Abstract

The use of tricyclic antidepressants is associated with an increased risk of hip fracture. Despite a better side effect profile, this adverse effect has also been reported for selective serotonin reuptake inhibitors. To determine whether these findings result from bias arising from the case-control method, the authors have performed a case-control analysis and a self-controlled case-series analysis using 1987–1999 diagnosis data for 16,341 cases of hip fracture and 29,889 controls drawn from the United Kingdom General Practice Research Database. Both analyses showed an association between hip fracture and antidepressant treatment, and this was most marked during the first 15 days of treatment. The estimates from the case-control study were larger than those from the case-series analysis: The odds ratios for fracture within the first 15 days of a prescription for tricyclic antidepressants and serotonin reuptake inhibitors were 4.76 (95% confidence interval (CI): 3.06, 7.41) and 6.30 (95% CI: 2.65, 14.97), whereas the equivalent incidence ratios were 2.30 (95% CI: 1.82, 2.90) and 1.96 (95% CI: 1.35, 2.83). Tricyclic antidepressants and serotonin reuptake inhibitors are both associated with an independent increase in hip fracture incidence during the first weeks of treatment. The estimates from the case-series analyses were smaller than those from the case-control analyses, suggesting that the case-control method is subject to bias.

Received for publication September 23, 2002; accepted for publication January 16, 2003.

Each year in the United Kingdom more than 10 percent of people over the age of 60 years receive a prescription for an antidepressant drug (1). In most cases, this is a tricyclic drug (1), but prescriptions for selective serotonin reuptake inhibitors have increased rapidly over the last 10 years (2, 3). The two classes of antidepressants appear to have similar efficacy for the treatment of depression (4), but selective serotonin reuptake inhibitors have fewer anticholinergic side effects, appear to be better tolerated by patients (2), and are safer in overdose (5).

The use of tricyclic antidepressant drugs has been associated with hip fracture, possibly due to an increased tendency to fall as a result of orthostatic hypotension, sedation, and/or confusion (68). This increase in risk appears to be mainly associated with new prescriptions for tricyclic antidepressants, rather than continued use, suggesting that tolerance to this adverse effect occurs (8). Although fewer problems would be expected with selective serotonin reuptake inhibitors, recent evidence suggests that fall rates are similar among nursing home residents taking each class of antidepressant (9) and that the risk of hip fracture may even be higher with selective serotonin reuptake inhibitors (8). The reasons for these findings are not clear, but since these analyses are based on between-person comparisons, a selection bias may be present if physicians preferentially use selective serotonin reuptake inhibitors in frail patients at greater risk of falling. Furthermore, if depression itself is associated with an increased risk of fracture, the estimates for both classes of drug may be increased erroneously because of indication bias.

To establish more clearly whether tricyclic antidepressant drugs and/or selective serotonin reuptake inhibitors are associated with an independent increase in the risk of hip fracture, we have analyzed data from the United Kingdom (UK) General Practice Research Database (10). We performed a classic case-control analysis initially and followed this with a self-controlled case-series analysis (11, 12) to minimize the effects of selection and indication bias.

MATERIALS AND METHODS

The UK General Practice Research Database

The UK General Practice Research Database (GPRD) is the largest longitudinal primary care data set available in the United Kingdom and contains data for more than 7 million people. The primary care centers contributing data to the GPRD are based throughout the United Kingdom and in terms of demographics appear to be typical of UK general practices (10). The data included in the GPRD are collected by the UK Medicines Control Agency directly from the computerized recordings of routine primary care; for example, the data on drug exposures are derived from the software used to generate prescriptions. The validity of a wide range of drug exposures, including antidepressants, and of diagnoses has been tested within the GPRD and consistently found to be high (10). Data collection for the GPRD started in 1987 and is still ongoing.

Case-control analysis

We identified all patients within the GPRD with a recorded diagnosis of hip fracture or fractured neck of femur up to November 1999. When possible, we identified two controls that were individually matched to each case by age, sex, general practice, and duration of available GPRD data. Each case-control set was assigned an index date corresponding to the date of the first recorded hip fracture in the case. We then extracted information on all recorded prescriptions for antidepressants prior to the index date.

We estimated the impact of first prescriptions for tricyclic antidepressants or selective serotonin reuptake inhibitors dated from 0 to 14 days, from 15 to 42 days, and 43 or more days before the index date using conditional logistic regression (Stata version 7 software; Stata Corporation, College Station, Texas). We then examined the impact of a number of potential confounding variables, including previously recorded falls, cerebrovascular disease, body mass index, diastolic blood pressure, hormone replacement therapy, ischemic heart disease, eye disease, peripheral neuropathy, and prescriptions for hypnotics, opiates, nonopiod analgesics, antipsychotics, inhaled corticosteroids, and oral corticosteroids, by constructing a series of bivariate models as appropriate. If any of these variables led to a change in the initial odds ratio for antidepressants of more than 10 percent, these variables were retained in a multivariate model. We also fitted a model including all of the potential confounders, which we shall refer to as the “full” model. We then examined the impact of individual antidepressant drugs by estimating odds ratios for drugs with an exposure rate of at least 1 percent in the control group.

