Nomogram for prediction of prognosis in patients with initially unresectable colorectal liver metastases

Original article Nomogram for prediction of prognosis in patients with initially unresectable colorectal liver metastases K. Imai 1,2,5, M.-A. Allard 1,2,4, C. Castro Benitez 1,2,4,E.Vibert 1,3,4, A. Sa Cunha 1,2,4,D.Cherqui 1,3,4, D. Castaing 1,3,4, H. Bismuth 1,H.Baba 5 and R. Adam 1,2,4 1 Centre Hépato-Biliaire, Assistance Publique Hôpitaux de Paris, Hôpital Universitaire Paul Brousse, 2 Institut National de la Santé et de la Recherche Médicale (INSERM), Unité 935, 3 INSERM, Unité 785, and 4 Université Paris-Sud, Villejuif, France, and 5 Department of Gastroenterological Surgery, Graduate School of Life Sciences, Kumamoto University, Kumamoto, Japan Correspondence to: Professor R. Adam, Centre Hépato-Biliaire, AP-HP, Hôpital Universitaire Paul Brousse, 12 Avenue Paul Vaillant Couturier, F-9484 Villejuif, France (e-mail: rene.adam@pbr.aphp.fr) Background: Although recent advances in surgery and chemotherapy have increasingly enabled hepatectomy in patients with initially unresectable colorectal liver metastases (CRLM), not all such patients benefit from surgery. The aim of this study was to develop a nomogram to predict survival after hepatectomy for initially unresectable CRLM. Methods: Patients with initially unresectable CRLM treated with chemotherapy followed by hepatectomy between 199 and 212 were included in the study. A nomogram to predict survival was developed based on a multivariable Cox model. The predictive performance of the model was assessed according to the C-statistic, Kaplan Meier curve and calibration plots. Results: Of a total of 439 patients, liver and globally completed surgery was achieved in 38 (86 6 per cent) and 335 (76 3 per cent) patients respectively. The 5-year overall and disease-free survival rates were 39 9 and 1 per cent respectively. Based on the Cox model, the following five factors were selected for the nomogram and assigned specific scores: node-positive primary, 5; more than six metastases at hepatectomy, 7; carbohydrate antigen 19 9 level at hepatectomy above 37 units/ml, 1; disease progression during first-line chemotherapy, 9; and presence of extrahepatic disease, 4. The model achieved relatively good discrimination and calibration, with a C-statistic of 66. The overall survival rate for patients with a score greater than 16 was significantly worse than that for patients with a score of 16 or less (5-year survival rate 4 versus 46 3 per cent respectively; P < 1). Conclusion: The nomogram facilitates personalized assessment of prognosis for patients with initially unresectable CRLM treated with chemotherapy and with planned resection. Paper accepted 5 November 215 Published online 18 January 216 in Wiley Online Library (www.bjs.co.uk). DOI: 1.12/bjs.173 Introduction The liver is the most common site of metastasis of colorectal cancer, and nearly half of patients with colorectal cancer develop liver metastasis at some point during the course of their disease 1,2. Although surgical resection is currently the only treatment of choice for cure, patients with colorectal liver metastases (CRLM) suitable for surgical resection at the time of diagnosis account for less than 3 per cent of subjects 3 5. For patients with initially unresectable CRLM, the use of modern systemic chemotherapy with or without biological agents enables surgical resection with curative intent by downsizing the tumours 6. This so-called conversion chemotherapy is becoming increasingly effective in providing a sufficient response to allow subsequent surgical resection, and offers a long-term survival benefit 4,7 1.A recent systematic review 11 reported the curative resection rate as 22 5 per cent and the 5-year survival rate with conversion chemotherapy followed by hepatectomy as 33 5 per cent, with a median survival time of 45 months. However, even with surgery, some patients may not benefit because of early recurrence 12. To assess factors affecting the long-term outcomes of patients with initially unresectable CRLM after hepatectomy on an individual basis, it is crucial to offer surgery only to those expected to obtain a real benefit. Combining several prognostic factors in a scoring system would be helpful for clinical decision-making. 216 BJS Society Ltd BJS 216; 13: 59 599

