Longitudinal relation between weight change and quality of life in a community-based population: a prospective cohort study

Abstract

Background: Previous cross-sectional studies revealed that obesity is associated with lower health-related quality of life (HRQOL). This study aimed to investigate the longitudinal association between 5-year weight change and HRQOL. Methods: Data from 2414 Dutch men and women was collected. HRQOL was measured with the RAND-36. Weight change was calculated as change in weight between 1998 and 2003. Using generalized estimation equations, we primarily analyzed the influence of weight change on HRQOL for the total population and additionally, by change groups (weight losers, weight maintainers and weight gainers) using regression analysis. All analyses were stratified for gender. Results: After 5 years, 598 men (50%) and 646 women (54%) maintained their weight, 177 men (15%) and 163 women (14%) lost >2.5 kg and 410 men (35%) and 379 women (32%) gained >2.5 kg. Longitudinal associations of 5-year weight change and HRQOL were found for mental component score (MCS) in women (β = 0.13; 95% CI: 0.02–0.24), and physical component score (PCS) in men (β = −0.09; 95% CI: −0.17 to −0.00) and women (β = −0.10; 95% CI: −0.19 to −0.01). Categorizing for 5-year weight change showed that weight gainers and weight losers did not significantly differ from weight maintainers on both MCS and PCS for both men and women. Conclusion: Weight change over 5 years leaded to a slight, though significant reduction on the PCS in both genders. In women, we found a positive association between weight change and MCS.

Introduction

Worldwide, the prevalence of overweight and obesity is increasing rapidly. This increase in overweight is associated with higher risk of developing diseases and considerable cost of treatment. Overweight and obesity are associated with major chronic disorders, i.e. cardiovascular diseases, diabetes mellitus, certain types of cancer and arthritis.^1–3

Substantial evidence mainly from cross-sectional studies indicates that obesity seriously decreases people’s health-related quality of life (HRQOL). Indeed, in women, overweight was associated with lower scores on HRQOL.^4–9 However, cross-sectional studies do not provide clues for cause or effect of these associations. Longitudinal analyses may partly overcome this phenomenon. So far, few prospective studies exist on the association between weight change and HRQOL. Most of these studies are limited to obese subjects only.^5,10–12

However, Fine et al.¹³ investigated, prospectively, the impact of weight change on HRQOL in a large sample of women of the general population. They observed that weight gain was consistently associated with declines in physical functioning and vitality, as well as increased levels of pain among women of all ages and baseline levels. In addition, weight change was more strongly associated with the physical than the mental components of HRQOL. Another study also showed that weight change was associated with lower HRQOL in an older, predominantly female, population.¹⁴ Decrease in HRQOL was most noticeable in obese women who gain weight and non-obese women who lost weight.

This study expand on a previous paper that showed no beneficial effects of a community-based intervention, Hartslag Limburg, on peoples’ physical and mental HRQOL after 5 years of intervention.¹⁵ Based on these results, we were interested in the influence of weight change (weight loss, maintenance or gain) on peoples HRQOL over a period of 5 years. We hypothesized that weight gain would especially lead to a decrease of physical HRQOL. Therefore, the aim of this study is to investigate the longitudinal association between weight change and HRQOL among a large Dutch population-based sample.

