Women s Imaging Original Research

Women s Imaging Original Research Lakhman et al. Postoperative CT as Prognostic Biomarker in Gynecologic Cancer Women s Imaging Original Research Yulia Lakhman 1 Oguz Akin 1 Michael J. Sohn 1 Junting Zheng 2 Chaya S. Moskowitz 2 Revathy B. Iyer 3 Richard R. Barakat 4 Paul J. Sabbatini 5 Dennis S. Chi 4 Hedvig Hricak 1 Lakhman Y, Akin O, Sohn MJ, et al. Keywords: biomarkers, CT, operative, ovarian neoplasms, prognosis, surgical procedures DOI:1214/AJR.11.7257 Received May 20, 2011; accepted after revision December 4, 2011. 1 Department of Radiology, Memorial Sloan-Kettering Cancer Center, 1275 York Ave, Rm C-278, New York, NY 10065. Address correspondence to H. Hricak. 2 Department of Epidemiology-Biostatistics, Memorial Sloan-Kettering Cancer Center, New York, NY. 3 Department of Radiology, M. D. Anderson Cancer Center, Houston, TX. 4 Gynecologic Service, Department of Surgery, Memorial Sloan-Kettering Cancer Center, New York, NY. 5 Gynecologic Medical Oncology Service, Department of Medicine, Memorial Sloan-Kettering Cancer Center, New York, NY. AJR 2012; 198:1453 1459 0361 803X/12/1986 1453 American Roentgen Ray Society Early Postoperative CT as a Prognostic Biomarker in Patients With Advanced Ovarian, Tubal, and Primary Peritoneal Cancer Deemed Optimally Debulked at Primary Cytoreductive Surgery OBJECTIVE. The purpose of this article is to determine whether early postoperative CT provides prognostic information in patients with advanced ovarian, tubal, or primary peritoneal carcinoma with optimal debulking reported at primary cytoreduction. MATERIALS AND Methods. Our study included 63 patients who underwent primary cytoreductive surgery for presumed advanced ovarian cancer, who had optimal debulking (residual disease 1 cm) reported at surgery, and who underwent CT before and 7 49 days after surgery. Two radiologists independently retrospectively interpreted all postoperative CT scans and scored lesions on a 5-point scale, where 1 indicates normal and 5 indicates definitely malignant. Lesions larger than 1 cm with a CT score of 4 or 5 were considered suboptimally debulked residual disease. RESULTS. Suboptimally debulked residual disease on CT (range, 1.1 5.8 cm) was reported by reader 1 for 29 of 63 patients (46%) and by reader 2 for 31 of 63 patients (49%), with substantial interobserver agreement (κ = 0.75). Patients with suboptimally debulked residual disease on CT had significantly worse median progression-free survival (p = 01, both readers) and overall survival (p 10, both readers). By univariate and multivariate analyses, suboptimally debulked residual disease on CT remained a significant independent predictor of progressionfree survival (p = 01, both readers) and overall survival (p 06, both readers). CONCLUSION. Our study showed that residual disease larger than 1 cm was present on early postoperative CT in almost half of the patients deemed to have optimally debulked disease at primary cytoreduction. Residual disease larger than 1 cm detected on early postoperative CT was associated with significant decreases in both progression-free and overall survival. O varian cancer continues to be the leading cause of death from gynecologic cancer in the United States [1]. Primary cytoreductive surgery plays a well-established role in the treatment of patients with advanced ovarian cancer [2]. Currently, the Gynecologic Oncology Group (GOG) defines optimal cytoreduction as no residual tumor larger than 1 cm in the longest diameter after surgery [3, 4]. Distinguishing between optimal and suboptimal primary cytoreduction is important, because these designations have therapeutic and prognostic implications. Current data suggest that patients with optimally debulked tumors benefit from a combination of primary intraperitoneal and systemic chemotherapy, which offers a 16-month median survival advantage compared with systemic chemotherapy alone [4]. In contrast, patients with suboptimal debulking have a worse prognosis and are generally only offered systemic therapy. Historically, the determination of residual disease status has been based on the surgeon s intraoperative assessment, which includes both visual inspection and manual measurement of the largest remaining tumor at the conclusion of the