Background
Osteoporotic fracture is a critical health problem worldwide because it causes severe pain, disability, decreased quality of life, and increased mortality and health costs [
1,
2]. More than one-third of women and approximately one-tenth of men aged 50 years will sustain a major osteoporotic fracture in their remaining lifetime [
3]. Similar to fracture, depression was a common disorder in modern society, with a high prevalence among the general population [
4]. The lifetime incidence of depression is approximately 16% among adults in the United States [
5]. Globally, the total number of individuals with depression was around 300 million in 2015, which is equivalent to 4.4% of the world’s population [
6].
During the past decades, several studies have assessed the association between depression and the risk for fracture. However, the results were conflicting [
7‐
9]. Some studies have reported that depression, was often complicated with decreased bone mineral density (BMD) and bone loss and is associated with a significant increased risk in fracture, but other studies have not found such risk. A previous meta-analysis pooled results from 10 studies published before 2009 as a secondary analysis and showed the association between depression and an increased risk for fracture [
10]. However, this previous study had several limitations. First, the potential sources of heterogeneity were not explored, with limited description on heterogeneity across studies. Second, its subgroup analysis was limited to sex and whether the study controlled for the use of antidepressants and lack of subgroup analyses stratified by other important study and participant characteristics. For example, because the prevalence of depression varied in different regions, investigating the geographical region differences in the depression–fracture association is of interest. In addition, more studies conducted in various countries have been published in recent years, which allowed for a more detailed analysis of the relationship between depression and the risk for fracture. Given the limitations of previous review and the recent publication of numerous large cohort studies, it is necessary for us to assess the effect of depression on the risk for fractures via an updated meta-analysis based on cohort studies.
Methods
Search strategy
This review was conducted in accordance with the Meta-analysis of Observational Studies in Epidemiology (MOOSE) guidelines [
11], with reference to the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) [
12]. A literature search on prospective or retrospective cohort studies showing the association between depression and fracture in Web of Science, PubMed, Embase, China National Knowledge Infrastructure (CNKI) and WanFang database was conducted from inception to February 2018.
The following search terms were used to identify relevant citations: (“depression” [Mesh] or “depressive disorder” [Mesh] or “depressive disorder, major” [Mesh] or “mood disorders” [Mesh]) and (“fractures, bone” [Mesh] or “fracture” or “bone fracture”) along with (“cohort studies” or “longitudinal studies” or “follow-up studies”). In addition, the reference lists of the original and relevant review articles were also assessed to further identify relevant studies. Papers published in English or Chinese were considered.
Selection criteria
Studies were included in the meta-analysis based on the following inclusion criteria: (1) the cohort comprised non-institutionalized adults; (2) the exposure of interest was depression; (3) the outcome was fracture; (4) the risk estimates with the corresponding 95% confidence intervals (CI) of depression related to fracture were reported. Studies were excluded if (1) the study had a case control design or retrospective design; (2) Reviews, letters, commentaries were excluded; (3) Lack of any information that allowed to calculate effect estimates and corresponding 95% CI. For cohorts with several reports, we tried using data from non-overlapping follow-up periods of each report, or publications with the longest follow-up periods were selected.
Data extraction and quality assessment
We extracted the following information from each eligible study: last name of the first author, year of publication, country where the study was performed, number of participants, characteristics of the participants (sex, age range, and mean age), follow-up time, depression measures, fracture type, and covariates adjusted in the multivariable analysis.
Quality assessment was performed according to the Newcastle–Ottawa scale (NOS) [
13], which is a nine-point scale allocating points based on the quality of selection (0–4 points), comparability (0–2 points), and the outcomes of the study participants (0–3 points). In the NOS, poor, fair, and good quality were scored 0–3, 4–6, and 7–9, respectively. Two investigators (L.Q and Q.Y) independently performed the literature search, selected eligible studies and assessed their quality, and extracted data; disagreements or uncertainties were resolved via discussion with an additional investigator (X.X.Y).
Statistical analysis
Hazard ratio (HRs) and their 95% CI were used to quantify the association between depression and fracture, and the reported relative risk (RRs) were considered equivalent to HRs. Any study results stratified by sex and fracture type were considered as independent reports. The heterogeneity of HRs across the studies was evaluated using the Cochran Q test (
P value < 0.10 was considered an indication of statistically significant heterogeneity) and the
I2 statistic (values of 25%, 50%, and 75% representing low, moderate, and high heterogeneity, respectively) [
14,
15]. A fixed-effect model was used if no or low heterogeneity was detected; otherwise, the random-effects model was adopted [
16].
