stable. Trends in WHR with reference to potential explanatory factors 1968=1969 ± 1981=1982 It may be hypothesized that these changes in WHR over time are dependent on the occurrence of lifestylerelated changes. This was examined using data collected between the rst two examinations only. We focused on the rst two examinations, because of the large change during this period, larger sample sizes and greatest similarity in questionnaire structure, with the following results. As mentioned above, changes in WHR were statistically independent of changes in BMI. Moreover, none of the lifestyle-related factors that were analysed in this study (listed in Table 2) could statistically account for the signi cant increase in WHR, when multivariate models were used. To further illustrate this point, we have divided the populations into subgroups, that is, smokers=non-smokers; drinkers=nondrinkers, sedentary=non-sedentary, obese=non-obese, etc. The WHR changes between 1968=1969 and 1980=1981 are shown in Table 2 to be similar in

4 Secular increases in waist± hip ratio Table 2 Group speci c waist ± hip ratio (WHR) (s.e.m.) in 1968=1969 and 1980=1981, and 12 y trends. Distribution of strati cation factors given in percentages. Signi cance levels in parentheses do not refer to WHR but rather, to chi-square test for secular change in strati cation variables 1119 Stratification variables % MeanWHR1968=1969 % MeanWHR1980=1981 Trend* between1968 ^ 1981 Age 38 y ( < 0.1) Current smokers Never- and ex-smokers Age 50 y ( < 0.05) Current smokers Never- and ex-smokers Consumes alcohol weekly a Others Age 50 y ( < 0.3) Consumes alcohol weekly a Others Sedentary b Non-sedentary Age 50 y ( < 0.001) Sedentary b Non-sedentary Age 38 y ( < 0.5) Obese (BMI > 30 kg=m 2 ) Non-obese Age 50 y ( < 0.8) Obese (BMI > 30 kg=m 2 ) Non-obese Age 38 y ( < 0.6) Taller ( > 165 cm) Smaller ( 165 cm) Age 50 y ( < 0.2) Taller ( > 165 cm) Smaller ( 165 cm) Oral contraceptives, ever c Never *Secular trends in WHR were signi cantly different from zero at P < , in all strati cation groups. a 1 serving wine or hard liquor=week; b Both at work and during leisure time; c Subjects aged 50 y not included due to small cell sizes at rst examination. BMI ˆ body mass index. magnitude, and always statistically signi cant (P < ), regardless of which sub-group is examined. Table 2 also includes frequencies for the different lifestyle related variables at two points in time. There was a trend towards more use of alcohol and oral contraceptives among those aged 38 y and more sedentary lifestyles in both age groups. Proportions of pre-vs post-menopausal subjects aged 50 y were not signi cantly different (not shown). Discussion Some years ago, it was recommended that WHR > 0.8 should be considered a cut-off value with which to identify women who are at high risk of CVD. 11,12 Using the 0.8 limit, our results suggest that the fraction of high risk women aged 38 y and 50 y, has increased from 8% and 14%, respectively, in the late 1960s, to 50% and 51%, respectively, in the early 1990s, with particularly sharp increases during the rst 12 y of this study. Alternatively, using waist circumference > 88 cm to identify high risk women, as now recommended by authorities in the US 13 and the UK, 14 the fraction of at-risk women aged 38 y and 50 y, respectively, more than doubled and tripled over the 24 y follow-up period. A number of factors should be taken into account when considering this dramatic increase, to understand whether it could be caused by certain plausible biases. One question is whether the results from this sample apply to the general female population. Although the sample is randomly selected and population-based, the number of subjects is small, so it will be of interest to re-examine this phenomenon in other representative female samples. It may also be pointed out that despite initial high participation rates, our population is likely to have become increasingly self-selected and less representative over time, and drop-outs are often assumed to have a worse health pro le than participants in population studies. However, we have recently observed that after 24 y surviving non-participants had a similar WHR and BMI at baseline as participants. 9 Moreover, since the initial WHR of subsequent dropouts was, if anything, higher (although not signi cantly so), this would in most likelihood represent a conservative bias that plays no explanatory role in our main nding of increased WHR.

