OUP user menu

Intra-individual changes in body weight in population-based cohorts during four decades: the Finnish FINRISK study

Pia Pajunen, Erkki Vartiainen, Satu Männistö, Pekka Jousilahti, Tiina Laatikainen, Markku Peltonen
DOI: http://dx.doi.org/10.1093/eurpub/ckq182 107-112 First published online: 1 December 2010


Background: The aim of the study was to evaluate trends in intra-individual weight changes over the last four decades in Finland. Methods: Within the eight FINRISK surveys conducted during the years 1972–2007, we identified individuals who had participated, by chance in at least two surveys. First, individuals aged 25–54 years who had undergone a health examination within 10 years of the first examination (n = 2033) were categorized into five cohorts covering different time periods, and the annual weight change was calculated at follow-up. Second, for each individual aged 25–69 years the longest possible follow-up time (5–35 years) was identified (n = 3443), and the corresponding annual weight change was evaluated. Results: Comparing the five cohorts from different time periods, the annual increase in body weight among men between 25- and 54-year old remained stable in the range of 0.27–0.47 kg per year over the decades, whereas women belonging to the most recent cohorts (the 1990s) gained weight at 0.53–0.63 kg per year, which is more than double the oldest cohort’s weight gain of 0.24 kg. Overall, the annual increase in weight was around 0.3 kg in both sexes. The younger participants had a stronger tendency to gain weight than the older, and those with BMI < 25 kg/m2 gained more than overweight or obese individuals. Conclusion: The amount of intra-individual annual increase in body weight among men remained stable over the decades, while it increased among women. The tendency to gain weight was most pronounced among the youngest and among those initially leanest.


Overweight and obesity are the greatest contributors to the global epidemic of cardiovascular diseases (CVD),1,2 fatty liver disease3 and type 2 diabetes.4 Body mass index (BMI) has been shown to be associated with the risk of death and metabolic diseases.5,6 Risk of type 2 diabetes, in particular increases progressively with increasing levels of BMI and with the duration of obesity and overweight7 and decreases with weight loss.1 In addition, weight gain is associated with decreased health-related quality of life regardless of baseline weight and age group.8

The prevalence of obesity has continuously increased in the USA; however, most recent data show promising trends.9 In the European countries, the prevalence of obesity varies widely.10 In Finland, obesity has steadily increased during recent decades11,12 and based on 2007 data, of the Finnish population aged 25–74 years, ∼20% are obese (BMI ≥ 30 kg/m2) and 50–65% are overweight (BMI ≥ 25 kg/m2).13

Several cross-sectional studies have observed that body weight is increasing,10 but longitudinal data are scarce on individual weight changes within population over time. Danish,14 Norwegian15 and Australian16,17 study groups have previously published data on intra-individual weight changes. In Finland similar analyses have been conducted in an older cohort.18

In this study, we have utilized standardized weight measures on a large population of Finnish adults, who have participated in at least two FINRISK health surveys between the years 1972 and 2007. We constructed two different cohorts: (i) in order to evaluate in five different cohorts with 5- to 10-year follow-up whether trends in intra-individual weight changes are different between the cohorts covering last four decades and (ii) in order to estimate the intra-individual weight changes over a long follow-up period (maximum 35 years) in a population-based natural cohort aged 25–69 years.


The FINRISK surveys

In Finland, cross-sectional population FINRISK surveys have been conducted every fifth year since 1972.11,19 The first two surveys in 1972 and 1977 covered the provinces of North Karelia and Kuopio in eastern Finland and included both urban and rural areas. The south western Finland area was included from 1982 and it incorporates urban and rural municipalities. Helsinki, the capital, was included from 1992 and Oulu province in Northern Finland, a territory mostly comprised of rural areas, since 1997. The latest survey was conducted in 2007. Approximately half of the Finnish total population of 5.5 million lives in the sampled areas.

The survey methods have been carefully standardized during the study years and comply with international recommendations.20,21 For each survey, an independent-random sample was drawn from the population register. Currently, the survey covers age groups 25–74 years. The samples were stratified according to geographical area, sex and 10-year-age groups (25–34 35–44, 45–54, 55–64 and 65–74 years). Any person who had already participated in one of the surveys may have been selected by random sampling to participate in another survey.

