Regular Exercise and Weight-Control Behavior Are Protective Factors against Osteoporosis for General Population: A Propensity Score-Matched Analysis from Taiwan Biobank Participants

The rising prevalence of osteoporosis, which can lead to osteoporotic fractures, increases morbidity, mortality, and socioeconomic burden. Multiple factors influencing bone mass have already been identified. The aim of this study was to investigate whether exercise habits and weight-control behaviors can lower the incidence of osteoporosis in the general population. This retrospective study recruited all participants aged 35–70 years who underwent dual-energy X-ray absorptiometry (DXA) from Taiwan Biobank (TWB). The final analysis consisted of 3320 eligible participants divided into two groups; demographic characteristics, prevalence of clinical symptoms, comorbidities, and daily behavior were collected using a self-reported questionnaire. After propensity score matching with a 1:1 ratio, 1107 out of 2214 individuals were classified into the osteoporosis group. Age, body fat rate, body shape, diabetes mellitus, and social status were found to affect the incidence of osteoporosis. Subjects with a habit of regular exercise and weight-control behavior showed decreased odds of osteoporosis. (odds ratio: 0.709 and 0.753, 95% confidence interval: 0.599–0.839 and 0.636–0.890). In the general population, regular exercise or weight-control behavior lowers the incidence of osteoporosis.


Introduction
Osteoporosis is a major public health problem, and a projected rising prevalence is expected in an aged society [1]. However, it is often asymptomatic until osteoporotic fractures occur [2]. According to statistics from the International Osteoporosis Foundation (IOF), 158 million adults aged 50 years and above are at high risk of osteoporotic fracture worldwide, and the number is predicted to double by 2040. Osteoporosis is prevalent in women and the elderly, but it also occurs in all populations and at all ages, and osteoporotic fractures cause significant morbidity, mortality and psychosocial and financial consequences [3]. In the UK, the National Health Service is estimated to spend GPB 4.4 billion per year because of fragility fractures [4]. In the USA, the cost of treatment for osteoporotic fractures is expected to reach USD 95 billion annually by 2040 [5]. Each fracture case costs more than ing health data, questionnaires, and biological specimens, released through de-identified personal information. A total of 3320 individuals were recruited to our study cohort.

Study Design and Data Collection
The participants were included in the present study if they met the following criteria: (1) participants aged 35-70 years and (2) participants underwent dual-energy X-ray absorptiometry (DXA) for BMD measurement. Individuals with a history of cancer were excluded from the study. The questionnaire-based information included epidemiological characteristics, systemic disease, symptoms, lifestyle, social status, and nutritional factors (dietary and use of mineral or vitamin supplements). As part of the questionnaire, participants were also asked if they exercised regularly. Regular exercise was defined as a participant having an amount of exercise activity over 30 min each time, at least 3 times per week. A current alcohol drinker was defined as drinking 150 mL of alcohol per week for more than six months. Participants who smoked for more than six months were defined as voluntary smokers.
Participants were then divided into two groups for comparison according to their T-scores. The osteoporosis group was defined as a T-score ≤ −2.5. Participants with a T-score > −2.5 were put in the non-osteoporosis group. A total of 1660 participants with osteoporosis diagnosed by DXA were first retrieved from Taiwan Biobank. After that, another 1660 participants without osteoporosis were randomly selected from the database following the same sex ratio as osteoporosis group. Finally, 3320 individuals underwent analysis.

Taiwan Biobank (TWB)
The TWB is conducted by the Taiwanese government and focuses on providing researchers with collaboration opportunities [36]. Several cohorts are included for population diversity, and the database is updated gradually, including a large-scale community-based cohort and hospital-based cohorts. The community-based cohort study was planned to recruit 200,000 volunteers between 30 and 70 years of age with no history of cancer. Until July 2021, 148,567 volunteer participants were recruited from three collecting sites evenly allocated in northern, southern, and eastern Taiwan. In contrast, the hospital-based cohort study aimed to recruit 100,000 patients with the most common chronic diseases in Taiwan, including lung, breast, oral cavity and colorectal cancers, hepatitis, cardiovascular disease, diabetes, chronic kidney disease, stroke, Alzheimer's disease, endometriosis, and asthma, from cooperating medical centers. Thus far, 7306 volunteer participants have been included in the project. Each subject in the TWB signed an approved informed consent form. Data collection was performed according to relevant guidelines and regulations. Apart from blood samples and physical examination, TWB researchers also conducted face-to-face interviews with participants and completed a structured questionnaire on personal information and lifestyle factors. According to official statistics, a total of 121,004 valid and available samples are in TWB currently. For demographic characteristics, the age group of 50-59 makes up the largest proportion of the age group distribution, accounting for 29.92% of the total. The male-to-female ratio is 1:1.8. In all, 57.94% subjects have studied in college, including those who already graduated. Overall, 60.49% female participants have habit of exercising, and 57.69% males do.

