Effects of selective serotonin reuptake inhibitors and other antidepressant drugs on the risk of hip fracture: a case–control study in an elderly Mediterranean population

Abstract

Objectives

To describe the association between exposure to different antidepressant drugs and hip fracture in an elderly Mediterranean population.

Methods

Cases were all patients aged 50–95 years admitted to the emergency room of our hospital with hip fracture not related to a high intensity trauma during 2010. For each case, four controls were identified from primary care electronic medical records matched by age (±3 years), gender, date of consultation at the primary care centre (±1 month) and primary care centre. Pharmacological treatments received within the previous 5 years were retrieved from the prescription records. Crude and adjusted risks associated with exposures were calculated by conditional logistic regression. ORs were adjusted by matching variables and by significant risk factors identified in the bivariate analysis (prescription of ≥4 drugs, osteoporosis, diabetes mellitus and previous fracture).

Results

136 cases and 544 controls were analysed. Adjusted OR (95% CI) for hip fracture associated with exposure to any antidepressants was 2.42 (1.24 to 4.73); for selective serotonin reuptake inhibitors (SSRIs) it was 3.52 (1.67 to 7.41), for non-selective monoamine reuptake inhibitors 1.07 (0.18 to 6.46) and for other antidepressants 0.82 (0.27 to 2.48). Sertraline (OR 3.88 (1.15 to 13.09)) was the only active principle with significant adjusted risk. When only exposures >6 months were considered, significant risks persisted for SSRIs (OR 2.64 (1.10 to 6.37)).

Conclusions

The results of this study are coincident with other studies in which SSRIs, but not other types of antidepressants, are associated with an increased risk of hip fracture in our setting.

Introduction

It has been reported that antidepressants increase the risk of fragility fractures,¹ but the mechanism by which they do so may vary according to the group of antidepressant. It has been proposed that non-selective monoamine reuptake inhibitors may increase the risk of falls mainly due to their effects of sedation, orthostatic hypotension and/or confusion, especially at the beginning of treatment.^{2 3} Selective serotonin reuptake inhibitors (SSRIs) also increase the risk of falls, but can also affect bone metabolism as functional receptors and serotonin transport systems have been identified in osteoblasts, osteoclasts and osteocytes, on which SSRIs may have an effect that favours osteopenia.^{4 5} Inhibition of serotonin transport systems in the bone could reduce the stimulating effect of serotonin on the proliferation of precursor cells of osteoblasts, favouring a loss of bone mineral density (BMD) and an increased risk of fractures. However, the biochemical nature of serotoninergic pathways and their impact on bone metabolism are not yet fully understood.⁶

Moreover, polypharmacy may be a risk factor for fall-related fractures. Some studies have found a dose–response relationship between the number of medicines and the risk of hip fracture in older people.^{7 8}

The prevalence of antidepressant use in our setting is high and increasing, especially in the population over 75 years old,^{9 10} an age group with an increased risk of fragility fractures of the hip. Considering the high morbidity and mortality associated with a hip fracture in this population,¹¹ along with their increased susceptibility and exposure to antidepressants, it is important to know the additional risk posed by the use of these drugs when deciding their prescription. On the other hand, it is of interest to confirm if risk associations previously described are also present in our Mediterranean region, as the population risk is lower than in other regions where studies describing the risks associated with SSRIs have been performed.¹² It is also noteworthy that there is significant diversity in the rate of hip fracture between the different regions of Spain. It is difficult to be able to establish the causes, taking into account the fact that there are no notable differences in population structure, lifestyle or eating patterns between the different geographical zones.¹³ Moreover, there is a scarcity of data comparing consumption of antidepressants across multiple countries and their possible contribution to population-based hip fracture rates. In observational epidemiological studies, antidepressant use has been associated with a 3–7% contribution to the population rate of hip fracture. The magnitude of reported relative risk estimates varies considerably between studies, reflecting—at least in part—differences in study design, inclusion/exclusion criteria, exposure and outcome definitions and availability and classification of data on potential confounders.¹⁴ The aim of this study is to describe the associations between exposure to antidepressant drugs and hip fracture in our population, and to observe whether there is a differential effect between SSRIs and other types of antidepressant drugs on the increased risk of fractures in this setting.

