CnJPsychitry 2015;60(1):9 13 Perspective The STAR*D Tril: It Is Time to Reexmine the Clinicl Beliefs Tht Guide the Tretment of Mjor Depression H Edmund Pigott, PhD 1 1 Chief Clinicl Officer, Positive Brin Trining, LLC, Juno Bech, Florid; Psychologist, Guiding Light Wellness Center, Wellington, Florid. Correspondence: 430 North Lyr Circle, Juno Bech, FL 33408; pthwre@erols.com. Key Words: depression, Sequenced Tretment Alterntives to Relieve Depression, STAR*D, selective outcome reporting Received June 2014, revised, nd ccepted July 2014. The STAR*D tril is the lrgest nd most consequentil ntidepressnt study ever conducted, with over 120 journl rticles published by study investigtors, innumerble cittions of STAR*D s findings by other reserchers, nd extensive coverge in the medi, thereby giving it n oversized impct on the tretment of depression, worldwide. Funded t cost of 35-million US dollrs, STAR*D enrolled 4041 ptients who screened positive for mjor depression while seeking routine medicl or psychitric cre. In contrst to most industry-funded trils, STAR*D included depressed ptients with comorbid conditions, thereby incresing the generlizbility of its findings nd further, provided 12 months of free continuing cre to monitor the durbility of tretment effects. STAR*D provided up to 4 tretment steps per ptient nd ws designed to give guidnce in selecting the best next-step tretment for the mny ptients who fil to get sufficient relief from their initil AD. Ech step consisted of 12-week, open-lbel tril, with n dditionl 2 weeks for ptients deemed close to remission. ADs were dministered using system of mesurement-bsed cre tht involved ssessing symptoms nd side effects t ech visit to guide ggressive mediction dosing to ensure tht the likelihood of chieving remission ws mximized nd tht those who did not rech remission were 1, p 30 truly resistnt to the mediction. STAR*D llowed ptients to select tretment options for rndomiztion in steps 2 to 4 to empower ptients, strengthen the therpeutic llince, optimize tretment dherence, nd improve outcome 2, p 483 nd evluted the reltive effectiveness of 11 phrmcologiclly distinct drug drug combintion tretments in 5 hed-to-hed comprisons. CT ws lso vilble s switch or drug ugmenttion option in step 2, but too few ptients included it s n cceptble tretment, resulting in only 101 contributing dt fter rndomiztion. Therefore, CT ws excluded from the step 2 switch nd ugmenttion nlyses. 3,4 Ptients who chieved remission during ny step were encourged to enter the 12 months of free follow-up cre, s were responder ptients who filed to ttin remission but did not wnt to continue to the next tretment step s this would involve chnge in mediction. Remission ws defined s score of less thn 8 on the HRSD, the study s prespecified primry outcome mesure, nd response s 50% or greter reduction in depressive symptoms on it. The follow-up protocol strongly recommended tht prticipnts continue the previously effective cute tretment mediction(s) t the doses used in cute tretment but treting physicins were llowed to mke ny psychotherpy, mediction, or mediction dose chnge 5, p 1908 they deemed necessry to sustin positive outcome during follow-up, including scheduling dditionl visits if depressive symptoms returned nd (or) intolerble side effects emerged. 6 www.thecjp.c The Cndin Journl of Psychitry, Vol 60, No 1, Jnury 2015 W 9
