ireeecabilidad or tac

7/29/2019 Ireeecabilidad or Tac

1/6

Multi-Institutional Reciprocal Validation Study ofComputed Tomography Predictors of Suboptimal PrimaryCytoreduction in Patients With Advanced Ovarian CancerAllison E. Axtell, Margaret H. Lee, Robert E. Bristow, Sean C. Dowdy, William A. Cliby, Steven Raman,John P. Weaver, Mojan Gabbay, Michael Ngo, Scott Lentz, Ilana Cass, Andrew J. Li, Beth Y. Karlan, andChristine H. Holschneider

A B S T R A C T

PurposeIdentify features on preoperative computed tomography (CT) scans to predict suboptimal primarycytoreduction in patients treated for advanced ovarian cancer in institution A. Reciprocally crossvalidate the predictors identified with those from two previously published cohorts from

institutions B and C.Patients and MethodsPreoperative CT scans from patients with stage III/IV epithelial ovarian cancer who underwentprimary cytoreduction in institution A between 1999 and 2005 were retrospectively reviewed byradiologists blinded to surgical outcome. Fourteen criteria were assessed. Crossvalidation wasperformed by applying predictive model A to the patients from cohorts B and C, and reciprocallyapplying predictive models B and C to cohort A.

ResultsSixty-five patients from institution A were included. The rate of optimal cytoreduction ( 1 cmresidual disease) was 78%. Diaphragm disease and large bowel mesentery implants were the onlyCT predictors of suboptimal cytoreduction on univariate (P .02) and multivariate analysis(P .02). In combination (model A), these predictors had a sensitivity of 79%, a specificity of 75%,and an accuracy of 77% for suboptimal cytoreduction. When model A was applied to cohorts Band C, accuracy rates dropped to 34% and 64%, respectively. Reciprocally, models B and C hadaccuracy rates of 93% and 79% in their original cohorts, which fell to 74% and 48% in cohort A.

ConclusionThe high accuracy rates of CT predictors of suboptimal cytoreduction in the original cohorts couldnot be confirmed in the cross validation. Preoperative CT predictors should be used with cautionwhen deciding between surgical cytoreduction and neoadjuvant chemotherapy.

J Clin Oncol 25:384-389. 2007 by American Society of Clinical Oncology

INTRODUCTION

Ovarian cancer remains the leading cause of

mortality from a gynecologic malignancy in the

United States with 16,210 deaths annually.1

Themajority of cases are advanced stage (stage III/

IV) at the time of diagnosis. Current front-line

therapy consists of cytoreductive surgery and

platinum-based chemotherapy.

Optimal primary cytoreduction has beendem-

onstrated for the past 30 years to be a highly signifi-

cantpredictor of outcomein patientswith advanced

ovarian cancer. A recent meta-analysis of maximal

cytoreduction and survival in 81 published patient

cohorts demonstrated that cohorts in which there

was a high proportion of maximal cytoreduction

( 75%) had a 50% increase in median survival

compared with those with a less than 25% maximal

cytoreduction rate (33.9 v22.7 months).2 A higher

response rate to chemotherapy and improved sur-

vival in patients with optimalcytoreduction,definedby residualdisease1 cm, continues to be observed

with contemporary first-line chemotherapy with a

platinum compound and taxane. This can be illus-

trated by a comparison of two Gynecologic Oncol-

ogy Group (GOG) studies, GOG 1113 and GOG

158.4 In GOG 111, patients with suboptimally cy-

toreduced ovarian cancer demonstrated a 26%

complete pathologic response rate to cisplatin and

paclitaxel as determined by second-look surgery. In

GOG 158, where optimally cytoreduced patients

were treated with paclitaxel and either carboplatin

From the University of California Los

Angeles (UCLA) Medical Center; Olive

View-UCLA Medical Center; Cedars-

Sinai Medical Center; Kaiser Perma-

nente Sunset Medical Center, Los

Angeles, CA; Johns Hopkins Medical

Institutions, Baltimore, MD; and the

Mayo Clinic, Rochester, MN.

Submitted June 7, 2006; accepted

November 17, 2006.

Presented in part during plenary presen-

tations at the 34th Annual Meeting of the

Western Association of Gynecologic

Oncologists, Santa Fe, NM, June 15-18,

2005; and at the 37th Annual Meeting of

the Society of Gynecologic Oncologists,

Palm Springs, CA, March 22-26, 2006.

Authors disclosures of potential con-

flicts of interest and author contribu-

tions are found at the end of this

article.

Address reprint requests to Christine H.

Holschneider, MD, Olive View-UCLAMedical Center, Department of Obstet-

rics and Gynecology, 14445 Olive View

Dr, Rm 2B-163, Sylmar, CA 91342;

e-mail: [email protected].

