DOI: 10.1111/1471-0528.12992 www.bjog.org
Radiological predictors of cytoreductive outcomes in patients with advanced ovarian cancer J Borley,a C Wilhelm-Benartzi,a J Yazbek,b R Williamson,b,c N Bharwani,b,c V Stewart,b,c I Carson,b E Hird,b A McIndoe,b A Farthing,a,b S Blagden,a,b S Ghaem-Maghamia,b a Department of Surgery and Cancer, Imperial College London, London, UK b West London Gynaecology Cancer Centre, Imperial College NHS Trust, London, UK c Department of Radiology, Imperial College Healthcare NHS Trust, London, UK Correspondence: Dr S Ghaem-Maghami, 2nd floor Hamm House, Hammersmith Hospital, Du Cane Road, London W12 0NN, UK. Email [email protected]
Accepted 10 May 2014. Published Online 5 August 2014.
Objective To assess site of disease on preoperative computed
tomography (CT) to predict surgical debulking in patients with ovarian cancer. Design Two-phase retrospective cohort study. Setting West London Gynaecological Cancer Centre, UK. Population Women with stage 3 or 4, ovarian, fallopian or
primary peritoneal cancer undergoing cytoreductive surgery. Methods Preoperative CT images were reviewed by experienced radiologists to assess the presence or absence of disease at predetermined sites. Multivariable stepwise logistic regression models determined sites of disease which were significantly associated with surgical outcomes in the test (n = 111) and validation (n = 70) sets. Main outcome measures Sensitivity and specificity of CT in
predicting surgical outcome. Results Stepwise logistic regression identified that the presence of
mesentery and small-bowel mesentery, and infrarenal para-aortic nodes were associated with debulking status. Logistic regression determined a surgical predictive score which was able to significantly predict suboptimal debulking (n = 94, P = 0.0001) with an area under the curve (AUC) of 0.749 (95% confidence interval [95% CI]: 0.652, 0.846) and a sensitivity of 69.2%, specificity of 71.4%, positive predictive value of 75.0% and negative predictive value of 65.2%. These results remained significant in a recent validation set. There was a significant difference in residual disease volume in the test and validation sets (P < 0.001) in keeping with improved optimal debulking rates. Conclusions The presence of disease at some sites on
preoperative CT scan is significantly associated with suboptimal debulking and may be an indication for a change in surgical planning. Keywords Computed tomography, ovarian cancer, preoperative assessment, surgical debulking.
lung metastasis, pleural effusion, deposits on the large-bowel Please cite this paper as: Borley J, Wilhelm-Benartzi C, Yazbek J, Williamson R, Bharwani N, Stewart V, Carson I, Hird E, McIndoe A, Farthing A, Blagden S, Ghaem-Maghami S. Radiological predictors of cytoreductive outcomes in patients with advanced ovarian cancer. BJOG 2014; DOI: 10.1111/1471-0528. 12992.
Introduction The management of women with epithelial ovarian cancer (EOC) remains one of the greatest challenges for the gynaecological oncology community. The majority of women present in advanced stage, eventually develop chemotherapy resistance and ultimately succumb to the disease. The volume of residual disease following primary debulking surgery remains the single best predictive factor in survival,1
ª 2014 Royal College of Obstetricians and Gynaecologists
with much recent emphasis placed on achieving complete macroscopic debulking (defined as the absence of any macroscopic residual disease).2–4 Supraradical debulking surgery (such as splenic and hepatic resection, diaphragm stripping and peritonectomy) is often required to achieve this. Results from multicentre randomised clinical trials (RCTs) suggest that UK debulking rates are inferior to those of other countries,5,6 with significantly lower numbers for large-bowel resection, pelvic and para-aortic
1
Borley et al.
lymphadenectomy when compared with some other countries. The site of disease prior to surgery has been shown to be a major predictor for complete tumour resection, with patients having disease at surgical locations of the umbilical, left flank and epigastrium regions having the highest risk of incomplete resection.7 It therefore appears logical that efforts to determine tumour burden preoperatively through imaging modalities such as computed tomography (CT), magnetic resonance imaging (MRI) and positron emission tomography (PET) would be beneficial and inform the surgeon of the likelihood of optimal versus suboptimal debulking and the need for bowel resection. This, in turn, would translate into better preoperative counselling and healthcare planning through improvements in operating list scheduling and specialised staff allocation. It may also allow improved identification of those who may benefit from neoadjuvant chemotherapy and delayed debulking instead of suboptimal primary debulking. Previous studies have examined the use of CT imaging to predict surgical success,8–17 but these studies relate to small patient numbers, changing definitions of optimal debulking and out-of-date debulking rates. Two more recent studies have tried to validate the previous studies, but with limited success.12,18 One of the main arguments in the failure to validate previous findings is that factors which accounted for unresectable disease 10 years ago are now irrelevant, as surgical paradigms change and debulking rates improve. Thus, the aim of this study was to assess the role of CT in the preoperative prediction of debulking outcome and requirement for bowel surgery in patients with advanced EOC in the UK. This was performed and validated with current data which are representative of surgical practices and debulking rates in the UK today.
Methods
CT scans were assessed against criteria previously reported as potential indicators for resectability by experienced gynaecological radiologists,9–12,14,16,19 plus the addition of our own anecdotal experience. The radiologists were blind to the surgical outcome of each patient. CT scans were from a variety of referring institutions. All scans were performed using thin slice multidetector CT and contrast enhancement. All scans were reconstructed in axial sections with the option for coronal section post-processing for radiological review. Images were read by three consultant radiologists who were core members of the Gynaecology–Oncology Multidisciplinary Team. In order to allow for reproducibility, disease was scored as being present or absent with the cut-off for the identification of disease being consistent with RECIST (Response Evaluation Criteria In Solid Tumours) 1.120 indicators of measurable disease, i.e. 10 mm being the minimum size for a positive result. Lymph nodes were measured in the short axis. Absolute measurements were not included. The following regions were assessed for disease: diaphragm/superior surface of the liver, lateral surface of the liver, liver parenchyma, inferior surface of the liver/porta hepatis, gastrosplenic ligament, lesser sac, large-bowel serosa, small-bowel serosa, large-bowel mesentery, small-bowel mesentery, pelvis/adnexa. Lymphadenopathy with a short axis >10 mm was assessed in the following regions: pericardiac, porta, small- and large-bowel mesentery, suprarenal para-aortic/aortocaval, infrarenal para-aortic, pelvic, inguinal. Non-measurable disease in the form of ascites on more than four axial sections and lung nodules of >7 mm21 were also recorded. The assessment of inter-observer variation did not form part of this study as, with the use of RECIST assessments, the cut-off size of 10 mm means that detection of the disease is consistent,20 with previous studies demonstrating intra- and inter-observer variation to be low.22,23 All studies included in the preliminary analysis were adequate for review.
