JAMA & ARCHIVES
Arch Fam Med
SEARCH
GO TO ADVANCED SEARCH
HOME  PAST ISSUES  TOPIC COLLECTIONS  CME  PHYSICIAN JOBS  CONTACT US  HELP
Institution: CLOCKSS  | My Account | E-mail Alerts | Access Rights | Sign In
  Vol. 8 No. 4, July 1999 TABLE OF CONTENTS
  Archives
  •  Online Features
  Original Contribution
 This Article
 •Abstract
 •PDF
 •Send to a friend
 • Save in My Folder
 •Save to citation manager
 •Permissions
 Citing Articles
 •Citation map
 •Citing articles on HighWire
 •Citing articles on Web of Science (37)
 •Contact me when this article is cited
 Related Content
 •Related articles
 •Similar articles in this journal
 Topic Collections
 •Psychiatry
 •Depression
 •Randomized Controlled Trial
 •Alert me on articles by topic

Long-term Outcomes of Initial Antidepressant Drug Choice in a "Real World" Randomized Trial

Gregory E. Simon, MD, MPH; John Heiligenstein, MD; Dennis Revicki, PhD; Michael VonKorff, ScD; Wayne J. Katon, MD; Evette Ludman, PhD; Louis Grothaus, MA; Edward Wagner, MD, MPH

Arch Fam Med. 1999;8:319-325.

ABSTRACT

Objective  To compare the long-term clinical, quality-of-life, and economic outcomes after an initial prescription for fluoxetine, imipramine hydrochloride, or desipramine hydrochloride.

Design  Randomized, controlled trial.

Setting  Primary care clinics of a staff-model health maintenance organization in the Seattle, Wash, area.

Patients  Four hundred seventy-one adults beginning antidepressant drug treatment for depression.

Intervention  Random assignment of initial medication (desipramine, fluoxetine, or imipramine), with treatment (dosing, medication changes or discontinuation, and follow-up visits) managed by a primary care physician.

Measurements  Interviews at baseline and at 6, 9, 12, 18, and 24 months examined medication use, clinical outcomes (Hamilton Depression Rating Scale and depression subscale of the Hopkins Symptom Checklist), and quality of life (Medical Outcomes Study SF-36 Health Survey). Medical costs were assessed using the health maintenance organization's accounting data.

Results  Patients assigned to fluoxetine therapy were significantly more likely to continue taking the initial antidepressant but no more likely to continue any antidepressant therapy. The fluoxetine group did not differ significantly from either tricyclic drug group on any measure of depression severity or quality of life. For 24 months, antidepressant drug costs were approximately $250 higher for patients assigned to fluoxetine therapy, but total medical costs were essentially identical.

Conclusions  Initial selection of fluoxetine or a tricyclic antidepressant drug should lead to similar clinical outcomes, functional outcomes, and overall costs. Differences in antidepressant prescription costs are blunted by the large minority of tricyclic-treated patients who switch to use of more expensive medications. Restrictions on first-line use of fluoxetine in primary care will probably not reduce overall treatment costs.



INTRODUCTION
 Jump to Section
 •Top
 •Introduction
 •Patients, materials, and methods
 •Results
 •Comment
 •Author information
 •References

COSTS OF newer antidepressant medications remain a source of concern and controversy. Annual US expenditures for serotonin reuptake inhibitor drugs now exceed $3 billion and continue to grow by approximately 25% per year.1

Cost concerns must be balanced against the large economic burden of untreated and undertreated depression. Depressive illness is associated with a large burden of functional impairment,2-3 decreased productivity, absenteeism, and increased use of health services.4-5 Among patients treated with antidepressant drugs in primary care, many discontinue taking the medication after only a few weeks, and few receive recommended dosage levels.6-7 Observational data7-8 suggest that use of newer antidepressant drugs may lead to greater adherence and more appropriate dosing. If use of new antidepressant drugs leads to more effective treatment, the higher purchase cost of newer drugs may be justified.

Available data do not resolve whether the advantages of taking newer antidepressant medications outweigh higher purchase costs. Results of traditional clinical trials9-11 consistently demonstrate no difference in efficacy between use of newer and older antidepressant drugs, but results of most trials9, 12-13 find that use of newer drugs leads to fewer adverse effects and lower rates of discontinuation. Several decision-analytic models14-16 based on randomized trial data conclude that newer drugs should have significant effectiveness and cost-effectiveness advantages in actual practice, although one recent reanalysis17 challenges this conclusion. A recent modeling study18 found initial treatment with fluoxetine to be more cost effective than initial treatment with imipramine hydrochloride followed by fluoxetine treatment in cases of intolerance or nonresponse. We previously reported19 short-term results from a large, primary care–based, randomized trial comparing patients initially treated with fluoxetine with those initially treated with older tricyclic antidepressant drugs (imipramine or desipramine hydrochloride). Our intent was to compare fluoxetine therapy with use of first- and second-generation tricyclic drugs. In that randomized trial, initial selection of fluoxetine led to fewer medication adverse effects and a lower likelihood of subsequent medication change but no significant differences in overall treatment adherence, short-term clinical outcomes, short-term functional outcomes, or total medical costs during the first 6 months of treatment.