BMD Measurement
DXA scans of the central skeleton at the lumbar vertebrae were used to measure the BMD for the TWB as the gold-standard method by World Health Organization. In this case, the participants with metal lumbar implant, severe joint degenerative disease, and atherosclerosis of aorta were advised not to receive the examination. In addition to diagnosing osteoporosis, it has the advantage of assessing patients' risk of fracture and monitoring response to treatment. Although some participants' BMD was also measured using quantitative ultrasound (QUS) devices because of easy administration and lower price, this type of measurement correlated poorly with central DXA [37].

Statistical Analysis
All statistical analyses were performed using SPSS version 22.0 software (IBM Corp., Armonk, NY, USA), and the significance level for the two-tailed tests was set at 0.05. Data were tabulated as mean and standard deviation (SD) for quantitative variables and as absolute numbers. The Kolmogorov-Smirnov test was used to analyze the distribution of the variables. Categorical variables are presented as numbers. Comparisons between categorical groups were made using the chi-square analysis. The variables of demographic and clinical characteristics, SES, and lifestyles, which were significantly different between osteoporosis group (n = 1660) and non-osteoporosis group (n = 1660), were used in the binary logistic regression model to generate the propensity score. Using the propensity score, each group of patients was matched in a 1:1 ratio using Greedy's nearest neighbor method (caliper of 0.2). Logistic regression analysis was performed to estimate the odds ratios (ORs) and 95% confidence intervals (CIs) of osteoporosis in relation to clinical characteristics, such as regular exercise and weight-control behavior.

Results
Of the 3320 subjects who received DXA for BMD measurement, 1660 people were diagnosed with osteoporosis according to a T-score ≤ −2.5. The remaining 1660 people were referred to the non-osteoporosis group. To minimize the differences in these two groups' baseline demographic and clinical characteristics and lifestyle, propensity score matching was used. The variables that we included in propensity score matching were age, weight, fat body rate, waistline, hipline, diabetes, vertigo, cataract, heart rate, red blood cell (RBC) count, hemoglobin (Hb), hematocrit (Hct), fasting blood sugar, and uric acid. For SES and lifestyle, we also included education level, income, marital status, residence, and smoking in our propensity score-matching procedure. In total, 2214 people were finally analyzed after matching. Table 1 and Figure 1 show a comparison of demographic data and comorbidities between the two groups. The average age of the participants in the osteoporosis group was significantly higher. Participants whose T-score > −2.5 had higher body weight, fat body rate, waistline, and hipline. No significant differences were observed for most comorbidities. The prevalence of diabetes mellitus (DM) in the osteoporosis group was 7.5%, which was significantly lower than that in the non-osteoporosis group. Clinical symptoms, including vertigo, joint stiffness, neck pain, sciatic pain, and headache, were present in 6.1%, 28.1%, 31.9%, 9.8%, and 20.5% of the cases, respectively. Although the prevalence of musculoskeletal problems was high, no significant difference between the two groups was observed.
Heart rate, blood pressure, and blood tests, including complete blood count and clinical biochemistry, were measured, and the results are shown in Table 2. The osteoporosis group had a faster heart rate, lower hemoglobin, and lower hematocrit (p = <0.001, <0.001, and <0.001, respectively). Regarding the lipid profile, people without osteoporosis had lower low-density lipoprotein levels although the data of both groups were within the acceptable normal range. The osteoporosis group also had a lower average level of uric acid in the plasma.
In Table 3, social status, including income, education, marital status, and place of residence, were also compared. Income and education levels showed significant differences. A higher percentage of individuals without osteoporosis had a college degree (41.3%) than those with osteoporosis (35.8%). The place of residence was related to the T-score. Participants with osteoporosis included a higher proportion of people living in northern Taiwan where urbanization is greater. A comparison of lifestyle habits, such as eating habits, smoking, alcohol drinking, dietary patterns, and exercise, is shown in Table 4. Over 60% participants cooked at home, and the frequency of eating out did not affect the incidence of osteoporosis. Details of dietary supplements were not collected using the structure questionnaire. There was no significant difference in the pattern of dietary supplement use between the two groups. Less than half of the patients who exercised regularly had osteoporosis. Osteoporosis occurred significantly less in individuals with regular exercise and weight control than in those who did not. After propensity score matching, individuals exercising regularly or controlling body weight had a lower risk of osteoporosis (Table 5, odds ratio: 0.709 and 0.753, respectively).