Data collection
Statistics
This study was approved by Imperial College NHS Trust (reference 1612) with no ethical objections to the study because of the retrospective nature of the data collection. Clinical data were collected retrospectively from the West London Gynaecological Cancer Centre database on patients with a diagnosis of stage 3 and 4 ovarian, fallopian and primary peritoneal cancer who underwent surgery. Data collected included age at surgery, histological subtype, stage and grade, timing of debulking surgery, details of operation received, including any surgery to the large or small bowel, and residual disease status. One-hundred and eleven patients undergoing treatment at the centre from August 2001 to June 2010 were randomly selected for the test set, and 70 patients undergoing treatment from June 2010 to December 2012 for the validation set. Their preoperative
The aim of this work was to determine which of the CT anatomical findings were associated with debulking success and requirement for bowel resection. Statistical analysis was performed in R (version 2.15). Permutation Kruskal–Wallis and Pearson chi-squared tests with 10 000 permutations were used to determine whether surgical or clinical characteristics were significantly different from one another in the test and validation datasets. A multivariable stepwise logistic regression model was first run associating debulking status, coded as suboptimal debulking (defined as residual disease > 10 mm) versus optimal debulking (residual disease ≤ 10 mm), against all available CT anatomical measures and age to find the optimal model based on statistical model fitting using the Akaike information criterion (AIC) (stepAIC MASS package R). A logistic regression was then
2
ª 2014 Royal College of Obstetricians and Gynaecologists
Radiological predictors in ovarian cancer
Box 1 Logistic regression model for prediction of surgical success and a worked example
If disease present, score = 1, or absent, score = 0, at each individual site. A = Lung metastasis B = Pleural effusion C = Large-bowel mesentery D = Small-bowel mesentery E = Inferior renal para-aortic lymph node
Log score = –1.15 + 16.394(A) + 1.9150(B) + 1.5965 (C) + 2.3020(D) + 1.0883(E) Log score = n Score = exp (n) A score of >1.563 indicates a high risk of suboptimal debulking; a score of 10 46 (41.4) Unresectable disease 13 (11.7) Optimal debulking 52 (46.8) (10 mm) Small bowel mesentry (>10 mm) Infrarenal paraaortic lymph node
1.15 16.394 1.915 1.597 2.302 1.088
(0.48) (1686.73) (0.77) (0.75) (1.17) (0.67)
improved from 46.8% in the test set to 75.7% in the validation set, and total macroscopic debulking from 30.6% to 54.3%, respectively. In order to establish whether findings on preoperative CT scans were associated with debulking outcome (optimal versus suboptimal debulking), a stepwise logistic regression model was performed to find the best model fit. Using this model, the presence of lung metastasis >7 mm, pleural effusion, deposits of >10 mm on the large-bowel mesentery and small-bowel mesentery, and infrarenal para-aortic nodes were associated with debulking status. A greater proportion of patients who were suboptimally debulked had disease in these sites relative to those optimally debulked. In the suboptimal group, 5.8% had lung metastasis, 38.5% had pleural effusions, 36.5% had large-bowel mesentery deposits, 21.2% had small-bowel mesentery deposits and 26.9% had inferior renal para-aortic lymph nodes relative to 0%, 14.3%, 16.7%, 4.8% and 21.4%, respectively, in the optimally debulked group (Table 2). The effect estimates (beta coefficients) from these covariates were calculated using logistic regression modelling (Table 2), which enabled a surgical predictive score to be generated for each individual. Our continuous predictive score was able to predict suboptimal debulking with an AUC of 0.749 (95% confidence interval [95% CI]: 0.652, 0.846) (Figure 1A). Patients were categorised into being above and below the median score (median, 1.563). A predictive score above the median was able to significantly predict suboptimal debulking (odds ratio [OR] = 5.62, n = 94, P = 0.0001) with a sensitivity of 69.2%, specificity of 71.4%, positive predictive value (PPV) of 75.0% and negative predictive value (NPV) of 65.2%. Box 1 illustrates how this logistical regression model is used with an example. Our predictive score was successfully validated in the validation set. As expected, a higher burden of disease was
4
0 6 7 2 9
(0.0) (14.3) (16.7) (4.8) (21.4)
Suboptimal debulking n = 52 (%)
3 20 19 11 14
(5.8) (38.5) (36.5) (21.2) (26.9)
Validation set (n = 70) Optimal debulking n = 53 (%)
4 8 24 5 9
(7.5) (15.1) (45.3) (9.4) (17.0)
Suboptimal debulking n = 17 (%)
2 7 13 4 4
(11.8) (41.2) (76.5) (23.5) (23.5)
again observed in the suboptimally debulked group, with more patients having disease on each site relative to those in the optimally debulked group (Table 2). The previous predictive score was used, as before from the model beta coefficients (Box 1), to generate an individual risk score for each case, and these were also grouped by the individual score being above or below the test set median score. In the validation set, a predictive score above the median was able to predict suboptimal debulking with a sensitivity of 64.7% and specificity of 67.9%. The continuous score generated an AUC value of 0.721 (95% CI: 0.594, 0.847) (Figure 1B). We also tested the ability of the predictive score to predict total macroscopic debulking (0 mm residual disease) relative to debulking with residual disease of ≥1 mm. This had a sensitivity of 65.6% and a specificity of 80% in the test set, and a sensitivity of 50.0% and a specificity of 68.4% in the validation set. Lastly, we used another stepwise logistic regression model to determine which covariates were associated with requirement for bowel resection at surgery. Twenty of the 111 patients (18%) had bowel resection as part of their debulking surgery. The model found that age, deposits on the large- and small-bowel mesentery, large-bowel serosa, and lateral and inferior surface of the liver, mesenteric or portal nodes and inferior renal para-aortic nodes were associated with requirement for bowel surgery. The beta coefficients were generated (data not shown) to provide an individual risk score for each patient. This score, however, was not found to be statistically significantly associated with bowel surgery in logistic regression models when the individual scores were used (n = 108, P = 0.789). There was, however, a trend towards a significant association with bowel surgery in the dichotomised model where the patients were coded depending on a score above or below the median (OR = 2.77, n = 108, P = 0.0533).
ª 2014 Royal College of Obstetricians and Gynaecologists
Radiological predictors in ovarian cancer
A
B
Figure 1. Receiver operating characteristic (ROC) curve for the logistic regression score to predict surgical debulking. (A) Test set. (B) Validation set. AUC, area under the curve.