Self-controlled case-series analysis

The case-series method is a self-controlled study design derived from the cohort method in which the incidence of an outcome during predefined periods of time after the exposure is compared in persons individually with the incidence during other baseline time periods (11, 12). The analysis is therefore based only on data from cases, and the estimates of relative fracture incidence for antidepressant exposures are derived from cases who were prescribed antidepressants alone. The major advantage of the case-series approach for this study is that the influence of confounding by factors that vary between individuals, such as frailty, bone mineral density, and presence and severity of depression, is removed.

For our analyses, we included all of the cases of hip fracture whether they had a recorded prescription for an antidepressant or not. Cases that were not prescribed antidepressants will contribute information to the impact of age on fracture but not to the relative impact of antidepressant exposure. We identified all prescriptions for antidepressants during this time and defined the hip fracture date as the first recorded diagnosis of hip fracture. We then divided the person-time into different time periods as follows: from –0 to 14 days and from 15 to 42 days after the start of treatment (high-risk periods) and the remaining treatment period. The remaining treatment period was defined as the time from day 43 of treatment to 31 days after the last prescription. We chose 31 days as a conservative estimate of duration of a prescription because, in our data set, the median duration of time between prescriptions for people with more than one prescription was 33 days (interquartile range, 26–61 days). We also included a washout period set arbitrarily at 6 months (182 days), which started at 31 days after the last prescription. Because the precise date when treatment stopped was not known, we divided the washout period into two equal periods of 91 days, assuming that although there might still be some drug exposure during the first washout period, there should not be any during the second period. All remaining person-time was used as the baseline (unexposed) comparison period. The division of the person-time is shown pictorially in figure 1.

We estimated the relative incidence ratios using conditional Poisson regression (12) within GLIM version 4.09 software (Numerical Algorithms Group, Ltd., Royal Statistical Society, Oxford, England) and in each analysis adjusted for the impact of age at fracture in 1-year bands. We first assessed the impact of exposure to any tricyclic antidepressant drug or selective serotonin reuptake inhibitor, but then examined the impact of specific drugs in a fashion similar to that of the case-control study. Because a significant proportion of patients received only one prescription for an antidepressant, we repeated our analyses restricting our data set to subjects treated for at least 12 months.

RESULTS

Case-control study

Our data set included 16,341 cases of hip fracture with two matched controls for 13,548 cases and one matched control for 2,793 cases. The mean age of the cases was 79 years (standard deviation, 12 years), and 12,854 (79 percent) were female. The median period of prescribing data prior to fracture was 2.7 years (interquartile range, 1.1–4.9 years).

Among the potential confounding variables, an increased risk of hip fracture was associated with a history of one or more falls (odds ratio (OR) = 4.67, 95 percent confidence interval (CI): 4.39, 4.95), cerebrovascular disease (OR = 1.89, 95 percent CI: 1.76, 2.03), being underweight (body mass index, <18.5; OR = 2.40, 95 percent CI: 2.02, 2.86), eye disease (OR = 1.11, 95 percent CI: 1.06, 1.17), and prescriptions for hypnotics (OR = 1.68, 95 percent CI: 1.60, 1.77), opioid analgesics (OR = 1.94, 95 percent CI: 1.81, 2.07), antipsychotics (OR = 2.79, 95 percent CI: 2.60, 2.99), oral corticosteroids (OR = 1.60, 95 percent CI: 1.50, 1.71), and inhaled corticosteroids (OR = 1.26, 95 percent CI: 1.17, 1.36). There was a dose-related decrease in risk of hip fracture with increasing diastolic blood pressure (OR per 10-mmHg increase in diastolic blood pressure = 0.92, 95 percent CI: 0.89, 0.95), and other protective factors included being mildly overweight (body mass index, 25–30; OR = 0.55, 95 percent CI: 0.51, 0.60), being obese (body mass index, >30; OR = 0.39, 95 percent CI: 0.34, 0.44), and having prescriptions for hormone replacement therapy (OR = 0.79, 95 percent CI: 0.66, 0.94). We found no evidence of associations between hip fracture and the presence of ischemic heart disease (OR = 1.00, 95 percent CI: 0.93, 1.07) or peripheral neuropathy (OR = 1.04, 95 percent CI: 0.96, 1.12).

Cases were more likely to be exposed to tricyclic antidepressants and selective serotonin reuptake inhibitors than were controls for all time periods before the index date, but these effects were particularly marked during the first 6 weeks of treatment (table 1). In our bivariate analyses, the confounders that appeared to have an impact on the association between exposure to tricyclic antidepressants and hip fracture were a history of falls and prescriptions for hypnotics and antipsychotic drugs. All of these confounders independently led to a reduction in the odds ratio for tricyclic antidepressants of approximately 10 percent, and so all were included in the final multivariate model (table 1). The results from the full model were similar to those from our multivariate model; for example, the odds ratio for “ever exposed” to antidepressants in the full model was 1.16 (95 percent CI: 1.09, 1.23). The same three confounders also attenuated the odds ratios for the association between selective serotonin reuptake inhibitors and hip fracture (table 1), and again the results from the full model were similar (OR for ever exposed to selective serotonin reuptake inhibitors = 1.35, 95 percent CI: 1.20, 1.51).