Nomogram for initially unresectable colorectal liver metastases 591 Nomograms, a statistical predictive model, were recently developed and validated for several types of cancer, including CRLM 13 16. Such nomograms create a simple graphical representation of a statistical predictive model that generates a numerical probability of a clinical event 17. Nomograms compare favourably with the traditional scoring system, and have thus been proposed as an alternative, or even a novel, standard tool 17 19. The objective of the present study was to develop a nomogram based on prognostic factors to predict long-term survival following hepatectomy in patients with initially unresectable CRLM. Methods Patients who underwent hepatectomy for CRLM between 199 and 212 at Hôpital Universitaire Paul Brousse, Villejuif, France, were identified from a prospectively maintained database. Only patients with unresectable tumours assessed according to well defined selection criteria were included 7. These criteria were based mainly on the impossibility of performing curative surgery with a single hepatectomy leaving at least 3 per cent of non-tumoral liver parenchyma, or 4 per cent when the patient received prolonged chemotherapy (more than 8 courses). The final decision to perform surgery after chemotherapy was made during a multidisciplinary meeting including surgeons, medical oncologists and radiologists. Preoperative management The response to preoperative chemotherapy was evaluated by means of CT after every four cycles of treatment according to the Response Evaluation Criteria in Solid Tumours (RECIST) 2. During chemotherapy, the need for liver surgery was reconsidered at a multidisciplinary meeting each time a response to chemotherapy was observed. If the estimated liver volume was thought to be insufficient, portal vein embolization was performed to induce compensatory hypertrophy of the remnant liver 21. Surgical treatment The objective of surgery was to resect all detectable lesions with tumour-free margins. If obtaining a tumour-free margin was not possible because of contact with major vascular or biliary structures, resection was still indicated provided that all tumours could be resected macroscopically. In patients with multiple bilateral tumours that could not be resected via single hepatectomy, hepatectomy combined with local ablation therapy (cryotherapy or radiofrequency ablation) or two-stage hepatectomy was performed 22,23. The presence of extrahepatic tumours was not considered a contraindication to hepatectomy if the lesions were limited and resectable. In principle, extrahepatic disease located in the abdominal cavity was resected at the same time as hepatectomy. For extrahepatic disease located outside the abdomen, resection was performed 2 3 months after hepatectomy if the disease remained controlled with interval chemotherapy. If the disease was not controlled with chemotherapy, resection was generally postponed and systemic chemotherapy restarted using another regimen; the response was then re-evaluated to assess the possibility of surgery in a multidisciplinary meeting. Postoperative management After treatment, all patients had regular follow-up examinations with imaging and estimation of tumour markers such as serum carcinoembryonic antigen (CEA) and carbohydrate antigen (CA) 19 9. Postoperative chemotherapy was recommended routinely, using the same protocol as that applied before surgery. Recurrence was treated surgically only when the overall strategy was considered to be potentially curative. Statistical analysis Survival data were analysed using the Kaplan Meier method and compared with the log rank test. Overall survival (OS) was calculated from the date of hepatectomy until death or last follow-up. Disease-free survival (DFS) was calculated from the date of hepatectomy or last potentially curative surgery for concomitant extrahepatic disease, if present, to the date of recurrence or death. Univariable analysis of prognostic factors for OS was performed using categorical variables. Cut-off values for continuous variables were determined based on the C-statistic, with a Cox regression model for survival data including censored patients. The estimated cut-off values for each variable were: tumour number, 1 1; tumour size, 2 8 mm; CEA level, 5 (higher limit of normal range), 1, 2, 5, 1, 2, 3, 4 ng/ml; CA19-9, 37 (higher limit of normal range), 5, 1, 2, 3, 4 units/ml; chemotherapy cycles, 4, 6, 8 2; and chemotherapy lines, 1, 2, 3. Variables for which the P value in univariable analysis was less than 5 were included in the multivariable Cox regression model. A nomogram was created based on the results of the multivariable analysis, and final model selection for the nomogram was performed using a stepwise backward elimination procedure with a threshold P < 5. The predictive performance of the nomogram was assessed by evaluating the degree of discrimination with the C-statistic, plotting Kaplan Meier curves over

