Early discontinuation and suboptimal dosing of duloxetine treatment in patients with major depressive disorder: analysis from a US third-party payer perspective

Abstract

Objective:

Little is known about early discontinuation of duloxetine therapy or the effect that initial dose has on discontinuation in patients with major depressive disorder (MDD).

Methods:

Data from a private payer insurance claim database included 6132 patients with MDD who started duloxetine between 7/1/2005 and 6/30/2006, had no prescription for duloxetine in the previous 3 months, and were enrolled for 12 months before and after initiation. Chi-square tests, t-tests, and logistic regression were used to compare demographic, clinical, and healthcare cost data stratified by length of continuation. Early discontinuation was defined as continuation ≤30 days. Healthcare costs, persistence, and adherence were compared between patients with suboptimal initial dose (<40 mg/day) and those with recommended initial dose (40–60 mg/day).

Results:

Discontinuation rates were 16.8% at ≤30 days, 16.7% at 31–90 days, 14.9% at 91–180 days, and 51.6% at >180 days. Suboptimal initial dose, younger age, male gender, prior benzodiazepine use, insomnia, psychiatric disorders, infectious diseases, digestive disorders, genitourinary disorders, and injury/poisoning increased the likelihood of early discontinuation (Odds ratios [ORs]: 1.18–2.16), while recent use of SSRIs or venlafaxine XR decreased the likelihood (ORs: 0.67–0.68). Compared with patients who persisted with therapy for >180 days, patients who discontinued early had more hospital admissions, longer hospital stays, and more ER visits during the 1-year follow-up (all p-values <0.01). Patients with an initial dose <40 mg/day had shorter persistence (p < 0.001) and lower rates of adherence (p < 0.001) compared with patients with an initial dose of 40–60 mg/day.

Limitations:

Limitations of this study were those of all analyses based on data from insurance claim databases.

Conclusions:

Early discontinuation was associated with increased healthcare utilization. Demographic and clinical predictors of early discontinuation were identified that may help target care for at-risk patients. Beginning therapy within the recommended dose range may improve persistence.

Key words::

• Major depressive disorder
• Attrition
• Duloxetine
• Predictor
• Health outcome

Introduction

Major depressive disorder (MDD) is among the most prevalent and costly psychiatric disorders. In a given year, roughly 14.8 million American adults, or ∼6.7% of the US population aged 18 and older, will experience depression1. Antidepressants are the mainstay of treatment for MDD, and antidepressant treatment using an adequate dose for a sufficient duration is critical for symptom improvement2. Generally, it is recommended that patients continue antidepressant treatment for 4–9 months to achieve full recovery and to reduce the likelihood of relapse3.

Discontinuation of antidepressant therapy within the first 12 weeks of therapy is common, and has been reported at rates of 20–40% in randomized clinical trials4–9. Patients who discontinue treatment early have higher rates of relapse and greater healthcare costs than patients who persist with treatment10–13. The identification and reduction of factors leading to early discontinuation have important implications for patients, providers, and healthcare payers. One primary contributor to early discontinuation is lack of efficacy, which can be related to initiation of therapy using a suboptimal dose14.

While several studies have examined the demographic and pre-treatment clinical predictors of early discontinuation during treatment with selective serotonin reuptake inhibitors (SSRIs)9,15,16, little is known about early discontinuation of duloxetine, a second-line antidepressant with proven efficacy, safety, and tolerability5,17,18. Duloxetine is currently recommended for treatment of MDD with an initial dose range of 40–60 mg/day, at a range of 40–60 mg/day during the acute phase of the illness, and at a maintenance dose of 60 mg/day extending for at least 6 months19. Initiation of treatment at a suboptimal dose, that is at a dose of <40 mg/day, is not uncommon20,21, but the extent to which this practice contributes to poor persistence with or adherence to therapy is not known.

The objectives of this analysis were to examine pre-treatment demographic and clinical predictors of early discontinuation from duloxetine treatment in patients with MDD treated in a typical clinical setting, and to assess healthcare outcomes of early treatment discontinuation in this population. Also, because a low initial dose of duloxetine was a modifiable risk for early discontinuation, demographic and clinical characteristics of patients whose initial dose of duloxetine was suboptimal were also examined, as were economic outcomes for this sub-group of patients.