Methods

We performed a case–control study. The protocol was reviewed and approved by the Research Ethics Committee of Corporació Parc Taulí. Cases were defined as all patients aged 50–95 years who, during 2010, were admitted to the emergency room of Parc Taulí Hospital Universitari due to femoral neck, femoral diaphysis, subtrochanteric or pelvic fractures not related to a high intensity trauma. Our hospital has exclusive coverage of an area of reference of about 400 000 inhabitants. Selected fractures were evaluated by a traumatologist, blind to the exposure of patients, to identify and validate bone fragility of fractures based on patient medical records and X-rays.

For each case, four controls were identified from primary care electronic medical records of the Catalan Health Service who had received assistance at least twice during the previous year in the same primary care centre as their matched case. Controls were selected from patients who consulted the primary care centre on the index date (defined as the date the matched case was admitted to the emergency room ±1 month) and also matched by age (±3 years) and gender. We excluded controls who visited the centre for diagnosis or treatment of fragility fractures, and those whose reason for visit on the index date included any diseases associated positively or negatively with exposures described as associated with increased risk (psychiatric, peptic disease or diabetes mellitus).

For both cases and controls, we retrieved data on personal history and osteoporosis risk factors from the patients’ electronic medical records at the hospital and at primary care, and on pharmacological treatments received within the previous 5 years from the Catalan Health Service invoicing system, coded according to the Anatomical Therapeutic Classification (ATC) of the World Health Organisation (http://www.whocc.no/atc_ddd_index/). In case of absence of valid records on the medical records or regarding prescriptions for a case or a control, we recorded the patient’s basic data and they were excluded from the study. We considered exposure the presence of at least one prescription of the studied active principles within the previous 5 years. We made a quantitative description of exposure times based on the start and end dates of treatment and, to consider the effects of chronic exposure, we analysed the risks excluding exposure durations of ≤6 months.

Statistical analysis

The use of antidepressant drugs was summarised by the number and proportion of cases and controls exposed with a 95% CI according to the Wilson method; SSRIs (ATC code N06AB) were distinguished from other antidepressant drugs (ATC codes N06AA, N06AF, N06AG, N06AX). The risks related to each exposure were calculated by multivariate methods based on conditional logistic regression models. We calculated models adjusted by matching parameters between cases and controls (crude model) and a model adjusted by matching variables and four factors significantly associated with fracture risk in the bivariate analyses (prior history of fracture, osteoporosis, diabetes and polypharmacy, defined as a number of drugs on the index date ≥4) (adjusted model). Associations were expressed as OR with 95% CI.

Sample size was calculated assuming a prevalence of use of SSRIs of around 20% in the control group.¹³ It was estimated that a sample of 166 cases and 664 controls would allow detection of an increased risk of fragility fracture of 1.9 or higher, with a type 1 error protection of 5% and a power of 80%, considering the possibility of 15% losses.

Results

We included a total of 165 patients with confirmed hip fragility fracture. Twenty-nine cases were excluded due to lack of valid medical and prescription records, for belonging to a health area different from the healthcare centre, or due to lack of eligible matching controls with valid medical and prescription records. Eventually, 136 cases of hip fracture were matched to 544 controls (figure 1).

Figure 1.

Selection of patients and controls. *Other locations than hip.

Table 1 shows the characteristics of cases and controls regarding sociodemographic and anthropometric variables, health-realted habits, diseases and number of treatments. Overall, comorbidities and health habits described as risk factors for fractures were more frequent among cases than among controls (diabetes, osteoporosis, malnutrition, smoking and alcohol habits, hyperthyroidism and number of prescriptions).

Table 1.