This article describes long-term outcomes in that randomized trial.19 As described elsewhere,20 this effectiveness study incorporated several elements intended to increase the generalizability to "real world" primary care practice. Although such a study design is ill suited for examining treatment efficacy, it is well suited for study of an effectiveness or policy question: Will first-line use of fluoxetine lead to better clinical, quality-of-life, or economic outcomes compared with a policy of reserving fluoxetine therapy for only those patients who have not responded to treatment with tricyclic medications?


PATIENTS, MATERIALS, AND METHODS
 Jump to Section
 •Top
 •Introduction
 •Patients, materials, and methods
 •Results
 •Comment
 •Author information
 •References

Study methods are described in detail in earlier publications19-20 and are summarized herein. Patients were enrolled from selected primary care clinics of Group Health Cooperative of Puget Sound, a staff-model health maintenance organization serving approximately 400,000 members in the Seattle, Wash, area. The study protocol was approved by Group Health Cooperative's human subjects review committee. Eligibility criteria were broad in order to enroll a sample representative of patients usually treated in primary care. At participating clinics, all primary care physicians were asked to refer any adult patient beginning antidepressant drug treatment for depression if the physician and the patient would consider random assignment of the initial medication. Need for antidepressant drug treatment was based strictly on the referring physician's judgment, regardless of medical comorbidity or severity of depression. Study personnel were immediately available (on-site or by telephone) to screen referrals, obtain written informed consent, and assess the following exclusion criteria: use of antidepressant drugs in the previous 90 days, current alcohol abuse, current psychotic symptoms, history of mania, recent use of lithium or antipsychotic medication, current pregnancy, or current use of medications that might contraindicate use of any study drug.

Eligible and consenting patients were randomly assigned (using computer-generated random numbers) to begin treatment with desipramine, fluoxetine, or imipramine; randomization was stratified according to the presence or absence of current major depression, determined by structured interview.21 Neither patients nor referring physicians were unaware of treatment assignment because this would preclude typical clinical management in primary care.

All decisions regarding clinical management (initial antidepressant drug dose, dosage changes, treatment discontinuation, switch to a different antidepressant drug, frequency of visits, and specialty referral) were made by patients and treating physicians, as in usual practice. This strategy was consistent with our objective to study the consequences of initial antidepressant drug choice under usual care conditions. Baseline assessment (conducted before randomization) included the following.

  • Current depression module of the Structured Clinical Interview for DSM-III-R21—a structured assessment of psychiatric diagnoses22
  • A structured interview rating of the 17-item version of the Hamilton Depression Rating Scale (HDRS)23-24
  • Anxiety and depression subscales of the Hopkins Symptom Checklist (SCL)—a standard self-rated measure of current psychiatric symptoms25
  • The Medical Outcomes Study SF-36 Health Survey—a self-report measure of health-related quality of life26-27
  • Questions concerning use of out-of-plan medical and mental health services

Each measure was repeated 1, 3, 6, 9, 12, 18, and 24 months after randomization. Follow-up assessments also assessed current medication use and adverse effects. Follow-up interviewers did not inquire about medication use (which might reveal treatment assignment) until after the assessment of clinical and quality-of-life outcomes. Follow-up data collection continued, regardless of patients' use of antidepressant medication or any other treatment.

Approximately 16% of baseline assessments and 97% of follow-up assessments were conducted by telephone (with the remainder conducted in person). Results of a test-retest reliability study28 showed excellent agreement between in-person and telephone administration of depression measures.

Computerized pharmacy records29 were used to examine subsequent antidepressant medication use (changes in medication, dose, and duration of use).30 Medical comorbidity was assessed using the Chronic Disease Score31—a measure of severity of medical illness computed from pharmacy records.

Actual budget-based costs (not charges) were calculated using Group Health Cooperative's utilization and cost accounting system, which tracks each unit of health care use (eg, outpatient visit, laboratory test, prescription, and inpatient stay).32 For this report, out-of-plan health services were assigned the costs of similar services delivered inside of Group Health Cooperative.

Primary data analyses were based on initial medication assignment or "intent to treat"33; patients discontinuing treatment or switching to a different antidepressant medication were included in the original treatment group. This strategy is most consistent with the study objective to examine the consequences of initial medication selection. This approach might be viewed as a randomized trial of a formulary or policy question: Should fluoxetine or a less expensive tricyclic drug be recommended as the initial antidepressant drug for primary care patients?

