Opioid Use and Survival at the End of Life: A Survey of a Hospice Population

532 Journl of Pin nd Symptom Mngement Vol. 32 No. 6 December 2006 NHPCO Originl Article Opioid Use nd Survivl t the End of Life: A Survey of Hospice Popultion Russell K. Portenoy, MD, Un Sibircev, BA, Rndll Smout, MS, Susn Horn, PhD, Stephen Connor, PhD, Ronld H. Blum, MD, Crol Spence, RN, nd Perry G. Fine, MD Deprtment of Pin Medicine nd Pllitive Cre (R.K.P., U.S.) nd Cncer Center (U.S., R.H.B.), Beth Isrel Medicl Center, New York, New York; Institute for Clinicl Outcomes Reserch (R.S., S.H.), Slt Lke City, Uth; Ntionl Hospice nd Pllitive Cre Orgniztion (S.C., C.S., P.G.F.), Alexndri, Virgini; nd Deprtment of Anesthesiology (P.G.F.), University of Uth Medicl Center, Slt Lke City, Uth, USA Abstrct Concern tht opioids hsten deth my be mong the resons tht pin is treted indequtely in popultions with dvnced illness. Studies tht ssess the true risks re needed. To determine whether survivl fter lst opioid dose chnge is ssocited with opioid dosing chrcteristics nd other fctors, dt from the Ntionl Hospice Outcomes Project, lrge prospective cohort study involving 13 U.S. hospice progrms, were nlyzed. Of 1,306 ptients, 725 received opioids nd underwent t lest one dose chnge before deth. Subsmples bsed on mximum opioid dose compred ptients receiving usul doses with those receiving high-dose therpy. Spermn rnk correltions exmined bivrite ssocitions between survivl fter finl dose chnge nd other vribles, including dose in morphine equivlent mg nd percentge dose increse. Multivrite lest squres regression nlyses determined ssocitions between survivl nd other vribles, including those significnt in bivrite nlyses. The men SD number of dys between finl dose chnge nd deth ws 12.46 23.11. Multivrite models demonstrted significnt ssocition between shorter survivl nd higher opioid dose, cncer dignosis, unresponsiveness, nd pin of <5 on0e10 scle, but none of these models explined >10% of the vrince in time till deth. Anlyses of subsmples did not revel dditionl effects of dose. This nlysis reveled tht opioid dosing ws ssocited with time till deth, but this fctor would explin very little of the vrition in survivl. In hospice popultion, survivl is influenced by complex fctors, mny of which my not be mesurble. Bsed on these findings, concern The Ntionl Hospice Outcomes Project ws supported by grnt from the Robert Wood Johnson Foundtion. Additionl support for sttisticl nlysis ws provided by the VistCre Foundtion. Neither funder ws involved in the design nd conduct of the study; the collection, mngement, nlysis, nd interprettion of the dt; or the preprtion, review, or pprovl of the mnuscript. Ó 2006 U.S. Cncer Pin Relief Committee Published by Elsevier Inc. All rights reserved. Address reprint requests to: Russell K. Portenoy, MD, Deprtment of Pin Medicine nd Pllitive Cre, Beth Isrel Medicl Center, First Avenue t 16th Street, New York, NY 10003, USA. E-mil: rportenoy@chpnet.org Accepted for publiction: August 28, 2006. 08-3924/06/$esee front mtter doi:10.1016/j.jpinsymmn.2006.08.003