surgery. However, such clinical assessment is difficult. When surgeons were asked to estimate the size of primary ovarian cancer and metastases in a simulated patient at laparotomy, significant measurement errors and high interobserver variability were found [5]. For a primary tumor of 13 cm, the surgeons estimates ranged from 4 to 20 cm, and for an 8.5-cm plaque in the right pelvic sidewall, their estimates ranged from 0 to 12 cm. More surgeons underestimated than overestimated tumor diameters. Despite its known limitations, intraoperative evaluation is not routinely substantiated by other methods such as cross-sectional imaging. The use of CT and MRI in preoperative staging of ovarian cancer and AJR:198, June 2012 1453

Lakhman et al. in predicting cytoreductive outcome is still evolving [6]. Little is known about the role of early postoperative CT in assessing residual disease status after primary debulking. However, prior research has suggested that residual disease status as assessed on early postoperative CT may have potential as a biomarker that could predict prognosis, guide treatment selection, and facilitate patient stratification for clinical trials [7, 8]. The purpose of our study was to determine the prognostic significance of residual disease status on early postoperative CT in patients with advanced ovarian, tubal, and primary peritoneal carcinoma for whom optimal debulking was reported at primary cytoreductive surgery. Our hypothesis was that the presence of residual disease larger than 1 cm on early postoperative CT, as determined with well-defined image interpretation criteria, would be associated with decreases in progression-free and overall survival. Materials and Methods Eligibility Our study was compliant with the HIPAA and was approved by our institutional review board. The study cohort was selected from the second part of a two-part institutional review board approved biinstitutional prospective clinical trial (clinicaltrials. gov identification no., NCT00587093). All patients 18 years old or older who were scheduled to undergo surgery by an attending gynecologic oncologist for presumed advanced ovarian cancer were eligible for inclusion in the first part of the trial, which aimed to assess the ability of preoperative cancer antigen 125 (CA-125) and CT to predict cytoreductive outcome [7, 8]. The aim of the second part of the trial was to compare surgeons intraoperative assessments of residual disease to residual disease findings on early postoperative CT in patients reported to have undergone optimal primary cytoreduction. The eligibility criteria for the second part were as follows: histologically confirmed stage III or IV ovarian, tubal, or primary peritoneal cancer; optimal cytoreduction (i.e., residual disease 1 cm) reported by the surgeon; preoperative CT performed 2 25 days before surgery; and postoperative CT performed 7 49 days after surgery and before initiation of chemotherapy. All patients enrolled in the trial gave informed consent for CT. The institutional review board granted a waiver of informed consent for retrospective review of the patients data and CT images. CT Technique All patients received oral (900 ml of dilute meglumine diatrizoate solution [Hypaque, Sanofi- Aventis]) and IV (150 ml of iohexol [Omnipaque 300, GE Healthcare]) contrast agents before the examination. CT scans were performed on helical scanners (GE LightSpeed, GE Healthcare). Images were reconstructed at 5- or 7.5-mm intervals and were sent to a PACS (Centricity, GE Healthcare). CT scans performed at outside institutions were digitized on a PACS, reviewed by one of the protocol radiologists, and included in the study only if they were judged to be of acceptable imaging quality. Image Analysis and Interpretation CT scans were independently and retrospectively interpreted by two radiologists with 3 and 6 years of experience in gynecologic cancer imaging. The readers were blinded to all clinical information other than the fact that all patients were deemed to have optimally debulked disease (i.e., residual disease 1 cm) at surgery. Qualitative Assessment Each reader assigned a qualitative assessment score for each lesion identified, to indicate the degree of certainty that it represented residual disease. The qualitative assessment scale