The individual antidepressant drugs with an overall exposure prevalence of 1 percent or greater in the control group were amitriptyline, dothiepin, lofepramine, fluoxetine, and paroxetine. All five drugs were associated with an acute increase in the risk of hip fracture if first prescribed within 14 days of the index date (table 2). The number of exposed cases and controls here was small and the confidence intervals were wide, but the strongest effects were seen with lofepramine (OR = 9.96, 95 percent CI: 3.24, 30.59) and fluoxetine (OR = 8.59, 95 percent CI: 2.42, 30.48).

Case-series analysis

The demographic details of the cases included in the case-series analysis are shown in table 3. Note that the case-series analysis uses information on the first exposure to antidepressants at any time, both before and after the fracture. Therefore, although the case population for this analysis is identical to that in the case-control study, the total number of cases who received a prescription for an antidepressant appears to be greater in the case-series analysis because some of the prescriptions were started after the fracture (table 3). The median duration of total observation period for the case-series analysis was greater than 5 years for each analysis. The mean age at first prescription was similar for all drugs, as was the proportion of people who had only one recorded prescription.

The relative incidence of hip fracture was increased after the first prescription for any tricyclic antidepressant drug, and this effect was most marked during the 15 days of treatment (incidence ratio (IR) = 2.30, 95 percent CI: 1.82, 2.90) (table 4). Thereafter, the relative incidence fell toward unity, but there was a marginal increase in fracture incidence during the remaining antidepressant exposure period (IR = 1.18, 95 percent CI: 1.06, 1.31), and there was also a small increase in fracture incidence during the first 91-day washout period. In each case, the size of the incidence ratios during the first 43 days of antidepressant exposure was smaller than the equivalent odds ratios from the case-control study; for example, for the first 15 days of treatment with a tricyclic antidepressant, the incidence ratio was 2.3 but the odds ratio was 4.8. A similar pattern was apparent for the three individual tricyclic drugs studied.

The relative incidence of hip fracture was also increased during the first 15 days of treatment with selective serotonin reuptake inhibitors, although this effect was not quite as marked as that seen for tricyclic drugs (IR = 1.96, 95 percent CI: 1.35, 2.83). In contrast to tricyclic antidepressants, however, the relative incidence of fracture with selective serotonin reuptake inhibitors did not decrease over the next 28 days (IR = 2.03, 95 percent CI: 1.53, 2.68) and remained raised during the remaining treatment period (IR = 1.54, 95 percent CI: 1.3, 1.78) (table 5). Again, the sizes of the relative incidences during the initial treatment periods with selective serotonin reuptake inhibitor exposure were lower than the equivalent odds ratio from the case-control study. The number of cases was small for individual selective serotonin reuptake inhibitors, but the acute impact of fluoxetine did appear to be more marked than that of paroxetine. Again, an increase in fracture rate during the first washout period was seen for selective serotonin reuptake inhibitors as a whole and for both individual drugs.

When the data set was restricted to patients who received 12 months of treatment, the statistical power was reduced and the resulting confidence intervals were wider, but the estimates of effect remained similar (table 6).

DISCUSSION

Our results show that the use of antidepressants is associated with an independent increase in the risk of hip fracture, that the size of this increase in risk is similar for tricyclic antidepressants and selective serotonin reuptake inhibitors, and that much of this increase occurs during the first 6 weeks of treatment. The large odds ratios derived from the case-control analysis, which suggest that the risk of hip fracture during the first 2 weeks of treatment may be increased by more than 300 percent, are probably overestimates of the true effect due to problems with selection and indication bias. The design of the case-series method will overcome these problems, and the more modest estimates from this analysis are likely to reflect the true adverse impact of antidepressants on hip fracture risk.

The GPRD is the largest longitudinal primary care data set currently available (10), and this data set represents one of the largest pharmacoepidemiologic studies of hip fracture reported to date. The accuracy of hip fracture diagnoses within the GPRD has been investigated previously and been found to be high (13). Because the data within the GPRD are derived from computerized general practice records and recorded as part of routine clinical care, they should reflect real life experience rather than the research setting. One disadvantage of these data, however, is that some important information is not available, for example, the severity of depression, the degree of frailty, and bone mineral density. This means that, with classic case-control analyses, it can be difficult to exclude bias and confounding.

