Abstract
Background:
Evidence on depressive symptoms and skeletal health is largely cross-sectional and rarely addresses competing risk, dose-response, or potential behavioral pathways. We examined the association between baseline depressive symptoms and incident osteoporosis in older adults, accounting for competing mortality and exploring mediation by physical activity.
Materials and Methods:
We analyzed 4920 adults aged ≥ 60 years from the English Longitudinal Study of Ageing with Wave 1 8-item Center for Epidemiologic Studies Depression Scale (CES-D-8) scores and follow-up osteoporosis ascertainment. Incident physician-diagnosed osteoporosis was the outcome. Cox models estimated hazard ratios (HRs) per 1-point higher CES-D-8 score. Restricted cubic splines assessed dose-response, Fine–Gray models accounted for death as a competing event, and exploratory natural-effects models assessed baseline physical activity as a potential mediator.
Results:
During a median follow-up of 16.0 years, 502 participants developed incident osteoporosis and 1138 deaths occurred. Higher CES-D-8 scores were associated with greater osteoporosis risk in the fully adjusted Cox model (HR, 1.05; 95% confidence interval [CI], 1.01–1.10; P = 0.020), with consistent results in Fine–Gray models (subdistribution HR, 1.05; 95% CI, 1.01–1.10; P = 0.019). The dose–response appeared approximately linear. Exploratory subgroup analyses did not show robust effect modification, and mediation by baseline physical activity was not statistically supported.
Conclusion:
Baseline depressive symptoms were associated with a small increase in incident osteoporosis risk among older adults. The individual-level effect size was modest, and mediation findings should be interpreted as exploratory rather than causal.
Keywords: Competing risks, depressive, English Longitudinal Study of Aging, mediation analysis, osteoporosis
INTRODUCTION
Osteoporosis is a major contributor to morbidity, disability, and health-care costs in aging populations.[1] Recent global assessments attribute a substantial burden of fractures and fracture-related disability to low bone mineral density (LBMD), with wide geographic heterogeneity and a growing impact as populations age.[2] Meta-analytic estimates further indicate a high prevalence of osteoporosis among older adults, particularly women, underscoring the need to clarify upstream, potentially modifiable risk pathways.[3]
Depressive symptoms are common in later life and have been linked to adverse skeletal outcomes.[4] Prospective meta-analysis suggests that depression is associated with increased fracture risk,[5] while updated syntheses report lower bone mineral density among individuals with depression.[6,7] Emerging genetic-epidemiologic evidence also supports a relationship between major depressive disorder and subsequent fractures, strengthening causal plausibility.[8] Several biological and behavioral mechanisms could connect depressive symptoms to bone loss and osteoporotic events. Dysregulation of the hypothalamic–pituitary–adrenal (HPA) axis and inflammatory signaling – features repeatedly observed in depression – may accelerate bone resorption and impair formation.[9] In parallel, depressive symptoms are associated with reduced physical activity, and higher activity levels are linked to lower depression risk; recent work suggests that physical activity may partially mediate associations between depression and skeletal outcomes.[10] These converging lines of evidence motivate an integrated evaluation of both association and pathway.
Evidence to date is heterogeneous and dominated by cross-sectional analyses, which limits temporal interpretation. Even in longitudinal work, death is infrequently handled as a competing event and may distort cumulative incidence in older adults. Few studies have examined the shape of the association or considered behavioral pathways. Leveraging the English Longitudinal Study of Aging (ELSA), we used time-to-event methods to estimate the association between baseline depressive symptoms and incident osteoporosis, incorporated competing-risk approaches, and explored whether physical activity might account for the part of the observed association.
MATERIALS AND METHODS
Study design and data source
We conducted a prospective cohort study using the ELSA, an ongoing, nationally representative survey of community-dwelling adults in England. Wave 1 served as the analytical baseline. For this analysis, we restricted eligibility to participants aged ≥ 60 years at baseline and constructed a de-identified cohort from the publicly available ELSA files. Baseline exclusions were prespecified and applied sequentially: missing baseline 8-item Center for Epidemiologic Studies Depression Scale (CES-D-8) score (n = 269), missing baseline osteoporosis status or prevalent osteoporosis at baseline (n = 359), and missing baseline covariates required for adjustment (n = 602), yielding n = 6016. We then excluded individuals without any osteoporosis ascertainment at follow-up (no osteoporosis measurement recorded at any subsequent wave; n = 1096), resulting in a final analytic cohort of 4920 participants [Figure 1].
Figure 1.
Flow of participants through the study: eligibility, sequential exclusions, and final analytic cohort
Depressive symptoms
Depressive symptoms at baseline were assessed using the CES-D-8, a brief, validated screening instrument that captures the frequency of core affective and somatic symptoms during the past week.[11] CES-D-8 scores range from 0 to 8, with higher scores indicating greater depressive symptom burden. For the primary Cox and Fine–Gray regression analyses, the CES-D-8 score was modeled as a continuous variable on its original 0–8 scale, and effect estimates were reported per 1-point higher CES-D-8 score. The exposure was measured at baseline only. Participants with missing baseline CES-D-8 scores were excluded during cohort assembly.
