Recently, the medical literature has paid attention to previously unrecognized adverse effects of commonly used medications.1, 2 A possibility for antidepressants to increase the risk of fracture has been debated in recent years.3 Antidepressants are one of the most commonly prescribed drugs in the western world; in 2002, 8.5% of the US population used antidepressants.4 Osteoporotic fractures are a major public health problem worldwide in the elderly, and the annual cost for this type of fracture is high.5, 6 In the United States alone, osteoporosis has affected more than 10 million people,7 and the cost of osteoporotic fractures in 2005 was estimated at $17 billion.8
Depression is a common ailment among the elderly.9 Selective serotonin reuptake inhibitors (SSRIs) are preferentially prescribed as the first-line therapy in elderly people instead of tricyclic antidepressants (TCAs) because they are safer and have low adverse effects.10, 11 TCAs could be related to fracture because of their adverse effects such as sedation or cardiac dysrhythmias leading to falls, but they might not have any intrinsic effect on the bone.12 SSRIs have been gaining recognition for increasing the risk of fractures as reported in previous studies.13–23
A few possible underlying mechanisms point to the biological plausibility of these observations. One explanation is that increased fracture risk is mediated simply by falling.24, 25 Another explanation shows the influence of serotonin on the bone.26 SSRIs may thus, besides the effects related to falls, share effects on bone turnover and bone mineral density (BMD).19
To our knowledge, no meta-analysis of such studies has been conducted to date. We investigated the association between the use of SSRIs and the risk of fracture by performing a meta-analysis.
Materials and Methods
Search strategy and data sources
We searched MEDLINE (PubMed), EMBASE, and the Cochrane Central Register of Controlled Trials (CENTRAL) in the Cochrane Library from their inception to October 20, 2010 without restrictions. We also searched bibliographies of relevant articles to identify any additional studies.
To identify studies, we used the following combinations in search terms: (“selective serotonin reuptake inhibitor” or “SSRI” or “serotonin” or “antidepressant”) and (“fracture” or “osteoporosis” or “osteopenia” or “bone mineral density” or “BMD” or “bone”).
We included studies that published original data relevant to a possible association between the use of SSRIs and incident bone fractures. We included any study that met all of the following criteria: randomized controlled trial, cohort study, or case-control study. We investigated the association between the use of SSRIs and the risk of fracture, and quantified the outcome with adjusted odds ratios (ORs), relative risks (RRs), and corresponding 95% confidence intervals (CIs).
Data extraction and quality assessment
Two investigators (CSE, HKL) independently evaluated the eligibility of all the studies retrieved from the databases on the basis of the predetermined selection criteria. They resolved any disagreements either by discussion or in consultation with the co-corresponding authors (SMP, KHC).
We assessed the methodological quality based on the Newcastle-Ottawa Scale.27 We conducted subgroup analyses according to the methodological quality (low-quality studies versus high-quality studies). Low quality was defined as a Newcastle-Ottawa Scale score <7.0 and high quality as a score ≥7.0 (maximum score 9).
Main and subgroup analyses
We investigated the association between the use of SSRIs and the risk of fracture by using adjusted data for the main analyses. We also performed subgroup analyses by type of study (cohort or case-control study), geographic location (latitude), the number of clinical risk factors used for statistical adjustment (≥4 versus <4), anatomical site of fracture (hip/femur, spine, or wrist/forearm), methodological quality of the study (high versus low), medication dose (high versus usual), exposure duration, age, and sex. The clinical risk factors of osteoporotic fracture were as follows: age, sex, body mass index (weight and height), previous fracture, smoking, alcohol, glucocorticoid, rheumatoid arthritis, bone mineral density, and depression or physical inactivity.28–31
We computed pooled OR and 95% CI from adjusted ORs/RRs and 95% CIs reported in the studies. We assumed that ORs approximated RRs and therefore combined the study estimates regardless of which measure of association was reported because the incidence of the outcomes of interest was sufficiently rare (<5% per year). We combined studies that provided stratum-specific estimates by using the inverse-variance method.
We examined heterogeneity in results across studies by using the Higgins I2 value, which measures the percentage of total variance in the summary estimate resulting from between-study heterogeneity.32
In light of the heterogeneity of study designs and population characteristics, we calculated the summary effect by means of the DerSimonian-Laird method33 for random-effects models.
We used meta-regression to determine whether there is a relation between fracture risk and potential effect modifiers including study design, quality assessment, latitude, and the number of adjusted variables for osteoporotic fracture.
We identified 2950 articles in the initial search, of which we reviewed 95 abstracts and 22 full articles. We included 12 of these 22 articles and one bibliography in this study. As a result, we included seven case-control and five cohort studies in the final analysis (Fig. 1).