2007 by American Society of Clinical

Oncology

0732-183X/07/2504-384/$20.00

DOI: 10.1200/JCO.2006.07.7800

JOURNAL OF CLINICAL ONCOLOGY O R I G I N A L R E P O R T

V O LU M E 2 5 N U MB ER 4 F E BR UA RY 1 2 0 07

384

Downloaded from jco.ascopubs.org on July 6, 2011. For personal use only. No other uses without permission.Copyright 2007 American Society of Clinical Oncology. All rights reserved.


2/6

or cisplatin, the complete pathologic response rate at second-look

surgery was almost twice as high (49%), with an associated difference

in median survivalof 15 months (38 months [GOG 111] v53 months

[GOG 158]). Striving for maximal primary cytoreduction becomes

even more important in light of recently published data from GOG

172,a randomizedtrial of patientswithoptimally cytoreducedovarian

cancer that compared intravenous paclitaxel plus cisplatin versus in-

travenous paclitaxel plus intraperitoneal cisplatin and paclitaxel. An

additional gain in median survival to 66 months was observed for

patients in the intraperitoneal treatment arm.5

For women with advanced ovarian cancer, rates of optimal pri-

mary cytoreduction vary widely from 25% to more than 90%.6 GOG

data have demonstrated that women whose tumors cannotbe cytore-

duced to smaller than 2 cm residual disease do not derive any signifi-

cant survival benefit from primary cytoreductive surgery.7 Thus,

patients who undergo suboptimal primary cytoreduction may incur

significant surgical morbidity without associated gain in survival.

This has prompted investigations into neoadjuvant chemother-

apy. Several trials have demonstrated decreased operative morbidity8

and improved maximal tumor debulking rates at interval cytoreduc-

tion after neoadjuvant chemotherapy.9-15

To date, there are no pub-lished randomized trials thatcomparesurvivalin patientswho receive

neoadjuvant chemotherapy with interval cytoreduction versus those

who undergo primary cytoreduction followed by chemotherapy. A

European Organisation for Research and Treatment of Cancer trial is

currently underway, which addresses this question.16 An extensive

review of retrospective and nonrandomized prospective studies of

neoadjuvant chemotherapy versus primary cytoreduction suggests

improved median survival with neoadjuvant chemotherapy over that

observed in patients who underwent suboptimal primary cytoreduc-

tion (26monthsv20 months).6 However, in the literature available to

date, survival of patients treated with neoadjuvant chemotherapy ap-

pears inferior to that observed after optimal primary cytoreduction

(26 months v51 months).6

Thus, the accurate pretreatment identifi-cation of patients whose disease is not optimally cytoreducible at

primary surgery becomes one of the most critical issues surrounding

neoadjuvant chemotherapy for ovarian cancer.

Investigators have attempted to identify specific preoperative

predictors of suboptimal cytoreduction. A number of studies have

demonstrated an association between the preoperative CA-125 level

and the inability to achieve optimal cytoreduction, yet the overall

accuracy rates at predicting surgical outcome (ie, optimal vsubopti-

mal cytoreduction) were only 50% to 78% with most studies using a

CA-125 cut off value of 500 U/mL.17-23 The two series with high

optimal cytoreduction rates ( 70%) found preoperative CA-125

levels completely lacking as predictors of surgical outcome.20-22 Five

small studies have been published to date,22,24-27 which have at-tempted to identify specific radiological predictors of suboptimal cy-

toreduction on preoperative computed tomography (CT) scans. For

example, Bristow et al26 created a model including 13 radiographic

features and performance status to calculate predictive index scores

with a 93% accuracy rate for optimal versus suboptimal primary

debulking. Dowdy and colleagues22 found only diffuse peritoneal

thickening to be an independent predictor of suboptimal surgical

cytoreduction. Interpretation of these studies published to date is

limited due to their retrospective nature, the highly variable rates of

optimal cytoreduction (33% to 80%), the fact that most2,24,25,27 in-

cluded patients with both early and advanced stage disease, and the

quite different combination of CT predictors that correlated with

suboptimal cytoreduction in each of the study cohorts; the latter

calling into question the applicability of identified CT predictors to

other patient populations.

We therefore undertook the currentstudy with thefollowing two

objectives: to identify radiologic features on preoperative CT scans

that predict suboptimal primary cytoreduction in a specific cohort of

patientswith advanced stage (III/IV) ovariancancer treatedat Univer-

sityof California,Los Angeles(LosAngeles,CA) and associated teach-

ing institutions, where the surgical practice is characterized by a strong

commitment to optimal primary cytoreduction (cohort A); and to

reciprocally cross validate the CT predictors identified in cohort A

withthose fromtwo previously published cohortsof patientwith stage

III/IV ovarian cancer who underwent primary cytoreduction at Johns

Hopkins Medical Institute26 (Baltimore, MD; cohort B) or the Mayo

Clinic22 (Rochester, MN; cohort C).

