Originl Article Ptient Survivl After Surgicl Tretment of Rectl Cncer Impct of Surgeon nd Hospitl Chrcteristics Dvid A. Etzioni, MD, MSHS 1,2 ; Toni M. Young-Fdok, MD, MS 1 ; Robert R. Cim, MD, MA 2,3 ; Nbil Wsif, MD 1 ; Robert D. Mdoff, MD 4 ; Jmes M. Nessens, SCD 2 ; nd Elizbeth B. Hbermnn, PhD 2,3 BACKGROUND: Surgeon nd hospitl fctors re ssocited with the survivl of ptients treted for rectl cncer. The reltive contribution of ech of these fctors towrd determining outcomes is poorly understood. METHODS: We used dt from the Surveillnce, Epidemiology, nd End Results Medicre dtbse to nlyze the outcomes of ptients ged 65 yers nd older undergoing opertive tretment for nonmetsttic rectl cncer, dignosed in the United Sttes between 1998 nd 2007. These dt were linked to registry to identify whether the treting surgeon ws bord-certified colorectl surgeon versus noncolorectl surgeon. Hospitl volume nd hospitl certifiction s Ntionl Cncer Institute designted Comprehensive Cncer Centers were lso nlyzed. The primry outcome of interest ws long-term survivl. RESULTS: Our dt source yielded 6432 ptients. Initil nlysis demonstrted improved long-term survivl in ptients treted by higher-volume colorectl surgeons, higher-volume hospitls, teching hospitls, nd Ntionl Cncer Institute (NCI) designted Comprehensive Cncer Centers. Bsed on n itertive pproch to modeling the interctions between these vrious fctors, we found robust effect of surgeon subspecilty sttus, hospitl volume, nd NCI designtion. ws not distinctly ssocited with long-term survivl. CONCLUSIONS: Ptients treted for rectl cncer by bord-certified colorectl surgeons in centers tht re higher volume nd/or NCI-designted Comprehensive Cncer Centers experience better overll survivl. These differences persist fter djustment for brod rnge of ptient nd contextul risk fctors, including surgeon volume. Ptients nd pyers cn use these results to identify surgeons nd hospitls where outcomes re most fvorble. Cncer 2014;120:2472-81. VC 2014 Americn Cncer Society. KEYWORDS: rectl neoplsms, surgery specilty, colon nd rectl, cncer cre fcilities. INTRODUCTION Rectl cncer is dignosed in more thn 40,000 individuls per yer in the United Sttes, nd ccounts for pproximtely 14,000 deths. 1 Although the cre of rectl cncer involves multiple modlities of tretment, surgicl resection is minsty of tretment. A significnt body of nonrndomized evidence highlights the importnce of surgicl pproch in the outcomes experienced by ptients dignosed with rectl cncer. The likelihood of locoregionl recurrence is notbly lower mong surgeons who hve specific trining in the techniques of totl mesorectl excision 2 Hospitl effects re lso importnt overll survivl rtes nd pproprite use of djuvnt therpy re notbly higher in higher-volume hospitls, nd those with more cdemic orienttion. 3-8 With this study, we nlyze the experience of Medicre beneficiries treted for nonmetsttic rectl cncer in the United Sttes. In this nlysis, we focus specificlly on chrcterizing the overlpping effects of surgeon fctors nd hospitl chrcteristics. MATERIALS AND METHODS Cohort Selection The Surveillnce, Epidemiology, nd End Results (SEER) Medicre linked dtbse served s the primry dt source regrding ptients surgiclly treted for rectl cncer between 1998 nd 2007. Our cohort includes ptients ged 65 yers or older t the time of dignosis who underwent proctectomy for stge I-III rectl denocrcinom. Continuous prend postopertive enrollment (for t lest 1 yer) in Medicre Prt A/B ws lso required. Corresponding uthor: Dvid A. Etzioni, MD, MSHS, Myo Clinic/Deprtment of Surgery, Myo Clinic Specilty Building, 3rd Floor; E Myo Blvd., Phoenix, AZ 85054; Fx: (480) 342-2866; etzioni.dvid@gmil.com 1 Deprtment of Surgery, Myo Clinic College of Medicine, Phoenix, Arizon; 2 Myo Clinic Center for the Science of Helthcre Delivery, Rochester, Minnesot; 3 Deprtment of Surgery, Myo Clinic College of Medicine, Rochester, Minnesot; 4 Deprtment of Surgery, University of Minnesot, Minnepolis, Minnesot DOI: 10.1002/cncr.28746, Received: Februry 3, 2014; Revised: Mrch 13, 2014; Accepted: Mrch 19, 2014, Published online My 6, 2014 in Wiley Online Librry (wileyonlinelibrry.com) 2472 Cncer August 15, 2014