Outcomes included categorical measures repeated over time (eg, remission of depression), continuous measures repeated over time (eg, HDRS score), and continuous measures summarizing the entire time (eg, health care costs). Repeated dichotomous measures were analyzed using generalized estimating equation (GEE) models with a logistic link via (SAS PROC GENMOD; SAS Inc, Cary, NC). Repeated continuous measures were analyzed using mixed-model analysis of covariance via SAS PROC MIXED (SAS Inc). Planned analyses included 2 comparisons for each study outcome (fluoxetine vs imipramine and fluoxetine vs desipramine), with an a priori {alpha} level of .02 for each comparison. Analyses of clinical and quality-of-life outcomes focused on detecting an overall difference in outcome from 6 to 24 months. For each outcome, we first examined a model including age, sex, baseline value, initial treatment group, time, and interaction of treatment group with time. In the absence of a significant group x time interaction, we examined the main effects of initial treatment assignment from 6 to 24 months. Presentation of cost data (means ± SDs) used untransformed data, but some significance tests required appropriate transformations: log transformation for categories that approximated a log-normal distribution (eg, total cost) and rank transformation for categories with a large proportion of nonusers (eg, inpatient costs).


RESULTS
 Jump to Section
 •Top
 •Introduction
 •Patients, materials, and methods
 •Results
 •Comment
 •Author information
 •References

Of 536 patients enrolled, 72% were women. Median age was 41 years (range, 18-90 years), and 7% of patients were 65 years or older. At the baseline interview, 358 patients (66.8%) satisfied Structured Clinical Interview for DSM-III-R criteria for current major depression, 36 (6.7%) met criteria for dysthymic disorder, and 142 (26.5%) met criteria for minor depression without dysthymia. Mean HDRS score was 13.3±2.8 and mean SCL depression subscale score was 2.05±0.78. Approximately 75% of patients reported previous depressive episodes (median, 3 episodes; range, 1-25 episodes), 33% reported previous episodes of antidepressant drug treatment, and 7% reported previous hospitalization for depression. The 3 randomization groups did not differ significantly in mean age, sex distribution, medical comorbidity, or any measure of baseline depression severity.

The number of patients completing each assessment was 471 (87.9%) at 6 months, 455 (84.9%) at 9 months, 435 (81.2%) at 12 months, 414 (77.2%) at 18 months, and 383 (71.5%) at 24 months. Compared with patients completing all follow-ups, those missing one or more assessments did not differ significantly in initial treatment assignment, sex distribution, mean age, medical comorbidity, or mean baseline scores on the HDRS or SCL depression subscale. The proportion of patients remaining enrolled in the health plan (ie, those with complete computerized pharmacy and cost and utilization records available) was 93.1% at 6 months, 84.8% at 12 months, and 75.3% at 24 months. Compared with patients remaining enrolled for the entire 24 months, those disenrolling had a significantly lower mean age (37.4 vs 44.2 years, t=5.27; P<.001) and a significantly higher mean baseline score on the SCL depression subscale (2.18 vs 2.00, t=2.23; P=.03). Disenrollment was not significantly associated with initial treatment assignment, sex distribution, medical comorbidity, or baseline HDRS score.

As shown in the left side of Figure 1, the proportion of patients continuing use of the original antidepressant drug was higher for the fluoxetine group than for either tricyclic drug group throughout follow-up. A GEE model found a significant decrease in probability of continuing use of the original medication over time ({chi}21=48.4; P<.001), a significant effect of randomization group ({chi}22=50.8; P<.001), and no significant group x time interaction ({chi}22=0.8; P=.65). In pairwise comparisons with the fluoxetine group, likelihood of continuing use of the original medication was significantly lower in patients assigned to receive either desipramine (odds ratio=0.32, Z=5.06; P<.001) or imipramine (odds ratio=0.47, Z=3.62; P<.001). An alternative method of expressing the difference in likelihood of continuing use of the original medication is that between 4 and 5 additional patients would need to begin treatment with fluoxetine (instead of imipramine or desipramine) to prevent an instance of medication switching or discontinuation during the first 6 months of treatment. As suggested by the left half of Figure 1, more than 60% of medication switching or discontinuation during follow-up occurred before 6 months. When analysis was confined to patients continuing use of the original medication for at least 6 months, those in the fluoxetine group were still significantly less likely to switch or discontinue medication use during the remainder of follow-up than those in either tricyclic drug group (P<.05 for both). The right side of Figure 1, however, demonstrates that initial treatment assignment did not affect likelihood of continuing treatment with any antidepressant drug. A GEE model found that likelihood of continuing use of any medication decreased over time ({chi}21=38.4; P<.001), without a significant difference between randomization groups ({chi}22=0.1; P=.95) or group x time interaction ({chi}22=1.1; P=.60). The contrast between both halves of Figure 1 is explained by the higher rate of medication switching among patients beginning treatment with tricyclic drugs. Medication adverse effects were the most frequently reported reason for patients' changing or discontinuing use of medication (data available on request).



View larger version (23K):
[in this window]
[in a new window]
Figure 1. Likelihood of continuing antidepressant medication treatment (as reported at follow-up interviews) according to initial treatment assignment. Left, Likelihood of continuing use of original medication. Right, Likelihood of continuing use of any antidepressant drug. See the "Results" section for the number of patients completing each follow-up assessment.


