Health Outcomes Research

Perioperative Outcomes for Laparoscopic andRobotic Compared With Open ProstatectomyUsing the National Surgical QualityImprovement Program (NSQIP) DatabaseJen-Jane Liu, Bryan G. Maxwell, Periklis Panousis, and Benjamin I. Chung

OBJECTIVE To examine contemporary outcomes of minimally invasive radical prostatectomy (MIRP)

Financial Disclosure: The authoFrom the Department of Urolo

CA; and the Department of AStanford, CAReprint requests: Jen-Jane Liu,

Medical Center, 300 Pasteur Djenjaneliu@gmail.comSubmitted: December 19, 201

ª 2013 Elsevier Inc.All Rights Reserved

compared with open prostatectomy, using a national, prospective perioperative databasereflecting diverse practice settings.

METHODS The National Surgical Quality Improvement Program database was queried from 2005 to 2010 for

laparoscopic or robotic prostatectomy (Current Procedural Terminology code 55866) and openretropubic prostatectomy (Current Procedural Terminology codes 55840, 55842, 55845). Peri-operative outcomes examined were surgical and total operation duration, transfusion rates, lengthof stay, major morbidity (cardiovascular, pulmonary, renal, and infectious), and mortality.

RESULTS The study identified 5319 radical prostatectomies: 4036 MIRP and 1283 open. Although oper-

ative time was significantly longer in the MIRP group, there were significantly fewer perioperativeblood transfusions and shorter mean length of stay. Major postoperative morbidity and mortalitywere 5% in the MIRP group and 9% in the open group (P <.001). Age, body mass index,presence of medical comorbidities, and open surgical technique were all independently predictiveof major complications and mortality on multivariate analysis.

CONCLUSION In a nationwide database of diverse medical centers, MIRP was associated with longer operative

time, but a significantly decreased rate of blood transfusions, length of stay, perioperativecomplication rate, and mortality compared with open prostatectomy. The minimally invasivesurgical approach was independently associated with significantly fewer complications and deathson multivariate analysis. Compared with other administrative databases that capture only inpa-tient events, the National Surgical Quality Improvement Program identifies complications up to30 days postoperatively, providing more detailed characterization of complications after prosta-tectomy. These data reflect contemporary practice patterns and suggest that MIRP can be per-formed with low perioperative morbidity. UROLOGY 82: 579e583, 2013. � 2013 Elsevier Inc.

he introduction of laparoscopic urologic surgicalapproaches and subsequent development of

Trobotic techniques have dramatically increased

the use of minimally invasive radical prostatectomy(MIRP). The resulting shift in practice patterns has nowmade the robotic approach themost practiced technique inthe United States.1,2 Cited advantages of the roboticapproach include reduced blood loss and patientmorbidity,including shorter hospital stay and lower analgesiarequirements.3 Aggressive marketing and patient expec-tations drove the initial adoption of MIRP,4 rather than

rs declare that they have no relevant financial interests.gy, Stanford University School of Medicine, Stanford,nesthesia, Stanford University School of Medicine,

M.D., Department of Urology, Stanford Universityrive, S-287, Stanford, CA 94305-5118. E-mail:

2, accepted (with revisions): March 12, 2013

h

population-based confirmatory data detailing equivalent orsuperior outcomes. As such, there are few population-basedstudies that assess contemporary perioperative outcomes.

Using the Nationwide Inpatient Sample (NIS), Huet al showed that MIRP was associated with decreasedlength of stay, blood transfusion requirement, complica-tion rate, and overall mortality compared with openradical prostatectomy.5 Although that administrativedatabase contains a large sample of patients, datareporting can be challenging to interpret without stan-dardized definitions, data collection, and data manage-ment. Administrative databases determine complicationsby diagnosis codes that are entered during hospitalization,which can result in miscoding or missing data.6 In addi-tion, because only inpatient complications are identified,there may be incomplete capture of complications fora procedure with a short postoperative stay such asprostatectomy. Furthermore, readmissions within 30 dayscannot be captured by administrative databases.

0090-4295/13/$36.00 579ttp://dx.doi.org/10.1016/j.urology.2013.03.080

Figure 1. Adoption of minimally invasive radical prostatec-tomy (MIRP) over time in National Surgical Quality Improve-ment Program (NSQIP).

