Weight loss is recommended as part of the effort to reduce the prevalence of obesity-related diseases, including type 2 diabetes mellitus (DM). Lifestyle interventions frequently lead to transient weight loss, with repeated and interrupted attempts often resulting in significant weight fluctuations, a process referred to as weight cycling. High body weight variability (BWV) is associated with many metabolic and cardiovascular diseases (CVDs) in adults.^1-3 In the past few years, a series of prospective studies, including results from meta-analyses, have reported direct associations between high BWV and increased risk of CVD, higher incidence of type 2 DM, and greater mortality.^1-4 However, the mechanisms linking high BWV to detrimental health effects are unclear. Moreover, the factors that influence BWV are unknown. We hypothesized that factors influencing BWV may underlie the association between high BWV and risk of detrimental health effects.

Some sociodemographic factors such as a low level of education may contribute to the development of obesity and regaining weight postdiet.⁵ Current economic difficulties are associated with weight gain among 40 to 60-year-old employees.⁵ Previous research has shown that fewer weight loss attempts, low weight loss expectations, and consistent exercise are key factors in long-term weight loss success and maintenance of stable weight. Identifying baseline characteristics that predict stable weight maintenance or unstable weight change will foster individualized interventions based on these characteristics.⁶

The objective of this study was to determine baseline characteristics associated with high BWV. We included variables to capture socioeconomic characteristics that could be expected to influence high BWV. We conducted a large population-based study of more than 70,000 Koreans who underwent health examinations annually for 5 years.

Participants

We used the Korean National Health Insurance Service (NHIS) data sets for claims and health checkups from January 2009 to December 2013. The Korean NHIS is a single-payer insurance organization managed by the Korean government that covers nearly all residents in Korea. The NHIS claims database includes a deidentified research data set of demographic information, primary and secondary diagnoses classified according to the International Classification of Diseases-10th Revision (ICD-10), prescriptions, procedures, hospital arrival routes, date of admission, and duration of hospitalization for all residents of Korea.^7-9 The NHIS consists of employee subscribers and regional insurance subscribers. All examinees are requested to receive annual or biannual health checkups. The health checkup results are compiled into data sets of preventive health checkups and constitute the largest nationwide cohort database with laboratory information in Korea. Further details of this database have been provided in previous reports.^7-9

In this study, individuals aged ≥20 years who underwent national health checkups between January 2009 (baseline year) and December 2013 were selected. Of the 521,993 individuals, 84,961 underwent health examinations every year from baseline to 2013 and received a total of five health checkups. We excluded subjects with prior or incident cancer (n=1,558). A total of 5,979 individuals with missing data for at least one variable also were excluded. Finally, 77,424 subjects were eligible for inclusion in the analysis. BWV was calculated using the body weight (BW) recorded in the health examination data collected over 5 years, and the characteristics affecting high BWV among baseline characteristics were investigated. The study was approved by the Institutional Review Board of Yeouido St. Mary’s Hospital, The Catholic University of Korea (No. SC22ZISE0058). Deidentified information was used for analysis; therefore, informed consent was not required.

Body weight variability

Three indices of variability were used: (1) coefficient of variation (CV); (2) variability independent of the mean (VIM); and (3) average successive variability (ASV). VIM was calculated as described previously.^3,10,11 To calculate ASV, the absolute value of the difference between two consecutive measurements of BW was obtained. Accordingly, ASV of BW was calculated as (D1+D2+D3+D4)/4, where each D is the absolute difference between two consecutive measurements of BW. High BWV was defined as the highest quartile (Q4) of BWV.

Weight change was calculated as the difference in BW over 5 years, corresponding to the period between the first (2009) and fifth (2013) health checkups. The stable weight group was defined as those with <5% weight change. We categorized weight change into three groups defined by a 5% increase or decrease in weight as follows: weight loss ≥−5%; weight change <5%; and weight gain ≥5%.¹²

Definitions of covariates

The date of the first general health checkup of each subject was designated as their index date. Demographic data; anthropometric data; history of hypertension, dyslipidemia, or DM; and data on lifestyle behaviors were obtained. Hospitals in which health examinations were performed were certified by the NHIS and subject to regular quality control. Blood samples for the measurement of serum glucose and lipid levels were obtained after an overnight fast. Body mass index was calculated as weight in kilograms divided by the square of height in meters. Information on smoking and alcohol consumption (heavy alcohol consumption defined as ≥30 g/day) was obtained using a questionnaire. Regular exercise was defined as >30 minutes of moderate physical activity at least five times per week or >20 minutes of strenuous physical activity at least three times per week. We categorized smoking status as non-smokers, ex-smokers, or current smokers. Household income was assessed using the national health insurance premium and classified into quintiles from the lowest (Q1) to the highest (Q5). The presence of DM was defined according to the presence of at least one claim per year under ICD-10 codes E10–14 and at least one claim per year for prescription of antidiabetic medication or a fasting blood glucose level ≥126 mg/dL.⁷ The presence of hypertension was defined as at least one claim per year under ICD-10 code I10 or I11 and at least one claim per year for prescription of antihypertensive agents or systolic/diastolic blood pressure (BP) ≥140/90 mmHg.⁷ The presence of dyslipidemia was defined as at least one claim per year under ICD-10 code E78 and at least one claim per year for prescription of a lipid-lowering agent or total cholesterol concentration ≥240 mg/dL.⁷