592 K. Imai, M.-A. Allard, C. Castro Benitez, E. Vibert, A. Sa Cunha, D. Cherqui et al. the quartiles of the nomogram-predicted score and examining calibration plots with bootstrapped samples. All statistical analyses were performed using the JMP software program (SAS Institute, Cary, North Carolina, USA) and R version 3 1.1 (http://www.r-project.org). Results Of 1163 consecutive patients who underwent initial hepatectomy for CRLM between 199 and 212, 444 (38 2 per cent) were identified as having undergone hepatectomy for initially unresectable CRLM. Five patients (1 1 per Table 1 Demographic and clinical characteristics of study patients No. of patients* (n = 439) Age (years) 58 6(1 5) (6; 31 84) Sex ratio (M : F) 26 : 179 Primary T category (n = 382) T1 2 44 (11 5) T3 4 338 (88 5) Primary N category (n = 382) N 19 (28 5) N1 2 273 (71 5) Primary tumour location (n = 438) Colon 338 (77 2) Rectum 1 (22 8) Primary tumour in situ at hepatectomy 92 (21 ) Primary tumour resection Before hepatectomy 347 (79 ) During hepatectomy 52 (11 8) After hepatectomy 27 (6 2) Eventually not resected 13 (3 ) Timing of liver metastasis Synchronous 345 (78 6) Metachronous 94 (21 4) Tumour no. at diagnosis 6 9(5 8) (5; 1 35) Tumour size at diagnosis (mm) 55 6(37 6) (5; 1 3) Localization of liver metastases Unilobar 94 (21 4) Bilobar 345 (78 6) Main cause of initial unresectability Multinodular 21 (47 8) Large size 93 (21 2) Difficult vascular location 92 (21 ) Extrahepatic disease 44 (1 ) CEA at diagnosis (ng/ml) 63 4(267 7) (52 3; 4 37 ) CA19-9 at diagnosis (units/ml) 166 9(649 6) (7; 4 6 ) Tumour no. at hepatectomy 6 3(5 4) (5; 33) Tumour size at hepatectomy (mm) 41 5(29 5) (35; 3) CEA at hepatectomy (ng/ml) 75 3(364 9) (6; 4 672) CA19-9 at hepatectomy (units/ml) 228 1(1573 2) (18; 4 27 348) Concomitant extrahepatic disease 148 (33 7) *With percentages in parentheses unless indicated otherwise; values are mean(s.d.) (median; range). Diagnosed before, during, or within 3 months of colorectal resection. CEA, carcinoembryonic antigen; CA, carbohydrate antigen. cent) died from postoperative complications within 9 days and were excluded from this study, leaving 439 patients eligible for the final analysis. The demographic and clinical characteristics of these patients are depicted in Table 1. Ninety-two patients (21 per cent) had a primary tumour present at hepatectomy. Primary tumour resection was performed before hepatectomy in 347 patients (79 per cent), duringin52(11 8 per cent) and after in 27 (6 2 per cent). Thirteen patients (3 per cent) were eventually unable to undergo primary tumour resection, because the liver resection would not have been curative (6 patients), concomitant extrahepatic disease was not resected (4) or there was early liver recurrence after hepatectomy (3). Liver metastasis was diagnosed synchronously (before, during or within 3 months after colorectal resection) in 345 patients (78 6 per cent). Concomitant extrahepatic disease was present in 148 patients (33 7 per cent). Detailed information regarding concomitant extrahepatic disease is summarized in Table S1 (supporting information). The main cause of initial unresectability was multinodular liver disease in 21 patients (47 8 per cent), large size of liver metastases in 93 (21 2 per cent), difficult vascular location Table 2 Chemotherapy and perioperative features No. of patients* (n = 439) Chemotherapy before hepatectomy 439 (1) No. of chemotherapy cycles 11 2(6 5) (1; 1 49) No. of chemotherapy lines 1 4( 7) (1; 1 5) Use of biological agents 17 (39 1) Response to first-line chemotherapy (n = 431) Complete 7 (1 6) Partial 285 (66 1) Stable disease 17 (24 8) Progressive disease 32 (7 4) Response to last-line chemotherapy (n = 432) Complete 6 (1 4) Partial 317 (73 4) Stable disease 19 (25 2) Progressive disease () Portal vein embolization 174 (39 6) Major hepatectomy ( 3 segments) 257 (58 5) Two-step approach 139 (31 7) Concomitant use of local ablation therapy 45 (1 3) Red blood cell transfusion 173 (39 4) Major complication 11 (25 1) Liver surgery completed 38 (86 6) Globally completed surgery# 335 (76 3) *With percentages in parentheses unless indicated otherwise; values are mean(s.d.) (median; range). Including conventional two-stage hepatectomy and associated liver partition with portal vein ligation for staged hepatectomy (ALPPS); including cryotherapy and radiofrequency ablation; indicates that potentially curative surgery was performed at least for liver disease; #indicates that potentially curative surgery was performed for both liver and concomitant extrahepatic disease.