Population characteristics

	Controls n=544	Cases n=136	P value
Female	432 (79.4%)	108 (79.4%)
Age (years), mean (SD)	82.3 (8.6)	82.3 (8.7)	0.934
51–65 years	184 (33.8%)	46 (33.8%)
66–81 years	196 (36.0%)	49 (36.0%)
82–95 years	164 (30.1%)	41 (30.1%)
Weight (kg), mean (SD)	67.5 (13)	65 (11.61)	0.195
Height (cm), mean (SD)	155.5 (8)	152.4 (7.8)	0.033
BMI (kg/m2), mean (SD)	27.5 (4.9)	27.5 (4.9)	0.819
Referred smoking habit			<0.0001
Non-smoker/not stated	522 (96.0%)	121 (89.0%)
Active smoker	16 (2.9%)	6 (4.4%)
Former smoker	6 (1.1%)	9 (6.6%)
Referred alcoholic habit			0.042
Non-drinker/not stated	482 (88.6%)	128 (94.1%)
Light drinker	59 (10.8%)	6 (4.4%)
Moderate drinker	3 (0.6%)	2 (1.5%)
Rheumatoid arthritis	5 (0.9%)	1 (0.7%)	0.838
Previous fracture	2 (0.4%)	58 (42.6%)	<0.0001
Diabetes mellitus	123 (22.6%)	46 (33.8%)	0.007
Diabetes mellitus type 1	4 (0.7%)	0 (0.0%)	0.010
Diabetes mellitus type 2	119 (21.9%)	46 (33.8%)	0.010
Osteoporosis	13 (2.4%)	19 (14.0%)	0.000
Neoplasia	65 (11.9%)	13 (9.6%)	0.434
Hyperthyroidism	3 (0.6%)	4 (2.9%)	0.014
Malnutrition	1 (0.2%)	3 (2.2%)	0.006
Malabsorption	1 (0.2%)	0 (0.0%)	0.617
Chronic liver disease	7 (1.3%)	5 (3.7%)	0.058
Number of active prescriptions, median (25th and 75th percentiles)	3 (0/5)	8 (5/14)	<0.0001
Patients with ≥4 active prescriptions	229 (42.1%)	104 (76.5%)

BMI, body mass index.

A history of osteoporosis was associated with an increased crude risk of fracture of 8.46 (95% CI 3.67 to 19.53) and diabetes mellitus of 1.82 (95% CI 1.19 to 2.79). Crude OR for active prescription of ≥4 medications was 4.76 (95% CI 3.05 to 7.45). These factors were included in the multivariate model. There were also significant differences in the history of hyperthyroidism, malnutrition and chronic liver disease, but with a very low prevalence, so these variables were not included in the multivariate model.

The results of the analysis of risks associated with exposure to different groups of antidepressant drugs are shown in table 2. The prevalence of use of drugs of the N06A group (antidepressants) was 37.5% in cases and 12.3% in controls, with an adjusted OR showing an increase of more than twofold in the risk of fractures for patients exposed to any antidepressant.

Table 2.

Exposure to antidepressant drugs and associated risk

Exposure	Controls (n=544) n (%)	Cases (n=136) n (%)	OR (95% CI); p value Crude model	OR (95% CI); p value Adjusted model*
ATC N06A Antidepressants	67 (12.3%)	51 (37.5%)	4.56 (2.88 to 7.23); p≤0.001	2.42 (1.24 to 4.73); p=0.009
ACT N06AA Non-selective monoamine reuptake inhibitors	9 (1.6%)	4 (2.9%)	1.83 (0.54 to 6.14); p=0.329	1.07 (0.18 to 6.46); p=0.945
ATC N06AB Selective serotonin reuptake inhibitors	41 (7.5%)	37 (27.2%)	4.89 (2.87 to 8.33); p≤0.001	3.52 (1.67 to 7.41); p≤0.001
ATC N06AX Other antidepressants	21 (3.9%)	15 (11.0%)	2.98 (1.51 to 5.90); p=0.002	0.82 (0.27 to 2.48); p=0.719
All non-SSRI antidepressants	48 (8.8%)	31 (22.8%)	3.26 (1.93 to 5.49); p≤0.001	1.71 (0.79 to 3.71); p=0.172
Exposure lasting more than 6 months
ATC N06AB Selective serotonin reuptake inhibitors	31 (5.7%)	27 (19.9%)	4.68 (2.54 to 8.61); p≤0.001	2.64 (1.10 to 6.37); p=0.030
All non-SSRI antidepressants	40 (7.4%)	19 (14.0%)	2.12 (1.16 to 3.84); p=0.014	0.95 (0.38 to 2.38); p=0.907
Active principles
Paroxetine	13 (2.4%)	6 (4.4%)	1.88 (0.70 to 5.05); p=0.208	1.14 (0.28 to 4.58); p=0.857
Sertraline	10 (1.8%)	8 (5.9%)	3.56 (1.32 to 9.60); p=0.012	3.88 (1.15 to 13.09); p=0.029
Citalopram	8 (1.5%)	10 (7.4%)	5.00 (1.97 to 12.67); p=<0.001	1.48 (0.36 to 6.08); p=0.591
Escitalopram	10 (1.8%)	8 (5.9%)	3.20 (1.26 to 8.11); p=0.014	3.23 (0.88 to 11.87), p=0.077
Fluoxetine	5 (0.9%)	5 (3.7%)	4.00 (1.16 to 13.82); p=0.028	2.47 (0.39 to 15.79), p=0.338
Venlafaxine	7 (1.3%)	5 (3.7%)	3.05 (0.92 to 10.12); p=0.068	1.79 (0.31 to 10.40), p=0.517
Trazodone	10 (1.8%)	8 (5.9%)	3.20 (1.26 to 8.11); p=0.014	0.46 (0.07 to 2.86), p=0.402
Amitriptyline	7 (1.3%)	1 (0.7%)	0.57 (0.07 to 4.64); p=0.601	0.10 (<0.01 to 16.73), p=0.381
Mirtazapine	5 (0.9%)	2 (1.5%)	1.60 (0.31 to 8.25); p=0.574	0.17 (<0.01 to 4.90), p=0.303
Mianserine	3 (0.6%)	1 (0.7%)	1.3 3 (0.14 to 12.82); p=0.803	0.06 (<0.01 to 15.99), p=0.327