Table 1 summarizes the general characteristics of the 12 studies that were included in the analysis.12–18, 20–23, 34 The mean quality score was 6.91 for the studies (maximum score 9) (Table 2).
Table 1. Characteristics of Studies Included in the Final Analysis of SSRIs and Fracture Risk
Anatomical site of fracture
No. of SSRI users/nonusers
SSRI = selective serotonin reuptake inhibitor; NR = not reported; DMARDs = disease-modifying antirheumatic drugs; NSAIDs = nonsteroidal anti-inflammatory drugs; COPD = chronic obstructive lung disease; CVD = cardiovascular disease; BMD = bone mineral density; GPRD = General Practice Research Database.
Note: Studies assessed by the Newcastle-Ottawa Scale, where full score = 9.
Current use of an antidepressant other than SSRI, use in the past 3 months of a benzodiazepine, use in the past 6 months of oral corticosteroids, hormone-replacement therapy, antipsychotics, beta-blockers, opioids, anticonvulsants, drugs for diabetes, more than two dispensings of an NSAID, DMARDs, and metoclopramide, and a history of malignant neoplasms, mental disorders, cerebrovascular diseases, obstructive airway diseases, or inflammatory bowel diseases
Income, region of residence, medications (SSRIs, other monoamine antidepressants, lithium, typical antipsychotics, atypical antipsychotics, benzodiazepines, anticonvulsants, diuretics, anticoagulants, and thyroid hormones), physical conditions (diabetes, ischemic heart disease, myocardial infarction, hypertension, epilepsy, rheumatoid arthritis, solid organ transplant, COPD, and home care use), and mental disorders (substance abuse, depression, schizophrenia, and dementia)
Anxiolytics and sedatives, neuroleptics, antidepressants, Charlson index, ever used any corticosteroid and antiepileptic drugs, alcoholism, working, income, living with someone, prior fracture, manic-depressant states, other psychoses, schizophrenia, eating disorders, ever treated with lithium
Nonspine fracture: age, health status, 1 or more medical conditions, walking for exercise, functional impairment, fall in previous year, cognitive function, weight change, gait speed, inability to rise from chair, and femoral neck bone density
Follow-up visit: 194/7933
Hip fracture: age, functional impairment, cognitive function, weight change, gait speed, inability to rise from chair, and femoral neck bone density
The selected studies were published between 1998 and 2010. Seven articles were case-control studies13–16, 21–23 and five articles were cohort studies.12, 17, 18, 20, 34 Of the 12 studies, eight were high-quality studies12–14, 16, 17, 20, 22, 34 and four were low-quality studies.15, 18, 21, 23
Meta-analyses showed significant positive associations between use of SSRIs and risk of fracture (adjusted OR = 1.69, 95% CI 1.51–1.90, I2 =89.9%).
In subgroup meta-analyses by type of the study design, we observed a significant positive association between the use of SSRIs and risk of fracture for case-control studies (adjusted OR = 1.74, 95% CI 1.51–2.01, I2 = 93.5%) and cohort studies (adjusted OR = 1.59, 95% CI 1.24–2.03, I2 = 50.9%) (Fig. 2).
Subgroup analyses by geographic location (latitude) showed a tendency of increasing risk for fracture with increasing latitude. The highest latitude was in Canada (42°∼83°N, median 62.5°N), where the fracture risk was higher (adjusted OR = 1.92, 95% CI 1.28–2.88, I2 = 93.8%), followed by western Europe (50°∼58°N, median 54°N; adjusted OR = 1.73, 95% CI 1.49–2.02, I2 = 91.2%) and the US (25°∼49°N, median 37°N; adjusted OR = 1.31, 95% CI 1.21–1.42, I2 = 0.0%) (Table 3). However, there was no statistically significant difference because of the widely overlapping CI of these estimates.
Table 3. Subgroup Analyses for Use of SSRIs and Risk of Fracture Using Random-Effects Model
No. of studies
Summary adjusted OR (95% CI)
Heterogeneity, I2 (%)
SSRIs = selective serotonin reuptake inhibitors; OR = odds ratio; CI = confidence interval.
Age, sex, body mass index (weight and height), previous fracture, smoking, alcohol, glucocorticoid, rheumatoid arthritis, bone mineral density, and depression or physical inactivity.
Assessed by the Newcastle-Ottawa Scale, where full score = 9.