Osteoporosis
The primary endpoint was incident osteoporosis during the follow-up. In ELSA, physician-diagnosed chronic conditions are ascertained at each wave via interviewer-administered questionnaires. Participants are asked whether a doctor has ever told them they have osteoporosis; responses are recorded as yes/no at the wave level. For this analysis, osteoporosis was treated as a binary status at each wave, and incident events were defined as the first postbaseline report of osteoporosis among participants free of osteoporosis at baseline. Event time was assigned to the interview date of the first positive report; participants without osteoporosis were censored at their last completed interview, and deaths were treated as censoring in the primary Cox models and as a competing event in Fine–Gray analyses. To ensure valid outcome ascertainment, individuals with no osteoporosis information recorded at any follow-up wave were excluded during cohort construction.
Covariates
All covariates were measured at baseline and treated as fixed. Variables and coding were as follows: age (years; modeled continuously in primary analyses and categorized as 60–69, 70–79, and ≥80 for prespecified subgroup analyses); gender (male/female); race (White vs. non-White); education (less than high school; high school or equivalent; some college or above); marital status (married or partnered vs. other marital status); physical activity (inactive, moderate, or vigorous); alcohol consumption (never drinkers vs. ever drinkers); smoking status (never smokers vs. ever smokers); and physician-diagnosed diabetes (yes/no) and cancer (yes/no).
Statistical analysis
Baseline characteristics were summarized as mean ± standard deviation (SD) for continuous variables and counts (percentages) for categorical variables. Follow-up duration was calculated from baseline to the first postbaseline report of osteoporosis, death, or the last wave with available osteoporosis information, whichever occurred first. Total person-years were calculated by summing individual follow-up time, and the incidence rate of osteoporosis was expressed per 1000 person-years. Time-to-event associations between the baseline CES-D-8 score and incident osteoporosis were estimated using Cox proportional hazards models with time since baseline as the time scale. Covariate adjustment followed a prespecified hierarchy: Model 1 adjusted for age, gender, and race; Model 2 additionally adjusted for education and marital status; Model 3 additionally adjusted for physical activity, drinking, smoking, diabetes, and cancer. Prespecified subgroup analyses were performed by restricting the sample to strata of interest (age 60–69, 70–79, ≥80; sex; and other a priori strata) and re-fitting Model 3 within each stratum. These analyses were considered exploratory. Formal interaction tests were performed by adding CES-D-8 score × subgroup interaction terms to the fully adjusted Cox model, with additional consideration of multiple subgroup testing. To assess functional form, we fitted restricted cubic splines for the CES-D-8 score within Model 3 (3–5 knots at empirical percentiles; primary specification: 5th, 35th, 65th, and 95th), plotted adjusted hazard ratios (HRs) with 95% confidence bands across the observed CES-D-8 score range, and tested overall nonlinearity using a joint Wald test. As death may preclude the occurrence of osteoporosis, we applied competing-risk methods. For graphical cumulative incidence analyses, CES-D-8 score was categorized as 0–3 versus 4–8, with scores ≥ 4 representing elevated depressive symptom burden based on prior ELSA studies.[12,13] This categorization was used only for cumulative incidence visualization and Gray test comparison. The primary Cox and Fine–Gray regression models analyzed CES-D-8 score as a continuous variable on its original 0–8 scale. Fine–Gray subdistribution hazard models were fitted to obtain subdistribution HRs (sHRs) with 95% confidence intervals (CIs), treating death as the competing event. To examine the hypothesized behavioral pathway, we conducted exploratory mediation analysis using CMAverse with baseline physical activity as the mediator. Natural-effect components were estimated for a CES-D-8 score contrast of 6 versus 1, including the total effect (TE), natural direct effect (NDE), natural indirect effect (NIE), and proportion mediated (PM), with 95% CIs via nonparametric bootstrap. These analyses were considered exploratory and were not intended to establish causal mediation. Statistical significance was defined as two-sided P < 0.05. All analyses were conducted using R, version 4.4.0 (R Foundation for Statistical Computing, Vienna, Austria).
RESULTS
Participant characteristics at baseline by incident osteoporosis status
The analytic cohort comprised 4920 participants at baseline (mean age 70.36 years, SD 7.36; 51.9% women), of whom 502 individuals experienced incident osteoporosis during the follow-up [Table 1]. During a median follow-up of 16.0 years (interquartile range, 10.0–16.0; range, 2.0–16.0), the cohort contributed 64,396.0 person-years of observation, corresponding to an incidence rate of 7.80 osteoporosis cases per 1000 person-years. Overall, 1138 deaths occurred during follow-up and were treated as competing events in the Fine–Gray analyses [Supplementary Table S1]. Compared with those who remained free of osteoporosis, participants who developed osteoporosis were slightly older at baseline (70.89 vs. 70.30 years) and were disproportionately female (82.7% vs. 48.4%) [both differences reflected in Table 1]. They were also less often married or partnered (55.4% vs. 66.2%), more frequently never-smokers (42.8% vs. 34.4%), and more likely to report inactive physical activity (46.8% vs. 40.7%). Baseline depressive symptoms were higher among those who later developed osteoporosis (mean CES-D-8 score, 1.90 vs. 1.49). Distributions of race, education, diabetes, and cancer were broadly similar between groups.