Vol. 32 No. 6 December 2006 Opioid Use nd Survivl t the End of Life 533 bout hstening deth does not justify withholding opioid therpy. J Pin Symptom Mnge 2006;32:532e540. Ó 2006 U.S. Cncer Pin Relief Committee. Published by Elsevier Inc. All rights reserved. Key Words Hospice, terminl illness, deth nd dying, opioids, risk of hstened deth Introduction Although pin is highly prevlent in popultions with dvnced illness nd the need for ggressive opioid therpy is widely ccepted, undertretment is common. 1 Fctors tht contribute to undertretment re complex 2 nd my include clinicin reluctnce to prescribe becuse of concern bout the potentil for serious dverse drug effects, which my hsten deth. The criticl blnce between the need to reduce suffering through ggressive phrmcotherpy nd the potentil to do hrm hs been discussed extensively in both the medicl nd ethics literture. 3e12 Idelly, ongoing discussion surrounding the risks, benefits, nd ethicl foundtion for the use of opioids t the end of life should be informed by empiricl observtions tht specificlly ssess opioid-relted risk in terms of the key concerndthe potentil to hsten deth when these drugs re used to tret pin. Remrkbly, there is very little evidence of this type. The few studies extnt suggest tht opioids ctully pose little risk of hstened deth in popultions with dvnced illness. 13e16 Additionl studies re needed. The Ntionl Hospice Outcomes Project (NHOP), prospective longitudinl survey of ptients dmitted to hospice progrms, collected extensive dt pertining to clinicl outcomes during end-of-life cre. These dt were nlyzed to explore the reltionship between opioid use nd survivl. Methods The NHOP ws conducted by the Ntionl Hospice nd Pllitive Cre Orgniztion during 2001e2003. Dt were collected from 1,306 ptients dmitted to 13 hospices cross the United Sttes. Bsed on clinicl prctice improvement methodology, the survey included review of hospice medicl records nd direct ptient ssessments by clinicins. Methodologicl detils re described elsewhere. 17 Cse Definition From 1,163 ptients who received n opioid drug (89% of the totl smple), 725 ptients (55.5% of the totl smple) who underwent t lest one recorded opioid dose chnge prior to deth were selected for study. The intervl between the lst opioid dose chnge nd deth ws the key vrible of interest. This intervl ws nlyzed in reltion to chrcteristics of the opioid regimen nd vriety of other fctors. Subsmples bsed on mximum dily opioid dose dministered during hospice dmission were selected to provide more interpretble informtion bout ptients who re commonly encountered in prctice nd to seprtely nlyze subgroups tht received high-dose opioid therpy. Ptients were tking different opioids, nd for ll nlyses, opioid doses were converted into milligrms equivlent to intrvenous morphine. 18 Totl dily dose ws the sum of the fixed scheduled dose plus s needed doses. One subsmple ws defined to include ptients (88.3% of 725) who received mximum dose of #200 intrvenous morphine equivlent mg per dy (IVME). This mximum dose, which is roughly equivlent to #600 orl morphine equivlent mg per dy, 18 is within the rnge of doses widely encountered in the U.S. hospice community. 13,14 In bivrite nlyses, this group ws compred to the remining group of ptients (11.7% of 725), who received mximum dose of >200 IVME. Another subsmple isolted smll group of ptients (2.6%) who required extremely high opioid doses nd ws defined to include those who were receiving >600 IVME (roughly equivlent in orl dosing to >1800 mg morphine per dy 18 ). These ptients were