Perspective Prior Criticisms of Apprent Bis in the Reporting of STAR*D Findings Pigott nd collegues 7,8 hve previously criticized the investigtors for the following: 1) not reporting in their summry rticle 5 remission nd response rtes using the prespecified HRSD but insted using the QIDS-SR, nonblinded, clinic-dministered ssessment tht ws excluded from use s reserch mesure in the Reserch Protocol 9 ; 2) excluding from nlysis ptients who strted on citloprm in their bseline visit nd then dropped out without tking the exit HRSD despite the investigtors sttement tht our primry nlyses clssified ptients with missing exit HRSD scores s nonremitters priori 1, p 43 nd these erly dropout ptients therefore should hve been counted s tretment filures s prespecified; 3) sserting in their summry rticle theoreticl cumultive remission rte of 67% with the unrelistic provisos tht this estimte ssumes no dropouts, nd it ssumes tht those who exited the study would hve hd the sme remission rtes s those who styed in the protocol 5, p 1910 ; nd 4) other indictors of bis (see etble 1). As Pigott et l 7 document, the investigtors ssumptions in clculting their theoreticl remission rte of 67% re simply not true in the rel world, nd were certinly not true in STAR*D, s more ptients dropped out in ech step thn remitted. Tody, STAR*D investigtors provisos concerning their theoreticl remission rte re sometimes dropped when portrying its findings. For exmple, recent editoril in the AJP sttes STAR*D found, fter four optimized, well-delivered tretments, pproximtely 70% of ptients chieve remission 10, p 580 s though this is fctul sttement of wht occurred. Figure 7 in STAR*D s Reserch Protocol hs step-bystep predictions of ptient drop out nd the number of ptients who would hve stisfctory response nd enter follow-up. 9, p 35 The Protocol ws obtined from NIMH in Abbrevitions AD ntidepressnt AJP Americn Journl of Psychitry CCJM Clevelnd Clinic Journl of Medicine CT cognitive therpy HRSD Hmilton Rting Scle for Depression NIMH Ntionl Institute of Mentl Helth PCP primry cre physicin QIDS-SR Quick Inventory of Depressive Symptoms Self Report RCT rndomized controlled tril STAR*D Sequenced Tretment Alterntives to Relieve Depression 10 W L Revue cndienne de psychitrie, vol 60, no 1, jnvier 2015 Freedom of Informtion Act request. The predictions reflect the predicted ggregte outcomes for STAR*D s sequentil tret-to-remission model of mesurement-bsed cre. STAR*D s investigtors found no significnt group differences between ny of the 11 drug drug combintion tretments. Further, no post hoc secondry nlyses hve reported significnt predictors of outcomes between the phrmcologiclly distinct tretments. Therefore, STAR*D provides no next-step guidnce to give hope for improving outcomes beyond tht found in the study itself. As Brbui et l 11 note, AD study completion rtes provide hrd mesure of tretment effectiveness nd cceptbility p 296 nd STAR*D s investigtors essentilly ignore this fct when reporting study outcomes. Elsewhere, Pigott 8 hs detiled the extensive efforts STAR*D mde to keep ptients in tretment nd mximize their likelihood of chieving remission nd mintining it during the 12 months of follow-up cre s well s completing the scheduled reserch ssessments; thus it is essentil to incorporte dropout ptients into the evlution of study outcomes. Methods The uthor bstrcted from the Reserch Protocol nd summry rticle s 5 published tbles nd figures to reclculte STAR*D s remission nd relpse rtes using 4041 s the denomintor, s ll ptients were strted on citloprm in their bseline visit. This pproch seems more relistic thn the summry rticle 5 tht used 3671 s the denomintor, thereby excluding from nlysis the 370 erly dropout ptients who did not return for subsequent tretment visits. The Dt Figure 1 compres the cumultive step-by-step predicted per cent of ptients who would be successfully treted nd enter follow-up nd STAR*D s theoreticl remission rte to wht occurred in the study. In the summry rticle, 5 STAR*D s reserchers used the nonblinded QIDS-SR to report cute nd follow-up cre outcomes insted of the protocol-specified HRSD. This ssessment ws dministered t ech clinic visit nd monthly by telephone during follow-up. The QIDS-SR provided more complete dt set s 690 ptients exited the study in step 1 lone, without tking the exit HRSD, 152 of whom hd QIDS-SR defined remission. 