Nomogram for initially unresectable colorectal liver metastases 593 No. at risk OS DFS 1 1 2 3 4 5 Time after surgery (years) 439 439 357 132 259 71 176 43 OS DFS 129 36 94 28 Fig. 1 Overall (OS) and disease-free (DFS) survival in patients who underwent hepatectomy for initially unresectable colorectal liver metastases in 92 (21 per cent) and presence of extrahepatic disease in 44 (1 per cent) (Table 1). Chemotherapy and perioperative features are summarized in Table 2. Preoperative chemotherapy was administered to all patients, with a median of 1 (range 1 5) lines and 1 (1 49) cycles. Disease progression was observed in 32 patients (7 4 per cent) during first-line chemotherapy and in no patient during last-line chemotherapy. Major hepatectomy was performedin 257 patients (58 5 per cent). Additional local ablation therapy, including radiofrequency ablation or cryotherapy, was used in 45 patients (1 3 per cent). A two-stage approach including conventional two-stage hepatectomy or associated liver partition with portal vein ligation for staged hepatectomy (ALPPS) was required in 139 patients (31 7 per cent). Liver surgery was completed (R + R1 liver resection with or without local ablation therapy) in 38 patients (86 6 per cent), and of the remaining 59 patients, in whom liver surgery was not completed, 45 dropped out of the approach of two-stage hepatectomy. When also considering patients with extrahepatic disease, globally completed surgery (potentially curative resection for both liver and extrahepatic disease) was achieved in 335 patients (76 3 per cent). Long-term outcomes The median (range) length of follow-up was 42 1 (5 251) months after the diagnosis of liver metastases and 29 5 ( 1 225) months after hepatectomy. Cumulative OS rates at 1, 3 and 5 years after hepatectomy were 89 8, 59 and 39 9 per cent respectively (median 46 1 months) (Fig. 1). From the diagnosis of liver metastasis, the cumulative OS rates at 1, 3 and 5 years after hepatectomy were 98 6, 74 3 and46 6 per cent respectively (median 56 7 months) (Fig. S1, supporting information). Cumulative DFS rates at 1, 3 and 5 years after surgery were 31 3, 11 8 and1 per cent respectively, on an intention-to-treat basis (Fig. 1). Among patients who had globally completed surgery, the cumulative DFS rates at 1, 3 and 5 years after surgery 1 Globally completed Globally not completed 1 Liver completed Liver not completed No. at risk Globally completed Globally not completed 335 14 293 62 228 31 a Survival according to global curability 1 2 3 4 5 Time after surgery (years) 158 18 121 8 92 2 1 2 3 4 5 Time after surgery (years) No. at risk Liver completed 45 33 23 15 Liver not completed 59 29 8 3 b Survival in patients without global completion 6 2 1 1 Fig. 2 a Overall survival after hepatectomy according to the potential global curability of surgery. b Overall survival after hepatectomy in patients in whom globally completed surgery was not achieved, divided into two groups: liver surgery completed but concomitant extrahepatic disease not resected, and liver surgery not completed. P < 1 for all comparisons (log rank test)

594 K. Imai, M.-A. Allard, C. Castro Benitez, E. Vibert, A. Sa Cunha, D. Cherqui et al. Table 3 Univariable analysis of factors associated with overall survival after hepatectomy n 5-year overall survival (%) P Age (years) > 5 34 38 7 59 5 99 44 Sex M 26 41 1 742 F 179 38 1 Primary T category T3 4 338 37 8 5 T1 2 44 6 Primary N category N1 2 273 35 6 15 N 19 52 3 Primary tumour location Colon 338 37 5 25 Rectum 1 47 4 Timing of liver metastasis Synchronous* 345 38 9 628 Metachronous 94 43 Tumour no. at diagnosis > 4 243 33 4 < 1 4 188 48 7 Tumour size at diagnosis (mm) > 7 14 47 4 142 7 318 37 8 Localization of liver metastasis Bilobar 345 36 6 Unilobar 94 55 CEA at diagnosis (ng/ml) > 2 249 31 4 2 2 129 49 4 CA19-9 at diagnosis (units/ml) > 1 19 3 2 31 1 142 39 4 No. of preoperative chemotherapy cycles > 1 193 32 7 3 1 238 45 8 No. of preoperative