Table 1.
Baseline characteristics of the English Longitudinal Study of Aging cohort at wave 1, overall and stratified by incident osteoporosis during follow-up
| Characteristics | Overall (n=4920), n (%) | Osteoporosis at follow-up | P | |
|---|---|---|---|---|
|
| ||||
| No (n=4418), n (%) | Yes (n=502), n (%) | |||
| Age (years), mean±SD | 70.36±7.36 | 70.30±7.37 | 70.89±7.29 | 0.050 |
| Gender | ||||
| Female | 2553 (51.89) | 2138 (48.39) | 415 (82.67) | <0.001 |
| Male | 2367 (48.11) | 2280 (51.61) | 87 (17.33) | |
| Race | ||||
| Nonwhite | 91 (1.85) | 85 (1.92) | 6 (1.20) | 0.3 |
| White | 4829 (98.15) | 4333 (98.08) | 496 (98.80) | |
| Education | ||||
| Less than high school | 2885 (58.64) | 2572 (58.22) | 313 (62.35) | 0.2 |
| High school or equivalent | 734 (14.92) | 663 (15.01) | 71 (14.14) | |
| Some college or above | 1301 (26.44) | 1183 (26.78) | 118 (23.51) | |
| Marital status | ||||
| Married or partnered | 3201 (65.06) | 2923 (66.16) | 278 (55.38) | <0.001 |
| Other marital status | 1719 (34.94) | 1495 (33.84) | 224 (44.62) | |
| Smoking | ||||
| Never smokers | 1735 (35.26) | 1520 (34.40) | 215 (42.83) | <0.001 |
| Ever smokers | 3185 (64.74) | 2898 (65.60) | 287 (57.17) | |
| Drinking | ||||
| Never drinkers | 614 (12.48) | 544 (12.31) | 70 (13.94) | 0.3 |
| Ever drinkers | 4306 (87.52) | 3874 (87.69) | 432 (86.06) | |
| Physical activity | ||||
| Inactive | 2031 (41.28) | 1796 (40.65) | 235 (46.81) | 0.029 |
| Moderate | 2121 (43.11) | 1925 (43.57) | 196 (39.04) | |
| Vigorous | 768 (15.61) | 697 (15.78) | 71 (14.14) | |
| Diabetes | ||||
| No | 4508 (91.63) | 4041 (91.47) | 467 (93.03) | 0.2 |
| Yes | 412 (8.37) | 377 (8.53) | 35 (6.97) | |
| Cancer | ||||
| No | 4601 (93.52) | 4136 (93.62) | 465 (92.63) | 0.4 |
| Yes | 319 (6.48) | 282 (6.38) | 37 (7.37) | |
| CES-D-8 score, mean±SD | 1.53±1.90 | 1.49±1.88 | 1.90±2.06 | <0.001 |
SD=Standard deviation; CES-D-8=8-item Center for Epidemiological Studies Depression Scale
Supplementary Table 1.
Follow-up duration, person-years, osteoporosis events, deaths, and incidence rates
| Characteristics | Overall (n=4920) | CES-D-8 score category | |
|---|---|---|---|
|
| |||
| 0–3 (n=4152) | 4–8 (n=768) | ||
| Participants, n | 4920 | 4152 | 768 |
| Follow-up duration (years), median (IQR) | 16.0 (10.0–16.0) | 16.0 (10.0–16.0) | 16.0 (8.0–16.0) |
| Follow-up duration (years), range | 2.0–16.0 | 2.0–16.0 | 2.0–16.0 |
| Total person-years | 64396.0 | 54930.0 | 9466.0 |
| Incident osteoporosis, n | 502 | 402 | 100 |
| Deaths during follow-up, n | 1138 | 926 | 212 |
| Incidence rate of osteoporosis, per 1000 person-years | 7.80 | 7.32 | 10.56 |
CES-D-8=8-item Center for Epidemiologic Studies Depression Scale; IQR=Interquartile range
Association between depressive symptoms and incident osteoporosis
In the primary time-to-event analyses, higher baseline depressive symptoms were associated with an increased hazard of incident osteoporosis. For each 1-point increase in the CES-D-8 score, the adjusted HR was 1.06 (95% CI, 1.01–1.10; P = 0.009) in Model 1, 1.06 (95% CI, 1.01–1.10; P = 0.014) in Model 2, and 1.05 (95% CI, 1.01–1.10; P = 0.020) in Model 3 [Table 2]. The proportional hazards assumption was satisfied for all three models, with no evidence of violation for either the CES-D-8 score or the global tests (all P > 0.05) [Supplementary Table S2].