Discussion
ascites and bowel mesentery involvement were often associated with suboptimal debulking status. We found that the presence of lung metastasis, pleural effusion, deposits on the large-bowel mesentery and small-bowel mesentery, and infrarenal para-aortic nodes were associated with suboptimal debulking in our patient group. Our model predicted suboptimal debulking with a sensitivity of 64.7–69.2% and specificity of 67.9–71.4% in two independent patient sets (AUC = 0.721–0.749). Why these particular sites of disease are more strongly associated with suboptimal debulking than others is unclear. It seems logical that deposits of >10 mm on the large- and small-bowel mesentery may be associated with higher rates of suboptimal debulking if disease at this site is so extensive that disease clearance would require multiple (or complete) bowel resection. The involvement of other sites may simply be an indication of extensive and biologically aggressive disease, making optimal resection within the abdomen less likely. As the debate with regard to tumour biology determining surgical outcomes persists,24,25 there may be inherent biological factors within these tumours which dictate specific spread and invasion to these sites. The appearance of infrarenal para-aortic nodes over other nodal disease being strongly associated with suboptimal debulking is also of curiosity. Although there was a larger proportion of cases with mesenteric or portal vein nodes in the suboptimal group than in the optimal group (12/52 [23.1%] versus 6/42 [14.3%]), the model did not find this to be an indicator. There were only five cases of suprarenal para-aortic node involvement in total, and so we suggest that the study was not adequately powered to determine a specific association for this disease site. In contrast with previous studies, we did not find that upper abdominal disease was associated with suboptimal debulking in the test set. This may be a result of our increasing use of extensive upper abdominal resections and diaphragm/peritoneal stripping in the centre. We found a non-significant trend towards increased risk of requiring bowel resection in patients who had disease present at sites over the large- and small-bowel mesentery, large-bowel serosa, and lateral and inferior surface of the liver, ascites, and mesenteric or portal nodes and inferior renal para-aortic nodes. The relationship between disease on the bowel and requirement for resection seems obvious; however, the volume of disease, number of deposits, specific location and invasiveness of the tumour into the bowel will impact on the need for resection. Thus, our study was not adequately designed to address this issue.
Main findings
Strengths and limitations
Much emphasis has been placed on finding a preoperative method to accurately predict the debulking status of suspected EOC. In previously published studies,7–18 the presence of disease in the upper abdomen, large volume of
Our study uses one of the largest datasets published and robust statistical methods to determine a prediction model. It is also the first study in more than 10 years that uses a current validation set. Stepwise logistic regression was used
ª 2014 Royal College of Obstetricians and Gynaecologists
5
Borley et al.
to determine the best predictive model for surgical debulking from which the risk score was determined. This method allows the creation of a small, readily interpretable model containing the most important predictors of surgical debulking from a list of CT anatomical findings. In addition, as we used RECIST 1.1 indicators of measurable disease (disease ≥ 10 mm), this ensures that the results are consistent with limited inter-observer variability. A potential limitation of this method lies in the instability of the result when adding or removing patients, and the statistical power of the study may be insufficient to select true predictors. However, given that our model was validated in an independent set, we believe that it is robust to these potential problems. One of the key issues in predicting debulking is the ever-changing surgical goal with inherent variation on what accounts for truly resectable disease. Factors that reduce the likelihood of achieving successful cytoreduction previously cited include: diffuse small-bowel involvement, infiltration into the porta hepatis and diffuse involvement of the right hemidiaphragm, resulting in fixation of the liver2 or large liver metastasis.24 Patient age, comorbidities and intensive care availability will also impact on outcome. Furthermore, what was deemed as unresectable disease in the year 2000 has now become resectable through the development of, and emphasis on supraradical surgery.26 This means that previous models of prediction have become obsolete. Our findings were validated in an independent dataset from patients recently treated at the centre, suggesting that it is more current and clinically relevant. During the time course of this study, we achieved a significant improvement in surgical debulking rates, with an improvement of optimal debulking from 46.8% to 75.7% and complete debulking from 30.4% to 54.3%. The model remains significant despite this change, indicating that these prediction markers are likely to be truly representative of an inability to achieve optimal debulking irrespective of surgical effort. We believe that our debulking rates at the centre are in line with those of other UK-based cancer centres, both historically5,27,28 and in the current era of supraradical surgery.29 This adds evidence that our prediction model can be translated to other centres that perform similar numbers and types of surgery. Current evidence now highlights the need to perform complete macroscopic debulking, whenever possible. However, at the time of surgery in our test set, optimal debulking was the primary goal and thus we used this as our primary outcome measure. In subset analysis, the model performed less well when predicting any amount of residual disease versus complete debulking. This demonstrates the limitations of CT scans to define disease which is present, but ≤10 mm, and highlights that the statistical model should not be used interchangeably. It appears that disease
6
that predicts suboptimal from optimal surgery may be different from that which predicts complete debulking. We strongly believe that it is still clinically relevant to determine the likelihood of suboptimal debulking prior to surgery, especially when making multidisciplinary team meeting treatment decisions with regard to upfront versus delayed debulking surgery.
Interpretation This statistical model and score can be readily calculated and interpreted for any patient who undergoes staging CT preoperatively for ovarian cancer (see Box 1). In the test set, only 25% (12/48) of patients with a high risk score ultimately achieved optimal debulking, and 39% (11/27) of patients in the validation set with a high risk score achieved optimal debulking. We argue that patients with a high risk score may be stratified into receiving neoadjuvant chemotherapy and interval debulking. This approach will not compromise survival and may lead to an increase in debulking rates.6
Conclusions Clearly, a well-designed prospective RCT is needed to address the issues raised in the Discussion section and to further validate our findings. National Institute for Health and Clinical Excellence (NICE) guidance28 also recommends further research into the evaluation of the role of MRI in addition to CT in the staging of ovarian cancer and prediction of surgical success, and this should also be incorporated into future studies. The relevance of being able to correctly predict surgical outcome has not been formally evaluated to date. It is therefore necessary to determine how an accurate prediction of surgical outcomes influences patient survival, quality of life and the impact on NHS service planning and provision. Only then will these markers of suboptimal debulking truly be able to be incorporated into the care and management of patients with advanced ovarian cancer.
Disclosure of interests The authors report no declaration of interest.
Contribution to authorship JB and SGM wrote the manuscript. JY, RW, AMI, AF, SB and SGM were responsible for the conception, design and critique of the study. JB, IC, EH and JY performed the data collection. JB and CWB performed the statistical analysis. NB, VS and RW reported the CT scans. All authors reviewed the manuscript and contributed to the final submission.
Details of ethics approval Ethical approval from Imperial College NHS Trust (ref. 1612).
ª 2014 Royal College of Obstetricians and Gynaecologists
Radiological predictors in ovarian cancer
Funding JB is funded by Cancer Research UK/National Institute for Health Research (CRUK/NIHR). CWB is funded by CRUK.