PATIENTS AND METHODS

Objective 1

Institutional review board approval of the study protocol was obtainedfrom all participating institutions. Patients who underwent primary surgeryfor stage III and stage IV epithelial ovarian cancer between 1999 and 2005 atone of four teaching institutions affiliated with the University of California,Los Angeles Gynecologic Oncology training program (patient cohort A) wereidentified through pathology databases and institutional tumor registries.Only patients who had preoperative CT scans performed within 4 weeksbefore primary cytoreductive surgery and whose CT films were available forreview were included in the study.

CT scanning protocols varied given the inclusion of four different insti-tutions.In general,all imageswereobtained using 5 mmto 10mm collimationthrough theabdomen andpelvis with PO andIV contrast unlessthe latter wasmedically contraindicated. Patients whose preoperative CT scans had beenperformed at outlying institutions were not eligible for this study given the

large variation in the imaging techniques used and the fact that images hadgenerally been returned to the originating institution. CT scans performed atthe study institutions were systematically re-reviewed by study radiologistswho were blinded to surgical outcomes. Fourteen radiologic criteria werechosen from pertinent positive predictors gatheredfrom previous studies22,26

and supplemented with potential predictors from clinical experience. Thesecriteriaincluded largevolumeascites, pleuraleffusion,diffuseperitoneal thick-ening, omental caking, omental extension to spleen or stomach, suprarenallymph nodes larger than 1 cm, infrarenal or inguinallymph nodes largerthan2 cm, and tumor implants larger than 2 cm on small and large bowel mesen-tery, peritoneum, diaphragm, liver, or portahepatis.Large volume ascites wasdefined as the presence of ascites on two thirds of abdominopelvic CT scancuts. Diffuseperitoneal thickening was defined as peritoneal thickening to4mminvolving at least twoof thefive following areas: lateralcolic gutters,lateralconal fascia,anteriorabdominal wall, diaphragm, andpelvic peritoneal reflec-

tions, as described by Dowdy.22

Demographic data, surgical findings, and pathologic data were retro-spectively obtained from the medical record. All surgeries were performed byoneof 12gynecologiconcologistsat oneof thefourstudyinstitutions. Optimalcytoreduction wasdefined as1 cm residual disease.The American Society ofAnesthesiologists (ASA) physical status classification was obtained from theanesthesia record. Laboratory values collected included preoperative CA-125and serum albumin levels obtained within 4 weeks before surgery.

Univariate comparisons of the percentage of patients who underwentsuboptimal cytoreduction werecarried outusing Fishersexact testsfor each ofthe14 potential radiologic predictors. TheWilcoxon ranksum testwas usedtocompare median age, albumin, ASA status, and CA-125 levels between pa-tients with optimal versus suboptimal cytoreduction. The Kruskal-Wallis testwasused to study theassociationbetweenincreasing ASAclassificationandthe

CT Predictors of Ovarian Cancer Cytoreduction

www.jco.org 385



3/6

proportion of patients with suboptimally cytoreduced ovarian cancer. Simul-taneous multivariate assessment of all 14 radiologic and four clinical variableswas carried out using backward stepwise logistic regression with a liberalP .15 variable retention criterion and a model predictive of suboptimalcytoreduction wasdeveloped (predictivemodel A).Sensitivity,specificity,andaccuracy of the model were calculated based on receiver operating curveanalysis. Stratification of the data by surgeon and/or procedures performedwould havebeen desirable since optimal cytoreduction depends in parton thesurgeons practice, philosophy,and skill set. However, in this study, thenum-

berof patientsper surgeon and/orper radical surgical techniquewas toosmallto allow for a stratified analysis.

In order to safeguardagainst statistical overfitting, we performed a leaveoneout(statistical)cross validation.In short, an observationwasleft outof thedata set, the process of identifying (variable selection) and fitting the modelwasperformedon all theremainingdata, andthenthefittedmodelwas used topredict the probability of a suboptimal cytoreduction for the observation leftout. These steps were repeated for each observation in the data set, a receiveroperating curve curve was formed, and resultant sensitivity, specificity, andaccuracy reported.

