Study population
This cross-sectional study population was collected from Jiangsu Province as part of a national osteoporosis epidemiological survey (2017) in China between January 2017 and April 2018. The national osteoporosis epidemiological survey (2017) was conducted in 11 provincial administrative units in China. Each administrative unit selected 4 regions, a total of 44 regions were investigated. Jiangsu Province is located in the eastern of China, with a predominantly plain terrain, and is an economically developed region of China. Four cities in Jiangsu Province were selected to represent urban (Nanjing-Liuhe District and Nantong-Gangzha District) and rural (Suzhou-Wujiang District and Taizhou-Jingjiang) areas respectively. Eligible participants were aged ≥ 20 years and had complete BMD measurement data. The exclusion criteria were as: (1) participants diagnosed with metabolic bone disease such as hyperthyroidism, hyperparathyroidism, renal failure, malabsorption syndrome, alcoholism, chronic colitis, multi-myeloma, leukemia, and chronic arthritis; (2) pregnant participants.
Sample size and sampling
The sample size calculation of the national survey (2017) was used for this study. The sample population was divided into 20–39 years and 40 years and above based on age, with the 20–39 years group being used to investigate peak bone mass in the Chinese population and the 40 years and above group being used to assess the prevalence of osteoporosis.
The sample size for people aged 40 years and above was calculated using the prevalence of osteoporosis:
$$N=deff\frac{{u}_{\alpha }^{2}p(1-p)}{{d}^{2}}$$
According to previous research [17], the estimated p value of the prevalence of osteoporosis in this study is 0.132. The value of α is 0.05 (two-sided), the value of uα is 1.96, the value of d is 0.0198 (relative error = 0.15, d = 0.15*0.132), and the design effect is 3. In addition, the stratification factors gender (male and female) and region (urban and rural) were considered in the sample size calculation. According to the formula, the average sample size of each layer (4 layers) is 3,369 people. Taking into account the above stratification factors and the 80% response rate, the minimum total sample size was calculated to be 16,845 people for the 40 years and above group, and each provincial administrative unit (11 units) sampled 1,532 people.
The sample size for people aged 20–39 years was calculated using peak BMD:
$${N=\left(\frac{{u}_{\alpha }\sigma }{\delta }\right)}^{2}$$
The value of α is 0.05 (two-side), and the value of uα is 1.96. σ is the overall standard deviation, and according to previous studies [18, 19], the standard deviation of BMD in people aged 20–39 years ranged from 0.090 g/cm2 to 0.196 g/cm2, and σ was taken as 0.196 for this study. δ is the allowable error and was taken as 25% of the standard deviation (taken as 0.090). Taking into account the gender, region, and age (20–29, 30–39 years) factors and the 80% response rate, the minimum total sample size was calculated to be 2,790 people for the 20–39 years group, and each provincial administrative unit was sampled 254 people. Therefore, the minimum total sample size required for each provincial administrative unit was 1,786, and a total of 2,710 participants were included in this study to meet the adequate sample size. In addition, our sample size was sufficient for statistical analysis according to the events per variable (EPV) rule [20].
The sampling method of this study was multistage and stratified cluster random sampling. In each survey area (4 areas), 4 towns/streets were randomly selected by cluster sampling method proportional to population size (PPS), and 2 administrative villages/neighborhood committees were randomly selected from each town/street (PPS sampling). One resident group for each administrative village/neighborhood committee was selected at random (each resident group should include at least 50 participants aged 40 years and above and 8 participants aged 20–39 years).