Because of the shortcomings in documenting compli-cations using administrative databases, we used theAmerican College of Surgeons National Surgery QualityImprovement Program (NSQIP) to better understand thecomplications of MIRP and open radical prostatectomy.NSQIP was developed specifically to assess the quality ofsurgical care,7 and prospectively collects perioperativedata on 135 perioperative variables. In 2010, there were258 participating sites and more than 1.3 million caseshave been contributed to date. NSQIP records compli-cations for 30 days postoperatively (even those recordedunder a separate admission) and has been shown to morereliably detect complications and mortality comparedwith administrative databases or traditional institutionalmorbidity and mortality reporting.8-10 We sought tocharacterize contemporary perioperative outcomes ofMIRP compared with open radical prostatectomy ona national level using the NSQIP database to assess thesafety of each technique.

METHODS

The Stanford University Institutional Review Board granted anexemption from review, as this study uses deidentified data.NSQIP datasets for the years 2005 through 2010 were queriedfor MIRP (Current Procedural Terminology [CPT] code 55866;includes laparoscopic and robotic assisted radical prostatectomy)or open retropubic prostatectomy (CPT codes 55840, 55842,and 55845). Demographics (age, body mass index [BMI], race,American Society of Anesthesiology physical status), baselinecomorbidities, surgical and total operation duration, presence ofresident during operation, perioperative blood transfusion,length of hospital stay, major postoperative complications, andmortality were examined. Coding for blood transfusion from2005 to 2009 in the NSQIP database was only for intraoperativetransfusions, whereas starting in 2010, blood transfusions werecoded for the entire perioperative period (intraoperative and 30days postoperatively). Major postoperative complications, listedas follows, were grouped according to organ system. Cardio-vascular complications included cardiac arrest, myocardialinfarction, or cerebrovascular accident. Pulmonary complica-tions included pneumonia, need for postoperative reintubation,and need for ventilatory support >48 hours. Deep venousthrombosis and pulmonary embolism were grouped together.Postoperative renal failure was defined as a rise in serum creat-inine >2 mg/dL from preoperative value, but not requiringdialysis. Deep organ space infection was grouped separately fromsurgical site infection. Because surgical site infection codingincluded superficial and deep fascial infections, this wasconsidered as a major complication.

Continuous outcomes were compared using a 2-sample t test.Binary outcomes were compared using Fisher’s exact test orPearson’s chi-square test and odds ratios with 95% confidenceintervals.Multivariate logistic regressionmodels were constructedfor a composite outcome ofmortality andmajormorbidity to assessfor factors contributing to postoperative major complication ormortality, including age, BMI, smoking status, presence of dia-betes mellitus, presence of hypertension, presence of othermedical comorbidities, race, presence of resident during surgery,and surgical approach (robotic or laparoscopic vs open). Analyseswere performed using SAS (SAS 9.3, SAS Institute, Cary, NC).

580

RESULTSThe study identified 5319 prostatectomies: 4036 (76%)MIRP and 1283 open. Adoption of MIRP increased overtime, accounting for 75% of all radical prostatectomies in2010 (Fig. 1). Baseline characteristics of the 2 groups arelisted in Table 1. Patients undergoing MIRP had slightlylower median American Society of Anesthesiologyphysical status and fewer major medical comorbidities,although rates of diabetes and hypertension were similarbetween the 2 groups. There were also fewer smokers(26% vs 30%, P ¼ .004) in the MIRP group. Presence ofa resident during surgery was more common in MIRP(55% vs 50%, P ¼ .005).

Total operating room time was 21 minutes longer forMIRP (Table 2). The increased time was because ofincreased surgical time (270 minutes vs 252 minutes,P <.0001) for MIRP. Rates of blood transfusion weresignificantly higher in the open group (21% vs 1.3%,P <.0001). When transfusions were required, the mediannumber of units transfused was 1.0 in MIRP comparedwith 1.5 in open prostatectomy. Length of stay wassignificantly shorter after MIRP (1.8 vs 3.2 days).

