Introduction

The use of antidepressant has been steadily increasing, not only in the elderly but also in the young and middle-aged populations [1–3]. Among all antidepressants, selective serotonin reuptake inhibitors (SSRIs) are the most frequently used, because of their favorable adverse effect profile [3, 4]. However, some studies demonstrated that the use of SSRIs is associated with decreased bone mineral density (BMD) and increased risk of fracture [5–11]. A systematic review and meta-analysis of 12 studies showed that SSRI use is associated with an increased risk of fractures (adjusted odds ratio [aOR] 1.69, 95 % confidence interval (CI) 1.51–1.90) [12]. Recently, a meta-analysis of 34 studies further reported a significant increased risk of fractures associated with the SSRI use (risk ratio [RR] 1.39, 95 % CI 1.32–1.47) [13].

A possible mechanism for the increased risk of fracture associated with SSRI is that SSRIs influence the serotonin transporter system in bone tissue [14]. There are several serotonin receptors that have been identified in osteocytes [15–17]. Nevertheless, most studies have focused on SSRI’s impact on bone metabolism. Very few studies have assessed the effect on the risk of fractures of serotonin–norepinephrine reuptake inhibitors (SNRIs), a frequently used class of antidepressants with similar serotonergic properties like SSRIs. Shea et al. examined the effect of SNRIs on biomarkers of bone turnover and suggested that venlafaxine (a commonly used SNRI) was associated with increased bone resorption [18]. Additional studies are therefore needed to assess the impact of this potential effect on the risk of clinically meaningful fractures. Furthermore, almost all the studies used Western subjects, and therefore, the findings cannot be easily generalized to non-Western populations.

In order to gather more clinical information on this important drug safety issue, we conducted a nested case–control study and assessed the association between the timing and the intensity of serotonergic antidepressant use (including SSRIs and SNRIs) and the risk of fracture in an Asian population. In order to enhance the application of our study findings to clinical settings, we included in the study all frequently used individual SSRIs and SNRIs.

Method

Data source

The National Health Insurance Research Database (NHIRD) is population-based claims data of Taiwan’s mandatory National Health Insurance (NHI) program. The NHI program was launched in 1995 and covers over 99 % of Taiwan’s population (approximately 23 million residents). The database provides abundant data including ambulatory care, inpatient care, prescriptions, and medications and has been commonly used for pharmacoepidemiology research in Taiwan [19]. Our study includes the subset of the NHIRD, the Longitudinal Health Insurance Database (LHID), which contains approximately 3 million individuals, randomly sampled from the Registry for Beneficiaries of the NHIRD.

Study cohort

We identified in the LHID all patients who were 20 years old and older and who were given at least three prescriptions of antidepressants (at least one prescription of SSRI or SNRI) between January 1st 2002 and December 31st 2010. The date of the first prescription of the SSRI or SNRI was assigned as the cohort entry date. To ensure a new user design, patients who received antidepressants 2 years prior their cohort entry date were excluded.

Cases and controls

Among our study cohort, we identified patients who were firstly diagnosed with fracture (ICD-9-CM codes 800-829) between 2002 and 2010. Only the first visit for fracture was included in our study. The date of first visit for fracture was defined as the index date. Using incidence-density sampling, cases were matched with four controls by age (±2 year), sex, and cohort entry date (±30 day). The index date of the control is the index date of the corresponding case. All patients were followed from the cohort entry date to the index date, and this period is the observational period.

Exposure to SSRIs and SNRIs

All antidepressant prescriptions received within the observational period by our cases and controls were retrieved from the LHID. In the primary analysis, different types of antidepressants were considered as one group, including SSRIs (fluoxetine, paroxetine, sertraline, citalopram, escitalopram, and fluvoxamine) and SNRIs (venlafaxine, duloxetine, and milnacipran). We assessed the association of timing (current, former, and non-user), intensity of antidepressant use, and the individual component use, with the risk of fracture.

A current user was defined as one whose last day of using antidepressants was within 30 days before the index date. A former user was one whose last day of using prescribed antidepressants was between 31 and 365 days before the index date. A non-user was one with no prescriptions of antidepressants within 365 days before the index date. Pre-planned sensitivity analyses were conducted by varying the definition of current user of SSRI/SNRI as patients whose last day of using SSRI/SNRI was within 60 and 90 days before the index date.

The intensity of exposure to antidepressants was calculated by the medication possession ratio (MPR) during the follow-up period as the total antidepressant prescribing days divided by the total number of days of the follow-up period [20]. Patients were classified into three intensity level based on their MPR: low (MPR <0.2), intermediate (MPR ≥0.2 to MPR <0.8), and high (MPR ≥0.8). Furthermore, we assess the use of individual component of SSRI/SNRI with the risk of fracture.