Objective 2Cross validation of thepredictors of suboptimal cytoreduction identified

in this study (predictive model A) with those identified from two previouslypublished studies was performed by applying predictive model A to patientcohortsB andC andreciprocally applyingpredictivemodels B andC to patientcohort A. Patient cohorts B26 and C22 were chosen from the five previouslypublished studies on CT predictors of cytoreduction,22,24-27 as those were theonly studies restricted, similar to cohort A, to patients with advanced-stageovarian cancer. Patient cohort B was comprised of 41 previously publishedpatients with stage III/IV disease who underwent preoperative CT scans fol-lowed by cytoreductive surgery at Johns Hopkins Medical Institutions or theMassachusetts General Hospital with a 48% optimal cytoreduction rate.26

Predictive model B entailed an additive predictive index score 4 calculatedbased on 13 radiologic criteria and performance status with each variablepresent being allotteda weightedpointvaluebetween1 and2. PatientcohortCconsists of 87 patients from a previously published study22 who underwentpreoperative CT scanning followed by cytoreductive surgery for stage III/IVovarian cancer at the Mayo Clinic with a 71% optimal cytoreduction rate.

Predictive model C was comprised of two radiologic factors identified on preop-erative CT scansdiffuse peritoneal thickening and large volume ascites. For allcrossvalidationswereportsensitivity,specificity,andunweightedaccuracy.Sensi-tivity reports the percentageof those patients with suboptimal debulking that arepredictedsuboptimal bythe model.Specificityreportsthepercentageof thosewithoptimal debulking thatare predicted optimal. Unweighted accuracy is defined asthe unweightedaverageof sensitivityand specificity.

RESULTS

Sixty-five consecutive patients frominstitutionA met study inclusion

criteria. Demographic and clinical data are described in Table 1.

Eighty-eight percent of patients had International Federation of Gy-

necology and Obstetrics staging system (FIGO) stage III disease while12% of patients had FIGO stage IV disease. Fifty-one patients (78%)

were optimally cytoreduced to 1 cm residual disease at the time of

primary surgery.

There were no statistically significant differences between the

median age (optimal: 62 years; range, 33 to 82 years; suboptimal: 56

years; range, 34 to 87 years; P .92), ASA status (optimal: 2; range, 2

to 4; suboptimal: 3; range, 2 to 4; P .12), serum albumin (optimal:

3.5 g/dL; range, 2.2 to 4.7 g/dL; suboptimal, 3.4 g/dL; range, 2.7 to 4.0

g/dL; P .62) and CA-125 levels (optimal: 860 U/mL; range, 15 to

7,960 U/mL; suboptimal: 780 U/mL; range, 105 to 2,866 U/mL);

P .42) of individuals who were optimally cytoreduced when com-

pared withthosewhoweresuboptimallydebulked. An associationwas

noted between the ASA classification and the proportion of patients

who underwent suboptimal cytoreduction, but this did not reach

statistical significance: 13% of patients with an ASA classification of 2

underwent suboptimal cytoreduction, this increased to 25% with an

ASA class of 3 and to 40% with an ASA classification of 4 (P .12).

Table 2 presents the percentage of patients who underwent sub-

optimal debulking for each of the 14 preoperative CT variables. Dia-

phragmatic disease larger than 2 cm (P .02) and large bowel

mesentery implants larger than 2 cm (P .02) were the only statisti-cally significant univariate predictors of suboptimal cytoreduction.

Forty-seven percent of women who were positive for diaphragm dis-

ease and 50% of those with disease of the large bowel mesentery on

preoperative CT scan were suboptimally debulked.

All 14 radiologic and four clinical criteria were candidates for

predicting suboptimal debulking in the backwardmultivariate logistic

analysis. Of these 18 potential predictors, the logistic regression iden-

tified only diaphragm disease (risk ratio [RR], 5.69; P .01) andlarge

bowel mesentery implants (RR, 6.07; P .011) as significant predic-

tors of suboptimal cytoreduction. While ASA status and omental

extension to stomach and spleen demonstrated borderline signifi-

cance on univariateanalysis, theylost anytrend toward significance on

multivariate analysis. Using a model where only the presence of bothdiaphragm disease and large bowel mesentery implants is considered

predictive of suboptimal cytoreduction (predictive model A), the

nominalsensitivitywas 79%,the specificity was75%,and theaccuracy

was 77% in patient cohort A. These nominal sensitivity, specificity,

and accuracy rates were confirmed by a leave one out statistical vali-

dation. In this analysis, both diaphragm disease and large bowel mes-

entery implants continued to be theonlysignificant predictors.Across

the 65 runs the mean sensitivity for suboptimal cytoreduction was

78.6% (95% CI, 56.6% to 100%), the mean specificity was 74.5%

(95% CI, 62.5 to 86.5%), and the mean accuracy was 76.5% (95% CI,

51.5% to 100%).

