predicting risk of_malignancy_in_adnexal_masses.4

Original Research

Predicting Risk of Malignancy in AdnexalMassesJohn M. McDonald, MD, Stacey Doran, BS, Christopher P. DeSimone, MD, Fred R. Ueland, MD,Paul D. DePriest, MD, Rachel A. Ware, MD, Brook A. Saunders, MD, Edward J. Pavlik, PhD,Scott Goodrich, MD, Richard J. Kryscio, PhD, and John R. van Nagell Jr, MD

OBJECTIVE: To estimate the accuracy of preoperativeultrasonography, serum CA 125, and patient demograph-ics as a means of predicting risk of malignancy in womenwith a ultrasonographically confirmed adnexal mass.

METHODS: Tumor morphology derived from ultrasono-graphic images, tumor size, tumor bilaterality, serum CA125, and patient demographics were evaluated preoper-atively in 395 patients undergoing surgery from 2001 to2008. Tumor morphology was classified as complex,solid, or cystic. Preoperative findings were comparedwith tumor histologic findings at the time of surgery.Multivariable classification and regression tree analysiswere used to identify a group of patients at high risk ofovarian malignancy.

RESULTS: One hundred eighteen patients had ovariancancer, 13 patients had ovarian tumors of borderlinemalignancy, and 264 had benign ovarian tumors. Multi-variable classification and regression tree analysis definedwomen at high risk of ovarian malignancy as those withan adnexal mass having complex or solid morphologyand a serum CA 125 value greater than 35 units/mL. Thisdefinition had a positive predictive value of 84.7% and anegative predictive value of 92.4% and correctly identi-fied 77.3% of patients with stage I and stage II ovarian

cancer and 98.6% of patients with stage III and stage IVovarian cancer.

CONCLUSION: Patients with solid or complex ovariantumors and an elevated serum CA 125 level (greater than35 units/mL) are at high risk of ovarian malignancy.(Obstet Gynecol 2010;115:687–94)

LEVEL OF EVIDENCE: II

Ovarian cancer remains the leading cause ofdeath from gynecologic cancers in the United

States. It was estimated that in 2009 there would be21,554 new cases of ovarian cancer, and that 14,600women would die as a result of disease.1 Case–controlstudies have indicated that women with ovarian can-cer commonly experience a pattern of symptoms thatinclude bloating, pelvic/abdominal pain, difficultyeating/feeling full quickly, and urinary urgency orfrequency.2,3 These symptoms were found to be morecommonly associated with ovarian cancer, when theywere newly experienced, and occurred more that 12times per month.4 Recently, consensus groups haverecommended that women who experience symp-toms suggestive of ovarian cancer should undergo acomplete physical examination, and in certain cases,transvaginal ultrasonography and CA 125 testing.Although the majority of patients with these symp-toms will not have ovarian cancer, those who do willrequire complete surgical staging and aggressive tu-mor debulking to maximize their chances of surviv-al.5–7 In this regard, it is important to establish riskprofiles of patients with ultrasonographically con-firmed adnexal tumors so that they can receive ap-propriate treatment and, when necessary, referral forspecialty cancer care. The authors postulate that theaddition of ultrasonographically generated tumormorphology to patient demographics and serum bi-omarker profiles could improve prediction of malig-nancy in a clinically detectable adnexal mass. This

See related editorial on page 680.

From the Division of Gynecologic Oncology, Department of Obstetrics andGynecology, and Departments of Statistics and Biostatistics, University ofKentucky Chandler Medical Center–Markey Cancer Center, Lexington,Kentucky.

Supported by research grants from the Kentucky Department of Health andHuman Services and the Telford Foundation.

Corresponding author: John Rensselaer van Nagell Jr, MD, Division ofGynecologic Oncology, Department of Obstetrics and Gynecology, Universityof Kentucky Medical Center, 800 Rose Street, Lexington, KY 40536; e-mail:[email protected].

Financial DisclosureThe authors did not report any potential conflicts of interest.

© 2010 by The American College of Obstetricians and Gynecologists. Publishedby Lippincott Williams & Wilkins.ISSN: 0029-7844/10

VOL. 115, NO. 4, APRIL 2010 OBSTETRICS & GYNECOLOGY 687

investigation was undertaken to estimate the accuracyof a combination of patient demographics, ultrasono-graphically generated tumor morphology, and serumCA 125 values in predicting risk of malignancy inadnexal masses.

