Disclosures The authors have disclosed that they are full-time employees of Lundbeck at the time of the study.
Delphine Saragoussi, H. Lundbeck A/S, Global Outcomes Research Division, 37-45 Quai du Président Roosevelt, 92445 Issy-les-Moulineaux Cedex, France Tel.: +33 (0)1 79 41 29 22 Fax: +33 (0)1 79 41 29 08 Email: firstname.lastname@example.org
Aims: To investigate switching patterns of major antidepressant treatments and associated factors in a primary care adult population with major depressive disorder (MDD) using data from the General Practitioner Research Database (GPRD).
Methods: A retrospective cohort study was conducted using the GPRD. The study included patients with MDD, aged [18–70], with a new prescription for amitriptyline, citalopram, escitalopram, fluoxetine, paroxetine, sertraline or venlafaxine between January 1, 2001 and September 30, 2003 and having no antidepressant prescription in the 6 months preceding index date. Switching of antidepressant treatment was defined as a prescription of a different antidepressant among all available marketed antidepressant treatment at this time (no restriction of compound) from 1 month before up to 2 months after the calculated end of the previous antidepressant treatment. Survival analysis techniques were used to describe switching of antidepressant and time to switch. Profiles of switchers were described and by-treatment analyses performed.
Results: Data from over 59,000 patients showed that 16% switched antidepressants. Seventy-two per cent of switches appeared within 3 months after treatment initiation. Within switchers, median time to switch was 53 days. Switching patients had generally a more severe psychiatric profile, including more previous episodes of depression or other psychiatric disorders. They also had a higher proportion of concurrent psychiatric disorders (especially anxiety) and concomitant prescription of anxiolytics or hypnotics. Patients initially prescribed amitriptyline were almost twice as likely to switch (27%) as patients prescribed venlafaxine (17%) or an SSRI (15%).
Conclusions: This population-based study confirmed that antidepressant switch is more likely to occur within the first 3 months of treatment and in patients with a more severe psychiatric profile. A particular attention paid to these patients within the early phase of treatment may therefore help to improve their management.
•According to international recommendations, antidepressant treatments should continue for up to 9 months after resolution of initial symptoms.
•Mainly because of the occurrence of adverse events or lack of treatment efficacy, a substantial proportion of patients stops their treatment too early or require to switch to another antidepressant.
•This real-life study, including 59,000 major depressive disorder patients from the GPRD, showed that almost three-fourths of switches occurred within the 3 months following treatment initiation.
•Switch was more frequent in amitriptyline users.
•The proportion of non-SSRIs treatments increased after a switch. However, SSRIs remained the most prescribed postswitch drugs but at a lower extent than before the switch
•The major predictive characteristics for switching were initial treatment with amitriptyline and younger age. Other predictive characteristics were concurrent psychiatric therapy, concurrent psychiatric disorder and gender.
Major depressive disorder (MDD) is a common primary care disorder with a European estimated lifetime prevalence of 12.8% (1). In the United Kingdom (UK), prescription of antidepressants by general practitioners (GPs) has steadily increased over the past 20 years (2). This upward trend was especially marked for selective serotonin reuptake inhibitors (SSRIs) and serotonin and norepinephrine reuptake inhibitors (SNRIs). These drugs became the cornerstone of MDD pharmacotherapy, with decreasing prescriptions of tricyclic antidepressants (TCAs) by GPs (3). Despite these quantitative and qualitative changes, few assessments of antidepressant utilisation patterns were performed (3–9).
International guidelines recommend that antidepressant treatment continues for four to 9 months after initial symptom resolution (10–13). But a significant number of treatment discontinuations or early terminations occur within the first 6 months (4,14–16). Most of them are related to adverse events or lack of efficacy. For example, approximately one third of patients treated for depression do not respond to initial treatment, with lack of tolerance reported as a frequent cause of treatment failure or discontinuation (17). Numerous treatment options are available to these patients. This includes dose adjustment, antidepressant combination strategies and switching between or within drug classes (18). Some clinicians favour combination of antidepressants because of ease of implementation (e.g. no titration needed) whereas others favour switching to a different antidepressant, particularly to one of a different class. The advantages of switching to a different class of monotherapy compared with combination therapy could include reduced medication costs, fewer drug interactions, better adherence and fewer consultations (19). Moreover, the landmark STAR*D pragmatic trial has highlighted a number of important findings in terms of antidepressant switch as for example a low rate of remission among switchers in second line of treatment.