Height was measured without shoes and hair ornaments using a wall-attached stadiometer. Height was recorded to an accuracy of 0.1 cm. Weight was measured in light clothing using a beam balance scale. Weight was recorded to an accuracy of 100 g. BMI was computed as weight/height × height (kg/m2).

In total, 68 397 individual health examinations have been conducted during the eight FINRISK surveys between 1972and 2007 (overall participation rate 75.0%). Of these, 61 319 examinations were conducted on individuals who had participated in only one of the surveys during these years. The rest of the health examinations (7078), were conducted among a total of 3443 individuals who had participated in at least two FINRISK surveys. Of these, 3262 individuals had participated in two of the eight surveys; 171 individuals had participated in three surveys; nine individuals had participated in four surveys; and one individual had participated in five surveys.


From the 3443 individuals who had participated in at least two FINRISK surveys, we constructed two different study cohorts for two different purposes:

Cohort 1

Among the 3443 individuals, we identified those who had undergone a repeated health examination within either 5 or 10 years of their first health examination (Supplementary Figure S1). The total number of such individuals (between 25 and 54 years of age) was 2033, and they were categorized in the following five cohorts to ensure a reasonably large number of participants in each (Supplementary Figure S1).

  • A = Participants in the 1972 survey with follow-up either in 1977 or 1982.

  • B = Participants in the 1977 survey with follow-up either in 1982 or 1987.

  • C = Participants in the 1982 survey with follow-up either in 1987 or 1992.

  • D = Participants in the 1987 survey with follow-up either in 1992 or 1997.

  • E = Participants in the surveys 1992, 1997 or 2002, with 5- or 10-year follow-up either in 1997, 2002 or 2007.

For each individual in these cohorts, we calculated the weight change, expressed as kilograms per year, during the 5- or 10-year follow-up period. The purpose of these cohorts was to evaluate whether there had been any change in intra-individual weight over time. In these cohorts, the age of the participants was restricted to those between 25 and 54 years at the first health examination, in order to ensure comparability between the cohorts.

Cohort 2

For each of the 3443 individuals aged 25–69 years, we identified the two health examinations with the longest possible follow-up time (Supplementary Figure S2). The follow-up periods varied from 5 to 35 years. Then, for each individual, we calculated the weight change during the follow-up period, expressed as kilograms per year. The purpose of this cohort was to estimate intra-individual weight change over a long follow-up period in a population-based ‘natural’ cohort, i.e. a cohort based on surveys that are representative of the population of the areas surveyed (randomly selected from a ‘normal population’). In addition, effects of sex, age and initial BMI level on the intra-individual weight changes were analysed within this cohort.

The cohorts, to some extent, include different persons because of the different age ranges of the cohorts, and because of the different periods of follow-up. We give two examples to clarify this: (i) a person who participated in the surveys in years 1972, 1982 and 1997 would contribute to Cohort 1 with data from the 1972 and 1982 surveys, whereas in Cohort 2 the data from the 1972 and 1997 surveys would be used to create the ‘longest follow-up period possible’; (ii) a person who participated in the surveys in 1972 and 2002 would not be included in Cohort 1 at all, but would contribute to Cohort 2 with 30 years of follow-up data.

Statistical methods

Annual weight changes, expressed as kilograms per year, were calculated for all individuals based on the follow-up time. Mean weight changes with corresponding 95% confidence intervals (CIs) were calculated, and statistical differences in these were assessed with analysis of covariance (ANCOVA). From the ANCOVA analysis, we report adjusted means with confidence intervals, and P-values for the linear trend test. Assumption of normal distribution was assessed using graphical methods. Adjustment for possible confounders (age, BMI at the baseline, follow-up time, smoking) was also done with ANCOVA. Interaction analyses between two independent variables were conducted including the corresponding product-term in the ANCOVA model. All analyses were conducted separately for both sexes. Analyses were performed with Stata software (release 10.1; STATA, College Station, TX, USA).


Cohort 1

Supplementary Figure S1 displays the number of individuals and the corresponding weight changes in the study years from which these five cohorts were constructed, comprising those individuals aged between 25 and 54 years with repeated health examination with either 5 or 10 years of follow-up in the FINRISK surveys (A, B, C, D and E). Supplementary Table S1 shows the baseline characteristics in these cohorts.