MATERIALS AND METHODSThis investigation was undertaken after approvalfrom the University of Kentucky Human SubjectsInstitutional Review Board. Study participants werewomen referred to the University of Kentucky–Mar-key Women’s Cancer Center with a diagnosis of anadnexal mass on pelvic examination who underwentsurgery at this institution from 2001 to 2008.

The following demographic data were obtainedfor all study patients: age, height, weight, gravidity,family history of breast or ovarian cancer, personalhistory of cancer, and menopausal status. Postmeno-pausal was defined as the absence of menses for aminimum of 12 months. Family history was consid-ered positive if the patient had a first-degree relative(ie, mother, sister, or daughter) or a second-degreerelative (ie, grandmother, granddaughter, aunt, orniece) with documented ovarian or breast cancer. Allwomen underwent pelvic examination, transvaginalultrasonography, and serum CA 125 determinationwithin 2 weeks before surgery. Transvaginal ultra-sonography was performed using General Electric(Milwaukee, WI) Logic 400 or Voluson ProV ultra-sound units with a 5-mHz vaginal probe as describedpreviously.8 Tumor dimensions from ultrasono-graphic images were recorded, and tumor morphol-ogy was classified as cystic, solid, or complex (con-taining both solid and cystic components). Allultrasonograms were reviewed by at least one of theauthors. All tumors classified as cystic were unilocu-lar, whereas cystic tumors with septations were in-cluded in the complex group. Ascites was defined asfree fluid more than 60 mL in the abdomen/pelvisconfirmed by ultrasonography. Serum CA 125 deter-minations were performed using a two-site sandwichparamagnetic particle chemiluminescent immunoen-zymatic assay with a normal value less than 35units/mL. Serum samples with values exceeding5,000 units/mL were diluted to end point for a finalresult.

After surgical removal, the dimensions of eachtumor were recorded, and frozen section histologicevaluation was performed. Tumors were classifiedhistologically according to the World Health Organi-zation system. Patients with a histologic diagnosis ofovarian malignancy underwent immediate tumor de-bulking and surgical staging according to the Interna-

tional Federation of Gynecology and Obstetrics sys-tem. Data were entered into a MEDLOG database(MEDLOG Systems, Crystal Bay, NV) and exportedinto an Excel (Microsoft Corp., Redmond, WA)spreadsheet for analysis using WinSTAT (R. FitchSoftware, Bad Krozingen, Germany), SPSS (SPSS,Inc., Chicago, IL), and PC-SAS with the EnterpriseMiner statistical software (SAS Institute, Inc., Cary, NC).

Proportions were compared using �2 statistics orFisher exact tests. Means were compared using two-sample t tests. Statistical significance was determinedat the .05 level. Multivariable analyses used theclassification and regression tree procedure, a non-parametric method that defines or accepts cut pointsand uses training and validation sets to optimize rulesfor the analysis.9 The training set is formed by split-ting the entire sample in half after stratification formalignant cases and benign controls.

RESULTSBetween July 2001 and December 2008, 399 patientsreferred to the outpatient clinic of the University ofKentucky–Markey Women’s Cancer Center for eval-uation of an adnexal mass on pelvic examination

Table 1. Patient Demographics, Tumor Variables,and Biomarker Profiles in PatientsStudied (N�395)

Mean Range

Age (y) 51.6 (�0.8) 10–86Height (in) 64 (�0.02) 55–72Weight (lb) 173 (�2.8) 78–384Gravidity 2.4 (�0.1) 0–21Family history

Ovarian cancer 48 (12)Breast cancer 64 (16)

Menopausal statusPremenopausal 176 (45)Postmenopausal 219 (55)

Tumor morphologyCystic 123 (31)Complex 236 (60)Solid 36 (9)

Tumor diameter10 cm or less 291 (74)More than 10 cm 104 (26)

AscitesPresent 54 (14)Absent 341 (86)

CA 125 (units/mL)Less than 35 247 (62)35–59 38 (10)60–120 23 (6)More than 120 87 (22)

Data are mean (�standard deviation) or n (%).