In Europe, switching has only been partially studied in large real-life settings populations (17,19–23) and a better understanding of the factors associated with this strategy may (i) improve the use of antidepressants, (ii) influence second treatment choice and (iii) result in achieving higher rates of therapeutic success.
The objectives of this study were to assess switching patterns of major antidepressant treatments and associated factors in an adult population with MDD in primary care in the UK using data extracted from the General Practice Research Database (GPRD).
This study was a retrospective pharmaco-epidemiological cohort study investigating switching patterns of main antidepressant treatments in a primary care adult population diagnosed with MDD using data from the GPRD.
The UK GPRD is the world’s largest computerised database of anonymous longitudinal clinical records from primary care, comprising over 62 million patient-years worth of data collected from approximately 10 million patients since its creation in 1987 (24). The data from the GPRD are generated by GP practices throughout the UK and participating GPs are equipped with a specific software to record patient data into the database. Participating clinicians agreed to provide data for research purposes to the GPRD. Currently, data are being collected from about 4.8 million patients in 590 primary care practices. These data cover 8% of the UK population. The patient population captured in the database is broadly representative of the overall UK population. Prescriptions are directly generated by computer, thus ensuring a complete recording of prescriptions written by the GP (25). Diagnoses were coded using Read or Oxford Medical Information System (OXMIS) codes and are entered by the GP at the time of consultation on his own account. The principal information recorded, and available for analysis, consist of physician practice characteristics, patient demographics, prescriptions of medicine, clinical diagnoses, referrals to hospitals or specialists and results of laboratory tests (24,25). The external validity of the GPRD was demonstrated in several studies (26,27).
The study population was a sample of GPRD patients from 18 to 70 years old having filled at least one prescription for the following antidepressants: amitriptyline, citalopram, escitalopram, fluoxetine, paroxetine, sertraline or venlafaxine between January 1, 2001 and September 30, 2003. The first prescription of an antidepressant was defined as the index prescription and its date corresponded to the index date, which defined the entry into the cohort. In addition, a diagnosis of MDD was required within a 90-day window preceding or following the index date and, to ensure that only incident antidepressant treatments episodes were identified, all patients had to be free of antidepressant use for a minimum of 6 months before the index date.
Amitriptyline, citalopram, escitalopram, fluoxetine, paroxetine, sertraline and venlafaxine were singled out as being among the most commonly prescribed antidepressants in the UK within the study period and belong to the main three pharmacological classes available within depression care: TCAs (amitriptyline), SSRIs (citalopram, escitalopram, fluoxetine, paroxetine, sertraline) and SNRIs (venlafaxine). Antidepressant treatment exposure was stratified according to these pharmacological classes.
Multiple entries of given patients could occur more than once as long as these patients met the 6-month antidepressant prescription-free period criterion prior to the index date. Similarly, patients with a prescription of more than one antidepressant could enter more than one class cohort as long as they had met the 6-month antidepressant prescription-free period.
The primary focus outcome of the study was made on the switch of antidepressant pharmacotherapy. The switch of antidepressant was defined as the prescription of a different antidepressant (‘post-switch’ treatment) during a timeframe of 1 month before and up to 2 months after the calculated end of the initial treatment (‘pre-switch’ treatment). The end of the initial treatment prescription was calculated and determined using the quantity of product prescribed and the daily number of pills. Time to switch was the duration from initiation of the preswitch antidepressant to initiation of the postswitch antidepressant. Early switch was defined as a switch occurring during the first 3 months after treatment initiation. Although preswitch treatments were restricted to the seven selected antidepressants of interest, postswitch treatments were not restricted to these, and were classified as: SSRIs, venlafaxine, TCAs and ‘others’ (e.g. N06AX codes according to the World Health Organization ATC classification, excluding venlafaxine).
Although several switches can occur within a depressive episode, a focus was made on the first switch only, i.e. between the index prescription and the second antidepressant prescribed within the same episode.
Switchers and non-switchers were compared in terms of (i) age, gender, region of practice of the physician, (ii) concurrent or previous history of MDD and/or other psychiatric disorders (anxiety disorders, alcohol dependence, drug dependence, bipolar disorders, schizophrenia and other psychotic disorders, obsessive-compulsive disorders, conversion disorders, anorexia nervosa, bulimia nervosa), (iii) severity of the current episode of MDD, rated ‘mild’, ‘moderate’ or ‘severe’, according to a published GPRD depression grading classification (28), and (iv) concurrent psychiatric treatment (antidepressants, anxiolytics, hypnotics, antipsychotics and antimanics).