Acknowledgements Conduct of the study was supported by the Experimental Cancer Medicine Centre, the National Institute for Health Research (NIHR), Biomedical Research Centre and the Ovarian Cancer Action Centre, Imperial College London, UK. &
References 1 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;20:1248–59. 2 Pomel C, Barton DP, McNeish I, Shepherd J. A statement for extensive primary cytoreductive surgery in advanced ovarian cancer. BJOG 2008;115:808–10. 3 Elattar A, Bryant A, Winter-Roach BA, Hatem M, Naik R. Optimal primary surgical treatment for advanced epithelial ovarian cancer. Cochrane Database Syst Rev 2011;8:CD007565. 4 Redman C, Duffy S, Bromham N, Francis K, Group GD. Recognition and initial management of ovarian cancer: summary of NICE guidance. BMJ 2011;342:d2073. 5 Crawford SC, Vasey PA, Paul J, Hay A, Davis JA, Kaye SB. Does aggressive surgery only benefit patients with less advanced ovarian cancer? Results from an international comparison within the SCOTROC-1 Trial. J Clin Oncol 2005;23:8802–11. 6 Vergote I, Trope CG, Amant F, Kristensen GB, Ehlen T, Johnson N, et al. Neoadjuvant chemotherapy or primary surgery in stage IIIC or IV ovarian cancer. N Engl J Med 2010;363:943–53. 7 Fotopoulou C, Richter R, Braicu EI, Schmidt SC, Lichtenegger W, Sehouli J. Can complete tumor resection be predicted in advanced primary epithelial ovarian cancer? A systematic evaluation of 360 consecutive patients. Eur J Surg Oncol 2010; 36:1202–10. 8 Ibeanu OA, Bristow RE. Predicting the outcome of cytoreductive surgery for advanced ovarian cancer: a review. Int J Gynecol Cancer 2010;21:S1–11. 9 Nelson BE, Rosenfield AT, Schwartz PE. Preoperative abdominopelvic computed tomographic prediction of optimal cytoreduction in epithelial ovarian carcinoma. J Clin Oncol 1993;11:166–72. 10 Meyer JI, Kennedy AW, Friedman R, Ayoub A, Zepp RC. Ovarian carcinoma: value of CT in predicting success of debulking surgery. Am J Roentgenol 1995;165:875–8. 11 Bristow RE, Duska LR, Lambrou NC, Fishman EK, O’Neill MJ, Trimble EL, et al. A model for predicting surgical outcome in patients with advanced ovarian carcinoma using computed tomography. Cancer 2000;89:1532–40. 12 Axtell AE, Lee MH, Bristow RE, Dowdy SC, Cliby WA, Raman S, et al. Multi-institutional reciprocal validation study of computed tomography predictors of suboptimal primary cytoreduction in patients with advanced ovarian cancer. J Clin Oncol 2007;25:384–9.
ª 2014 Royal College of Obstetricians and Gynaecologists
13 Dowdy SC, Mullany SA, Brandt KR, Huppert BJ, Cliby WA. The utility of computed tomography scans in predicting suboptimal cytoreductive surgery in women with advanced ovarian carcinoma. Cancer 2004;101:346–52. 14 Qayyum A, Coakley FV, Westphalen AC, Hricak H, Okuno WT, Powell B. Role of CT and MR imaging in predicting optimal cytoreduction of newly diagnosed primary epithelial ovarian cancer. Gynecol Oncol 2005;96:301–6. 15 Salani R, Axtell A, Gerardi M, Holschneider C, Bristow RE. Limited utility of conventional criteria for predicting unresectable disease in patients with advanced stage epithelial ovarian cancer. Gynecol Oncol 2008;108:271–5. 16 Risum S, Høgdall C, Loft A, Berthelsen AK, Høgdall E, Nedergaard L, et al. Prediction of suboptimal primary cytoreduction in primary ovarian cancer with combined positron emission tomography/ computed tomography – a prospective study. Gynecol Oncol 2008;108:265–70. 17 Jung DC, Kang S, Kim MJ, Park SY, Kim HB. Multidetector CT predictors of incomplete resection in primary cytoreduction of patients with advanced ovarian cancer. Eur Radiol 2010;20:100–7. 18 Gemer O, Lurian M, Gdalevich M, Kapustian V, Piura E, Schneider D, et al. A multicenter study of CA 125 level as a predictor of non-optimal primary cytoreduction of advanced epithelial ovarian cancer. Eur J Surg Oncol 2005;31:1006–10. 19 Byrom J, Widjaja E, Redman CW, Jones PW, Tebby S. Can pre-operative computed tomography predict resectability of ovarian carcinoma at primary laparotomy? BJOG 2002;109:369–75. 20 Eisenhauer EA, Therasse P, Bogaerts J, Schwartz LH, Sargent D, Ford R, et al. New response evaluation criteria in solid tumours: revised RECIST guideline (version 1.1). Eur J Cancer 2009;45:228–47. 21 Swensen SJ, Jett JR, Hartman TE, Midthun DE, Mandrekar SJ, Hillman SL, et al. CT screening for lung cancer: five-year prospective experience. Radiology 2005;235:259–65. 22 Muenzel D, Engels HP, Bruegel M, Kehl V, Rummeny EJ, Metz S. Intra- and inter-observer variability in measurement of target lesions: implication on response evaluation according to RECIST 1.1 Radiol Oncol 2012;46:8–18. 23 Sun JM, Ahn MJ, Park MJ, Yi JH, Kim TS, Chung MJ, et al. Accuracy of RECIST 1.1 for non-small cell lung cancer treated with EGFR tyrosine kinase inhibitors. Lung Cancer 2010;69:105–9. 24 Chi DS, Schwartz PE. Cytoreduction vs. neoadjuvant chemotherapy for ovarian cancer. Gynecol Oncol 2008;111:391–9. 25 Borley J, Wilhelm-Benartzi C, Brown R, Ghaem-Maghami S. Does tumour biology determine surgical success in the treatment of epithelial ovarian cancer? A systematic literature review. Br J Cancer 2012;107:1069–74. 26 Chi DS, Eisenhauer EL, Zivanovic O, Sonoda Y, Abu-Rustum NR, Levine DA, et al. Improved progression-free and overall survival in advanced ovarian cancer as a result of a change in surgical paradigm. Gynecol Oncol 2009;114:26–31. 27 Perren TJ, Swart AM, Pfisterer J, Ledermann JA, Pujade-Lauraine E, Kristensen G, et al. A phase 3 trial of bevacizumab in ovarian cancer. N Engl J Med 2011;365:2484–96. 28 Naik R, Galaal K, Alagoda B, Katory M, Mercer-Jones M, Farrel R. Surgical training in gastrointestinal procedures within a UK gynaecological oncology subspecialty programme. BJOG 2010;117:26–31. 29 Naik R, Edmondson R, Galaal K, Hatem M, Godfrey K. A statement for extensive primary cytoreductive surgery in advanced ovarian cancer. Br J Obstet Gynaecol 2008;115:1713–14.