688 McDonald et al Risk of Malignancy in an Adnexal Mass OBSTETRICS & GYNECOLOGY

were included in this investigation. Four patients hadtheir first CA 125 determination after surgery andwere excluded from further evaluation. Demographicdata, biomarker profiles, and tumor characteristics ofthe patients evaluated are listed in Table 1. Fifty-fivepercent of patients were postmenopausal and 40%were aged 55 years or older. Sixty-four patients (16%)had a family history of breast cancer, 48 patients(12%) had a family history of ovarian cancer, and onepatient was BRCA1 positive. Two hundred thirty-sixmasses (60%) were ultrasonographically complex(Fig. 1A), 123 (31%) were cystic (Fig. 1B), and 36 (9%)were solid (Fig. 1C). Five of the 236 complex adnexalmasses (2.1%) were cystic ovarian tumors with thicksepta but no solid areas. Radiologic evidence ofascites was present in 54 patients (14%). Serum CA125 level was elevated (more than 35 units/mL) in148 patients (38%) and was more than 120 units/mLin 87 patients (22%).

At the time of surgery, 264 patients (67%) werefound to have benign ovarian tumors, 118 patients(30%) had ovarian cancer, and 13 patients (3%) hadovarian tumors of borderline malignancy. The stagedistribution of the patients with ovarian tumors ofborderline malignancy and ovarian cancer was asfollows: stage I, n�38; stage II, n�17; stage III,n�74; and stage IV, n�2. The most common celltypes of ovarian malignancy were adenocarcinoma,

followed by serous cystadenocarcinoma, mucinouscystadenocarcinoma, and clear-cell carcinoma.

The relationship of demographic, biomarker, andtumor variables to risk of malignancy in patients withan adnexal mass is presented in Table 2. Variablesstatistically related to risk of malignancy were tumormorphology, ascites, serum CA 125 level, tumor size,tumor bilaterality, menopausal status, and age.

Tumor morphology from ultrasonographicallygenerated images was related directly to risk of ma-lignancy. There were 236 complex adnexal masses,and 120 (51%) were malignant. None of the fivecomplex masses with septal morphology without solidareas were malignant. There were 36 solid adnexalmasses, and 11 (32%) were malignant. In contrast,there were 123 cystic adnexal masses, and none wereovarian tumors of borderline malignancy or invasiveovarian cancers (P�.001). The finding of purely cysticmorphology in an adnexal mass was associated with anegative predictive value (NPV) for malignancy of 100%.

Fifty-four patients with an adnexal mass hadradiologically confirmed ascites, and all of these pa-tients had invasive epithelial ovarian cancer (stage IC,n�2; stage IIC, n�1; stage IIIC, n�49; and stage IV,n�2). Therefore, the finding of documented ascitesin a patient with a complex or solid adnexal masshad a positive predictive value (PPV) for malig-nancy of 100%.

Fig. 1. Patterns of adnexal mor-phology. A. Complex adnexalmass, 12 cm in diameter; the cys-tic periphery is marked with cleararrows and the internal solidcomponent is marked with solidarrows (histology: clear-cell carci-noma). B. Cystic adnexal mass,9.4 cm in diameter, periphery ismarked with clear arrows (histol-ogy: ovarian serous cystade-noma). C. Solid adnexal mass, 5cm in diameter; the periphery ismarked with clear arrows (histol-ogy: ovarian adenocarcinoma).McDonald. Risk of Malignancy in anAdnexal Mass. Obstet Gynecol2010.

VOL. 115, NO. 4, APRIL 2010 McDonald et al Risk of Malignancy in an Adnexal Mass 689

Serum CA 125 values were related directly to riskof malignancy in women with an adnexal mass. Only19 (7.7%) of 247 patients with a normal serum CA 125value (less than 35 units/mL) had ovarian cancer.Conversely, 13 of 83 patients (34.2%) with a serumCA 125 value of 35–59 units/mL, 17 of 23 patients(73.9%) with a serum CA 125 value of 60–120units/mL, and 82 of 87 patients (86.8%) with a CA 125value more than 120 units/mL had borderline ormalignant ovarian tumors (P�.001). Serum CA 125values related to specific histologic diagnoses in all

patients with complex and solid adnexal masses arelisted in Table 3. The only benign histologic findingconsistently associated with an elevated serum CA125 value was ovarian endometriosis. Thirteen (48%)of 27 women with an ovarian endometrioma had anelevated serum CA 125 value (more than 35 units/mL), and 5 (18%) had a serum CA 125 value morethan 60 units/mL. There were 114 women with otherbenign complex or solid adnexal masses, and only 1patient (with an ovarian fibroma) had a CA 125 levelmore than 60 units/mL. As expected, serum CA 125values were related directly to stage of disease inpatients with ovarian malignancies. Serum CA 125values were elevated (more than 35 units/mL) in30.7% of patients with an ovarian tumor of borderlinemalignancy, 77.2% of patients with stage I and stage IIovarian cancer, and 98.6% of patients with stage IIICand IV ovarian cancer. All 54 patients with asciteshad an elevated (more than 35 units/mL) serum CA125 level, and 52 (96.2%) had a CA 125 level morethan 60 units/mL.