534 Portenoy et l. Vol. 32 No. 6 December 2006 compred with the remining subsmple of ptients (97.4% of 725), who received #600 IVME. Mesures Dt included demogrphics nd disese-relted informtion, signs nd symptoms, tretments nd processes of cre, nd ptient nd fmily outcomes. 17 In ddition to opioid dosing informtion, the present study evluted hospice length of sty, demogrphics, type of disese, disese severity score (determined with the Comprehensive Severity Index, which yields higher scores in ptients with greter disese severity,20 ), pin nd level of consciousness t the time of lst opioid dose chnge, nd lst recorded performnce sttus (mesured with the Pllitive Performnce Scle (PPS), on which higher scores indicte better functioning 21 ). Sttisticl Anlyses The gol of the nlyses ws to determine the extent to which opioid dosing chrcteristics or other clinicl fctors (the independent vribles) were ssocited with time till deth fter lst opioid dose chnge (the dependent vrible). Relevnt vribles were first compred cross the entire smple nd defined subsmples. Anlysis of vrince, Wilcoxon rnk sum tests, nd Spermn rnk correltions were used to exmine bivrite ssocitions between survivl fter finl dose chnge nd other vribles. Opioid vribles tht were nlyzed included opioid dose in morphine equivlent mg nd percentge increse in dose. These nlyses were performed on the totl smple nd ech of the subsmples. Multivrite lest squres regression nlyses then were pplied to determine ssocitions between the intervl between lst opioid dose chnge nd deth, nd vriety of independent vribles, including vribles tht hd significnt ssocitions in bivrite nlyses. We llowed the lgorithm in the mximum R 2 selection procedure for ordinry lest squres regressions to select independent vribles to enter nd leve ech model. Seprte models were constructed in n effort to explore the reltionship between opioid dosing chrcteristics nd survivl, nd to optimize their explntory vlue (R 2 ). All nlyses were performed with SAS sttisticl softwre Relese 8.2 (SAS Institute, Cry, North Crolin). Results Ptient Chrcteristics Chrcteristics of the entire smple (n ¼ 725) nd ech subsmple re presented in Tble 1. Mximum dily opioid dose vried significntly by length of sty in hospice nd by ptient ge. Men SD length of sty in hospice ws 30.25 37.8 dys. Ptients receiving >200 IVME hd length of sty significntly longer thn those receiving less thn this mount (46.6 45.4 dys vs. 28.1 36.1, P ¼ 0.0005), nd ptients receiving >600 IVME hd length of sty longer thn those receiving less (54.7 48.8 vs. 29.6 37.3, P ¼ 0.0042). Men SD ge for the entire group ws 76.6 13.4 yers. Those ptients receiving >200 IVME were younger thn those receiving #200 IVME (67.1 16.8 vs. 77.9 12.4 yers, P < 0.0001), nd those receiving >600 IVME were younger thn those receiving less thn this dose (69.2 15.8 vs. 76.8 13.3 yers, P < 0.0150). Mximum dily opioid dose did not vry by other demogrphic fctors (Tble 1). Forty-two percent of the smple popultion hd cncer. The proportion with this dignosis ws significntly higher in the subsmples defined by totl dily dose of >200 IVME nd >600 IVME, respectively. Men disese severity scores did not vry significntly mong subsmples nd rnged from 17.4 21.4 to 20.2 22.0 (Tble 1). Men SD PPS score for the entire smple ws 33.2 14.1 nd ws significntly higher