Survivl After Surgery for Rectl Cncer/Etzioni et l TABLE 1. Ptient/Surgeon/Tretment/Hospitl Chrcteristics n 5 6432 Figure 1. Inclusion/Exclusion Criteri To ccount for ptients who my hve presented cutely with rectl cncer (eg, obstruction, bleeding, perfortion), we excluded ny ptient who ws hospitlized urgently between 30 dys before dignosis nd the dte of first tretment (Fig. 1). In the United Sttes, tretment for cute colorectl conditions is predominntly provided by noncolorectl surgeons, nd we were concerned tht these ptients might represent higher-risk cohort nd possibly bis the findings. 9,10 We lso excluded ptients for whom we were unble to identify the treting surgeon. Ptients with inflmmtory bowel disese or ny Interntionl Clssifiction of Disese (ICD) dignosis code of ileostomy/colostomy before the dte of dignosis were excluded s well. To homogenize the type of rdicl resection tht ws performed, ptients who hd n exentertive procedure were excluded. An exentertive procedure ws defined on the bsis of specific Current Procedurl Terminology (CPT) code, or in situtions in which nother pelvic orgn ws resected on the sme dy s ptient s first rdicl resection, with n ccompnying dignosis of rectl cncer. Opertive nd Adjuvnt Tretment The opertive tretment provided to ech ptient ws identified through nlysis of CPT codes within physicin clims. Ech opertion ws ctegorized s sphincterspring proctectomy (SSP) or n bdominoperinel resection (APR). The use nd timing of djuvnt therpy were identified through n nlysis of physicin nd hospitl clims bsed on CPT nd ICD coding (Tble 1). Ptient chrcteristics Men ge t dignosis (y) 75.6 Femle sex 44.9% Nonwhite rce (%) 12.8% Modified Chrlson Score 0 58.0% 1 14.0% 2 16.2% 31 11.9% Urbn/rurl geogrphy Big metropolitn 52.6% Metropolitn 29.9% Urbn, less urbn, or rurl 17.5% Medin income Qurtile 1 (<$34,311/yer) 24.7% Qurtile 2 ($34,311-$44,795) 24.7% Qurtile 3 ($44,796-$60,130) 24.7% Qurtile 4 (>$60,130) 24.7% Missing 1.0% Tumor chrcteristics AJCC stge 36.8% I 30.2% II 32.9% III Surgeon chrcteristics Low (1-2 cses/yer) 39.5% Mid (3-4 cses/yer) 25.2% High (51 cses/yer) 35.2% Surgeon 5 colorectl surgeon 36.8% Hospitl chrcteristics Hospitl volume Low (1-12 cses/yer) 33.8% Mid (13-24 cses/yer) 33.3% High (251 cses/yer) 32.8% NCI-designted Comprehensive Cncer Center 6.5% Yer of surgery 1998-1999 11.2% 2000-2001 24.3% 2002-2003 23.4% 2004-2005 21.2% 2006-2007 19.6% Tretment Opertion type 5 bdominoperinel resection 33.8% Rdiotherpy Used 58.1% b Preopertive rdiotherpy (only) 33.9% b Postopertive rdiotherpy (only) 25.2% b Postopertive chemotherpy used 39.1% b Guideline-concordnt djuvnt tretment c 45.4% Abbrevitions: NCI, Ntionl Cncer Institute; CCC, Comprehensive Cncer Center. Determined bsed on medin income within zip code in 2000 Census. b Smple size restricted to 4,062 ptients with stge II/III rectl cncer only. c Composite vrible denotes use of no djuvnt tretment for stge I rectl cncer, chemordiotherpy (either pre-opertively or post-opertively) plus post-opertive chemotherpy for stge II/III rectl cncer. Documenttion of rdition therpy within the SEER dtbse ws lso included. Receipt of chemotherpy nd/ or rdition therpy between the dte of dignosis nd the dte of rdicl resection ws considered preopertive therpy. Postopertive therpy ws considered djuvnt Cncer August 15, 2014 2473
Originl Article therpy if the first documented tretment ws delivered within 120 dys fter rdicl surgery. Surgeon Fctors To identify the subspecilty sttus of the treting surgeon, Medicre clims were linked to the Americn Medicl Assocition Physicin Msterfile using n unencrypted Unique Physicin Identifier Number. After this linkge, the nme nd dte of birth of ech surgeon were used to link to registry of bord-certified colorectl surgeons provided by the Americn Bord of Colon nd Rectl Surgery. ws clculted s the number of rdicl proctectomies performed by surgeon within the SEER-Medicre progrm during the yer in which the ptient hd his or her opertion by the surgeon. Hospitl Fctors We tllied the number of proctectomies performed by the hospitl in the yer of the index opertion in the SEER- Medicre progrm. The hospitl where the ptient hd her or his first rdicl opertion ws identified, nd spects of this hospitl were glened through linkge to SEER- Medicre registry. On the bsis of existing literture, our primry hospitl fctor of interest ws whether the index hospitl ws Ntionl Cncer Institute designted Comprehensive Cncer Center (NCI-CCC). 