Co-morbidities and co-medications

As many diseases and drugs other than serotonergic antidepressants may affect the fracture risk [21, 22], we retrieved these information using claims data 1 year before the index date for our cases and controls. The co-morbidities taken into consideration included hypertension, diabetes mellitus, osteoporosis, history of falling, cardiac disorders, chronic obstructive lung disease (COLD), urinary incontinence (UI), Parkinson disease, chronic mental disease (CMD), dementia, depression, liver disease, peripheral vascular disease (PVD), cerebrovascular disease (CVD), arthritis, chronic kidney disease (CKD), and glaucoma. The co-medications consisted of opiate, non-opioid analgesics, antipsychotics, anxiolytics, sedatives, corticosteroid, diuretic, hormone replacement therapy (HRT), antiepileptics, and tricyclic antidepressants (TCAs).

Statistical analysis

McNemar tests were used to compare categorical variables between cases and controls. Multivariate conditional logistic regressions were used to evaluate the associations between the exposure to antidepressants and risk of fracture. All models were adjusted for co-morbidities and co-medications. The associations were presented as adjusted odds ratios (aORs) and 95 % confidence intervals (CIs). The p values were two sided, with p < 0.05 considered statistically significant. All data management and analyses were performed using SAS 9.3 for Windows (SAS Institute, Cary, NC, USA).

Results

We identified 8250 cases and 33,000 matched controls by age, sex, and cohort entry date. Overall, the distribution of age, sex, and cohort entry year in cases and controls were well matched. Cases were more likely to have co-morbid conditions and co-medications than controls (Table 1).

Table 1 Baseline characteristic of the study cohort
Full size table

After adjusting for all co-morbidities and co-medications, only current users were associated with an increased risk of fractures (current; aOR 1.16 [95 % CI 1.07–1.25], p < 0.001). However, higher risks of fracture was found in patients with osteoporosis (aOR 3.05 [2.73–3.42], p < 0.001), history of falling (aOR 6.13 [3.41–11.0], p < 0.001), chronic mental disease (aOR 1.61 [1.30–2.00], p < 0.001), and arthritis (aOR 1.58 [1.43–1.73], p < 0.001). Patients who were exposed to drugs that are potentially associated with fractures within 1 year before the index date also had a higher risk of fracture, especially those exposed to opiate analgesics (aOR 2.20 [1.87–2.60], p < 0.001) (Table 2).

Table 2 Crude and adjusted risks for fractures and exposure to SSRI/SNRIs
Full size table

Among current users, we found a temporal relationship between use of SSRI/SNRI and risk of fracture. The risk of fracture appeared to decrease as the MPR increase (MPR <0.2: aOR 1.19 [1.01–1.41], p = 0.031; 0.2 ≤ MPR < 0.8: aOR 1.15 [1.02–1.29], p = 0.0217; MPR ≥0.8: aOR 1.10 [0.98–1.22], p = 0.893). The risk of fracture was greater for patients who used combination of SSRI and SNRI (with aOR of 1.16 [95 % CI 1.05–1.28], p = 0.0025), but not in patients who used SSRI or SNRI alone (Table 3). Separate ORs were estimated for exposure to each SSRIs or SNRIs. Compared to non-users, there were no differences in the risk of fractures found for individual SSRIs or SNRIs (Table 4).

Table 3 Crude and adjusted risks for fractures and intensity or types of antidepressant; current users of antidepressant
Full size table
Table 4 Adjusted risks for fracture among each SSRI or SNRI
Full size table

We also performed a sensitivity analysis by varying the cut-off for current and past user of SSRI/SNRI. We found that only current users of SSRI/SNRI were associated with increased risks of fracture (cut-off = 60 days, current user: aOR 1.14 [1.06–1.20], p = 0.001, and cut-off = 90 days, current user: aOR 1.12 [1.04–1.21], p = 0.0023) (Table 5).

Table 5 Sensitivity analyses: different cut-off points of current and past users of SSRI/SNRI and risk of fracture
Full size table

Discussion

To our knowledge, this is the first study investigating the association between the use of SSRIs and SNRIs and the risk of fracture using a nested matched case–control design in a nationwide representative cohort of ethnic Chinese. We found that current use of SSRI/SNRIs increased fracture risk by 16 %. Our sensitivity analyses by varying the definition of current users yield similar results. These findings are consistent with several previous studies. Brand et al. found that the risk of hip/femur fractures is increased in current users of antidepressant and declines rapidly after discontinuation of use [9]. Liu et al. and Rabenda et al. also showed that current users of antidepressants are associated with a higher risk of hip fractures and non-vertebral fractures [23, 24]. In addition, the prescription prevalence of antidepressants rose significantly after hip fractures and are known to have a negative effect on bone and on the risk of falling, and finally on re-fracturing [25].