Data collection
The primary outcome of this study was prevalence of osteoporosis. All participants received a face-to-face interview and physical examination, which was conducted by investigators trained in standard research protocols. A standardized questionnaire was used to assess risk factors for osteoporosis, including sociodemographic factors, lifestyle factors, dietary intake, physical activity, and family history of osteoporosis or fragility fracture. Information of participants were collected including gender (male and female), age (20–29, 30–39, 40–49, 50–59, 60–69, 70–79, and 80–89 years), area (urban and rural), body mass index (BMI), systolic blood pressure (SBP), diastolic blood pressure (DBP), heart rate, BMD levels [lumbar spine (L1, L2, L3, L4), the greater trochanter, and total hip], ethnicity (Han and others), education levels (< high school, high school, and college and above), marital status (unmarried, married, cohabitation, widowed, and divorced), income, expenditure, smoking (everyday, not every day, smoking before but not present, and never), drinking (never, sometimes, often but not exceeding the norm, often and beyond the norm), family history of osteoporosis (yes, no, and unknown), diet (rice/pasta, tuber, pork, aquatic product, vegetables, and eggs), physical activity (high-intensity and moderate-intensity), activity duration, sleep duration, fasting plasma glucose, triglyceride, total cholesterol, low density lipoprotein cholesterol (LDL-C), high density lipoprotein cholesterol (HDL-C), calcium, phosphorus, 25-hydroxyvitamin D (25(OH)D), β-crosslaps (β-CTX), and procollagen type I N-terminal propeptide (PINP). BMI was divided into three types, including underweight (BMI < 18.5 kg/m2), normal weight (18.5 kg/m2 ≤ BMI < 24.0 kg/m2), and overweight (BMI ≥ 24.0 kg/m2) [21]. Hypertension was defined as SBP ≥ 140 mm Hg, and/or DBP ≥ 90 mm Hg, and/or use of antihypertensive medications within the past two weeks [22]. Hyperglycemia was defined as fasting plasma glucose ≥ 6.11 mmol/L [23]. Dyslipidemia was defined based on current lipids levels or the use of anti-dyslipidemia medications within the past two weeks. The cut-off values were 6.22 mmol/L for higher total cholesterol, 4.14 mmol/L for higher LDL-C, 1.04 mmol/L for lower HDL-C, and 2.26 mmol/L for higher triglyceride [24].
BMD measurements and the definition of osteoporosis
BMD was measured by the method of dual energy x-ray absorptiometry (DXA) using GE Lunar DXA scanners (Prodigy or iDXA; GE Healthcare, Waukesha, WI, USA). The measurement of BMD was performed first by scanning the lumbar spine, and then by scanning the left proximal femur including the femoral neck, total hip, and greater trochanter. The quality control process was carried out based on the manufacturer’s operating manual. In addition, the unified European spine phantom (ESP) was scanned 10 times to calibrate each DXA scanner and repositioned for each scan.
Osteoporosis and low BMD were defined according to World Health Organization criteria [25]. Osteoporosis was defined as a T-score ≤ -2.5 standard deviation (SD), and low BMD was defined as a -1 SD < T-score < -2.5 SD. T-scores were calculated as (measured BMD—peak BMD)/SD. The peak BMD was defined as the maximal sex-specific mean BMD. In addition, T-scores were calculated based on peak bone mass determined for males and females, respectively.
Laboratory testing
Blood biochemistry and bone turnover indicators of participants including fasting plasma glucose, triglyceride, total cholesterol, LDL-C, HDL-C, calcium, phosphorus, 25(OH)D, β-CTX, and PINP were measured by the third-party laboratory according to the relevant technical manual. Fasting venous blood of participants was collected using 5 ml vacuum coagulant tube and 2 ml Na-F anticoagulant tube, respectively. Blood samples from 2 ml Na-F anticoagulant tube were directly centrifuged and 0.6–1.0 ml of plasma was collected and dispensed into 1.5 ml blood glucose testing tube and frozen at -20℃ for fasting blood glucose testing. Blood samples in 5 ml vacuum coagulant tube were used to test other indexes, centrifuged after 45 min at room temperature, and the serum was collected and divided into two tubes, one for testing (at least 1.5 ml serum) and the other for storage, and both frozen at -20℃.
Statistical analysis
Continuous variables were described as mean and standard error (S.E.), and weighted independent samples t-test was used for comparison between groups. Categorical variables were expressed as numbers and percentages (n (%)), and the comparison between groups used weighted Chi-square test. All percentages were weighted results due to the sampling method of multistage and stratified cluster random sampling. Multivariate logistic regression analysis was utilized to analyze the factors that may be associated with osteoporosis in urban and rural populations. Variables with statistically significant differences (P < 0.05) on binary analysis were included in multivariate logistic regression analysis using stepwise regression method (backward). Variables with P ≥ 0.05 in stepwise regression were excluded step by step in each fitting process. Statistical analysis was performed by SAS 9.4 software (SAS Institute Inc., Cary, NC, USA), and bar chart were drawn using GraphPad Prism 8 software (GraphPad Software, San Diego, California, USA). P < 0.05 was considered statistically significant. Adjusted odds ratio (AOR) with 95% confidence interval (CI) were used for association assessment.