Table 2.
Association between baseline depressive symptoms and incident osteoporosis
| Model | HR (95% CI) | P |
|---|---|---|
| M1 | 1.06 (1.01–1.10) | 0.009 |
| M2 | 1.06 (1.01–1.10) | 0.014 |
| M3 | 1.05 (1.01–1.10) | 0.020 |
CI=Confidence interval; HR=Hazard ratio
Supplementary Table 2.
Proportional hazards assumption tests based on Schoenfeld residuals
| Model | CES-D-8 score test | Global test | ||
|---|---|---|---|---|
|
|
|
|||
| χ 2 | P | χ 2 | P | |
| Model 1 | 0.68 | 0.410 | 9.32 | 0.592 |
| Model 2 | 0.74 | 0.389 | 11.85 | 0.457 |
| Model 3 | 0.91 | 0.340 | 14.26 | 0.356 |
CES-D-8=8-item Center for Epidemiologic Studies Depression Scale
Prespecified exploratory subgroup analyses showed broadly similar patterns when refitting the fully adjusted model within each stratum [Figure 2]. Point estimates were positive in most strata, with statistically significant associations in women (HR 1.06, 95% CI 1.01–1.11; P = 0.026) and in participants aged 60–69 years (HR 1.07, 95% CI 1.00–1.14; P = 0.036). Estimates in men and in older age groups were imprecise and not statistically significant. Among racial groups, the estimate for White participants (HR 1.05, 95% CI 1.01–1.10; P = 0.025) was statistically significant, whereas the non-White stratum showed a larger but imprecise estimate (HR 1.74, 95% CI 0.93–3.26; P = 0.082), consistent with smaller numbers. Patterns across education, marital status, lifestyle, and comorbidity strata were largely concordant with the overall result. Formal interaction tests did not provide robust evidence that the association between CES-D-8 score and incident osteoporosis differed across prespecified subgroups. A nominal interaction was observed for marital status (P for interaction = 0.021), but this did not remain evident after false discovery rate adjustment (FDR =0.215); all other interaction tests were non-significant [Supplementary Table S3]. Therefore, subgroup-specific findings should be interpreted cautiously as exploratory and hypothesis-generating rather than confirmatory evidence of effect modification.
Figure 2.
Exploratory subgroup analysis of the association between baseline 8-item Center for Epidemiologic Studies Depression Scale (CES-D-8) and incident osteoporosis. Fully adjusted Cox models were re-fit within each prespecified stratum. Points denote hazard ratios (HRs) per 1-point higher CES-D-8 score; horizontal bars show 95% confidence intervals; P values are from Wald tests within strata. Because multiple subgroup analyses were performed, subgroup-specific findings should be interpreted cautiously. The vertical dashed line marks HR = 1.0
Supplementary Table 3.
Interaction tests for prespecified subgroup analyses
| Subgroup variable | Interaction χ2 | P for interaction | FDR |
|---|---|---|---|
| Age | 1.07 | 0.584 | 0.848 |
| Gender | 0.32 | 0.573 | 0.848 |
| Race | 0.04 | 0.847 | 0.848 |
| Education | 1.01 | 0.603 | 0.848 |
| Marital status | 5.29 | 0.021 | 0.215 |
| Physical activity | 1.71 | 0.426 | 0.848 |
| Drinking | 0.09 | 0.759 | 0.848 |
| Smoking | 0.04 | 0.848 | 0.848 |
| Diabetes | 0.17 | 0.676 | 0.848 |
| Cancer | 0.05 | 0.821 | 0.848 |
FDR=False discovery rate
Restricted cubic spline analysis supported an approximately monotonic increase in osteoporosis risk with higher CES-D-8 score [Figure 3]. The overall association was statistically significant (P = 0.035) with no evidence of nonlinearity (P for nonlinearity = 0.216). Confidence bands widened at distributional extremes, reflecting sparse data at very low and very high CES-D-8 scores.
Figure 3.
Restricted cubic spline for baseline 8-item Center for Epidemiologic Studies Depression Scale (CES-D-8) score and hazard of incident osteoporosis. Spline terms for the CES-D-8 score were fit within the fully adjusted Cox model. The solid curve shows the adjusted hazard ratios (HR) across the observed CES-D-8 score range; the shaded area indicates the 95% confidence interval. The dashed horizontal line marks HR = 1.0
Competing-risk analyses
Accounting for death as a competing event, cumulative incidence functions for osteoporosis diverged by baseline depressive symptoms (CES-D-8 score categories 0–3 vs. 4–8), with a higher cumulative incidence in the higher-symptom group; the Gray test was significant (P = 0.004) [Figure 4]. In Fine-Gray subdistribution hazard models, each 1-point higher CES-D-8 score was associated with a higher subdistribution hazard of osteoporosis: sHR, 1.06 (95% CI, 1.01–1.10; P = 0.009) in Model 1; sHR, 1.06 (95% CI, 1.01–1.10; P = 0.013) in Model 2; and sHR, 1.05 (95% CI, 1.01–1.10; P = 0.019) in Model 3 [Table 3]. These estimates are directionally consistent with the cause-specific Cox results, indicating that higher depressive symptom burden is associated with a greater cumulative risk of osteoporosis over follow-up when accounting for the competing risk of death.