*Adjusted by age, sex, previous fracture, osteoporosis, diabetes mellitus and number of active drugs 4 or higher.

.ATC, Anatomical Therapeutic Classification; SSRI, selective serotonin reuptake inhibitors.

The drugs most commonly used were SSRIs, particularly citalopram, paroxetine and escitalopram. The second most used group of drugs was 'other antidepressants', in particular trazodone and venlafaxine. Non-selective reuptake inhibitor antidepressants were used infrequently (table 2).

In the analysis of hip fracture risk associated with different active principles, sertraline shows the greatest increased risk in the adjusted model (OR 3.88 (95% CI 1.15 to 13.09)) and escitalopram shows a non-significant trend (OR 3.23 (95% CI 0.88 to 11.87)) (table 2).

Discussion

We observed a high prevalence of antidepressant use in our setting, consistent with the data previously described.^{9 10} Exposure to SSRIs was associated with an increased risk of hip fracture of 3.5 times compared with unexposed subjects, after adjusting for other risk factors. When we analysed the risk for chronic exposure (duration of exposure longer than 6 months), we observed a persistent increase in fracture risk of 2.6 times compared with unexposed subjects. For other antidepressants we observed a non-significant increased risk (adjusted estimator 1.71, non-significant).

In our study the prevalence of exposure to SSRIs was 27.2% for cases and 7.5% for controls. This exposure is higher than those described in other population-based studies,2 3 15 16 but consistent with national consumer data which show that the use of antidepressants in the Spanish and Catalan population is generally high and has been growing in recent years, especially in the older age groups.^{10 17} Observed data suggest that SSRIs increase significantly the risk of hip fracture, and this risk persists in prolonged treatments, so it is consistent with the hypothesis that, once past the initial increase in risk attributable to secondary falls and potential sedative and hypotensive effects, SSRIs may interfere in the longer term on bone metabolism leading to fragility, which may increase the risk of fracture in low-intensity trauma.

It has previously been mentioned that depression itself may increase the risk of falls and fractures.¹⁸ Some studies have reported an association between depression and falls, independent of medication,¹⁹ but other studies have not observed such an association.^{3 20} Depression is often accompanied by additional risk factors for a low BMD such as lack of physical activity, hypovitaminosis D, anorexia and weight loss or a lower intake of calcium.²¹ Therefore, it is possible that the pathology itself can act as a confounding factor both with respect to the reduction of BMD and to increased falls. However, several studies have reported increases in the risk of fractures by antidepressant drugs. In a meta-analysis of 33 observational studies that evaluated the risk of fractures in different locations associated with the use of antidepressants, both SSRIs and tricyclic antidepressants were associated with a moderate increase in fractures of all kinds (RR 1.39; 95% CI 1.32 to 1.47) and, in particular, of hip fracture (RR 1.47; 95% CI 1.36 to 1.58), and the risk persisted even when the analysis was restricted to studies that could be adjusted to the independent effect of a diagnosis of depression (RR 1.30; 95% CI 1.15 to 1.47).¹

In particular, the use of SSRI antidepressants was associated independently with decreased BMD and increased risk of hip fracture in different studies after adjustment for several potential confounders,^{2 3 15} and two meta-analyses concluded that they represent an independent risk factor for any type of fracture^{1 22} and, in particular, hip fracture with a RR of 1.64 (95% CI 1.42 to 1.89).¹

The magnitude of the increased risk of hip fracture that we observed in our study due to exposure to SSRIs is greater than those described in other studies. This difference could be due to the fact that our model used few adjustment factors, since data sources used in our study limited the amount of information available. For example, we lacked indicators of physical and cognitive impairment, nutrition and physical activity. More moderate risks of hip fracture have been described associated with SSRIs when adjustments considered physical and cognitive impairment and the degree of disability in activities of daily living, with an OR of 1.8 (95% CI 1.54 to 2.10).²¹ Also, data on comorbidities available in our study were limited because the data source used is not specifically designed for research but for healthcare purposes. We therefore cannot exclude some residual confounding, which may result in an overestimation of risks in our study.