Table 1. Clinical Data and Tumor Characteristics of Study Cohort A (N 65)

Characteristic

Patients

No. %

Age, years

50 17 26

50-59 18 27

60-69 19 29

70-79 9 14 80 2 4

FIGO stage

III A/B 4 6

III C 53 82

IV 8 12

Histologic subtype

Serous 52 80

Clear cell 6 9

Mixed 6 9

Endometrioid 1 1

Optimal cytoreduction 51 78

Suboptimal cytoreduction 14 22

Abbreviation: FIGO, International Federation of Gynecology and Obstetrics

staging system.

Axtell et al

386 JOURNAL OF CLINICAL ONCOLOGY



4/6

Those patients with no disease on the diaphragm or large bowelmesentery had the lowest rates of suboptimal cytoreduction (7%),

patients with one, but not both risk factors had intermediate rates of

suboptimal cytoreduction (42% to 44%), whereas those with both

diaphragm and large bowel mesentery implants on preoperative CT

had the highest rates of suboptimal cytoreduction (67%). Thus, pa-

tients with both CT predictors were 9.11 times more likely to be

suboptimally debulked than those with neither predictor (Table 3).

In order to study the applicability of our thus identified CT

predictors of suboptimal cytoreduction to other patient cohorts,

we applied predictive model A to patient cohorts B andC (Table4).

The resultant sensitivity of predictive model A fell to 15% and 72%,

specificity to 32% and 56%, and accuracy to 34% and 64% when

applied to patient cohorts B and C, respectively. For reciprocal cross

validation, we applied the CT predictors previously identified in pa-

tient cohorts B (predictive modelB) andC (predictive modelC) to the

present patient cohort A (Table 5). When CT predictor models B and

C were applied to patient cohort A, the original accuracy rates of 93%

and 79% fell to 74% and 48%, respectively. The CT predictors identi-

fied in three additional studies in the literature were not used as

primary outcome measure for cross validation as the original study

cohortsincludedearly-stageovariancancer.23,24,26However,whenthe

predictors from these three studies were applied to cohort A for com-

parison purposes, a similar loss in accuracywas observed from 79%to

88%intheoriginalcohortsto51%to62%incohortA.Inotherwords,

64% to 86% of patients in cohort A predicted to undergo suboptimalcytoreduction using any of the five published models were actually

optimally cytoreduced.

DISCUSSION

Our current study identifies diaphragm disease and large bowel

mesentery implants as being the only statistically significant pre-

dictors of suboptimal cytoreduction. Despite this statistical signif-

icance, the clinical relevance of this finding remains questionable

because approximately half of our own patients whose preopera-

tive CT scans were positive for either of these predictors actually

underwent optimal cytoreduction at the time of primary surgery.

Even when applying our multivariate model, which required the

presence of both, diaphragm disease and disease of the large bowel

mesentery on the preoperative CT scan in order to be predictive of

suboptimal cytoreduction, there was still a 33% false positive rate.

Despite optimizing our model, one of every three patients pre-

dicted to undergo suboptimal tumor debulking could actually be

left with minimal residual disease after operation with therapeutic

intent. If we had followed our CT predictor model, each of these

patients would have been deprived of the potential survival advan-

tage associated with optimal primary cytoreductive surgery.

As becomesevident fromthe reviewof allpublishedCT predictor

data to date, each of the retrospective studies,22,24-27

including ourown, identified a different set of predictors of cytoreductive surgery

outcome, raising the question as to their applicability to patient

Table 2. Univariate Analysis of Computed Tomography Predictors of Suboptimal Cytoreduction

Predictor

Patients

With Variable Present With Variable Absent

PNo. Suboptimal (%) No. Suboptimal (%)

Diaphragm disease 2 cm 15 46 50 14 .02

Large bowel mesentery implants 2 cm 12 50 53 14 .02

Omental extension to stomach or spleen 23 35 42 14 .07

Omental cake 46 16 19 11 .20

Small bowel mesentery implants 2 cm 11 36 54 19 .23

Suprarenal lymph nodes 1 cm 5 0 60 33 .35

Porta hepatis or gall bladder fossa disease 6 33 59 20 .60

Diffuse peritoneal thickening 12 17 53 33 .73

Pleural effusion 28 25 37 19 .76

Large volume ascites 40 20 25 24 .76

Peritoneal implants 2 cm 24 21 41 22 .99

Liver implants 8 25 57 21 .99

Inguinal canal disease 2 0 63 22 .99

Infrarenal lymph nodes 2 cm 2 0 63 22 .99

Table 3. Multivariate Risk Factors for Suboptimal Cytoreduction

DiaphragmLarge Bowel

Mesentery

Suboptimal

Risk RatioNo. %

No No 41 7 1.00

No Yes 9 44 6.07

Yes No 12 42 5.69

Yes Yes 3 67 9.11

Using absence of diaphragm and large bowel mesentery disease asthe referent.