Table 1 shows that the annual increase in body weight (kg) among men aged between 25 and 54 years remained stable over the decades, varying from 0.27 to 0.47 kg per year. In women, there was a statistically significant increase in weight gain over the years. Women belonging to the most recent cohorts (the 1990s) gained weight at 0.53–0.63 kg per year, which is more than double the weight gain of the oldest cohort (0.24). This difference remained statistically significant despite adjustment for age, BMI, length of follow-up time and smoking status. A model adjusted for height gave similar results (data not shown).

View this table:
Table 1

Changes in body weight (kg) per year among the 25–54-year-old FINRISK—participants with repeated health examination with 5- or 10-year follow-up (Cohort 1, n = 2033)

Baseline year(s) → end of follow-up−72 → −77, −82−77→ −82, −87−82→ −87, −92−87→ −92, −97−92, −97, −02 →−97, −02, −07
Men, n:483230107180106
Change/year (kg)
        95% CI0.29–0.460.20–0.390.25–0.610.29–0.640.06–0.49
        95% CI0.28–0.450.22–0.410.27–0.630.32–0.660.07–0.49
        Age, BMI0.360.310.460.510.290.608
            95% CI0.27–0.440.22–0.400.28–0.650.34–0.670.09–0.49
        Age, BMI, FU-time0.360.310.460.510.290.605
            95% CI0.27–0.440.21–0.400.28–0.640.35–0.680.08–0.49
        Age, BMI, FU-time, smoking0.320.290.440.510.270.260
            95% CI0.23–0.410.19–0.390.25–0.620.34–0.670.07–0.48
Women, n483230107180106
Change/year (kg)
        95% CI0.17–0.320.30–0.490.29–0.590.48–0.780.35–0.72
        95% CI0.17–0.320.29–0.480.33–0.630.48–0.770.33–0.70
        Age, BMI0.240.380.470.630.52<0.001
            95% CI0.17–0.320.29–0.480.32–0.630.48–0.770.33–0.70
        Age, BMI, FU-time0.240.380.470.620.52<0.001
            95% CI0.17–0.320.29–0.490.32–0.630.48–0.770.33–0.70
        Age, BMI, FU-time, smoking0.250.400.500.620.53<0.001
            95% CI0.17–0.330.29–0.500.35–0.660.47–0.760.34–0.71
  • FU-time = follow-up time. P-values are for test of trend

To further elucidate the trends in weight change, interaction analyses were conducted for age and time period, and for BMI and time period, respectively. No significant interactions were observed in women (P-values 0.252 and 0.326 for age-time period and BMI-time period interactions, respectively). In men <40 years, there was a tendency towards an increase in body weight over time as compared to men over 40 years (P = 0.079 for the interaction term). The interaction between BMI and time period was not significant in men (P = 0.303).

Cohort 2

The mean follow-up time was 11.4 years for men and 11.3 years for women. In Supplementary Figure S2, the follow-up times for the FINRISK Study participants with repeated health examinations with longest possible follow-up times are shown. Supplementary Table S2 represents the baseline demographics of the participants with repeated health examination by age groups.

Changes in the body weight of participants in kilograms per year are given in table 2. The total yearly increase in weight was 0.31 kg (95% CI 0.27–0.35) in men and 0.35 kg (95% CI 0.32–0.39) in women. The younger participants had a significantly stronger tendency to gain weight. Leaner participants gained weight more quickly than overweight or obese individuals. Results remain similar regardless of the length of follow-up. The annual weight change was greater in smokers (table 2). The tallest women gained most weight (P for trend = 0.023) while the association between height and annual mean weight change was insignificant among men (table 2).