Methods

Data source

This was a retrospective analysis of medical, pharmacy, and enrollment data from the PharMetrics Integrated Outcomes Database, a private payer insurance claims database comprised of claims from 58 million commercially insured patients enrolled in more than 70 different managed care organizations. All records considered for inclusion were from those patients aged 18–64 years who received a first prescription for duloxetine between 7/1/2005 and 6/30/2006. Relative to the index prescription, patients also had to have ≥2 outpatient or ≥1 inpatient claims with a diagnosis code for MDD (International Classification of Diseases, Ninth Revision, Clinical Modification [ICD-9-CM] codes: 296.2 and 296.3) in the prior year, no prescription for duloxetine in the prior 6 months, and continuous enrollment in the previous 12 months and the following 12 months.

Definitions

Based on work done by Warden et al.4, early discontinuation of treatment was defined as discontinuation within 30 days of the initial prescription. Patients who met this criterion comprised the Early Discontinuation cohort. Three other treatment discontinuation cohorts were identified: those who discontinued treatment between days 31 and 90, days 91 and 180, and beyond day 180. Doses of duloxetine that were <40 mg/day were defined as sub-optimal, and patients whose initial dose was <40 mg/day comprised the Suboptimal Initial Dose cohort. Those whose initial dose was in the range of 40–60 mg/day comprised the Recommended Initial Dose cohort.

A measure of adherence, the medication possession ratio (MPR) was calculated as the fraction of days the medication was taken in the first 6 months after initiation. Patients were defined as adherent if they had an MPR of ≥0.8 in the 6 months following the index dose. Persistence was defined as the length of therapy without exceeding a 15-day gap in the 6 months following the index dose.

Variables

Demographic variables included age, gender, geographic region, health plan type, and provider type. Clinical variables included other diagnoses identified from claims for service related to 17 categories of systemic disease from ICD-9-CM codes submitted in the year prior to the index dose. Selected psychiatric, chronic pain, and sleep disorders were examined in greater detail, as was use psychotropic medication and medication to treat insomnia and pain. Itemized healthcare expenses in the 6 months before and 12 months following duloxetine initiation were also examined, as was healthcare utilization (inpatient stays, ER visits, office visits, home visits, other outpatient services, and lab services) in the 12 months following duloxetine initiation.

Statistical analysis

Patients were stratified into cohorts based on whether they discontinued treatment within the first 30 days, between 31–90 days, between 91–180 days, or after 180 days. Baseline characteristics, prescriber specialty, prior healthcare costs, and initial duloxetine dose; presence of systemic illness; rates of specific psychiatric disorders and pain conditions; and prior use of psychiatric and pain medication were compared between these cohorts using chi-square tests for categorical variables and t-tests for continuous variables. Due to the skewness of some continuous variables, a non-parametric bootstrap analysis was conducted to confirm the t-test results.

Using data from the cohort that persisted beyond 180 days as the reference cohort, multiple logistic regression analysis was used to identify associations between early discontinuation and the following variables: baseline demographics, prescriber specialty, systemic and psychiatric comorbidities, prior medication use, healthcare utilization, healthcare costs, and initial duloxetine dose. Results of the logistic regression analysis were evaluated for consistency with a model of association derived from ordinal response logistic regression in which all four discontinuation cohorts were included.

Healthcare cost and utilization outcomes in the 12 months following initiation of duloxetine were examined between these cohorts using t-tests and non-parametric bootstrapping. Group comparisons were not meant to show causality and, for that reason, no adjustment for baseline differences between cohorts was incorporated in the analysis. A sensitivity analysis was conducted using only data from those patients with at least one refill of duloxetine.

Because initial duloxetine dose was found to be associated with treatment duration and, importantly, because this factor can be modified at the discretion of the clinician, further assessments of initial dose were conducted. Demographic characteristics; the incidence of systematic diseases, specific psychiatric disorders, and pain disorders; and prior resource use and healthcare costs were compared between patients in the Suboptimal Initial Dose and Recommended Initial Dose cohorts. Healthcare cost and utilization outcomes were also compared between dosing cohorts using t-tests and non-parametric bootstrapping. MPR and the percentage of patients who were adherent to therapy in the 6 months after the initial duloxetine dose were compared using t-tests and chi-square tests. Persistence was compared using Kaplan-Meier analysis. All tests of difference in this analysis were considered significant at the p < 0.05 level.