Figure 2 displays the time course of mean HDRS scores according to initial treatment assignment. All 3 groups showed large decreases in HDRS scores during the first 6 months, followed by slower declines during the next 18 months. Repeated-measures analysis using data from months 6 through 24 found a significant effect of time (F1=44.79; P<.001) but no significant effect of randomization group (F2=1.46; P=.23) and no significant group x time interaction (F2=0.93; P=.40). Across this entire interval, mean±SE HDRS score (adjusted for age, sex, baseline HDRS score, and history of previous depression treatment) was 8.10±0.26 for the fluoxetine group, which did not differ significantly from the adjusted mean±SEs of 7.81±0.25 for the desipramine group (P=.35 for comparison with fluoxetine) and 8.43±0.25 for the imipramine group (P=.42 for comparison with fluoxetine). Remission of depression was defined as fewer than 5 DSM-IV34 depressive symptoms (ie, below the major depression diagnostic threshold) and an HDRS score of 7 or less. The proportion of patients in remission increased from 38% to 50% between 6 and 24 months; this rate of increase was similar in the 3 randomization groups (Figure 3). A GEE model predicting likelihood of remission from 6 to 24 months found a significant increase over time ({chi}21=17.7; P<.001), a marginally significant effect of randomization group ({chi}22=5.9; P=.05), and no significant group x time interaction ({chi}22=0.1; P=.94). In pairwise comparisons, likelihood of remission among patients starting treatment with fluoxetine did not differ from likelihood in the desipramine (odds ratio=1.22, Z=1.29; P=.20) or imipramine groups (odds ratio=0.97, Z=.24; P=.81).



View larger version (11K):
[in this window]
[in a new window]
Figure 2. Course of Hamilton Depression Rating Scale (HDRS) scores according to initial treatment assignment. See the "Results" section for the number of patients completing each follow-up assessment.




View larger version (15K):
[in this window]
[in a new window]
Figure 3. Proportion of patients in clinical remission over time according to initial treatment assignment. Remission was defined as resolution of a major depressive episode and a Hamilton Depression Rating Scale score of 7 or less.


Analysis of SCL depression subscale scores (Figure 4) yielded similar results: a sharp decline during the first 6 months, followed by a gradual decrease. Repeated-measures analysis also showed the same pattern: a significant decrease over time (F1=17.97; P<.001) but no significant effect of randomization group (F2=1.61; P=.20) or difference in time effect across groups (F2=0.53; P=.59). Adjusted mean±SE SCL depression subscale score for the entire period was 0.59±0.04 for the fluoxetine group, which was not significantly different from the adjusted mean±SEs of 0.54±0.04 for the desipramine group and 0.63±0.04 for the imipramine group.



View larger version (11K):
[in this window]
[in a new window]
Figure 4. Course of Hopkins Symptom Checklist (SCL) depression subscale scores according to initial treatment assignment. See the "Results" section for the number of patients completing each follow-up assessment.


Figure 5 displays mean values for the SF-36 Mental Component Summary score at each time. All 3 groups began with mean scores more than 2 SD below the general population mean of 50, and finished with mean scores near the general population mean. Repeated-measures analysis found significant effects of time (F1=22.78; P<.001) and randomization group (F2=3.99; P=.02), but not of group x time interaction (F2=0.03; P=.97). Adjusted mean±SE score for the 18-month period was 47.32±0.64 for the fluoxetine group, which did not differ significantly from the adjusted mean±SEs of 49.13±0.62 for the desipramine group and 46.76±0.61 for the imipramine group.



View larger version (11K):
[in this window]
[in a new window]
Figure 5. Course of SF-36 Mental Component Summary score according to initial treatment assignment. See the "Results" section for the numbers of patients completing each assessment.


Because our sample included a substantial number of patients with mild depression, we repeated the previous analyses in the subgroup of patients satisfying criteria for major depression at the baseline assessment (approximately 65% of the total). Figure 6 displays mean HDRS scores at each follow-up assessment in this subgroup. Repeated-measures analyses using data from 6 through 24 months yielded identical results to those in the full sample: no difference in average score among randomization groups and no difference in rate of improvement over time. Probability of remission in this subgroup was slightly lower than in the full sample (33% at 6 months and 46% at 24 months), but GEE analyses comparing probability of remission found no significant main effect of initial treatment assignment and no group x time interaction. Parallel analyses were performed in the subgroup with baseline HDRS scores of 15 or more (approximately 40% of the sample). Results in this subgroup were also essentially identical to those in the entire sample.



View larger version (11K):
[in this window]
[in a new window]
Figure 6. Course of Hamilton Depression Rating Scale (HDRS) scores according to initial treatment assignment and limited to patients satisfying criteria for major depression at the baseline assessment.