was as follows: 1, definitely no residual tumor; 2, probably no residual tumor; 3, indeterminate; 4, probably residual tumor; and 5, definitely residual tumor. The readers agreed to assign scores of 4 or 5 if a lesion of similar attenuation and shape, as well as the same or smaller size, was present in an identical location on the preoperative CT scan; or if a lesion was only seen on postoperative CT but had well-defined borders, was either solid or mixed cystic and solid in composition, and was situated away from the surgical resection beds, surgical clips, or surgical anastomoses. Quantitative Assessment Both readers recorded bidimensional measurements and locations for all parenchymal and peritoneal lesions, as well as enlarged lymph nodes (defined by a short axis > 1 cm). The long-axis measurement was used for parenchymal and peritoneal tumor nodules; the short-axis measurement was used for lymph nodes. Any lesion larger than 1 cm with a score of 4 or 5 was considered suboptimally debulked residual disease on CT. Residual disease locations were tabulated as follows: perihepatic region (right subdiaphragmatic region, right and left perihepatic spaces, and Morison pouch), hepatic parenchyma, perisplenic region (left subphrenic region and anterior and posterior perisplenic spaces), splenic parenchyma, small bowel (small-bowel mesentery and serosa and lesser sac), large bowel (large-bowel mesentery and serosa and both paracolic gutters), pelvis, supradiaphragmatic and retrocrural adenopathy, upper abdominal adenopathy (gastrohepatic, porta hepatis, and suprarenal paraaortic regions), and lower abdominal and pelvic adenopathy (infrarenal paraaortic, iliac, and inguinal regions). Data Collection Demographic, clinical, surgical, and pathologic data were collected for all patients. Date of progression was determined by follow-up CT scan or CA- 125 level or both. When the date of progression was diagnosed on follow-up CT, it was defined as the first appearance of one or more new tumors or enlargement of existing lesions after completion of adjuvant chemotherapy [9]. When it was determined by CA-125 level, the date of progression was defined as the first date when the CA-125 level was at least twice its nadir value or twice the upper limit of normal [10]. When a subsequent CT scan confirmed that a single increase in CA-125 level indicated disease progression, date of progression was defined as the date of CA-125 increase. Progression-free survival was defined as the time interval between the date of surgery and the date of first documented recurrence or progression of disease. If there was no documented recurrence, progression-free survival was calculated as the time from the date of surgery to the date of last follow-up or death. Overall survival was defined as the time interval between the date of surgery and the date of death or last follow-up. Statistical Methods Interobserver agreement was analyzed with the Cohen kappa statistic, where a kappa value less than 0 indicates no agreement, 0 0 indicates slight agreement, 1 0 indicates fair agreement, 1 0 indicates moderate agreement, 1 0 indicates substantial agreement, and 1 indicates almost perfect agreement [11]. Univariate analyses were performed to analyze the effects of 10 potential prognostic factors on progression-free and overall survival [12, 13]. These 10 variables were age at the time of the surgery, presence of ascites, stage (III vs IV), grade (high vs other), histology (serous versus other), primary site of disease (ovary vs other), residual disease at surgery (none vs 0.1 0.5 cm vs 1 cm), days between surgery and postoperative CT, intraperitoneal chemotherapy as a part of initial treatment, and presence of residual disease larger than 1 cm on early postoperative CT. Age at surgery was analyzed as a binary variable either above or below the median age. Because all postoperative CT scans were performed within 49 days of surgery and the number of days between surgery and CT scan was one of 10 variables analyzed, a landmark analysis was used with a landmark time of day 50 after surgery. Progressionfree and overall survival were summarized using the Kaplan-Meier method starting from the date of surgery [14]. All 63 patients were included in the over- 1454 AJR:198, June 2012