Table 4. Validation Set for the Prediction of Suboptimal Cytoreduction(predictive model A applied to patient cohorts A, B, and C)

Patient Cohort Sensitivity (%) Specificity (%) Accuracy (%)

A 79 75 77

B 15 32 34

C 72 56 64


www.jco.org 387



5/6

cohorts other than the one they were developed in. The lack of gener-

alizability of any of the CT predictive models identified to date is

illustratedby our cross validation studies. Reciprocal accuracy rates inthe validation cohorts were poor across all models and cohorts stud-

ied. Most importantly, there was an unacceptable overprediction of

suboptimal cytoreduction in the reciprocal cross validation scenarios

of 64% to 86%. Which is to say that 64% to 86% of those patients

predicted in the cross validation cohorts to undergo suboptimal cy-

toreduction using any of the published prediction models actually

underwent optimal primary tumor cytoreduction, leaving the patient

withminimal residualdisease. Thishighrateof overcallof suboptimal

cytoreduction by CT predictors becomes particularly concerning

in light of the well-documented survival advantage associated with

intraperitoneal chemotherapy in patients with optimally cytore-

duced cancer,5,28,29 especially when juxtaposed to data from a

recent meta-analysis of neoadjuvant chemotherapy which suggests

a 4.1-month reduction in survival for each cycle of chemotherapy

given before surgical tumor debulking.30

Oneof theprincipledifficulties inthedevelopment of anyreliable

predictivemodel of surgicaloutcome forpatientswith advancedovar-

iancancer isthechallengeof factoring in the significant impact of each

individual surgeonsphilosophy, effort, and ability to utilizeadvanced

surgical techniques to achieve maximal cytoreduction.31 Subset anal-

ysis of the suboptimally debulked patients by individual surgeon was

not possible in our study due to the low number of patients with

suboptimal cytoreduction for each of the 12 surgeons in institution

cluster A. However, an overall optimal cytoreduction rate of 78% in

patient cohort A is an attestation to the surgical practice at Universityof California, Los Angeles and affiliated teaching institutions that is

characterized by a strong commitment to optimalprimary cytoreduc-

tionand theinclusionof advancedsurgical techniquesin thesurgeons

armamentarium to achieve resection of the tumor.

In summary, identification of risk factors for suboptimal cy-

toreduction in small retrospective populations such as ours and all

previously published cohorts, are not reproducible in alternate

populations. Until prospective, randomized trials have demonstrated

that neoadjuvant chemotherapy followed by interval cytoreduction is

equivalent in terms of survival outcomes to primary optimal cyto-

reduction followedby chemotherapy, extremecautionshould be used

when applying preoperative imaging predictors to decide between

primary surgical exploration and neoadjuvant chemotherapy in the

medically fit patient. Only the patient who is the most unlikely toundergo optimalcytoreduction should be offered neoadjuvantchem-

otherapy, unless her medical condition renders her unsuitable for

primary surgery. Ultimately, only a multi-institutional prospective

trial would answer the question of whether or not an accurate and

reproducible preoperative model using CT or other imaging modali-

ties could be developed for the prediction of surgical outcome. Such a

model would have to take into consideration the impact of variable

surgical practices and the surgeons philosophical commitment to

aggressive tumor debulking and skills in advanced cytoreductive sur-

gical techniques.

AUTHORS DISCLOSURES OF POTENTIAL CONFLICTS

OF INTEREST

The authors indicated no potential conflicts of interest.

AUTHOR CONTRIBUTIONS

Conception and design: Allison E. Axtell, Margaret H. Lee, Christine H.

HolschneiderAdministrative support: Allison E. Axtell, Christine H. Holschneider

Provision of study materials or patients: Robert E. Bristow, Sean C.

Dowdy, William A. Cliby, Scott Lentz, Andrew J. Li, Beth Y. Karlan,

Christine H. Holschneider

Collection and assembly of data: Allison E. Axtell, Margaret H. Lee,Steven Raman, John P. Weaver, Mojan Gabbay, Michael Ngo, Ilana Cass,

Christine H. Holschneider

Data analysis and interpretation: Allison E. Axtell, Robert E. Bristow,Sean C. Dowdy, William A. Cliby, Steven Raman, John P. Weaver, Scott

Lentz, Ilana Cass, Andrew J. Li, Beth Y. Karlan, Christine H.