chemotherapy lines > 1 136 35 8 7 1 32 41 8 Use of biological agents Yes 17 45 2 136 No 265 37 6 Response to first-line chemotherapy Progression 32 22 < 1 Response/stable 399 4 9 Tumour no. at hepatectomy > 6 161 26 6 < 1 6 278 46 8 Tumour size at hepatectomy (mm) > 4 16 32 2 6 4 274 44 5 CEA at hepatectomy (ng/ml) > 1 175 27 5 < 1 1 258 47 7 CA19-9 at hepatectomy (units/ml) > 37 13 2 4 < 1 37 29 48 5 Concomitant extrahepatic disease Yes 148 31 5 12 No 291 44 3 Portal vein embolization Yes 174 32 6 < 1 No 265 44 4 Major hepatectomy ( 3 segments) Yes 257 4 9 898 No 182 38 7 Two-step approach Yes 139 31 4 5 No 3 43 5 Concomitant use of local ablation therapy Yes 45 4 743 No 394 39 8 Red blood cell transfusion Yes 173 32 6 75 No 266 46 Major complication Yes 11 25 3 4 No 329 44 8 *Indicates diagnosed before, during, or within 3 months of colorectal resection; including cryotherapy and radiofrequency ablation. CEA, carcinoembryonic antigen; CA, carbohydrate antigen. Log rank test.

Nomogram for initially unresectable colorectal liver metastases 595 Table 4 Multivariable cox regression analysis of factors associated with overall survival after hepatectomy Hazard ratio Primary N category (N1 2) 1 51 (1 11, 2 1) 4 Response to first-line 2 11 (1 28, 3 27) 4 chemotherapy (progression) Tumour no. at hepatectomy (> 6) 1 76 (1 3, 2 36) < 1 CA19-9 at hepatectomy 2 36 (1 73, 3 18) < 1 (> 37 units/ml) Concomitant extrahepatic disease (present) 1 42 (1 6, 1 89) 19 Values in parentheses are 95 per cent c.i. CA, carbohydrate antigen. were 41 1, 15 4 and 13 1 per cent respectively (data not shown). According to the potential curability of surgery, the 5-year OS rate after hepatectomy was significantly higher in patients who had globally completed surgery (48 4 per cent; median 56 6 months) than in those who did not (4 7 per cent; median 22 2 months) (P < 1) (Fig. 2a). From the diagnosis of liver metastasis, the 5-year OS rate was 55 per cent (median 67 4 months) in patients P who had globally completed surgery and 14 4 per cent (median 35 2 months) in those who did not (P < 1) (Fig. S2, supporting information). Among patients who did not have globally completed surgery, the 5-year OS rate for those in whom liver surgery was completed, although concomitant extrahepatic disease was not resected, was significantly higher than that for patients who did not have completed liver surgery (R2 resection) (5 versus 6 per cent respectively; median 3 1 versus 14 7months) (P < 1) (Fig. 2b). Univariable and multivariable analysis of factors associated with overall survival In the univariable analysis, 17 variables, including primary T and N categories, tumour number at diagnosis, localization of liver metastases, serum CEA and CA19-9 levels at diagnosis, preoperative chemotherapy cycles and lines, response to first-line chemotherapy, tumour number and size at hepatectomy, serum CEA and CA19-9 levels at hepatectomy, presence of concomitant extrahepatic disease, portal vein embolization, two-stage approach and Factor Primary tumour N category N N1 2 Tumour no. at hepatectomy 6 > 6 CA19-9 level at hepatectomy (units/ml) 37 > 37 Response to first-line chemotherapy Response/stabilization Progression Concomitant extrahepatic disease No Yes a Prognostic factor scores Score 5 7 1 9 4 Total points 5 years 7 3 years 5 1 year 95 1 7 5 9 3 15 5 2 5 3 1 25 75 7 3 1 35 5 b Nomogram to predict survival after hepatectomy Fig. 3 Nomogram to predict survival after hepatectomy in patients with initially unresectable colorectal liver metastases. a Multivariable Cox regression analysis identified five variables as independent prognostic factors of survival; specific scores for each variable were assigned as follows: node-positive primary, 5; tumour number at hepatectomy greater than six, 7; carbohydrate antigen (CA) 19 9 level at hepatectomy above 37 units/ml, 1; disease progression during first-line chemotherapy, 9; and presence of concomitant extrahepatic disease, 4. b Nomogram predicting 1-, 3- and 5-year survival probabilities. The sum of the scores for each variable was plotted on the total points axis (left side), and the estimated probabilities of survival at 1, 3 and 5 years were obtained by drawing a line horizontally from the plotted total point axis straight to the survival axis (right side)