The results from our case-control study are consistent with those from the study by Liu et al. (8) in showing that both tricyclic antidepressants and selective serotonin reuptake inhibitors are associated with an increased risk of hip fracture, and that the main increase in fracture occurs shortly after the start of treatment. These similarities were expected because the two data sets have many similarities. The data are not consistent with the adverse effect profiles of these two classes of drug, however. Tricyclic antidepressants have a greater anticholinergic activity than selective serotonin reuptake inhibitors and are less well tolerated by patients, so they might be expected to be associated with a greater risk of hip fracture (2). One explanation for the findings of case-control studies is that the elevated risk of fracture associated with the use of selective serotonin reuptake inhibitors occurred because physicians preferentially prescribe these drugs to people who, in their opinion, will have adverse effects with tricyclic antidepressants, for example, frail patients. This selection bias would lead to an overestimate of the impact of selective serotonin reuptake inhibitors and an underestimate for tricyclic antidepressants. In addition to selection bias, an indication bias may also overinflate the odds ratios in our case-control study because depression per se may be a risk factor for fracture. A previous study has demonstrated an association between depression and decreased bone mineral density (14), and other factors, such as agitation, may also increase the risk of fracture in patients with depression.

The case-series method was originally developed to allow a rapid assessment of vaccine safety in response to concerns about adverse effects (11), but the advantages of the method have been exploited in other settings (15, 16). Its particular advantage for this study is that the influence of factors that vary between individuals, such as frailty and severity of depression, is removed. Our findings of an increased incidence in fracture shortly after the start of treatment with both tricyclic antidepressants and selective serotonin reuptake inhibitors are, therefore, highly suggestive of an independent acute adverse effect of both drug groups. Furthermore, the smaller size of these acute effects in the case-series analysis suggests that the results from the case-control analysis are overestimates resulting from bias.

In our case-series analysis, the tricyclic antidepressants were associated with a marginally higher early risk of fracture than were the selective serotonin reuptake inhibitors, in keeping with the more marked sedation and anticholinergic adverse effects associated with these drugs. There was, however, some heterogeneity of effect within each drug class. Among the tricyclic antidepressants, the early increase in fracture incidence was more marked with amitriptyline and lofepramine than with dothiepin. The better safety profile of dothiepin in comparison with amitriptyline is in keeping with meta-analyses of the results of prospective clinical trials, which suggest that dothiepin is better tolerated by patients (17, 18). The reason for the marked increase in hip fracture with lofepramine is not clear since, although fewer data are available, it appears to have a better side effect profile than amitriptyline does (19). Among the selective serotonin reuptake inhibitors, fluoxetine was associated with a larger relative incidence of hip fracture than paroxetine during the first 2 weeks of use, which may reflect the increased risk of agitation in the first few days of treatment with this drug (19).

As the duration of treatment with tricyclic antidepressants increased, the relative incidence of hip fracture fell progressively toward unity in the case-series analysis, and much of the increase in risk had gone after 6 weeks of starting treatment. This pattern was similar when the data set was restricted to patients treated for at least 12 months, in keeping with a progressive tolerance to their anticholinergic effects rather than a survivor bias (20). For selective serotonin reuptake inhibitors, the increase in relative fracture incidence was sustained during the first 6 weeks of treatment and remained high during the remaining treatment period for fluoxetine. The reasons why selective serotonin reuptake inhibitors increase the risk of fracture are still not clear but they, and fluoxetine in particular, may be associated with bradycardia, orthostatic hypotension, and syncope (2123). Similar results were seen in the case-control study for antidepressant prescriptions that started more than 43 days before the outcome, although these odds ratios are not directly comparable with the incidence ratios from the case-series analysis, because the case-control analysis does not take account of whether the medication was stopped or whether the risk of fracture was increased during the washout period.

Both classes of drug were associated with an increased risk of fracture during the first washout period, and it is possible that this was due to withdrawal symptoms (20). An alternative explanation, however, is that the antidepressant treatment was stopped because the patient developed another risk factor for fracture, for example, a prescription for another drug associated with postural hypotension or a new comorbid illness such as a stroke.

In summary, our findings suggest that tricyclic antidepressants and selective serotonin reuptake inhibitors are both associated with an increased risk of fracture in the first few weeks of treatment. The results from the case-series approach are likely to estimate these effects more accurately, because they remove the problems of selection and indication bias present in the case-control study. The clinical implication is that older patients starting treatment with either tricyclic antidepressants or selective serotonin reuptake inhibitors will have a transient doubling in their risk of hip fracture and that patients should be aware of this increased risk so that they can take appropriate precautions.

ACKNOWLEDGMENTS

The Wellcome Trust funded this project.

The authors thank Alan Dean, Hassy Devalia, Sam Rowlands, and Alison Bourke from the Epidemiology and Pharmacology Information Core for their expert advice in using the General Practice Research Database.

Correspondence to Dr. Richard Hubbard, Division of Respiratory Medicine, Clinical Sciences Building, Nottingham City Hospital, Nottingham NG5 1PB, United Kingdom (e-mail: Richard.hubbard@nottingham.ac.uk).

FIGURE 1. Pictorial representation of the case-series approach used to analyze data from the 16,341 patients with hip fracture included within the United Kingdom General Practice Research Database (GPRD) (data for diagnoses between 1987 and 1999).

FIGURE 1. Pictorial representation of the case-series approach used to analyze data from the 16,341 patients with hip fracture included within the United Kingdom General Practice Research Database (GPRD) (data for diagnoses between 1987 and 1999).