Methods

Study population

The study population consists of men and women participating in a cohort study conducted by the Dutch National Institute for Public Health and the Environment (RIVM).¹⁶ In this cohort study, weight change and HRQOL were monitored over a 5-year period and measured at baseline (1998), 2-year (2000) and 5-year (2003) follow-up. The study population originated from a former monitoring study, the MORGEN study, conducted by the RIVM.¹⁷ The former study consisted of 13 184 men and women living in the Maastricht region. From this group, a gender- and age-stratified sample of 4500 subjects was selected. This was done because the aim was to include at total of 3000 subjects in the baseline measurement. A response rate of at least 65% was anticipated based on the previous experiences. Of the selected 4500 samples, 441 men and women were excluded because they had moved to another region. The remaining 4059 subjects were invited to participate. A total of 3232 (80%) wished to participate in the study, but for economical and logistical reasons, we were forced to include 3000 subjects only. However, as the aim was to include 3000 participants, 232 participants were informed that their participation was no longer required. The 3000 included participants were exposed to a community-based intervention project ‘Hartslag Limburg’. Hartslag Limburg, Dutch for Heartbeat Limburg, started in 1998. The aim of that project was to decrease the prevalence of cardiovascular diseases in the general population of the Maastricht region (population 185 000) by encouraging the inhabitants to become physically active, reduce their fat intake and quit smoking.¹⁶ A detailed description of the project and the study population is described elsewhere.^16,18 Of the 3000 subjects that were included, 2190 (73%) participated in the 2-year follow-up measurement (2000) and 2414 (81%) participated in the 5-year follow-up measurement in 2003. As we were interested in the influence of weight change on peoples HRQOL, we included those people (n = 2414) who still participated at the 5 years follow-up, because for those people (n = 586) who were lost to follow-up, it was not possible to measure the change in weight over 5 years.

Data collection

Baseline data collection and follow-up measurements started in August and lasted until February the next year. The measurements consisted of a physical examination and a self-administered questionnaire. The self-administered questionnaire consisted of questions on demographics, health status, HRQOL, smoking, physical activity, diet and chronic diseases. During the physical examination, blood pressure (systolic and diastolic), height, weight, waist circumference and total and HDL cholesterol concentration were measured.

HRQOL

HRQOL was measured by the Dutch version of the RAND-36 Health Survey (RAND-36),¹⁹ which was translated from the standardized SF-36 Health Survey.²⁰ The RAND-36 consists of 36 questions which comprise 8 multi-item scales: physical functioning, social functioning, role limitations due to physical health problems, role limitation due to emotional problems, general mental health, vitality, bodily pain and general health perception. For the RAND-36 scales, which have dimensions with four items or more, missing data were handled by computing an individual mean value of the non-missing responses if they have responded to at least 50% of the scale items. Otherwise, the total scale was coded as missing. For the RAND-36 scales, which have dimensions with fewer than four items, no missing values were allowed. All scales were scored from 0 to 100 with higher scores indicate a better HRQOL.²¹ In addition, two summary scores representing physical (PCS) and mental health (MCS) are generated. These summary score were standardized with a linear t-score with a mean and standard deviations (SDs) derived from the Dutch population. This results that the summary scores have a mean (SD) of 50 (10). The cut-off point is <50 for poor and >50 for better HRQOL. The RAND-36 comprises the same items as the SF-36;²² however, the methodology to derive the scores is different, but the effect on the final score is minimal.²⁰ The RAND-36 is a validated, reliable and responsive measure with good psychometric properties.¹⁹ It is suggested that a minimum of three to five points on any given scale may be considered clinically important.²³

Body mass index categories

During the medical examination, respondents were weighted wearing light indoor clothing after they had taken off their shoes and emptied their pockets. Body mass index (BMI) was calculated as actual measured weight divided by height squared (kg/m²). In this calculation, 1 kg was subtracted from the measured weight to adjust for the light indoor clothing. Participants were divided in three groups based on their BMI (Normal weight: BMI <25.0 kg/m²; Overweight: BMI 25.0–29.9 kg/m²; Obesity: BMI >30.0 kg/m²). Cut-off scores of BMI corresponded to recommendations proposed by the World Health Organization.²⁴

Weight change groups

Longitudinal weight change was measured as the amount of weight change between 1998 and 2003. Subjects were categorized according to their 2- and 5- year weight change as weight losers, gainers or maintainers. Weight losers included individuals who lost ≥2.5 kg. Weight gainers included individuals who gained ≥2.5 kg. Weight maintainers included individuals who maintained their weight within 2.5 kg of their original weight at baseline. Individuals who belonged to two different categories after 2 and 5 years (yoyo-effect) were excluded (n = 9) for analysis as it was not possible to categorize them correctly into one of the three groups. The categorization of groups of weight change correspond to the methods also used by Fine et al.¹³ The cut-off score of 2.5 kg was based on the average increase in weight of the Dutch population with ∼0.5 kg/year.