Changes in overall health care costs over time are displayed in Figure 7. As shown in the left half of the figure, total costs showed considerable variability among groups and over time, reflecting the large variability of inpatient costs. Total outpatient costs (right half of Figure 7) showed considerably less variability. Mixed-model analysis of covariance was used to compare total costs over time for the 3 treatment groups after adjustment for age and sex. Costs in all 3 groups declined gradually over time (F=20.5; P<.001), but there was no main effect of treatment group (F=0; P=.99) or group x time interaction (F=0.04; P=.96). Adjusted mean±SE costs (per 6 months) from this model were $1768±$178 for the fluoxetine group compared with $1780±$171 for the desipramine group and $1788±$175) for the imipramine group.



View larger version (24K):
[in this window]
[in a new window]
Figure 7. Time course of direct health care costs according to initial treatment assignment. Time labels represent the end of each 6-month period for which costs were calculated (n=494 for months 1-6, n=452 for months 7-12, n=434 for months 13-18, and n=397 for months 19-24).


Table 1 displays mean direct health care costs for the entire 24 months of follow-up in specific categories. Costs for antidepressant drug prescriptions were approximately $250 higher (during 24 months) for the fluoxetine group than for either tricyclic drug group. Antidepressant drug costs in the tricyclic drug groups reflect the substantial minority of patients switching medication, and costs for all 3 groups reflect the increasing proportion of patients discontinuing antidepressant drug treatment over time. For all other categories of cost, the 3 groups were approximately equal. Although the fluoxetine group had higher mean medical inpatient costs than the imipramine group ($1584 vs $1400), the comparison of rank-transformed costs indicated significantly lower cost in the fluoxetine group (P=.02). The apparent discrepancy between untransformed and rank-transformed analyses reflects the effect of a few high-cost outliers in the fluoxetine group. No other pairwise comparisons exceeded the .02 level of statistical significance.


View this table:
[in this window]
[in a new window]
Direct Health Care Costs During 24 Months of Follow-up According to Initial Treatment Assignment*



COMMENT
 Jump to Section
 •Top
 •Introduction
 •Patients, materials, and methods
 •Results
 •Comment
 •Author information
 •References

These results are broadly consistent with short-term outcomes reported earlier.19 Patients beginning treatment with fluoxetine were more likely to continue taking the original medication but not more likely to continue treatment with any antidepressant medication. Initial selection of fluoxetine led to essentially identical clinical and quality-of-life outcomes compared with initial selection of imipramine or desipramine. Beginning treatment with fluoxetine led to significantly higher antidepressant drug prescription costs, but overall costs of care did not differ in the 3 randomization groups.

Our study design intended to examine the consequences of initial antidepressant drug selection. Initial medication selection was randomly assigned, but all subsequent treatment (including switching or discontinuing antidepressant medication use) was allowed to proceed as usual. Approximately 35% of patients beginning treatment with tricyclic antidepressant drugs switched to fluoxetine or other newer antidepressant drugs at some time during follow-up. Our study design regards such medication switches as important outcomes of initial treatment choice. Because initial selection of a tricyclic antidepressant led to a higher proportion of medication switches, excluding or censoring these "crossovers" could introduce significant bias. Medication switches are important outcomes of initial treatment choice because such switches may affect clinical outcomes and treatment costs.

We did not attempt to control or affect treating physicians' decisions regarding medication dosing, criteria for medication change or discontinuation, or frequency of follow-up visits. We cannot say whether a stricter protocol for treatment would yield different results, but we can say that results from such a protocol would be less generalizable to real world practice.

Clinical and quality-of-life outcomes were generally similar in the 3 randomization groups. On average, patients experienced the greatest improvement early in treatment, with slow but statistically significant improvement continuing after 6 months. Although most patients experienced considerable improvement, fewer than half of the patients at each assessment achieved remission of depression (as defined in most specialty trials). Differences in clinical outcome among the 3 randomization groups were neither clinically nor statistically significant. Standard error estimates from the repeated-measures models (0.25 for the HDRS score, 0.04 for the SCL depression subscale score, and 0.60 for the SF-36 Mental Component Summary score) indicate that our sample allowed sufficient power to detect small differences in clinical outcome. The differences observed among groups (eg, 0.3 HDRS points and 2 points on the SF-36 Mental Component Summary scale) were not clinically significant, suggesting that our finding of no significant difference between groups is not a consequence of inadequate sample size.

The analyses reported previously do not address whether patients continuing treatment with fluoxetine experience different outcomes than those continuing treatment with tricyclic drugs. Analyses restricted to patients continuing use of the original medication yield results that are similar to the intent-to-treat analyses reported previously (details available on request). Analyses according to treatment received suggested that patients switching medication use experienced slightly delayed recovery compared with those continuing the original treatment. We suggest, however, that the intent-to-treat analyses reported herein address the most important policy and clinical questions. Policy makers must consider whether fluoxetine or other serotonin reuptake inhibitor drugs should be recommended as first-line treatments or reserved for second-line use. Restricting or prohibiting medication switches (even for those not responding to treatment with tricyclic drugs) is not a viable policy option. Similarly, a clinician recommending antidepressant treatment must choose an initial medication based on clinical efficacy, tolerability, and cost. This clinical decision should incorporate the knowledge that medication switches are common in everyday practice and the likelihood of switching varies among antidepressant medications.7, 19 As shown in Figure 1, patients beginning treatment with tricyclic drugs are as likely to switch medication (primarily to fluoxetine) by 12 months as they are to continue tricyclic treatment.