1, p 34 The cumultive per cent of ptients who hd remission determined by their finl QIDS-SR fter up to 4 tretment trils ws 45.9%. By step 4, the cumultive per cent of such ptients who entered follow-up ws only 37.6%. While both clcultions used 4041 s the denomintor, these findings improve little when using 3671 ptients s the denomintor (50.5% nd 41.4%, respectively). The dt STAR*D investigtors provide for ccessing the durbility of tretment gins re even more discourging. For step 1, only 17.8% of citloprm-treted ptients hd remission s determined by their lst clinic-dministered www.lrcp.c
The STAR*D Tril: It Is Time to Reexmine the Clinicl Beliefs Tht Guide the Tretment of Mjor Depression Figure 1 Comprison between predicted, theoreticl, nd ctul step-by-step success rtes 80 70 Predicted % to enter follow -up 60 Theoreticl remission rteb Rte, % 50 40 30 20 c Remitted ptients determined by lst QIDS-SR who entered follow-up d Remitted ptients e Remitted ptients without confirmed relpse 10 Without relpse nd (or) drop out f 0 Step 1 Step 2 Step 3 Step 4 Step Clculted from Figure 7 in STAR*D s Reserch Protocol predicting how mny ptients would hve stisfctory response nd enter follow-up. b Clculted by STAR*D s uthors in the summry rticle. 5 c Clculted from the summry rticle s Tble 3 Remission t ech step row but using 4041 s the denomintor s ll ptients were strted on citloprm in their bseline visit (tht is, (1346 + 439 + 53 + 16)/4041 = 45.9% by step 4). d Clculted from the summry rticle s Tble 5, column 2, rows 3, 6, 9, nd 12, by dding the number of remitted ptients entering follow-up nd dividing by 4041 (tht is, (1085 + 383 + 35 + 15)/4041 = 37.6% by step 4). e Clculted from the summry rticle s Tble 5, column 6, rows 3, 6, 9, nd 12. For ech step, subtrct the per cent relpse rte from 1.0 nd then mulitply by the number of remitted ptients entering follow-up (for exmple, step 1: 1.0 0.335 = 0.665 x 1085 = 722 remitted ptients who did not hve confirmed relpse during follow-up). Repet for steps 2 to 4. Add the remitted ptients without confirmed relpse for steps 1 to 4 nd divide by 4041. f Clculted from the summry rticle Figure 3 s tble, column 5, by dding the number of surviving remitted ptients without relpse nd (or) drop out during months 10 to 12 nd dividing by 4041 (tht is, (84 + 20 + 2 + 2)/4041 = 2.7% by step 4). QIDS-SR = Quick Inventory of Depressive Symptoms Self Report; STAR*D = Sequenced Tretment Alterntives to Relieve Depression QIDS-SR nd during follow-up did not hve confirmed relpse on 1 or more of the 12 monthly dministered telephonic QIDS-SR ssessments. After up to 4 rounds of AD drug drug combintion tretments, the cumultive rte of ptients who did not hve confirmed relpse improved to only 23.5%. When drop out is dded, the durbility of tretment effects is even pltrier. Only 2.7% of ptients hd QIDS-SR determined remission fter up to 4 rounds of AD drug cre nd neither relpsed nor dropped out s evidenced by tking t lest 1 of the months 10-to-12 QIDS-SR telephonic ssessments nd not scoring s hving relpsed in ny of the 12 monthly dministered ssessments. Figure 2 presents the rtes of intolerble side effects nd drop out by tretment step tken from the summry rticle s 5 Tble 3 nd Figure 1. It demonstrtes how ech