Results

A total of 6132 patients met inclusion criteria for this analysis. Patients discontinued duloxetine within the first 30 days of treatment at a rate of 16.8% (1033/6132), and slightly more than half of patients persisted with treatment for >180 days (51.6%; 3167/6132).

Demographic and clinical characteristics

Univariate summary statistics for baseline characteristics, prescriber specialty, prior healthcare costs, and initial duloxetine dose; presence of systemic illness; rates of specific psychiatric disorders and pain conditions; and prior use of psychiatric and pain medication are shown by time of discontinuation in Tables 1–4. The ratio of females to males was ∼3:1 for all groups. Patients in the >180 day cohort were significantly older than those who discontinued early, both by age cohort (p < .001) and by mean age (47.2 vs 44.2, p < 0.001). Nearly half of patients in each cohort were enrolled in a PPO, and HMOs were the second most common heath plan type. Proportionately greater numbers of patients in the Early Discontinuation cohort were from the East or South, rather than the Midwest or West compared to patients in the >180 day cohort (p < 0.001). Duloxetine was prescribed by a psychiatrist for roughly half of patients in each cohort (range of 48.1–51.4%), with no statistical differences noted between cohorts. In the 6 months preceding the initiation of duloxetine, patients who discontinued treatment early had spent less on prescriptions (p < 0.001) and more on emergency services (p = 0.002) than patients who had continued treatment beyond 180 days (Table 1).

Table 1.  Characteristics of patients with major depressive disorder treated with duloxetine, stratified by time of discontinuation. Demographic, prior healthcare cost, and initial dosing characteristics.

	Discontinuation cohort				Comparison
	≤30 days (n = 1033)	31–90 days (n = 1021)	91–180 days (n = 911)	> 180 days (n = 3167)	≤30 vs 31–90 days	≤30 vs 91–180 days	≤30 vs >180 days
Female gender, (%)	72.3	71.7	73.3	75.5	0.76	0.62	0.04
Age cohort, (%)					0.94	0.98	<0.001
18–25 years	8.4	7.4	8.5	5.0
26–35 years	14.3	14.9	13.5	9.5
36–45 years	25.6	25.4	25.1	24.3
46–55 years	33.9	34.1	35.1	37.6
56–65 years	17.8	18.2	17.8	23.7
Age in years, mean (SD)	44.2 (11.5)	44.7 (11.3)	44.7 (11.5)	47.2 (10.6)	0.35	0.31	<0.001
Health plan type, %					0.05	0.01	0.06
HMO	24.6	27.9	24.8	25.1
Indemnity	7.6	6.8	7.1	6.1
POS	17.3	14.6	14.3	17.1
PPO	46.0	48.0	51.5	48.7
Other	4.5	2.7	2.3	3.0
US Region %					0.05	0.03	<0.001
Midwest	36.0	39.9	39.5	42.3
East	26.8	21.2	21.8	19.4
South	22.3	21.4	19.8	17.8
West	15.9	17.6	18.9	20.5
Physician specialty, %					0.89	0.43	0.09
General/family	11.5	12.0	10.1	9.8
Internal medicine	4.3	4.1	4.0	3.4
Mental health	9.4	10.7	11.2	11.2
Psychiatrist	49.0	48.1	51.0	51.4
Other physician	25.9	25.2	23.7	24.1
Prior healthcare costs in US dollars, mean (SD)
Total	17,135 (27,095)	18,083 (34,767)	16,598 (28,677)	16,667 (28,871)	0.47	0.69	0.66
Pharmacy	3516 (5,725)	3427 (5323)	3378 (4,705)	4359 (6,655)	0.74	0.62	<0.001
Medical	13,619 (24,688)	14,656 (33,352)	13,220 (27,580)	12,308 (26,751)	0.40	0.75	0.19
Office visits	2261 (3,183)	2071 (2601)	2073 (2,395)	2078 (2,589)	0.11	0.12	0.06
ER services	702 (1,910)	703 (2,143)	569 (1,632)	496 (1,872)	0.99	0.12	0.002
Hospitalization	5886 (19,050)	7426 (29,875)	6087 (23,776)	5341 (21,507)	0.13	0.85	0.51
Home visits	148 (668)	138 (785)	128 (847)	166 (842)	0.77	0.58	0.53
Lab services	324 (939)	272 (492)	316 (626)	325 (1,321)	0.27	0.88	0.98
Other outpatient	4299 (8,058)	4046 (7,431)	4046 (7,628)	3902 (8,055)	0.47	0.48	0.16
Initial duloxetine dose, %	35.5	29.1	26.9	25.4	0.002	<0.001	<0.001
20–30 mg/day	59.3	63.6	65.8	67.8
40–60 mg/day	5.1	7.4	7.4	6.8
>60 mg/day