11,12 Risk Adjustment Multivrite risk djustment ws plnned on the bsis of ptient demogrphics, comorbidities, cncer stge, nd opertion type (SSP vs APR). Demogrphics included wege, sex, socioeconomic sttus (bsed on zip code linked medin income, prsed into qurtiles), nd rce/ethnicity (white/non-hispnic vs other). Comorbidity ws estimted using modified Chrlson score, bsed on dignosis codes incurred in both inptient nd outptient contexts within the 6 months (180 dys) before dignosis. Cncer stge ws modeled s stge I, II, or III bsed on discretionry cncer stging vribles present within the SEER dt set. This vrible is bsed on preopertive nd pthologic stging dt (the higher stge is considered the ptient s ctul stge). We lso modeled whether the ptient lived in n urbn versus rurl loction (qulified s big metropolitn, metroprolitn, nonmetropolitn bsed on preexisting definitions within the SEER- Medicre dt source 13 ). Sttisticl Anlysis The primry outcome of interest ws mortlity, defined s the period between dignosis nd deth (s reported by the SEER registry). A Cox proportionl hzrds model ws used to nlyze multivrite predictors of long-term mortlity over time for ptients undergoing rdicl resection; December 31, 2009, ws used s the lst dte of follow-up (nd therefore censoring), for the ptients in our cohort. Bivrite reltions were tested using chi-squre test. Institutionl Review This study ws reviewed nd pproved by the institutionl review bord of the Myo Clinic College of Medicine, s well s by the SEER-Medicre Progrm. RESULTS Cohort Our pproch to identifying ptients resulted in cohort of 6432 ptients undergoing surgicl tretment for rectl denocrcinom, fter pplying our exclusion criteri (Fig. 1). Ptients hd men ge of 75.6 yers nd were predominntly mle (55.1%). In this cohort, minority of ptients were nonwhite (12.8%), nd the vst mjority resided in mjor metropolitn res. (Tble 1) During the period of our study, there were totl of 3021 deths (47.0%), with minimum follow-up of 2 yers (medin follow-up, 4.3 yers). Chrcteristics of Surgeons nd Hospitls Our nlysis included opertions performed by 2125 unique surgeons. Colorectl surgeons performed 2369 (36.8 %) of ll rdicl resections (Tble 1). The opertions performed by colorectl surgeons were more likely to hve been performed by high-volume surgeons (68.7% vs 31.3%, P <.001). Also, opertions by colorectl surgeons were more likely to hve been performed t highvolume institution (57.7% vs 42.3%, P <.001) nd t NCI-CCC (8.6% vs 5.2%, P <.001). Our nlysis included opertions performed t 839 hospitls. A smll minority of ptients (415 ptients, 6.5%) were treted t NCI-CCCs (29 centers). These hospitls were not higher-volume centers when compred with non-nci-cccs (P 5.18). Opertions t NCI- CCCs were performed more often by high-volume surgeons (41.0% vs 34.9%, P <.001) nd by colorectl surgeons (49.2% vs 36.0%, P <.001) compred with non- NCI-CCCs. Tretment Approximtely one-third of ptients (33.8%) in our cohort hd n APR, with the reminder undergoing SSP. Ptients operted on by higher-volume surgeons nd by colorectl surgeons were more likely to hve SSP (Tble 2). Hospitl volume nd NCI designtion were not 2474 Cncer August 15, 2014
Survivl After Surgery for Rectl Cncer/Etzioni et l ssocited with the proportion of ptients undergoing APR (Tble 2). Of ptients with stge II/III rectl cncer, mjority of ptients were treted with rdiotherpy (58.1%), but minority received postopertive chemotherpy (39.1%); see Tble 1. Preopertive rdiotherpy ws used more commonly thn postopertive rdiotherpy. To estimte the use of pproprite djuvnt therpy, we relied on Ntionl Comprehensive Cncer Network guidelines. We defined composite vrible tht ws specific to the ptient s cncer stge. Guideline-concordnt djuvnt tretment ws considered to hve been delivered TABLE 2. Opertion Type Versus Surgeon/Hospitl Fctors SSP (%) Surgeon chrcteristics Low (1-2 cses/yer) 63.0% <.001 Mid (3-4 cses/yer) 67.7% High (51 cses/yer) 68.8% Surgeon type Noncolorectl surgeon 63.8% <.001 Colorectl surgeon 70.4% Hospitl chrcteristics Hospitl volume Low (1-12 cses/yer) 65.9%.06 Mid (13-24 cses/yer) 64.7% High (251 cses/yer) 68.1% NCI-designted CCC xttus Non-NCI-CCC 66.2%.93 NCI-CCC 66.0% Abbrevitions: NCI, Ntionl Cncer Institute; CCC, Comprehensive Cncer