The STAR"D Project Results:A Comprehensive Review of Findings

Diane'Warden, PbD, MBA, A. John Rusb, MD,Madhukar H. Triuedi, MD, Maurizio Faua, MD,

and Stephen R. .Wisniewski,

PhD

Corresponding authorDiane Warden. PhD. MBADepartment of Psychiatry, The University of Texas SouthrvesternMedical Center at Dallas, 5323 Harry Hines Boulevard,Dallas, TX 75390-9086. LISA.E-rnai l: If iane.Warden@UTSouthwestern.ecluCurrent Psychiatry Reports 2007, 9:449-459Current Medic ine Croup LLC ISSN 1523-3812Copyright @ 2007 by Current Medicine Crr>up LLC

The Sequcnced Treatment Alternatives to Relicve Dcpres-sion trial enrolled outpatients with nonpsychotic rnajordcpressivc disorder treated prospectively in a serics ofrandomized controlled trials. These were conducted inrepresentative primary and psychiatric practices. Remis-sion rates for treatment steps 1 to 4 bascd on the 16-item

Quick Inventory of Depressivc Symptomatology-Self-report were 37Yo, 31"/o,

'l{oh, and 13olo, rcspectively.There were no differences in rcmission rates or timesto remission among mcdication switch or among medi-cation augmentation strategies at any treatmcnt level.Participants who required increasing numbers of trcat-ment steps showed greater dcprcssivc illness burdenand increasingly greater relapse ratcs in the naturalisticfollow-up period (40"/o-71"h\. Prognosis was bettcr at alllcvcls for participants who cntcrcd follow-up in remissionas oppnsed to those who cntcred with response withoutremission. These results highlight the prevalencc of treafment-resistant depression and suggest potential benefitfor using more vigorous treatmonts in the earlier steps.

IntroductionThe Wor ld Heal th Organizat ion has pro jected thatmajor depressive d isorder (MDD) wi l l be the second-leading cause of d isabi l i ty wor ldwide by 2020 [ ] . Thelifetirne risk of MDD is 7o/" to l2o/" for men and 207oto 25"/ " for wornen [2] , and i ts annual cost was est i -ma ted to be $83 .1 b i l l i on i n 2000 [31 . Mos t i nd i v i dua l swi th MDD have a chronic or recurrent course, of tenwi th considerable symptomatology and d isabi l i ty evenbetween episodes t4-6) .

Remission is the accepted goal of acute MDD treat-ment [28] because remitters function better [9] andhave a better prognosis [10] than those who respond(have a symptom reduction) without renitt ing. Efficacytrials with symptomatic volunteers report remissionrates of 22oh to 40% [11]. Horvever, these results maynot p ieneral ize to par t ic ipants typ ical ly seen in c l in ica lpractices because efficacy trials often exclude patientswith chronic depression or rnultiple general medical orpsychiatric comorbid conditions 112,131. In effectivenessstudies with more representative populations, remissionrates are low (LI"/"-30"/") even after 8 ro 12 rnonths oftreatment U4-I71. Furthermore, relapse rates of 10ol' ro45o1, are found within 1 year or less of remission duringmaintenance treatrnent for chronic or recurrent depres-sions [18]. In spite of extensive evidence supporting theefficacy of antidepressants in patients with MDD andthe recognition of modest remission rates with first-stepantidepressants, gaps in our knowledge remain aboutthe choice of second- or third-step treatments if the firststep does not work. Furthermore, previous research hasfocused on strict, protocol-driven, research setting-basedstudies, thus l irnit ing the generalizabii ity to "real world"patients in clinicai practice with attendant medical andpsychiatric cornorbiditres.

The Sequenced Treatment Alternatives to RelieveDepression (STAR.D) trial 119,20,21ro1 is the largestprospective, randomized treatment study to date of outpa-tients with MDD recruited from real world psychiatric andprimary care settings who did not receive adequate ben-efit with initial and, if needed, subsequent antidepressanttreatments. To date, more than 50 manuscripts have beenpublished from the STAR"D sample, with rnore in prepara-tion or under review (see http://www.star-d.org for a l istof current publications). This review summarizes ST;\R*Dfindings to date that address the following questions:

l. What arc the remission and response rates andtime to remission with an init ial selectivc scrotonirlreuptake inhibitor (SSRI)?

2. What participant characteristics are related toremission with an init ial SSRI?

450 Mood Disorders

c r l i c r + c t t

I i - BUP-SR, SERT,3 iT l+ BUP-SR,SERIVEN-XR, or CIT ft | VEN-XR, or CIT

MIRT +VEN-XR

Figure 1.-freatnrent strategies and options in Levels 1 to 4. BUP-SR-bupropion sustained release; BUS-buspirorre; Cl-f-ci talopram;CT-cognit ive therapy; Li- l i thiunr; MIRT-mirtazapirre; NTP-nortr iptyl ine; SERT-sertral ine; T,-tr i iodothyronine; TCP-tranylcypronrine;VEN-X R-venlafaxi ne extended relea se.

6 .

3. Do part icipants dif fer in their acceptance of dif fer-

ent t leatment strate€l ies or options at subsequent

treatment steps?

4. What are the remission and response rates and

times to remission and response for subsequent

treatment stePs?

5. How does cognitive therapy compare with rnedica-

tion treatments at the second treatment step?

7 .

\Vhat part icipant characterist ics are associated with

the ne ed for a greater number of treatment steps?

rWhat arc relapsc rares in those who respond or

remit aftcr onc to four trcacment steps?

8. ! f lhat characterist ics dist inguish part icipants who

leave treatment prematurely?

9. Whatcharacterist ics dist inguish part icipants

treated in primary versus psychiatr ic care?

STAR"D Study OverviewSTAR*D was designed to determine which treatrnents aremost effective following nonretnission or intolerance to aninit ial SSRI or to any of a series of subsequent randomizedtreatments. Over a 37-month period, STAR*D enrolled4041 outpatients aged 18 to 75 years with nonpsychoticMDD at 41 clinical sites across the United States (18 pri-rnary and 23 psychiatric care settings). The study enrolledonly treatmenr-seeking patients (as opposed to symptom-atic volunreers) rvith a clinical diagnosis of nonpsychotic

MDD confirmed with a DSM-IV checklist and a scoregreater than or equal to 14 on the 17-item HamiltonRatirrg Scale of Depression (HRSD,-) [22]. To maximizcgeneralizabil ity of results, the study included patients

with most concurrent psychiatric and general medicalconditions, including those with active substance abuse orsuicidality. as long as outpatient care was appropriate.

Figure 1 depicts the treatment options at each of the four

steps in the series of rials. All participants began with citalo-pram treatment (Level 1). Those intctlerant ro or not remittingwith citalopram could enter Level 2, which included switchingto bupropion-sustained release (SR), cognitive therapn sertra-[ine, or venlafaxine-extended release (XR), or attgmentationof citalopram with bupropion-SR, buspirone, or cognitivetherapy. Those without adequate benefit from medication-only treatments in Level 2 could proceed to Level 3. For those

with inadequate benefit from cognitive therapy as a switch or

augmentation in Level 2, the next step (Level 2A) was a switch

to a second rnedication (bupropion-SR or venlafaxine-XR) to

ensure that all Level 3 enrollees had received and not obtained

adequate benefit from at least two prior rnedication trials.