HMO, health maintenance organization; POS, point of service; PPO, preferred provider organization; SD, standard deviation.

Table 2.  Characteristics of patients with major depressive disorder treated with duloxetine, stratified by time of discontinuation. Systemic disease (%).

	Discontinuation cohort				Comparison
	≤30 days (n = 1033)	31–90 days (n = 1021)	91–180 days (n = 911)	>180 days (n = 3167)	≤30 vs 31–90 days	≤30 vs 91–180 days	≤30 vs >180 days
Pregnancy/birth complication	2.1	2.7	1.7	1.5	0.37	0.44	0.13
Congenital	5.7	3.6	4.6	4.9	0.03	0.28	0.32
Blood	13.1	11.6	12.2	13.0	0.30	0.56	0.94
Neoplasms	20.9	20.4	23.4	21.8	0.76	0.19	0.54
Infectious	26.0	22.5	19.8	21.0	0.06	0.001	<0.001
Skin	30.9	35.1	29.8	31.8	0.04	0.59	0.58
Digestive	41.7	40.6	37.8	37.9	0.59	0.08	0.03
Circulatory	42.4	42.3	41.7	45.0	0.97	0.76	0.14
Injury/poisoning	47.6	45.1	43.0	41.5	0.24	0.04	<0.001
ENMIDs	49.4	51.7	54.7	54.7	0.29	0.02	0.003
Genitourinary	51.2	46.3	47.8	46.7	0.03	0.13	0.01
Nervous	54.7	53.8	51.3	54.6	0.67	0.13	0.94
Respiratory	58.1	55.7	56.8	53.8	0.28	0.55	0.02
Musculoskeletal	70.0	66.5	68.3	69.5	0.09	0.41	0.75
Mental	80.7	78.8	74.6	74.9	0.29	0.001	<0.001
Ill-defined condition	84.3	83.3	82.8	81.9	0.51	0.36	0.07

ENMIDs, Endocrine, nutritional and metabolic, immunity disorders.

Table 3.  Characteristics of patients with major depressive disorder treated with duloxetine, stratified by time of discontinuation. Specific psychiatric illness and pain disorders (%).

	Discontinuation cohort				Comparison
	≤30 days (n = 1033)	31–90 days (n = 1021)	91–180 days (n = 911)	>180 days (n = 3167)	≤30 vs 31–90 days	≤30 vs 91–180 days	≤30 vs >180 days
Schizophrenia	1.1	2.1	1.5	1.2	0.07	0.36	0.73
Neuropathic pain	2.4	2.0	1.5	2.8	0.48	0.17	0.47
Rheumatoid arthritis	2.9	2.7	1.8	2.8	0.83	0.10	0.87
Alcohol dependence	4.7	6.5	5.1	4.3	0.09	0.76	0.54
Obstructive sleep apnea	5.7	5.2	4.6	6.1	0.60	0.28	0.65
Hypersomnia	6.3	7.7	6.6	8.3	0.20	0.79	0.03
Organic psychosis	7.4	6.6	5.6	5.8	0.48	0.12	0.06
Bipolar disorder	12.3	12.3	9.4	11.3	0.97	0.04	0.37
Insomnia	12.7	11.3	11.6	9.8	0.32	0.48	0.008
Osteoarthritis	13.7	12.5	14.3	15.6	0.46	0.69	0.12
Drug dependence	17.8	16.8	14.5	12.8	0.52	0.05	<0.001
Dysthymic disorder	18.0	17.4	15.0	16.0	0.73	0.08	0.12
Fibromyalgia	21.2	16.9	16.8	18.4	0.01	0.01	0.05
Headaches	27.6	28.2	26.0	24.9	0.76	0.44	0.09
Lower back pain	41.9	37.4	37.7	37.1	0.04	0.06	0.005
Anxiety disorders	42.2	42.7	40.1	36.9	0.82	0.34	0.003

Table 4.  Characteristics of patients with major depressive disorder treated with duloxetine, stratified by time of discontinuation. Psychotropic and pain medications (%).