Figure 4.
Cumulative incidence of osteoporosis and death by baseline 8-item Center for Epidemiologic Studies Depression Scale (CES-D-8) score. Cumulative incidence functions (CIFs) for osteoporosis and death are shown by baseline CES-D-8 score category. CES-D-8 scores were categorized as 0–3 versus 4–8 for visualization, with scores ≥ 4 representing elevated depressive symptom burden. The Gray test compares osteoporosis CIFs between CES-D-8 score categories. Numbers at risk and cumulative events by category are displayed beneath the plot
Table 3.
Association of depressive symptoms with incident osteoporosis in competing-risk models
| Model | sHR (95% CI) | P |
|---|---|---|
| M1 | 1.06 (1.01–1.10) | 0.009 |
| M2 | 1.06 (1.01–1.10) | 0.013 |
| M3 | 1.05 (1.01–1.10) | 0.019 |
CI=Confidence interval; sHR=Subdistribution hazard ratio
Mediation analysis
In exploratory natural-effects decomposition for a CES-D-8 score contrast of 6 versus 1, with baseline physical activity specified as the mediator, the TE on incident osteoporosis was HR, 1.38 (95% CI, 1.09–1.73; P = 0.008). The pure NDE was HR, 1.30 (95% CI, 1.02–1.62; P = 0.024), and the total NDE was HR, 1.34 (95% CI, 1.06–1.68; P = 0.010). In contrast, the pure and total NIEs were not statistically significant: the pure natural indirect effect (PNIE) was HR, 1.03 (95% CI, 0.98–1.08; P = 0.194), and the TNIE was HR, 1.07 (95% CI, 0.97–1.16; P = 0.196). The controlled direct effect at the prespecified mediator level was HR, 1.38 (95% CI, 1.00–1.83; P = 0.042). The PM was 22% but was imprecisely estimated, with a wide CI crossing zero (95% CI, −17%–72%; P = 0.192) [Table 4]. Therefore, the exploratory mediation analysis did not provide statistically supported evidence that baseline physical activity mediated the association between the CES-D-8 score and incident osteoporosis. These estimates should not be interpreted as causal evidence of mediation. Effect estimates are shown in Figure 5a, the PM estimate in Figure 5b, and the conceptual pathway diagram in Figure 5c.
Table 4.
Mediation by physical activity in the association between baseline depressive symptoms and incident osteoporosis
| Effect | HR (95% CI) | P |
|---|---|---|
| CDE | 1.38 (1.00–1.83) | 0.042 |
| PNDE | 1.30 (1.02–1.62) | 0.024 |
| TNDE | 1.34 (1.06–1.68) | 0.010 |
| PNIE | 1.03 (0.98–1.08) | 0.194 |
| TNIE | 1.07 (0.97–1.16) | 0.196 |
| TE | 1.38 (1.09–1.73) | 0.008 |
| PM (%) | 22 (−17–72) | 0.192 |
CDE=Controlled direct effect; PNDE=Pure natural direct effect; TNDE=Total natural direct effect; PNIE=Pure natural indirect effect; TNIE=Total natural indirect effect; TE=Total effect; PM=Proportion mediated; CI=Confidence interval; HR=Hazard ratio
Figure 5.
Exploratory mediation analysis of physical activity: natural effect components, proportion mediated, and pathway diagram. (a) Hazard ratios (log scale) with 95% confidence intervals for the total effect, pure/total natural direct effects (pure natural direct effect, total natural direct effect), and pure/total natural indirect effects (pure natural indirect effect [PNIE], total natural indirect effect [TNIE]), estimated for a 8-item Center for Epidemiologic Studies Depression Scale (CES-D-8) score contrast of 6 versus 1 using the fully adjusted time-to-event specification. (b) Proportion mediated (PM) with 95% confidence interval obtained by nonparametric bootstrap. (c) Conceptual path diagram for the CES-D-8 score → physical activity → osteoporosis pathway annotated with effect estimates. These mediation analyses were exploratory and should not be interpreted as causal evidence of mediation; lack of statistical significance for PNIE/TNIE and PM indicates no statistically supported mediation by baseline physical activity
DISCUSSION
In this nationally representative aging cohort, higher baseline depressive symptoms were associated with a modestly increased hazard of incident osteoporosis, with broadly consistent estimates across hierarchical adjustment and robustness to competing-risk methods. The dose–response appeared approximately linear. In exploratory mediation analysis, baseline physical activity did not show statistically supported mediation of the association between depressive symptoms and incident osteoporosis.