Total medical costs were essentially identical for the 3 treatment groups. Antidepressant drug prescription costs for 24 months were $250 higher among patients beginning treatment with fluoxetine. This difference reflects actual patterns of use (including medication switching and discontinuation) and is smaller than one would predict by simply projecting differences in drug purchase costs for 24 months. As indicated by the SDs in Table 1, our sample size would not allow detection of small-to-moderate differences in total cost. Although our best estimate is that total costs in the fluoxetine group were essentially identical to those in the tricyclic groups, we cannot definitively exclude differences of up to several hundred dollars in either direction. The observed pattern of costs in the group starting treatment with fluoxetine (higher medication costs balanced by slightly lower general medical costs) might have different implications in different health care systems.

Our findings differ from those of several decision-analytic studies14-16 that found a cost-effectiveness advantage of using newer antidepressant drugs. These models typically use lower discontinuation rates for newer drugs to project higher clinical response rates and lower subsequent costs for nonmedication components of treatment. Our data (including short-term results published previously19) show that patients treated with fluoxetine experienced fewer adverse effects and were less likely to discontinue use of the original medication. Unlike several decision-analytic models, we did not find differences in overall treatment adherence or clinical outcomes. Although we found that total treatment costs were no higher in the fluoxetine group, we cannot link this finding to any difference in clinical outcome. Specifically, we cannot say that the small decrease in nonantidepressant drug costs in the fluoxetine group results from more effective depression treatment or improved clinical outcomes. Our data are consistent with results from other observational32 and modeling35 studies finding that differences in antidepressant drug costs are relatively small compared with the overall medical care costs of treating depressed patients.

Our findings suggest that administrative restrictions on first-line use of fluoxetine will not reduce overall costs of care. This study design might be considered a randomized trial of a formulary policy: restricting use of fluoxetine to patients who have not responded to initial treatment with a tricyclic antidepressant. As seen in these data and our earlier report of short-term outcomes,19 requiring the initial selection of a tricyclic drug does not significantly affect clinical or quality-of-life outcomes compared with initial selection of fluoxetine. Initial choice of a tricyclic drug, however, can lead to more frequent medication adverse effects and a higher likelihood of medication switches. Restricting fluoxetine to second-line treatment can lead to moderate savings in antidepressant drug prescription costs but not to significant overall savings.


AUTHOR INFORMATION
 Jump to Section
 •Top
 •Introduction
 •Patients, materials, and methods
 •Results
 •Comment
 •Author information
 •References

Reprints: Gregory E. Simon, MD, MPH, Center for Health Studies, Group Health Cooperative, 1730 Minor Ave, Suite 1600, Seattle, WA 98101-1448 (e-mail: simon.g{at}ghc.org).

Accepted for publication June 16, 1998.

Supported by a grant from Lilly Research Laboratories, Indianapolis, Ind.

We thank Mark Hornbrook, PhD, Thomas Inui, MD, MPH, Thomas Koepsell, MD, MPH, Bryan Luce, PhD, and H. C. Schulberg, PhD, for providing valuable advice on study design and data analysis, and Chester Pabiniak, MS, and Do Peterson, MA, for assistance with computer programming and data analyses.

From the Center for Health Studies, Group Health Cooperative (Drs Simon, VonKorff, Ludman, and Wagner and Mr Grothaus), and the Department of Psychiatry and Behavioral Sciences, University of Washington (Drs Simon and Katon), Seattle, Wash; Lilly Research Laboratories, Indianapolis, Ind (Dr Heiligenstein); and MEDTAP International, Bethesda, Md (Dr Revicki).


REFERENCES
 Jump to Section
 •Top
 •Introduction
 •Patients, materials, and methods
 •Results
 •Comment
 •Author information
 •References