TABLE 1.

Case-control analysis: association between antidepressant exposure and hip fracture, UK* General Practice Research Database, 1987–1999

Exposure Cases  Controls  Univariate analysis  Multivariate analysis 
 (n = 16,341)  (n = 29,889)  Odds ratio 95% confidence interval  Odds ratio† 95% confidence interval 
 No.  No.       
Any tricyclic antidepressant            
Exposed before hip fracture 2,908 17.8  3,544 11.9  1.61 1.53, 1.70  1.22 1.15, 1.29 
First prescription (days before fracture)            
0–14  78 0.5  31 0.1  5.15 3.39, 7.84  4.76 3.06, 7.41 
15– 42  118 0.7  76 0.3  3.15 2.34, 4.22  2.58 1.88, 3.53 
≥43  2,712 16.6  3,437 11.5  1.55 1.46, 1.63  1.15 1.08, 1.23 
Any selective serotonin reuptake inhibitor            
Exposed before hip fracture 955 5.8  892 3.0  2.01 1.82, 2.21  1.42 1.28, 1.58 
First prescription (days before fracture)            
0–14  29 0.2  0.02  8.02 3.50, 18.38  6.30 2.65, 14.97 
15–42  54 0.3  16 0.05  6.38 3.64, 11.18  4.30 2.39, 7.74 
≥43  872 5.3  869 2.9  1.88 1.70, 2.07  1.32 1.19, 1.48 
Exposure Cases  Controls  Univariate analysis  Multivariate analysis 
 (n = 16,341)  (n = 29,889)  Odds ratio 95% confidence interval  Odds ratio† 95% confidence interval 
 No.  No.       
Any tricyclic antidepressant            
Exposed before hip fracture 2,908 17.8  3,544 11.9  1.61 1.53, 1.70  1.22 1.15, 1.29 
First prescription (days before fracture)            
0–14  78 0.5  31 0.1  5.15 3.39, 7.84  4.76 3.06, 7.41 
15– 42  118 0.7  76 0.3  3.15 2.34, 4.22  2.58 1.88, 3.53 
≥43  2,712 16.6  3,437 11.5  1.55 1.46, 1.63  1.15 1.08, 1.23 
Any selective serotonin reuptake inhibitor            
Exposed before hip fracture 955 5.8  892 3.0  2.01 1.82, 2.21  1.42 1.28, 1.58 
First prescription (days before fracture)            
0–14  29 0.2  0.02  8.02 3.50, 18.38  6.30 2.65, 14.97 
15–42  54 0.3  16 0.05  6.38 3.64, 11.18  4.30 2.39, 7.74 
≥43  872 5.3  869 2.9  1.88 1.70, 2.07  1.32 1.19, 1.48 

* UK, United Kingdom.

† Adjusted for history of falls and prescriptions for hypnotics and antipsychotics.

TABLE 2.

Case-control analysis: association between hip fracture and specific antidepressants for prescriptions starting 14 days or less before diagnosis, UK* General Practice Research Database, 1987–1999

Exposure No. of cases  No. of controls Univariate analysis  Multivariate analysis 
 (n = 16,341) (n = 29,889) Odds ratio 95% confidence interval  Odds ratio† 95% confidence interval 
Tricyclic antidepressants        
Amitriptyline 29 13 4.30 2.23, 8.29  3.73 1.88, 7.42 
Dothiepin 22 13 3.19 1.60, 6.37  2.72 1.29, 5.73 
Lofepramine 22 10.31 3.53, 30.09  9.96 3.24, 30.59 
Selective serotonin reuptake inhibitors        
Fluoxetine 16 10.85 3.16, 37.25  8.59 2.42, 30.48 
Paroxetine 2.90 0.94, 8.94  2.01 0.61, 6.62 
Exposure No. of cases  No. of controls Univariate analysis  Multivariate analysis 
 (n = 16,341) (n = 29,889) Odds ratio 95% confidence interval  Odds ratio† 95% confidence interval 
Tricyclic antidepressants        
Amitriptyline 29 13 4.30 2.23, 8.29  3.73 1.88, 7.42 
Dothiepin 22 13 3.19 1.60, 6.37  2.72 1.29, 5.73 
Lofepramine 22 10.31 3.53, 30.09  9.96 3.24, 30.59 
Selective serotonin reuptake inhibitors        
Fluoxetine 16 10.85 3.16, 37.25  8.59 2.42, 30.48 
Paroxetine 2.90 0.94, 8.94  2.01 0.61, 6.62 

* UK, United Kingdom.

† Adjusted for history of falls and prescriptions for hypnotics and antipsychotics.

TABLE 3.