in subsmples distinguished by reltively high opioid doses (Tble 1). Thirty-six percent of the totl smple ws unresponsive t the time of lst opioid dose chnge nd this proportion ws similr cross subsmples defined by mximum dily dose. Opioid Doses Tble 1 lso depicts the intervl in dys between the time of lst opioid dose chnge nd deth, the mximum opioid dose prior to deth, nd the percentge by which the dose hd been chnged to chieve this finl dose. For the entire smple, the men SD number of dys between finl dose chnge nd deth ws 12.46 23.11 (medin 5 dys; rnge 0e231 dys). Ptients

Vrible Totl Smple Ptients w/mx IVME of up to 200 Tble 1 Ptient Chrcteristics Ptients w/mx IVME of 200 nd bove P-vlue Ptients w/mx IVME of up to 600 Ptients w/mx IVME of 600 nd bove No. of ptients 725 (88.3) (11.7) (97.4) (2.6) Age 76.6 13.4 77.9 12.4 67.1 16.8 <0.0001 76.8 13.3 69.2 15.8 0.015 Gender Femle 421 (58) 380 (59) 41 (48) 0.0609 408 (58) 13 (68) 0.4811 Mle 304 (42) 260 (41) 44 (52) 298 (42) 6 (32) Ethnicity White 626 (86) 552 (86) 74 (87) 0.1301 608 (86) 18 (95) 0.5736 Blck 32 (4) 25 (40) 7 (8) 31 (4) 1 (5) Other 8 (1) 8 (1) 0 (0) 8 (1) 0 (0) Unknown 59 (8) 55 (9) 4 (5) 59 (8) 0 (0) Length of sty in 30.25 37.8 28.1 36.1 46.6 45.4 0.0005 29.6 37.3 54.7 48.8 0.0042 hospice (dys) Disese severity score, lst.9 21.9 20.2 22.0 17.4 21.4 0.259 20.0 22.0 17.7.1 0.6633 Dignosis Cncer 307 (42) 248 (39) 59 (69) <0.0001 294 (42) 13 (68) 0.1026 CHF/COPD 132 (18) 127 (20) 5 (6) 131 () 1 (5) Mentl lt 105 (14) 102 (16) 3 (4) 104 (15) 1 (5) Combintions of cncer, 181 (25) 163 (25) 18 (21) 177 (25) 4 (21) CHF/COP, nd/or mentl lt Mximum totl IVME dose 164.3 1221 65.7 45.6 906.3 3493 0.0292 89.5 91.7 2944 7159 0.0993 Chnge from mximum 40.6 951.7 1.9 9.6 332 2776 0.2764 2.6 16.5 1450 54 0.2955 dose to finl Dys from dose chnge 12.5 23.1 12.7 23.9 10.8 16.2 0.3538 12.5 23.4 10.1 11.2 0.3723 to deth Finl dose before deth 123.7 363.7 63.8 45.1 574.5 944.3 <0.0001 86.8 90.5 1494 1722 0.0022 Finl dose chnge percent 291.9 1240, 188.4 300.1, 815.3 2937, 0.0707 243.8 1038, 1682 3, 0.1554 n ¼ 448 (61.7) n ¼ 374 (58.5) n ¼ 74 (87.1) n ¼ 433 (61.3) n ¼ 15 (78.9) PPS score (lst) 33.2 14.15, 32.7 13.9, 37.5 15.45, 0.07 32.9 14.05, 44.2 14.4, 0.0063 n ¼ 511 (70.5) n ¼ 459 (71.7) n ¼ 52 (61.2) n ¼ 499 (70.7) n ¼ 12 (63.2) Level of consciousness No. of ptients 615 (84.3) 540 (84.4) 75 (88.2) 599 (84.8) 16 (84.2) Full 171 (28) 150 (28) 21 (28) 0.7353 167 (28) 4 (25) 0.3760 Drowsy 97 (16) 82 (15) 15 (20) 92 (15) 5 (31) Confused 127 (20) 113 (21) 14 () 124 (21) 3 () Unble to respond 220 (36) 5 (36) 25 (33) 216 (36) 4 (25) Dt re presented s men SD or number (percent). w/mx ¼ with mximum; IVME ¼ intrvenous morphine equivlent mg per dy; CHF ¼ congestive hert filure; COPD ¼ chronic obstructive pulmonry disese; Mentl lt ¼ mentl ltertion; PPS ¼ Pllitive Performnce Scle. Bold numbers indicte P # 0.05. Smple size decreses due to known dose chnge but unknown previous dose (n ¼ 123) or previous dose ws 0 (n ¼ 154). P-vlue Vol. 32 No. 6 December 2006 Opioid Use nd Survivl t the End of Life 535