Level 3 included switches to mirrazapine or norrriptyline and

augmentation of the Level 2 or 2A medication with thyroid

hormone (Tr) or lithium. Level 4 treatments were a switch ro

tranylcypromine or to venlafaxine-XR plus mirtazapine.An innovarive study design fearure-an equipoisc

stratif ied randomized design [23]-permitted participanrs,

as in clinical practice, to accept or decline switch or aug-

rnent treatment strategies as long as sufficient options for

The STAR"D Project Results: A Comprehensive Review of Findings Warden et al. 45'J.

randornization rernained. For example, participants couldallolv randomization to all available treatment optiorls,exclude randomizatic.rn to all switches or to all augmelltsat Level 2 (or 3), exclude or include cognitive therapy as aswitch and/or augment, or exclude all medication srvitchesand augments at Level 2 to guarantee cognitive therapy.

The protocol recommended that treatrlen! visirs ateach level occur at baseline and at weeks 2, 4, 6,9, and12. with an additional visit at week 14 if ueeded. Parrici-pants who rernirted or received an adequate lrenefit couldenter a 1-year naturalistic follow-up period, although allwho did not r:emit were encouraged to lnove to the nextlevel. Participants could rnove to the next level wheneverintolerable side effects were encountered or dosing hadbeen maximized, but substantial symptoms remainedafter several weeks.

In the naturalistic follow-up phase, visits were rec-ornnrended every 2 months, and clinicians were urged tocontinue the acute treatment at the sarne dosage found tobe effective previously. Outcomes in follc.rw-up were gath-ered using an interactive voice response system [24].

Tl're primary research outcome was remission definedby a score of 7 or less on the HRSD,r, obtained by blindedtelephone-based assessors. The secondary outcofi le wasremission defined by a score of 5 or less on the QuickInventory of Depressive Symptomatology-Self-report(QIDS-SRr6) 125-271 obtained at each treatmenr visit.Response was defined as a 50o% or greater reduction in thebaseline QIDS-SR16 score. HRSD,, remission rates weregenerally lower than QIDS-SRl5 rates because participantsrvithout an exit HRSD.- score were deerned nonremitters.Irrtolerance was definei a priori as discontinuation beforeweek 4 for any reason or discontinuation thereafter forirrtoierable side effects.

To ensure adequate dosing for an appropriate periodof t ime, treatment was delivered using measuremellt-based care (MBC) 128ot,291. Thus, nonremission couldrlot be attributed to an inadequate medication trial. N{BCincluded guided but f lexible dosing recommendations atcrit ical decision points based on syrnptom and side effectmeasurements at each treatlnent visit using the QuickInventory of Depressive Syrnptomatology-Clinician-rated(QIDS-C16) I2S-271 and the Frequencv, Intensity, andBurden of Side Effects Rating [30]. A web-based treatmentmonitoring system 129,31) was used to flag and follow upwith the study clinicians when dosing deviations werefound. Clinical research coordinators assisted cliniciansand participallts with study assessments and treatmentdelivery at each clinical site.

STAR*D FindingsPretreatment characteristics of participants in Level 1Of 4041, consenting participants, 607 had HRSD,- scoresless than 14 when rated by the telephone assessors, and234 did not return after the baseline visit. The remain-

ing evaluable sample of 2876 participants entered Level 1.Racial and ethnic distribution was consistent with rhe USCensus. ln this effectiveness sample, participants averaged3.3 concurrent general medical conditions, and abouttwo thirds had at least one concurrent Axis I psychiatricdisorder. The sarnple averaged rnoderately severe depres-sion (HRSD,, = 21.8). More than one fourth had chronicdepression (index episode > 2 years), and three fourthshad recurrent depression. This is in stark contrast to thesamples typically studied in efl icacy trials, in rvhich strictinclusion and exclusion criteria often exclude Axis I or IIIcomorbidit ies, chronic depression, and substance alruse,thus l imiting the generalizabil ity of f indings.

Treatment with citalopram: Level 1Figure 2 shows outcomes for all medication treatments. C)f2876 participants, approximately 28% (n = 790) remittedbased on the HRSD,', and approximately 337o remittedbased on the QIDS-SR,, [28. .1.The response rate, whichincluded those who remitted, was 47o/". Remission andresponse rates and times to renrission or response were notdifferent between those treated in primary and psychiatriccare settings. These remission rates are comparable to the22o/o to 407o remission rates found in 8-week randomizedcontrolled efficacy trials [11] that typically recruit lesscomplicated symptornatic volunteers.

Mean time to QIDS-SR* remission for those whorernitted within the L4-week period was 6.7 weeks. Aboutone half of the remissions occurred after week 6 (Fig. 3).Mean time to response among those who responded was5.7 weeks. Of those who responded, one third did so afterweek 6. Mean dose of ciralopram at exit was 41.8 mg/d.an adequate dose. Mean dose was siruilar in those whoreached remission and those who did not.

Which participant characteristics are related toremission with an init ial SSRI?Participants who were white, female, married, more edu-cated, had higher income, had private insurance, or ivereernployed had significantly higher remission rates withcitalopram. Those disadvantaged by increased numbersof concurrent general rnedical and psychiatric disorders,longer current episodes, and poorer funcrion and qualityof l i fe had significantly lower remission rates [28r.1.

Level 2Of the participants in Level I, L439 were intolerant tocitalopram or received inadequate benefit and movedon to Level 2.

Acceptability of Leuel 2 treatmerxtsThe equipoise stratified randomization design resulted insomewhat unexpected findings for the 1439 participantswho agreed to enter Level2 [321. Only 1o/" accepted random-ization to all seven treatment options. Approximately 50'%accepted medication augment, a*d 57Y" accepted medica-

452 Mood Disorders

39 .0(n = 10B)

32.9h = 943)

26 .6(n = 63)25.0

\n = 62)24.7

(n = 1B )

r 3 .8( n = 8 )12.4

( n = l 5 )

8.0(rt = 9)

{augment} (switch) (switch)(n = 142) (n = 235) {n = 'l 09)

(switch)ln = 727t

( n = 9 )

N

NN

ctT BUS BUP-SR VEN-XR BUP-SR SERT T3 Li NTF MIRT VEN-XR TCP( n = 1 n = 2 8 6 ) ( n = 2 7 i l ( n = 2 5 0 ) 1 n = 2 3 9 ) ( n = 2 3 8 ) t n = 7 3 ) { n = 6 9 ) ( n = 1 2 1 ) ( n = 1 1 4 ) + M I R T ( n = 5 8 )2876) (n = 5t)

LLeve l l JLLeve l 2JLLeve l 2JLLeve l 3 l l Leve l 3JLLeve l 4J{n = {augment)2876i, (n = 565)

( n = 6 1 )

I

Figure 2. Diminishing rerniss ion rates wi th greater levels of t reatnrent resistance. Rernission rates at t reatnrent exi t were based on the . l6- i tenr

Quick Inventory of Depressive Syrnptomatology-Sel f - report . BUP-SR-bupropion sustained re lease; BUS-buspirone; CIT-c i ta loprarn;t , i - l i th ium; MIRT-nr i r tazapine; NTP-nortr ipty l ine; StRT-sert ra l ine; T.- t r i iodothyronine; TCP-tranylcypromine; VEN-XR-venlafaxineextended release.