7
Radiological predictors of cytoreductive outcomes in patients with advanced ovarian cancer J Borley,a C Wilhelm-Benartzi,a J Yazbek,b R Williamson,b,c N Bharwani,b,c V Stewart,b,c I Carson,b E Hird,b A McIndoe,b A Farthing,a,b S Blagden,a,b S Ghaem-Maghamia,b a Department of Surgery and Cancer, Imperial College London, London, UK b West London Gynaecology Cancer Centre, Imperial College NHS Trust, London, UK c Department of Radiology, Imperial College Healthcare NHS Trust, London, UK Correspondence: Dr S Ghaem-Maghami, 2nd floor Hamm House, Hammersmith Hospital, Du Cane Road, London W12 0NN, UK. Email [email protected]
Accepted 10 May 2014. Published Online 5 August 2014.
Objective To assess site of disease on preoperative computed
tomography (CT) to predict surgical debulking in patients with ovarian cancer. Design Two-phase retrospective cohort study. Setting West London Gynaecological Cancer Centre, UK. Population Women with stage 3 or 4, ovarian, fallopian or
primary peritoneal cancer undergoing cytoreductive surgery. Methods Preoperative CT images were reviewed by experienced radiologists to assess the presence or absence of disease at predetermined sites. Multivariable stepwise logistic regression models determined sites of disease which were significantly associated with surgical outcomes in the test (n = 111) and validation (n = 70) sets. Main outcome measures Sensitivity and specificity of CT in
predicting surgical outcome. Results Stepwise logistic regression identified that the presence of
mesentery and small-bowel mesentery, and infrarenal para-aortic nodes were associated with debulking status. Logistic regression determined a surgical predictive score which was able to significantly predict suboptimal debulking (n = 94, P = 0.0001) with an area under the curve (AUC) of 0.749 (95% confidence interval [95% CI]: 0.652, 0.846) and a sensitivity of 69.2%, specificity of 71.4%, positive predictive value of 75.0% and negative predictive value of 65.2%. These results remained significant in a recent validation set. There was a significant difference in residual disease volume in the test and validation sets (P < 0.001) in keeping with improved optimal debulking rates. Conclusions The presence of disease at some sites on
preoperative CT scan is significantly associated with suboptimal debulking and may be an indication for a change in surgical planning. Keywords Computed tomography, ovarian cancer, preoperative assessment, surgical debulking.
lung metastasis, pleural effusion, deposits on the large-bowel Please cite this paper as: Borley J, Wilhelm-Benartzi C, Yazbek J, Williamson R, Bharwani N, Stewart V, Carson I, Hird E, McIndoe A, Farthing A, Blagden S, Ghaem-Maghami S. Radiological predictors of cytoreductive outcomes in patients with advanced ovarian cancer. BJOG 2014; DOI: 10.1111/1471-0528. 12992.
Introduction The management of women with epithelial ovarian cancer (EOC) remains one of the greatest challenges for the gynaecological oncology community. The majority of women present in advanced stage, eventually develop chemotherapy resistance and ultimately succumb to the disease. The volume of residual disease following primary debulking surgery remains the single best predictive factor in survival,1
ª 2014 Royal College of Obstetricians and Gynaecologists
with much recent emphasis placed on achieving complete macroscopic debulking (defined as the absence of any macroscopic residual disease).2–4 Supraradical debulking surgery (such as splenic and hepatic resection, diaphragm stripping and peritonectomy) is often required to achieve this. Results from multicentre randomised clinical trials (RCTs) suggest that UK debulking rates are inferior to those of other countries,5,6 with significantly lower numbers for large-bowel resection, pelvic and para-aortic
1
Borley et al.
lymphadenectomy when compared with some other countries. The site of disease prior to surgery has been shown to be a major predictor for complete tumour resection, with patients having disease at surgical locations of the umbilical, left flank and epigastrium regions having the highest risk of incomplete resection.7 It therefore appears logical that efforts to determine tumour burden preoperatively through imaging modalities such as computed tomography (CT), magnetic resonance imaging (MRI) and positron emission tomography (PET) would be beneficial and inform the surgeon of the likelihood of optimal versus suboptimal debulking and the need for bowel resection. This, in turn, would translate into better preoperative counselling and healthcare planning through improvements in operating list scheduling and specialised staff allocation. It may also allow improved identification of those who may benefit from neoadjuvant chemotherapy and delayed debulking instead of suboptimal primary debulking. Previous studies have examined the use of CT imaging to predict surgical success,8–17 but these studies relate to small patient numbers, changing definitions of optimal debulking and out-of-date debulking rates. Two more recent studies have tried to validate the previous studies, but with limited success.12,18 One of the main arguments in the failure to validate previous findings is that factors which accounted for unresectable disease 10 years ago are now irrelevant, as surgical paradigms change and debulking rates improve. Thus, the aim of this study was to assess the role of CT in the preoperative prediction of debulking outcome and requirement for bowel surgery in patients with advanced EOC in the UK. This was performed and validated with current data which are representative of surgical practices and debulking rates in the UK today.
Methods
CT scans were assessed against criteria previously reported as potential indicators for resectability by experienced gynaecological radiologists,9–12,14,16,19 plus the addition of our own anecdotal experience. The radiologists were blind to the surgical outcome of each patient. CT scans were from a variety of referring institutions. All scans were performed using thin slice multidetector CT and contrast enhancement. All scans were reconstructed in axial sections with the option for coronal section post-processing for radiological review. Images were read by three consultant radiologists who were core members of the Gynaecology–Oncology Multidisciplinary Team. In order to allow for reproducibility, disease was scored as being present or absent with the cut-off for the identification of disease being consistent with RECIST (Response Evaluation Criteria In Solid Tumours) 1.120 indicators of measurable disease, i.e. 10 mm being the minimum size for a positive result. Lymph nodes were measured in the short axis. Absolute measurements were not included. The following regions were assessed for disease: diaphragm/superior surface of the liver, lateral surface of the liver, liver parenchyma, inferior surface of the liver/porta hepatis, gastrosplenic ligament, lesser sac, large-bowel serosa, small-bowel serosa, large-bowel mesentery, small-bowel mesentery, pelvis/adnexa. Lymphadenopathy with a short axis >10 mm was assessed in the following regions: pericardiac, porta, small- and large-bowel mesentery, suprarenal para-aortic/aortocaval, infrarenal para-aortic, pelvic, inguinal. Non-measurable disease in the form of ascites on more than four axial sections and lung nodules of >7 mm21 were also recorded. The assessment of inter-observer variation did not form part of this study as, with the use of RECIST assessments, the cut-off size of 10 mm means that detection of the disease is consistent,20 with previous studies demonstrating intra- and inter-observer variation to be low.22,23 All studies included in the preliminary analysis were adequate for review.