Postoperative CT as Prognostic Biomarker in Gynecologic Cancer Enrolled Patients (n = 285) Exclusions (n = 222) Withdrew from the prospective study (n = 27) Did not have ovarian, tubal, or peritoneal cancer (n = 42) Did not have advanced-stage disease (n = 16) Underwent suboptimal cytoreduction (n = 48) Declined postoperative CT or did not have it within 7 49 days (n = 74) Missing either preoperative, postoperative, or both CT scans (n = 15) Study Population (n = 63) Memorial Sloan-Kettering Cancer Center (n = 60) M. D. Anderson Cancer Center (n = 3) all survival analysis. Two patients who experienced disease progression within 50 days after surgery were removed from the progression-free survival analysis. Therefore, 61 of 63 patients were included in the progression-free survival analysis. A Cox regression based on the landmark analysis was used to assess the association between survival (progression-free and overall survival) and days between surgery and postoperative CT. The log-rank test based on the landmark analysis was used to examine the univariate associations between survival and each of the other nine potential prognostic factors. The multivariate Cox proportional-hazards model based on the landmark analysis was applied to analyze the effect of residual disease larger than 1 cm on early postoperative CT on survival, controlling for the covariates that had p values less than 0.100 in the univariate analysis. The effects of residual disease larger than 1 cm on CT were analyzed separately for each reader. All progressionfree survival, overall survival, and Cox model results should be interpreted as quantities conditional on patients surviving at least 50 days after surgery. Results Patients Between January 2001 and September 2006, 285 patients were enrolled in the prospective clinical trial; of these patients, 78 were eligible for the second part of the trial (Fig. 1). Fifteen of the 78 patients were excluded from our retrospective study because preoperative or postoperative CT scans were not available for review (Fig. 1); thus, our study included a total of 63 patients. The patient and tumor characteristics are summarized in Table 1. Discrepancy Between Early Postoperative CT and Surgical Residual Disease Assessment The surgical and imaging characterizations of residual disease status are summarized in Fig. 1 Flowchart summarizing study cohort. Table 2. The surgeon reported no evidence of residual disease or microscopic residual disease (< 0.1 cm) in 27 of 63 (43%) cases, residual disease of 0.1 0.5 cm in 22 of 63 (35%) cases, and residual disease of 1 cm in 14 of 63 (22%) cases. Reader 1 disagreed with the surgical assessment of optimal debulking in 29 of 63 (46%) cases, and reader 2 disagreed with the surgical assessment in 31 of 63 (49%) cases. Substantial agreement was noted between the two radiologists for detecting residual disease larger than 1 cm on CT (κ = 0.75). In patients with residual disease larger than 1 cm on CT, the median lesion size was 1.8 cm (range, 1.1 5.8 cm) for both readers. Figure 2 provides examples of suboptimally debulked disease on CT. The locations of residual tumors larger than 1 cm on CT are summarized in Table 3. For both readers, the most common locations were the perihepatic region, large-bowel serosa, upper abdominal lymph nodes, supradiaphragmatic and retrocrural lymph nodes, lower abdominal or pelvic lymph nodes, and the perisplenic region. In 60 of 63 patients, all residual disease lesions were present on both preoperative and postoperative CT scans. In three patients, in addition to residual disease lesions that were present on preoperative CT, new metastatic lesions were detected on postoperative CT. In one of these three cases, the patient had hepatic metastases that increased in number as well as size, and in the other two, in addition to residual lymphadenopathy, newly developed upper abdominal lymphadenopathy was found. Survival Analysis At the time of the analysis, 41 of 63 patients were deceased. For reader 1, 23 of 29 patients with residual disease larger than 1 cm on CT and 18 of 34 patients without residual disease larger than 1 cm on CT were dead. For reader 2, 25 of 31 patients with residual disease larger than 1 cm on CT and 16 of 32 patients without residual disease larger than 1 cm