Managing Antidepressant Drug Therapy in an Evolving Marketplace

July 1, 2004
Jeffrey McCombs, PhD

Supplements and Featured Publications, Current Trends for the Management and Treatment of Depression, Volume 10, Issue 6 Suppl

The Health Employer Data and InformationSet (HEDIS) plays an importantrole in the marketing ofmanaged care plans to potential enrollees.HEDIS guidelines include several prescriptiondrug—related quality-of-care reporting requirementsdeveloped by the National Committeefor Quality Assurance.1,2 In particular, HEDISguidelines for the treatment of depressionrequire plans to report data on duration ofantidepressant drug therapy. The importanceof plan performance in meeting drug therapygoals will only increase with the implementationof the Medicare prescription drug benefitdesigned to encourage Medicare recipients toenroll in managed care plans. Specifically,Medicare beneficiaries use a disproportionateshare of all prescription drugs, and successfulmanagement of their chronic medical conditionsis critical to the financial survival ofMedicare managed care plans.3 This articleaddresses 2 issues. First, the current researchon the impact of quality-of-care report cardson the marketing of managed care plans isreviewed. This review is followed by a discussionof the data needed by managed careplans to effectively manage antidepressanttherapy in an evolving marketplace.

Several factors related to the market forand use of antidepressant medications are ofparticular importance. First, many of themost popular selective serotonin reuptakeinhibitor (SSRI) antidepressants are nowavailable from generic manufacturers at significantreductions in price. This radicalchange in the relative pricing structurebetween alternative products has significantlyaltered the type of data needed toeffectively manage antidepressant drugtherapy. Second, new research has emergedthat suggests that improved access to SSRIantidepressants may expand the clinicalconditions for which physicians prescribeantidepressant therapy.4 This expansion ofantidepressant use is likely to make it increasinglydifficult for managed care plans toachieve the treatment goals for depressionas expressed in the HEDIS guidelines. Third,the research designs and statistical methodsemployed in earlier retrospective databaseanalyses has resulted in a confusing array ofconflicting research results on the relativeperformance of alternative antidepressants.Any attempt to glean the data needed fortoday's formulary decisions from these studiesrequires a solid understanding of theirlimitations and potential biases. Finally, newdata are emerging concerning the use ofantidepressant medications in children, adolescent,and geriatric patients and patientswith cardiovascular comorbidities that alsoaffect the development of a treatment protocolfor antidepressant therapy.

Quality-of-Care Report Cards and theMarketing of Managed Care

The managed care marketplace is movingrapidly, if reluctantly, to implement systemsfor measuring and reporting the quality ofcare data. For example, health maintenanceorganizations (HMOs) and managed careplans are increasingly required (throughmarketplace pressures) to adhere to quality-of-care reporting standards, such as HEDIS.Individual purchasers of health insurancewill use information on quality of care andpatient satisfaction if these data are providedduring open enrollment periods. Forexample, in 1994, the California Public Employees'Retirement System4 (CalPERS)published HEDIS quality-of-care informationand its own consumer satisfaction rankingsfor all contract plans in its annual openenrollment material. CalPERS then surveyedmembers who switched plans toassess which sources of data were influentialin the member's plan selection decision. Theresults indicated that advertising was theleast influential source of information, usedby 10% of plan switchers, followed by healthfair information (27%), friends and family(64%), and the CalPERS Quality of CareReport (66%). Chernew and Scanlon5 foundthat employees from one large firm didrespond to plan performance measuresrelated to preventive services. Spranca andcolleagues6 found that data from theConsumer Assessment of Health Plans Study(CAHPS) survey were instrumental in shiftingprivately insured consumers in LosAngeles to lower priced but higher qualityplans, even when these plans covered fewerservices. Schultz and colleagues7 found thatconsumers who changed physician groupswere more likely to use report card informationin their selection. Beaulieu8 found thatemployees were more likely to switch fromplans with low reported quality of care.