Discussion
Any reason that leads to an imbalance between bone resorption and formation causes osteoporosis. Factors for osteoporosis can be classified as modifiable or non-modifiable. Reducing modifiable risk factors, such as inadequate nutritional absorption, lack of physical activity, weight loss, cigarette smoking, alcohol consumption, air pollution, or stress, has been proven to prevent osteoporosis [38]. In this study, propensity score matching was used for known non-modifiable factors and a few modifiable factors for osteoporosis. As a result, regular exercise and weight control are both independent factors in preventing osteoporosis.
Evidence from many RCTs and meta-analyses showed that exercise training, such as weight-bearing impact exercise and progressive resistance training, can promote the bone health of children and adolescents, pre-and postmenopausal women, and older men [39][40][41][42]. In this study, the exercises were not limited to resistance or weight-bearing training. On the contrary, "exercise" was defined as activities during which the participant breathed deeply or perspired in the questionnaire. An open-ended question was asked to determine the exercises done by the participants. Most people chose aerobic exercise as a routine exercise. However, the participants could write down a maximum of three exercises.
Therefore, it was difficult to further investigate which exercise was potentially effective in preventing osteoporosis. Although regular exercise and body weight control were shown to be protective factors for osteoporosis, the regularity and self-discipline were not defined or further evaluated. Participants who stood on the positive side for these two questions cared more about their health condition than did others. As shown in Supplementary Figure S1, a variety of exercises were performed. Although most exercises were classified as weight-bearing exercises, other aerobic exercises, such as swimming, biking, and tai-chi, which specifically improve cardiovascular function, balance, and body strength, were also options for participants.
An association was found between body weight and bone mineral density though the relationship was still controversial. Jensen et al. and Villareal et al. pointed out that weight loss is associated with a decrease in BMD [43,44]. As obesity increases, bone mineralization increases, leading to a reduced risk of osteoporosis and related diseases [45]. A recent systematic review and meta-analysis in 2020 also found obesity as a protective factor to osteoporosis [46]. On the contrary, some studies also have indicated that low BMD is strongly associated with increased percentage body fat (PBF) in adults with obesity [47][48][49]. To date, the mechanism of this correlation remains unclear although several explanations have been proposed. The acceptable hypothesis is that bone mass increases to accommodate the greater mechanical load from a larger body mass. However, weight loss does not adversely affect BMD as long as the muscle is maintained with exercise training. In contrast, changes in BMD were correlated with changes in thigh muscle volume [50]. Muscle was more strongly related to BMD than fat or body weight. Maintenance of lean tissue and prevention of fat mass gain may be important for maintaining bone health. According to a longitudinal cohort study, a loss of lean tissue (0.9%) and gain in fat (9%) occurred concomitantly with a decrease in bone mass (1.6%) over 5 years in men aged 25 to 96 years [51].
Type 2 diabetes mellitus is significantly associated with osteoporosis risk. A study by Lin et al. from Taiwan concluded a positive association between the presence of type 2 DM and the incidence of osteoporosis [52]. Another small sample, cross-sectional study revealed no significant difference in bone density between diabetic and nondiabetic subjects [53]. Nevertheless, DM was more common among participants without osteoporosis, in conflict with previous evidence. This could be explained by the fact that the comorbidities of participants were self-reported. Prediabetes could sometimes be mistaken for diabetes because we did not observe differences in HbA1C levels between obese individuals. According to the mean value of HbA1C, over half of the participants are in the status of prediabetes. This might affect the statistical results. Moreover, although fasting sugar in the non-osteoporosis group was higher, clinical significance was not observed, and the test is less reliable compared to HbA1C test due to human error.
In postmenopausal women, adipocytes contribute to the production of estrogen, which inhibits bone resorption by osteoclasts [54]. Increases in adipose tissue with increasing BMI in postmenopausal women result in increased estrogen production, osteoclast suppression, and a resultant increase in bone mass. Previous study also verified our findings, as it reported that people with a higher body fat rate had a lower percentage of osteoporosis [14,54].