on deceased. The median patient followup time was 75 months (range, 2 105 months; 95% CI, 63 85 months), with a median progression-free survival of 19 months (95% CI, 16 27 months) and a median overall survival of 51 months (95% CI, 37 70 months). The median progression-free survival times for patients with optimally and suboptimally debulked residual disease on CT were 29 months (95% CI, 19 65 months) and 16 months (95% CI, 10 24 months), respectively, for reader 1 (p = 01) and 36 months (95% CI, 19 66 months) and 16 months (95% CI, 12 19 months), respectively, for reader 2 (p = 01) (Fig. 3). The median overall survival times for patients with optimally and suboptimally debulked residual disease on CT were 70 months (95% CI, 38 months to not achieved) and 44 months (95% CI, 32 70 months), respective- TABLE 1: Demographic and Tumor Characteristics of 63 Patients Characteristic Value Age (y), median (range) 60 (39 81) Surgical stage IIIA 1 (2) IIIC 47 (74) IV 15 (24) Tumor grade 1 4 (6) 2 4 (6) 3 52 (83) Unknown 3 (5) Histology Serous 58 (92) Clear cell 1 (2) Mixed 4 (6) Primary disease site Ovary 50 (79) Peritoneum 8 (13) Fallopian tube 5 (8) Ascites No 9 (14) Yes 54 (86) Note Except where noted otherwise, data are no. (%) of patients. AJR:198, June 2012 1455

Lakhman et al. TABLE 2: Residual Disease in 63 Patients, According to Surgeon s Assessment and Postoperative CT Scans Assessment Residual Disease Surgeon Reader 1 Reader 2 Residual disease 1 cm 63 (100) 34 (54) 32 (51) No or microscopic 27 (43) 32 (51) 32 (51) 0.1 0.5 cm 22 (35) 0 0 cm 14 (22) 2 (3) 0 Residual disease > 1 cm 0 29 (46) 31 (49) 1.1 2.0 cm 0 13 (21) 12 (19) > 2.0 cm 0 16 (25) 19 (30) Note Data are no. (%) of patients. A C E B D F ly, for reader 1 (p = 10) and 77 months (95% CI, 49 months to not achieved) and 43 months (95% CI, 30 62 months), respectively, for reader 2 (p = 05) (Table 4; Fig. 3). No significant difference in median progression-free survival (reader 1, p = 23; reader 2, p = 09) or overall survival (reader 1, p = 0.999; reader 2, p = 0.912) was present between patients with residual disease of 1.1 2 cm and patients with residual disease larger than 2 cm on CT. The 10 potential predictors of progression-free and overall survival were analyzed (Tables 4 and 5). In the univariate analysis for progression-free survival, only residual disease larger than 1 cm on early postoperative CT was a statistically significant prognostic factor (p = 01, both readers). It retained its significance in the multivariate analysis (p = 01, both readers). In the univariate analysis for overall survival, the presence of residual disease larger than 1 cm on CT was also the only significant predictor (reader 1, p = 10; reader 2, p = 05). In the multivariate analysis, residual disease larger than 1 cm on CT remained an important predictor of overall survival (reader 1, p = 01; reader 2, p = 06), after adjusting for stage and days between surgery and postoperative CT. Discussion The current standard for determining residual disease status after primary cytoreduction relies on clinical intraoperative assessment [5, 15]. However, the clinical assessment of tumor size is difficult and not error proof. A study by Aletti et al. [15] found that patients with advanced ovarian cancer who underwent diaphragmatic surgery had a significantly greater chance of 5-year survival than did patients who did not. Those authors postulated that this survival difference was due to incomplete tumor exposure and visualization with occult diaphragmatic dis- Fig. 2 Examples of suboptimally debulked residual disease on early postoperative CT scans of three different patients. A and B, 58-year-old woman with primary ovarian cancer. Preoperative (A) and postoperative (B) CT scans show enlarged left paraaortic lymph nodes (arrows). C and D, 60-year-old woman with primary ovarian cancer. Preoperative (C) and postoperative (D) CT scans show right posterior hepatic surface implant (arrows). E and F, Different 60-year-old woman with primary ovarian cancer. Preoperative (E) and postoperative (F) CT scans show splenic implant (arrows). 1456 AJR:198, June 2012