Center; SSP, sphincter-spring proctectomy. P in the following situtions: 1) ptient with stge I rectl cncer received no djuvnt chemotherpy or rdition, or 2) ptient with stge II/III rectl cncer received chemordiotherpy (either preopertively or postopertively) plus postopertive chemotherpy. On the bsis of this definition, our criteri s mnifested in this composite mesure were met in 45.4% of ptients (Tble 3). Guideline-concordnt djuvnt therpy use vried by surgeon subspecilty, hospitl designtion, nd stge of disese. Colorectl surgeons nd NCI- CCCs were more likely to provide nonconcordnt cre to ptients with stge I disese. Ptients with stge III disese were more likely to receive concordnt cre t NCI- CCCs. Hospitl nd surgeon volume did not distinctly correlte with ptterns of djuvnt therpy. Of note, the dt displyed in Tble 3 re not n nlysis of the reltion between djuvnt therpy use nd survivl. Survivl Anlysis Initil djusted nlyses We conducted n initil explortory nlysis to understnd the reltionship between surgeon/hospitl/tretment fctors nd survivl. In this nlysis (Tble 4), we performed Cox regression nlysis tht ws djusted for ptient/tumor chrcteristics, s well s the yer of surgery. Ech surgeon, tretment, or hospitl predictor vrible shown in Tble 4 ws considered individully to provide n estimte of the contribution of tht single fctor. The impct of ech of these fctors ws distinct nd sttisticlly significnt. High-volume surgeons nd TABLE 3. Use of Guideline-Concordnt Adjuvnt Therpy Stge I n 5 2370 Stge II n 5 1945 Stge III n 5 2117 All Stges n 5 6432 % Use P % Use P % Use P % Use P Surgeon chrcteristics Low (1-2 cses/yer) 70.8%.33 23.9%.77 35.2%.49 45.0%.22 Mid (3-4 cses/yer) 70.8% 25.6% 38.4% 47.0% High (51 cses/yer) 67.8% 25.0% 36.7% 44.2% Surgeon type Noncolorectl surgeon 72.9% <.001 24.2%.51 36.7%.85 46.6% <.01 Colorectl surgeon 64.0% 25.5% 36.3% 42.9% Hospitl chrcteristics Hospitl volume Low (1-12 cses/yer) 69.8%.70 22.9%.21 34.9%.25 44.0%.33 Mid (13-24 cses/yer) 70.8% 24.4% 35.7% 45.4% High (251 cses/yer) 68.8% 27.0% 38.9% 46.2% NCI-designted CCC sttus Non-NCI-CCC 70.3%.03 24.3%.13 35.9%.02 45.0%.25 NCI-CCC 61.9% 31.0% 45.0% 48.0% Abbrevitions: NCI, Ntionl Cncer Institute; CCC, Comprehensive Cncer Center. Composite vrible denotes use of no djuvnt tretment for stge I rectl cncer nd chemordiotherpy (either preopertively or postopertively) plus postopertive chemotherpy for stge II/III rectl cncer. Cncer August 15, 2014 2475
Originl Article TABLE 4. Cox Regression Anlyses Adjusted Hzrd Rtio of Mortlity (95% Confidence Intervl) Tumor chrcteristics Stge I Stge II 1.54 (1.40-1.69) Stge III 2.25 (2.06-2.45) Surgeon chrcteristics Low (1-2 cses/yer) Mid (3-4 cses/yet 0.96 (0.88-1.05) High (51 cses/yer) 0.86 (0.79-0.93) Surgeon type 5 colorectl surgeon 0.84 (0.78-0.91) Tretment Opertion type 5 bdominoperinel resection 1.35 (1.25-1.45) Rdiotherpy Preopertive or postopertive rdiotherpy 0.71 (0.64 0.78) Preopertive rdiotherpy (only) 0.88 (0.80 0.97) Postopertive rdiotherpy (only) 0.81 (0.74 0.90) Postopertive chemotherpy 0.63 (0.57 0.69) Guideline-concordnt djuvnt tretmen b 0.79 (0.73 0.85) Hospitl chrcteristics Hospitl volume Low (1-12 cses/yer) Mid (13-24 cses/yer) 0.85 (0.78-0.93) High (251 cses/yer) 0.86 (0.79-0.94) Hospitl 5 Ntionl Cncer Institute 0.76 (0.64-0.90) Comprehensive Cncer Center All models djusted for ptient chrcteristics, stge, nd yer of surgery (see Tble 1), except for models exmining tumor chrcteristics, which were not djusted for tumor stge. b Smple size restricted to 4062 ptients with stge II/III rectl cncer only. colorectl surgeons hd better long-term survivl. Hospitls with higher nnul cse volumes nd NCI-CCCs showed similr correltions with higher survivl rtes. Tretment fctors were lso correlted with longterm survivl. Ptients undergoing n APR hd higher hzrd of mortlity thn those undergoing SSP. Also, ptients receiving chemotherpy or rdiotherpy (preopertive or postopertive) showed better survivl. Our composite mesure denoting receipt of guideline-concordnt djuvnt tretment lso showed positive correltion with long-term survivl. Multivrite nlysis We nlyzed long-term survivl with the intent of understnding the impct of surgeon fctors nd hospitl fctors, controlling for fctors pertinent to the ptient, tumor chrcteristics, nd tretment. Given the correltions