Descriptive analyses were performed using summary statistics. Categorical variables were expressed as counts and percentages, and continuous data as means and standard deviations. Comparisons were assessed using chi-squared test for categorical variables and Student t-test for continuous variables.
Kaplan–Meier methods were used to describe cumulative probability curves of time to switch. The identification of the major significant predictors of switching antidepressant treatments was performed including all potential factors described in the literature in a multivariate Cox proportional hazard model using a stepwise procedure (threshold of entry p ≤ 0.05, removal p ≤ 0.05). As the primary variable of interest, the treatment class was forced into the final model. The effect of each predictor on the risk of switching was expressed as adjusted hazard ratios (HR) and corresponding 95% confidence intervals. Statistical analyses were performed using the sas software version 9.1 (SAS Institute Inc; Cary, NC, USA).
The study protocol was approved by the GPRD’s Scientific Ethical and Advisory Group.
Characteristics of depressive episodes
Table 1 presents the characteristics of episodes according to the occurrence of an antidepressant switch. 59,389 patients with a total of 65,613 eligible depressive episodes starting within the study window were selected. Index treatment was a SSRI in 59,356 episodes (90.5%), venlafaxine in 3259 episodes (5.0%) and amitriptyline in 2998 episodes (4.6%). A switch from the index prescription occurred for 16% (10,314) of these episodes. The proportion of switch was almost twice higher in episodes which initial treatment was amitryptiline (27.4%) compared with episodes initiated with a SSRI (15.1%) or venlafaxine (16.8%). The proportion of women was slightly lower in the switch group than in the non-switch group (67.9% vs. 69.5%, p = 0.001). However, age did not differ between switch and non-switch groups. In 45.8% of the episodes in the switch group, the index MDD diagnosis was the first one compared with 49.6% of the episodes in the non-switch group (p< 0.0001); Moderate and severe episodes represented 44.7% of episodes in the switch group compared with 42.5% in the non-switch group (p < 0.0001). In the switch group, the frequency of previous history of anxiety disorders was 29.8% compared with 25.0% in the non-switch group (p < 0.0001). Moreover, switchers showed a markedly higher frequency of concurrent psychiatric disorders and especially concurrent anxiety disorders (27.3% vs. 16.8%, p < 0.0001). More frequent concurrent psychiatrics treatments, particularly anxiolytics and hypnotic treatments, were observed in the switch group (28.7% vs. 12.6%, p < 0.0001).
Table 1. Description of the depressive episodes with switch (n = 10,314) or without (n = 55,299)
All (n = 65,613)
Switch (n = 10,314)
No switch (n = 55,299)
Otherwise indicated, results are presented as n (%). *p-value from Student t-test. †p-value for ‘First episode vs. Previous episode(s)’. ‡p-value for ‘Mild vs. Moderate or Severe’. §Patients could have more than one concurrent psychiatric treatment. MDD, major depressive disorder; SNRI, serotonin and norepinephrine reuptake inhibitor; SSRI, selective serotonin reuptake inhibitor; TCA, tricyclic antidepressant.
Index antidepressant treatment
Mean ± SD (years)
39.2 ± 12.7
39.2 ± 12.6
39.2 ± 12.7
< 40 years old
Region of practice
Eastern regional office
London regional office
North West Regional Office
Northern and Yorkshire regional office
South East regional office
South West regional office
Trent regional office
West Midlands regional office
Previous history of MDD
One previous episode
Higher than one previous episode
Severity of the index MDD diagnosis
Previous history of psychiatric disorder(s)
Other than depression
Anxiety disorders only
Concurrent psychiatric disorders
Anxiety disorders only
Concurrent psychiatric treatment§
Characteristics of switches
The cumulative probability of switch was estimated using the Kaplan–Meier method. Comparison of survival curves stratified according to the index treatment class showed statistical significant differences between treatments (Log-rank p-value = 0.001) (Figure 2). At 3 months, the percentage of patients who had not switched from their initial treatment was 86.0% [85.7–86.3%] for patients treated with SSRIs, 84.2% [82.7–85.7%] for those treated with venlafaxine and 62% [59.6–64.4%] for patients treated with amitriptyline. These proportions remained quite stable (79.7%, 78.8% and 55.7%, respectively, at 6 months), indicating that most switches occurred within the first 3 months after treatment initiation and were more frequent in the amitriptyline group.
