Varying Patterns of BMI Increase in Sex and Birth Cohorts of Korean Adults


Department of Epidemiology, School of Public Health and Institute of Health and Environment, Seoul National University, Jongno-gu Yeongun-dong 28, Seoul 110-460, Republic of Korea. E-mail:


Objective: The Korean population has recently experienced a rapid increase in obesity associated with lifestyle changes arising from economic growth. We examined trends in BMI by analyzing sex-specific birth cohorts using 3,400,727 measurements from 1,662,477 Korean adults.

Research Methods and Procedures: Birth cohort data were collected from the employees of government organizations and schools and their dependents, 20 to 65 years of age, who participated in health examinations provided by the Korean National Health Insurance Corporation in 1992, 1996, and 2000.

Results: The prevalence of obesity (BMI ≥ 30 kg/m2) was 0.8% among men and 0.3% among women in 1992, but by 2000, it had increased 2.5-fold to 2.0% in men and 2.3-fold to 0.7% in women. Over the 8-year period, the mean BMI increased 0.8 kg/m2 in men and 0.3 kg/m2 in women. The rate of BMI increase over the 8 years varied markedly among the sex-specific birth cohorts, with the steepest slopes representing the youngest men (0.2 kg/m2 per year).

Discussion: National health promotion activities should target younger men to prevent an increase in obesity-related morbidity and mortality.


The prevalence of obesity has been increasing dramatically worldwide (1, 2). This global epidemic has been attributed to an obesogenic environment, which includes high-fat and energy-dense diets and sedentary lifestyles (1, 2). Secular trend studies show that the BMI has increased over time across all ages (3, 4). However, the time trend of BMI is difficult to interpret because of confounding by cohort effects. This is particularly true for populations experiencing rapid socioeconomic changes caused by the great diversity of lifestyles between generations. South Korea has seen rapid economic growth over the past 30 years, with a 57-fold increase in per capita gross national income to $14,000 in 2004 (5). With economic growth, health indicators have also improved; life expectancy rose from 63 years in 1973 to 76 years in 2001 (6), and the infant mortality rate dropped from 4.5% in 1970 to 0.62% in 1999 (7). These demographic changes provide an opportunity for an examination of the characteristics of BMI change among different birth cohorts in the Korean population. In addition, few studies aimed at examining birth cohort differences in the BMI transition have been conducted among Korean or other Asian populations. Our study examined BMI trends by sex-specific birth cohorts of men and women.

Research Methods and Procedures

The Korean National Health Insurance Corporation provides biennial health examinations for the insured and their dependents. Details of the cohort selection method (Korean National Health Service study), the health examination, and the brief questionnaire have been published previously (8, 9). Briefly, all subjects had health examinations at medical institutions equipped with standardized, high-quality laboratories authorized by the Korea Association of Clinical Pathology and the Korea Association of Quality Control. A trained nurse at each medical institution measured subjects’ height and weight while they wore light clothing and no shoes.

For this study, we used health examination data from 1992, 1996, and 2000. Participants 20 to 65 years of age at the time of the examination, consisting of employees of government organizations and schools and their dependents, were included in the analysis. We used 3,400,727 measurements from 1,662,477 Korean adults (781,291 men and 269,280 women in 1992; 910,677 men and 315,625 women in 1996; and 805,141 men and 318,713 women in 2000). From the successive cross-sectional data, the aggregate birth cohort from 1933 to 1972 was restructured into 10 groups, each spanning 4 years, following the standard approach to cohort analysis (10). A secondary analysis limited to 607,005 subjects who provided measures for all three time periods was also conducted (Table 1). The subjects were employed continuously from 1992 to 2000. Those who did not provide measures three times included employees who resigned before 2000 and new employees hired after 1992. The age-adjusted BMI in 1992 was similar for those with all three measures (23.22 kg/m2 in men and 21.02 kg/m2 in women) and those without (23.30 kg/m2 in men and 21.10 kg/m2 in women).

Table 1.  Number of subjects in each birth cohort for three surveys and mean BMI in men and women by birth cohort and survey year (20 to 65 years old at the time of examination)
 199219962000 Secondary analysis 
Birth cohortNMean BMIPercent obese (BMI ≥ 30 kg/m2)NMean BMIPercent obese (BMI ≥ 30 kg/m2)NMean BMIPercent obese (BMI ≥ 30 kg/m2)Mean BMI change from 1992 to 2000 (Y1)*Mean BMI change from 1992–2000 (y1)*NCohort no. (X1)*
  • *

    Y1 = −0.14X1 + 1.7, y1 = −0.17X1 + 1.9 in men; Y1 = 1, y1 = 1 in women.

  • Secondary analysis using the people who provided measures on all three occasions.