Risk of malignancy in an adnexal mass was signif-icantly higher in women older than 55 years than inyounger women (P�.001), in postmenopausal womencompared with premenopausal women (P�.001), inwomen with bilateral compared with unilateral ovariantumors (P�.01), and in women whose adnexal masseswere more than 10 cm in diameter compared withsmaller tumors (P�.001). A family history of ovariancancer or breast cancer in a woman with an adnexalmass was not associated statistically with an increasedrisk of malignancy in the population studied.

Although many variables were correlated with ma-lignancy, only a small number had acceptable positiveor NPVs (Table 4). Classification and regression treemultivariable analysis considered age, gravidity, post-menopausal status, weight, family history of ovariancancer or breast cancer, tumor morphology, ascites,tumor bilaterality, maximum tumor diameter, and CA125 value. This analysis found the most accurate signif-icance of interactions to declare a high risk of malig-nancy if a patient had an adnexal mass with complex orsolid morphology and a serum CA 125 value more than35 units/mL. The statistics for the training and validationsets indicated uniformly high performances for sensitiv-ity, specificity, and predictive values across both thetraining and validation sets.

Statistical parameters associated with the high-riskdefinition are listed in Table 5. Using the stated high-riskparameters resulted in a sensitivity of 30.8% for ovariantumors of borderline malignancy, 77.3% for earlystage (I and II) ovarian cancer, and 98.6% for ad-vanced stage (III and IV) ovarian cancer. Increasing

Table 2. Demographic, Tumor, and BiomarkerVariables Related to Risk of Malignancy(N�395)

Variable Total Malignant P

Age (y)55 or younger 239 58 (24.3) �.001Older than 55 156 73 (46.8)

Gravidity1 or less 125 41 (32.8) .92More than 1 270 90 (33.3)

Menopausal statusPremenopausal 176 43 (24.4) �.001Postmenopausal 219 88 (40.2)

Weight (lb)200 or less 314 116 (36.9) �.02More than 200 81 15 (18.5)

Family history of ovariancancer

Yes 48 16 (33.3) .98No 347 115 (33.1)

Family history of breastcancer

Yes 64 20 (31.3) .72No 331 111 (33.5)

Tumor morphologyCystic 123 0 (0) �.001*Complex 236 120 (50.8)Solid 36 11 (30.6)

AscitesNegative 341 77 (22.6) �.001Positive 54 54 (100)

Tumor lateralityUnilateral 375 119 (22.6) �.01Bilateral 20 12 (60)

Tumor diameter10 cm or less 291 67 (23) �.001More than 10 cm 104 64 (61.5)

CA 125 (units/mL)Less than 35 247 19 (7.7) �.001†

35–59 38 13 (34.2)60–120 23 17 (73.9)More than 120 87 82 (94.2)

Data are n or n (%).* Cystic compared with complex or solid tumor morphology.† CA 125 less than 35 units/mL compared with 35–59, 60–120,

or more than 120 units/mL.


the cutoff value of CA 125 from 35 units/mL to 60units/mL and keeping the other parts of the definitionthe same increased specificity from 92.4% to 96.6%but lowered sensitivity from 84.7% to 80.9% andmissed 13 patients with stage I or stage II ovariancancer. Therefore, a cutoff value of 35 units/mL forCA 125 was used in the final high-risk definition. Thisdefinition correctly identified 34 of 44 patients withstage I and stage II ovarian cancer and 93 of 94patients with stage III and stage IV ovarian cancer.Also, it correctly excluded 244 of the 264 patientswith benign ovarian tumors.

DISCUSSIONThe observation that the occurrence of certain symp-toms may precede the clinical diagnosis of ovariancancer has resulted in the recommendation thatwomen experiencing the recent onset of bloating,pelvic/abdominal pain, feeling full quickly after eat-ing, or urinary urgency/frequency should consult aphysician and undergo a complete physical examina-tion. Women having a clinically palpable abnormalityin the pelvis or those with persisting symptoms in thepresence of a normal pelvic examination are advisedto undergo transvaginal ultrasonography and CA 125testing.