chnge in tretment resulted in n incresed rte of intolerble side effects from the newly prescribed mediction, compred with the prior step, incresing from 16.3% in step 1 to 30.1% by step 4. This sme trend is seen for study drop out, with it incresing from 28.1% in step 1 to 42.3% by step 3. Both mesures demonstrte n incresing risk to ptient cre tied to ech chnge in AD drug drug combintion tretment s ptients progressed from step 1 through steps 2 to 4. The Need for Reexmintion Following publiction of STAR*D s steps 1 to 4 nd summry results in 2006, 1 5,12 14 its investigtors hve continued publishing t high rte of over 120 journl rticles, including the recent AJP one which precipitted the editoril cited bove. 15 In ddition to not disclosing devitions from the prespecified criteri, often in these rticles the investigtors interpret wht they do report in wys tht led to inccurte conclusions nd potentilly hrmful recommendtions for ptient cre. For exmple, in their CCJM rticle s Key Points section, the investigtors stte: With persistent nd vigorous tretment, most ptients will enter remission: bout 33% fter one step, 50% fter two steps, 60% fter three steps, nd 70% fter four steps (ssuming ptients sty in 16, p 57 tretment). The investigtors then open this rticle stting: DEPRESSION cn be treted successfully by primry cre physicins under rel world conditions. Furthermore, the prticulr drug or drugs used re not s importnt s following rtionl pln: giving ntidepressnt medictions in dequte www.thecjp.c The Cndin Journl of Psychitry, Vol 60, No 1, Jnury 2015 W 11
Perspective Figure 2 Rtes of intolerble side effects nd drop out by tretment step 45 40 35 30 Rte, % 25 20 15 10 5 Intolerble side effects Study drop out b 0 Step 1 Step 2 Step 3 Step 4 Step From the Intolerble side effects row, Tble 3, of the summry rticle. 5 b Clculted from Figure 1 by dding ll ptients who exited the study in steps 1 to 3 nd dividing by the number enrolled in ech step. doses, monitoring the ptient s symptoms nd side effects nd djusting the regimen ccordingly, nd switching drugs or dding new drugs to the regimen only fter n dequte tril. These re mong the lessons lerned from the Sequenced Tretment Alterntives to Relieve Depression study, the lrgest prospective clinicl tril of tretment of mjor 16, p 57 depressive disorder ever conducted. There re severl noteworthy points in this rticle s summry of STAR*D s lessons lerned for PCPs who prescribe most ADs. First when summrizing their results, the highly theoreticl 67% remission rte is rounded up to 70%. Second, the summry rticle s proviso tht this rte ssumes tht those who exited the study would hve hd the sme remission rtes s those who styed in the protocol 5, p 1910 is dropped, leving only the ssumption of no dropouts. Third, nowhere in this rticle nor elsewhere do the investigtors cknowledge tht through step 3, 43% of ptients hd in fct dropped out despite their best efforts in designing the study so s to minimize this hrd mesure of tretment (in)effectiveness 11, p 296 from occurring. Fourth, while correctly stting tht the prticulr drug or drugs used re not s importnt 16, p 57 s STAR*D provides no next-step guidnce to improve outcomes, the investigtors strongly endorse here s elsewhere their system of mesurement-bsed cre s the mens to improve sid outcomes. For exmple, the investigtors stte: high qulity of cre ws delivered (mesurementbsed cre)... Consequently, the outcomes in this report my exceed those tht re presently obtined in dily prctice wherein neither symptoms nor 12 W L Revue cndienne de psychitrie, vol 60, no 1, jnvier 2015 side-effects re consistently mesured nd wherein prctitioners vry gretly in the