between surgeons, hospitls, nd opertive tretment (APR vs SSP) nd use of djuvnt therpy, it proved conceptully difficult to incorporte surgeon, tretment, nd hospitl fctors into single multivrite regression model. Therefore, we chose n nlytic pproch in which we modeled surgeon nd hospitl fctors both with nd without the presence of vribles pertining to tretment. In Tble 5, we present 7 types of models, ech of which ccounts for different set of hospitl nd tretment fctors. Within ech model type, we generted 3 model subtypes: 1) surgeon subspecilty, 2) surgeon volume, or 3) both subspecilty nd volume. In our initil bse model, ptients treted by colorectl subspecilists nd/or by high-volume surgeons hd lower hzrd rtio for long-term mortlity (0.84; 95% CI, 0.78-0.91, for colorectl surgeons; nd 0.86, 95% CI,, 0.79-0.93, for high-volume surgeons). When surgeon subspecilty nd volume were considered jointly in ech of our models, the importnce of surgeon volume becme sttisticlly nonsignificnt. Conversely, surgeon subspecilty retined importnce s positive predictor of long-term survivl. Adjustment for tretment nd hospitl type/volume hd little effect on the importnce of colorectl subspecilty. Tble 6 shows similr nlysis undertken to exmine hospitl fctors. With this nlysis 7 regression models were pplied, ech ccounting for different set of surgeon nd tretment fctors. Within ech model type, we generted 3 model subtypes: 1) NCI-CCC designtion, 2) hospitl volume, or 3) both. In our initil bse model, both hospitl volume nd NCI-CCC designtion were correlted with long-term survivl. The importnce of these 2 fctors ws ltered only slightly by the inclusion of different combintions of surgeon volume, surgeon subspecilty sttus, nd tretment. Sensitivity nlysis In post hoc sensitivity nlysis, we explored lterntive strtegies to represent the type(s) of djuvnt tretment delivered to ptients in our cohort. These strtegies were 1) including only preopertive therpy, 2) including only rdiotherpy, 3) modeling rdiotherpy nd chemotherpy individully, nd 4) developing stge/tretment composite vrible. None of these lterntive pproches hd significnt impct on our findings. DISCUSSION The cre of ptients with rectl cncer necessrily involves the input of multiple disciplines, including surgery, medicl oncology, nd rdition oncology. Our findings highlight the importnce of hospitl nd surgeon fctors in determining outcomes. In this study of more thn 6400 Medicre beneficiries treted within the United Sttes, ptients treted in high-volume hospitls, NCI-CCCs, nd by bord-certified colorectl surgeons experienced better long-term survivl. 2476 Cncer August 15, 2014
Survivl After Surgery for Rectl Cncer/Etzioni et l TABLE 5. Anlysis of Surgeon Fctors Impct on Long-Term Survivl Model 1: Includes No Additionl Adjustment Model 1A Model 1B Model 1C Surgeon 5 colorectl surgeon 0.84 (0.78-0.91) 0.87 (0.80-0.96) Low (1-2 cses/yer) Mid (3-4 cses/yer) 0.96 (0.88-1.05) 0.99 (0.90-1.08) High (51 cses/yer) 0.86 (0.79-0.93) 0.92 (0.84-1.02) Model 2: Includes Adjustment for Hospitl Volume Model 2A Model 2B Model 2C Surgeon 5 colorectl surgeon 0.86 (0.79-0.93) Low (1-2 cses/yer) Mid (3-4 cses/yer) High (51 cses/yer) 0.88 (0.80-0.96) 0.98 (0.90-1.08) 1.00 (0.92-1.10) 0.89 (0.81-0.97) 0.95 (0.86-1.05) Model 3: Includes Adjustment for Hospitl NCI Sttus Model 3A Model 3B Model 3C Surgeon 5 colorectl surgeon 0.85 (0.78-0.91) Low (1-2 cses/yer) Mid (3-4 cses/yer) High (511 cses/yer) 0.88 (0.80-0.96) 0.97 (0.89-1.06) 0.99 (0.90-1.08) 0.86 (0.79-0.93) 0.92 (0.84-1.02) Model 4: Includes Adjustment for Tretment Model 4A Model 4B Model 4C Surgeon 5 colorectl surgeon 0.85 (0.78-0.92) Low (1-2 cses/yer) Mid (3-4 cses/yer) High (51 cses/yer) 0.88 (0.80-0.96) 0.96 (0.88-1.05) 0.98 (0.90-1.08) 0.88 (0.80-0.95) 0.93 (0.84-1.02) Model 5: Includes Adjustment for Tretment, Hospitl Volume Model 5A Model 5B Model 5C Surgeon 5 colorectl surgeon 0.86 (0.80-0.94) Low (1-2 cses/yer) Mid (3-4 cses/yer) High (51 cses/yer) 0.88 (0.80-0.96) 0.99 (0.90-1.08) 1.01 (0.92-1.10) 0.90 (0.82-0.99) 0.96 (0.87-1.06)\ Model 6: Includes Adjustment for Tretment, Hospitl NCI Sttus Model 6A Model 6B Model 6C Surgeon 5 colorectl surgeon 0.85 (0.79-0.92) 0.89 (0.81-0.97) Cncer August 15, 2014 2477