1. Glaser M. Annual Rx survey. Drug Top. 1997;141:45-53.
2. Ormel J, VonKorff M, Ustun TB, Pini S, Korten A, Oldehinkel T. Common mental disorders and disability across cultures. JAMA. 1994;272:1741-1748. FREE FULL TEXT
3. Spitzer R, Kroenke K, Linzer M, et al. Health-related quality of life in primary care patients with mental disorders. JAMA. 1995;274:1511-1517. FREE FULL TEXT
4. Simon G, Ormel J, VonKorff M, Barlow W. Health care costs associated with depressive and anxiety disorders in primary care. Am J Psychiatry. 1995;152:352-357. FREE FULL TEXT
5. Unutzer J, Patrick DL, Simon G, et al. Depressive symptoms and the cost of health services in HMO patients age 65 and over: a four-year prospective study. JAMA. 1997;277:1618-1623. FREE FULL TEXT
6. Keller MB, Lavori PW, Klerman GL, et al. Low levels and lack of predictors of somatotherapy and psychotherapy received by depressed patients. Arch Gen Psychiatry. 1986;43:458-466. FREE FULL TEXT
7. Simon G, VonKorff M, Wagner EH, Barlow W. Patterns of antidepressant use in community practice. Gen Hosp Psychiatry. 1993;15:399-408. FULL TEXT | ISI | PUBMED
8. Katon W, VonKorff M, Lin E, Bush T, Ormel J. Adequacy and duration of antidepressant treatment in primary care. Med Care. 1992;30:67-76. FULL TEXT | ISI | PUBMED
9. Song F, Freemantle N, Sheldon T, et al. Selective serotonin reuptake inhibitors: meta-analysis of efficacy and acceptability. BMJ. 1993;306:683-687.
10. Workman E, Short D. Atypical antidepressants versus imipramine in the treatment of major depression: a meta-analysis. J Clin Psychiatry. 1993;54:5-12.
11. Greenberg R, Bornstein RF, Zborowski MJ, Fisher S, Greenberg MD. A meta-analysis of fluoxetine outcome in the treatment of depression. J Nerv Ment Dis. 1994;182:547-551. ISI | PUBMED
12. Montgomery S, Henry J, McDonald G, et al. Selective serotonin reuptake inhibitors: meta-analysis of discontinuation rates. Int Clin Psychopharmacol. 1994;9:47-53. ISI | PUBMED
13. Pande A, Sayler ME. Adverse events and treatment discontinuations in fluoxetine clinical trials. Int Clin Psychopharmacol. 1993;8:267-269. ISI | PUBMED
14. LePen C, Levy E, Ravily V, Beuzen JN, Meurgey F. The cost of treatment dropout in depression. J Affect Disord. 1994;31:1-18. FULL TEXT | ISI | PUBMED
15. Jonsson B, Bebbington PE. What price depression: the cost of depression and the cost-effectiveness of pharmacological treatment. Br J Psychiatry. 1994;164:665-673. FREE FULL TEXT
16. Bentkover JDFJ. Cost analysis of paroxetine vs imipramine in major depression. PharmacoEconomics. 1996;8:223-232.
17. Woods S, Rizzo JA. Cost-effectiveness of antidepressant treatment reassessed. Br J Psychiatry. 1997;170:257-263. FREE FULL TEXT
18. Revicki D, Brown RE, Keller MB, Gonzales J, Culpepper L, Hales RE. Cost-effectiveness of newer antidepressants compared with tricyclic antidepressants in managed care settings. J Clin Psychiatry. 1997;58:47-58. ISI | PUBMED
19. Simon G, VonKorff M, Heiligenstein JH, et al. Initial antidepressant selection in primary care: effectiveness and cost of fluoxetine vs.tricyclic antidepressants. JAMA. 1996;275:1897-1902. FREE FULL TEXT
20. Simon G, Wagner E, VonKorff M. Cost-effectiveness comparisons using "real world" randomized trials: the case of new antidepressant drugs. J Clin Epidemiol. 1995;48:363-373. FULL TEXT | ISI | PUBMED
21. Spitzer RL, Williams JBW, Gibbon M, First MB. User's Guide for the Structured Clinical Interview for DSM-III-R. Washington, DC: American Psychiatric Press; 1990.
22. American Psychiatric Association. Diagnostic and Statistical Manual of Mental Disorders, Third Edition, Revised. Washington, DC: American Psychiatric Association; 1987.
23. Hamilton MA. A rating scale for depression. J Neurol Neurosurg Psychiatry. 1960;23:56-62.
24. Williams JBW. A structured interview guide for the Hamilton Depression Rating Scale. Arch Gen Psychiatry. 1988;45:742-747. FREE FULL TEXT
25. Derogatis L, Rickels K, Uhlenhuth EH, Covi L. The Hopkins Symptom Checklist: a measure of primary symptom dimensions. In: Pichot P, ed. Psychological Measurements in Psychopharmacology: Problems in Psychopharmacology.Basel, Switzerland: Kargerman; 1974:79-110.
26. Ware J, Snow KK, Kosinski M, Gandek B. The MOS SF-36 Health Survey: Manual and Interpretation Guide. Boston, Mass: The Health Institute, New England Medical Center; 1993.
27. Ware J, Kosinski M, Keller SD. SF-36 Physical and Mental Component Summary Scales: A User's Manual. Boston, Mass: The Health Institute, New England Medical Center; 1994.
28. Simon G, VonKorff M, Revicki D. Telephone assessment of depression severity. J Psychiatr Res. 1993;27:247-252. FULL TEXT | ISI | PUBMED
29. Saunders K, Stergachis A, VonKorff M. Group Health Cooperative of Puget Sound. In: Strom B, ed. Pharmacoepidemiology.New York, NY: John Wiley & Sons Inc; 1994.
30. Simon G, Lin EHB, Katon W, et al. Outcomes of "inadequate" antidepressant treatment in primary care. J Gen Intern Med. 1995;10:663-670. ISI | PUBMED
31. Clark D, VonKorff M, Saunders K, Baluch WM, Simon GE. A Chronic Disease Score with empirically derived weights. Med Care. 1995;33:783-795. FULL TEXT | ISI | PUBMED
32. Simon G, VonKorff M, Barlow W. Health care costs of primary care patients with recognized depression. Arch Gen Psychiatry. 1995;52:850-856. FREE FULL TEXT
33. Sackett D, Gent M. Controversy in counting and attributing events in clinical trials. N Engl J Med. 1979;301:1410-1412. ISI | PUBMED
34. American Psychiatric Association. Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition, Revised. Washington, DC: American Psychiatric Association; 1994.
35. Hylan T, Kotsanos JG, Andersen JS, et al. Comparison of a decision analytic model with results from a naturalistic economic clinical trial: an application to evaluating alternative antidepressants. Am J Managed Care. 1996;2:1211-1223.