Data collection
Statistics
This study was approved by Imperial College NHS Trust (reference 1612) with no ethical objections to the study because of the retrospective nature of the data collection. Clinical data were collected retrospectively from the West London Gynaecological Cancer Centre database on patients with a diagnosis of stage 3 and 4 ovarian, fallopian and primary peritoneal cancer who underwent surgery. Data collected included age at surgery, histological subtype, stage and grade, timing of debulking surgery, details of operation received, including any surgery to the large or small bowel, and residual disease status. One-hundred and eleven patients undergoing treatment at the centre from August 2001 to June 2010 were randomly selected for the test set, and 70 patients undergoing treatment from June 2010 to December 2012 for the validation set. Their preoperative
The aim of this work was to determine which of the CT anatomical findings were associated with debulking success and requirement for bowel resection. Statistical analysis was performed in R (version 2.15). Permutation Kruskal–Wallis and Pearson chi-squared tests with 10 000 permutations were used to determine whether surgical or clinical characteristics were significantly different from one another in the test and validation datasets. A multivariable stepwise logistic regression model was first run associating debulking status, coded as suboptimal debulking (defined as residual disease > 10 mm) versus optimal debulking (residual disease ≤ 10 mm), against all available CT anatomical measures and age to find the optimal model based on statistical model fitting using the Akaike information criterion (AIC) (stepAIC MASS package R). A logistic regression was then
2
ª 2014 Royal College of Obstetricians and Gynaecologists
Radiological predictors in ovarian cancer
Box 1 Logistic regression model for prediction of surgical success and a worked example
If disease present, score = 1, or absent, score = 0, at each individual site. A = Lung metastasis B = Pleural effusion C = Large-bowel mesentery D = Small-bowel mesentery E = Inferior renal para-aortic lymph node
Log score = –1.15 + 16.394(A) + 1.9150(B) + 1.5965 (C) + 2.3020(D) + 1.0883(E) Log score = n Score = exp (n) A score of >1.563 indicates a high risk of suboptimal debulking; a score of 10 46 (41.4) Unresectable disease 13 (11.7) Optimal debulking 52 (46.8) (10 mm) Small bowel mesentry (>10 mm) Infrarenal paraaortic lymph node
1.15 16.394 1.915 1.597 2.302 1.088
(0.48) (1686.73) (0.77) (0.75) (1.17) (0.67)
improved from 46.8% in the test set to 75.7% in the validation set, and total macroscopic debulking from 30.6% to 54.3%, respectively. In order to establish whether findings on preoperative CT scans were associated with debulking outcome (optimal versus suboptimal debulking), a stepwise logistic regression model was performed to find the best model fit. Using this model, the presence of lung metastasis >7 mm, pleural effusion, deposits of >10 mm on the large-bowel mesentery and small-bowel mesentery, and infrarenal para-aortic nodes were associated with debulking status. A greater proportion of patients who were suboptimally debulked had disease in these sites relative to those optimally debulked. In the suboptimal group, 5.8% had lung metastasis, 38.5% had pleural effusions, 36.5% had large-bowel mesentery deposits, 21.2% had small-bowel mesentery deposits and 26.9% had inferior renal para-aortic lymph nodes relative to 0%, 14.3%, 16.7%, 4.8% and 21.4%, respectively, in the optimally debulked group (Table 2). The effect estimates (beta coefficients) from these covariates were calculated using logistic regression modelling (Table 2), which enabled a surgical predictive score to be generated for each individual. Our continuous predictive score was able to predict suboptimal debulking with an AUC of 0.749 (95% confidence interval [95% CI]: 0.652, 0.846) (Figure 1A). Patients were categorised into being above and below the median score (median, 1.563). A predictive score above the median was able to significantly predict suboptimal debulking (odds ratio [OR] = 5.62, n = 94, P = 0.0001) with a sensitivity of 69.2%, specificity of 71.4%, positive predictive value (PPV) of 75.0% and negative predictive value (NPV) of 65.2%. Box 1 illustrates how this logistical regression model is used with an example. Our predictive score was successfully validated in the validation set. As expected, a higher burden of disease was
4
0 6 7 2 9
(0.0) (14.3) (16.7) (4.8) (21.4)
Suboptimal debulking n = 52 (%)
3 20 19 11 14
(5.8) (38.5) (36.5) (21.2) (26.9)
Validation set (n = 70) Optimal debulking n = 53 (%)
4 8 24 5 9
(7.5) (15.1) (45.3) (9.4) (17.0)
Suboptimal debulking n = 17 (%)
2 7 13 4 4
(11.8) (41.2) (76.5) (23.5) (23.5)
again observed in the suboptimally debulked group, with more patients having disease on each site relative to those in the optimally debulked group (Table 2). The previous predictive score was used, as before from the model beta coefficients (Box 1), to generate an individual risk score for each case, and these were also grouped by the individual score being above or below the test set median score. In the validation set, a predictive score above the median was able to predict suboptimal debulking with a sensitivity of 64.7% and specificity of 67.9%. The continuous score generated an AUC value of 0.721 (95% CI: 0.594, 0.847) (Figure 1B). We also tested the ability of the predictive score to predict total macroscopic debulking (0 mm residual disease) relative to debulking with residual disease of ≥1 mm. This had a sensitivity of 65.6% and a specificity of 80% in the test set, and a sensitivity of 50.0% and a specificity of 68.4% in the validation set. Lastly, we used another stepwise logistic regression model to determine which covariates were associated with requirement for bowel resection at surgery. Twenty of the 111 patients (18%) had bowel resection as part of their debulking surgery. The model found that age, deposits on the large- and small-bowel mesentery, large-bowel serosa, and lateral and inferior surface of the liver, mesenteric or portal nodes and inferior renal para-aortic nodes were associated with requirement for bowel surgery. The beta coefficients were generated (data not shown) to provide an individual risk score for each patient. This score, however, was not found to be statistically significantly associated with bowel surgery in logistic regression models when the individual scores were used (n = 108, P = 0.789). There was, however, a trend towards a significant association with bowel surgery in the dichotomised model where the patients were coded depending on a score above or below the median (OR = 2.77, n = 108, P = 0.0533).
ª 2014 Royal College of Obstetricians and Gynaecologists
Radiological predictors in ovarian cancer
A
B
Figure 1. Receiver operating characteristic (ROC) curve for the logistic regression score to predict surgical debulking. (A) Test set. (B) Validation set. AUC, area under the curve.