2 4 6 B 1 0 1 ? > 1 , 3

Weeks

Figure 3. Tinre to remission wit lr ci talopram (l ,evel 1) for remitters.Remission was defined as a score of 5 or less on the 16-itenr QuickI nventory of Depressive Symptomatology-Self-report.

tion switch. Only 7o/" accepted randomization to medication

switch and augment. Not surprisingly, those who accepted

switch treatments had experienced greater intolerance orless improvement with citalopram than those who acceptedaugrnent treatments. Those with recurrent MDD and con-current drug abuse preferred augments. Approxirnately 26%\,vere willing to accept randomization that included cognitivetlrerapy, and 3%" accepted only cognitive therapy (switch oraugment). Those who accepted cognitive therapy had moreyears of education and rnore frequent family history of rnooddisorder than those who did not.

As rnost participants elected to allow randomizationto switch or augment strategies (not both), the study wasnot adequately powered to compare outcotnes for switchversus augment treatments. This finding suggests thatpatients have ciear preferences about switch versus aug-ment as a second step.

Let'el 2 medication switchThe medication switches at Level 2 were designed tocornpare medications with different pharmacologiceffects (ie, sertraline, a second SSRI; bupropion-SR, anonserotonin active agent; venlafaxine-XR, a reuptakeblocker of both norepinephrine and serotonin) [19,201.

The STAR"D Project Results: A Comprehensive Review of Findings 'Warden

et al. 453

Participants entering a medication switch started witha mean HRSDI" of 18.9, and about one fourth of theseparticipants rernitted (QIDS-SRl6) [33..1. Remission raresfor bupropion-SR, sertraline, and venlafaxine-XR werenot differerrt; neither were time to remissiou or t irne roresponse. Mean time to remission for those who remittedranged frorn 5.4 to 6.2 weeks.

These remiss ion and response rates are lowerthan would be expected f rom open- label t r ia ls wi thmedicat ion swi tch 133oo,34- j61. This is l ike ly c lueto STAR' 'D 's inc lus ion of par t ic ipants wi th generalmedical or psychiat r ic comorbid i t ies or chronic depres-sion, as well as the study's longer first-step treatment.These low remiss ion rates do not appear to resul t f ror-ninadequate t reatment , as the doses and durat ions oftreatment seemed rolrust, except for a somewhat lowermean dose of venlafaxine-XR (- 194 rng/d) than thepotent ia l maximuur dose.

Those intolerant to citalopram (Level 1) toleratedser t ra l ine and bupropion-SR equal ly wel l , and a lackof citalopram efficacy did not portend a lack of efficacywi th ser : r ra l ine. The dual-act ion agent (venlafax ine-XR)did not produce significantly higher rernission rates.Thus, ir seems reasonable to switch u'ithin class, out ofc lass, or to a dual -act ion rnedicat ion as a second step.

Let,el 2 medication augmentNo prior randomized controlled trials in real world set-tings have directly compared second-step auglnentation*'ith nontricyclic antidepressants [37]. Participants ran-dornized to augmentation had gained sorne benefit withthe previous citaloprarn monotherapy before entry into arnedicat ion augment (mean HRSDTz = 15.8) [38. .1.

Both rnedication augrnentatior)s appear to be help-fui, with some advantages for bupropion-SR. About onethird of participants remitted with bupropion-SR or bus-pirone, with no differences between groups in remission,response, or t irnes to rernission or response. Participantstreated with bupropion-SR showed greater baseline-to-exit symptorn improvement, lou'er exit syrnptom severity,and fewer dropouts due to intolerance (12.5o/" vs 20.60/").Both rnedications were delivered at adeouate doses foradequate time periods.

Leuel 2 cogttitiue therapy su'itcb or augmentOverall, 147 participants received a cognitive therapyswitch or augment. Sixty-five cognitive therapy augmentparticipants were compared with L17 parttcipants whoreceived rnedication augmentation, and 36 cognitivetherapy switch participants lvere compared with 85 par-ticipants who received rnedication switch [39o1. Aboutone third of cognitive therapy augment and medicationaugment participants rernitted. No significant differenceswere found in rernission or response rates, tolerabil ity, ornumbers of weeks in treatment. There was a statisticallysignificant di{ference in mean tirne to remission (55 days

for cognitive therapy augment, 40 days for medicationaugment). If speed of remission is important, medicationauglnent has an advaDtage.

i\ little more than one fourth of cognitive therapyswitch and rnedication switch participants remitted.There were no differences in time in treatment. remission.response, or t ime to remission or response between thoseswitching to cognitive therapy and those switching tomedication. Of participants who switched to medication,487o experienced at least moderate side effects, whereasnone of those who switched to cognitive therapy reportedside effects, although there were no significant differencesin discontinuations due to intolerance.

Stat is t ica l power for cogni t ive therapy coutpar isonswas low given the smal ler par t ic ipant samples. Thislorv acceptabil ity of cognitive therapy (267:" acceptedthe possib i l i ty of randon-r izat ion) is not consistent wi thpr ior f ind ings [40-42] . Perhaps the use of cogni t ivetherapy in th is s tudy as a second rather than f i rs t t reat-ment dissuaded patients who preferred psychotherapvfrom enrol l ing in STAR" 'D. Al ternat ive ly , an increasedco-pay burden (because of the increased number of v is-its) or the need to travel to another iocation to see atherapist rnay have played a role.

Level 3Of the par t ic ipants in Level 2 and 21 ' ,377 exper iencedintolerance or received inadequare benefit and movedon to Level 3.

Leuel 3 sruitcbThis study [43o] is the first to compare two medicatiorlswitch freatments as a third step following intolerance orlack of remission, prospectiveiy determined, to an SSRIand a second antidepressant. The two switch treatments(mirtazapine ln = 1l4l and nortriptyline [n = 121J) havedifferent phannacologic mechanisms of action. Nor-triptyline represented a third reuptake blocker, whereasmirtazapine does not directly block reuptake. The rneanHRSD,, score at entry was 19.2.

There rvere no signi6cant differences in the modestrernission rates between mirtazapine and nortriptylineby the HRSD T,- (12.3" / " , I9 .8%) or the QIDS-SR'6 (8.0%,

1,2.4%). or in response rares or t i r l res to remiss ion orresponse. Medication tolerabil ity in the previous stepwas unrelated to outcome. Both medications rvere dosedadequately for adequate time periods.

Leuel 3 augmetttThis study [44r] is the first to compare rnedicatiolr aug-mentatiorl treatments as a third step following two trialswith newer antidepressants. Augmerrtation consisted ofadding l ithium (n = 69\ or T, (n = 73) to one of the rnedi-cation switch options in Level 2 or 24. or to cifalopram(for those who had received a Level 2 augmentatiou treat-meltt). Mean HRSDI" score at entry was 18.1.

454 Mood Disorders

90

BO

ITotalS$S Not in remission

, , : , Rcmissiorr

58.7

IW;

83 .3

c

€Ec

4

64.6

50

40

30

Third step

Remission rates were very modest . There were nosigni f icant d i f ferences in rern iss ion rates for l i th iurnor T. augment by the HRSD'7 05.9y" , 24.7%l or the

QIDS-SR'" (13.2o/ , ' , 24.7"1,) , or in response rates ort ime to remiss ion or response. Remiss ion rates were r lo ts igni f icant ly d i f ferent for those using any of the rnedi -cat ions in Level 2.