596 K. Imai, M.-A. Allard, C. Castro Benitez, E. Vibert, A. Sa Cunha, D. Cherqui et al. 1 Apparent predictive accuracy Bootstrap-corrected estimates 1 Actual 3-year survival (proportion) Actual 5-year survival (proportion) 1 Nomogram-predicted 3-year survival 1 Nomogram-predicted 5-year survival a Probability of survival at 3 years b Probability of survival at 5 years Fig. 4 Calibration plots comparing nomogram-predicted and actual survival probabilities at a 3andb 5 years. The dotted line is the reference indicating ideal prediction 1 Quartile 1 ( 5) Quartile 2 (6 1) Quartile 3 (11 16) Quartile 4 ( 17) 1 Quartile 1 3 ( 16) Quartile 4 ( 17) No. at risk Q1 Q2 1 2 3 4 5 Time after surgery (years) 124 58 111 52 86 36 64 3 Q3 14 86 68 43 Q4 7 46 21 8 a Survival according to nomogram-predicted score quartiles 52 24 29 3 38 18 18 1 No. at risk Q1 3 Q4 1 2 3 4 5 Time after surgery (years) 286 7 249 46 19 21 b Quartiles 1 3 versus quartile 4 137 8 15 3 74 1 Fig. 5 Overall survival after hepatectomy according to a quartiles of the nomogram-predicted score and b quartiles 1 3 versus quartile 4. a,b P < 1 (log rank test) major complications ( grade III 24 ), were related to a decreased OS (P < 5) (Table 3). Based on the results of this univariable analysis, 14 variables were included in the subsequent multivariable Cox regression model. Five independent prognostic factors for OS were identified: node-positive primary (hazard ratio (HR) 1 51; P = 4), tumour number at hepatectomy greater than six (HR 1 76; P < 1), CA19-9 level at hepatectomy above 37 units/ml (HR 2 36; P < 1), disease progression during first-line chemotherapy (HR 2 11; P = 4) and presence of concomitant extrahepatic disease (HR 1 42; P = 19) (Table 4). Creation of a prognostic nomogram for overall survival A prognostic nomogram for OS after hepatectomy with point scales for the above five factors was constructed

Nomogram for initially unresectable colorectal liver metastases 597 subsequently (Fig. 3). Based on the multivariable Cox model, these factors were assigned a specific score as follows: node-positive primary, 5; tumour number at hepatectomy greater than six, 7; CA19-9 level at hepatectomy above 37 units/ml, 1; disease progression during first-line chemotherapy, 9; and presence of concomitant extrahepatic disease, 4 (Fig. 3a). The sum of the scores for each variable was plotted on the total points axis (left side), and the estimated probabilities of survival at 1, 3 and 5 years were obtained by drawing a line horizontally from the plotted total points axis straight to the survival axis (right side) (Fig. 3b). Total points for the scores ranged from to 35, and the C-statistic for OS prediction was 66. A calibration plot for the probability of survival at 3 years (Fig. 4a) and 5 years (Fig. 4b) demonstrated good calibration between the prediction by the nomogram and the actual observation. The survival curves stratified by quartiles of the nomogram-predicted score are shown in Fig. 5.Patients with the lowest predicted OS (quartile 4, total score above 16) exhibited substantially worse survival (5-year OS rate 4 per cent) than those in quartile 1 (total score 5; 5-year OS rate 58 per cent), quartile 2 (score 6 1; 44 per cent) and quartile 3 (score 11 16; 34 7 per cent) (Fig. 5a). OS for patients in quartile 4 was significantly worse than that for patients in quartiles 1 3 (5-year OS rate 4 versus 46 3 per cent respectively; P < 1) (Fig. 5b). Discussion In this study a nomogram was developed to predict survival for patients with initially unresectable CRLM downsized by conversion chemotherapy followed by hepatectomy. Multivariable Cox regression