Smoking is known to cause osteoporosis and bone fracture. Cadmium contained in tobacco has been shown to play a critical role in the underlying mechanism. Smoking is therefore a major source of cadmium exposure in smokers. Wallin et al. indicated that even low cadmium exposure increases the risk of low BMD and fractures. They found that 10 pack-years of smoking could shorten the time to first hip fracture by 3% [55]. Although the amount of smoking was not analyzed in our study, people who smoked voluntarily had a higher percentage of osteoporosis.
A recent meta-analysis reported that supplementation with calcium and vitamin D significantly reduced the risk of fractures and hip fractures by 15% and 30%, respectively [56]. A prospective study in Lyon Hospital proved that combined treatment with vitamin D and calcium supplementation is beneficial for postmenopausal women [37]. However, maintaining the correct dietary calcium phosphate is also emphasized. High calcium intake can cause the calcium paradox [57]. Satisfying the vitamin D requirements is important to obtain the best response for BMD. Vitamin D supplementation, with or without calcium, can increase bone mineral density (BMD) and have a positive anti-fracture effect. Vitamins and minerals are the main supplements taken by people in Taiwan [58]. However, the highest consumption was for vitamin E (19.4%), followed by vitamin C (9.7%) in the vitamins category. The most popular item was calcium (20.9%) in the mineral category. In people who took vitamins and minerals, only 2.1% consumed vitamin D with calcium [31]. The low usage of vitamin D plus calcium may explain why people using dietary supplements had no lower osteoporosis incidence.
A strong connection was found between socioeconomic status (SES) and health [59]. People with lower education and lower income have a greater risk of fracture [60,61]. A study based on the Korean adult population revealed that osteoporosis was more prevalent among men with a lower household income and women living in rural areas [62]. In a cohort study by Varenna et al., a protective role was played by increases in formal education [23]. An increase in educational status was associated with a significantly reduced risk of osteoporosis. Nevertheless, the study was limited because the researchers focused on a low educational level. They did not further stratify samples with more than 9 years of formal education. The results of our study showed that people with more than a high school education comprise a large proportion of osteoporosis patients. This is because the nine-year education program from primary to junior high schools started to be implemented in Taiwan since 1968. However, the percentage of osteoporosis between the high school and college education subgroups was not significantly different.
A study on osteoporosis between urban and rural areas was also conducted by Kang et al. The prevalence of osteoporosis in rural residents was higher than that in urban area residents. This tendency was also observed in our results. The urbanization level is highest in northern Taiwan, which leads to a lower percentage of osteoporosis [63].
There are some limitations to this study. Although this is a large database conducted by the Taiwanese government, over 95% of Taiwan's population of 23.4 million consists of Han Chinese. Second, data collection for this cohort was based on a structured and rough questionnaire. Personal information and lifestyle factors were mostly self-reported. Common comorbidities except cancers were included without consideration of the nature of diseases. The severity of diseases could potentially impact osteoporosis. Moreover, the interaction between drug and osteoporosis was not considered. In addition, some of the questions were too general, imprecise, or indefinite. More than one option of exercise could be chosen by the participant, and the intensity among these exercises was also different. Therefore, the frequency and the intensity discrepancies between different exercise cannot be controlled during analysis in this study to make our result more convincing. These limitations should be noticed in future study.

Conclusions
After propensity score matching, most covariates and known risk factors were balanced. For the general population in Taiwan, regular exercise or weight-control behavior is an independent factor for lowering the incidence of osteoporosis.
Supplementary Materials: The following are available online at https://www.mdpi.com/article/ 10.3390/nu14030641/s1, Figure S1: Count of different exercises (maximum of three exercises by each participant).  Informed Consent Statement: Informed consent was obtained from all recruited subjects of Taiwan Biobank Project.

Data Availability Statement:
The data presented in this study are available from the corresponding authors upon reasonable request.