Covariates

The covariates were assessed by questionnaire. Social–economic status (SES) was defined by the highest level of education that was completed. Education was categorized into: low (primary school, lower occupational education or less), medium (secondary level education) and high education (university, higher occupational or corresponding education). Physical activity was assessed with questions covering time spent on leisure time activities, such as walking, bicycling, odd jobs, sports and gardening. Weekly hours of physical activity was derived as a continuous variable. Presence of chronic diseases at baseline was based on self-reported prevalence of one of the following diseases: myocardial infarction, stroke, cancer or diabetes mellitus type 2. The occurrence of chronic diseases between baseline and follow-up is determined by the absence of a disease at baseline and the self-reported presence of one or more of the above mentioned diseases at follow-up.

Statistical analysis

Descriptive data (means, SD and percentage) of the baseline characteristics were presented for men and women separately. First, we used generalized linear models to explore in more detail whether differences at baseline existed between the three BMI categories and HRQOL as measured with the eight scales plus the two summary scores of the RAND-36 at baseline. Baseline BMI was used as the independent variable, HRQOL was the dependent variable.

Next, generalized estimating equation (GEE) analyses were conducted to longitudinally analyse if 5-year weight change was associated with all eight scales and the summary scores of the RAND-36. We stratified these analyses for gender and the three baseline BMI categories. Five-year weight change was used as the independent variable and again HRQOL was used as dependent variable. Results were presented as an outcome coefficient of interest (β) and their 95% confidence interval (CI), which shows the magnitude of the longitudinal relationship between weight change and each item of the RAND-36.²⁵

Finally, we used regression analyses to explore if the weight change groups (weight loss, weight maintenance and weight gain) were associated with each dimension of HRQOL. For these analyses, we used the group of weight maintainers as the referent group. All analyses were stratified for men and women. Results were presented as mean change and their SD. These multivariate regression analyses were adjusted for age SES, physical activity measured at baseline, presence or occurrence of chronic diseases measured at baseline and 5-year follow-up and the mean of baseline and follow-up measurement of the variable under study to neutralize possible effects of regression to the mean.²⁶ For all statistical testing, we used two-sided hypothesis testing with an α level of <0.05. Data were analysed using SAS software version 9.1.

Results

Study population

Baseline characteristics, measured in 1998, of participants who completed follow-up in 2003 by sex are shown in table 1. Mean age of men and women was, respectively, 50.7 and 50.4 years. Women had a lower level of education then men and a higher prevalence of diabetes mellitus type 2. Men more often had overweight, spent more time on leisure time activities, had a higher prevalence of a myocardial infarction and cancer, and scored higher on all domains of HRQOL except for general health perception.

Additional analysis showed that responders (n = 2414) (the total number of subjects with a baseline and 5-year measurement) compared with non-responders (n = 586) of the follow-up measurement scored slightly higher, though significantly on baseline PCS (50.4 vs. 49.4) and MCS (50.2 vs. 48.6). Furthermore, responders were significantly older and had a lower BMI.