Those taking l ithiurn had rnore frequent side effectsand were more likely to leave the study due to intolerableside effects despite moderare dosing of l i thium. T, rnay bethe preferential treatment due to its lower intolerance rateand lack of a necessity ro check blood levels, althoughlong-terrn safety data are lacking.

Level 4This is the 6rsr available report [45r] of a fourtir random-ized medication trial based on prospective observations.Participants with intolerance or lack of rernission withthree prior medication treatments were randomized totranylcypromine, a monoarnine oxidase inhibitor (z = 58),or venlafaxine-XR plus mirtazapine (n = 51). The meanHRSDr- score at entry into Level 4 was 19.6.

Rernission rates were remarkably low and sirnilarbetween tranylcypromine and venlafaxine-XR pius rnir-tazapine on the HRSD'7 rc.9'A. 13.7%) or the QIDS-SR,,(13.8%, 15.7%). However, the t ranylcypromine meandose of approximately 37 mgld did not approach the pro-tocol-recommended maximum dosage of 60 mg/d. Therewere no differences between the medications in response,time to rernission, or side effects, although participanrstaking tranylcypromine w'ere rnore l ikely to leave the studyprematurely and to leave due to side effects. The 2-weekwashout period requirement for tranylcyprornine, duringwhich five participants dropped out, also may have beena barrier. There were no differences in remission basedon intolerance in Level 3. The combination of venlafax-

Figure 4. Relapse rate increases wi th eachtreatment step. Relapse rate was calculatedfrom those pat ients who made at least onepostbasel ine cal l to the interact ive voiceresponse system. Treatrnent step pairrvisecompar isons showed only Step 1 tc l d i f fers igni f icant ly f ronr the rest (P < 0.0001).

ine-XR and mirtazapine seems to be a better option thantranylcypromine, given better acceptance and tolerabil ityand lack of d ierary rest r ic t ions.

Overall outcomes and treatment resistanceTo provide an overall evaluation of outcomes, enrolleeswere divided into groups based on the totai number ofacute t reatment s teps (Levels '1 ,2,2A, 3, 4 l [45-481. Ofthe 4041 participants, only the 370 without a postbase-line visit were excluded. Those with an HRSD,- entryscore iess than 14 were included, which resulted in 357Ievaluable participanrs.

For the intent-to-freat group, QIDS-SR'" remissionrates were approximately 37% for step 1. 3IaA for step2, I4y" for step 3, and 13"/" for step 4. Treatrnent ir.rtol-erarlce was 15% for step 1,20"/o for step 2,26o for step3, and 34u/o for step 4.

At study entry, those who received more acute treaf-ment steps had greater general medical i l lness burden,inc luding longer i l lness durat ion (15-20 years) , longerindex episodes (25-42 months) , a greater proport ionwitlr anxious features (45%-57%), and a higher meanHRSD,, score at s tudy entry (19.9-23.3) . They a lsohad a larger proportion of those with at least one AxisI conrorbidity (61%-72%) and more concurrent generalmedical condi t ions (3.0*3.9) .

Longer-term outcomesFigure 4 shows relapse rates during the l-year naturalisticfollow-up by number of acute treatment steps received.Overall, when rnore acute treatment steps were used,higher relapse rates (40%-71%l were found. Participantsin remission at fbllow-up entry were less likely to relapse(34%-50%) than those nor ir remission at entry into fol-low-up (59%-83%1. Participar-rts and clinicians may beincreasingly willing to accept higher levels of s,vrnptorns as

The STAR"D Project Results: A Comprehensive Review of Findings 'Warden

et al. 455

the number of treatment attempts increases. Mean rirne torelapse at all steps rvas short for remitters and those not inrernission, ranging from 2.5 to 4.5 rnonths. Time to relapsewas shorter for those requiring multiple treatment steps.

These long-term results highlight the need to achieveremission rvith acute treatment (as opposed to response)and indicate the need to aggressively achieve the desiredoutcome as early as possible.

Who drops out of treatment?Tl're large STAR"D sample provided a unique opportLr-n i ty to descr ibe w.h ich pat ients drop out of t reat lnentand when they do so. Cost d id not p lay a major ro le,as t reatments and t reatment v is i ts were providecl atno cost . Dropour is an important problem, but l i t t leresearch has been publ ished regarding the phenom-enon. Of the or ig inal sample of 4041 par t ic ipants whoentered c i ta lopram t reatment , 1034 (26%) le f t the studyfor nonmedical reasons [49] in Level 1. About one th i rdof these dropped c lut af ter only a basel ine v is i t . Theseirnmediate at t r i ters were youngier , less educated, andperceived their rnental health functioning to be betterthan those remain ing in f featment . Those who had atleast one postbasel ine v is i t but who dropped out before12 weeks were more l ike ly to be b lack, younger, andless educated. Hispanics were more l ike ly to drop outaf ter return ing once. Having publ ic insurance and hav-ing more psychiat r ic comorbid condir ions were re latedto greater overal l a t t r i t ion. Of note, par t ic ipants wi threcurrent depression were rnore l ike ly to remain int reatment than those wi th a s ingle episode. These datasuggest rhat these indiv iduals rnay benef i t f rom focusedretent ion ef for ts . Al though cer ta in character is t ics areassociated wi th h igher dropout rates, the overal l a t t r i -t ion frorn the study at all levels of treatment indicatesa need to institute preventive procedures involvingpat ient educat ion and at t r i t ion-rnoni tor ing appr<-rachesfc l r a l l pat ients.

Participants in primary versus psychiatric carePrimary care patients have been reported to have lowerseverity of depression [50] and a rnilder course of illness [51]than psychiatric care petienrs.

Based on a comparison of the first 1500 STAR*Dparticipants, rninirnal differences were found in clinicalpresentation of depression between patients in these twosettings [52], including depression severity. Not surpris-ingly, pri inary care participants did have more gerreralmedical cornorbidit ies.

In pr imary care, onset of the f i rs t episode occurredrnore f requent ly af ter age 18 years, t ime s ince onsetof the f i rs t major depressive episode (MDE) was lon-ger , and the current episode was longer. ln specia l tycare, par t ic ipants were rnore l ike ly to have previouslyat tempted suic ide and to have current su ic idal ideat ion.

Analyses of the subsequent 2-541 participants confirmedthese f ind ings [ -531.

Our finding of rninimal differences in clinical presen-tation between primary care and specialty care patientssupports the use of the same methods for screening andmeasuring treatment outcomes in both settings.

STAR"D: Conclusions and ImplicationsSTAR*D addresses questions of substantial public healthsignificance related to treatment for increasingly treat-ment-resistant MDD. Results are generalizable to tvpicalprimary and psychiatric clinical practices.

Chronicity and comorbidity are commonFor these participants, MDD was longstanding, with anaverage length of i l lness of 15 years. The course of i l l-ness was most frequently chronic andlor recurrent andtypically associated with other psychiatric and medicalcclmorbid conditions. Increasing chronicity and numberof comorbid conditions are related to treatment resistance.These participants are not usually included in efficacy tri-als but may represent a substantial proportion of thoseseen in clinical practice. The results of efficacy trials arelikely to be more generalizable when criteria for inclusioncan be safe ly expanded.

