This retrospective cohort study included 14,166 individuals ≥ 50 years from the NHANES database. NHANES is conducted by the Centers for Disease Control and Prevention’s National Center for Health Statistics (NCHS) biennially [14]. NHANES is a nationally representative survey assessing the health and nutritional status of the non-institutionalized civilian population in USA through collecting the questionnaire, physical examination data, and biospecimens. The study was approved by the NCHS Research Ethics Review Board, and informed consents were collected from the subjects. In the present study, people without survival information, those without data on BMD, dietary intake (energy, calcium, vitamin D, and dietary supplements), physical activity, and subjects with implausible energy intakes were excluded. Finally, 5,452 participants were included, among them 4,724 people were survived and 728 were dead. The requirement of ethical approval for this was waived by the Institutional Review Board of  The Affiliated Jiangning Hospital of Nanjing Medical University, because the data was accessed from NHANES (a publicly available database). Written informed consent was not required as this study was based on publicly available data. All methods were performed in accordance with the relevant guidelines and regulations.

Osteopenia was evaluated based on the data of BMD. BMD was detected using dual-energy X-ray absorptiometry (DXA) at the femur neck and total femur to calculate the T-score [15]. Osteopenia was defined as femur neck or total femur BMD T-score ≤   − 1.

The MD contains nine food groups (a total of 9 points), and the intakes of the groups were dichotomized by sex-specific median values. When subjects consumed presumed beneficial foods (whole grains, vegetables (excluding potatoes), fruit (including juice), nuts, legumes, fish, and the ratio of monounsaturated fatty acids-to-saturated fatty acids) above the median level and consumed presumed detrimental foods (red and processed meat) below the median level, a score of 1 point was assigned, and 0 point was assigned for all other. A score of 1 was assigned to men who consumed alcohol between 10 and 25 g/day and to women who consumed between 5 and 15 g/day, versus a score of 0 [16, 17].

All-cause mortality of subjects was the outcome in this study, and the follow-up was from October 1, 2006, to December 31, 2019. The median survival time of patients was 81 months.

Age (years), gender, race (White or other races), education (high school of above or high school or below), marital status (married or other), poverty-to-income ratio (PIR), drinking (yes or no), smoking (yes or no), metabolic equivalents (METs), hypertension (yes or no), diabetes (yes or no), dyslipidemia (yes or no), CVDs (yes or no), parents previous fracture (yes or no), previous fracture (yes or no), glucocorticoid use (adrenal cortical steroids) (yes or no), anti-osteoporosis therapy (bisphosphonates and miscellaneous bone resorption inhibitors) (yes or no), BMI (kg/m2; obesity, overweight, or underweight & normal),waist circumference (cm), total 25-hydroxyvitamin D (mcg), energy (kcal), calcium (mg), vitamin D (mcg), dietary supplements taken (yes or no), hormonetherapy (androgens and anabolic steroids, estrogens, gonadotropins, progestin, sex hormone combinations, miscellaneous hormones, and gonadotropin-releasing hormone and analogs) [inapplicable (male), yes or no], and menopausal [inapplicable (male), yes or no] were potential confounders in the current study.

The PIR was calculated by dividing family (or individual) income to the poverty guidelines specific for each survey year. Physical activity was converted into energy consumption based on the questionnaire in the database. Energy consumption (MET × min) = recommended MET × exercise time of corresponding activity (min), which can be converted into weekly energy consumption. According to the questionnaire, those who smoked at least 100 cigarettes during their lifetime was regarded as smoking. CVD was defined based on the answer of “Yes” to variable MCQ160D (Ever told you had angina or heart failure?), MCQ160E (Ever told you had heart attack?), MCQ160C (Has a doctor or other health professional ever told you that you had coronary heart disease?), MCQ160F (Ever told you had a stroke?), MCQ160B (Ever told had congestive heart failure?), or those received CVD drugs based on 40-CARDIOVASCULAR AGENTS-41, 43, 44, 45, 46, 50, 51, 52, 53, 54, 56, 303, 340, 342, 430, 433, 483. Dyslipidemia was defined based on total cholesterol ≥ 200 mg/dL (5.2 mmol/L) or triglyceride ≥ 150 mg/dL (1.7 mmol/L) or low-density lipoprotein cholesterol ≥ 130 mg/dL (3.4 mmol/L) or high-density lipoprotein cholesterol ≤ 40 mg/dL (1.0 mmol/L), previous physician-diagnosed hypercholesterolemia (BPQ080) or receiving cholesterol-lowering treatment (BPQ090D) or lipid-lowering drugs (358-metabolic agents-19-antihyperlipemic agents). Hypertension was defined as systolic blood pressure ≥ 130 mmHg and or diastolic blood pressure ≥ 80 mmHg or previous physician-diagnosed hypertension (BPQ020) or take blood pressure medications (BPQ040A or drug code 40-CARDIOVASCULAR AGENTS-42, 47, 48, 49, 482, 55). Diabetes was diagnosed based on glycated hemoglobin ≥ 6.5%, fasting glucose ≥ 126 mg/dL, 2 h oral glucose tolerance test blood glucose ≥ 200 mg/dL, previous physician-diagnosed diabetes [DIQ010 (Doctor told you have diabetes)], insulin use (DIQ050) or antidiabetic agents (DIQ070 or 358-metabolic Agents-99-antidiabetic agents). Energy was calculated based on energy intake from Day 1 dietary recall, and the variable in the NHANES database from 1999 to 2002 was DRXTKCAL and from 2003 to 2004 was DR1TKCAL.

Mean ± standard deviation (mean ± SD) was used to describe the measurement data with normal distribution, and t-test was used to compare the differences between the two groups. Median and quartiles [M (Q1, Q3)] were used to describe the measurement data with abnormal distribution, and Wilcoxon rank sum test was used to compare the differences between the two groups. The number of cases and percentages [n (%)] was used to describe the enumeration data. Chi-square test or Fisher’s exact probability was used to compare the differences between the groups. The weighted univariate Cox proportional hazards model was used to screen out confounding factors. The associations of osteopenia and adherence to the MD with all-cause mortality, and then the interaction and moderating effects between the osteopenia and adherence to the MD were explored via univariate and multivariable Cox proportional hazards models. Model 1 was not adjusted and Model 2 adjusted for confounding factors including age, race, education, marital status, PIR, smoking, hypertension, CVDs, previous fracture, glucocorticoid use, anti-osteoporosis therapy, total 25-hydroxyvitamin D, energy, calcium, BMI, and menopausal. Considering that the application of random forest to fill in data with a missing proportion < 20% may result in bias (Additional file 1: Table S1), sensitivity analysis was conducted by comparing the data before and after interpolation to exclude the influence of random forest interpolation on the study results, and no significant difference was observed (Additional file 1: Table S2). Subgroup analysis was conducted in people in different gender, age, BMI groups and those with or without CVDs. Hazards ratio (HR) and 95% confidence interval (CI) were employed for evaluating the associations among adherence to the MD, osteopenia and the risk of all-cause mortality in general population. Missing value interpolation was performed using Python 3.7.4. Sensitivity analysis and difference comparisons were performed by SAS 9.4 (SAS Institute Inc., Cary, NC, USA). The weighted univariate/multivariate Cox proportional hazards model modeling and subgroup analysis were completed by R version 4.2.0 (2022–04-22 ucrt).