Treatment patterns, healthcare resource utilization, and costs following first-line antidepressant treatment in major depressive disorder: a retrospective US claims database analysis

Major depressive disorder (MDD) is a complex, multifaceted psychiatric condition characterized by a variety of symptoms, including a persistent state of sadness and hopelessness, anhedonia, sleep disturbance, indecision, reduced ability to concentrate, and recurrent suicidal ideation [1, 2]. MDD is one of the most prevalent mental disorders in the United States (US) [3] and a leading cause of global disability [4]. According to the 2014 National Survey on Drug Use and Health, 6.6% of all adults, or 15.7 million people, in the US experienced at least one major depressive episode in the prior 12-month period [5]. The high prevalence and significant functional impact of MDD [6‐9] result in considerable economic consequences at all levels of society. Recent evidence shows that the total economic burden of MDD in the US has increased notably in the last decade, up to $210 billion dollars in 2010, with workplace and healthcare costs contributing almost equally (48% and 47%, respectively) followed by suicide-related costs (5%) [10].

Symptoms of MDD are often managed by appropriate psychological and/or pharmacological therapy and should be addressed throughout the depression life-course [11]. Practice guidelines recommend that treatments be individually tailored and based on disease severity and history, level of functioning, and comorbid psychiatric and physical conditions [6, 7, 12, 13]. Antidepressant drugs constitute the standard of care for MDD [6, 7, 14], whereby most patients will receive a selective serotonin reuptake inhibitor (SSRI) as first-line pharmacotherapy [15]. However, response to first-line antidepressant treatment is often not optimal [16‐19].

Management of MDD following non-optimal response to pharmacotherapy can be summarized by four key treatment steps as presented in current US guidelines: (1) dose optimization to maximize therapeutic benefit, (2) therapy switch (i.e., discontinuation of the current pharmacotherapy in favor of an alternative pharmacotherapy), (3) combination therapy (i.e., adding a second antidepressant drug to the initial one), or (4) augmentation therapy (i.e., augmenting the initial antidepressant drug with a non-antidepressant pharmacotherapy) [11, 14, 18]. Treatment regimens prescribed in the real-world setting do not always conform to practice guidelines or on-label use of treatment for MDD. Therapeutic decisions are influenced by a number of factors including safety and tolerability, adherence, and patient-physician dynamics [20]. Nonetheless, identification of the most suitable treatment for each patient as early as possible remains a key challenge and represents a barrier to the improvement of long-term treatment outcomes in MDD [7, 20].

An assessment of long-term real-world treatment patterns in MDD can provide empirical evidence on patient burden and illustrate the economic impact associated with different pharmacological approaches taken following initiation of first-line therapy. In this study, we describe long-term real-world treatment patterns among MDD patients in the US and quantify healthcare resource utilization (HCRU) and costs associated with different treatment strategies following first-line antidepressant therapy.

This retrospective study provides the first comprehensive characterization of real-world, long-term treatment patterns following initiation of first-line antidepressant therapy among patients with MDD using data from a large US administrative database. This work further represents a novel contribution to the scientific literature by reporting on the economic burden associated with recommended treatment patterns commonly observed in clinical practice in the US.

The study findings show that, in real-world practice, despite the fact that US guidelines recommend treatments to be administered for four to eight weeks before considering a patient as responsive or unresponsive to treatment, a sizable number of patients experience multiple treatment changes after a treatment duration that may not be sufficient to adequately evaluate treatment response [11, 23]. Hence, this study further underscores the need for improved long-term, personalized psychiatric management of patients aimed at yielding the best treatment outcomes possible.

Although most practice guidelines state that pharmacological therapy of patients with MDD typically includes three phases – acute phase (6–8 weeks), continuation phase (16–20 weeks), and maintenance phase for chronic and/or recurrent MDD [23] – our results show that, in real-world practice, a considerable proportion of patients do not reach the end of the continuation phase due to treatment discontinuation. Consequently, a significant number of patients could benefit from continuing therapy over a longer period of time assuming no major safety and tolerability issues are of concern.

The high rate of discontinuation in our study is consistent with the high discontinuation rate reported in a recent study by Jung et al. [24]. In this study, over 70% of patients discontinued antidepressants after 6 months (including patients who switched). While some patients may discontinue therapy due to having achieved remission, treatment side effects, inadequate patient follow-up, patient concern about medications and fear of addiction, physician-patient miscommunication, misperception of the benefits of treatment, or patient perception that treatment is no longer needed are also common reasons for discontinuation [25‐31]. Indeed, as response to first-line therapy is often less than optimal [16‐19], it is likely that the discontinuation cohort in our study includes non-responders and non-remitters. Furthermore, some of the symptoms that characterize MDD – such as difficulty planning, disorganization, distraction, and other cognitive impairments – are also likely to contribute to poor persistence on first-line antidepressant and frequent cycling on and off treatment. Although the reasons for therapy discontinuation were not available, and that it is possible that a number of patients discontinued due to successful treatment, the high discontinuation rates and the relatively short period of time within which patients discontinued their first-line antidepressant also suggest that there may be a need to improve long-term treatment persistence in some patients with MDD.