Major complication rates for cardiovascular, deepvenous thrombosis or pulmonary embolism, deep organspace infection, and surgical site infection were allsignificantly lower in the MIRP group (Table 2). Therewas a trend toward fewer respiratory complications. Onlypostoperative renal failure was similar between the 2groups. Overall complication rate after MIRP was 5%compared with 9% in open prostatectomy (P <.0001).Mortality was low in both groups, although it wassignificantly lower in the MIRP group (0.05%)compared with the open prostatectomy group (0.4%,P ¼ .01).

Multivariate analysis revealed that age, BMI, medicalcomorbidities other than diabetes or hypertension, andopen surgical approach were all independently associatedwith postoperative complications and mortality(Table 3). Similar results were obtained with multivariate

UROLOGY 82 (3), 2013

Table 1. Patient demographics

Variable MIRP (n ¼ 4036) Open Radical Prostatectomy (n ¼ 1283) P Value

Age (y, mean � SD) 61.7 � 7.3 62.8 � 7.7 <.0001Caucasian 2903 (82%) 930 (83%) .416BMI (kg/m2; mean � SD) 29.9 � 5.3 29.9 � 4.9 .91ASA physical status1 140 (3%) 53 (4%) .00092 2606 (65%) 749 (58%)3 1250 (31%) 465 (36%)4 31 (0.8%) 16 (1%)

DM 177 (4%) 67 (5%) .220HTN 2066 (51%) 664 (52%) .749Other medical comorbidity* 208 (5%) 86 (7%) .042Smoker 1064 (26%) 390 (30%) .005Resident present during surgery 703 (55%) 2022 (50%) .004

ASA, American Society of Anesthesiology; BMI, body mass index; DM, diabetes mellitus; HTN, hypertension; MIRP, minimally invasiveradical prostatectomy; SD, standard deviation.

Values are number (percent) unless otherwise noted.* Includes baseline dyspnea, chronic obstructive pulmonary disease (COPD), congestive heart failure (CHF), previous myocardial infarction(MI), previous cardiac stenting, angina, peripheral vascular disease, claudication, esophageal varices, hemodialysis, history of cerebro-vascular accident or transient ischemic attach, hemiplegia, baseline renal insufficiency (creatinine �2 mg/dL).

Table 2. Perioperative variables and complications

Variable MIRP Open Radical Prostatectomy P Value

Total OR time (min; mean � SD) 211 min � 1.31 190 min � 2.87 <.0001Induction to incision (min; mean � SD) 40 min � 0.32 41 min � 0.53 .025Surgical time (min; mean � SD) 270 min � 1.17 252 min � 2.54 <.0001Need for blood transfusion 54 (1.34%) 274 (21.36%) <.0001Length of stay (d, mean � SD) 1.81 � 0.04 3.16 � 0.09 <.0001ComplicationCardiovascular (arrest, CVA, MI) 15 (0.37%) 14 (1.09%) .01Respiratory (pneumonia, ventilator >48 h, reintubation) 19 (0.47%) 12 (0.94%) .06DVT or PE 33 (0.82%) 23 (1.79%) .003Renal failure 24 (0.59%) 9 (0.7%) .68Deep organ space infection 14 (0.35%) 11 (0.86%) .032Surgical site infection 60 (1.49%) 71 (5.54%) .001Composite 201 (4.98%) 116 (9.04%) <.0001Death 2 (0.05%) 5 (0.39%) .01

CVA, cerebrovascular accident; DVT, deep venous thrombosis; MI, myocardial infarction; OR, operating room; PE, pulmonary embolism;other abbreviations as in Table 1.

Values are number (percent) unless otherwise noted.

logistic regression models for nonfatal morbidity alone asthe modeled outcome (data not shown).

COMMENTMinimally invasive prostatectomy has rapidly beenadopted without population-based evidence demon-strating superiority in terms of complications and onco-logic outcomes. Our data confirm that the incidence ofmajor postoperative complications and mortality, lengthof hospital stay, and need for blood transfusion are lowerafter MIRP compared with open prostatectomy. Therewere some baseline differences between the patientcohorts, likely because of the patient selection bias.However, on multivariate analysis, the minimally inva-sive surgical approach was independently associated withdecreased postoperative complications and mortality aftertaking into account these baseline differences on vari-ables contained in the NSQIP database. Differences inother unmeasured characteristics may exist between