Grouping studies by the number of adjustment variables for osteoporotic fracture risk factors revealed a higher increased fracture risk in those adjusting for fewer than four variables (adjusted OR = 1.83, 95% CI 1.57–2.13, I2 = 88.0%) than in those adjusting for four or more variables (adjusted OR = 1.38, 95% CI 1.27–1.49, I2 = 46.1%). The group for fewer than four adjusted numbers had greater heterogeneity than the four or more adjusted-variable group (Table 3).
In subgroup analyses by anatomical site of fracture, we observed significant positive associations between fracture to the hip or femur (adjusted OR = 2.06, 95% CI 1.84–2.30, I2 = 62.3%), spine (adjusted OR = 1.34, 95% CI 1.13–1.59, I2 = 48.5%), and wrist or forearm (adjusted OR = 1.51, 95% CI 1.26–1.82, I2 = 76.6%). Hip or femur showed higher risk of fracture than spine and wrist or forearm (Table 3).
In subgroup analyses by methodological quality, we observed significant positive associations between fractures to the high-quality studies (adjusted OR = 1.75, 95% CI 1.49–2.07, I2 = 80.1%) and low-quality studies (adjusted OR = 1.63, 95% CI 1.36–1.96, I2 = 95.5%). However, there was no statistically significant difference between high- and low-quality studies (Table 3).
Subgroup analyses by dose indicated that high doses of SSRIs showed a significant association with the risk of fracture (adjusted OR = 1.89, 95% CI 1.18–3.05, I2 = 87.4%) and usual doses of SSRIs showed a significant association with the risk of fracture (adjusted OR = 1.65, 95% CI 1.05–2.61, I2 = 67.3%) (Table 3).
Subgroup analyses by exposure duration revealed that the strength of association decreased with a longer window of SSRI administration before the index date. The risk of fracture was greater with SSRIs administered within 6 weeks before the index date (adjusted OR = 3.83, 95% CI 1.96–7.49, I2 = 41.5%) than 6 weeks or more (adjusted OR = 1.60, 95% CI 0.93–2.76, I2 = 63.1%), in which the associations were not statistically significant (Table 3).
Subgroup analyses by age of the study population showed that the pooled effect of risk appear similar for three age groups: age ≥18 years (adjusted OR = 1.84, 95% CI 1.33–2.55, I2 = 95.4%), age ≥50 years (or 55 years) (adjusted OR = 1.56, 95% CI 1.33–1.81, I2 = 83.8%), and age ≥65 years (adjusted OR = 1.71, 95% CI 1.20–2.45, I2 = 90.6%) (Table 3).
Subgroup analyses by sex showed that the risk estimates appear similar for both sexes (women: adjusted OR = 1.68, 95% CI 1.04–2.73, I2 = 93.9%; men: adjusted OR = 1.70, 95% CI 1.56–1.85, I2 = 0.0%) (Table 3).
We found no statistically significant effect on fracture risk in a meta-regression for latitude, quality assessment, study design, and the number of key adjusted variables for osteoporotic fracture risk factors (data not shown).
We found no statistically significant asymmetry in funnel plots, and the result of the Egger test was not significant. The p value for bias in the Egger test was 0.051 (data not shown).
Our results suggest that the use of SSRIs is associated with an increased risk of fracture. Given the widespread use of SSRIs, the implications of this increased risk are serious. If we assume that 36 cases of fracture occur for every 1000 individuals not receiving SSRIs,19 and if we also assume a 1.69-fold increase in the risk of fracture owing to SSRIs, as determined in this study, 60 cases of fracture can be expected for every 1000 recipients of these drugs. This translates to about one case of fracture for every 42 patients treated with SSRIs.
A few lines of evidence point to the biological plausibility of these observations. The increased fracture risk may be explained through cardiovascular effects such as syncope and orthostatic hypotension.35 One explanation is that increased fracture risk is mediated simply by falling.24 The likelihood of the effect, however, is much less than with TCAs.36, 37
Another explanation may be the influence of serotonin on the bone. Recent research reveals a potentially important role for the serotonergic system in bone physiology.26 Functional serotonin receptors and transporter systems have been found in osteoblasts, osteoclasts, and osteocytes in animal models.38–42 Thus, increased serotonergic activity may directly affect bone metabolism, causing excessive bone loss.13, 43, 44 As SSRIs increase systemic serotonin levels, they may contribute to an increased risk for bone fractures through serotonin pathways.26, 41, 45 Recent studies on humans have added to this concern, by showing a reduction in bone mineral density (BMD) and an increased loss rate for BMD.4, 20, 46–48 Regarding fracture risk, both TCAs and SSRIs share a common risk of falls, whereas the SSRIs possess an additional risk of decreasing the BMD by affecting the bone metabolism. Two previous studies have pointed toward a higher risk of fractures with SSRIs than with TCAs.21, 23 SSRIs may thus, besides the effects related to falls, share effects on bone turnover and BMD.19
A high I2 of 89.9% was present for overall fracture risk, and we attempted to investigate the reasons for this heterogeneity by subgroup meta-analysis and meta-regression.