The impact of quality of care and patientsatisfaction information on plan selection isnot uniform across all populations. Farleyand colleagues9 found that Medicaid recipientsin New Jersey chose HMOs with higherCAHPS scores conditional on the recipienthaving read the materials. McCormack andcolleagues10 found only limited effects of thehandbook entitled "Medicare and You,1999" and that the impact depended on howthe information was presented. Hibbard andcolleagues11 found that most Medicare beneficiariesdid not understand the differencesin their managed care and fee-for-serviceoptions to make well-informed purchasingdecisions. Regardless of these latter setbacks,the growing amount of literature onthe impact of information on plan choicesuggests that consumers will sanction poorplan performance and low quality in a competitivemarket for health insurance if thisinformation is provided.

The reporting of quality-of-care measuresrelated to the effective use of prescriptiondrugs subjects the formulary decisions ofmanaged care plans to external review. Caremust be taken to achieve therapeutic treatmentgoals in a cost-conscience manner.Antidepressant therapies are an importantcase in point because HEDIS includes durationof antidepressant therapy as one of itsguidelines. However, the generic availabilityof well-established medications, such as fluoxetineand paroxetine, has radically alteredthe information needed by managed careplans to effectively manage antidepressanttherapy.

Effective Antidepressant Therapy in anEvolving Market

Starting with the introduction of fluoxetinein the late 1980s, managed care planswere inundated with a confusing array ofresearch based on real-world data that eitherdocumented the advantages of SSRI antidepressantsover older alternatives or trumpetedthe advantages of various brandedantidepressants over their competitors. Theavailability of generic SSRIs has made muchof this research irrelevant because thegeneric SSRIs are being granted preferredproduct status by most managed care formularies.However, much can be learnedfrom a quick review of this literature asbranded antidepressants battle to maintainan appropriate role in managed care's drugarsenal.

Patient Selection Criteria. The majorityof published observational research comparingpatient outcomes across alternative antidepressantsis focused on new episodes oftreatment, defined as the lack of antidepressanttherapy for 6 or more months beforethe initiation of treatment. While this focuswas well suited for the selection of a fewbrand name medications for preferred formularystatus, these studies are much lessrelevant after the introduction of genericSSRIs. Assuming that generic SSRIs will nowbe given preferred formulary status for newtreatment episodes, data are needed on theperformance of generic and branded antidepressantsin 3 understudied patient groups.First, will generic SSRIs perform well inpatients with a recurring episode of depressionwho responded well when previouslytreated with a nonpreferred branded antidepressant? Second, can long-term depressedpatients who are currently stabilized on anonpreferred branded product be successfullyswitched to a generic SSRI? Finally, arethere identifiable subgroups of depressedpatients who are at high risk of treatmentfailure using the preferred generic SSRIs?The current research literature providesvery little information with which to addressthese issues.

While the SSRIs have generally performedwell compared with conventionalantidepressants in most published researchstudies, very little data are available on howthe cost effectiveness of newer antidepressantsmay vary by diagnosis. For example,McCombs and colleagues12 restricted theiranalyses to major depression, whereasHylan and colleagues13 included neuroticdepression and brief and prolonged depressivereaction. Patient outcomes data by diagnosisare particularly important becausephysicians are likely to expand the range ofpatients treated with SSRIs with the availabilityof less expensive generic products.Specifically, McCombs and colleagues14 founda significant and permanent shift in the numberof MediCal recipients treated for depressionafter open access to fluoxetine andparoxetine was granted in May 1996. Moreimportant, the expansion of antidepressanttherapy resulted in a significant decrease incompliance across all antidepressants, notjust the SSRIs granted preferred formularystatus.15 Managed care plans can expect similarreductions in their HEDIS performanceindicators unless the expansion of antidepressanttherapy to less severely ill patientscan be clinically justified.

Most of the available research excludedpatients who had a history of comorbid mentaldisorders or who used psychotropic medicationsthat would indicate either refractorydepression (eg, antipsychotic drugs) orcomorbid conditions (eg, mood stabilizers forbipolar disorders). Granting preferred formularystatus to the generic SSRIs will createbarriers to branded antidepressants for theseimportant and costly patient populationsthat may or may not have an adverse impacton patient outcomes. Clearly, more researchis needed on the impact of newer antidepressantsby diagnosis.