The findings of this investigation indicate thatanalysis of data concerning patient demographics,

tumor morphology obtained from ultrasonographicimages, and serum CA 125 levels is useful in estimat-ing the risk of malignancy in women with an adnexalmass. For example, the risk of neoplasia in unilocularcystic ovarian tumors is very low. This morphologicpattern was present in 123 (31%) of 395 adnexaltumors, and no patient had either a borderline or aninvasive ovarian malignancy. This confirms the ob-servation by Roman and colleagues10 who, in a sum-mary of the literature, reported a 0.7% rate of malig-nancy in 569 unilocular cystic ovarian tumors 10 cmor less in diameter. Similarly, Modesitt and col-leagues11 followed more than 3,200 women withunilocular cystic ovarian tumors less than 10 cm indiameter for an average of 6 years without operativeintervention. No patient developed ovarian cancer,and 69% of these tumors resolved spontaneously overthe period of observation. There is no doubt thatsome unilocular cystic ovarian tumors grow to signif-icant size and require surgical removal. However, therisk of malignancy even in larger cystic lesions is low,particularly in women with a normal CA 125 level. Inthe present study, there were 27 unilocular cysticovarian tumors more than 10 cm diameter, and allwere benign.

In contrast, adnexal masses with complex or solidmorphology are associated with a significant risk ofmalignancy. There were 272 patients with a complex

Table 3. CA 125 Related to Histology in Ultrasonographically Complex or Solid Adnexal Tumors(n�272)

CA 125

Histology nLess Than

35 Units/mL35–59

Units/mL60–120

Units/mLMore Than

120 Units/mL

Fibroma 15 12 2 1 0Cystadenofibroma 27 25 2 0 0Endometrioma 27 14 8 1 4Serous cystadenoma 22 22 0 0 0Cystic teratoma 20 20 0 0 0Mucinous cystadenoma 15 15 0 0 0Granulosa cell tumor 4 3 1 0 0Pedunculated myoma 7 6 1 0 0Sertoli-Leydig cell tumor 1 0 1 0 0Brenner tumor 3 3 0 0 0Mucinous LMP 5 3 1 0 1Serous LMP 8 5 1 1 0Clear-cell carcinoma 8 3 2 3 0Carcinosarcoma 4 0 0 0 4Mucinous cystadenocarcinoma 5 2 0 1 1Serous cystadenocarcinoma 8 1 0 0 7Adenocarcinoma 93 5 8 12 68

LMP, low malignant potential.Data are n.


or solid adnexal mass, and 131 (48%) had an ovarianmalignancy. These women form the basis of a high-risk group for ovarian cancer. In the present investi-gation, documented ascites in a woman with a com-plex or solid adnexal mass was uniformly predictiveof ovarian cancer. There were 54 patients with thisfinding, and all had epithelial ovarian cancer. Theseresults are consistent with the findings of Im andcolleagues12 who, in a multiinstitutional review, re-ported that 79% of postmenopausal women with aclinically detectable pelvic mass and ascites had anovarian malignancy.

The use of serum CA 125 as a method ofpredicting risk of malignancy in patients with a clin-ically detectable pelvic mass was suggested in 2002 bya joint publication of the American College of Obste-

tricians and Gynecologists and the Society of Gyne-cologic Oncologists.13 This report stated that 69.8% ofpostmenopausal women with an adnexal mass and aserum CA 125 value more than 35 units/mL had anovarian malignancy and that an elevated serum CA125 value in a patient with a clinically detectablepelvic mass could be used as one indication forpatient referral for subspecialty care. In the presentinvestigation, more than three fourths of women witha complex or solid adnexal mass and a CA 125 valuemore than 35 units/mL had either borderline orinvasive ovarian cancer.