MBC may help to improve outcomesClinicians in practice settings typically use patient and cli-nician subjective judgrnent regarding, rreatment efficacy andtolerabil ity rather than a measurement-based approach. Useof this approach could be related to the hig,h rates of treat-ment inadequacy found in clinical practice setrings [2].

Using objective measurernents of symptorns and sideeffects may be helpful when making adequate dosingand time frame determinations to maximize symptolnreduction and rninirnize side effects. MBC may identifysymptorn severity changes that would be less apparent indiscussion with a patient. In this study, MBC rnay haveimpacted the similarity in outcornes betweerr participantsin pr imary and psychiat r ic care.

MBC, including the use of participant self-report rat-ings of symptoms and side effects, was easy to use andacceptable in clinical practice settings. Self-ratings alsornay assist participants in learning to monitor and man-age their own disease.

Remission can take timer\ large percentage of participants at each treatment step didnot reach remission by week 6 or 8. Sorne rerritted in week!4 or later. In the context of acceptable side eff-ects, clini-cians may want to consider at least 8 weeks of treatrnentbefore rnaking a treatment change due to lack of ef6cacy.Study periods of greater than 8 weeks also are needed forclinical trials that have rernission as the erldpoint.

456 Mood Disorders

Many steps may be needed to reach remissionTwo thirds of participants did not remit with init ialcitalopram treatment based on the QIDS-SR,,,. Additionaltreatments resulted in decreasing rernission probabil it ies.The current study was the first to make this determina-tion based on rernission rates.

However. the curnulative rernission rate after twosteps was approximately 50%. With all steps included,alnrost 70"/" oi participants who remained in the stud,vexper ienced remiss ion. Pat ients and c l in ic ians areencouraged to not give up. Sharing specific expecrarionsat the beginning o{ t reatment regarding the probabi l i tyof remission with each subsequent step may help retainpatients in treatrnent without the discouragement thatcan come f ronr unreal isr ic expecrat ions.

Remission is less l ikely for participants with a longertime since first-episode onset, a longer length of currentIVIDE, or more medical and psychiatric comorbidit ies.These participants require particular focus in rreatrnent.

The modest remissic;n rates that result from mulripletreatments, especially the third or fourth medication tri-als, along with prernature treatl l lent discontinr.ration bya large percentage of participants, suggest the need formore effective treatnlents.

Treatmcnt choices for increasinglytrcatmcnt-resistant participantsRemarkably, there were no statistical or meaningful clini-ca ld i f ferences in rern iss ion rates, response rates, or t i rnesto remission or response among any of rhe medicationscompared in this study. All medications were safe andwell-tolerated. Bupropion-SR had sorne advantages overbuspirone as a second-step medication augment agent,including greater irnprovement in symptoms and fewerdropouts due to intolerance. There were also advantagesfor T, over l i thium as a third medication treatlnent dueto fewer side effects, and advantages for venlafaxine-XR plus mirtazapine over tranylcypromine as a fourthmedication treatrnent due to fen'er side effects and lackof need for d ietary rest r ic t ions.

These results did not support the following commonlyheld notions: 1) the advantage <lf a dual-action switchagent as a second-step treatment, 2) lack of efficacy whensrvitching to a second SSRI after an init ial tr ial with anSSRI resulting in intolerance or lack of efficacl', and 3) thegreater efficacy of a second-step switch to an antidepres-sant with a different mechanism of action cornpared witha switch to an agent with the same mechanism. However,based on the very rnodest outcomes in Level 3 switchtreatments, three consecutive monotherapies do not pro-duce efficacious resuits.

Similarly, we found no differences in outcorne whencogr.rit ive therapy as a second-step switch or augmentwas corl]pared with medication switch or aug[tent,respectively, although remission occurred more slowly

with cognitive therapy augment to citalopram than withrnedication augment to citalopram. Surprisingly, rnedica-tion augment was almost as well tolerated as cognitivetherapy augment. When speed of remission is irnportanr,medication augment has an advantage.

When selecting rnedication options, clinicians mustconsider efficacy, side effect burden and tolerability, con-venience of dosing, drug interactions, and parricipanrfidelity. Clinicians must weigh the possibil i ty of rernissionwith continuation of a current trearmerlt against the prob-able efficacy and side effect burden of the other oprions.Adequate dosing for an adequate duration of t ime isimportant for a well-toleratecl treatment. However, dos-ages needed to reach remission rnay be higher than thoseoften administered in clinical practice.

In c l in ica l pract ices, pat ients and thei r : physic iansselect acceptable treattrent approaches. In this study,it was striking that only loh of participants found itacceptable to be randomized to all seven treatrnentoptions at Level 2. However, it is not surprising thatthose with greater side effect burden or less symptomaticimprovement with citalopram preferred a switch ratherthan an augmentation approach. The extent of patientpreferences found in the study raises questions about thegeneralizabil ity of f indings from randomized clinical tri-als without an acceptabil ity component.

Remission results in better long-term outcomesRemission is not a safe haven g iven the h igh re lapserates for participants in remission who entered follow-up and rhe decreased time to relapse found in thosewho required more treatment steps. However, responsewi thout remiss ion is even more precar ious g iven thehigher probabi l i ty of a re lapse, especia l ly as addi t ionalsteps are needed. This argues for a l l reasonable ef for tsto be made to assis t par t ic ipants in reaching rern iss ionand for the continuation of careful clinical monitorinsbeyond rhe t i rne of remiss ion or respor lse.

Participant attrition is highParticipant attrit ion frorn antidepressant treatrnent is asubstantial public health concern, alrhough it has beenaddressed in only a l imited fashion. In the current study,about one fourth of participants left init ial treatmentwith citaloprarn. Participants with sociodemographicdisadvantages and more psychiatric comorbidit ies andrninorit ies were lnore l ikely to drop out, whereas thosewho had prior experience with depressive episodes weremore l ikely to remain. In both clinical trials and practice,it may be advantageous to direct outreach efforts towardparticipants from populations at high risk for attrit ionand those with i l lness-related features associated withattrit ion. Elicit ing and addressing individual barriers toremaining in treatment uray be helpful, as may educationalintervention regarding depression, its chronic or recurrent

The STAR"D Project Results: A Comprehensive Review of Findings Warden et al. 457

course, expectations about improvement in treatment, theimportance of oblective assessment of irnprovement, andthe consequences of dropping out of treatment.

ConclusionsSTAR*D has begun to answer specific questions aboutdepression and i ts t reatment i r r representat ive pract iccsetti l lgs, but many questions remain. Rernissron rates atthe end of two treatrnent steps were approxir-nately 50%and thr.is are encouraging when compared with otherchronic medical i l lnesses. Also, the modest and decreas-ing remission rates found with the third and fourth stepslead to speculation that combination treatrnents maybe valuable ear l ier in t reatment . Concerns about cost .adverse effects, drug-drug interactions, and long-terrneffects of such cornbinations sug€iest the need to pro-spectively evaluate the usefulness of such a strategy.