View this table:
Table 2

Mean changes in body weight (kg) per year and 95% CI among the FINRISK-participants with repeated health examinations over thewith longest possible follow-up (Cohort 2) time by sex, age, baseline BMI, baseline height, smoking status at baseline and length of follow-up

nChange/year, kilogram95% CInChange/year, kilogram95% CI
Age group (year)
BMI (kg/m2)
Height, tertilesa
Length of follow-up, year
  • P-values, test of trend

  • a: Height tertiles, men: I < 1.700, II 1.705–1.760, III ≥ 1.765; women: I < 1.57, II 1.575–1.62, III ≥ 1.625

The joint effects of age and BMI at baseline on intra-individual annual weight changes are presented in table 3. Changes in body weight were greatest in the youngest overweight women (0.52 kg per year) and in the youngest and leanest men (0.48 kg per year). Less weight gain appeared in the oldest and most obese group of individuals. The age–BMI interaction test was statistically significant in men (P = 0.048) but not in women (P = 0.290). Additionally, interactions between age and smoking, age and height, and BMI and smoking were tested, but none of these were statistically significant.

View this table:
Table 3

Annual changes in body weight (kg) with 95% CI by age and BMI among the FINRISK-participants with repeated health examination over thewith longest possible follow-up time (Cohort 2, n = 3443)

BMI, kg/m2MenWomen
Age group (year)
    95% CI:(0.41–0.54)(0.31–0.49)(0.09–0.64)(0.41–0.52)(0.38–0.65)(0.20–0.61)
    95% CI:(0.21–0.38)(0.18–0.35)(0.11–0.63)(0.35–0.49)(0.27–0.43)(0.05–0.41)
    95% CI:(0.16–0.44)(0.01–0.21)(−0.40 to 0.10)(0.05–0.27)(−0.08–0.19)(−0.11–0.27)
  • P-values for age–BMI interaction: Men P = 0.048, women P = 0.290

Table 4 shows that roughly half of the individuals maintained their weight within ±5 kg during 10-year follow-up, when including estimation based on annual weight change. However, the vast majority of participants increased their weight during the follow-up. More pronounced increases were observed in the youngest age groups compared to the older age groups in both sexes. Some 20% of obese men and women lost >5 kg in weight over the 10-year follow-up period. Nevertheless, the majority of overweight and obese people increased in body weight over the follow-up period.

View this table:
Table 4

Distribution of changes in body weight per 10-years (%) by age and BMI among the FINRISK-participants with repeated health examinations over thewith longest possible follow-up time in the FINRISK surveys (Cohort 2, n = 3443)

10-year weight change
Kilograms−10 or less−9.9 to −5.0−4.9 to 00 to 4.95.0 to 9.010 or more
    Age group, year
BMI, kg/m2
    Age group, year
    BMI, kg/m2
  • The 10-year weight change is estimated on annual weight change during the 5–35-year follow-up period in Cohort 2


To our knowledge, this is the first study in which secular time trends in intra-individual weight changes has been assessed. The amount of intra-individual increase in body weight per year among the 25–54-year-old men remained stable over the decades, varying from 0.27 to 0.47 kg per year. In contrast, for women the more recent cohorts (i.e. the 1980s and 1990s) have considerably larger weight gains compared to the cohorts from the 1970s. These changes were not explained by the additional characteristics of the cohorts such as age, smoking and baseline BMI distribution.

Data from the cross-sectional analyses have revealed that the mean BMI level in the population has increased continuously from 1972 to 2002 among men. However, no increase in BMI was observed between 2002 and 2007.11 In women, the increase in mean BMI level started later compared to men, beginning in the early 1980s. Thus, it appears that in women, the changes in intra-individual weight in adulthood (age >25 years) are in agreement with population-level trends in BMI, whereas our results on weight changes in men indicate that at least a part of the increasing population-level BMI trends do originate before the age of 25 years. During the years 1993–2004, the mean body mass of Finnish conscripts aged 20 years increased from 70.8 to 75.2 kg (P < 0.01) without a significant increase in body height (0.6 cm).22

In the present study, we also report the weight development of 25–69-year-old adult Finns who have undergone two repeated standardized health examinations in the national FINRISK Surveys, i.e. individuals who have participated by chance in two different health surveys carried out every 5 years in Finland. Overall, we observed an annual increase in weight of 0.31 kg in men and 0.35 kg in women during a mean follow-up of 11 years (Cohort 2). Our results on intra-individual weight gain are of the same magnitude as those observed in the Norwegian HUNT surveys, in which 11-year weight change was 4.2 kg (0.38 kg/year) in men and 4.7 kg (0.43 kg/year) in women.15 A Danish study, which included only overweight individuals with attained BMI of ≥28 kg/m2 during follow-up, reported slightly higher individual yearly changes in body weight.14