536 Portenoy et l. Vol. 32 No. 6 December 2006 receiving mximum dily dose of >600 IVME hd n intervl slightly shorter thn those receiving less thn this dose (10.1 11.2 vs. 12.5 23.4, P ¼ 0.3723). Overll, ptients received 123.7 363.7 IVME (rnge 5.0e0 mg) immeditely prior to deth. In mny cses, the finl dose before deth ws not the mximum dose dministered during the time in hospice. Ptients in the entire smple received men SD mximum opioid dose of 164.3 1,221 IVME. Men SD mximum dose of opioid nlgesic vried from 65.7 45.6 mg per dy in the group receiving mximum dily dose <200 IVME to 2,944 7,159 mg per dy in the group receiving mximum dily dose of >600 IVME. For the totl smple, finl dose before deth verged 40.6 951.7 morphine equivlent mg less thn the mximum dose dministered t ny point during the hospice dmission; there were no significnt differences between mximum dose nd finl dose cross the vrious subsmples. Bivrite Anlyses To depict the reltionship between opioid dose nd survivl fter lst dose chnge, the smple ws divided into two sets of deciles, the first bsed on finl opioid dose nd the second bsed on percent finl dose chnge. These deciles were compred in terms of men survivl nd the proportion of ptients who died within two dys of the lst dose chnge (Tble 2). The only significnt reltionship ws tht ptients in the lowest decile of finl dose hd significntly longer men time till deth thn the other nine deciles. In other bivrite nlyses, the intervl between finl dose chnge nd deth ws significntly relted to rce/ethnicity nd severl disese-relted vribles (Tble 3). Cucsin rce ws ssocited with reltively shorter survivl fter lst dose chnge in the subsmple receiving <200 IVME (P ¼ 0.011) nd reltively longer survivl in the group receiving more thn this mount (P ¼ 0.041). Length of sty in hospice ws positively ssocited with survivl fter lst dose chnge in the totl smple nd the subsmples with mximum dily doses of #200 IVME nd #600 IVME, respectively (P # 0.001). Lst recorded PPS score nd full level of consciousness were ech significntly ssocited with longer survivl for the entire smple nd ech subsmple, except the group receiving >600 IVME. Being unble to respond ws correlted with shorter survivl in ll groups (Tble 3). Higher finl opioid dose ws significntly ssocited with shorter time till deth in the totl smple (P ¼ 0.010) nd in subsmples receiving #200 IVME (P ¼ 0.001) nd #600 IVME (P ¼ 0.008), respectively. Absolute dose chnge nd percent dose chnge were not ssocited with survivl. Tble 2 Reltionships Between Deciles of Lst Opioid Dose (Absolute nd % Chnge), nd Men Time Till Deth nd Proportion of Deths During Dys 1e2 After the Finl Dose Chnge Decile n Rnge of IVME Dose Finl Dose Men (SD) Time till Deth % Ptients Who Died Within 0e1 Dys b n Rnge of % Chnge % Finl Dose Chnge Men (SD) Time till Deth c % Ptients Who Died Within 0e1 Dys d 1 61 5e17 26.5 (43.1) 16.4 45 87 to 7 10.5 (16.6) 37.8 2 70 20e25 11.5 (21.6) 24.3 44 6 to 16 12.0 (16.7) 20.5 3 73 27e32 14.1 (21.9) 16.4 47 17e33 7.5 (7.9) 17.0 4 83 33e47 14.8 (29.7).3 44 34e56 8.7 (10.5) 11.4 5 66 48e58 13.9 (22.2) 22.7 44 57e87 10 (18.6) 22.7 6 81 59e78 7.4 (9.1) 29.6 46 88e120 9.1 (15.7) 32.6 7 71 80e97 9.7 (.7) 18.3 44 125e177 8.8 (11.3) 18.2 8 75 97e128 8.9 (13.8) 24.0 45 183e279 8.9 (13.9) 33.3 9 73 129e210 9.3 (15.7) 24.6 44 280e490 7.9 (13.3) 31.8 10 72 216e6,400 11.1 (17.1) 20.8 45 500e20,000 11.4 (.7) 13.3 IVME ¼ intrvenous morphine equivlent mg per dy. ANOVA P-vlue ¼ 0.0001. b Wilcoxon rnk sum test P-vlue ¼ 0.3545. c ANOVA P-vlue ¼ 0.9171. d Wilcoxon rnk sum test P-vlue ¼ 0.7984.