Our results are broadly consistent with evidence that depressive phenotypes co-occur with poorer skeletal metrics. A 26-study meta-analysis reported lower BMD among individuals with depression at the spine, total hip, and femoral neck, with more pronounced deficits in adults and women; complementary estimates suggested ≈ 5.1%, 4.1%, and 9.1% lower BMD at the spine, total femur, and femoral neck, respectively.[14] In clinical cohorts, depressive symptoms have been linked not only to lower BMD but also to adverse bone-turnover profiles and worse postoperative recovery after hip fracture in elderly men,[15] reinforcing biologic and functional correlates relevant to osteoporosis risk. Cross-sectional analyses in U.S. women further indicate that antidepressant exposure – across several classes – is associated with higher odds of osteoporosis (≈44%) and fractures (≈62%), with risk rising with longer duration of use, although such designs remain vulnerable to confounding by indication and temporal ambiguity.[16] Against this observational backdrop, Mendelian randomization using large European GWAS datasets found no causal effect of genetic liability to depression on BMD at forearm, femoral neck, lumbar spine, or heel, nor on fracture, highlighting potential confounding in conventional studies.[17] By contrast, a recent analysis in perimenopausal women combined NHANES data with MR and reported positive observational associations alongside MR signals suggestive of a causal link between major depressive disorder and osteoporosis, underscoring heterogeneity by sex, menopausal status, and ancestry and pointing to possible population-specific genetic architectures.[18]
Multiple, partially overlapping pathways plausibly connect depressive symptoms with adverse skeletal outcomes. Chronic stress-related HPA axis dysregulation – a well-described feature of depressive disorders – elevates glucocorticoid exposure that suppresses osteoblastogenesis, promotes osteoblast/osteocyte apoptosis, and prolongs osteoclast survival, shifting remodeling toward net resorption.[9,19] Contemporary reviews of glucocorticoid-induced osteoporosis and neuroendocrine–bone crosstalk underscore these cell-specific effects and their clinical relevance in ageing populations.[20] In parallel, low-grade inflammation – characteristic of many depressive states – amplifies osteoclastogenesis via RANK–RANKL signaling and nuclear factor kappa B activation, providing a mechanistic bridge from affective dysregulation to bone resorption.[21] Sleep disruption, common in late-life depression, is also linked to lower BMD and higher fracture risk, potentially through circadian, hormonal, sympathetic, and inflammatory pathways.[22]
Regarding behavioral pathways, depression is frequently accompanied by reduced physical activity and skeletal loading, which diminishes osteocytic mechanotransduction, increases sclerostin, suppresses Wnt/β-catenin signaling, and attenuates osteogenesis – features characteristic of disuse-related bone loss.[23] However, in our exploratory natural-effects decomposition, baseline physical activity did not show statistically supported mediation of the association between CES-D-8 score and incident osteoporosis. The indirect effect estimates were not statistically significant (PNIE 1.03 and TNIE 1.07), and the PM estimate was imprecise, with a wide CI crossing zero. Therefore, the PM point estimate should not be interpreted as evidence that physical activity causally explained a fixed proportion of the association. These results may reflect baseline-only assessment, coarse physical activity categorization, exposure-mediator temporal misalignment, limited precision, and residual confounding. Beyond behavior, serotonergic signaling offers a biologically plausible pharmacologic route: osteoblasts and osteoclasts express serotonin transporters/receptors, and experimental SSRI exposure impairs bone accrual.[24] Meta-analytic clinical data in 2024 associate SSRI use with lower BMD and greater fracture risk, though confounding by indication and falls remains a consideration.[25] These axes – neuroendocrine, immune–bone, mechanotransductive, circadian–autonomic, and serotonergic – likely operate concurrently rather than in isolation, with their relative contributions varying by age, sex, medication exposure, and comorbidity burden. Conceptual models of the brain–bone axis synthesize these pathways and support a systems-level link between affective dysregulation and skeletal remodeling.[26]
Taken together, our results indicate that greater depressive symptom burden is associated with a modest elevation in the risk of incident osteoporosis in later life. The magnitude of association was small at the individual level: an HR of approximately 1.05 per 1-point higher CES-D-8 score should not be interpreted as having strong stand-alone clinical predictive value. Rather, depressive symptom burden may serve as a complementary risk marker within broader osteoporosis risk assessment, alongside established determinants such as age, sex, prior fracture, bone mineral density, falls risk, comorbidities, medication use, nutrition, and physical function. Because depressive symptoms are common in older adults and CES-D-8 scores span a range from 0 to 8, the association may still be relevant from a population-health perspective, but its individual-level clinical significance is limited. In practice, this supports coordinating mental-health assessment with standard osteoporosis prevention strategies, without supplanting guideline-directed case finding, fall prevention, nutrition, or pharmacotherapy.