Postoperative CT as Prognostic Biomarker in Gynecologic Cancer TABLE 3: Sites of on Postoperative CT of 63 Patients Residual Disease Locations Reader 1 Reader 2 Perihepatic region 15 (24) 13 (21) Large-bowel region 12 (19) 13 (21) Upper abdominal lymphadenopathy 7 (11) 10 (16) Supradiaphragmatic and retrocrural lymphadenopathy 7 (11) 7 (11) Lower abdominal or pelvic lymphadenopathy 5 (8) 7 (11) Perisplenic region 4 (6) 9 (14) Small-bowel region 4 (6) 6 (10) Liver parenchyma 4 (6) 6 (10) Pelvis 2 (3) 4 (6) Spleen parenchyma 0 0 Note Data are no. (%) of patients. ease in patients presumed to have had optimal debulking. In present clinical practice, postoperative CT is not routinely used to document residual disease status after primary debulking, to provide a baseline before initiating postoperative chemotherapy, or to guide therapeutic decision making after surgery. Yet, prospective clinical studies, such as the GOG trials, frequently require a baseline postoperative CT scan before initiation of investigational treatment. We examined the value of using early postoperative CT to supplement surgical evaluation of residual disease status. Progression-Free Survival (%) Overall Survival (%) p = 01 p = 10 A C In our study, residual disease larger than 1 cm was found on early postoperative CT in up to 49% of patients whose disease was considered optimally debulked at primary cytoreduction on the basis of intraoperative clinical visualization alone. In identifying patients with residual disease larger than 1 cm, two radiologists displayed substantial agreement, suggesting that CT findings regarding residual disease status are highly reproducible. For each patient, a preoperative CT scan was available for comparison at the time of postoperative CT review; the preoperative CT was instrumental in distinguishing between residual disease (which, in Progression-Free Survival (%) Overall Survival (%) p = 01 p = 05 B D the vast majority of cases, was in the same site as disease seen on the initial scan) and postsurgical change. The most common locations of suboptimally debulked disease on CT (e.g., diaphragmatic surfaces, the portal region, the root of the mesentery, and paraaortic and pelvic adenopathy) were consistent with those reported by gynecologic oncologists to pose the greatest difficulties in achieving optimal cytoreduction [16]. Most important, the presence of residual disease larger than 1 cm on CT was associated with decreases in progression-free and overall survival in both univariate and multivariate analyses. These findings indicate that the assessment of residual disease on early postoperative CT based on specific interpretation criteria could aid in prognostic assessment and treatment selection and, importantly, serve as a biomarker for patient stratification in clinical trials. Our results expand on those of an initial analysis by Chi et al. [8], which assessed the prognostic significance of early postoperative CT findings in 67 patients from the second part of the same clinical trial (i.e., patients with advanced ovarian, tubal, and peritoneal cancer in whom optimal primary debulking was reported by the surgeon). Our patient cohort closely resembles that analyzed by Chi et al., with similar demographics and postsurgical treatments. In the analysis by Chi et al. [8], postoperative CT indicated residual disease larger than 1 cm in 43% of patients whose disease was deemed optimally debulked at surgery. In patients with residual disease larger than 1 cm with a qualitative assessment score of 4 or 5 on postoperative CT, median progression-free survival was reduced by 4 months, and median overall survival was reduced by 17 months, but the differences in survival were not statistically significant at the time of writing [8]. In the initial study by Chi et al. [8], each CT scan was read by one of five protocol radiologists, and the baseline and follow-up scans of each patient were not necessarily interpreted by the same reader. Specific imaging criteria for identifying residual disease were not outlined, and radiologists were guided by their variable levels of experience and their individual clinical judgment. In contrast, in the current study, two radiologists (neither of whom Fig. 3 Survival for patients with residual disease status determined on CT by two readers. A and B, Graphs show progression-free survival as determined by reader 1 (A) and reader 2 (B). C and D, Graphs show overall survival as determined by reader 1 (C) and reader 2 (D). AJR:198, June 2012 1457