Al though pharrracologic d i f ferences d id not t rans-late in to large c l in ica l d i f ferences, the quest ion remainshow to rnatch a treatment or treatment sequence toan indiv idual pat ienr us ing c l in ica l characrer is t ics (eg,anxious depression) or b iornarkers. Upcorning STAR*Danalyses are p lanned to address these issues.

AcknowledgmentsThis pro ject has received funding f rom the Nar ionalInst i tu te of Menta l Heal th, Nat ional Inst i tu tes ofHea l th , unde r con t rac t N01MH90003 to UT Sou th -western Medical Center at Dal las (pr inc ipal invest igatorDr. Rush). The content of th is publ icat ion does not nec-essar i ly ref lect the v iews or pol ic ies of the Departmentof Heal th and Human Serv ices. nor does the ment ionof t rade names, commercia l products, or organizat ionsimply end<lrsement by the US governrnent. The authorsrv ish to acknowledge the edi tor ia l support of Jon Ki lnerand the secretarial support of Fast rfford InformationProcessing, Inc. This work is supported by R01 MH-164062-0141, Computer ized Decis ion Support Systemfor Depression, awarded through the Nat ional Inst i tu teof Mental Health (principal investigator Dr. Trivedi)and T32 NIH Training Grant }t4}{067543-04 from theUT Southwestern Mood Disorders Cl in ica l In tervenr ionTraining Prograrn.

Bristol-Myers Squibb, Forest Laboratories, Inc., Glaxo-SmithKline, Inc., King Pharmaceuticals, lnc., Organon,Inc., Pfizer, Inc., and Vyeth provided medications er nocost for this trial.

Dr. Warden currently owns stock in Pfizer, Inc. andpreviously owned stock in Bristol-Myers Squibb.

Dr. Rush fras served as a cr:nsultanf for AdvancedNeuromodulation Systems, lr lc., Best Practice ProjectManagement, Jazz Pharmaceuticals, Irrc., Merck &Co., Inc., Neuronetics, Inc., Cyberonics, Inc., Forest

Pharmaceuticals, Inc., GlaxoSrnithKline, Inc., Pfizer,Inc., and Ono Pharmaceutical Co.; has served as aspeaker for Cyberonics, Inc., Forest Pharmaceuticals,Inc.. and GlaxoSmithKline. Inc.t has received researchsupport from the National Institute of Mental Health,the Robert \7ood Johnson Foundation, and the Stan-ley Medical Research Institute; and has owned stockin P6zer. Inc.

Dr. Trivedi has received research support frolrrBristol-Myers Squibb, Cephalon, Inc., Corcept Thera-peut ics, Cyberonics, Inc. , Forest Pharmaceut ica ls , Inc. ,the Nat ional Inst i tu te of Menta l Heal th, the Nat ionalAll iance for Research in Schizophrenia and Depression,Novartis, Predix Pharmaceuticals, and Wveth; has servedon the advisory board/as a consultant for AstraZenecaPharmaceuticals, LP, Bristol-Myers Squibb, Cephalon,Inc. , Cyberonics, Inc. , E l i L i l ly and Co. , Forest Phar-maceut ica ls , Inc. , GlaxoSmithKl ine, Inc, , Fabre KramerPharmaceut ica ls , Inc. , Neuronet ics, Inc. , Novart is , Van-tagePoint, and Wyeth; and has served as a speaker forBr is to l -Myers Squibb, Cephalon, Inc. , Cyberonics, Inc. ,El i L i l ly and Co. , Forest Pharmaceut ica ls , Inc. , Glaxo-SmithKline, Inc., and Wyeth.

Dr. Fava has received research supporr from AbbottLaboratories, Alkerme s, Inc., Aspect Medical Syr-tems, AstraZeneca Pharmaceuticals, LP, Bristoi-MyersSquibb, Cephalon, Inc. , E l i L i l ly and Co. , Forest Phar-maceut ica ls , Inc. , GlaxoSmithKl ine, Inc. , Johnson &

Johnson, Lichtwer Pharrna GmBH, Lorex Pharmaceu-t ica ls , Novart is , Orgarron USA, Inc. , ParnLab, LLC,Pfizer, Inc., Pharmavite, LLC, Roche, Sanofi/Synthelabo,Solvay Pharmaceuticals, and Wyeth; has served on theadvisory board/as a consultant for Aspect N{edical Sys-tems, AstraZeneca Pharrnaceuricals, LP, Bayer AG, BesrPractice Project Management, Biovail Pharmaceuticals,Inc. , Bra inCel ls , Inc. , Br is to l -Myers Squibb. Cephalorr .Inc. , Compel l is Pharmaceut ica ls , Inc. , Cypress Phar-maceut ica l , Inc. , DOV Pharmaceut ica l , Inc. , E l i L i l lyand Co., EPIX Pharmaceuticals, Inc., Fabre KramerPharmaceut ica ls , Inc. , Forest Pharmaceut ica ls , Inc. ,GlaxoSmithKl ine, Inc. , Grunenthal GmBH, JanssenPlrarnraceutica, Jazz Pharmaceuticals, Inc., Johnson Ec

Johnson, Knoll Pharmaceuticals, Lundbeck, MedAvante,Inc. , Merck & Co. , Inc. , Neuronet ics, Inc. , Novart is ,Nutr i t ion 21, Organon, Inc. , Pamlab, LLC, Pf izer , Inc. ,

PhamraStar, Pharrnavite, LLC, Roche, Sanofi/Synthe-labo, Sepracor, Inc., Solvay Pharmaceuticals, SomaxonPharmaceut ica ls , Somerset Pharmaceut ica ls , Inc. ,Takeda Pharmaceutical Co., and Wyeth; has served as

a speaker {ot AstraZeneca Pharmaceuticals' LP, Boeh-r inger Ingelheim, Br is to l -Myers Squibb, Cephalon, Inc. ,El i L i l ly and Co. , Forest Pharmaceut ica ls , luc. , Glaxo-SrnithKline, Inc., Novartis, Organon, Inc., Pfizer, Inc.,PharrnaStar, and \fyeth; and holds equity in CompellisPharmaceuticals, Inc. and Medi\vante, Inc.

458 Mood Disorders

Dr. Wisnielvski has served as a consultant for Cyberon-ics, Inc., lmaRx Therapeutics, Inc., Bristol-Myers Squibb,Organon, Inc., and Case'Western University.

References and Recommended ReadinePapers of part icular inrerest , publ ished recent ly ,have been highl ighted as:. Of importance. . Of rnajor in lportance

1. Niurray CJ, Lopez AD: Evidence-based heal th pol icy-lessons f rorn the Global Burden of Disease Study. S<: ience'1996,

274:740-743.2. Kcssler RC, Bcrglund P, Denr lcr O, et a l . : Thc epidcmiology

of major depressive disorder: results from the NationalComorbidity Survey Replication (NCS-R). JAMA 2003,289 : -3095 -3105 .

3. Grecnberg PE, Kesslcr RC, Birnbaum HCi, ct a l . : Theeconomic burden of depression in the United States: howdid i t change between 1990 and 2000? / CI in Ps;*chiatr l '2003, 64:1465-147 5.

1. Akiskal HS, Akiskal K: Cyclothymic, hyperthymic anddepressive temperaments as subaffective variant of mood dis-orders. In American Psychiatric Press Reuieu oi Psychiatry.F,dited by Tasntan A. Washington, I)C: Arnerican PsychiarricPress; 1992:43-52.