Several limitations temper inference. Osteoporosis was self-reported rather than densitometry-confirmed and was ascertained at discrete survey waves. Therefore, the exact date of osteoporosis onset was unknown, and event time was approximated by the interview date of the first positive report. The true onset may have occurred between the previous negative assessment and the first positive report, introducing interval-censoring-related timing uncertainty that may affect time-to-event estimates and cumulative incidence curves. Depressive symptoms and physical activity were measured only at baseline, which may have introduced exposure and mediator misclassification and limited assessment of temporal ordering. Depressive symptoms can fluctuate over time, and a single baseline CES-D-8 score assessment cannot capture persistent, recurrent, remitted, or newly developed depressive symptoms during follow-up. If such exposure misclassification was largely nondifferential, the estimated association may have been attenuated; however, differential changes related to comorbidity, functional decline, or subsequent skeletal risk cannot be excluded. The exploratory mediation analysis relied on strong assumptions, including no unmeasured exposure-outcome, exposure-mediator, or mediator-outcome confounding; these assumptions are unlikely to be fully satisfied in an observational cohort. Therefore, the mediation results, particularly the PM, should not be interpreted as causal evidence. Residual confounding by bone-active medications, falls, inflammatory or endocrine factors cannot be excluded. Complete-case construction and small numbers in some strata reduce precision and may limit generalizability, especially beyond predominantly White populations. Future work should pair repeated assessments of depressive symptoms with device-based physical-activity measures, adjudicated skeletal endpoints, and richer confounder data. Time-updated causal methods – such as marginal structural models or target-trial emulation – and interventional or quasi-experimental designs that treat depressive symptoms or increase activity are needed to test whether modifying upstream factors yields measurable skeletal benefit and to inform risk-prediction models that incorporate mental-health indicators.
CONCLUSION
In a nationally representative ageing cohort, higher baseline CES-D-8 scores were associated with a small, approximately linear increase in the risk of incident self-reported osteoporosis, with directionally consistent estimates after accounting for competing mortality. The individual-level effect size was modest, and exploratory mediation analyses did not provide statistically supported or causal evidence that baseline physical activity mediated this association. These findings suggest that depressive symptom burden may serve as a complementary marker within broader osteoporosis risk assessment, while underscoring the need for studies with repeated depressive-symptom assessments, adjudicated skeletal endpoints, and time-updated or interventional designs to clarify whether modifying depressive symptoms or related pathways can reduce skeletal risk.
Conflicts of interest
There are no conflicts of interest.
Funding Statement
Nil.
REFERENCES
- 1.Ayers C, Kansagara D, Lazur B, Fu R, Kwon A, Harrod C. Effectiveness and safety of treatments to prevent fractures in people with low bone mass or primary osteoporosis: A living systematic review and network meta-analysis for the American college of physicians. Ann Intern Med. 2023;176:182–95. doi: 10.7326/M22-0684. [DOI] [PubMed] [Google Scholar]
- 2.Hsieh E, Bryazka D, Ong KL, Rhinehart PA, Cousin E, Hagins H, et al. The global, regional, and national burden attributable to low bone mineral density, 1990-2020: An analysis of a modifiable risk factor from the global burden of disease study 2021. Lancet Rheumatol. 2025;7:e873–94. doi: 10.1016/S2665-9913(25)00105-5. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Salari N, Darvishi N, Bartina Y, Larti M, Kiaei A, Hemmati M, et al. Global prevalence of osteoporosis among the world older adults: A comprehensive systematic review and meta-analysis. J Orthop Surg Res. 2021;16:669. doi: 10.1186/s13018-021-02821-8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Ye C, Chen G, Huang W, Liu Y. Association between skeletal muscle mass to visceral fat area ratio and depression: A cross-sectional study based on the National Health and Nutrition Examination Survey. J Affect Disord. 2025;372:314–23. doi: 10.1016/j.jad.2024.12.041. [DOI] [PubMed] [Google Scholar]
- 5.Qiu L, Yang Q, Sun N, Li D, Zhao Y, Li X, et al. Association between depression and the risk for fracture: A meta-analysis and systematic review. BMC Psychiatry. 2018;18:336. doi: 10.1186/s12888-018-1909-2. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Lee DY, Yoo DK, Han SY, Lee K, Lee Y, Teopiz KM, et al. Association between depressive symptoms and bone density in elderly patients with non-dialysis dependent chronic kidney disease. J Affect Disord. 2022;319:549–54. doi: 10.1016/j.jad.2022.09.014. [DOI] [PubMed] [Google Scholar]