Moreover, polypharmacy is associated with risk of falls and fracture, even if taking multiple medications does not directly impact bone health.⁷ A case–control study found a dose–response relationship between the number of medicines and the risk of hip fracture in older people aged ≥65 years; older people taking 2 to 4 medicines had a 64% greater risk of hip fracture than older people taking 0–1 medicine.⁸ There is also evidence to suggest that the risk of polypharmacy on falls may be potentiated by the presence of depression.²³ In our study the polypharmacy is almost double in SSRI users. These factors suggest that patients with hip fracture receiving SSRI treatment are more fragile than non-users.

On the other hand, it is worth noting that, in our study, in the case of exposure to other non-SSRI antidepressants we observed a trend towards an increased risk that did not reach statistical significance in the adjusted model, which was not observed for exposures of duration >6 months. The absence of risk associated with chronic exposure for non-SSRI provides consistency to our findings, discarding confounding by an indication that would condition risk increases also in this group. These data are consistent with other studies which have found differential effects of the two types of antidepressants.^{24 25}

Some studies that have assessed the effect of treatment duration with antidepressants^{16 24 26} reported that the risk of fracture associated with SSRIs has an initial peak during the first 6–12 months of treatment and, despite a subsequent decrease, it remains high for exposures over 1.5 years. The profile of risk associated with exposure to tricyclic antidepressants shows a peak during the first 1–2 months and then decreases and normalises after 6–12 months of treatment. In both cases the increased risk was found to disappear between 3 and 12 months after treatment discontinuation.²⁷ Other studies, however, have found no differences in the timing of risk between the two groups of drugs,²⁸ suggesting that both have mainly an increased risk of falls related to effects on the level of awareness and coordination. A meta-analysis of 12 studies²⁹ shows that this increase in risk is more pronounced in the first 6 weeks of treatment (OR 3.83 (95% CI 1.96 to 7.49)) than in the period between 6 weeks and 6 months (OR 1.60 (95% CI 0.93 to 2.76)). The results of our study show an increased risk that persisted beyond 6 months for SSRIs, but only short term risk increase with non-SSRI. This may reflect differences in the main pharmacological mechanisms potentially involved in risk–that is, sedation in the case of non-SSRIs and a possible additional demineralising effect in the long term in the case of SSRIs.

In the analysis of risk associated with different active ingredients, we observed a significant increase in risk associated with sertraline, which would be consistent with the description, in terms of relative power, stating that sertraline would be the most potent SSRI inducing inhibition of the formation and resorption of osteoclasts, followed by fluoxetine, paroxetine and fluvoxamine.³⁰ Venlafaxine, included in the group of non-SSRIs which also have effects on serotonin reuptake, was not found to be associated with an increased risk.

In our study the inherent limitations derived from a retrospective observational study and its design did not allow us to adjust for some of the possible confounding factors such as different comorbidities. Therefore, the conclusions regarding the dose–response relationship, the temporal relationship or the underlying mechanism should be corroborated in future research studies.

In summary, in our study we observed a substantial increase in the risk of hip fracture associated with the chronic use of SSRI antidepressants, similar to other studies conducted in other countries. A risk that persists when exposures of less than 6 months are discarded, which suggests a chronic effect as a probable consequence of demineralisation by SSRIs in the long term. Finally, we must always take into account the risks of polypharmacy, especially among older adults, and we have to quantify the risk–benefit ratio of these medications, especially in frail older adults.

What this paper adds.

What is already known on this subject

• Antidepressants increase the risk of fragility fracture due to their effects on bone metabolism and sedation.

• Polypharmacy can be a risk factor for falls and fractures in the elderly population.

• Fragility fractures increase morbidity and mortality in this population.

What this study adds

• Our study confirms the risk of fracture associated with selective serotonin reuptake inhibitor antidepressants and not with other antidepressants in the Mediterranean population that has a lower risk of fragility fracture.