	Discontinuation cohort				Comparison
	≤30 days (n = 1033)	31–90 days (n = 1021)	91–180 days (n = 911)	>180 days (n = 3167)	≤30 vs 31–90 days	≤30 vs 91–180 days	≤30 vs >180 days
MAOIs	0.5	0.3	0.0	0.3	0.49	0.04	0.43
Lithium	3.6	5.0	2.2	3.4	0.11	0.07	0.83
Typical antipsychotics	4.0	2.8	4.8	3.4	0.16	0.35	0.40
Non-narcotics	6.3	5.3	4.1	5.3	0.33	0.03	0.21
Anxiolytics	10.3	12.6	9.7	10.8	0.09	0.66	0.63
Psychostimulants	13.1	15.1	15.2	16.0	0.19	0.19	0.02
TCAs	13.9	11.7	12.1	13.7	0.12	0.22	0.83
Venlafaxine	19.0	21.2	21.1	24.6	0.22	0.25	<0.001
Atypical antipsychotics	20.4	22.2	21.8	20.2	0.32	0.44	0.90
Muscle relaxants	26.7	24.9	27.1	24.5	0.34	0.85	0.15
Anticonvulsants	30.3	32.8	30.1	31.3	0.22	0.92	0.56
Anti-inflammatory	31.9	33.0	33.2	33.7	0.58	0.54	0.28
Hypnotics	33.5	32.5	32.9	31.1	0.64	0.79	0.16
SSRIs	57.6	60.8	60.9	63.4	0.14	0.14	<0.001
Benzodiazepines	58.7	59.8	54.0	54.7	0.62	0.04	0.02

MAOIs, monoamine oxidase inhibitors; SSRIs, selective serotonin reuptake inhibitors; TCAs, tricyclic antidepressants.

Compared to patients in all other discontinuation cohorts, patients who discontinued treatment early were more likely to have been prescribed an initial dose of duloxetine that was suboptimal (36% vs 25–29%) and less likely to have had an initial dose of 40–60 mg/day or >60 mg/day (59% vs 64–68% and 5% vs 7%, respectively). All comparisons by dose were significant at the p ≤ 0.002 level (Table 1).

Compared to patients who continued treatment for >180 days, patients who discontinued early were more likely to have had infectious diseases (p < 0.001); diseases of the respiratory (p = 0.02), digestive (p = 0.03), genitourinary (p = 0.01), to have been injured or poisoned (p < 0.001); and to have had a mental disorders other than MDD (p < 0.001) in the 12 months preceding the index dose of duloxetine. They were also less likely to have had endocrine/metabolic/nutritional disease, or immune system disease (p = 0.003) (Table 2).

Rates of comorbid psychiatric and pain disorders were also higher for patients who discontinued treatment early compared to those who persisted with treatment beyond 180 days. Specifically, higher rates of insomnia (p = 0.008), drug dependence (p < 0.001), lower back pain (p = 0.005), and anxiety (p = 0.003) were noted. They were less likely to have issues with hypersomnia (p = 0.03) (Table 3).

In the year preceding the index dose of duloxetine, patients in the Early Discontinuation cohort had greater use of benzodiazepines (p = 0.02) and lesser use of psychostimulants (p = 0.02), venlafaxine (p < 0.001), and SSRIs (p = 0.001) compared to patients who continued treatment for > 180 days (Table 4).