Originl Article TABLE 5. Continued Model 6: Includes Adjustment for Tretment, Hospitl NCI Sttus Model 6A Model 6B Model 6C Low (1-2 cses/yer) Mid (3-4 cses/yer) High (51 cses/yer) 0.97 (0.89-1.06) 0.99 (0.90-1.08) 0.87 (0.80-0.95) 0.93 (0.84-1.02) Model 7: Includes Adjustment for Tretment, Hospitl Volume, Hospitl NCI-CCC Sttus Model 7A Model 7B Model 7C Surgeon 5 colorectl surgeon 0.87 (0.80-0.95) Surgeon v 5 olume Low (1-2 cses/yer) Mid (3-4 cses/yer) High (51 cses/yer) 0.89 (0.81-0.97) 0.99 (0.91-1.09) 1.00 (0.92-1.11) 0.91 (0.83-0.99) 0.96 (0.87-1.06) Abbrevitions: NCI, Ntionl Cncer Institute; CCC, Comprehensive Cncer Center. Vrible not included. All models djusted for ptient/tumor chrcteristics nd yer of surgery (see Tble 1). All results reported s djusted hzrd rtios with 95% confidence intervls. We re not the first to exmine the role of hospitl nd surgeon fctors nd their impct on the outcomes of ptients with rectl cncer. A Cochrne review published in 2012 nlyzed the existing literture regrding these reltionships. 5 Similr to our study, this review found tht hospitl volume, surgeon volume, nd tretment by surgicl subspecilists to be importnt in determining 5-yer survivl. Our study hs severl strengths tht set it prt from the existing literture. The most importnt of these strengths ws tht we were ble to simultneously ccount for issues relted to tretment fctors, surgeon subspecilty, surgeon volume, hospitl specilty (NCI-CCC), nd hospitl volume. Our itertive pproch to chrcterizing the complex interction mong these fctors yielded importnt findings. Surgeon subspecilty hs greter importnce thn surgeon volume in determining long-term ptient outcomes. Hospitl fctors (volume nd NCI- CCC designtion) lso ply n importnt role. Of ll the fctors pertinent to surgeon nd hospitl, NCI-CCC sttus hd the lrgest effect size in determining overll survivl. Our findings re t lest prtilly contextul. We were ble to dichotomize surgeons s being subspecilty surgeons versus other surgeons bsed on distinct fctor: bord certifiction. This fctor is firly specific to the United Sttes nd therefore rises nturl question: wht specific ttribute of bord-certified colorectl surgeons is responsible for better outcomes? Possible nswers include technique, trining, experience, nd focus of prctice. Recent work in the United Kingdom hs lso demonstrted improved survivl in ptients with colorectl cncer who received cre by specilist surgeons, with specilist designtion pplied on the bsis of pttern of prctice. 14 In our study we nlyzed severl spects of the cre provided by subspecilists nd cncer-oriented hospitls tht my be responsible for differences in outcomes nd identified severl res in which tretment ptterns differed. Colorectl surgeons re less likely to perform APRs, which hve lower overll survivl. Ptients with stge III rectl cncer treted t NCI-CCCs were more likely to receive pproprite djuvnt therpy. Unexpectedly, the overuse of djuvnt therpy in ptients with stge I rectl cncer ws higher when tretment ws provided by colorectl surgeons nd/or NCI-CCCs. Our nlytic pproch, however, demonstrted tht djuvnt tretment did not significntly modify the reltion between surgeon nd hospitl subspecilty on outcomes. Ptients treted by surgeons nd hospitls with greter focus on cncer cre hve improved long-term survivl. Although we could not completely chrcterize these effects, these differences in outcomes my hve to do with vritions in technique (eg, totl mesorectl excision), 15 decision mking, follow-up, ptient selection, improved multidisciplinry cre coordintion, or other unmesured confounder. Our findings re limited primrily by the nture of the dt sources. The dt bstrcted by SEER registrrs re generlly considered to be high qulity nd subjected 2478 Cncer August 15, 2014