Holschneider

Manuscript writing: Allison E. Axtell, Robert E. Bristow, Sean C.Dowdy, William A. Cliby, Christine H. Holschneider

Final approval of manuscript: Allison E. Axtell, Margaret H. Lee, Robert

E. Bristow, Sean C. Dowdy, William A. Cliby, Steven Raman, John P.

Weaver, Mojan Gabbay, Michael Ngo, Scott Lentz, Ilana Cass, Andrew J.Li, Beth Y. Karlan, Christine H. Holschneider

Table 5. Cross Validation Set for the Prediction of Suboptimal Cytoreduction (predictive models B and C applied to patient cohort A) and ComparativeApplication of the Three Other Published CT Predictor Models to Patient Cohort A for the Prediction of Suboptimal Cytoreduction

Study and Model

Original Cohort (%) Cohort A (%)

Sensitivity Specificity Accuracy Sensitivity Specificity Accuracy

Cross validation set

Bristow26 (model B) 100 85 93 93 55 74

Dowdy22 (model C) 52 90 71 7 88 48

Comparative application of other published CTpredictor models

Nelson24 92 79 86 79 45 62

Meyer25 58 100 79 57 45 51

Qayyum27 76 99 88 50 65 58

NOTE. Number of patients per cohort: model A, N 65; model B, N 41; model C, N 87; Nelson: N 42; Meyer: N 28; Qayyum: N 137.Abbreviation: CT, computed tomography.

Includes performance status in predictive index score.All include patients with early-stage disease.Includes CT and magnetic resonance imaging.

Axtell et al

388 JOURNAL OF CLINICAL ONCOLOGY



6/6

REFERENCES

1. American Cancer Society. Cancer Facts &

Figures 2005. http://www.cancer.org/downloads/

STT/CAFF2005f4PWSecured.pdf

2. Bristow RE, Tomacruz RS, Armstrong DK, et al:

Survival effect of maximal cytoreductive surgery for

advanced ovarian carcinomaduring theplatinum era: A

meta-analysis. J Clin Oncol 20:1248-1259, 2002

3. McGuire WP, Hoskins WJ, Brady MF, et al:Cyclophosphamide and cisplatin compared with

paclitaxel and cisplatin in patients with stage III and

stage IV ovarian cancer. N Engl J Med 334:1-6, 1996

4. Ozols RF, Bundy BN, Greer BE, et al: Phase III

trial of carboplatin and paclitaxel compared with

cisplatin and paclitaxel in patients with optimally

resected stage III ovarian cancer: A Gynecologic

Oncology Group study. J Clin Oncol 27:3194-3200,

2003

5. Armstrong DK, Bundy B, Wenzel L, et al:

Intraperitoneal cisplatin and paclitaxel in ovarian can-

cer. N Engl J Med 354:34-43, 2006

6. Holschneider CH, Berek JS: Cytoreductive

surgery: Principles and rationale, in Bristow RE,

Karlan BY, Montz FJ (eds): Surgery for Ovarian

Cancer: Principles and Practice. Lancaster, United

Kingdom, The Parthenon Publishing Group, 2005, pp

141-143

7. Hoskins WJ, McGuire WP, Brady MF, et al:

The effect of diameter of largest residual disease on

survival after primary cytoreductive surgery in pa-

tients with suboptimal residual epithelial ovarian

carcinoma. Am J Obstet Gynecol 170:974-979, 1994

8. Schwartz PE, Rutherford TJ, Chambers JT, et

al: Neoadjuvant chemotherapy for advanced ovarian

cancer: Long term survival. Gynecol Oncol 72:93-99,

1999

9. Kayikciog F, Rose MF, Boran N, et al: Neoad-

juvant chemotherapy or primary surgery in advanced

epithelial ovarian cancer. Int J Gynecol Cancer 11:

466-470, 2001

10. Fanfani F, Ferrandina G, Corrado G, et al:

Impact of interval debulking surgery on clinical out-

come in primary unresectable FIGO stage IIIC ovar-

ian cancer patients. Oncology 65:316-322, 2003

11. Morice P, Dubernard G, Rey A, et al: Results

of interval debulking surgery compared with primary

debulking surgery in advanced stage ovarian cancer.

J Am Coll Surg 197:955-963, 2003

12. Shibata K, Kikkawa F, Mika M, et al: Neoadju-

vant chemotherapy for FIGO stage III or IV ovarian

cancer: Survival benefit and prognostic factors. Int J

Gynecol Cancer 13:587-592, 2003

13. Ansquer Y, LeBlanc E, Clough K, et al: Neo-

adjuvant chemotherapy for nonresectable ovarian

carcinoma: A French multicenter study. Cancer 91:

2329-2334, 2001

14. Kuhn W, Rutke S, Spathe K, et al: Neoadju-vant chemotherapy followed by tumor debulking

prolongs survival for patients with poor prognosis in

International Federation of Gynecology and Obstet-

rics stage IIIC ovarian carcinoma. Cancer 92:2585-

2591, 2001

15. Vergote I, DeWever I, Tjalma W, et al: Neoad-

juvant chemotherapy or primary debulking surgery

in advanced ovarian carcinoma: A retrospective anal-

ysis of 285 patients. Gynecol Oncol 71:431-436,

1998

16. European Organization for Research and Treat-

ment of Cancer: Randomized phase III study compar-

ing upfront debulking surgery versus neo-adjuvant

chemotherapyin patients with stage IIIc or IV epithelial

ovarian cancer (EORTC protocol 55971). http://www

.eortc.be/protoc/Details.asp?Protocol5597117. Chi DS, Venkatraman ES, Masson V, et al: The

ability of preoperative serum CA-125 to predict

optimal primary tumor cytoreduction in stage III

epithelial ovarian carcinoma. Gynecol Oncol 77:227-

231, 2000

18. Gemer O, Segal S, Kopmar A, et al: Preoper-

ative CA-125 level as a predictor of non-optimal

cytoreduction of advanced epithelial ovarian cancer.

Acta Obstet Gynecol Scand 80:583-585, 2001

19. Saygili U, Guclu S, Uslu T, et al: Can serum

CA-125 levels predict the optimal primary cytoreduc-

tion in patients with advanced ovarian carcinoma?

Gynecol Oncol 86:57-61, 2002

20. Cooper BC, Sood AK, Davis CS, et al: Preop-

erative CA-125 levels: An independent prognostic

factor for epithelial ovarian cancer. Obstet Gynecol100:59-64, 2002

21. Memarzadeh S, Lee SB, Berek JS, et al:

CA-125 levels are a weak predictor of optimal

cytoreductive surgery in patients with advanced

epithelial ovarian cancer. Int J Gynecol Cancer

13:120-124, 2003

22. Dowdy SC, Mullany SA, Brandt KR, et al: The

utility of computed tomography scans in predicting

suboptimal cytoreductive surgery in women with

advanced ovarian carcinoma. Cancer 101:346-352,

2004

23. Brockbank EL, Ind TE, Barton DP, et al: Pre-

operative predictors of suboptimal primary surgical

cytoreduction in women with clinical evidence of

advanced primary epithelial ovarian cancer. Int J

Gynecol Cancer 14:42-50, 2004

24. Nelson BE, Rosenfield AT, Schwartz PE: Pre-operative abdominopelvic computed tomographic

prediction of optimal cytoreduction in epithelial ovar-

ian carcinoma. J Clin Oncol 11:166-172, 1993

25. Meyer JI, Kennedy AW, Friedman R, et al:

Ovarian carcinoma: Value of CT in predicting suc-

cess of debulking surgery. Am J Roentgenol 165:

875-878, 1995

26. Bristow RE, Duska LR, Lambrou NC, et al: A

model for predicting surgical outcome in patients

with advanced ovarian carcinoma using computed

tomography. Cancer 89:1532-1540, 2000

27. Qayyum A, Coakley FV, Westphalan AC, et

al: Role of CT and MRI imaging in predicting

optimal cytoreduction of newly diagnosed primary

epithelial ovarian cancer. Gynecol Oncol 96:301-

306, 200528. Alberts DS, Liu PY, Hannigan EV, et al: Intra-

peritoneal cisplatin plus intravenous cyclophospha-

mide versus intravenous cisplatin plus intravenous

cyclophosphamide for stage III ovarian cancer.

N Engl J Med 335:1950-1955, 1996

29. Markman M, Bundy BN, Alberts DS, et al:

Phase III trial of standard-dose intravenous cisplatin

plus paclitaxel versus moderately high-dose carbo-

platin followed by intravenous paclitaxel and intra-

peritoneal cisplatin in small-volume stage III ovarian

carcinoma: An intergroup study of the Gynecologic

Oncology Group, Southwestern Oncology Group,

and Eastern Cooperative Oncology Group. J Clin

Oncol 19:1001-1007, 2001

30. Bristow RE, Chi DS: Platinum-based neoadju-

vant chemotherapy and interval cytoreduction foradvanced ovarian cancer: A meta-analysis. Gynecol

Oncol 103:1070-1076, 2006

31. Aletti GD, Gostout BS, Podratz KC, et al:

Ovarian cancer surgical resectability: Relative im-

pact of disease, patient status, and surgeon. Gy-

necol Oncol 100:33-37, 2006

Acknowledgment

We thank Jeffrey Gornbein, PhD, Department of Biostatistics, David Geffen School of Medicine at University of California, Los Angeles,

for assistance with the statistical analysis.


www.jco.org 389


ireeecabilidad or tac

Documents