Cross-sectional association of baseline BMI and HRQOL

Cross-sectional associations between baseline BMI and HRQOL are shown in table 2. Descriptive statistics show that in both men and women HRQOL declined with higher BMI categories, most pronounced between overweight and obese subjects. For men, being obese at baseline was significantly associated with lower scores on physical functioning, physical role limitations, vitality, bodily pain, social functioning, general health perception and PCS compared with overweight and normal weight. For women, obesity was significantly associated with lower mean scores on the scales physical functioning, general health perception and PCS in comparison with people with normal weight. These associations were considered as clinical relevant.²³

Longitudinal association of weight change and HRQOL

Multivariate analyses between 5-year weight change and HRQOL, stratified by gender showed that weight change over 5 years leaded to a reduction in PCS for all men (β = −0.09, 95% CI: −0.17 to −0.00) and all women (β = −0.10, 95% CI: −0.19 to −0.01), indicating that if women gain 1 kg their PCS will reduce with 0.10 points, whereas in men it will reduce with 0.09 points. Moreover, in all women a positive association between weight change and MCS (β = 0.13, 95% CI 0.02–0.24) was found, mainly caused by positive associations, which were found on the subscales social functioning and emotional role limitations (tables 3 and 4).

When stratified for baseline BMI categories, our multivariate analyses of 5-year weight change showed that in women, an increase in weight lead to a lower score on physical functioning for all three BMI categories (normal weight: β = −0.41, 95% CI: −0.64 to −0.18; overweight: β = −0.56, 95% CI: −0.83 to −0.29; obese: β = −0.56, 95% CI: −0.85 to −0.27). Yet, a positive association was found for normal weighed women with MCS (β = 0.20, 95% CI: 0.03–0.38), social functioning (β = 0.51, 95% CI: 0.14–0.88) and emotional role limitations (β = 0.77, 95% CI: 0.14–1.40). In obese men, weight gain leaded to a lower score on the scales PCS (β = −0.28, 95% CI: −0.44 to −0.12), bodily pain (β = −0.57, 95% CI: −1.08 to −0.06) and general health perception (β = −0.50, 95% CI: −0.83 to −0.17). Yet, in obese men, weight gain lead to a higher score on vitality β = 0.33, 95% CI: 0.01 to 0.65).

Weight change groups and HRQOL

The multivariate associations between groups of weight change over the 5-year period (weight losers, weight maintainers and weight gainers) and mean change in HRQOL is presented in table 5. A total of 598 men (50%) and 646 women (54%) maintained their weight, 177 men (15%) and 163 women (14%) lost >2.5 kg and 410 men (35%) and 379 women (32%) gained >2.5 kg over the 5-year period. During the 5-years of follow-up, HRQOL deteriorate for all individuals including those who maintained their weight. In both genders, weight gainers showed a greater decline with respect to the mean change of the physical components (physical functioning, physical role limitations, vitality and bodily pain) of HRQOL compared with weight maintainers. Yet, these differences were not statistically significant. A significant difference was found between men who gained weight compared with weight maintainers on the scales physical functioning [mean (SD) change −1.7 (14.5)] and general health perception [−3.7 (15.2)]. For both genders, weight losers did not significantly differ from the weight maintainers on any subscale, nor on the PCS and MCS.

Discussion

To our knowledge, this study is one of the first to show the longitudinal association between weight change and HRQOL among a large population-based sample of all ages. We found an inverse association between weight gain and PCS (in all women and obese men) and physical functioning (women). Furthermore, weight gain was also positively associated with emotional role limitations, social functioning and MCS in women. The latter two only restricted to normal weight women. However, observed changes were rather small and none of them were above the minimal important differences of 3–5 points.²³

Few studies have reported on the longitudinal association between weight change and HRQOL. However, two studies were consistent with our results.^13,14 Also, analysis stratified for BMI showed that the association between weight change and HRQOL was not consistently for all BMI categories. For men, most associations were found in obese, whereas in women associations were mostly found among normal weight. Yet, Fine et al.,¹³ consistently found an association between weight gain and the physical domains of HRQOL among women of all BMI levels.

In this study, we have found a small positive association between weight change and MCS among women, but it was not above the minimal important differences. Previous studies have found contradictory results on the association of weight and the mental domains of HRQOL.^9,13–14,27 Some studies have reported that obesity does not result in lower scores on the mental domains.^28,29 In our study, it seems that women do not struggle with their excess body weight in their social life. Though, Lopez-Garcia et al.²⁸ have stated that in future research the generic HRQOL questionnaire, like the RAND-36, should be accompanied by an obese-specific questionnaire, which includes topics on the impact of excess weight that are not included in the RAND-36.