Logistic regression

Results of the multiple logistic regression analysis are shown in Figure 1. The following illnesses were identified as predictors of early discontinuation: a history of injury or poisoning, digestive disease, infectious disease, psychiatric illness other than MDD, fibromyalgia, and insomnia (odds ratios [ORs] range from 1.26–1.36). Having been treated with an SSRI, venlafaxine, or other antidepressants made early discontinuation less likely (OR range from 0.73–0.85), while benzodiazepine use made discontinuation more likely (OR 1.29). Male gender rather than female gender was a risk factor for early discontinuation (OR 0.78). Using the age cohort 18–35 years as the comparator, younger age was a significant risk factor for early discontinuation, with ORs increasing in a stepwise fashion with each comparison to an older age cohort (36–45 years, 1.39; 46–55 years, OR 1.61; 56–64 years, 1.94). Being in the youngest rather than the oldest age cohort was, in fact, the strongest predictor of early discontinuation. A regional predictor of early discontinuation was living in the Northeast portion of the US as compared to the West (OR 1.75) or the Midwest (OR 1.52). Having received an initial dose of duloxetine that was suboptimal (<40 mg/day) was also predictive of early discontinuation compared to having been started within the recommended initial dose range of 40–60 mg/day (OR 1.63), or above that range (OR 1.86). An ordinal response logistic regression analysis using the same variables produced similar results.

Figure 1.  Demographic and clinical predictors of early discontinuation (≤30 days) based on a multivariate logistic regression analysis using values from patients who continued treatment for ≥180 days as the reference. An odds ratio >1 with confidence intervals that do not cross 1 indicates a greater likelihood of early discontinuation, while an odds ratio <1 with confidence intervals that do not cross 1 indicates a lesser likelihood of early discontinuation.

Healthcare cost and utilization outcomes

In the 12 months following initiation of duloxetine, rates of hospitalization and use of ER services responded to time of discontinuation monotonically and were statistically significantly greater for patients in the Early Discontinuation cohort compared to patients who persisted with treatment for >180 days (22.3% vs 17.4% and 35.9% vs 27.0%, respectively; p < 0.001 for both). The average number of inpatient days was also significantly greater for patients in the Early Discontinuation cohort compared to the >180 day cohort (3.0 days vs 1.9 days, p < 0.001).

Component healthcare costs in the 12 months following initiation of duloxetine are shown stratified by time of discontinuation in Figure 2. Mean expenditures for hospitalization, ER services, and office visits were significantly greater for patients in the Early Discontinuation cohort compared to patients in the >180 day cohort ($6771 vs $5052, p = 0.03; $807 vs $470, p < 0.001; $2422 vs $2132, p = 0.04, respectively). While medical costs were significantly greater for patients in the Early Discontinuation cohort ($14,874 vs $11,992; p = 0.005), pharmacy costs were significantly less ($3990 vs $5506, p < 0.001), so that total costs (medical + pharmacy) were not significantly different between the two groups ($18,864 for the Early Discontinuation cohort vs $17,498 for the >180 day cohort, p = 0.22).

Figure 2.  Healthcare costs in the 12 months following initiation of duloxetine therapy in patients with major depressive disorder, stratified by time of treatment discontinuation.

Sensitivity analysis

In a sensitivity analysis which included only those patients who had refilled at least one duloxetine prescription, results of a multivariate logistic regression were similar to the original results. Suboptimal initial dose, younger age, male gender, West and Midwest region, prior benzodiazepine use, anxiety disorders, drug abuse disorders, and headache increased the likelihood of early discontinuation. Recent use of SSRIs, venlafaxine XR and TCAs, and congenital disorders decreased the likelihood of discontinuation. Also, compared to patients who persisted with therapy for >180 days, patients in this population who discontinued early had more hospital admissions, longer hospital stays, and more emergency department visits during the 1-year follow-up period. Likewise, although their pharmacy costs were less, patients who discontinued early had increased expenses related to hospitalization, ER services, medical services, and total healthcare costs in the year following initiation of duloxetine.

Analysis of suboptimal initial dose cohort

Because an initial dose of duloxetine <40 mg/day was a strong predictor of early discontinuation, and because this is a modifiable factor, further analysis of suboptimal initial dosing was undertaken. Of the 6132 patients included in this analysis, 1714 (28%) had an initial duloxetine prescription that was <40 mg/day, compared to 4008 (65%) patients whose initial dose was in the recommended range.

The Suboptimal Initial Dose and Recommended Initial Dose cohorts did not differ with regard to age, gender, or healthcare plan type. Greater proportions of patients in the Suboptimal Initial Dose group were from the West (23% vs 18%) and smaller proportions were from the South (16% vs 20%), a difference significant at the p < 0.001 level. Patients in the Suboptimal Initial Dose cohort were less likely to have been treated by a family physician or a psychiatrist and more likely to have been treated by an internal medicine physician (p = 0.007). Healthcare costs for the initial dose groups in the 12 months following initiation of duloxetine are shown in Figure 3. Patients in the Suboptimal Initial Dose group had significantly higher costs for emergency services compared to patients who were initiated at the recommended dose ($694 vs $540, p = 0.01).