Details of populations for case-series analysis, UK* General Practice Research Database, 1987–1999

Drug First exposed before hip fracture (no.) First exposed after hip fracture (no.) Duration (median years) of available data  Mean age at first exposure (years) Median no. of prescriptions  Subjects with only one prescription  Duration (median years) of antidepressant exposure 
      No.  
Any tricyclic antidepressant 2,908 1,050 5.92 (3.83–7.85)† 77 7 (2–24)† 815 21 1.18 (0.16–3.35)† 
Amitriptyline 933 475 6.25 (4.50–8.08) 77 4 (1–15) 440 31 0.51 (0.08–2.36) 
Dothiepin 1,079 379 6.08 (4.03–8.03) 76 5 (1–19) 382 26 0.64 (0.08–2.45) 
Lofepramine 656 322 5.86 (3.66–7.72) 79 3 (1–11) 294 30 0.35 (0.08–1.47) 
Any selective serotonin reuptake inhibitor 955 718 6.69 (4.57–8.18) 74 5 (2–14) 392 23 0.59 (0.11–1.91) 
Fluoxetine 489 344 6.75 (4.73–8.18) 77 3 (1–9) 241 29 0.36 (0.08–1.29) 
Paroxetine 358 293 7.00 (4.83–8.40) 77 4 (1–13) 203 31 0.39 (0.08–1.33) 
Drug First exposed before hip fracture (no.) First exposed after hip fracture (no.) Duration (median years) of available data  Mean age at first exposure (years) Median no. of prescriptions  Subjects with only one prescription  Duration (median years) of antidepressant exposure 
      No.  
Any tricyclic antidepressant 2,908 1,050 5.92 (3.83–7.85)† 77 7 (2–24)† 815 21 1.18 (0.16–3.35)† 
Amitriptyline 933 475 6.25 (4.50–8.08) 77 4 (1–15) 440 31 0.51 (0.08–2.36) 
Dothiepin 1,079 379 6.08 (4.03–8.03) 76 5 (1–19) 382 26 0.64 (0.08–2.45) 
Lofepramine 656 322 5.86 (3.66–7.72) 79 3 (1–11) 294 30 0.35 (0.08–1.47) 
Any selective serotonin reuptake inhibitor 955 718 6.69 (4.57–8.18) 74 5 (2–14) 392 23 0.59 (0.11–1.91) 
Fluoxetine 489 344 6.75 (4.73–8.18) 77 3 (1–9) 241 29 0.36 (0.08–1.29) 
Paroxetine 358 293 7.00 (4.83–8.40) 77 4 (1–13) 203 31 0.39 (0.08–1.33) 

* UK, United Kingdom.

† Numbers in parentheses, interquartile range.

TABLE 4.

Case-series analysis for tricyclic antidepressants: association between exposure and hip fracture, UK* General Practice Research Database, 1987–1999

Drug and exposure period Fractures (no.) Incidence ratio 95% confidence interval 
Any tricyclic antidepressant    
Unexposed period 1,926  
0–14 days† 78 2.30 1.82, 2.90 
15–42 days‡ 118 1.88 1.55, 2.28 
Other exposed period§ 1,465 1.18 1.06, 1.31 
Washout 1¶ 240 1.54 1.33, 1.77 
Washout 2# 131 1.18 0.98, 1.42 
Amitriptyline    
Unexposed period 848  
0–14 days 29 2.56 1.77, 3.72 
15–42 days 35 1.66 1.17, 2.33 
Other exposed period 375 1.10 0.91, 1.32 
Washout 1 75 1.39 1.09, 1.77 
Washout 2 46 1.13 0.83, 1.53 
Dothiepin    
Unexposed period 852  
0–14 days 22 1.77 1.15, 2.72 
15–42 days 27 1.18 0.80, 1.74 
Other exposed period 419 1.21 1.01, 1.45 
Washout 1 87 1.42 1.12, 1.79 
Washout 2 51 1.11 0.83, 1.48 
Lofepramine    
Unexposed period 614  
0–14 days 22 2.44 1.59, 3.76 
15–42 days 40 2.44 1.75, 3.39 
Other exposed period 218 1.29 1.03, 1.60 
Washout 1 54 1.37 1.03, 1.83 
Washout 2 30 0.96 0.66, 1.40 
Drug and exposure period Fractures (no.) Incidence ratio 95% confidence interval 
Any tricyclic antidepressant    
Unexposed period 1,926  
0–14 days† 78 2.30 1.82, 2.90 
15–42 days‡ 118 1.88 1.55, 2.28 
Other exposed period§ 1,465 1.18 1.06, 1.31 
Washout 1¶ 240 1.54 1.33, 1.77 
Washout 2# 131 1.18 0.98, 1.42 
Amitriptyline    
Unexposed period 848  
0–14 days 29 2.56 1.77, 3.72 
15–42 days 35 1.66 1.17, 2.33 
Other exposed period 375 1.10 0.91, 1.32 
Washout 1 75 1.39 1.09, 1.77 
Washout 2 46 1.13 0.83, 1.53 
Dothiepin    
Unexposed period 852  
0–14 days 22 1.77 1.15, 2.72 
15–42 days 27 1.18 0.80, 1.74 
Other exposed period 419 1.21 1.01, 1.45 
Washout 1 87 1.42 1.12, 1.79 
Washout 2 51 1.11 0.83, 1.48 
Lofepramine    
Unexposed period 614  
0–14 days 22 2.44 1.59, 3.76 
15–42 days 40 2.44 1.75, 3.39 
Other exposed period 218 1.29 1.03, 1.60 
Washout 1 54 1.37 1.03, 1.83 
Washout 2 30 0.96 0.66, 1.40 

* UK, United Kingdom.