Mental Health Costs, Total Costs, andDrug Use Patterns. All studies using retrospectivepaid claims data have sufficient datato investigate a wide range of patient outcomes.When comparing alternative antidepressants,this range of outcomes shouldinclude duration of therapy, future changesin therapy, and a comprehensive breakdownof total cost by type of service. Even withthese data in hand, studies often elect tofocus on mental health costs. There are 3 limitationsassociated with measuring mentalhealth—related costs of which formulary decisionmakers must be aware. First, patientswith depression often seek treatment for awide range of other healthcare problems.16Therefore, the successful treatment of depressionmay be associated with significant reductionsof nonmental healthcare costs. Second,it may be difficult to disentangle mentalhealthcare costs from other costs based on thereported diagnosis or type of provider as thecompleteness and validity of these data elementscan be questioned. Third, limiting thecost analysis to mental health—related costs,including the cost of the antidepressant used,tends to be a disadvantage for more expensivemedications and medications with longeraverage duration of therapy by not includingall of the potential cost-offsets that may accrueto that medication's use. Any study using aretrospective data set consisting of paid claimsor healthcare utilization data has access todata on total costs, and failure to report outcomesrelated to total costs must be evaluatedcautiously.

Multivariate Modeling of Cost. Multivariatecost models can be difficult foruntrained formulary decision makers toevaluate. Three potential problems are ofparticular concern in models that comparepatient outcomes across alternative antidepressants.First, patients are not randomlyassigned to treatment, resulting in potentiallyimportant differences across treatmentgroups that could affect costs (treatmentselection bias). The first line of defenseagainst treatment selection bias is to makemaximum use of available data to control forimportant differences across treatmentgroups. Multivariate models with a parsimoniouslist of independent variables should beevaluated with caution. For example, Russelland colleagues17 specified cost models comparingfluoxetine, paroxetine, and sertralinein which available data on the patient's prioruse of nonmental health—related costs bytype of service, the patient's depression diagnosis,type of insurance coverage, and mix ofcomorbid conditions were not used. Anyantidepressant taken more frequently inrecurrent major depression or by patientswith higher than average prior use of nonmentalhealth services would appear to beless effective when these variables are omittedfrom the cost models.

There are several statistical methodsdesigned to control for unobserved treatmentselection bias that may persist evenafter an exhaustive array of independentvariables have been included in the multivariatemodels of patient outcomes. Forexample, Hylan and colleagues13 apply the 2-staged analytic approach developed by Lee18to adjust the cost estimates associated withdrug selection using an adjustment factorthat reflects the probability that eachpatient selected the treatments under study.The results derived using methods designedto adjust for treatment selection bias are difficultfor formulary decision makers to evaluate,especially if they are not presented incomparison with the parallel results usingmore traditional approaches.

The third problem concerns the inclusionof endogenous variables for duration,switching, or augmentation of the initialantidepressant regimen in the cost modelsdesigned to compare outcomes across alternativeproducts. This approach will likelygenerate results that are very misleadingwith respect to formulary decisions. Forexample, longer duration of antidepressanttherapy has been associated with lower totaltreatment costs and better patient outcomes.If duration of therapy is used as anexplanatory variable in a cost model, thepotential savings associated with drugs withbetter compliance will be allocated to duration,thus reducing the estimated differenceacross medications in favor of those antidepressantswith shorter duration. For example,Sclar and colleagues19 and Skaer andcolleagues20 included a variable for havingreceived 3 consecutive prescriptions in costmodels comparing SSRIs with tricyclic antidepressants.It was not surprising that filling3 sequential antidepressant prescriptionswas associated with significantly higher drugcosts with a trend toward reduced hospitalcost. Likewise, switching and augmentationhave been found to increase treatment costs.Including future switches in therapy or augmentationas an independent variable incost models comparing alternative antidepressantswould reduce the costs associatedwith using a less effective medication to theswitch and augmentation variables.