UROLOGY 82 (3), 2013

patients in the MIRP and open groups. Early endeavorsbased on Surveillance Epidemiology and End Results(SEER) analyses initially demonstrated a higher rate ofgenitourinary complications with MIRP.3 Subsequentpopulation-based studies based on administrative data-bases demonstrated that there were fewer respiratory,wound, and vascular complications, and also mortality inpatients undergoing robotic assisted radical prostatectomycompared with the open approach.11

Administrative databases such as NIS and SEER relyon coding data to analyze complications, which can resultin inaccurate characterization of complications.12

NSQIP, however, uses detailed complication definitionsentered by trained clinical nurses at each institution.Recent studies document high accuracy in data entry witha discordance rate between institutions of 1.8%.7 Addi-tionally, only inpatient events are coded in administrativedatabases, limiting their ability to identify complicationsthat occur after the postoperative hospital stay. As lengthof stay for radical prostatectomy is short, data based on

581

Table 3. Multivariate logistic regression for factors pre-dicting major complications and mortality

Variable

Major Complicationsand Mortality

OddsRatio

ConfidenceInterval

Age 1.04 1.02-1.06Race (white vs nonwhite) 0.96 0.66-1.40BMI 1.03 1.01-1.05HTN 1.15 0.85-1.56DM 0.96 0.51-1. 82Smoker 1.35 0.995-1.83Other medical comorbidities* 2.11 1.34-3.21Robotic or laparoscopic vs open 0.56 0.41-0.75Resident present 1.08 0.81-1.44

Abbreviations as in Table 1.* Including dyspnea, COPD, CHF, previous MI, previous cardiacstenting, angina, peripheral vascular disease, claudication,esophageal varices, hemodialysis, history of CVA/TIA, hemiplegia,baseline renal insufficiency (Cr >1.99).

inpatient coding might not accurately capture perioper-ative outcomes. NSQIP captures data for 30 days post-operatively, ensuring thorough collection of complicationdata. Because of the longer data collection period, onemight expect a higher complication rate compared withother studies using inpatient data alone; however, thiswas not observed. In fact, rates of serious complicationswere lower in our study compared with those reportedfrom NIS.11

Lower complication and mortality rates observed inNSQIP could be the result of several factors. NSQIP hasstandardized definitions of complications, and severalstudies have shown that NSQIP is more sensitive andaccurate for detection and characterization of relevantcomplications compared with administrative databases.Reliance on diagnosis coding by administrative databasescan lead to reporting of complications that are not clinicallysignificant.10 Thus, NSQIP does not simply identify morecomplications, but rather, identifies those complicationsthat are most predictive of mortality.8,9 As participation inNSQIP is voluntary and requires resources to maintainpersonnel for data entry, it may select for larger, high-volume institutions with more experience performingMIRP. High volume centers are known to have lowercomplication rates after MIRP.11 The large percentage ofMIRP (76%) compared with open prostatectomy in ourstudymay also be a reflection of this bias. The contemporarynature of the data in NSQIP (2005-2010) may account forthe low complication andmortality rate, as itmay reflect thewidespread adoption of MIRP beyond the initial learningcurve. Other studies using NIS and Medicare that reporthigher complication rates after MIRP spanned earlier timeperiods (2003-2007),whichmayhave reflected the learningcurve of MIRP.5,13 Thus, if NSQIP reflects contemporarypractice patterns at high volume centers, one can expecta comparatively low rate of complications and death.

In many quality databases, procedural coding produceslimitations. As CPT codes do not distinguish laparoscopicvs robotic prostatectomy, we cannot separate these 2

582

techniques in the MIRP group. However, because thedataset contains data from 2005 to 2010, it is likely thatthe vast majority of procedures coded as MIRP representrobotic cases, as a recent NIS study in 2008 using ICD-9procedure codes that differentiate between laparoscopicand robotic assisted prostatectomy demonstrated thatonly 3% of MIRPs were being performed laparoscopi-cally.5 Furthermore, CPT coding may not provide reliableinformation with regard to pelvic lymph node dissection(PLND), which could contribute to operative time,transfusion rates, and complications. For open prostatec-tomy, there are separate bundled CPT codes that includeradical retropubic prostatectomy and PLND that wereused. However, for MIRP there is 1 code for prostatec-tomy, and a separate one for PLND. Thus, undercoding ofPLND is more likely to be encountered in the MIRPgroup. As coding of PLND was imbalanced (and poten-tially unreliable) in this dataset, we did not stratify ouranalysis by the performance of PLND. Databases withbetter information on pathologic data could be used toassess whether PLND affects quality outcomes.