Dose escalation and switching to another antidepressant or to an atypical antipsychotic were found to be the most common treatment changes among patients who continued treatment after first-line therapy. This finding suggests a physician’s preference for antidepressant drug monotherapy versus adding or combining therapies. Although the choice of the next treatment step is likely to depend on the initial treatment response (i.e., partial vs. no response) [11, 32], a physician’s preference for monotherapy in the US could reflect an effort to avoid potential drug interactions and to maximize compliance by minimizing the number of MDD pharmacotherapies prescribed.

In terms of HCRU and costs, our results indicate that treatment patterns representing a cycling on and off treatment in the switch cohort were associated with the greatest total healthcare costs PPPY over the long-term follow-up. This may suggest that patients who commonly switch and continually cycle through different treatments might be particularly difficult-to-treat patients, thus driving medical costs up. It should, however, be noted that the treatment patterns observed in this study, along with their outcomes, may not reflect optimal treatment paths and can have grave consequences on patient responses and treatment costs. For example, it may be more appropriate to switch a patient from their first line of treatment rather than escalating their initial dose; this may, for some patients, result in lower burden and reduced long-term costs if treatment selection is appropriately determined early in the treatment path, i.e., it may be more costly and burdensome if a patient remains on a treatment that fails to be effective rather than switching to a different, more optimal, treatment [33]). Results for HCRU and costs also suggest that a relatively high proportion of the medical services and costs are not related to mental health. These results are consistent with prior studies. For example, Simon et al. (1995) reported that 66% to 71% of the total direct healthcare costs were not related to antidepressant or mental health-related visit costs [34]. These findings may reflect HCRU and costs associated with the management of other concurrent comorbidities, complications, or unrelated conditions worth evaluating in future studies. Despite the high direct healthcare costs PPPY reported in this study, these costs likely reflect an underestimate of the overall economic burden reported for MDD in the US. Indeed, a previously published US study estimated that only half of the economic burden of MDD was attributable to direct healthcare costs, with the other half attributable to worker absenteeism and presenteeism [10].

Mention of the methodological limitations of this study is warranted. First, treatment cohorts and patterns were identified based on claims for a filled prescription, which does not guarantee that the medication was consumed by the patient or account for changes in prescribed therapy. Second, a patient’s medical history may influence treatment decisions and subsequent treatment patterns and outcomes. However, disease severity, clinical characteristics and reasons underlying specific treatment patterns are not available in claims databases. Further analyses using a different data source or data linkage with patients’ medical records is warranted to better understand treatment decisions resulting in the patterns observed. Third, the study sample was limited to privately insured employees and their dependents. All patients included in our study sample were required to meet pre-determined inclusion and exclusion criteria (e.g., to have at least two billing claims for MDD within a pre-specified time period and to initiate the index antidepressant in monotherapy as first-line treatment). Accordingly, the results from our study may not be generalizable to the overall MDD patient population in the US. Fourth, our study was descriptive in nature and no statistical comparisons between treatment cohorts/patterns were conducted. However, it is possible that patients across treatment cohorts/patterns have differential disease profiles. Patients may have also undergone concomitant non-pharmacological therapy, such as psychotherapy, which may influence treatment decisions and outcomes. Non-pharmacological treatment like cognitive behavioral therapy was not evaluated as part of the current study. Further analyses are warranted to better understand reasons for specific treatment changes/patterns and to compare outcomes between treatment cohorts/patterns, including cohorts receiving concomitant psychotherapy. Fifth, the current study identifies dose escalation but did not assess whether the dosage prescribed was within the therapeutic range (a determination that would require that the dose prescribed is indeed suitable for a particular patient). Additional analyses would be needed to analyze the impact of adequate/inadequate dosing on outcomes and costs. Sixth, this study was designed to evaluate MDD patients with treatment patterns involving antidepressant and antipsychotic therapies. It is, however, possible that some patients with MDD also receive other treatments such as lithium and thyroid hormones as part of their augmentation therapy. Other therapies such as lithium and thyroid hormones, for example, were not analyzed in our study as patients augmenting with these agents may require them for the treatment of non-MDD conditions (e.g., lithium for bipolar disorder and thyroid hormones for conditions related to hormonal imbalance) – claims data do not include information to confirm the reason for which a treatment is prescribed. Finally, administrative claims databases are subject to coding errors and data omissions.