The predictors of switch were identified using the Cox proportional hazards model with treatment class forced in the model and adjusted for other baseline predictors (Table 2). Initiating amitriptyline treatment at index date was associated with an increased probability of switching compared with initiating a SSRI treatment [HR = 2.55 (2.37–2.74), p < 0.0001], whereas initiating venlafaxine treatment was not [HR = 1.04 (0.96–1.14), p= 0.36]. Increasing age was associated with a decreasing probability of switching, as compared with the 18–30 years reference class [from HR = 0.86 (0.81–0.90) for 30–40 years to HR = 0.69 (0.63–0.75) for 60–70 years, p < 0.0001]. Finally, the presence of concomitant psychiatric pharmacotherapies and concomitant psychiatric disorders was associated with a significant risk of switching antidepressant treatment [respectively, HR = 1.24 (1.18–1.30), p < 0.0001; and HR = 1.13 (1.08–1.19), p < 0.0001].
Table 2. Cox proportional hazards model for predicting switch
Hazard ratio [95% CI]
MDD, major depressive disorder; SNRI, serotonin and norepinephrine reuptake inhibitor; SSRI, selective serotonin reuptake inhibitor; TCA, tricyclic antidepressant.
Index antidepressant treatment
Region of practice
West Midlands regional office
Northern & Yorkshire regional office
Personal history of MDD
Severity of the first episode
Moderate or severe
Personal history of psychiatric disorder(s)
Concurrent psychiatric disorder(s)
Concurrent psychiatric therapy(ies)
Among patients who switched, median time to switch to any antidepressant treatment was estimated at 53 days. Median time to switch was estimated at 55 days for SSRIs users, 39 days for venlafaxine users and 42 days for amitriptiline users. The proportion of early switches was higher among amitriptiline users as compared with SSRI or venlafaxine users (n = 729/821, 88.8%; n = 6328/8946, 70.7%; and n = 405/547, 74.1%, respectively, p < 0.0001) (Figure 3).
Independently of the index treatment class, most treatments were switched to a SSRI although the proportion of non-SSRI antidepressants as postswitch treatments was increased (Figure 4). Among episodes initiated with SSRIs, 50.3% switched to another SSRI, 24.1% to venlafaxine, 23.5% to a TCA and 2% to another class of antidepressant. Of the episodes initiated with venlafaxine, 75.1% switched to SSRIs, 22.5% to TCAs and 2.4% to ‘others’. Among episodes initiated with amitriptyline, 76.9% switched to SSRIs, 16.7% to another TCAs, 5.0% to venlafaxine and 1.3% to ‘others’.
Whereas the pharmaco-epidemiological studies of antidepressants in Europe mostly focused on the safety associated with their use, few studies have been published on antidepressant utilisation and treatment patterns of MDD (3–9). Using the GPRD, the present study analysed the real-life characteristics of antidepressant switch over 59,000 MDD patients in the real-life UK setting. The results showed that about 1/6 of patients switched, most of them within the first 3 months following treatment initiation. During the study period, SSRIs were already the most prescribed drugs as preswitch or postswitch antidepressant treatment (although at a lower extent after switch, with a higher representation of non-SSRIs treatment), as later recommended by the NICE in 2004 for the management of depression. The major predictive characteristics for switching were initial treatment with amitriptyline and younger age. Other predictive characteristics were concurrent psychiatric therapy, concurrent psychiatric disorder and gender.
The overall clinical profile of switching patients showed that they presented a more severe clinical profile: multiple history of MDD episodes, frequent psychiatric comorbidities, concomitant psychiatric treatment and more severe current MDD episode. The particularly higher switching rate in patients with concurrent psychiatric treatments such as anxiolytics, hypnotics or antipsychotics has previously been reported in the literature (29,30). In addition, increasing age was found to be associated with a decreasing probability of switching. One hypothesis could be that an increased compliance to treatment is observed with increasing age. Another hypothesis could be that lower doses are prescribed in older patients, thus limiting the number of adverse reactions that could lead to a switch.
In the US, the published global switch rates varied from 3.3% (31) to 8.6% (32). However, higher proportions of switch were also observed for initial TCA users with switch rates from 13.5% (33) to 26.6% (32). In Europe, the switch rates were reported to be 3.8% in the UK (34), 4% in the Netherlands (35) and 6.4% in Sweden (36). These rates are below those observed in the present study in which switch rates were estimated to 15% for SSRI users, 17% for SNRI users and up to 27% for TCA users. These differences are substantial but may derive from the diversity in observation periods, studied countries, type of database and inclusion criteria. Most of the previously published studies were performed in the mid 1990s, in other countries than the UK, with different data sources and study-specific inclusion criteria. In the same time period, the evolution of depression management and guidelines, and the marketing of new antidepressants in the 2000s, could have played a significant role in the differences observed with the present results.