Vol. 32 No. 6 December 2006 Opioid Use nd Survivl t the End of Life 537 Vrible Tble 3 Bivrite Anlyses Between Ptient nd Opioid Tretment Chrcteristics, nd Time Between Finl Dose Chnge nd Deth Ptients w/o Mximum IVME Limittion (n ¼ 725) Ptients w/mx IVME of <200 (n ¼ ) Ptients w/mx IVME of > 200 (n ¼ ) Ptients w/mx IVME of <600 (n ¼ ) Ptients w/mx IVME of >600 (n ¼ ) n r nd P n r nd P n r nd P n r nd P n r nd P Age 725 Gender, femle 725 Ethnicity White 725 Blck 725 Other 725 0.10, 0.011 0.22, 0.041 b Length of sty 725 0.52, <0.001 0.56, <0.001 Disese severity 725 0.09, 0.018 Dignosis Cncer 725 CHF/COPD 725 Mentl lt 725 Combintions of cncer, 725 CHF/COPD, nd/or mentl lt Finl dose before deth 725 0.10, 0.010 0.13, 0.001 Finl dose chnge, bsolute 602 523 79 Finl dose chnge, % chnge 448 374 74 0.53, <0.001 0.09, 0.014 0.10, 0.008 5 17 433 15 Lst PPS score 511 0.24, <0.001 459 0.23, <0.001 52 0.28, 0.045 499 0.23, <0.001 12 Level of consciousness Full 615 0.32, <0.001 540 0.30, <0.001 75 0.49, <0.001 599 0.32, <0.001 16 Drowsy 615 540 75 599 16 Confused 615 0.09, 0.011 540 0.11, 0.004 75 599 0.1, 0.01 16 Unble to respond 615 0.23, <0.001 540 0.22, <0.001 75 0.33, 0.002 599 0.23, <0.001 16 0.59, 0.008 Bold numbers indicte P # 0.05. w/o ¼ without; w/mx ¼ with mximum; IVME ¼ intrvenous morphine equivlent mg per dy; CHF ¼ congestive hert filure; COPD ¼ chronic obstructive pulmonry disese; Mentl lt ¼ mentl ltertion; PPS ¼ Pllitive Performnce Scle. P-vlue greter thn 0.056. b No ptients in the group with rce designtion. b b Multivrite Anlyses Exmintion of smple chrcteristics nd the bivrite nlyses did not revel meningful differences in the groups receiving reltively higher doses. Indeed, there ws no significnt reltionship between high-dose opioid tretment (i.e., in the smple receiving >200, or >600, IVME) nd the intervl between finl dose chnge nd deth. Accordingly, multivrite nlyses were performed on the subsmple tht received mximum dily dose of <200 IVME, the group tht both most reflected the opioid dose rnge commonly encountered in prctice nd demonstrted reltionship between finl dose nd shorter survivl (r ¼ 0.13, P ¼ 0.001, Tble 3). In the first multiple regression model constructed, the vribles entered included ll those found to hve significnt bivrite reltionships with survivl, plus ge nd pin level t time of finl dose chnge. The ltter vribles were included becuse of clinicl observtions suggesting tht they re likely to influence the potentil for dverse effects during opioid therpy. Of ptients receiving <200 IVME, 360 (56.25%) could provide dt on ll these vribles. The model (Tble 4) demonstrted tht pin score of <5 on 0e10 numeric scle (ssocited with longer survivl) nd lck of consciousness (ssocited with shorter survivl) remined significnt when controlling for other vribles, but together explined only bout 10% of the vrince in time till deth fter finl dose chnge (R 2 ¼ 0.0999). Becuse PPS score ws not ssocited with survivl in the initil model, nd missing vlues for this vrible hd reduced the smple size for nlysis, it ws deleted to provide ccess to lrger smple (n ¼ 523 or 81.7% of

538 Portenoy et l. Vol. 32 No. 6 December 2006 Tble 4 Multiple Regression Anlyses of Time to Deth After Finl Dose Chnge in Ptients Receiving <200 IVME Model Informtion Vrible Coefficient t-vlue P-vlue n ¼ 360, R 2 ¼ 0.0992 Age 0.074 0.79 0.4272 Finl dose chnge 0.005 0.1 0.8795 Finl dose in IVME 0.054 1.72 0.0861 Lst PPS score 0.123 1.58 0.116 Dignosis Cncer 3.801 1.37 0.1730 CHF/COPD 3.545 1.18 0.2377 Mentl lt 1.748 0.52 0.6008 Level of consciousness ¼ unble to respond 4.969 2.14 0.0333 Pin <5 11.042 3.36 0.0009 n ¼ 360, R 2 ¼ 0.0836 Finl dose of IVME 0.059 6.3 0.0125 Level of consciousness ¼ unble to respond 5.218 5.51 0.05 Pin <5 12.147 14.61 0.0002 n ¼ 523 b, R 2 ¼ 0.0586 Finl dose in IVME 0.048 5.59 0.01 Dignosis Cncer 3.821 3.88 0.0493 Level of consciousness ¼ unble to respond 5.958 8.71 0.0033 