5 1120 Secular increases in waist ± hip ratio A further issue is whether the results can be attributed to measurement-related biases. In 1981, when the largest secular increase was observed, there was no signi cant between-observer difference in WHR values in the entire sample of 1422 women, after adjusting for confounding by different age distributions. Of the examinations 99% were undertaken by three examiners, whose age-adjusted WHR measures were 0.799, and 0.809, respectively. Thus, the within-examiner differences are much smaller than the secular increases. The single examiner at baseline did not participate in subsequent examinations, so it is not possible to examine this formally. The observation that changes in WHR and waist circumference were independent of not only relative weight, but also of subcutaneous skinfold thicknesses, is of potential interest in relation to the metabolic syndrome and associated comorbidities, since this result implies that visceral fat might be increasing independent of increases in subcutaneous fat. This phenomenon could have strong public health implications, particularly if veri ed in other studies employing more advanced body composition techniques. The substantial increases in both skinfold thicknesses also provide indirect evidence against a methodological bias over time that is speci c to waist and=or hip circumference measurements. These ndings may in part re ect a historic trend that has been referred to as the `civilization syndrome', 5 which consists of decreased physical activity, increased stress and anxiety, increased smoking and alcohol consumption, and other changes which we are only able to partially capture in surveys such as this. In this population, there was some indication that subjects were drinking more wine and liquor, and exercising less, as their WHR was increasing, but none of the available lifestyle variables could statistically account for the increased WHR. However, residual confounding is an issue here; for instance the physical activity questions did not re ect increased use of motorized transport; likewise `usual' alcohol consumption is an imprecise measurement that is believed to be substantially underreported. Similarly, `stress' as measured in this study is a very subjective phenomenon that may not re ect important secular changes of modern times, for example, women's increasing participation in the workforce. Although our data on self-reported stress and employment status are not described in this paper, we found that these variables were unrelated to the observed WHR trends. Nevertheless, while we are unable to explain the large observed changes on the basis of many available lifestyle-related variables, this does not exclude them as possible explanatory factors. Given the direct association between WHR and CVDs, ecological reasoning might predict that rates of disease would increase in parallel with WHR among Swedish women during the period of time covered in this study. In fact, the rates of heart disease and stroke have decreased in Swedish women, 15,16 a trend that is commonly attributed to improved medical treatments and changing lifestyles. It may, however, be speculated that CVD might have declined even more rapidly in the female population, had body composition and fat distribution remained stable. Acknowledgements Funding was received from the Swedish Medical Research Council 11653, 10146, References 1 World Health Organization. Physical Status: The use and interpretation of Anthropometry. WHO: Geneva, Seidell JC. Obesity in Europe: scaling an epidemic. Int J Obes 1995; 19 (Suppl. 3): S1 ± S4. 3 Bengtsson C, BjoÈrkelund C, Lapidus L, Lissner L. Associations of serum lipid concentrations and obesity with mortality in women: 20 year follow up of participants in prospective population study in GoÈteborg, Sweden. BMJ 1993; 307: 1385 ± Larsson B, Bengtsson C, BjoÈrntorp P, Lapidus L, SjoÈstroÈm L, SvaÈrdsudd K, Tibblin G, Wedel H, Welin L, Wilhelmsen L. Is abdominal body fat distribution a major explanation for the sex difference in the incidence of myocardial infarction? Am J Epidemiol 1992; 135: 266 ± BjoÈrntorp P. Visceral Obesity: A ``Civilization Syndrome''. Obes Res 1993; 1: 206 ± Han T, Seidell J, Currall J, Morrison C, Deurenberg P, Lean M. The in uences of height and age on waist circumference as an index of adiposity in adults. Int J Obes 1997; 21: 83 ± Lapidus L, Bengtsson C,Larsson B, Pennert K, Rybo E, SjoÈstroÈm L. Distribution of adipose tissue and risk of cardiovascular disease and death: a 12 year follow up of participants in the population study of women in GoÈteborg, Sweden. BMJ 1984; 289: 1257 ± Bengtsson C, Blohme G, Hallberg L, HaÈllstroÈm T, Isaksson B, Korsan-Bengtsen K, Rybo G, Tibblin E, Tibblin G, Westerberg H. The study of women in GoÈteborg 1968 ± 1969 ± a population study. Acta Med Scand 1973; 193: 311 ± Bengtsson C, Ahlqwist M, Andersson K, BjoÈrkelund C, Lissner L, SoÈderstroÈm M. The Prospective Population Study of Women in GoÈteborg, Sweden, 1968 ± 69 to 1992 ± 93. A 24- year follow-up study with special reference to participation, representativeness and mortality. Scand J Prim Health Care 1997; 15: 214 ± Keys A. Recommendations concerning body measurements for characterization of nutritional status. In: Joseph Brozek, ed. Body measurements and human nutrition. Wayne State University Press: Detroit, Mich: 1956, pp 1 ± BjoÈrntorp P. Regional patterns of fat distribution. Ann Intern Med 1985; 103: 994 ± USDA. Dietary Guidelines for Americans. US Department of Agriculture: Washington, NHLBI. Statement on First Federal Obesity Clinical Guidelines. National Heart, Lung, and Blood Institute: Bethesda, Scottish Intercollegiate Guidelines Network (SIGN). Integrating Prevention and Management of Overweight and Obesity. Scottish Intercollegiate Guidelines Network Secretariat: Edinburgh, Peltonen M, Asplund K. Age-period-cohort effects on stroke mortality in Sweden 1969 ± 1993 and forecasts up to the year Stroke 1996; 27: 1981 ± Wilhelmsen L, Rosengren A, Johansson S, Lappas G. Coronary heart disease attact rate, incidence and mortality 1975 ± 1994 in GoÈteborg, Sweden. Eur Heart J 1997; 18: 572 ± 581.