Survivl After Surgery for Rectl Cncer/Etzioni et l TABLE 6. Anlysis of Hospitl Fctors Impct on Long-Term Survivl Model 1: Includes No Additionl Adjustment Model 1A Model 1B Model 1C NCI-designted CCC 0.76 (0.64-0.90) Hospitl volume Low (1-12 cses/yer) Mid (13-24 cses/yer) High (251 cses/yer) 0.76 (0.64-0.90) 0.85 (0.78-0.93) 0.85 (0.78-0.93) 0.86 (0.79-0.94) 0.86 (0.78-0.94) Model 2: Includes Adjustment for Surgeon Subspecilty Model 2A Model 2B Model 2C NCI-designted CCC 0.78 (0.66-0.92) Hospitl volume Low (1-12 cses/yer) Mid (13-24 cses/yer) High (251 cses/yer) 0.78 (0.66-0.92) 0.87 (0.79-0.94) 0.87 (0.79-0.95) 0.91 (0.83-1.00) 0.90 (0.82-0.99) Model 3: Includes Adjustment for Surgeon Volume Model 3A Model 3B Model 3C NCI-designted CCC 0.77 (0.65-0.91) Hospitl volume Low (1-12 cses/yer) Mid (13-24 cses/yer) High (251 cses/yer) 0.77 (0.65-0.91) 0.87 (0.80-0.95) 0.87 (0.80-0.95) 0.90 (0.82-0.99) 0.90 (0.82-0.98) Model 4: Includes Adjustment for Tretment Model 4A Model 4B Model 4C NCI-designted CCC 0.76 (0.64-0.90) Hospitl volume Low (1-12 cses/yer) Mid (13-24 cses/yer) High (251 cses/yer) 0.76 (0.64-0.90) 0.84 (0.77-0.91) 0.84 (0.77-0.92) 0.86 (0.79-0.94) 0.86 (0.79-0.94) Model 5: Includes Adjustment for Surgeon Subspecilty, Tretment Model 5A Model 5B Model 5C NCI-designted CCC 0.77 (0.65-0.91) Hospitl volume Low (1-12 cses/yer) Mid (13-24 cses/yer) High (251 cses/yer) 0.77 (0.65-0.91) 0.85 (0.78-0.93) 0.85 (0.78-0.93) 0.91 (0.83-0.99) 0.90 (0.82-0.99) Model 6: Includes Adjustment for Surgeon Volume, Tretment Model 6A Model 6B Model 6C NCI-designted CCC 0.76 (0.64-0.90) 0.76 (0.64-0.90) Cncer August 15, 2014 2479
Originl Article TABLE 6. Continued Model 6: Includes Adjustment for Surgeon Volume, Tretment Model 6A Model 6B Model 6C Hospitl volume Low (1-12 cses/yer) Mid (13-24 cses/yer) High (251 cses/yer) 0.86 (0.78-0.94) 0.86 (0.78-0.94) 0.90 (0.82-0.99) 0.90 (0.81-0.99) Model 7: Includes Adjustment for Surgeon Subspecilty, Surgeon Volume, Tretment Model 7A Model 7B Model 7C NCI-designted CCC 0.77 (0.65-0.91) Hospitl volume Low (1-12 cses/yer) Mid (13-24 cses/yer) High (251 cses/yer) 0.77 (0.65-0.91) 0.86 (0.78-0.94) 0.86 (0.78-0.94) 0.92 (0.83-1.01) 0.91 (0.83-1.00) Abbrevitions: NCI, Ntionl Cncer Institute; CCC, Comprehensive Cncer Center. Vrible not included. All models djusted for ptient/tumor chrcteristics nd yer of surgery (see Tble 1). All results reported s djusted hzrd rtios, with 95% confidence intervls. to extensive qulity control procedures. Some importnt spects regrding cncer evlution nd tretment re notbly missing, however. Our nlysis of rtes of sphincter-spring (vs APR; Tble 2) is superficil, s cncer registries do not cpture informtion regrding the loction of the rectl cncer reltive to ptient s sphincter complex. It is worth noting tht the SEER registry is geogrphiclly bsed smple, nd there re vlid concerns tht the smple is not perfectly representtive of the ntion s whole. 16 Medicre dt re lso distinctly limited in terms of their bility to chrcterize the types of processes tht re provided by hospitls nd surgeons. This limittion is borne from Medicre dt being essentilly dministrtive in nture nd therefore not cpturing rich clinicl informtion. These dt, however, re the bsis for reimbursement to physicins inccurte or misleding informtion my result in penlties, nd therefore the dt re considered to hve good ccurcy The gretest limittion of this study is the possibility of selection bis in our nlysis. We controlled for brod rnge of fctors relted to ptient demogrphics nd comorbidities, stge of tumor, nd environment. Although these djustments were importnt, they were lso potentilly incomplete. There is good reson to believe, however, tht selection bis within our study might bis our findings towrd the null hypothesis. Within ny community in which specilized versus subspecilized cre is vilble, referring physicins nd ptients hve choice. It is likely tht when this choice is vilble tht the more complex nd techniclly chllenging cses will be directed to subspecilty cre; the reverse sitution would be difficult to comprehend. The possibility of reverse distnce bis ws found in recent study of 6938 undergoing mjor surgery t tertiry-cre fcility. 17 In this study, ptients who trveled frther for their cre hd outcomes tht were worse thn expected. Our findings my therefore underestimte the importnce of surgeon/hospitl subspeciliztion in determining ptient outcomes. How should these results be trnslted into chnge? Access to cre is clerly importnt. The distribution of subspecilized cre is not perfectly mtched to the US popultion. There is only 1 colorectl surgeon ech in Alsk nd Wyoming, wheres there re pproximtely 150 in New York Stte ( 10-fold disproportion fter ccounting for popultion). 18 In previous work, we found tht fewer thn hlf of opertions for rectl cncer were performed in hospitls with colorectl surgeon vilble. 9 There re only 61 NCI-CCCs in the United Sttes, nd the geogrphic distribution of these centers is imperfect in our study only 6.5% of rectl cncer opertions occurred in these centers. Access is not the only issue, however. From the ptient perspective, choices regrding tretment re complicted, nd not directly relted to n estimtion of outcomes. Finlyson et l noted tht ptients were often 2480 Cncer August 15, 2014