Finally, healthcare costs are often not normallydistributed, thus violating one of theassumptions of ordinary least squares (OLS)regression analysis. While OLS regressionprovides the best linear unbiased estimatorsin large samples, many investigators transformcosts into log-costs to improve theextent to which the dependent cost variablemeets the normality requirements. Thisapproach is not without its own set of issues.First, most cost models comparing alternativetreatments are conducted using data forpatients being actively treated for a specificdisease. Therefore, the typical disease-specificstudy population does not exhibit theusual, troublesome cost distribution wheremost patients are clustered at or near zerocosts. Second, a log-transformation of costseffectively devalues the impact of patientswith extremely high cost values. It is oftenquite plausible that the impact of new treatmentalternatives will be expressed as areduction in the likelihood that patientsexperience catastrophic, costly events, suchas future hospitalizations. In the case of antidepressanttherapy, better medications arelikely to reduce attempted suicides andfuture hospitalizations, both very costlyevents. Therefore, cost models based on log-transformedcosts may underestimate thetrue impact of more effective antidepressants.Given the risks and benefits of costmodels based on either actual costs or thelog-transformation of costs, it is often usefulfor researchers to conduct both models inparallel and report the extent to which theirresults are robust across both approaches.

Emerging Clinical Concerns

Cardiac Effects of Antidepressants. Ingeneral, the newer antidepressants are consideredsafe in terms of adverse cardiaceffects. Only 2 medications warrant concern.Warrington et al21 report that trazadone cancause significant postural hypotension andmay be associated with prolonged QTc intervals.Mucci22 and Holm and Spencer23 reportthat reboxetine, which is not yet approvedfor use in the United States, has been associatedwith significant increases in heart rateand with atrial and ventricular ectopic beats,especially in the elderly.

Suicide Risk in Children and Adolescents.Several studies outside the UnitedStates have found an inverse correlationbetween the increased use of SSRI antidepressantmedications and the risk of suicideusing aggregate data on the number of prescriptionsdispensed and suicide rates.24-26However, these results appear to be correlatedwith age with some evidence that suiciderisk has actually increased in adolescentsand young adults.27 In the United States, theFood and Drug Administration (FDA) issueda public health advisory on March 22, 2004,that directed manufacturers of 10 newerantidepressants to add a warning statementto their product labeling recommendingclose observation of both adult and pediatricpatients for worsening depression or theemergence of suicidality.28


The availability of generic SSRI antidepressantshas simplified formulary decisionsrelated to how best to treat uncomplicateddepression. However, several new challengeshave arisen. Better data are needed to guideformulary guidelines concerning the reuse ofnonpreferred antidepressants for patientswith recurrent episodes of depression andthe risks associated with switching longtermpatients to generic SSRIs. Managedcare organizations need more informationon the appropriateness of expanding the useof generic SSRIs to treat less severe forms ofdepression. An expansion of the use of antidepressanttherapy will likely reduce HEDIScompliance performance, which couldadversely affect the marketing of the plan topotential enrollees. Plans must also considerFDA concerns that the newer antidepressantsmay increase suicidality in a smallnumber of patients. Finally, more researchis needed to determine if the generic SSRIsadequately meet the therapeutic needs ofmore severely ill patients, such as treatmentrefractory depressed patients and patientswith bipolar disorders.


1. Epstein AM. Rolling down the runway: the challengesahead for quality report cards. . 1998;279:1691-1696.

Health Aff (Millwood)

2. Thompson JW, Bost J, Ahmed F, Ingalls CE, Sennett C.The NCQA's quality compass: evaluating managed carein the United States. . 1998;17:152-158.

Drugs and the Elderly

3. Lipton HL, Lee PR. . Stanford,Calif: Stanford University Press; 1998:85-117.

4. CalPERS Health Plan Decision Guide: Survey Results.California Public Employees' Retirement System, HealthProgram Development. Sacramento, Calif; 1995.


5. Chernew M, Scanlon DP. Health plan report cardsand insurance choice. . 1998;35:9-22.

Health Serv Res

6. Spranca M, Kanouse DE, Elliott M, Short PF, FarleyDO, Hays RD. Do consumer reports of health plan qualityaffect health plan selection? . 2000;35:933-947.

Health Serv Res

7. Schultz J, Thiede Call K, Feldman R, Christianson J.Do employees use report cards to assess health careprovider systems? . 2001;36:509-530.