The younger participants gained significantly more weight than the older among both sexes. The 25–39-year-old men gained 0.44 kilograms per year which is in line with earlier data from Finland in a study conducted in the late 1960’s and early 1970’s.18 However, in the age groups between 40 and 69 years the weight gain was less common than in those younger. Moreover, the mean yearly weight gain was greater than in the earlier study in which mean weight increased with age until the fourth decade in men and then levelled off.18 Our findings agree with a population-based study of Danish adults which observed a significantly greater weight gain in 30–40-year-old individuals than in those aged 50–60 years.14 Results are also consistent with Australian data.16,17

Mid-life obesity predicts health problems in later life.23–25 A recent Finnish study showed that those men who had normal weight in later life, but had been overweight in mid-life, had the greatest mortality risk in old age. In contrast, the risk of those men who did not become overweight until after middle age did not differ from that of men with constantly normal weight.26 In old age groups (65–102 years), a higher BMI has even been associated with increased survival.27 This may, however, be true only for those who were not overweight in early adult life.26 Sufficiently, early and effective preventive measures are warranted as the behaviours or lifestyle factors associated with weight control, or lack of them, are established by late adolescence.28

In agreement with previous observations,14,29 the mean yearly weight change was greatest for those who were initially leanest. Our results remain similar regardless of the length of follow-up. The WHO/FAO Expert Consultation has recommended that individuals should avoid a weight gain >5 kg per annum during adult life, and maintain a BMI in the range 18.5–24.9 kg/m2.30 Still, the majority of individuals increased their weight during the follow-up period, especially in the younger age groups which is in line with the increasing prevalence of overweight and obesity observed repeatedly in cross-sectional studies.12 Prevalence of weight gain of 5 kg or more over 10 years was remarkably similar, at just over a third, in all BMI categories. However, not all individuals gained weight. In a fifth of obese individuals, weight decreased during the 10-year follow-up. However, it was not possible to explore the factors behind or determinants of weight changes. A large change in weight (both in terms of increase and decrease) and weight fluctuation are common phenomena and may have negative health consequences including increased mortality, independent of obesity and the trend of body weight over time.24,31,32

Several limitations of this study should be acknowledged. As we have only two measurements for each individual, we cannot estimate possible weight fluctuation occurring during the follow-up period. We did not measure the body composition of the participants and we cannot be sure as to what extent the observed weight changes correlate to changes in the amount of body fat. BMI is not a reliable indicator of adiposity, especially in older adults, because it does not differentiate lean mass from adipose tissue.33 Lean body mass is decreasing with age, but is preserved better in women than in men and thus changes in body weight better reflect changes in body fat in women than in men.34 Furthermore, the participation rate has been declining over the years and individuals who currently participate may represent the more health-conscious part of the population. Thus our result may even underestimate the real extent of the obesity problem. Selection bias in this study can arise due to various factors. We have no possibility of gathering information about possible drop-outs, since we identified only people who had participated in pre-existing surveys. Some individuals may have died prior to participation in the second survey, and only the healthiest participants may thus be included in the analyses. Furthermore, prior participation may induce a healthier life-style. We did not have information on young adults under 25 years, who are at particular risk of gaining weight.35

In conclusion, we have demonstrated that in Finland over the past four decades there have been significant changes in the intra-individual weight changes in women but not in men. Women belonging to the most recent cohorts gained on average more weight per year during a 5–10-year period than women belonging to the earlier cohorts. The increase in yearly weight gain doubled from 1970s to 1990–2000s. Findings of widespread weight gain particularly in early adulthood emphasize the need for effective preventive measures.

Supplementary data

Supplementary data are available at EURPUB online.


The Academy of Finland (grant 118065).

Conflicts of interest: None declared.

Key points

  • We evaluated trends in intra-individual weight changes over the last four decades in Finland.

  • The amount of intra-individual annual increase in body weight among men remains stable over the decades while it increased among women.

  • The tendency to gain weight was most pronounced among the youngest and among those initially leanest.


View Abstract