Survivl After Surgery for Rectl Cncer/Etzioni et l inclined to receive complicted surgicl cre t locl center, even with the knowledge tht their likelihood of mortlity ws distinctly lower with trvel to regionl center. 19 Despite this, pproprite ptient choice cn only be empowered by the findings of our study. Given our findings, tht both surgeon nd hospitl fctors re importnt drivers of outcomes, orgnizing certified Centers of Excellence is n ttrctive option. The concept often becomes mired in prgmtic considertions regrding criteri for certifiction nd the propensity of pyers to restrict pyment bsed on certifiction. Historiclly, the pproch hs not worked well within the helth cre free mrket tht currently exists in the United Sttes. Regrdless of the perspective held, our study holds distinct importnce in informing ptients, pyers, nd providers who need to mke decisions regrding the tretment of rectl cncer. Those who re motivted to obtin the best possible outcomes re resonbly directed to centers nd surgeons with focused expertise. FUNDING SUPPORT No specific funding ws disclosed. CONFLICT OF INTEREST DISCLOSURES The uthors mde no disclosures. REFERENCES 1. Siegel R, Nishdhm D, Jeml A. Cncer sttistics, 2013. CA Cncer J Clin. 2013;63:11-30. 2. den Dulk M, vn de Velde CJ. Qulity ssurnce in surgicl oncology: the tle of the Dutch rectl cncer TME tril. J Surg Oncol. 2008;97:5-7. 3. Stewrt DB, Hollenbek C, Deshrnis S, et l. Rectl cncer nd teching hospitls: hospitl teching sttus ffects use of neodjuvnt rdition nd survivl for rectl cncer ptients. Ann Surg Oncol. 2013;20:1156-1163. 4. Scerdote C, Bldi I, Bertetto O, et l. Hospitl fctors nd ptient chrcteristics in the tretment of colorectl cncer: popultion bsed study. BMC Public Helth. 2012;12:775. 5. Archmpong D, Borowski D, Wille-Jorgensen P, et l. Worklod nd surgeon s specilty for outcome fter colorectl cncer surgery. Cochrne Dtbse Syst Rev. 2012;3:CD005391. 6. Finks JF, Osborne NH, Birkmeyer JD. Trends in hospitl volume nd opertive mortlity for high-risk surgery. N Engl J Med. 2011; 364:2128-2137. 7. vn Gijn W, Gooiker GA, Wouters MW, et l. Volume nd outcome in colorectl cncer surgery. Eur J Surg Oncol. 2010;36(Suppl 1):S55-S63. 8. Hodgson DC, Zhng W, Zslvsky AM, et l. Reltion of hospitl volume to colostomy rtes nd survivl for ptients with rectl cncer. J Ntl Cncer Inst. 2003;95:708-716. 9. Etzioni DA, Cnnom RR, Mdoff RD, et l. Colorectl procedures: wht proportion is performed by Americn bord of colon nd rectl surgery-certified surgeons? Dis Colon Rectum. 2010;53:713-720. 10. McArdle CS, Hole DJ. Emergency presenttion of colorectl cncer is ssocited with poor 5-yer survivl. Br J Surg. 2004;91:605-609. 11. In H, Neville BA, Lipsitz SR, et l. The role of Ntionl Cncer Institute-designted cncer center sttus: observed vrition in surgicl cre depends on the level of evidence. Ann Surg. 2012; 255:890-895. 12. Pulson EC, Mitr N, Sonnd S, et l. Ntionl Cncer Institute designtion predicts improved outcomes in colorectl cncer surgery. Ann Surg. 2008;248:675-686. 13. http://helthservices.cncer.gov/seermedicre/boutdt/pedsf.pdf. Accessed My 7, 2013. 14. Oliphnt R, Nicholson GA, Horgn PG, et l. Contribution of surgicl speciliztion to improved colorectl cncer survivl. Br J Surg. 2013;100:1388-1395. 15. Peeters KC, Mrijnen CA, Ngtegl ID, et l. The TME tril fter medin follow-up of 6 yers: incresed locl control but no survivl benefit in irrdited ptients with resectble rectl crcinom. Ann Surg. 2007;246:693-701. 16. Jcobson M, Erle CC, Newhouse JP. Geogrphic vrition in physicins responses to reimbursement chnge. N Engl J Med. 2011;365:2049-2052. 17. Etzioni DA, Fowl RJ, Wsif N, et l. Distnce bis nd surgicl outcomes. Med Cre. 2013;51:238-244. 18..http://www.bcrs.org/. Accessed My 17, 2013. 19. Finlyson SR, Birkmeyer JD, Tosteson AN, et l. Ptient preferences for loction of cre: implictions for regionliztion. Med Cre. 1999;37:204-209. Cncer August 15, 2014 2481