In subgroup analyses by the adjusted number of key risk factors for osteoporotic fracture, the lower adjusted number group had greater heterogeneity than the higher adjusted number group. Therefore, one possible reason for heterogeneity may be a lower adjusted number for key osteoporotic risk factors.
Epidemiologically, incidence of radiographic vertebral fracture is more than the hip and distal forearm fracture in osteoporotic fractures.49 However, in our study, the risk of hip fracture is overwhelmingly higher than other fractures. This finding may be associated with a combined effect of serotonin on the bone and falls. There is an important finding that vertebral (spine) fracture under the relatively less influence of falls has a significant positive association in our study.15 In addition, heterogeneity of vertebral (spine) fracture is lowest.
Interestingly, in this study, current use of SSRIs was associated with a significantly increased risk of fractures. The onset of this effect was already visible within the first 6 weeks of treatment and decreased with longer duration of SSRIs use. The increase in risk of fractures was more pronounced with a shorter rather than an extended duration. This effect with time points to an increased risk of falls shortly after initiation of SSRIs, an effect that decreases with time probably as a result of tachyphylaxis.19 This depletion of susceptible effect has been considered in other pharmacoepidemiological studies of adverse events.50
In subgroup analysis by sex, we observed that the heterogeneity of men was lower than women's. This finding means that women may be affected by the influence of other factors, such as menopause, but men may be more consistent regarding the effect of SSRI on fractures.
One issue for the findings of case-control studies is that the elevated risk of fracture associated with the use of SSRIs suggested the possibility of selection bias from clinicians preferentially prescribing SSRIs rather than TCAs in their frail patients at higher risk of falling.22 In addition to selection bias, an indication bias may also overinflate the odds ratios in a case-control study because depression itself is associated with increased risk for fracture. A previous study has shown an association between depression and decreased bone mineral density.51 In this sense, one of the most persistent critiques is that depression may be confounded with SSRI use. Only a few of the underlying studies attempted to control for depression, and even then statistical controls are a poor substitute for randomized controlled trials. However, the result of the cohort study, as well as in the case-control study, is also elevated risk of fracture to the same degree. Nevertheless, even small increases in these values for common diseases may have important public health implications. Clinicians should therefore carefully consider their decision to prescribe SSRIs for patients at elevated risk for fracture, especially women older than 65 years of age.52, 53
Strengths and limitations
Our meta-analysis has several strengths. First, it is the most comprehensive meta-analysis for the use of SSRIs and risk of fracture to date. Second, it examines the associations in greater detail by stratifying into type of study design, latitude, the number of adjusted variables for osteoporotic fracture risk factors, anatomical site of fracture, methodological quality, age of the study population, and sex, as well as by dose and duration of exposure to use of SSRIs. Despite these strengths, our study has some limitations. First, observational studies are considered to have potential for bias. To compensate for this limitation, we conducted subgroup analyses according to various components. Second, all studies included in the analyses were from Western countries; therefore, our findings cannot be easily generalized to non-Western populations. More research in non-Western populations is required. Finally, we did not have access to individual data on other nutrients that might affect the risk of fracture. However, it is unlikely that other nutrients would have altered the validity or magnitude of the associations between SSRIs and risk of fracture.
Elderly persons are at increased a risk of developing osteoporosis and depression. The SSRIs used to treat depression in this population may increase the risk of subsequent fractures. The relatively high risk of fracture with SSRI use may have a significant clinical impact. These risks must be balanced against the benefits gained by the treatment for depression. Clinicians should carefully consider BMD screening before prescribing SSRIs, especially for patients who are already at a risk for fracture, and proper management for the high-risk population.
All authors state that they have no conflicts of interest.
This work was supported by a National Research Foundation of Korea (NRF) grant funded by the Korean Government (grant number 2011-0005475).
Authors' roles: C-SE was responsible for initial plan, study design, data collection, data extraction, data interpretation, manuscript drafting, statistical analysis, and conducting the study. HKL and SY were responsible for data collection, extraction, and critical revision of the manuscript. SMP and KHC were responsible for data interpretation, manuscript drafting, supervision, and critical revision of the manuscript for important intellectual contents. This article's contents are solely the responsibility of the authors. C-SE is the guarantor for this article and has full responsibility for this study.