We have also investigated the relationship between weight change groups (weight losers, maintainers and weight gainers) and HRQOL. Even though we found that, in both genders, weight gainers showed a greater reduction in the mean change of the physical components of HRQOL compared with weight maintainers. We have only found significant difference in the mean change of the domains physical functioning and general health perception between men weight gainers and weight maintainers. In contrast to our results, Burns et al.⁴ and Fine et al.¹³ consistently found a reduction in HRQOL with weight gain. This study may be different from previous studies, because of the small number of participants in our groups of weight change to detect clinically relevant differences.

Furthermore, one study have found associations in older women.⁴ Since of small numbers, we did not stratify our results for age. Yet, the impact of weight change on HRQOL in different age groups might be dissimilar. Indeed, other studies have showed dissimilar impacts on HRQOL and weight among young and middle aged.^8,29 Previous research also have found that the reduction in HRQOL was only apparent in women with weight gain greater than 10%.⁴ Yet, ∼50% of all weight gainers and weight losers in our study population, gained or losed >5 kg. So, the magnitude of weight change in our study seems sufficient to have found possible effects in HRQOL.

Consistently with previous studies, this study also confirms that obese people report greater decrements on all domains of HRQOL than persons with a lower BMI.^4–8,30 Moreover, obesity was more strongly associated with the physical domains than the mental domains of HRQOL.^6,8,9,29,31 Previous research also have reported on gender differences in the relation between BMI and HRQOL. Some studies have showed that obese women report more problems than men.^27,29 Yet, we have only found a stronger relation between BMI and HRQOL for men.

The strengths of our study are its longitudinal design, using data of baseline, 2-year and 5-year follow-up. The present results were obtained in a large community-based sample and not selected based on people’s weight. Also, our data were based on measured weight and height instead of self-report. This is important, because men and women tend to underestimate their weight and overestimate their height.^32,33 For our longitudinal analyses, we have used GEE analyses. This method was very suitable in our study, because GEE takes into account that repeated measurements within one individual are not independent. Besides that, in GEE, subjects with incomplete data were not excluded from analyses.²⁵ Finally, we have used the RAND-36, which is a validated, reliable and responsive questionnaire to measure HRQOL.¹⁹

This study also has some limitations that should be addressed. First, the study population consists of participants from the intervention region of the community-based prevention programme ‘Hartslag Limburg’. They were exposed to an intervention programme aimed at reducing the risk of cardiovascular diseases. The level of exposure of this community-based programme might be dissimilar for each individual. Subsequently, this may reduce the generalizability of this study. Secondly, this study lacked information whether or not weight change was intentional. Hence, we did not differentiate between intentional and unintentional weight loss. Thirdly, because only information about selected chronic diseases (cardiovascular diseases, diabetes and cancer) were assessed in the questionnaires we were not able to control for all possible chronic diseases.

In conclusion, this study contributes to a better understanding of this association and shed a better light upon the possibilities of developing effective strategies to prevent the adverse effects of excess weight on the physical domains of HRQOL. We have found that weight gain was inversely associated with the physical domains of HRQOL in women and obese men. Conversely, in women, weight gain was positively associated with the mental domains of HRQOL. No differences between weight losers and weight gainers were found in HRQOL compared with weight maintainers. Future research is needed and should address with the possible diverse effects of young and older aged study populations. Furthermore, the possible positive influence of gaining weight on the mental domains of HRQOL should be explored more deeply, because this could have important implications for people’s benefits of losing weight. Finally, future research should also use an obese-specific questionnaire to measure HRQOL to assess the association between weight change and HRQOL.

Conflicts of interest: None declared.

References