Lakhman et al. TABLE 4: Univariate Analysis for Progression-Free Survival and Overall Survival Univariate Progression-Free Survival Analysis Univariate Overall Survival Analysis Survival a (Months), Survival b (Months), Variables No. Median (95% CI) p No. Median (95% CI) p Days between surgery and CT, median (range) 15 (4 49) 19 (16 27) 0.123 c 15 (4 49) 51 (37 70) 98 d Site Ovary 48 19 (16 29) 94 50 51 (38 70) 74 Other 13 18 (10 NA) 13 NA (17 NA) Stage III 46 21 (17 32) 0.507 48 53 (44 84) 82 IV 15 15 (13 NA) 15 30 (23 NA) Grade Other 8 36 (12 NA) 39 8 70 (53 NA) 62 High 50 18 (16 27) 52 44 (36 70) Histology Serous 56 19 (16 29) 0.195 58 53 (38 77) 41 Other 5 14 (10 NA) 5 43 (26 NA) Postsurgery residual disease None 26 19 (15 66) 0.531 26 53 (43 NA) 0.397 0.1 0.5 cm 22 23 (16 41) 22 51 (28 NA) cm 13 16 (10 NA) 15 37 (27 NA) Ascites No 9 21 (17 NA) 0.531 9 NA (53 NA) 0.117 Yes 52 19 (15 29) 54 49 (36 70) Intraperitoneal chemotherapy Yes 38 18 (14 29) 0.171 40 44 (29 70) 0.132 No 23 21 (18 65) 23 70 (50 NA) Age at surgery (y) < 60 28 24 (16 41) 26 29 62 (43 NA) 07 60 33 18 (14 29) 34 49 (28 70) Residual disease on CT, reader 1 1 cm 33 29 (19 65) 01 34 70 (38 NA) 10 > 1 cm 28 16 (10 24) 29 44 (32 70) Residual disease on CT, reader 2 1 cm 32 36 (19 66) 01 32 77 (49 NA) 05 > 1 cm 29 16 (12 19) 31 43 (30 62) Note Except where noted otherwise, data are no. of patients. NA = not achieved. a In patients who survived without progression for at least 50 days after surgery. b In patients who survived at least 50 days after surgery. c p value was derived from the Cox regression for those patients who survived at least 50 days without progression; other p values were calculated using the log-rank test. d p value was derived from the Cox regression for those patients who survived at least 50 days; other p values were calculated using the log-rank test. participated in the initial study) independently interpreted all patients CT scans using mutually agreed on criteria. The prognostic significance of CT findings was analyzed separately for each reader, and interobserver variability was assessed. Furthermore, although the initial analysis used a median follow-up time of 50 months, the current study benefited from a longer median follow-up time of 75 months. Although prospective clinical trials are considered the reference standard for biomarker validation, such trials are typically both expensive and lengthy. Retrospective analysis of diagnostic noninterventional data from a well-designed prospective clinical trial the approach used in the present study has been recognized as an essential means for validating new biomarkers in a timely manner [17]. The limitations of retrospective studies are, however, acknowledged. For example, in an ACRIN/GOG trial comparing CT and MRI for the assessment of early invasive cervical cancer, prospective readings of both modalities were found to be more accurate than retrospective readings [18]. Our study is hypothesis generating and can serve as a starting point for the design and testing of new risk-adjusted 1458 AJR:198, June 2012

Postoperative CT as Prognostic Biomarker in Gynecologic Cancer TABLE 5: Multivariate Analyses for Both Readers Reader 1 Reader 2 Multivariate Survival Analysis No. of Patients HR (95% CI) p HR (95% CI) p Progression-free survival: residual disease on CT 1 cm 33 1 (reference) 01 1 (reference) 01 > 1 cm 28 2.54 (1.44 4.49) 2.50 (1.42 4.37) Overall survival Days between surgery and CT 2 (0.15 1.20) 0.105 5 (0.16 1.28) 0.135 Stage III 48 1 (reference) 16 1 (reference) 0.125 IV 15 2.54 (1.19 5.42) 1.76 (6 3.64) Residual disease on CT 1 cm 34 1 (reference) 01 1 (reference) 06 > 1 cm 29 3.29 (1.62 6.70) 2.48 (1.30 4.73) Note All analyses were based on the landmark analyses for those patients who were event free at 50 days. HR = hazard ratio. treatment protocols and further comparisons of outcomes in patients with optimally and suboptimally debulked disease on CT. In summary, our study showed that residual disease larger than 1 cm was present on early postoperative CT in almost half of the patients deemed to have optimally debulked disease at primary cytoreduction. This finding was reproducible, because the two radiologists displayed