Discussion
ascites and bowel mesentery involvement were often associated with suboptimal debulking status. We found that the presence of lung metastasis, pleural effusion, deposits on the large-bowel mesentery and small-bowel mesentery, and infrarenal para-aortic nodes were associated with suboptimal debulking in our patient group. Our model predicted suboptimal debulking with a sensitivity of 64.7–69.2% and specificity of 67.9–71.4% in two independent patient sets (AUC = 0.721–0.749). Why these particular sites of disease are more strongly associated with suboptimal debulking than others is unclear. It seems logical that deposits of >10 mm on the large- and small-bowel mesentery may be associated with higher rates of suboptimal debulking if disease at this site is so extensive that disease clearance would require multiple (or complete) bowel resection. The involvement of other sites may simply be an indication of extensive and biologically aggressive disease, making optimal resection within the abdomen less likely. As the debate with regard to tumour biology determining surgical outcomes persists,24,25 there may be inherent biological factors within these tumours which dictate specific spread and invasion to these sites. The appearance of infrarenal para-aortic nodes over other nodal disease being strongly associated with suboptimal debulking is also of curiosity. Although there was a larger proportion of cases with mesenteric or portal vein nodes in the suboptimal group than in the optimal group (12/52 [23.1%] versus 6/42 [14.3%]), the model did not find this to be an indicator. There were only five cases of suprarenal para-aortic node involvement in total, and so we suggest that the study was not adequately powered to determine a specific association for this disease site. In contrast with previous studies, we did not find that upper abdominal disease was associated with suboptimal debulking in the test set. This may be a result of our increasing use of extensive upper abdominal resections and diaphragm/peritoneal stripping in the centre. We found a non-significant trend towards increased risk of requiring bowel resection in patients who had disease present at sites over the large- and small-bowel mesentery, large-bowel serosa, and lateral and inferior surface of the liver, ascites, and mesenteric or portal nodes and inferior renal para-aortic nodes. The relationship between disease on the bowel and requirement for resection seems obvious; however, the volume of disease, number of deposits, specific location and invasiveness of the tumour into the bowel will impact on the need for resection. Thus, our study was not adequately designed to address this issue.
Main findings
Strengths and limitations
Much emphasis has been placed on finding a preoperative method to accurately predict the debulking status of suspected EOC. In previously published studies,7–18 the presence of disease in the upper abdomen, large volume of
Our study uses one of the largest datasets published and robust statistical methods to determine a prediction model. It is also the first study in more than 10 years that uses a current validation set. Stepwise logistic regression was used
ª 2014 Royal College of Obstetricians and Gynaecologists
5
Borley et al.
to determine the best predictive model for surgical debulking from which the risk score was determined. This method allows the creation of a small, readily interpretable model containing the most important predictors of surgical debulking from a list of CT anatomical findings. In addition, as we used RECIST 1.1 indicators of measurable disease (disease ≥ 10 mm), this ensures that the results are consistent with limited inter-observer variability. A potential limitation of this method lies in the instability of the result when adding or removing patients, and the statistical power of the study may be insufficient to select true predictors. However, given that our model was validated in an independent set, we believe that it is robust to these potential problems. One of the key issues in predicting debulking is the ever-changing surgical goal with inherent variation on what accounts for truly resectable disease. Factors that reduce the likelihood of achieving successful cytoreduction previously cited include: diffuse small-bowel involvement, infiltration into the porta hepatis and diffuse involvement of the right hemidiaphragm, resulting in fixation of the liver2 or large liver metastasis.24 Patient age, comorbidities and intensive care availability will also impact on outcome. Furthermore, what was deemed as unresectable disease in the year 2000 has now become resectable through the development of, and emphasis on supraradical surgery.26 This means that previous models of prediction have become obsolete. Our findings were validated in an independent dataset from patients recently treated at the centre, suggesting that it is more current and clinically relevant. During the time course of this study, we achieved a significant improvement in surgical debulking rates, with an improvement of optimal debulking from 46.8% to 75.7% and complete debulking from 30.4% to 54.3%. The model remains significant despite this change, indicating that these prediction markers are likely to be truly representative of an inability to achieve optimal debulking irrespective of surgical effort. We believe that our debulking rates at the centre are in line with those of other UK-based cancer centres, both historically5,27,28 and in the current era of supraradical surgery.29 This adds evidence that our prediction model can be translated to other centres that perform similar numbers and types of surgery. Current evidence now highlights the need to perform complete macroscopic debulking, whenever possible. However, at the time of surgery in our test set, optimal debulking was the primary goal and thus we used this as our primary outcome measure. In subset analysis, the model performed less well when predicting any amount of residual disease versus complete debulking. This demonstrates the limitations of CT scans to define disease which is present, but ≤10 mm, and highlights that the statistical model should not be used interchangeably. It appears that disease
6
that predicts suboptimal from optimal surgery may be different from that which predicts complete debulking. We strongly believe that it is still clinically relevant to determine the likelihood of suboptimal debulking prior to surgery, especially when making multidisciplinary team meeting treatment decisions with regard to upfront versus delayed debulking surgery.
Interpretation This statistical model and score can be readily calculated and interpreted for any patient who undergoes staging CT preoperatively for ovarian cancer (see Box 1). In the test set, only 25% (12/48) of patients with a high risk score ultimately achieved optimal debulking, and 39% (11/27) of patients in the validation set with a high risk score achieved optimal debulking. We argue that patients with a high risk score may be stratified into receiving neoadjuvant chemotherapy and interval debulking. This approach will not compromise survival and may lead to an increase in debulking rates.6
Conclusions Clearly, a well-designed prospective RCT is needed to address the issues raised in the Discussion section and to further validate our findings. National Institute for Health and Clinical Excellence (NICE) guidance28 also recommends further research into the evaluation of the role of MRI in addition to CT in the staging of ovarian cancer and prediction of surgical success, and this should also be incorporated into future studies. The relevance of being able to correctly predict surgical outcome has not been formally evaluated to date. It is therefore necessary to determine how an accurate prediction of surgical outcomes influences patient survival, quality of life and the impact on NHS service planning and provision. Only then will these markers of suboptimal debulking truly be able to be incorporated into the care and management of patients with advanced ovarian cancer.
Disclosure of interests The authors report no declaration of interest.
Contribution to authorship JB and SGM wrote the manuscript. JY, RW, AMI, AF, SB and SGM were responsible for the conception, design and critique of the study. JB, IC, EH and JY performed the data collection. JB and CWB performed the statistical analysis. NB, VS and RW reported the CT scans. All authors reviewed the manuscript and contributed to the final submission.
Details of ethics approval Ethical approval from Imperial College NHS Trust (ref. 1612).
ª 2014 Royal College of Obstetricians and Gynaecologists
Radiological predictors in ovarian cancer
Funding JB is funded by Cancer Research UK/National Institute for Health Research (CRUK/NIHR). CWB is funded by CRUK.