timing nd level of 5, p 1914 dosing [emphses dded]. As Pigott et l 7 noted, STAR*D s sequentil tret-toremission model of mesurement-bsed cre my hve been detrimentl to ptient cre leding to worse outcomes thn wht occurs in dily prctice. STAR*D encourged ll ptients who did not chieve remission bsed on number to enter the next tril despite the filure of the QIDS/HRSD to differentilly weigh core depressive symptoms (e.g. mood, guilt, suicidl idetion, or nhedoni) nd ccessory ones (e.g. ppetite, insomni, or gittion) nd ptients self-ssessments of the reltive importnce of ech. p 277 Fifth, STAR*D s investigtors my hve overstted the benefits of remission reltive to ptients with merely substntil improvement s both groups hd high rtes of confirmed relpse during follow-up nd STAR*D s symptom-driven pursuit of more ggressive tretment ws not shown to be better thn usul cre. More importntly, by encourging PCPs whose ptients chieve merely substntil improvement from one AD to switching drugs or dding new drugs to the regimen 16, p 57 in their pursuit of less thn 8 HRSD score, STAR*D s investigtors fil to cknowledge the risks tht cme with ech such chnge; tht is, the step-by-step incresing rtes of drug intolernce nd study drop out. The mnner in which STAR*D s nlyses were conducted nd results communicted hve creted nrrtive round its ggressive tret-to-remission model of mesurement-bsed cre tht is not ligned with the ctul results of the study. www.lrcp.c
The STAR*D Tril: It Is Time to Reexmine the Clinicl Beliefs Tht Guide the Tretment of Mjor Depression STAR*D ws filed tril with results tht were not shown to be superior to tretment-s-usul. Despite this lck of evidence though, some STAR*D investigtors continue to dvocte strongly for the superiority of their mesurementbsed model of cre 17 19 even rguing for the doption of finncil incentives nd 9 other policy chnges to mke it the stndrd-of-cre for psychitric prctice. 20 The STAR*D dt set re now vilble from NIMH for independent nlysis by university-bsed reserchers. 21 Reserchers re urged to request this dt set for independent nlysis of the steps 1 to 4 nd follow-up outcomes ccording to STAR*D s prespecified criteri. Only then will the lessons lerned from this once-in--genertion study be fully known nd mde vilble to inform clinicl prctice nd guide subsequent reserch efforts to improve on the sme. Acknowledgements In the pst 3 yers, Dr Pigott hs consulted for Amen Clinics, CNS Response, nd the Interntionl Society of Neurofeedbck nd Reserch. He lso owns stock options in CNS Response, compny tht mrkets quntittive electroencephlogrm dtbse to predict mediction response, nd is prtner in Positive Brin Trining, LLC, neurofeedbck prctice. He ffirms tht the mnuscript is n honest, ccurte, nd trnsprent ccount of his nlysis of wht he hs derived from the STAR*D investigtors published tbles nd figures reporting their findings nd then compring them to the Reserch Protocol he obtined from NIMH through the Freedom of Informtion Act. Editor s Note References 22 nd 23 cn be found in etble 1. References 1. Trivedi MH, Rush AJ, Wisniewski SR, et l. Evlution of outcomes with citloprm for depression using mesurement-bsed cre in STAR*D: implictions for clinicl prctice. Am J Psychitry. 2006;163:28 40. 2. Fv M, Rush AJ, Trivedi MH, et l. Bckground nd rtionle for the Sequenced Tretment Alterntives to Relieve Depression (STAR*D) study. Psychitr Clin North Am. 2003;26:457 494. 3. Rush AJ, Trivedi MH, Wisniewski SR, et l. Bupropion-SR, sertrline, or venlfxine-xr fter filure of SSRIs for depression. N Engl J Med. 2006;354:1231 1242. 4. Trivedi MH, Fv M, Wisniewski SR, et l. Mediction ugmenttion fter the filure of SSRIs for depression. N Engl J Med. 2006;354:1243 1252. 