Although the range of switch rates differed from those previously reported, the results of the present analysis confirmed the higher proportion of antidepressant treatment switch following the initiation of a TCA such as amitriptyline. This finding is supported by database analyses performed in Sweden (7% for TCAs vs. 5% for SSRIs) (36) and in the US (32,33). The present results also confirmed that initiating a TCA such as amitriptyline was associated with shorter treatment duration than initiating an SSRI or an SNRI (37–39). These data might be related to the fact that TCAs are more likely prescribed to patients with moderate or severe depression, who are at higher risk of switching, although statistical analyses have been designed to reduce this confounding bias. Considering the specific safety profile of TCA antidepressants such as amitriptyline, the role of adverse events in switching from amitriptyline might also be taken into consideration and could participate in the higher frequency of treatment switch as compared with SSRIs or venlafaxine users. This supports the major role of tolerance in the discontinuation of antidepressant treatments (17). In addition, the titration of TCA treatments as amitriptyline, frequently needed, could have impacted user’s compliance and favoured a switch to other recent antidepressant such as SSRIs. However, data available in the GPRD did not allow extrapolation of the reasons for treatment change.
The occurrence of switches within the first 3 months of treatment is consistent with current national guidelines on the treatment of depression, which recommend to increase dose, switch to another antidepressant or a psychotropic drug, or add a second antidepressant in case of inadequate response to treatment within 4–8 weeks of pharmacotherapy (10–12,40). However, to avoid relapse, national guidelines also recommend that antidepressants be continued 4–9 months after symptom resolution, i.e. after 2 or 3 months of treatment. As also suggested by Claxton et al. (34), early switch may therefore be responsible for a higher frequency of relapse or recurrence of depression, and, as a consequence, for increased costs associated with antidepressant treatment switch.
Studies based on ‘real-life’ observations are a useful tool providing data that cannot derive from clinical trials. The main strength of this study resides in the use of the GPRD database, which allowed extracting and analysing data from almost 60,000 patients followed in primary care routine practice. The GPRD is the largest and most validated medical records database in Europe and was created in 1987. It allows long periods of follow-up and observations on patients in a real-life setting. The accuracy of prescription data contained in this database is an asset for drug utilisation analyses such as the present study.
The limitations of the present study are similar to those of many database analyses. The major limitation was that patients with MDD were selected based on the diagnosis reported by the physician. It was not indicated whether the diagnosis was made using a validated depression scale or only based on the physician’s clinical judgment, so that reliability of the diagnoses could not be ascertained. Moreover, as diagnoses are entered by GPs at the time of consultation, on their own account, there may be cases where depressed patients did not have a recorded diagnosis, thus creating an underestimation of cases. This may have an impact on switch rates observed in this study. In addition, somatic comorbidities were not considered in this analysis. The analyses are also subject to the information available in the database, and unmeasured confounding cannot be ruled out.
Moreover, because of limited clinical data collected in the GPRD, the clinical description of switchers could not be exhaustive. Consequently, clinical progress of switchers could not be assessed. This study is also based on specific UK data extracted from the GPRD which might not be fully generalised to other countries with different healthcare systems. In addition, this study is based on data collected from 2001 to 2003; this should be taken into account as regards the interpretability of data over time.
The results of this study indicate that the prevalence of antidepressant switch is probably higher than previously estimated in the UK. The SSRI use was the dominant drug strategy prescribed both as pre or postswitch treatment. However, this study was not designed to bring any results regarding the factors driving the choice of one drug class compared with another. We can only speculate that this is because of the efficacy and tolerability profile of the drugs, as stated in later treatment guidelines. This study confirmed that the probability of switching was mainly associated with age and antidepressant class. The psychiatric profile of depressed patients was also found to be a major factor. The probability to switch was increased in patients with more severe depressive episodes and in those with concomitant psychiatric disorder(s) or concomitant psychiatric pharmaco-therapies. These data suggest that a better knowledge of the clinical profile of patients switching to another antidepressant would be relevant in improving the management of MDD patients.
This study was funded by H. Lundbeck A/S.
DS participated in: concept/design, data analysis/interpretation, drafting article, critical revision of article and approval of article. JC participated in: concept/design, data analysis/interpretation, drafting article, critical revision of article and approval of article. SB participated in: concept/design, data analysis/interpretation, statistics, drafting article, critical revision of article and approval of article. YC participated in: data analysis/interpretation, statistics, critical revision of article and approval of article. DM participated in: concept/design, critical revision of article and approval of article.