RELATED ARTICLES

The Archives of Family Medicine Continuing Medical Education Program
Arch Fam Med. 1999;8(4):291-293.
FULL TEXT  

TCAs vs SSRIs: Same Bang for Whose Buck?
Joseph J. Gallo
Arch Fam Med. 1999;8(4):326-327.
FULL TEXT  


THIS ARTICLE HAS BEEN CITED BY OTHER ARTICLES

Antidepressant Switching Among Adherent Patients Treated for Depression
Marcus et al.
Psychiatr. Serv. 2009;60:617-623.
ABSTRACT | FULL TEXT  

Evidence-based guidelines for treating depressive disorders with antidepressants: A revision of the 2000 British Association for Psychopharmacology guidelines
Anderson et al.
J Psychopharmacol 2008;22:343-396.
ABSTRACT  

Implications of part d for mentally ill dual eligibles: a challenge for medicare.
Morden and Garrison
Health Aff (Millwood) 2006;25:491-500.
ABSTRACT | FULL TEXT  

Continuity of Antidepressant Treatment for Adults With Depression in the United States
Olfson et al.
Am. J. Psychiatry 2006;163:101-108.
ABSTRACT | FULL TEXT  

Cost-effectiveness of Evidence-Based Pharmacotherapy or Cognitive Behavior Therapy Compared With Community Referral for Major Depression in Predominantly Low-Income Minority Women
Revicki et al.
Arch Gen Psychiatry 2005;62:868-875.
ABSTRACT | FULL TEXT  

Change in Psychosocial Functioning and Quality of Life of Patients With Body Dysmorphic Disorder Treated With Fluoxetine: A Placebo-Controlled Study
Phillips and Rasmussen
Psychosomatics 2004;45:438-444.
ABSTRACT | FULL TEXT  

Quantitative analysis of sponsorship bias in economic studies of antidepressants
Baker et al.
Br. J. Psychiatry 2003;183:498-506.
ABSTRACT | FULL TEXT  

Venlafaxine: the relationship between dose, plasma concentration and clinical response in depressive patients.
Charlier et al.
J Psychopharmacol 2002;16:369-372.
ABSTRACT  

Two-Year Effects of Quality Improvement Programs on Medication Management for Depression
Unutzer et al.
Arch Gen Psychiatry 2001;58:935-942.
ABSTRACT | FULL TEXT  

Antidepressant selection and economic outcome: a review of methods and studies from clinical practice
CROWN
Br. J. Psychiatry 2001;179 :s18-s22.
ABSTRACT | FULL TEXT  

Amitriptyline: still efficacious, but at what cost?
THOMPSON
Br. J. Psychiatry 2001;178:99-100.
FULL TEXT  

The Electronic Medical Record: A Randomized Trial of Its Impact on Primary Care Physicians' Initial Management of Major Depression
Rollman et al.
Arch Intern Med 2001;161:189-197.
ABSTRACT | FULL TEXT  

Managing Depression in Medical Outpatients
Whooley and Simon
NEJM 2000;343:1942-1950.
FULL TEXT  

More people with depression continued treatment with fluoxetine than with desipramine or imipramine
Anderson
Evid. Based Med. 2000;5:51-51.
FULL TEXT  

More people with depression continued treatment with fluoxetine than with desipramine or imipramine
Anderson
Evid. Based Ment. Health 2000;3:23-23.
FULL TEXT  

Evidence-based guidelines for treating depressive disorders with antidepressants: a revision of the 1993 British Association for Psychopharmacology guidelines
Anderson et al.
J Psychopharmacol 2000;14:3-20.
ABSTRACT  

TCAs vs SSRIs: Same Bang for Whose Buck?
Gallo
Arch Fam Med 1999;8:326-327.
FULL TEXT  




HOME | CURRENT ISSUE | PAST ISSUES | TOPIC COLLECTIONS | CME | PHYSICIAN JOBS | HELP
CONDITIONS OF USE | PRIVACY POLICY | CONTACT US | SITE MAP
 
© 1999 American Medical Association. All Rights Reserved.