J Health Econ

8. Beaulieu ND. Quality information and consumerhealth plan choices. . 2002;21:43-63.

Health Serv Res

9. Farley DO, Short PF, Elliot MN, Kanouse DE, BrownJA, Hays RD. Effects of CAHPS health plan performanceinformation on plan choices by New Jersey Medicaidbeneficiaries. . 2002;37:985-1007.

Health Serv Res

10. McCormack LA, Garfinkel SA, Hibbard JH, NortonEC, Bayen UJ. Health plan decision making with newMedicare information materials. . 2001;36:531-554.

Health Aff (Millwood)

11. Hibbard JH, Jewett JJ, Engelmann S, Tusler M. CanMedicare beneficiaries make informed choices? . 1998;17:181-193.

Value in Health

12. McCombs JS, Nichol MB, Stimmel GL. The role ofSSRI antidepressants for treating depressed patients in theCalifornia Medicaid (Medi-Cal) program. . 1999;2:269-280.

J Affect Disord

13. Hylan TR, Crown WH, Meneades L, et al. Tricyclicantidepressant and selective serotonin reuptakeinhibitors antidepressant selection and health care costsin the naturalistic setting: a multivariate analysis. . 1998;47:71-79.

Health Policy

14. McCombs JS, Shi L, Croghan TW, Stimmel GL.Access to drug therapy and substitution between alternativeantidepressants following an expansion of theCalifornia Medicaid formulary. . 2003;65:301-311.

Clin Ther

15. McCombs JS, Shi L, Stimmel GL, Croghan TW. Aretrospective analysis of the revocation of prior authorizationrestrictions and the use of antidepressant medicationsfor treating major depressive disorder. . 2002;24:1939-1959.

Gen Hosp Psychiatry

16. Katon W, Von Korff M, Lin E, et al. Distressed highutilizers of medical care. DSM-III-R diagnoses and treatmentneeds. . 1990;12:355-362.

Am J Manag Care

17. Russell JM, Berndt ER, Miceli R, Colucci S,Grudzinski AN. Course and cost of treatment for depressionwith fluoxetine, paroxetine, and sertraline. . 1999;5:597-606.


18. Lee LF. Generalized models with selectivity. . 1983;51:507-512.

Clin Ther

19. Sclar DA, Robison LM, Skaer TL, et al. Antidepressantpharmacotherapy: economic outcomes in a healthmaintenance organization. . 1994;16:715-730.

Curr Ther Res

20. Skaer TL, Sclar DA, Robison LM, et al. Economicvaluation of amitriptyline, desipramine, nortriptyline andsertraline in the management of patients with depression.. 1995;56:556-567.

Psychol Med

21. Warrington SJ, Padgham C, Lader M. The cardiovasculareffects of antidepressants. . (Monograph Supplement 16); Cambridge, Mass:Cambridge University Press; 1989.

J Psychopharmacol

22. Mucci M. Reboxetine: a review of antidepressant tolerability.. 1997;11(suppl 4):S33-S37.

CNS Drugs

23. Holm KJ, Spencer CM. Reboxetine: a review of itsuse in depression. . 1999;12:65-83.

Pharmacoepidemiol Drug Saf

24. Carlsen A, Waern M, Ekedahl A, Ranstram J.Antidepressant medication and suicide in Sweden.. 2001;10:525-530.

Acta Psychiatry Scandia

25. Isacsson G. Suicide prevention–a medical breakthrough?. 2000;102:113-117.

Eur Psychiatry

26. Rihmer Z. Can better recognition and treatment ofdepression reduce suicide rates? A brief review. . 2001;16:406-409.


27. Hall WD, Mant A, Mitchell BP, Rendle VA, Hickie IB, McManus P. Association between antidepressantprescribing and suicide in Australia, 1991-2000: trend analysis. . 2003;326:1008-1012.

28. US Food and Drug Administration Web site. Centerfor Drug Evaluation and Research. Worsening depression and suicidality in patients being treated with antidepressantmedications. Available at: AccessedJune 23, 2004.