Study population
Data used in our study was collected over the last decade (2007 to 2018) by the NHANES, one of the largest and most comprehensive databases in the United States. Due to the COVID-19 outbreak, data from 2019 to 2020 was not collected, as it was considered incomplete and not nationally representative. Finally, data over four periods (2007–2008, 2009–2010, 2013–2014, and 2017–2018) were selected. Two periods (2011–2012 and 2015–2016), without the osteoporosis project, were excluded. The data of 59842 participants, over the age of 20 were assessed. The NHANES database selects the representative population of each region. Different weights are matched for each person, to better estimate the characteristics of the overall population data. We extracted the following demographic data for each participant: body mass index (BMI), physical activity information, sedentary time per day (min), smoking, alcohol drinking, hypertension, diabetes, serum folate, and osteoporosis. Participants with missing information were excluded. Finally, participants (n=8,429) were selected for our study and included in the final analysis. Both red blood cell (RBC) folate and serum folate can be used as biomarkers, we chose to evaluate serum folate in consideration of its wider clinical application. [14] The survey protocol was approved by the National Centre for Health Statistics (NCHS) Research Ethics Review Board, and all participants provided written informed consent.
Serum folate
More than a decade ago, serum folate levels were measured using the Quantaphase II Folate radio assay kit (Bio-Rad Laboratories, Hercules, CA). Subsequently, the microbiologic assay (MBA) was developed to test folate content, which is considered more accurate [15]. Since 2011, total serum folate has been calculated as the sum of 5 forms of folate, namely 5-methyltetrahydrofolate, folic acid, 5-formyltetrahydrofolate, tetrahydrofolate, and 5, 10-methenyltetrahydrofolate by isotope-dilution high performance liquid chromatography coupled with tandem mass spectrometry (LC-MS/MS) [16]. We classified folate concentration into three grades (according to the interquartile range): low, medium, and high. Q1 was considered low grade; Q2–Q3, intermediate grade; and Q4, high grade.
Osteoporosis
Information about osteoporosis was screened from the questionnaire data and the dual-energy X-ray absorptiometry (DXA) test. Responses of ‘Yes’ to family questions such as: ‘Have you ever been diagnosed with osteoporosis/brittle bones?’ was accepted as osteoporosis. Additionally, DXA is the most commonly used method for BMD measurement in a clinic, and has the advantages of high speed and low radiation. According to the World Health Organization criteria, osteoporosis should be considered if bone mineral density of the lumbar spine or femoral neck is < 2.5 standard deviation (SD) of healthy person of the same sex. [17].
Covariates
In the NHANES database, we selected the following as covariates: age, race, sex, BMI, physical activity, sedentary time, smoking, alcohol consumption, hypertension, and diabetes. BMI uses international units (kg/m2). Physical activity was defined as: ‘How much time do you spend in a typical day doing moderate-intensity activity?’ Sedentary time was defined as ‘How much time do you usually spend in a typical day sitting?’. Three levels distinguished the degree of smoking: never (the participant has smoked <100 cigarettes during their entire life), former smoker (the participant smoked in the past but has completely stopped), and current smoker (the participant has smoked >100 cigarettes during their life and currently smoke, either on some days or daily). Three levels distinguished the degree of drinking: never (the participant has hardly consumed alcohol in their life), light (the participant consumes alcohol occasionally throughout the year), and moderate to heavy (weekly or daily consumption). The diagnostic criteria for hypertension were: ‘Have you ever been diagnosed with hypertension by a doctor?’ or ‘Are you currently using any antihypertensive medication?’. The diagnostic criteria for diabetes were: ‘Have you ever been diagnosed with diabetes by a doctor?’ or ‘Are you currently using insulin, or any other anti-diabetic medication?’ The above information can be found in the questionnaire data or examination data.
Statistical analysis
R software (version 4.1.2) was used for data statistics and analysis. R is a language and environment for statistical computing and graphics, developed at Bell Laboratories (formerly AT&T, now Lucent Technologies). The Student t-test was used to compare normally distributed measurement data, and the χ2 test was used to compare the count data. Serum folate concentration was divided into three grades according to the quartile (low grade: Q1, middle grade: Q2–Q3, high grade: Q4). Taking the lower grade as reference, logistic regression analysis was performed to analyse the association between folate and osteoporosis. Stratified analysis was conducted for age, sex, race, and the results were expressed as odds ratios (ORs) and corresponding 95% confidence intervals (CIs). The restricted cubic spline (RCS) model was used to analyse the nonlinear association between folate and osteoporosis. Due to the large sample size, four points (5th, 35th, 65th, and 95th) of folate concentration were used as knots, and the mean was used as the reference point [18-19]. The NHANES adopted complex multi-stage sampling, and our data was analysed after matching with appropriate weights. Unless otherwise specified, all data were considered statistically significant with a two-tailed P value < 0.05.