5. Rush AJ, Trivedi MH, Wisniewski SR, et l. Acute nd longer-term outcomes in depressed outptients requiring one or severl tretment steps: STAR*D report. Am J Psychitry. 2006;163:1905 1917. 6. Trivedi MH, Stegmn D, Rush AJ, et l. STAR*D clinicl procedures mnul [Internet]. Pittsburgh (PA): University of Pittsburgh; 2002 Jul 31 [dte cited unknown]. Avilble from: http://www.edc.gsph.pitt.edu/strd/public/study_mnuls.html. 7. Pigott HE, Leventhl AM, Alter GS, et l. Efficcy nd effectiveness of ntidepressnts: current sttus of reserch. Psychother Psychosom. 2010;79:267 279. 8. Pigott HE. STAR*D: tle nd tril of bis. Ethicl Hum Psychol Psychitry. 2011;13(1):6 28. 9. Ntionl Institute of Mentl Helth (NIMH). Sequenced Tretment Alterntives to Relieve Depression (STAR*D) reserch protocol. Revised ed. Wshington (DC): NIMH; 2002. Received October 1, 2010, from NIMH following request under the Freedom of Informtion Act. Also vilble from: http://www.mdinmeric. com/wp-content/uplods/2011/12/star-d%20protocol%20 with%20anlytic%20pln.pdf. 10. Greden JF. Workplce depression: personlize, prtner, or py the price. Am J Psychitry. 2013;170:578 581. 11. Brbui C, Furukw TA, Ciprini A. Effectiveness of proxetine in the tretment of cute mjor depression in dults: systemtic re-exmintion of published nd unpublished dt from rndomized trils. CMAJ. 2008;178:296 305. 12. Fv M, Rush AJ, Wisniewski SR, et l. A comprison of mirtzpine nd nortriptyline following two consecutive filed mediction tretments for depressed outptients: STAR*D report. Am J Psychitry. 2006;163:1161 1172. 13. Nierenberg AA, Fv M, Trivedi MH, et l. A comprison of lithium nd T3 ugmenttion following two filed mediction tretments for depression: STAR*D report. Am J Psychitry. 2006;163:1519 1530. 14. McGrth PJ, Stewrt JW, Fv M, et l. Trnylcypromine versus venlfxine plus mirtzpine following three filed ntidepressnt medictions trils for depression: STAR*D report. Am J Psychitry. 2006;163:1531 1541. 15. Trivedi MH, Morris DW, Wisniewski SR, et l. Increse in work productivity of depressed individuls with improvement in depressive symptom severity. Am J Psychitry. 2013;170:633 641. 16. Gynes BN, Rush AJ, Trivedi MH, et l. The STAR*D study: treting depression in the rel world. Clev Clin J Med. 2008;75:57 66. 17. Morris DW, Trivedi MH. Mesurement-bsed cre for unipolr depression. Curr Psychitry Rep. 2011;13(6):446 458. 18. Shelton RC, Trivedi MH. Using lgorithms nd computerized decision support systems to tret mjor depression. J Clin Psychitry. 2011;72(12):e36. 19. Trivedi MH. Evluting nd monitoring tretment response in depression using mesurement-bsed ssessment nd rting scles. J Clin Psychitry. 2013;74(7):e14. 20. Hrding KJ, Rush AJ, Arbuckle M, et l. Mesurement-bsed cre in psychitric prctice: policy frmework for implementtion. J Clin Psychitry. 2011;72(8):1136 1143. 21. Ntionl Institute of Mentl Helth (NIMH). Avilble limited ccess dtsets from NIMH clinicl trils: Sequenced Tretment Alterntives to Relieve Depression (STAR*D). Wshington (DC): NIMH; [yer of publiction unknown] [cited 2013 Aug 26]. Avilble from: http://www.nimh.nih.gov/trils/dtsets/ imh-procedures-for-requesting-dt-sets.shtml. 22. Ntionl Institute of Mentl Helth (NIMH). New strtegies help depressed ptients become symptom-free [Internet]. Wshington (DC): NIMH; 2006 [cited 2013 Aug 26]. Avilble from: http://www.nimh.nih.gov/science-news/2006/new-strtegieshelpdepressed-ptients-become-symptom-free.shtml. 23. Hmilton M. Development of rting scle for primry depressive illness. Br J Soc Clin Psychol. 1967;6:278 296. www.thecjp.c The Cndin Journl of Psychitry, Vol 60, No 1, Jnury 2015 W 13