Statistical analysis

General characteristics for subjects were analyzed by one-way analyses of variance tests for continuous variables and χ² test for nominal variables and were presented as mean±standard deviation (SD) for continuous variables and number (percentage, %) for categorical variables according to quartile of BWV. BWV was measured using CV, which was shown as the main result. VIM and ASV were also measured. High BWV was defined as the highest quartile (Q4). Multiple logistic regression analysis was performed to obtain odds ratios (ORs) and 95% confidence intervals (CIs) for high BWV. A multivariable model was adjusted for age; sex; income; smoking; alcohol intake; regular exercise; and presence of DM, hypertension, or dyslipidemia. Stable weight change was defined as change within ±5%. We further analyzed the factors affecting high BWV according to weight change. A P-value <0.05 for two-tailed t-tests was considered statistically significant. All statistical analyses were performed using SAS software version 9.3 (SAS Institute).

Baseline characteristics

The characteristics of the participants according to quartile of CV in BW are listed in Table 1. High BWV was defined as the highest quartile (Q4) of CV in BW. The CV values of BW in the Q1–Q4 groups were 1.35%±0.37%, 2.23%±0.22%, 3.12%±0.31%, and 5.38%±1.94%, respectively. Subjects with high BWV were younger, more commonly female, less likely to have a high income, and current smokers. Among those with high BWV, 23% showed weight loss ≥−5% and 53% experienced weight gain ≥+5%. Among those with low BWV (Q1), 99.9% had weight change <±5%.

Factors affecting high body weight variability

Younger age was an independent factor of high BWV in both male and female. Young people under the age of 40 years were more than twice as likely to have high BWV compared to those older than 65 years (OR, 2.17; 95% CI, 1.88 to 2.50) (Table 2). The incidence of high BWV was higher in female than in male (OR, 1.52; 95% CI, 1.44 to 1.59) (Table 2).

The analysis was conducted separately on male and female. Low income was a factor influencing high BWV in male. The lowest income (Q1) was associated with a 1.9-fold higher risk of high BWV compared with the highest income (Q5: OR, 1.97; 95% CI, 1.81 to 2.13) (Table 2). Compared to non-smokers, current smoking (OR, 1.24; 95% CI, 1.18 to 1.30) was associated with high BWV in male. Comorbidity with DM was associated with high BWV in male (OR, 1.17; 95% CI, 1.07 to 1.28). For female, the effect of income status on BWV does not appear to be as great as that for male. The risk of high BWV increases by 20% to 30% in subjects with lower incomes compared with those with the highest income status (Table 2). Heavy alcohol intake (OR, 1.50; 95% CI, 1.17 to 1.91) and current smoking (OR, 1.97; 95% CI, 1.67 to 2.33) were associated with high BWV in female compared to non-drinkers and non-smokers, respectively.

The results were similar when the variability of BW was determined using VIM and ASV, except for the relationship of sex with ASV (Supplementary Tables 1 and 2).

Factors affecting high body weight variability according to weight change

Among subjects with weight gain ≥+5%, factors indicating high BWV were young age (OR, 1.77), low income (lowest income: OR, 1.41), comorbidity with DM (OR, 1.35), female sex (OR, 1.29), and current smoking (OR, 1.27) (Table 3).

Among subjects with weight loss ≥−5%, factors influencing high BWV were female sex (OR, 1.68), young age (OR, 1.59), low income (lowest income: OR, 1.28), and current smoking (OR, 1.18) (Table 3). The results were similar among subjects with a weight change <±5%, with significant factors of female sex, young age, low income, and current smoking.

In our study, low income and unhealthy behaviors of alcohol use and smoking were independently associated with high BWV. Low income, current smoking, and DM influence high BWV in male, while heavy alcohol intake and current smoking were associated with higher BWV in female.

In a cohort of relatively young adults, income variability and income decrease during a 15-year period were independently associated with nearly twofold increased risk of CVD and all-cause mortality.¹³ Income volatility implies episodes of lower income, which is associated with an array of unhealthy behaviors, such as alcohol use, smoking, and inadequate physical activity. Behavioral changes, such as increased smoking and poor dietary habits, may occur as an adaptation or coping response to psychological distress. Therefore, they are potentially important intermediate factors in disease processes.^13,14 Psychosocial stress primarily activates the hypothalamic–pituitary–adrenocortical axis and sympathetic nervous system, which can trigger pathophysiological mechanisms that include inflammation, hemostasis, and altered metabolic control.^13,14 The association between psychological distress and CVD risk is largely explained by behavioral processes.¹⁴ Treatment of psychological distress that aims to reduce CVD risk should primarily focus on health behavior change.¹⁴