Our study is also the first study to use MPR for exploring the association between intensity of SSRI/SNRI use and risk of fracture. Recently, Zucker et al. used proportion of days covered (PDC) as an indicator of adherence to investigate the association between adherence to SSRI treatment and risk of bone loss-related events [15]. They suggest that a higher adherence to SSRI treatment is significantly associated with an increased risk of bone loss-related events. In contrast, we found the risk of fracture decreased as the MPR increased. The differences can be explained by the following: First, this temporal association may be supported by the study done by Brand et al. They found that the risk of hip and femur fractures is increased in current users of antidepressant and declines rapidly after discontinuation of use [9]. Furthermore, patients may discontinue their SSRI/SNRI due to drug-associated adverse events (such as fracture) and resulting in a lower MPR of drug exposure. However, more researches are warranted to clarify such findings.

Secondly, the confounding factors included for adjustment may further explain this discrepancy. Our study included 17 co-morbidities and 10 co-medications associated with the risk of fracture. However, the study done by Zucker et al. included only 10 associated confounding factors [15]. Our study produced similar results when we only included co-morbidities for adjustments. However, when all the confounding factors are included for adjustments, the relationship of intensity and the outcome become not statistically significant. In addition to timing of SSRI/SNRI use, our findings were consistent with existing evidence that underlying diseases and co-medications are influential on the risk of fracture. For example, Hubbard R. et al. found that history of falling has an important impact on the association between SSRIs and hip fracture [26].

Our results extend the current evidence by evaluating the association between SNRI and risk of fracture. No study has addressed this issue before even SNRIs have similar serotonergic properties like SSRIs. We found that the risks of fracture were comparable in patients who used SSRI or SNRI alone. Furthermore, the risk of fractures was not significantly different among each SSRIs or SNRIs. However, we found that combined use of SSRIs and SNRIs was associated with an increased risk of fracture. Such association could be explained by the severity of depressive symptoms. According to clinical guidelines, patients who do not respond to one antidepressant may use a combination of two different kinds of antidepressants to control their disease [27]. As depression itself has been reported to be associated with increased risks of fall-related events, such as fractures, those who used a combination of SSRI and SNRI may represent a group of patients vulnerable to fracture [28].

Our study has some limitations. First, as this is a claims-based database, we cannot retrieve information not routinely collected in claims database such as smoking, body mass index (BMI), and severity of depressions. In particular, depressions were also suggested to increase the risk of fall-related events [28]. Our study thus includes those who received at least three prescriptions of antidepressants to minimize the variation of underlying clinical situations. Secondly, our study endpoint was all fracture instead of fractures more likely to be osteoporosis related. Therefore, the results may not be specific to SSRI-related bone resorption-induced fractures. The reason that we included all fractures is because there are two possible mechanisms of action regarding SSRI-related fractures. One possible mechanism concerns SSRI/SNRI-mediated osteoporosis events. This pathway takes time to develop the osteoporosis-related fracture risk. Existing evidence suggests that it takes about three or more months [29]. However, there are also some studies that show that the fracture is more likely to happen within the first 15 days after initiation of SSRI treatment [26]. The cause of this early effect on fracture may be associated with the SSRI-related syncope, postural hypotension, or dizziness [9, 30, 31]. Therefore, in order to catch all possible causes of SSRI-related fractures, we did not limit our study endpoint to “osteoporosis-related fractures.” However, we have done a subgroup analysis focusing on the association between the use of serotonergic antidepressant in postmenopausal women (age ≥55) and risk of fragility fracture (including hip, spine, wrist, and femoral fractures) and yielded similar results.

There are certain strengths to our study. To our knowledge, this study is the first to investigate the association between the use of SSRI/SNRI and the risk of fracture using a nested matched case–control study design in a nationwide representative cohort of ethnic Chinese. In an aging society, the problems of osteoporosis-related disabilities create both medical economic burdens seen in many Asian countries which make these studies clinically significant [32, 33]. In particular, a higher risk of fracture associated with current use of serotonergic antidepressant was found in the elderly when stratifying our study cohort into two age groups (aOR; ≥65 years old 1.45 vs. <65 years old 1.11). Secondly, the use of antidepressants is increasing among the younger generation. Therefore, our study population was not limited to the elderly and the results can be adapted to broad-range age groups. Thirdly, to enhance the application of our study results to clinical settings, we examine the fracture risk among individual SSRIs or SNRIs. Finally, we included many clinical-related confounding factors to evaluate the relationship between SSRI/SNRI use and the risk of fracture. By doing so, we have found that several important diseases potentially increase the risk of fractures. These results convey the need for more studies to be done on those high-risk populations.

Conclusion

The current use of SSRI/SNRI is associated with risk of fractures. Furthermore, patients with osteoporosis and history of falling have increased risk of fractures.

Screening of bone mineral density and supplementation with calcium and vitamin D are recommended during the SSRI/SNRI treatment period in high-risk patients.