Pin <5 8.200 7.97 0.0049 Bold numbers indicte P # 0.05. IVME ¼ intrvenous morphine equivlent mg per dy; CHF ¼ congestive hert filure; COPD ¼ chronic obstructive pulmonry disese; Mentl lt ¼ mentl ltertion; PPS ¼ Pllitive Performnce Scle. Only three vribles included. b PPS score removed, constructed to mximize R 2. ptients receiving <200 IVME). When the vribles used in the initil model (Tble 4) were forced into this regression model, finl opioid dose, cncer dignosis, nd unresponsiveness were significntly correlted with shorter survivl, nd pin score <5 ws significntly ssocited with longer time till deth. This model, however, ccounted for only 7% of the vrition in survivl time. To further explore the impct of opioid dose, nother nlysis entered only the two vribles tht were predictive in the initil model, plus finl opioid dose (Tble 4). All three vribles were significntly predictive of survivl, but together explined only bout 8% of the vrince (R 2 ¼ 0.0836). In this model, higher finl opioid dose nd being unresponsive t time of finl dose chnge were ssocited with shorter time till deth, nd pin score of <5 ws ssocited with longer survivl. Finlly, in n ttempt to improve the bility to explin the vrince in survivl fter finl opioid dose chnge, multiple regression nlysis evluted finl opioid dose, dignosis, lck of consciousness, nd pin score <5 in 523 ptients (Tble 4). All were significnt, but explined only 6% of the vrince in survivl (R 2 ¼ 0.059). Discussion Although pin due to dvnced illness usully cn be mnged with n opioid regimen, undertretment is common. 1 Clinicin concern bout the potentil for serious toxicity in mediclly fril ptients my contribute to undertretment. Although robust literture hs generlly endorsed the ppliction of the ethicl principle of double effect s the morl justifiction for ggressive therpy, 3,5,6 this my not lly concerns. This literture typiclly ssumes substntil risk of hstened deth due to the opioid. Opioid use t the end of life lso hs been discussed in the literture on physicin-hstened deth. 5,8,9,12 These writings lso ssume tht opioid toxicity is substntil during endof-life cre nd tht these drugs my be used intentionlly to shorten life. These ssumptions contrst with the common clinicl experience of specilists in pin mngement or pllitive medicine, who typiclly use opioid drugs in whtever doses re needed to chieve nlgesi nd rrely encounter scenrio consistent with primry opioid-relted deth. This snguine experience is supported by smll number of studies 9,13e15 tht together suggest reltively low risk of serious opioid toxicity.

Vol. 32 No. 6 December 2006 Opioid Use nd Survivl t the End of Life 539 To reconcile these opposing views, studies must specificlly ssess the extent to which opioid use ffects survivl mong those with dvnced illness. The NHOP nlyses demonstrte tht finl opioid dose, but not percent chnge in dose, ws one of severl fctors ssocited with survivl, but the ssocition is very wek, nd in multivrite nlyses, this nd other relevnt fctors explin only very smll percentge of vrition in survivl. The impliction tht opioid dose poses n extremely smll risk of hstened deth in this popultion ws supported further by the reltively long intervls between finl dose chnge nd deth, nd the lck of higher opioid risk in subsmples receiving high doses. These findings resonte with clinicl experience nd findings of other surveys. 