Underweight and risk of fractures in adults over 40 years using the nationwide claims database

We aimed to investigate how underweight affects the incidence of fractures, as well as the influence of cumulative, longitudinal periods of low body mass index (BMI) and changes in body weight on fracture development. Data on adults aged 40-year and over who had three health screenings between January 1, 2007, and December 31, 2009 were used to determine the incidence of new fractures. The hazard ratios (HRs) for new fractures depending on BMI, total cumulative number of underweight, and weight change over time were calculated using Cox proportional hazard analysis. In this study, 15,955 (2.8%) of the 561,779 adults were diagnosed with fractures more than once over three health examinations. The fully adjusted HR for fractures in underweight individuals was 1.173 (95% Confidence interval [CI] 1.093–1.259). Underweight individuals diagnosed only once, twice, or three times had an adjusted HR of 1.227 (95%CI 1.130–1.332), 1.174 (95%CI 1.045–1.319), and 1.255 (95%CI 1.143–1.379), respectively. Although the adjusted HR was higher in adults who consistently had underweight (HR; 1.250 [95%CI 1.146–1.363]), those with underweight had an increased risk of fractures regardless of weight change (HR; 1.171 [95%CI 1.045–1.312], and 1.203[95%CI 1.075–1.346]). Underweight is a risk factor for fractures in adults over the age of 40 years, even if they returned to normal weight.


Methods
Data source, study design and population. The study protocol was approved by the Institutional Review Board of Korea University Ansan Hospital (Approval No. K2021-2601-001). The ethics committees of Korea University Ansan Hospital have waived the requirement to obtain informed consent as the register data analysed in this study are in anonymised and deidentified format. This study was performed in accordance with the tenets of the Declaration of Helsinki, and all research methods were carried out in accordance with appropriate regulations and guidelines.
The KNHIS database contains health information of the entire Korean population (approximately 50 million people), including diagnoses (ICD-10) and prescriptions as well as procedures 19 . All insured Koreans aged 40 years and older and all workers aged 20 years and older must undergo regular health screening examinations one or two years 20 . Among the information contained in these health-screening records include anthropometric measurements and lifestyle questionnaires, socioeconomic data and records of prescriptions and hospitalizations as well as outpatient records and death dates for the insured Korean population.
Data on adults over the age of 40 who had three consecutive general health tests between January 1, 2007, and December 31,2009 was collected from this database and used to establish a long-term cohort study. Patients who previously suffered from osteoporotic fractures and with incomplete information were excluded from the study. The impact of being underweight was amplified by applying a one-year time lag after the screening process had been carried out. In total, this research included 561,779 participants (Fig. 1). Fracture cases were tracked in this cohort from the time of initial health assessment to the end of the cohort's designated follow-up period (December 2018) or the participant's death. Fractures were defined as any fracture that resulted in a claim for hospitalization or outpatient treatment after the index general health-screening date.
Evaluation of body weight. This information was taken from the general health screening results. BMI was calculated as: weight in kilograms divided by their height in meters squared (kg/m 2 ). Underweight (< 18.5), normal (≥ 18.5 and < 23), overweight (≥ 23 and < 25), obesity (≥ 25 and < 30), and severely obesity (≥ 30) were defined by the WHO Asia-Pacific regional guidelines 21,22 . The cumulative number of underweight diagnosed at each health screening examinations (0 to 3 times) was counted and divided into four groups.
Each time a patient was screened, their body weight status was reported. The total number of people who were identified to be overweight or obese as a result of routine health examinations was used to calculate the number of people who were actually underweight. As part of our study, we evaluated the diagnoses of underweight status at the first and final health screenings to see how BMI changes over time could affect fractures. There were four groups of people in the study: underweight to underweight (U-to-U), underweight to non-underweight (U-to-N), non-underweight to underweight (N-to-U), and non-underweight to non-underweight (N-to-N).
Operational definitions of fractures. We utilized ICD-10, procedure, and radiographic study codes to search all the fracture cases from the insurance claim database 1,23,24   www.nature.com/scientificreports/ Covariates and measurements. In this study, baseline demographic data were defined as those from the most recent health screening. Socioeconomic data, laboratory results (cholesterol, fasting glucose, blood pressure, triglyceride), responses to lifestyle questionnaires (regular exercise, smoking, alcohol consumption), anthropometric measurements (height, weight, waist circumference), and medical histories, which included hypertension, diabetes, dyslipidemia, and chronic kidney disease (CKD), comprised these fundamental characteristics 25 . Regarding medical history, comorbidities were provided if a record at the health screening or past medical claim data indicated their presence. Non-smokers, former smokers, and current smokers were distinguished by their smoking status. According to the amount of alcohol consumed daily, participants were categorized as non-drinkers, light drinkers (less than 30 g/day), or heavy drinkers (more than 30 g/day). Regular exercise was defined as at least 20 min of vigorous exercise on at least three days per week or 30 min of moderate to intense exercise on at least five days per week. The income was classified as low if it fell within the bottom 20 percent of the yearly income, or as normal otherwise. Appendix I is a listing of the ICD-10 codes utilized for this investigation.