5. Kocsis JH, Klein DN: Diagnosis and Treatment of Chronic

r ; Dcpre ssior t . Nerv York: Gui l ford Pressl 1995.(6. ' ' Schulberg HC, Katon W, Simon GFi, Rush AJ: Treat ingv maior dcpression in pr imary care pract ice: an update of

the Agency for Health Care Policy and Research PracticeGuidel i r res, Arch Gen Psycbiatry 1.998, 55:1,721-1127.

7. tsal lenger: JC: Cl in ical guidel ines for establ ishing remissionin patients with depression and anxicty. / CIin Psvcbiatry1 999, 60(Suppl 22)t29-34.

8. Tr ivedi MH, DeBatt is ta C, Fawcett J , et a1. : Developingtreatnlent algorithms for unipolar depression in cyberspace:International Psychopharmacology Algorithm Projccc(IPAP). Psl ,cDo ph a rma col B ul l 1.998, 34:355-359.

9. Hirschfeld RM, Dunner DL, Kei tner G, et a l . : Doespsychosocial functionirrg irnprove independent of depressivesymptoms? A comparison of nefazodone, psychotherapy,and their combinat ion. Bio l Psychiatry 2002, 51:123-133.

10. Rush AJ, Kr lemer HC, Sackeinr HA, et a l . : Report by theACNP Task Force on Response and Rernission in MajorDepressive Disorder. Newopsychopbarmacology 2006,3 1 : i 841 -1853 .

11. Depression Guidel ine Panel : Cl in ical Pract ice Guidel ine,Number 5: Depression i r r . Pr imary Care: Volut te 2.Treatment of Major I )epressictn. Rockvi l le , MD: t lSDepartment of Heal th and Hurnan Services, Publ icHeal th Service, Agency for Healrh Care Pol icy andR e s e a r c h ; 1 9 9 3 .

12. Kessler RC, lv ' [cGonagle KA, Zhao S, et a l . : L i fet ime and12-month prevalcnce of DSM-I I I -R psychiatr ic d isorders inthe United States, Results from the National ComorbiditySurvey, Arch Gen Psycbiatry 1991,51t8-19.

13. Weissman MM, l -eaf PJ, Bruce ML, Flor io l . : Theepidemiology of dysthymia in five communities: rates,r isks, comorbid i ty , and t reatment. Am J Psychiatry 1988,145 :815*8 19 .l - in EH, Katon WJ, Simon G!, , et a l . : Achieving guidel inesfor the t rcatment of depression in pr imary care. Is physic ianeducat ion enough? Med Carc 1997,35:831-842.Rost K. Nutt ing P, Smith JL, et a l . : Marraging depressiou asa chronic d isease: a randomised t r ia l of ongoing t reatmentin pr imary c$e, I iMJ 2002,325:934.

Rush AJ, Trivedi M, Carmody TJ, et al.: One-year clinicaloutcomes of depressed public sector outpatients: a bencbmarkfor subscquent studies. Blol Psychiatrv 2004,56:46-53.Schulberg HC, Block MR, Madonia M.] , et a l . : The'usualcare ' of maior depression in pr inrary care pract ice.A rch Fam Med 1997 ,6 ;3 i 4 -339 .Grcdcn JF: Clinical prevention of recurrent depression.The need for paradigm shifts. In Tredtment of RecurrentDepression. Edited by Greden JF. Washington, DC: AnrericanPsychiarric Press; 2001 :143-l 70.Fava M, Rush AJ, Trivedi I\,1H, et al.: Background andrationale for the sequenced treatment alternatives to relievedepression (STAR+D) study, Psychiatr Clin North Am2003,26t457-494.Rush AJ, Fava M, Wisniewski SR, et al.: Sequenced Treat,ment Alternatives to Relieve Depression (STAR*D): rarionaleand design. Control Clin Trials 2004,25:119-142.

ro Rush AJ, Tr : ivedi MH, Wisnrervski SR, er a l . : Acute andlonger- term outcomes in depressed outpat ients rcquir ingone or several t reatment steps: a STAR*D report .Am J Psychiafty 2006, t6321.905-L91.7.

As additional acute treatment steps are required, especially threeor four steps. there are lower rates of rernission and higher ratesof relapse in follow-up. Relapse rales are lower for those enteringfollow-up in rernission as compared with those entering u'ith ar :esponse bur not remission.22. Hami l ton M: A rat ing scalc for depression. / Neurol

N e urosurg P sych iatry 19 60, 23: 5 6-62.23. Lavor i PW' Rush AJ, Wisniervski SR, et a l . : Strengthening

clinical effectiveness trials: equipoise-stratified randomiza-tion. Biol Psychiatry 200 l, 50:792-80 l.

24. Mundt. f C: Interact ive voicc responsc sysrems in c l in icalresearch and t reatment. Psychian Seru 1997,48:6l l -612.

25. Rush AJ, Tr ivedi MH, ibrahirn Hi \ { , et a l . : The 16^l tem

Quick Inventory of Depressive Symptomatology (QIDS),c l in ic ian rat ing (QIDS-C), and sel f - report (QIDS-SR): apsychometric evaluation in patients with chronic maiordepression. Bio l P$,chiatry 2003, 54:573-583. IPubl ishederratum appears in Biol Psychiafiy 2003,.54:.58.5.1

26. Rush AJ, Bernstein IH, Tr ivedi MH, et a l . : An evaluat ionof the Quick Inventory of Depressive Symptomatology andthe Hami l ton Rat ing Scale for Depression: a SequencedTreatment Alternativcs to Relievc Depression trial rcport.Biol P sych i atry 200 6, 59:49 3 -501..

27. Tr ivedi MH, Rush A| , Ibrahim HM, et a l . : The Inventoryof Depressive Symptomatology, Cl in ic ian Rat ing ( IDS-C)and Sel f -Report ( IDS-SR), and the Quick Inventory ofDepressive Symptomatology, Cl in ic ian Rat ing (QIDS-C)and Sel f -Report (QIDS-SR) in publ ic sector pat ients wi thmood disorders: a psychometr ic evaluat ion. Psychol Nied2004,34:73-82.

28.. . Tr ivedi MH, Rush A.f , Wisniewski SR, et a l . : Evaluat ion ofoutcomes with citalopram for depression using measure-ment-based care in STAR*D: implications f.or clinicalpfacrice. Am I Psychiatr.v 2006,163:28-40.

About one third of participants rernitted rvith an initial treatmentrv i th c i ta lopram. This paper a lso de scr ibes MBC.29 . T r i ved i MH, Rush : ! , Gaynes BN , e t a l . : Max i r n i z i ng

the adequacy of medicat ion t reatment in contro l led t r ia lsand c l in ical pract ice: STAR*D measurement-based care.Neuropsychopharmacology- 2007 [Epub ahead of pr int ] .

30. 'Wisniewski

SR, Rush AJ, Balasubramani GK, et al.: Self-rated global measure of the frequency, intensity, and burdenof sidc cffects. / Psychiatr Pract 20Q6,12177-79.

31. l f isniewski SR, Eng H, Meloro L, et a l . : Web-basedcomlnunicat ions and management of a mul t i -center c l in icalt r ia l ; the Sequenced Treatment Al ternat ives to Rel ieveDeprcssion (STARtD) proiect. Clin Trials 200.4, 1:387-398.