 1927 to 19284,33923.030.46          
 1929 to 193215,15523.200.645,05823.411.13       
 1933 to 193650,71823.390.8127,17823.551.161,16724.101.370.710.4092510
 1937 to 194070,74623.550.9256,81323.671.209,84624.141.740.590.498,9269
 1941 to 194480,85323.550.8475,27623.741.1636,40924.051.560.500.5632,9228
 1945 to 194884,43023.580.9384,85323.871.3067,03524.191.740.610.6558,1507
 1949 to 195289,79723.480.8492,06623.841.2679,93824.221.850.740.7766,6506
 1953 to 1956112,51823.320.79120,16723.771.29107,40524.231.920.910.9183,5855
 1957 to 1960118,69423.040.75129,47723.561.23120,01824.081.841.041.0687,3974
 1961 to 196499,24122.730.66120,47723.401.27113,93124.,4223
 1965 to 196850,00922.340.50103,51423.201.29103,80323.942.381.601.6932,9482
 1969 to 19724,79121.790.7769,49622.690.8887,89223.632.291.842.002,7921
 1973 to 1976   26,30222.230.3859,33623.161.58    
 1977 to 1980      18,36122.640.83    
Total781,29123.220.78910,67723.51.20805,14123.941.93  445,717 
 Age adjusted 23.220.76    23.961.95    
 1927 to 192822623.882.65          
 1929 to 19321,21023.742.0740024.172.00       
 1933 to 19364,00823.691.852,15824.133.2013324.492.260.801.078410
 1937 to 19406,47323.511.544,99023.892.381,04524.23.060.690.878379
 1941 to 19449,26423.150.928,81323.562.013,87824.022.680.870.873,2038
 1945 to 194811,91322.640.7011,86323.121.378,49323.612.180.970.887,0407
 1949 to 195219,29422.140.5319,65922.71.0416,27223.231.621.090.9513,4406
 1953 to 195625,93521.490.3227,03622.130.6924,92622.721.261.231.0919,4085
 1957 to 196039,15221.000.1939,81421.580.4938,75922.20.951.201.0428,9594
 1961 to 196458,77120.640.1354,19921.180.3754,18121.730.651.091.0041,2093
 1965 to 196864,29720.180.0658,27420.870.2753,68621.430.631.251.1635,7772
 1969 to 197228,73720.210.1166,42920.300.1759,71320.960.450.751.0311,3311
 1973 to 1976   21,99020.420.3150,11720.410.37    
 1977 to 1980      751020.650.67    
Total269,28021.030.29315,62521.230.53318,71321.600.77  161,288 
 Age adjusted 21.210.33    21.470.74    

BMI was calculated as weight in kilograms divided by height in meters squared. To assess the change during the data period (1992 to 2000), the mean BMI and prevalence of obesity were age-standardized, using stratum-specific weights from the mean population between 1992 and 2000. We used regression models to assess whether the 8-year BMI changes were significantly different among birth cohorts.


The mean BMI increased by 0.8 kg/m2 in men (from 23.2 to 24.0 kg/m2) and by 0.3 kg/m2 in women (from 21.2 to 21.5 kg/m2) from 1992 to 2000. Over the same period, the prevalence of obesity (BMI ≥ 30 kg/m2) increased 2.5-fold in men (from 0.8% to 2.0%) and 2.3-fold in women (from 0.3% to 0.7%). The percentage of overweight individuals (25 kg/m2 ≤ BMI < 30 kg/m2) increased by ∼42% in men (from 22.4% to 32.2%) and 34% in women (from 6.7% to 9.0%). The 8-year change in BMI was fitted to regression models, with a significant linear term in men (p < 0.001) and a constant term (p < 0.001) in women (Table 1). These results indicate that BMI changes differed significantly among the birth cohorts in men but not in women.

Figure 1 shows the cross-sectional change in BMI with increasing age for both men and women between 1992 and 2000. Compared with 1992, BMI in men was higher in 2000 by a similar magnitude across all age groups, whereas in women, BMI in 2000 was only slightly higher in the young and elderly age groups. Regarding the BMI patterns by age for both 1992 and 2000, BMI in men increased with age, peaking at 50 years of age, and decreased thereafter. In women, BMI showed a J-shaped pattern, with the nadir located between 20 and 25 years of age.

Figure 1.

The change in the mean BMI by sex for the cross-sectional data for 1992 and 2000 and the change in the mean BMI by sex and by birth cohort for cohorts born between 1933 and 1972.

Figure 1 also shows the change in mean BMI in male and female birth cohorts. In men, the younger birth cohorts showed a greater increase in BMI. In women, the slope of the increase was constant for all birth cohorts. The youngest female cohort exhibited an increased BMI, from 20.2 to 20.3 kg/m2 over the 4-year age interval, that showed a decreasing pattern in the cross-sectional graph.