Individual hospital and payer characteristics are notavailable within NSQIP, limiting the ability to assess theeffects of hospital volume and geography. Additionally, asNSQIP is a quality database, information on oncologic andother prostatectomy-specific outcomes (eg, urinary leak,potency, and continence rates) is not available. Althoughthe complications are tightly defined, there is limitedinformation on certain aspects of the patient’s post-operative stay, which prevents identification of a stan-dardized complication classification system such as theClavien system. Despite these limitations, multiple studieshave demonstrated that NSQIP is an excellent predictor ofpostoperative morbidity and mortality. Risk calculators formorbidity andmortality have been generated and validatedin the general surgery literature.14

CONCLUSIONOur study is a prospective, population-based study thatdemonstrates lower transfusion rates, length of hospitalstay, and serious postoperative complication andmortality rates for MIRP compared with open prostatec-tomy. We used a highly accurate, quality improvementdatabase to provide contemporary evidence that MIRPcan be safely performed with low perioperative morbidityand extremely low mortality.

References

1. Lee DI. Robotic prostatectomy: what we have learned and where weare going. Yonsei Med J. 2009;50:177-181.

2. Stitzenberg KB, Wong Y-N, Nielsen ME, et al. Trends in radicalprostatectomy: centralization, robotics, and access to urologic cancercare. Cancer. 2012;118:54-62.

3. Hu JC, Gu X, Lipsitz SR, et al. Comparative effectiveness ofminimally invasive vs open radical prostatectomy. JAMA. 2009;302:1557-1564.

4. Alkhateeb S, Lawrentschuk N. Consumerism and its impacton robotic-assisted radical prostatectomy. BJU Int. 2011;108:1874-1878.

UROLOGY 82 (3), 2013

5. Yu H, Hevelone ND, Lipsitz SR, et al. Use, costs and comparativeeffectiveness of robotic assisted, laparoscopic and open urologicalsurgery. J Urol. 2012;187:1392-1398.

6. Jollis JG, Ancukiewicz M, DeLong ER, et al. Discordance of data-bases designed for claims payment versus clinical informationsystems. Implications for outcomes research. Ann Intern Med. 1993;119:844-850.

7. Shiloach M, Frencher SK, Steeger JE, et al. Toward robust infor-mation: data quality and inter-rater reliability in the AmericanCollege of Surgeons National Surgical Quality ImprovementProgram. J Am Coll Surg. 2010;210:6-16.

8. Davenport DL, Holsapple CW, Conigliaro J. Assessing surgicalquality using administrative and clinical data sets: a directcomparison of the University Health System Consortium ClinicalDatabase and the National Surgical Quality Improvement Programdata set. Am J Med Qual. 2009;24:395-402.

9. Cima RR, Lackore KA, Nehring SA, et al. How best to measuresurgical quality? Comparison of the Agency for Healthcare Researchand Quality Patient Safety Indicators (AHRQ-PSI) and theAmerican College of Surgeons National Surgical QualityImprovement Program (ACS-NSQIP) postoperative adverse ev.Surgery. 2011;150:943-949.

10. KochCG,Li L,HixsonE, et al.What are the real rates of postoperativecomplications: elucidating inconsistencies between administrativeand clinical data sources. J Am Coll Surg. 2012;214:798-805.

11. Yu H-Y, Hevelone ND, Lipsitz SR, et al. Hospital volume, utiliza-tion, costs and outcomes of robot-assisted laparoscopic radicalprostatectomy. J Urol. 2012;187:1632-1637.

12. Berthelsen CL. Evaluation of coding data quality of the HCUPNational Inpatient Sample. Top Health Inf Manage. 2000;21:10-23.

13. Kowalczyk KJ, Levy JM, Caplan CF, et al. Temporal national trendsof minimally invasive and retropubic radical prostatectomyoutcomes from 2003 to 2007: results from the 100% Medicaresample. Eur Urol. 2012;61:803-809.