Subgroup and sensitivity analyses were conducted to explore potential heterogeneity across studies, and the differences among subgroups were tested via meta-regression analysis. Publication bias was assessed via visual inspection of the funnel plot and using the Begg [
17] and Egger tests [
18]. The Duval and Tweedie nonparametric trim-and-fill method was used to adjust the potential publication bias [
19]. Data were statistically analyzed with STATA version 11.0 (StataCorp, College Station, Texas, the USA). All statistical tests were two-sided with a 0.05 significance level.
Patient involvement
No Patients were involved in determining the research question or the outcome measures or in developing plans for the design or implementation of the study. In addition, no patients were required to provide advice on the interpretation or writing of the results. There were no plans of disseminating the research results to the study participants or the relevant patient population.
Discussion
Data from 16 cohort studies with 25 independent reports about 414, 686 participants were used, and our meta-analysis showed that depression was associated with a significantly increased risk for fracture. In addition, the associations remained statistically significant in the groups adjusted for fracture type, sex, study region, and other studies and characteristics of the participants; therefore, our results are robust and suggest that depression is prospectively associated with a significantly increases risk of fracture.
Our results showed that the pooled HR was 1.24 (95% CI: 1.14–1.35), which was slightly higher than that of a previous meta-analysis of 10 studies published before 2009 (HR: 1.17; 95% CI: 1.00–1.36) [
10]. The current meta-analysis included 25 independent reports with sample sizes that are 4 times larger, which significantly enhanced statistical power and provided more accurate and comprehensive estimates of the association between depression and the risk for fracture. More importantly, compared with the previous meta-analysis, heterogeneity was thoroughly assessed, and subgroup analyses were conducted.
Our subgroups analyses identified an important finding. That is, the associations between depression and fractures varied between populations when subgroups analyses conducted by continents. The association was stronger in individuals from European countries than those from America and Asia. This may be attributed to the variation in health care systems in different geographic regions, availability of health services and other factors that are currently unknown. Since the studies included in the current meta-analysis were conducted in high-income countries(areas), such as those in Europe, North America, and Taiwan (research from Asian countries included only two studies conducted in Taiwan), these results should be cautiously generalized to developing countries. To apply this finding to a wider population, more studies must be conducted in other populations from Asia, Africa, and South America.
Depression may contribute to fracture via several possible mechanisms. First, depressive disorders are associated with the deregulation of the hypothalamic–pituitary–adrenocortical axis [
33], chronic low-grade inflammatory response [
34,
35], and increased oxidative and nitrosative stress [
36]. These neuroendocrine and inflammatory effects caused by depressive disorders had implications for osteoporosis, which ultimately increased the risk for fracture [
37,
38]. Second, some studies have shown that neuropathological lesions in certain regions of the brain in patients with depressive disorders can influence these patients’ balance, judgment, gait, and coordination, thus increasing the likelihood of falls, which in turn increases the risk for fracture [
21,
39,
40]. Third, depression was associated with other major comorbidities, such as hypertension [
41,
42] and diabetes [
43]; these comorbidities were considered as risk factors for fracture, which were confirmed in two recent meta-analyses [
44,
45]. Lastly, epidemiological findings showed that antidepressants, in particular selective serotonin reuptake inhibitors, may have direct effects on bone metabolism and decreased bone strength [
46,
47], thereby increasing the risk for fracture [
48]. However, the role of antidepressants should be cautiously interpreted because drug use can be a sign of severe depression, and numerous studies lacked information about the dose and duration of drug use.
This meta-analysis has several limitations. First, a moderate level of heterogeneity across studies was observed, which might result from the differences in the characteristics of the participants, sample sizes, depression measures, and statistical adjustments for potential confounders. Although moderate heterogeneities were still observed in some subgroups, the pooled HRs consistently showed positive associations in all subgroups and the prediction interval remained significant. Second, the funnel plot indicated a possible publication bias; however, the trim-and-fill method was used to correct the bias, which did not materially change the positive association. Third, we conducted only limited subgroups analyses because most original studies included in this meta-analysis did not adjust for other confounders such as social-economic status, medical comorbidities or ethnicity; information about these factors was not provided in most original reports. Finally, because the measurement of depression was mainly based on self-reported symptom scales, misclassification of exposure was inevitable, and this might bias the actual association between depression and fracture.