A secondary analysis using the subjects who provided measures on all three occasions also showed patterns of BMI increase similar to those in the synthetic cohort, compared using the regression slopes (Table 1). Among men, although the slope appeared to be slightly steeper in the secondary analysis, the difference was not statistically significant according to a test for the interaction term (p = 0.29) in the regression model.


The apparent BMI decrease with age over certain age ranges in the cross-sectional graphs did not appear in the cohort graphs, indicating that this tendency reflects cohort effects arising from younger generations with greater BMI increases.

Cross-sectional data in many previous studies have shown a lower BMI among older people (3, 11, 12). This spurious BMI decrease with age arises from cohort effects, i.e., younger generations experienced a faster BMI increase than did older generations. In our study, the age-related BMI trend decrease among the oldest men and the youngest women in the cross-sectional graphs can be explained by the same mechanism. BMI increases slowly both in the oldest men and youngest women in the cohort graphs. Thus, earlier generations are replaced by the later generations with higher BMI.

Several factors could explain the steeper slope of the BMI change in the younger male birth cohort compared with that of older birth cohorts. Older generations may be slower to adopt new lifestyles than younger generations and more likely to maintain their old habits of diet and physical activity. Although evidence is scarce, younger men may have a greater biological susceptibility to obesogenic diets and physical inactivity than older men.

A sex difference in the changing pattern of BMI has been observed in other studies. In both Sweden (13) and Finland (10), women experienced a smaller increase in BMI over time than did men across all ages. This difference was attributed to greater physical activity on the part of women, owing to less use of motor vehicles and more time spent in housekeeping activities. Sternfeld et al. (14) conducted a longitudinal study of women 42 to 52 years of age and concluded that reduced physical activity with increasing age was the main reason for weight gain in midlife, rather than hormonal effects or aging itself. In our data, the BMI increase with age among women showed very similar patterns in both 1992 and 2000, with no significant differences by birth cohort. These results suggest that the age-related reduction in physical activity remained similar over time among women in our study, whereas such inactivity might have worsened in men during the same period. Overall, the sex and birth cohort differences seen in the increased BMI slopes in our study probably arose from varying patterns of change in diet and physical activity.

Our study results may have important implications at the population level. A more rapid increase in BMI among younger men may forecast an increasing burden of obesity-related morbidity and mortality in the future, even though their current BMI levels are not in the obese range. In a cohort study of American men (15), those whose BMI increased >3 kg/m2 over 30 years had a relative risk of 7.68 for type 2 diabetes compared with those with no BMI change. In our study, the youngest men experienced a BMI increase of ∼2 kg/m2 over 8 years, a change twice as rapid as that observed in the American study. The mortality from diabetes in the Korean population increased 200% between 1985 and 2004, from 6.8/100,000 to 24.3/100,000 (16). A rapid increase of BMI over a relatively short period, as observed in this study, may result in a marked increase of type 2 diabetes incidence and mortality in Korea.

Our study has several limitations. First, the study population included men and women employed in government organizations and schools, where employment security and socioeconomic position are higher than among the average working population in Korea. Our 2000 data showed a similar BMI level to that reported in 2001 in national data for men (24.0 vs. 23.7 kg/m2); however, the level was lower for women (21.5 vs. 23.4 kg/m2) (11). Because obesity among women is often associated with a lower socioeconomic position, the increase of BMI over time may have been underestimated. Second, the sample size varied across periods and birth cohorts. However, secondary analysis using the people who provided measures on all three occasions showed a very similar tendency. Therefore, the varying sample size did not seem to involve a major selection bias. Third, even though BMI is currently considered a valid and convenient proxy measure of body fat, its limitations have been recognized, given its inability to distinguish fat mass from lean mass. BMI may not have the same implications among Asian and whites or elderly populations (17, 18, 19). Waist circumference or waist-to-hip ratio may be a better indicator for assessment of elderly obesity; if lean body mass declines because of chronic disease or physical inactivity among elderly individuals, adiposity may increase without a change in BMI. Therefore, the BMI increases in our results may have underestimated actual changes in obesity among the older age groups. However, we did not have information on waist circumference.

In conclusion, BMI increased with both age and time in Korean men. The greatest BMI slope increase over time was among young men. National health promotion activities should target these specific groups to effectively prevent future obesity-related morbidity and mortality.


This study was supported by the Ministry of Health and Welfare, Korea (Grant 01-PJ1-PG1-01CH10-0007), and the 2nd stage Brain Korea 21 project.


  1. The costs of publication of this article were defrayed, in part, by the payment of page charges. This article must, therefore, be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.