Figure 3.  Healthcare costs in the 12 months following initiation of duloxetine therapy in patients with major depressive disorder, stratified by initial dosing cohort.

Table 5 shows persistence, MPR, and adherence rates for patients stratified by their initial dose group. Patients in the Suboptimal Initial Dose cohort had a shorter mean time to discontinuation (97 days vs 106 days, p < 0.001), lower MPR (53% vs 58%, p < 0.001), and a lower rate of adherence (36% vs 41%, p < 0.001) compared to patients in the Recommended Initial Dose cohort. A Kaplan-Meier curve of rates of persistence in the 6 months following initiation showed early and consistent separation of cohorts, with greater persistence seen in the Recommended Initial Dose cohort (Figure 4).

Figure 4.  Kaplan-Meier survival curves of time to discontinuation (persistence) for patients in the Suboptimal Initial Dose and Recommended Initial Dose cohorts.

Table 5.  Adherence to duloxetine treatment in patients with major depressive disorder who received an initial dose beneath the recommended range (<40 mg/day) or within the recommended range (40–60 mg/day).

	Initial dosing group
	Suboptimal (<40 mg/day) n = 1714	Recommended (40–60 mg/day) n = 4008	P Value
Persistence*, days	97	106	<0.001
Medication Possession Ratio^†, %	53	58	<0.001
Adherence^‡, %	36	41	<0.001

*Persistence = number of days from the index prescription to the earliest of the ending date of the last prescription, the date of the first gap of more than 15 days between prescriptions, or the end of the study period (6 months).

^†Medication Possession Ratio was calculated as the sum of the days’ supply of study medication within 182 days after the index date, divided by 182 days.

^‡Adherence defined as having an MPR ≥0.8.

Discussion

Predictors of early treatment discontinuation and subsequent health outcomes were assessed in this retrospective analysis of data from a large commercial health insurance claims database. Of the 6132 patients who were started on duloxetine therapy for treatment of MDD, over half continued therapy for at least 180 days, but one out of six discontinued treatment within the first 30 days. Distinct factors that significantly increased or decreased the likelihood of early discontinuation were identified within the categories of baseline demographic and clinical variables; prior systemic, psychiatric, and pain disorders; and recently used medications. During 12 months of follow-up, patients who discontinued treatment early had, on average, significantly greater use of healthcare resources and higher healthcare costs than those who persisted with treatment through study end (180 days).

Despite guidelines supporting the continued use of antidepressant therapy for 4–9 months following an acute episode of MDD, a significant proportion of patients do not reach this treatment goal. Our results confirm those of earlier studies showing discontinuation rates of 20–40% in the first 12 weeks of treatment4–9. Although our data came from commercially insured patients who were not enrolled in clinical trials, rates of discontinuation in this analysis resembled those reported in early clinical trials comparing duloxetine to placebo5,22. Discontinuation was less common for patients in this study compared to rates seen in a 6-week comparison study of duloxetine (24%) and venlafaxine (17%)23. When duloxetine (28%) was compared to paroxetine (25%)24, discontinuation rates at Week 8 of therapy fell between those seen at Weeks 4 and 12 of this study, as might be expected given the time point at which rates were measured. Although other direct comparisons between duloxetine and SSRIs have not been done, a meta-analysis of discontinuation rates from 42 studies found rates of 25% and 29%, although lack of a specific treatment duration prevents meaningful comparison to our data.

The majority of patients with MDD benefit from treatment with antidepressants, and the consequences of untreated or inadequately treated depression can be severe. Adherence to treatment is essential. There is clinical utility in knowing the baseline demographic and clinical factors that predict early discontinuation. Using this dataset, we constructed a prediction model and identified male gender, younger age, living in the Northeastern area of the US, a history of injury or poisoning, digestive disease, infectious disease, psychiatric illness other than MDD, fibromyalgia, insomnia, having used benzodiazepines, and having been prescribed an initial dose of duloxetine <40 mg/day as significant predictors of early discontinuation. These characteristics help identify patients who might benefit from additional attention and encouragement during the acute phase of treatment.