Acknowledgements Conduct of the study was supported by the Experimental Cancer Medicine Centre, the National Institute for Health Research (NIHR), Biomedical Research Centre and the Ovarian Cancer Action Centre, Imperial College London, UK. &
References 1 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;20:1248–59. 2 Pomel C, Barton DP, McNeish I, Shepherd J. A statement for extensive primary cytoreductive surgery in advanced ovarian cancer. BJOG 2008;115:808–10. 3 Elattar A, Bryant A, Winter-Roach BA, Hatem M, Naik R. Optimal primary surgical treatment for advanced epithelial ovarian cancer. Cochrane Database Syst Rev 2011;8:CD007565. 4 Redman C, Duffy S, Bromham N, Francis K, Group GD. Recognition and initial management of ovarian cancer: summary of NICE guidance. BMJ 2011;342:d2073. 5 Crawford SC, Vasey PA, Paul J, Hay A, Davis JA, Kaye SB. Does aggressive surgery only benefit patients with less advanced ovarian cancer? Results from an international comparison within the SCOTROC-1 Trial. J Clin Oncol 2005;23:8802–11. 6 Vergote I, Trope CG, Amant F, Kristensen GB, Ehlen T, Johnson N, et al. Neoadjuvant chemotherapy or primary surgery in stage IIIC or IV ovarian cancer. N Engl J Med 2010;363:943–53. 7 Fotopoulou C, Richter R, Braicu EI, Schmidt SC, Lichtenegger W, Sehouli J. Can complete tumor resection be predicted in advanced primary epithelial ovarian cancer? A systematic evaluation of 360 consecutive patients. Eur J Surg Oncol 2010; 36:1202–10. 8 Ibeanu OA, Bristow RE. Predicting the outcome of cytoreductive surgery for advanced ovarian cancer: a review. Int J Gynecol Cancer 2010;21:S1–11. 9 Nelson BE, Rosenfield AT, Schwartz PE. Preoperative abdominopelvic computed tomographic prediction of optimal cytoreduction in epithelial ovarian carcinoma. J Clin Oncol 1993;11:166–72. 10 Meyer JI, Kennedy AW, Friedman R, Ayoub A, Zepp RC. Ovarian carcinoma: value of CT in predicting success of debulking surgery. Am J Roentgenol 1995;165:875–8. 11 Bristow RE, Duska LR, Lambrou NC, Fishman EK, O’Neill MJ, Trimble EL, et al. A model for predicting surgical outcome in patients with advanced ovarian carcinoma using computed tomography. Cancer 2000;89:1532–40. 12 Axtell AE, Lee MH, Bristow RE, Dowdy SC, Cliby WA, Raman S, et al. Multi-institutional reciprocal validation study of computed tomography predictors of suboptimal primary cytoreduction in patients with advanced ovarian cancer. J Clin Oncol 2007;25:384–9.
ª 2014 Royal College of Obstetricians and Gynaecologists
13 Dowdy SC, Mullany SA, Brandt KR, Huppert BJ, Cliby WA. The utility of computed tomography scans in predicting suboptimal cytoreductive surgery in women with advanced ovarian carcinoma. Cancer 2004;101:346–52. 14 Qayyum A, Coakley FV, Westphalen AC, Hricak H, Okuno WT, Powell B. Role of CT and MR imaging in predicting optimal cytoreduction of newly diagnosed primary epithelial ovarian cancer. Gynecol Oncol 2005;96:301–6. 15 Salani R, Axtell A, Gerardi M, Holschneider C, Bristow RE. Limited utility of conventional criteria for predicting unresectable disease in patients with advanced stage epithelial ovarian cancer. Gynecol Oncol 2008;108:271–5. 16 Risum S, Høgdall C, Loft A, Berthelsen AK, Høgdall E, Nedergaard L, et al. Prediction of suboptimal primary cytoreduction in primary ovarian cancer with combined positron emission tomography/ computed tomography – a prospective study. Gynecol Oncol 2008;108:265–70. 17 Jung DC, Kang S, Kim MJ, Park SY, Kim HB. Multidetector CT predictors of incomplete resection in primary cytoreduction of patients with advanced ovarian cancer. Eur Radiol 2010;20:100–7. 18 Gemer O, Lurian M, Gdalevich M, Kapustian V, Piura E, Schneider D, et al. A multicenter study of CA 125 level as a predictor of non-optimal primary cytoreduction of advanced epithelial ovarian cancer. Eur J Surg Oncol 2005;31:1006–10. 19 Byrom J, Widjaja E, Redman CW, Jones PW, Tebby S. Can pre-operative computed tomography predict resectability of ovarian carcinoma at primary laparotomy? BJOG 2002;109:369–75. 20 Eisenhauer EA, Therasse P, Bogaerts J, Schwartz LH, Sargent D, Ford R, et al. New response evaluation criteria in solid tumours: revised RECIST guideline (version 1.1). Eur J Cancer 2009;45:228–47. 21 Swensen SJ, Jett JR, Hartman TE, Midthun DE, Mandrekar SJ, Hillman SL, et al. CT screening for lung cancer: five-year prospective experience. Radiology 2005;235:259–65. 22 Muenzel D, Engels HP, Bruegel M, Kehl V, Rummeny EJ, Metz S. Intra- and inter-observer variability in measurement of target lesions: implication on response evaluation according to RECIST 1.1 Radiol Oncol 2012;46:8–18. 23 Sun JM, Ahn MJ, Park MJ, Yi JH, Kim TS, Chung MJ, et al. Accuracy of RECIST 1.1 for non-small cell lung cancer treated with EGFR tyrosine kinase inhibitors. Lung Cancer 2010;69:105–9. 24 Chi DS, Schwartz PE. Cytoreduction vs. neoadjuvant chemotherapy for ovarian cancer. Gynecol Oncol 2008;111:391–9. 25 Borley J, Wilhelm-Benartzi C, Brown R, Ghaem-Maghami S. Does tumour biology determine surgical success in the treatment of epithelial ovarian cancer? A systematic literature review. Br J Cancer 2012;107:1069–74. 26 Chi DS, Eisenhauer EL, Zivanovic O, Sonoda Y, Abu-Rustum NR, Levine DA, et al. Improved progression-free and overall survival in advanced ovarian cancer as a result of a change in surgical paradigm. Gynecol Oncol 2009;114:26–31. 27 Perren TJ, Swart AM, Pfisterer J, Ledermann JA, Pujade-Lauraine E, Kristensen G, et al. A phase 3 trial of bevacizumab in ovarian cancer. N Engl J Med 2011;365:2484–96. 28 Naik R, Galaal K, Alagoda B, Katory M, Mercer-Jones M, Farrel R. Surgical training in gastrointestinal procedures within a UK gynaecological oncology subspecialty programme. BJOG 2010;117:26–31. 29 Naik R, Edmondson R, Galaal K, Hatem M, Godfrey K. A statement for extensive primary cytoreductive surgery in advanced ovarian cancer. Br J Obstet Gynaecol 2008;115:1713–14.
7