In our study, low income was associated with increased BWV, especially among male. Female seem to experience a smaller effect of income status on BWV than for male. According to a study conducted in Finland, economic difficulties in adults remained associated with weight gain over 5 kg among both female and male.⁵ Among female, low level of education and house renting were associated with weight gain. Among male, frequent economic difficulties remained associated with weight gain.⁵ Another study showed that male had higher mortality and were hospitalized more than female, and sex differences were most pronounced in people with stable low income and a downward income trajectory.¹⁵ Low income individuals in their late 50s were most likely to experience mortality, particularly among male.¹⁵ The study supports the notion that male are more vulnerable to income loss than female.¹⁵ Economic stability or support may be important for maintaining a stable weight especially in male. According to the obesity fact sheet in Korea, 2019, the prevalence of obesity was higher in correlation with lower levels of education and household income.¹⁶ This trend was observed more prominently in female.¹⁶ In terms of weight change in our study, female represented a larger proportion of those with weight loss greater than 5% than male (male vs. female; 9.7% vs. 13.3%) (Supplementary Table 3). Differences in obesity prevalence according to income were mainly observed in female,¹⁶ but the effect of income status on BWV appeared to be more prominent in male. There were sex differences in the effect of income status on obesity or BWV, and further study is needed.

In our study, unhealthy lifestyles of high alcohol intake and smoking were associated with high BWV in female. Smokers generally gain weight when they quit smoking, and this weight gain can lessen some of the health benefits of quitting smoking. Cross-sectional and prospective studies suggest that adolescents are much more likely to experiment with smoking if they are dieting and trying to lose weight.^17,18 It is possible that younger adults may use smoking as a means of weight control, and trying to quit smoking could lead to weight gain, which can lead to weight fluctuation.^17,18 It has also been reported that there is a significant relationship between smoking status and weight control efforts in a representative sample of United States adults.¹⁹ Smoking cessation and reduction of alcohol intake in young female may influence stable weight control.

Comorbidity with DM was associated with weight gain rather than weight loss in high BWV. Diabetes medications can affect changes in BW. Weight gain may occur because of treatment with insulin or other hypoglycemic agents such as sulfonylurea. It was reported from the Action to Control Cardiovascular Risk in Diabetes (ACCORD) trial that a higher percentage of participants with type 2 DM who gained the most weight also had a high BWV.¹⁹ In addition, a report by Rhee et al.³ showed that a high BWV was associated with increased risk of DM. Weight gain disrupts the metabolic steady state, which indicates hyperglycemia with reactive hyperinsulinemia. If weight loss follows this weight gain, hyperinsulinemia may be even more pronounced due to the decreased basal metabolic rate.²⁰

The Look Action for Health in Diabetes (AHEAD) study reported that higher BWV was associated with significantly increased risk of CVD outcomes and death in the control group but not in the intensive lifestyle intervention group.² The authors found that an intensive weight loss program attenuated the association of BWV with adverse CVD outcomes. While the exact mechanisms for this finding are unclear, a possible explanation relates to the effect of exercise on abdominal fat.² BWV associated with higher intensity exercise in the intensive lifestyle intervention group would be expected to be associated with a reduction in abdominal adiposity.² In our study, regular exercise had a weak association with high BWV or had no significance in male.

As opposed to SD, CV and VIM are unrelated to the effect of the average and should be preferred in assessment of variability.²¹ In our study, ASV of BW in male was higher than that in female (Supplementary Table 3), but the CV and VIM of BW in male were lower those in female. The ASV, which essentially averages the absolute differences of consecutive measurements, is a more reliable representation of time series variability than SD. However, ASV is related to the effect of the average and the number of measurements. For example, when measuring BP variability, ASV is more dependent on the number of ambulatory BP readings than the SD.²² Analysis of BWV through CV, VIM, and ASV showed almost identical results in analysis of factors affecting high BWV except for sex.

We acknowledge several limitations of our study. First, selection of study subjects who had received health examinations annually for 5 years might be a source of selection bias. Male and employee subscribers were more likely to participate in the regular health checkup. Therefore, the main analysis was performed after selection by sex. Second, there were no data on whether BW change was intentional or unintentional. In our study, all patients with prior or incident cancer were excluded. Last, the possible effects of unmeasured confounding variables such as dietary habits may be present.

In conclusion, our results show that high BWV is associated with a lower income status, young people, female, current smoking, and comorbidity of DM. Addressing income inequality or recommending a healthy lifestyle (smoking cessation and limiting alcohol intake) may be a way to overcome obesity and reduce BWV. Whether such associations between high BWV and adverse health outcomes are direct, indirect, due to shared risk factors, or a combination of these requires further investigation.

Supplementary materials can be found online at https://doi.org/10.7570/jomes22063.

The authors declare no conflict of interest.

This study was supported by the 2022 JOMES Research Grant (Grant No. KSSO-J-2022005) of the Korean Society for the Study of Obesity. The funders of the study had no role in study design, data collection, data analysis, data interpretation, or writing of the report.

Study concept and design: KH and MKK; acquisition of data: KH; analysis and interpretation of data: KH; drafting of the manuscript: MKK; critical revision of the manuscript: KH; statistical analysis: KH; obtained funding: MKK; administrative, technical, or material support: KH and MKK; and study supervision: KH and MKK.