13,14 Although experienced clinicins re wre tht serious toxicity, including risk of hstened deth, could be produced by rpid dose escltion, the usul pproch to opioid therpy incorportes incrementl dose chnges tht, on blnce, pper reltively sfe even if ptients re ner the end of life, require reltively high doses, or need frequent dose increses. These findings hve severl limittions. The results pply to hospice popultion, nd lthough the methodology incorported systemtic prospective dt collection, the study ws not primrily designed to ddress the question of opioid risk. Given missing dt nd dt collection by treting nurses, the possibility of observer bis, nd referrl nd selection bises, must be cknowledged. Finlly, vrition in opioid regimens my hve limited the bility to identify the signl of opioid toxicity during complex interply of mny other relevnt phenomen. These limittions notwithstnding, this nlysis suggests tht the timing of deth in popultions with fr dvnced illness involves complex interply of vribles, including importnt fctors tht were not ssessed in this study, nd tht opioid therpy should not be the focus of future reserch of this type. Eqully importnt, these dt should help dvnce the pproprite use of opioids for symptom control. Undertretment of pin is fr more pressing concern thn is the risk of hstened deth in those with dvnced disese, nd physicins should be encourged to use opioids effectively to relieve suffering t the end of life. References 1. Teno JM, Clrridge BR, Csey V, et l. Fmily perspectives on end-of-life cre t the lst plce of cre. JAMA 2004;291(1):88e93. 2. Prgeon KL, Hiley BJ. Brriers to effective cncer pin mngement: review of the literture. J Pin Symptom Mnge 99;18:358e368. 3. Krkuer EL, Penson RT, Truog RD, et l. Sedtion for intrctble distress of dying ptient: cute pllitive cre nd the principle of double effect. Oncologist 2000;5(1):53e62. 4. Kldjin LC, Jekel JF, Bernene JL, et l. Internists ttitudes towrds terminl sedtion in end of life cre. J Med Ethics 2004;30(5):499e503. 5. Cvnugh TA. The ethics of deth-hstening or deth-cusing pllitive nlgesic dministrtion to the terminlly ill. J Pin Symptom Mnge 96; 12:248e254. 6. Sulmsy DP, Pellegrino ED. The rule of double effect: clering up the double tlk. Arch Intern Med 99;159:545e550. 7. Quill TE, Dresser R, Brock DW. The rule of double effecte critique of its role in end-of-life decision mking. N Engl J Med 97;337:1768e1771. 8. Cntor NL, Thoms GC. Pin relief, ccelertion of deth, nd criminl lw. Kennedy Inst Ethics J 96;6:107e127. 9. vn der Heide A, vn der Ms PJ, vn der Wl G, Kollee LA, de Leeuw R. Using potentilly life-shortening drugs in neontes nd infnts. Crit Cre Med 2000;28:2595e2599. 10. Quill TE. The mbiguity of clinicl intentions. JAMA 93;329:1039e1040. 11. Brody H. Cusing, intending, nd ssessing deth. J Clin Ethics 93;4:112e117. 12. Wilson WC, Smedir NG, Fink C, McDowell JA, Luce JM. Ordering nd dministrtion of sedtives nd nlgesics during the withholding nd withdrwl of life support from criticlly ill ptients. JAMA 92;267:949e953. 13. Bercovitch M, Adunsky A. Ptterns of high-dose morphine use in home-cre hospice service: should we be frid of it? Cncer 2004;101:1473e1477. 14. Bercovitch M, Wller A, Adunsky A. High dose morphine use in the hospice setting. A dtbse survey of ptient chrcteristics nd effect on life expectncy. Cncer 99;86:871e877. 15. Thorns A, Sykes N. Opioid use in lst week of life nd implictions for end-of-life decision-mking. Lncet 2000;356:398e399.

540 Portenoy et l. Vol. 32 No. 6 December 2006 16. Sykes N, Thorns A. Sedtive use in the lst week of life nd the implictions for end-of-life decision mking. Arch Intern Med 2003;163:341e344. 17. Connor SR, Horn SD, Smout RJ, Gsswy J. The Ntionl Hospice Outcomes Project (NHOP): development nd implementtion of multi-site hospice outcomes study. J Pin Symptom Mnge 2005;29:286e296. 18. Americn Pin Society. Principles of nlgesic use in the tretment of cute pin nd cncer pin, 5th ed. Glenview, IL: Americn Pin Society, 2003.. Horn SD, Shrkey PD, Buckle JM, et l. The reltionship between severity of illness nd hospitl length of sty nd mortlity. Med Cre 91;29: 305e317. 20. Averill RF, McGuire TE, Mnning BE, et l. A study of the reltionship between severity of illness nd hospitl cost in New Jersey hospitls. Helth Serv Res 92;27(5):587e617. 21. Morit T, Tsunod J, Inoue S, et l. Vlidity of the Pllitive Performnce Scle from survivl perspective. J Pin Symptom Mnge 99;18:2e3.