32. Wisniewski SR, Fava M, Tr ivedi MH, et a l . : Acceptabi l i tyof second-step treatments to depressed outpatients: aSTAR*D report, Am I Psychiaty 2007,164:753-750.

18.

I Lo\

\tli

19.

20.

21

t4 .

/ l

!v.

The STAR"D Project Results: A Comprehensive Review of Findings Warden et al. 459

33.. . Rush AJ, Tr ivedi MH, Wisniewski SR, et a l . ; Bupropion-SR, sert ra l ine, or venlafaxine-XR af ter fa i lure of SSRIs fordepression. N Engl . l Med2006,35421231-1242.

About one fourth of participants remitted with a Level 2medication srvitch, rvirh no significant difference in ourcomes orto lerabi l i ty amorrg bupropion-SR, sert ra l ine, or vcnlafaxine-XR.34. Fava M, Papakostas GI, Petersen T, et a1. : Switching to

bupropion in lluoxetine-resistant major depressive disorder,Ann Cl in P sychiatr l , 2003, 1517 -22.

3.5. fava M, McGrath PJ, Sheu !fP: Switching to reboxetine: anefficacy and safety study in patients with maior depressivedisorder unresponsive to tluoxetine. I CIin Psychopharmacol2003,23t365-369.

36. Saiz-Ruiz J, lbanez. A, Diaz-Marsa M, et a l . : Ef f icacyof venlafaxine in maior depression resistant to selectiveserotonin reuptake inhibitors. P rog N e ur op s1,ch oplt t rnn colB iol P sy c h iatry 2002, 26t1129 -1134.

37. Thase MFi, Rush AJ: Treatment-resistant depression. InPsychopharmacology: Fourth Generat ion of Progress.Edi ted by Bloorn FE, Kupfer DJ. New York: Raven Press;1 995 :108 l - 1 097 .

38. . . Tr ivedi MH, Fava M, Wisniewski SR. et a l . : Medicat ionaugmcntat ion af ter the fa i lure of SSRIs for depression.N Engl J Met l 2006,354:1243-1.252.

About one third of participants rernitted with a Level 2 medica-t ion augment, wi th no s igni f icant d i f lerence in outcorres berweenbupropion-SR or buspironc augmcrrtation of citalopram. Ilupro-pion-SR r.vas better tolerated and was associated with a greaterreduction in number and severity of depressive symptoms.39. . Thase Mh., Fr iedrran ES, Biggs Mlv{ , et a l . : Cogni t ive

therapy versus medicat ion in augmcntat ion and swi tchstrategies as second-step treatments: a STAR"'D report.An J Psychiatry 2007,164:739-752.

At l ,evel 2, about one th i rd of cognir ive therapy augment andrnedication augment participants rcinittecl, with no significantdifferences in remission rates. A little more than one fourth ofcog-nitive therap.v su'itch and medicirtion switch participants remirted,wrth r to s igni f icant d i f fercnccs i r r rcmist ion mtcs.40. Hollon SD, DeRubeis RJ, Shclton RC, et al.: Prevention of

relapse following cognitive therapy vs medications in moderateto severe depression. Arch Gen Psycbiatr,- 2005,62:417-422.

41. Kel ler Ml) , McCul lough JP, Kle in DN, et a l . : A compar isonof nefazodone, thc cognitivc behavioral-analysis system ofpsychotherapy, and their combination for the treatment ofchronic depression. N Engl J Me d 2000,342:1,462-1470.

42. ' Ihase

MFi, Fr iedman F. lS, Bern-ran SR, et a l . : ls cogni t ivebehavior therapy just a 'nonspeci f ic ' in tervent ion fordepression? A retrospect ive compar ison of consecut ivecohorts t reated wi th cogni t ive behavior therapy orsupport ;ve counsel ing and pi l l p lacebo. I Af fect Disord2000, 57:63-7L.

43. . Fava M, Rush AJ, Wisniewski SR, et a l . : A compar ison ofmir tazapine and nort r ipty l ine fo l lowing two consecut ivefailed medication treatments for depressed outpaticnts: aSTAR*D rcpotl Atn J Psycbiatry 2006,163l.1161-1172.

Rernission rates were modest for Level 3 medicarion switch treat-mcnts, with no signi6cant differences in outcome s or tolerabilitybetween mirtazapine and nortriptyline.44.c Nierenberg AA, Fava M, Tr ivedi MH, et a l . : A compar ison

of l i th ium and T3 augmentat ion fo l lowing two fa i ledmedication tteatments for depression: a STARtI) report.An J Psych iatry 2006, 16.1:1519-1530.

Remission rates were modest for Level 3 rnedicat ion augmentat ion,with no significant differences in remission outcolres betweenlithium and T. augmentation. Gre ater tole rability and ease of useprovides some advantage for T,.45.. McGrath PJ, Steu'arr JW, Fava M, et irl.: Tranylcypromine

versus venlafaxine plus mirtazapine follorving three failedantidepressant medication trials for depression: a STARUDrcporr . Af t t .J Psychiatry 2006, 163:1531- i541.

Remission rates were very modest for Level 4 treatments, '*'ith nosigni6cant differences in ourcomes between tranylcvpromine andthe combinat ion of vcnlafaximc-XR and mir tazapine. Venlafaxine-XR plus rnirtazapine is favored dile to greater tolerability, ease ofuse, arrd lack of d ietary restr ic t ions.46. Fava M: Diagnosis and def n i t ion of t reatment-resistant

depression. Biol Psychiatry 2003, 53:649-659.47. Sackeim HA: The dcfinition and meaning of treatmcnt-rcsis-

tant depression . J C lin P sy c h i a tr y 200 7, 62(Suppl l6\t10 -17.

48. Souerv D, Arnsterdanr J, de lvlonrighny C, et al.: Treatmentresistant depression: methodological overview and opera-tional criteria. Eur N e ur op st- c h op h anna c o I 19 9 9, 9 t83 -9 1.

49. Warden D, Trivedi MH,'Wisniera.ski SR, et a1.: Pretreatmentpredictors of attrition during initial (citalopram) treatmentfor depression: a STAR*D report. An I Psychiatry 2007,164':1189-1197.

50. Wagner HR, Burns B[, Broadhead'S7E, et a1.: Minor depres-sion in family practice: functional rnorbidity, co-morbidity,service utilization and outcomes. Psychol llled 2000,30:1377-1,390.

51. Simon GE, L in EH, Katon W, et a l . : Outcomes of " inad-equate" ant idepressant t reatment. / Cen Intern Med 1995,10:663-670.

52. Caynes I lN, Rush AJ, Tr ivedi M, ct a l . : A di rect compar i -son of present ing character is t ics of depressed outpat ientsf tom pr imary vs. specia l ty care set t ings: prel iminary f ind-ings from the STAR*D clinical trial. Gen Hosp Psychiatry2005,27287-96.

53 . Gaynes BN , Rush AJ , T r i ved i MH. e t a l . : Ma io rdepression symptoms in pr i rnary care and psychiatr icca re se t t i ngs : a c ross - sec t i ona l ana l ys i s . Ann Fam l v4ed2 0 0 7 , 5 : 1 2 6 - 1 3 4 .