14. Cohen ME, Bilimoria KY, Ko CY, et al. Development of anAmerican College of Surgeons National Surgery Quality Improve-ment Program: morbidity and mortality risk calculator for colorectalsurgery. J Am Coll Surg. 2009;208:1009-1016.

EDITORIAL COMMENT

Over the last decade, robot-assisted radical prostatectomy(RARP) has become the most commonly performed surgery formen with localized prostate cancer. By 2013, RARP has beenestimated to be approximately 80% of all surgeries for prostatecancer.1 Many critics have argued that the major reason for therapid dissemination of RARP is the aggressive marketing.Indeed, several high profile litigations have recently been filedin various jurisdictions because of poor outcomes after RARP.

In this regard, the authors of the present study attempted tocompare the complication rates between minimally invasiveradical prostatectomy (MIRP) and open radical prostatectomyusing the database National Surgical Quality ImprovementProgram (NSQIP). Because NSQIP was established to captureall events within 30 days of surgery prospectively by examiningthe clinical data, NSQIP is inherently more accurate in assessingperioperative complications when compared with thecommonly used databases such as Nationwide Inpatient Sampleand Surveillance Epidemiology and End Results (SEER). Theresults revealed that MIRP took significantly longer to completebut had a dramatically lower complication rate. These resultsconfirm the previous report by Hu et al that suggested thatMIRP is associated with lower perioperative complications thanopen radical prostatectomy.2

Despite this strength, this study falls short in that a detailedanalysis of long-term outcomes is not possible. Indeed, owing to

UROLOGY 82 (3), 2013

the inherent nature of NSQIP, continence, potency, andoncologic results are not available. Previously, it has been re-ported that MIRP is associated with inferior outcome in conti-nence and potency.2 As surgical techniques have been furtherrefined and additional long-term data have accumulated, newpopulation-based studies on long-term outcome after MIRP/RARP are eagerly anticipated.

Isaac Yi Kim, M.D., Ph.D., Section of Urologic Oncology, TheCancer Institute of New Jersey, Robert Wood JohnsonMedical School, 195 Little Albany Street, #4565, NewBrunswick, NJ

References

1. Skarecky DW. Robotic-assisted radical prostatectomy after the firstdecade: surgical evolution or new paradigm. ISRN Urol. 2013;2013:1-22.

2. Hu JC, Gu X, Lipsitz SR, et al. Comparative effectiveness of mini-mally invasive vs open radical prostatectomy. JAMA. 2009;302:1557-1564.

http://dx.doi.org/10.1016/j.urology.2013.03.082UROLOGY 82: 583, 2013. � 2013 Elsevier Inc.

REPLY

There is no question that the advancement of robotic andlaparoscopic techniques has contributed to the dominance ofminimally invasive surgery for prostate cancer. Minimallyinvasive radical prostatectomy (MIRP) was adopted rapidly,before large population-based studies could be performed. Wesought to explore a unique dataset with comprehensive data oncomplications and mortality that encompasses the contemporarytime period over which MIRP has been adopted. Although theNational Surgical Quality Improvement Program cannotprovide data on oncologic outcomes, it is highly accurate inpredicting postoperative morbidity and mortality.1 The nextstep is to validate the long-term oncologic efficacy andmorbidity of MIRP for prostate cancer.Minimally invasive surgery is a tool for the surgeon. All

urologists agree that the goals of extirpative prostate cancersurgery remain the same in MIRP and that the fundamentalsteps of the operation are similar. With the large number ofdiverse patients being treated with radical prostatectomy,a randomized trial is the only way to determine if MIRP issuperior (or equivalent) to open radical prostatectomy; however,such a study is unlikely to occur given the current dominance ofMIRP and the long duration of follow-up necessary toadequately track oncologic outcomes.

Jen-Jane Liu, M.D., and Benjamin I. Chung, M.D.,Department of Urology, Stanford University School ofMedicine, Stanford, CA

Reference

1. Koch CG, Li L, Hixson E, et al. What are the real rates of post-operative complications: elucidating inconsistencies betweenadministrative and clinical data sources. JACS. 2012;214:798-805.

http://dx.doi.org/10.1016/j.urology.2013.03.083UROLOGY 82: 583, 2013. � 2013 Elsevier Inc.

583