Statistical analysis.
According to the total number of underweight patients, baseline parameters of the study population are reported as mean (± standard deviation) or counts (percentages). The incidence rate (IR) per 1,000 person-years (PY) with 95% confidence intervals (95%CIs) was used to define the IR. We calculated the hazard ratios (HRs) with 95%CIs for the incidence of fractures by the BMI at the time of the index health screening examination (3rd exam; 2009) and the cumulative numbers of underweight using Cox's regression analysis. The proportional-hazards assumption was assessed using the Schoenfeld residuals test, with a logarithm of the cumulative hazard functions based on Kaplan-Meier estimates 26 . Over time, there was no significant departure from proportionality in the hazards. To decrease covariate bias, we compared HRs for unadjusted and three adjusted models: Model 1 was adjusted for age and sex; Model 2 was adjusted for age, sex, and additional environmental factors including smoking, alcohol consumption, regular exercise, and low income; and Model 3 was fully adjusted for age, sex, additional environmental factors (smoking, alcohol consumption, regular exercise, and income), and comorbidities (diabetes, hypertension, dyslipidemia, and CKD). Statistical analysis was conducted with the SAS 9.3 program (SAS Institute, Cary, NC, USA). The analysis of variance for continuous variables and the chi-square test for categorical variables were utilized, and a two-sided p < 0.05 was regarded statistically significant.

Results
Baseline characteristics. Table 1 provides a summary of the baseline characteristics according to the cumulative number of underweight participants at each health screening examination. Of the total 561,779 participants, 545,824 (97.2%) had never been diagnosed as underweight. Regarding those who were underweight, 5,354 (1.0%) were diagnosed thrice, 3,672 (0.7%) were diagnosed twice, and 6,929 (1.2%) were diagnosed only once over the three health screenings. Except for age, the four groups of never-diagnosed, once-diagnosed, twice-diagnosed, and thrice-diagnosed individuals, indicated statistically significant differences in all categories investigated. Regardless of the duration of underweight status, those in the underweight group were more likely than those in the non-underweight group to be current smokers, to abstain from alcohol intake, to engage in regular exercise, and to have a low income.