When evaluating a number of variables, includ-ing patient demographics, tumor morphology, andCA 125 levels, as predictors of malignancy, multiva-riable classification and regression tree analysis de-

Table 4. Sensitivity, Specificity, Positive and Negative Predictive Values, and 95% Confidence Limits ofEach Variable Identifying or Ruling Out Ovarian Malignancy

Variable Sensitivity

Sensitivity(95% CL)

Specificity

Specificity(95% CL)

Lower Upper Lower Upper

Age older than 55 y 55.7 47.2 64.2 68.6 63.0 74.2Gravidity more than 1 68.7 60.8 76.6 31.8 26.2 37.4Postmenopausal 67.2 59.1 75.2 50.4 44.3 56.4Weight more than 200 lb 11.5 6.0 16.9 75.0 69.8 80.2Family history

Ovarian cancer 12.2 6.6 17.8 87.9 83.9 91.8Breast cancer 15.3 9.1 21.4 83.3 78.8 87.8

Tumor morphologyCystic 100.0 97.7 100.0 46.6 40.6 52.6Complex 91.6 86.9 96.4 56.1 50.1 62.0

Ascites 41.2 32.8 49.7 100.0 98.9 100.0Tumor bilaterality 9.2 4.2 14.1 97.0 94.9 99.0Tumor diameter

More than 10 cm 48.9 40.3 57.4 84.8 80.5 89.2CA 125 level more than 35 units/mL 85.5 79.5 91.5 86.4 82.2 90.5

CL, confidence limit; PPV, positive predictive value; NPV, negative predictive value.

Table 5. Statistical Parameters Associated with a High-Risk* Group for Ovarian Cancer as Defined byMultivariable Classification and Regression Tree Analysis

Criteria TP FP TN FN PPV NPV Sensitivity† Specificity‡

Total malignancies (n�131) 111 20 244 20 0.847 0.924 0.847§ 0.924§

Borderline malignancy (n�13) 4 20 244 9 0.167 0.964 0.308 0.924Stages I and II ovarian

malignancy (n�44)34 20 244 10 0.630 0.961 0.773 0.924

Stages III and IV ovarianmalignancy (n�74)

73 20 244 1 0.785 0.996 0.986 0.924

TP, true positive; FP, false positive; TN, true negative; FN, false negative; PPV, positive predictive value (TP/TP�FP); NPV, negativepredictive value (TN/TN�FN).

* High-risk defined as a complex or solid adnexal mass with a CA 125 value more than 35 units/mL.† Sensitivity, (TP/TP�FN).‡ Specificity, (TN/TN�FN).§ These figures vary within 1 standard error (�3.1% for sensitivity and �1.6% for specificity) when the sample is randomly split into

training and validation sets.


fined high risk as women with a complex or solidadnexal mass and a serum CA 125 value of more than35 units/mL. In the population studied, this definitionof high risk was associated with a PPV of 84.7% anda NPV of 92.4% and correctly identified 34 (77.3%) of44 patients with stage I or stage II ovarian cancer aswell as 73 of 74 patients (98.6%) with stage III or stageIV ovarian cancer. Thus, including tumor morphol-ogy significantly increases predictive values beyondthe PPV of 24.3% for stage I or II ovarian cancer and56.8% for stage III or IV ovarian cancer associatedwith the original American College of Obstetriciansand Gynecologists/Society of Gynecologic Oncolo-gists high-risk criteria.14 In the present study, raisingthe cutoff value of CA 125 from 35 units/mL to 60units/mL in the definition of high risk would haveincreased specificity for identifying ovarian cancercases but would have lowered sensitivity and resultedin missing a significant number of patients withearly-stage disease. This is important because severalstudies have reported as high as a 24% survivaladvantage for patients with early-stage ovarian cancerwho received comprehensive surgical staging andtreatment by a gynecologic oncologist.15,16

Although family history of ovarian cancer or breastcancer was not statistically associated with an increasedrisk of ovarian cancer in this investigation, a hereditarycancer risk assessment was not routinely performed onevery patient enrolled in this study. Clearly, womenwith a strong family history of ovarian or breast canceror women who are members of populations with a highprevalence of “founder mutations” should undergo ge-

netic testing,17 and this information should be taken intoconsideration in determining optimal treatment for apatient with an adnexal mass.

As an increasing number of women who havesymptoms suggestive of ovarian cancer are evaluated,clinicians will be asked to determine which patientsare at significant risk for ovarian cancer. Data fromthe present investigation suggest that the combinationof ultrasonographic tumor morphology and serumCA 125 value improves the discrimination of womenat risk of ovarian cancer from those with benignadnexal lesions. These findings should be helpful indetermining which patients can be followed withoutsurgery, which patients are likely to have a benignovarian tumor, and which patients are at high risk ofovarian malignancy and should be referred for sub-specialty care.

REFERENCES1. Jemal A, Siegel R, Ward E, Hao Y, Xu J, Thun M. Cancer

statistics 2009. CA Cancer J Clin 2009;59:225–49.

