Chinese subjects who received health screening at the Department of Health Examination Centre of Xiangya Hospital, Central South University (located in Changsha of Hunan, China) within the time window from October 2013 to December 2015, were recruited for the present research. The study design remained consistent with some of our previously-published works [21-25]. To acquire the demographic information and health-related habits, registered nurses were engaged to interview all the participants during the physical examination by referring to a standard questionnaire.
Prior to research implementation, the study protocol had been reviewed and approved by the Ethics Committees on Research of Xiangya Hospital, Central South University (No. 201312459), and informed consent had been collected from all the participants after explaining the research content in verbal and written. For sample screening, the following inclusion criteria applied: (1) ≥40 years old; (2) the data of average consumption of specific food items and drinks during the last 12 months could be retrieved from the semi-quantitative food frequency questionnaire (SFFQ); (3) basic characteristics such as age, gender, body mass index (BMI), smoking status, alcohol drinking status, waist circumference, exercise intensity, and history of hypertension/diabetes were available; and (4) participants were not diagnosed with any other musculoskeletal disorders (e.g., rheumatoid arthritis, osteochondroma, and other bone tumors).
A total of 31542 participants received routine physical examinations at aforementioned study center from October 2013 to December 2015, and 6471 of them were qualified by the inclusion criteria. Then, 200 participants were excluded due to the existence of other musculoskeletal disorders (e.g., rheumatoid arthritis, osteochondroma, and other bone tumors), and four participants were excluded due to the unavailability of dietary questionnaire data. Eventually, 6267 participants were included for final analysis.
A validated SFFQ which was adopted in some of our previously-published studies [23, 26] was referred to for the assessment of dietary intake. The SFFQ survey was conducted twice for all participants, with at least a one-week interval in between, to comparatively evaluate their reproducibility based on the calculation of dietary Se intake. The correlation coefficient was 0.64 (P < 0.001). Then, a subsample (n = 173) was created by random selection from the study cohort, and was used to validate the SFFQ by comparing the result derived from SFFQ with that obtained from the 24-h dietary recall method over the same sample. The correlation coefficient for dietary Se intake was 0.47 (P < 0.001). The results of validation showed that the overall performance of SFFQ in the present study was consistent with previous works [27, 28].
In this SFFQ, a total of 63 food items were included based on the general dietary habit in Hunan province of China, with the intention to understand the participants’ frequency of consumption for each food item (i.e., never; once per month, 2-3 times per month, 1-3 times per week, 4-5 times per week, once per day, twice per day, or >twice per day) and average amount of consumption in each time (<100 g, 100-200 g, 201-300 g, 301-400 g, 401-500 g, or >500 g) during the previous year. The SFFQ consisted of 63 commonly consumed local food items, including the main sources of dietary Se, which included meat, fish, eggs, bread, cereals and milk . Also included almost Se-free food sources, such as fruits and vegetables . To facilitate the participants in making accurate choices, pictures of food items showing the standard weight were provided alongside the SFFQ. The compositions of macro nutrients and micro nutrients were calculated based on the Chinese Food Composition Table for all the included food items.
Assessment of other exposures
The BMI was calculated based on the measurement of weight and height for each participant. The average frequency of physical activity (never, 1-2 times per week, 3-4 times per week or ≥5 times per week), average duration of each physical activity (<30 min, 30-60 min, 1-2 h, or >2 h), as well as the smoking and drinking status were all inquired and recorded during the interview. The fasting blood glucose (FBG) was detected by the Beckman Coulter AU 5800 (Beckman Coulter Inc., Brea, CA, US), and a participant would be diagnosed of diabetes if his/her FBG ≥ 7.0 mmol/L or if he/she was undergoing any anti-diabetic treatment. The blood pressure was measured by an electronic sphygmomanometer, and a participant would be diagnosed of hypertension if his/her systolic blood pressure ≥ 140 mmHg or diastolic blood pressure ≥ 90 mmHg, or if he/she was using any anti-hypertensive drug.
The BMD was detected by a compact radiographic absorptiometry (RA) system called Alara MetriScan (Alara Inc., Fremont, CA, US), at the middle phalanges of the second to fourth fingers on the non-dominant hand. To guarantee the accuracy of measurement, all participants were requested to take off accessories from the hand before testing. The RA system would capture a high-resolution radio graphic image at an intensity expressed in arbitrary units (mineral mass/area) based on the mean value. Then, by referencing to a manufacturer-provided database, the T-score, which was used to compare the measured BMD of a participant with the average BMD of young, healthy subjects of the same gender, would be computed [31-33]. The peripheral densitometry system used in this study was characterized by high portability, low cost, and low X-ray dose (< 0.02 μSV/test). Based on the collected measurements, the participants were classified in accordance with the recommendations specified by the World Health Organization. Specifically, the BMD level within 1 standard deviation (SD) vs. a young, healthy adult is regarded as normal; the BMD level ranged from 1-2.5 SD below a young, healthy adult is regarded as osteopenia; and the BMD level equal to or 2.5 SD below a young, healthy adult is regarded as OP . Participants classified as normal and osteopenia were both regarded as non-OP.
All the continuous data was presented as means ± standard deviations, and the differences were assessed by the one-way analysis of variance (data of normal distribution) or the Kruskal-Wallis H test (data of non-normal distributions). All the categorical data was presented as percentages, and the differences were assessed by the Pearson Chi-square test. Dietary Se intake was categorized, on the basis of quartile distribution of the study population, into four categories: ≤29.2 μg/day, 29.3-39.8 μg/day, 39.9-51.8 μg/day, and ≥51.9 μg/day. The odds ratio (OR) with 95% confidence interval (CI) was calculated for all the quartiles of Se intake, and the lowest quartile was considered as the reference. A total of three models were created for multivariable analysis: the first model targeted on the dietary energy intake (quartiles); the second model further incorporated the factors of age (40-49, 50-59, 60-69, ≥ 70 years), gender (male, female) and BMI (< 28, ≥ 28 kg/m2) on the basis of the first model; and the third model further incorporated the factors of smoking status (yes/no), alcohol drinking status (yes/no), physical activity intensity (continuous), nutritional supplements (yes/no), hypertension (yes/no), diabetes (yes/no), dietary calcium intake (quartiles) and dietary fibre intake (quartiles) on the basis of the second model. Then, subgroup analysis of gender was performed. Subsequently, restricted cubic splines regression, with three knots (29.2μg/day, 39.8μg/day,
51.8μg/day) defined by the quartile distribution of dietary Se was conducted to evaluate the dose-response relationship between dietary Se intake and the prevalence of OP[35, 36]. Statistical software SPSS 21.0 and STATA 11.0 were used for data analysis. P < 0.05 was equivalent to statistically significant.