Discussion
Based on our knowledge, this is the first large population-based cohort study to establish the risk of fractures related with the cumulative burden of underweight. This study determined that underweight status increases the risk of fractures in people over 40 years of age, and increasing cumulative number in underweight does not enhance the risk of further fracture. Despite the fact that the mechanism by which underweight increases the incidence of fractures is unknown, this study discovered that underweight is a risk factor for increased fractures 27 . Hypothesized to cause osteoporosis, being underweight in humans is frequently related with malnutrition. Malnutrition leads to bone deterioration and osteoporosis 28,29 . In addition, a low BMI is strongly associated with sarcopenia development. Previous research has demonstrated that malnourished people are more susceptible to sarcopenia 30 . Sarcopenia diminishes physical strength and muscular performance, leading to injuries that increase the probability of fracture 31,32 . Therefore, a lower BMI correlates with decreased BMD levels and diminished muscle strength. However, because Table 1. Baseline characteristics of this study according to the cumulative number of the presence of underweight. DM, diabetes mellitus; CKD, chronic kidney disease; BMI, body mass index; WC, waist circumference; BP, blood pressure; HDL, high-density lipoprotein; LDL, low-density lipoprotein; eGFR, estimated glomerular filtration rate; TG, triglyceride. Numeric parameters are expressed as mean ± standard deviation and categorical parameters are expressed as counts and percentages in parentheses. *Underweight was defined as body mass index under 18.5 kg/m 2 . † Cumulative number of underweight diagnosed at each health examination (0-3 times). ‡ Alcohol consumption was divided into 3 categories; Non (no alcohol consumption), Mild (under 30 g/day consumption), and heavy (over 30 g/day consumption). § Regular exercise is defined as performing over 30 min moderate intensity exercise over 5 times per a week or over 20 min vigorous intensity exercise over 3 times per a week. || Low income is defined as total household monthly income belongs to lower 20% group among Korean entire population.    www.nature.com/scientificreports/ this was a population-based cohort study utilizing the ICD-10 diagnostic, procedure and radiographic codes, actual skeletal muscle index and BMD scores were not available. Although this study cannot definitively explain the association between low BMI, BMD, and skeletal muscle index, the vast population database confirmed that low BMI is associated with fractures. After adjusting for a number of factors, the association between underweight and fractures was analyzed. Compared to individuals who never had underweight, those who had been underweight at least once had an increased risk of fracture. In other words, the risk of fracture remained to increase regardless of the duration of underweight or the status of underweight; however, the risk of fracture does not increase if an individual consistently maintains a weight above the normal range. Individuals who have shifted from underweight to normal weight or normal weight to underweight are considered to have a normal weight but close to being underweight. It is believed that these people had low bone density and diminished muscle function, which raises the risk for fractures. Even if body weight is regained to a non-underweight status, adults over the age of 40 who have been underweight may have a loss in bone density or muscular strength due to an increase in fat mass relative to muscle mass 33,34 . Thus, adults who have ever been underweight may be at a higher risk for fractures than adults with a normal or higher body weight. This is the only study to our knowledge that used a national database to analyze the risk of fracture in the general underweight population over 40 years. All citizens were enrolled in the national health insurance system, which is a substantial quantity of data. Furthermore, the database is regularly updated; hence, it yields substantial results that may be applied to the general population.
This study has some limitations. First, the BMD T-scores could not be directly validated. Underweight had an effect on the BMD score; however the exact effect was unknown in this investigation. Second, determining the precise number of fractures was difficult. Unlike other fractures, vertebral fractures are often asymptomatic and are more likely to be underestimated than actual fractures. Third, this study utilized a national database from one nation's national health insurance services, making it difficult to adapt to multiple ethnic groups. Because fractures were identified using the fracture diagnostic code in this analysis, we were unable to validate that all fractures were appropriately diagnosed. The best way to confirm the suggested algorithm of diagnostic codes is through validation studies. In order to identify fractures, the same operational definitions established in previous studies were utilized in this study 4,23 . To diagnose fractures as precisely as possible, we excluded individuals with previous fractures and employed a one-year lag time period after underweight diagnosis. It is highly probable that the incidence rate of fractures was significantly underestimated due to the implementation of the most conservative methodology in this study. Finally, we tried to analyze as many factors as possible. While analyzing and adjusting for confounding factors is an important to increase the reliability of a study, no study is ideal and there is always a possibility of unmeasured or unanalyzed confounding factors. Therefore, while efforts were made to adjust for confounding factors in this study to obtain more accurate results, future research is needed for better understanding, and its limitations should be taken into account.
In conclusion, this study investigated whether being underweight is an important factor that increases the risk of fracture in the Korean population over 40-year-old individuals using a nationwide population-based cohort. Adults over the age of 40 who were underweight had an increased risk of fractures, even if they returned to normal weight.

Data availability
All data generated or analyzed during this study are included in this published article.