2. Goff BA, Mandel LS, Melancon CH, Muntz HG. Frequency ofsymptoms of ovarian cancer in women presenting to primarycare clinics. JAMA 2004;291:2705–12.

3. Smith LH, Morris CR, Yasmeen S, Parikh-Patel A, Cress RD,Romano PS. Ovarian cancer: can we make the clinical diag-nosis earlier? Cancer 2005;104:1398–407.

4. Goff BA, Mandel L, Drescher CW, Urban N, Gough S,Schurman KM, et al. Development of an ovarian cancersymptom index. Cancer 2007;109:221–7.

5. Bristow RE, Tomacruz RS, Armstrong DK, Trimble EL,Montz FJ. Survival effect of maximal cytoreductive surgery foradvanced ovarian carcinoma during the platinum era: a meta-analysis. J Clin Oncol 2002;20:1248–59.

PPV

PPV(95% CL)

NPV

PPV(95% CL)

Lower Upper Lower Upper

46.8 39.0 54.6 75.7 70.3 81.233.3 27.7 39.0 67.2 59.0 75.440.2 33.7 46.7 75.6 69.2 81.918.5 10.1 27.0 63.1 57.7 68.4

33.3 20.0 46.7 66.9 61.9 71.831.3 19.9 42.6 66.5 61.4 71.6

48.2 42.2 54.1 100.0 97.6 100.050.8 44.5 57.2 93.1 89.1 97.0

100.0 94.4 100.0 77.4 73.0 81.960.0 38.5 81.5 68.3 63.6 73.0

61.5 52.2 70.9 77.0 72.1 81.875.7 68.8 82.6 92.3 89.0 95.6


6. Eisenhauer EL, Abu-Rustum NR, Sonoda Y, Aghajanian C,Barakat RR, Chi DS. The effect of maximal surgical cytore-duction on sensitivity to platinum-taxane chemotherapy andsubsequent survival in patients with advanced ovarian cancer.Gynecol Oncol 2008;108:276–81.

7. Eisenhauer EL, Abu-Rustum NR, Sonoda Y, Levine DA,Poynor EA, Aghajanian C, et al. The addition of extensiveupper abdominal surgery to achieve optimal cytoreductionimproves survival in patients with stages IIIC-IV epithelialovarian cancer. Gynecol Oncol 2006;103:1083–90.

8. van Nagell JR, DePriest PD, Ueland FR, DeSimone CP,Cooper AL, McDonald AM, et al. Ovarian cancer screeningwith annual transvaginal sonography: findings of 25,000women screened. Cancer 2007;109:1887–96.

9. Breiman L, Friedman J, Olshen R, Stone C. Classification andregression trees. New York (NY): Wadsworth; 1984.

10. Roman LD. Small cystic pelvic masses in older women: issurgical removal necessary? Gynecol Oncol 1998;69:1–2.

11. Modesitt SC, Pavlik EJ, Ueland FR, DePriest PD, van NagellJR Jr. Risk of malignancy in unilocular ovarian cystic tumorsless than 10 cm in diameter: a long-term follow-up study.Obstet Gynecol 2003;102:594–9.

12. Im S, Gordon A, Buttin B, Leath CA III, Gostout BS, Shah C,et al. Validation of referral guidelines for women with pelvicmasses. Obstet Gynecol 2005;105:34–41.

13. The role of the generalist obstetrician-gynecologist in the earlydetection of ovarian cancer. ACOG Committee Opinion No.280. American College of Obstetricians and Gynecologists.Obstet Gynecol 2002;100:1413–6.

14. Dearking AC, Aletti G, McGree M, Weaver AL, Sommer-field M, Cliby WA. How relevant are ACOG and SGOguidelines for referral of an adnexal mass. Obstet Gynecol2007;110:841– 8.

15. Le T, Adolph A, Krepart G, Lotocki R, Heywood MS. Thebenefits of comprehensive surgical staging in the managementof early-stage epithelial ovarian carcinoma. Gynecol Oncol1992;47:223–7.

16. Suh-Burgman E. Long-term outcomes following conservativesurgery for borderline tumor of the ovary: a large population-based study. Gynecol Oncol 2006;103:841–7.

17. Hereditary breast and ovarian cancer syndrome. ACOGCommittee on Practice Bulletins. Gynecol Oncol 2009;113:6 –11.


predicting risk of_malignancy_in_adnexal_masses.4

Health & Medicine