The higher rates of healthcare utilization and expense seen for patients who discontinued early in this analysis are consistent with previously reported results10–12, and highlight the importance that patient persistence with duloxetine therapy holds for third party payers and policy-makers. Even with the higher pharmacy costs incurred by patients treated for the full 180 days, lower expense related to inpatient care (both medical and psychiatric), office visits, and emergency services resulted in lower overall healthcare expense.

Many of the factors that predict early discontinuation identified in this analysis can be noted by the clinician, but not easily modified (gender, age, past medical history, etc.). The initial duloxetine dose, however, is a factor that is purely at the discretion of the clinician, and for this reason, was worthy of further exploration. Although an increase in persistence of 9 days may not appear clinically significant, when combined with improved adherence, we believe that initiating duloxetine at the recommended dose is a simple action that can be taken to improve outcomes for some patients.

We identified factors that describe patients likely to receive an initial dose <40 mg/day. Factors included living in the Western portion of the US, having been seen by an internist rather than a psychiatrist; and having had higher costs for emergency services in the 6 months prior to the first dose of duloxetine. Differences in regional prescribing patterns have been noted previously25 and may reflect differences in training or patient temperament. Patients seen by psychiatrists were likely to have differed from those seen by primary care physicians. Psychiatrists may have seen patients with more long-standing depression or depression that had been difficult to manage, while primary care physicians may have seen patients with a greater burden of medical comorbidity.

Study limitations

The limitations of this study were those of all analyses based on data from an insurance claims database. They include selection bias, the chance that results based on patients in the PharMetrics database may not apply to patients in another insurance claims database, patients who have federally funded healthcare, or who are self-pay; miscoding of information (data retrieval and coding for insurance purposes is not as carefully controlled as that gathered in clinical trials); and consequent biases in the estimations based on these data.

In addition, the daily dose based on the first prescription may not have been the initial dose taken by the patient. For example, a patient may have been started on a lower dose of duloxetine through use of office samples or instructions to cut tablets in half for a short period of time. Also, daily duloxetine doses were estimated based on the presence of claims for filled prescriptions rather than on recorded or observed consumption of medication. Finally, treatment strategies such as cross-titration of antidepressant treatment that might have explained initial dosing choices were beyond the scope of this analysis.

The database on which this analysis was based lacked detailed clinical data regarding the severity of depression, past illness history, duration of the current episode, and reason for discontinuation. Also, detailed demographic data regarding education, race, employment status, income, and family environment limited the factors that could be considered as possible predictors of early discontinuation.

Finally, no attempt was made to control for differences in baseline information when comparing outcomes by discontinuation cohort or by dosing cohort. Therefore, the conclusions of this analysis are limited to those of association rather than causation.

Conclusions

In this database, roughly half of the patients initiated on duloxetine persisted with therapy through 6 months of treatment. The sub-group of patients who discontinued therapy within the first 30 days of treatment had greater use of healthcare resources and greater healthcare expense. Multiple demographic and clinical variables were identified that could potentially be used to identify patients at risk for early discontinuation. Having received an initial dose of duloxetine less than the recommended dose (<40 mg/day) was a strong predictor of early discontinuation and one which is potentially modifiable. The frequency of early treatment discontinuation and its adverse consequences makes improved adherence a primary goal for clinicians and third party payers. Encouraging clinicians to provide an initial dose of duloxetine within the recommended dose range would be a reasonable, and likely fruitful approach.

Transparency

Declaration of funding

Financial support was provided by Eli Lilly and Company, Indianapolis, IN, USA. Employees of Lilly were involved in the study design, analysis of data, critical revision of the manuscript, and in the decision to submit the manuscript for publication.

Declaration of financial/other relationships

At the time this manuscript was written, ZC, DF, SG, DN, and XL were full-time employees of Eli Lilly and/or one of its subsidiaries and were minor stockholders of Eli Lilly and Company.

Acknowledgments

Appreciation is expressed to Tamara Ball, MD, for writing and editorial contributions. Dr Ball is a scientific writer employed full-time by i3, part of the inVentiv Health Company. Eli Lilly and Company contracted the technical writing of this manuscript with i3. Also acknowledged are: Steve Able and Baojin Zhu of Eli Lilly and Company for critical review of this manuscript, and Teri Tucker of i3 for editorial review of this manuscript.