substantial interobserver agreement. Residual disease larger than 1 cm detected on early postoperative CT was associated with significant decreases in both progression-free and overall survival. These findings support the premise that residual disease status as assessed on early postoperative CT could serve as a biomarker to supplement the clinical intraoperative evaluation of residual disease status by the surgeon. CT could play an important role in prognostic assessment, treatment selection, and patient allocation to (and within) clinical drug trials. The role of CT in this context could be especially important in large clinical trials involving multiple surgeons and institutions. Acknowledgment We thank Ada Muellner for editing the manuscript. References 1. Jemal A, Murray T, Ward E, et al. Cancer statistics. CA Cancer J Clin 2005; 55:10 30 [Erratum in CA Cancer J Clin 2005; 55:259] 2. Bristow RE, Tomacruz RS, Armstrong DK, Trimble EL, Montz FJ. Survival effect of maximal cytoreductive surgery for advanced ovarian carcinoma during the platinum era: a meta-analysis. J Clin Oncol 2002; 20:1248 1259 3. Ozols RF, Bundy BN, Greer BE, et al; Gynecologic Oncology Group. 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 2003; 21:3194 3200 4. Armstrong DK, Bundy B, Wenzel L, et al; Gynecologic Oncology Group. Intraperitoneal cisplatin and paclitaxel in ovarian cancer. N Engl J Med 2006; 354:34 43 5. Préfontaine M, Gelfand AG, Donavan JT, Powell JL. Reproducibility of tumor measurements in ovarian cancer: a study of interobserver variability. Gynecol Oncol 1994; 55:87 90 6. Bristow RE, Duska LR, Lambrou NC, et al. A model for predicting surgical outcome in patients with advanced ovarian carcinoma using computed tomography. Cancer 2000; 89:1532 1540 7. Chi DS, Ramirez PT, Teitcher JB, et al. Prospective study of the correlation between postoperative computed tomography scan and primary surgeon assessment in patients with advanced ovarian, tubal, and peritoneal carcinoma reported to have undergone primary surgical cytoreduction to residual disease 1 cm or less. J Clin Oncol 2007; 25:4946 4951 8. Chi DS, Barlin JN, Ramirez PT, et al. Follow-up study of the correlation between postoperative computed tomography scan and primary surgeon assessment in patients with advanced ovarian, tubal, and peritoneal carcinoma reported to have undergone primary surgical cytoreduction to residual disease of 1 cm or smaller. Int J Gynecol Cancer 2010; 20:353 357 9. Therasse P, Arbuck SG, Eisenhauer EA, et al. New guidelines to evaluate the response to treatment in solid tumors. European Organization for Research and Treatment of Cancer, National Cancer Institute of the United States, National Cancer Institute of Canada. J Natl Cancer Inst 2000; 92:205 216 10. Rustin GJ, Marples M, Nelstrop AE, Mahmoudi M, Meyer T. Use of CA-125 to define progression of ovarian cancer in patients with persistently elevated levels. J Clin Oncol 2001; 19:4054 4057 11. Landis JR, Koch GG. The measurement of observer agreement for categorical data. Biometrics 1977; 33:159 174 12. Mantel N. Evaluation of survival data and two new rank order statistics arising in its consideration. Cancer Chemother Rep 1966; 50:163 170 13. Cox DR. Regression models and life-tables. J R Statist B 1972; 34:187 200 14. Kaplan EL, Meier P. Nonparametric estimator from incomplete observations. J Am Stat Assoc 1958; 53:457 481 15. Aletti GD, Dowdy SC, Podratz KC, Cliby WA. Surgical treatment of diaphragm disease correlates with improved survival in optimally debulked advanced stage ovarian cancer. Gynecol Oncol 2006; 100:283 287 16. Eisenkop SM, Spirtos NM. What are the current surgical objectives, strategies, and technical capabilities of gynecologic oncologists treating advanced epithelial ovarian cancer? Gynecol Oncol 2001; 82:489 497 17. Mandrekar SJ, Sargent DJ. Clinical trial designs for predictive biomarker validation: theoretical considerations and practical challenges. J Clin Oncol 2009; 27:4027 4034 18. Mitchell DG, Snyder B, Coakley F, et al. Early invasive cervical cancer: tumor delineation by magnetic resonance imaging, computed tomography, and clinical examination, verified by pathologic results, in the ACRIN 6651/GOG 183 Intergroup Study. J Clin Oncol 2006; 24:5687 5694 AJR:198, June 2012 1459