† 0–14 days = first 15 days of treatment, including the day of prescription.

‡ 15–42 days = next 28 days after first prescription.

§ Other exposed period = time from 43 days after first prescription to 31 days after last prescription.

¶ Washout 1 = time from 32 days after last prescription to 123 days after last prescription.

# Washout 2 = time from 124 days after last prescription to 215 days after last prescription.

TABLE 5.

Case-series analysis for selective serotonin reuptake inhibitors: association between exposure and hip fracture, UK* General Practice Research Database, 1987–1999

Drug and exposure period Fractures (no.) Incidence ratio 95% confidence interval 
Any selective serotonin reuptake inhibitor    
Unexposed period 990  
0–14 days† 29 1.96 1.35, 2.83 
15–42 days‡ 54 2.03 1.53, 2.68 
Other exposed period§ 472 1.54 1.33, 1.78 
Washout 1¶ 87 1.57 1.25, 1.96 
Washout 2# 41 0.99 0.72, 1.36 
Fluoxetine    
Unexposed period 530  
0–14 days 16 2.31 1.41, 3.81 
15–42 days 24 1.93 1.27, 2.92 
Other exposed period 188 1.81 1.44, 2.26 
Washout 1 47 1.55 1.14, 2.11 
Washout 2 28 1.19 0.81, 1.75 
Paroxetine    
Unexposed period 429  
0–14 days 1.36 0.67, 2.77 
15–42 days 20 1.94 1.23, 3.06 
Other exposed period 130 1.28 0.99, 1.65 
Washout 1 48 2.08 1.53, 2.83 
Washout 2 16 0.93 0.56, 1.55 
Drug and exposure period Fractures (no.) Incidence ratio 95% confidence interval 
Any selective serotonin reuptake inhibitor    
Unexposed period 990  
0–14 days† 29 1.96 1.35, 2.83 
15–42 days‡ 54 2.03 1.53, 2.68 
Other exposed period§ 472 1.54 1.33, 1.78 
Washout 1¶ 87 1.57 1.25, 1.96 
Washout 2# 41 0.99 0.72, 1.36 
Fluoxetine    
Unexposed period 530  
0–14 days 16 2.31 1.41, 3.81 
15–42 days 24 1.93 1.27, 2.92 
Other exposed period 188 1.81 1.44, 2.26 
Washout 1 47 1.55 1.14, 2.11 
Washout 2 28 1.19 0.81, 1.75 
Paroxetine    
Unexposed period 429  
0–14 days 1.36 0.67, 2.77 
15–42 days 20 1.94 1.23, 3.06 
Other exposed period 130 1.28 0.99, 1.65 
Washout 1 48 2.08 1.53, 2.83 
Washout 2 16 0.93 0.56, 1.55 

* UK, United Kingdom.

† 0–14 days = first 15 days of treatment, including the day of prescription.

‡ 15–42 days = next 28 days after first prescription.

§ Other exposed period = time from 43 days after first prescription to 31 days after last prescription.

¶ Washout 1 = time from 32 days after last prescription to 123 days after last prescription.

# Washout 2 = time from 124 days after last prescription to 215 days after last prescription.

TABLE 6.

Case-series analysis for tricyclic antidepressants and selective serotonin reuptake inhibitors for patients with treatment duration of at least 12 months, UK* General Practice Research Database, 1987–1999

Drug and exposure period Fractures (no.) Incidence ratio 95% confidence interval 
Any tricyclic antidepressant    
Unexposed period 579  
0–14 days† 23 1.85 1.21, 2.82 
15–42 days‡ 34 1.45 1.02, 2.06 
Other exposed period§ 1,277 1.00 0.89, 1.14 
Washout 1¶ 115 1.67 1.35, 2.06 
Washout 2# 51 1.14 0.85, 1.53 
Any selective serotonin reuptake inhibitor    
Unexposed period 257  
0–14 days 0.98 0.37, 2.59 
15–42 days 16 2.07 1.24, 3.45 
Other exposed period 362 1.22 1.02, 1.47 
Washout 1 23 1.41 0.91, 2.19 
Washout 2 0.62 0.29, 1.34 
Drug and exposure period Fractures (no.) Incidence ratio 95% confidence interval 
Any tricyclic antidepressant    
Unexposed period 579  
0–14 days† 23 1.85 1.21, 2.82 
15–42 days‡ 34 1.45 1.02, 2.06 
Other exposed period§ 1,277 1.00 0.89, 1.14 
Washout 1¶ 115 1.67 1.35, 2.06 
Washout 2# 51 1.14 0.85, 1.53 
Any selective serotonin reuptake inhibitor    
Unexposed period 257  
0–14 days 0.98 0.37, 2.59 
15–42 days 16 2.07 1.24, 3.45 
Other exposed period 362 1.22 1.02, 1.47 
Washout 1 23 1.41 0.91, 2.19 
Washout 2 0.62 0.29, 1.34 

* UK, United Kingdom.