- 7.Jamshaid M, Heidari A, Hassan A, Mital D, Pearce O, Panourgia M, et al. Bone loss and fractures in post-menopausal women living with HIV: A narrative review. Pathogens. 2024;13:811. doi: 10.3390/pathogens13090811. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Zhao P, Ying Z, Yuan C, Zhang H, Dong A, Tao J, et al. Shared genetic architecture highlights the bidirectional association between major depressive disorder and fracture risk. Gen Psychiatr. 2024;37:e101418. doi: 10.1136/gpsych-2023-101418. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Cui L, Li S, Wang S, Wu X, Liu Y, Yu W, et al. Major depressive disorder: Hypothesis, mechanism, prevention and treatment. Signal Transduct Target Ther. 2024;9:30. doi: 10.1038/s41392-024-01738-y. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Pearce M, Garcia L, Abbas A, Strain T, Schuch FB, Golubic R, et al. Association between physical activity and risk of depression: A systematic review and meta-analysis. JAMA Psychiatry. 2022;79:550–9. doi: 10.1001/jamapsychiatry.2022.0609. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Schlechter P, Ford TJ, Neufeld SA. The eight-item center for epidemiological studies depression scale in the English longitudinal study of aging: Longitudinal and gender invariance, sum score models, and external associations. Assessment. 2023;30:2146–61. doi: 10.1177/10731911221138930. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Zwick L, Schmitz N, Shojaa M. Oral health-related quality of life and depressive symptoms in adults: Longitudinal associations of the English longitudinal study of ageing (ELSA) BMC Oral Health. 2023;23:1029. doi: 10.1186/s12903-023-03722-4. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Beach B, Steptoe A, Zaninotto P. Depression and anxiety in people with cognitive impairment and dementia during the COVID-19 pandemic: Analysis of the English longitudinal study of ageing. PLoS Med. 2023;20:e1004162. doi: 10.1371/journal.pmed.1004162. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Yuan S, Chen J, Zeng L, Zhou C, Yu S, Fang L. Association of bone mineral density and depression in different bone sites and ages: A meta-analysis. Food Sci Nutr. 2021;9:4780–92. doi: 10.1002/fsn3.2379. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Liu C, Shi L, He J, Wang L, Min N, Wang F, et al. Effects of depressive symptoms on osteoporotic bone metabolism and prognosis of joint replacement surgery in elderly male patients with femoral neck fracture. Asian J Surg. 2023;46:3687–92. doi: 10.1016/j.asjsur.2023.06.020. [DOI] [PubMed] [Google Scholar]
- 16.Rajha HE, Abdelaal R, Charfi K, Alemadi AO, Al-Sheraim AS, Al-Maadid MA, et al. Examining depression, antidepressants use, and class and their potential associations with osteoporosis and fractures in adult women: Results from ten NHANES cohorts. J Affect Disord. 2025;369:1223–32. doi: 10.1016/j.jad.2024.10.114. [DOI] [PubMed] [Google Scholar]
- 17.He B, Lyu Q, Yin L, Zhang M, Quan Z, Ou Y. Depression and osteoporosis: A Mendelian randomization study. Calcif Tissue Int. 2021;109:675–84. doi: 10.1007/s00223-021-00886-5. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Guo X, She Y, Liu Q, Qin J, Wang L, Xu A, et al. Osteoporosis and depression in perimenopausal women: From clinical association to genetic causality. J Affect Disord. 2024;356:371–8. doi: 10.1016/j.jad.2024.04.019. [DOI] [PubMed] [Google Scholar]
- 19.Menke A. The HPA axis as target for depression. Curr Neuropharmacol. 2024;22:904–15. doi: 10.2174/1570159X21666230811141557. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Chen M, Fu W, Xu H, Liu CJ. Pathogenic mechanisms of glucocorticoid-induced osteoporosis. Cytokine Growth Factor Rev. 2023;70:54–66. doi: 10.1016/j.cytogfr.2023.03.002. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Epsley S, Tadros S, Farid A, Kargilis D, Mehta S, Rajapakse CS. The effect of inflammation on bone. Front Physiol. 2020;11:511799. doi: 10.3389/fphys.2020.511799. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Swanson C. Sleep disruption and bone health. Curr Osteoporos Rep. 2022;20:202–12. doi: 10.1007/s11914-022-00733-y. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Qin L, Liu W, Cao H, Xiao G. Molecular mechanosensors in osteocytes. Bone Res. 2020;8:23. doi: 10.1038/s41413-020-0099-y. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Durham E, Zhang Y, LaRue A, Bradshaw A, Cray J. Selective serotonin reuptake inhibitors (SSRI) affect murine bone lineage cells. Life Sci. 2020;255:117827. doi: 10.1016/j.lfs.2020.117827. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Hu J, Xie K, Wu S, Chen Y. Osteoporosis and fracture risk associated with novel antidepressants: A systematic review and meta-analysis. Actas Esp Psiquiatr. 2024;52:334–46. doi: 10.62641/aep.v52i3.1560. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Shi H, Chen M. The brain-bone axis: Unraveling the complex interplay between the central nervous system and skeletal metabolism. Eur J Med Res. 2024;29:317. doi: 10.1186/s40001-024-01918-0. [DOI] [PMC free article] [PubMed] [Google Scholar]