† 0–14 days = first 15 days of treatment, including the day of prescription.

‡ 15–42 days = next 28 days after first prescription.

§ Other exposed period = time from 43 days after first prescription to 31 days after last prescription.

¶ Washout 1 = time from 32 days after last prescription to 123 days after last prescription.

# Washout 2 = time from 124 days after last prescription to 215 days after last prescription.

References

1.
Office for National Statistics. Key health statistics from general practice 1996. London, England: Office for National Statistics, 1998:44–7.
2.
Martin RM, Hilton SR, Kerry SM, et al. General practitioners’ perceptions of the tolerability of antidepressant drugs: a comparison of selective serotonin reuptake inhibitors and tricyclic antidepressants.
BMJ
 
1997
;
314
:
646
–51.
3.
Olfson M, Klerman GL. Trends in the prescription of antidepressants by office-based psychiatrists.
Am J Psychiatry
 
1993
;
150
:
571
–7.
4.
Geddes JR, Freemantle N, Mason J, et al. SSRIs versus alternative antidepressants in depressive disorder (Cochrane review). In: The Cochrane Library, issue 4. Oxford, England: Update Software, 1999.
5.
Henry JA, Alexander CA, Sener EK. Relative mortality from overdose of antidepressants.
BMJ
 
1995
;
310
:
221
–4.
6.
Ray WA, Griffin JP, Downey W. Benzodiazepines of long and short elimination half-life and the risk of hip fracture.
JAMA
 
1989
;
262
:
3303
–7.
7.
Ray WA, Griffin MR, Malcolm E. Cyclic antidepressants and the risk of hip fracture.
Arch Intern Med
 
1991
;
151
:
754
–6.
8.
Liu B, Anderson G, Mittmann N, et al. Use of selective serotonin-reuptake inhibitors or tricyclic antidepressants and risk of hip fractures in elderly people.
Lancet
 
1998
;
351
:
1303
–7.
9.
Thapa PB, Gideon P, Cost TW, et al. Antidepressants and the risk of falls among nursing home residents.
N Engl J Med
 
1998
;
339
:
875
–82.
10.
Walley T, Mantgani A. The UK General Practice Research Database.
Lancet
 
1997
;
350
:
1097
–9.
11.
Farrington P, Pugh S, Colville A, et al. A new method for active surveillance of adverse events from diphtheria/tetanus/pertussis and measles/mumps/rubella vaccines.
Lancet
 
1995
;
345
:
567
–9.
12.
Farrington CP, Nash J, Miller E. Case series analysis of adverse reactions to vaccines: a comparative evaluation.
Am J Epidemiol
 
1996
;
143
:
1165
–73.
13.
van Staa TP, Leufkens HGM, Abenhaim L, et al. Use of oral corticosteroids and risk of fractures.
J Bone Miner Res
 
2001
;
16
:
581
–8.
14.
Michelson D, Stratakis C, Hill L, et al. Bone mineral density in women with depression.
N Engl J Med
 
1996
;
355
:
1176
–81.
15.
Murphy TV, Gargiullo PM, Massoudi MS, et al. Intussusception among infants given oral rotavirus vaccine.
N Engl J Med
 
2001
;
344
:
564
–72.
16.
Kramarz P, DeStefano F, Gargiullo PM, et al. Does influenza vaccination exacerbate asthma? Analysis of a large cohort of children with asthma. Vaccine Safety Datalink Team.
Arch Fam Med
 
2000
;
9
:
617
–23.
17.
Barbui C, Hotopf M. Amitriptyline v. the rest: still the leading antidepressant after 40 years of randomised controlled trials.
Br J Psychiatry
 
2001
;
178
:
129
–44.
18.
Anderson IM. Selective serotonin reuptake inhibitors versus tricyclic antidepressants: a meta-analysis of efficacy and tolerability.
J Affect Disord
 
2000
;
58
:
19
–36.
19.
Anderson IM. Meta-analytical studies on new antidepressants.
Br Med Bull
 
2001
;
57
:
161
–78.
20.
Parfitt K. Antidepressants. In: The complete drug reference. London, England: Pharmaceutical Press, 1999:271–312.
21.
Pacher P, Ungvari Z. Selective serotonin-reuptake inhibitor antidepressants increase the risk of falls and hip fractures in elderly people by inhibiting cardiovascular ion channels.
Med Hypotheses
 
2001
;
57
:
469
–71.
22.
Rodriguez de la Torre B, Dreher J, Malevany I, et al. Serum levels and cardiovascular effects of tricyclic antidepressants and selective serotonin inhibitors in depressed patients.
Ther Drug Monit
 
2001
;
23
:
435
–40.
23.
Cherin P, Colvez A, Deville de Periere G, et al. Risk of syncope in the elderly and consumption of drugs: a case-control study.
J Clin Epidemiol
 
1997
;
50
:
313
–20.