analysis identified five variables as independent prognostic factors, and a nomogram was created. Because these five factors can be assessed before surgery, except when extrahepatic disease is diagnosed incidentally during hepatectomy, it is possible to identify patients who may benefit from subsequent surgery among those with lesions downsized by conversion chemotherapy for initially unresectable CRLM. It is beyond question that the primary objective of the strategy of conversion chemotherapy is to convert, that is, to be able to perform complete resection of liver disease as well as concomitant extrahepatic disease, if present. In fact, the OS rate after hepatectomy was significantly better in patients who had globally completed surgery than in those who did not in the present study. Interestingly, OS for patients who had completed liver surgery without resection of extrahepatic disease was significantly better than that observed for those in whom liver surgery was not completed. These findings suggest that curative hepatectomy of CRLM might be of benefit in prolonging survival of patients with both CRLM and extrahepatic disease compared with palliative chemotherapy, even when extrahepatic disease is not resected. Previously reported nomograms are considered to be complicated for use in daily practice because their total points range from zero to hundreds 13 15. The concept of the nomogram presented here is that it is simple and easy to use. For this purpose, numerical variables were converted to categorical variables with optimal cut-off values and rounded values were used to create the nomogram, resulting in a total score with a lower range of points. In this setting, the nomogram achieved a favourable performance of discrimination, higher than that of other models for CRLM described in the literature 13,14. Furthermore, internal calibration showed good predictive ability. Therefore, the authors believe that the proposed nomogram is simple and useful for determining the individual likelihood of survival after hepatectomy in patients with initially unresectable CRLM. The five factors identified here have been reported to be prognostic, and are commonly used in prognostic scoring for CRLM, except for the response to preoperative chemotherapy 13 16,25,26. The prognostic role of the tumour response to chemotherapy has been reported, and poor outcomes have been shown to be associated with disease progression 27 29. Unlike the present nomogram, the previous prognostic models did not take chemotherapy-related factors into account. This study has several limitations. First, the nomogram was developed based on data collected at a single institution. Second, the study population enrolled in the nomogram consisted of patients treated with hepatectomy. Therefore, this predictive model might be applicable only to potential candidates for hepatectomy after conversion chemotherapy. Third, a validation study using an external cohort is required to confirm the usefulness of this model. Finally, there is heterogeneity in the definition of initial unresectability between different institutions. At Hôpital Universitaire Paul Brousse, all patients with CRLM were evaluated by a multidisciplinary team including surgeons, oncologists and radiologists, and the determination of resectability was made by considering both technical and oncological criteria 3. To date, few studies have gathered both components together, and unresectability is often considered only on a technical basis 31. However, the definition of unresectability is changing over time with the development of chemotherapy regimens and surgical techniques. The creation of a standard definition of unresectability will be a critical issue in the near future.

598 K. Imai, M.-A. Allard, C. Castro Benitez, E. Vibert, A. Sa Cunha, D. Cherqui et al. Disclosure The authors declare no conflict of interest. References 1 Manfredi S, Lepage C, Hatem C, Coatmeur O, Faivre J, Bouvier AM. Epidemiology and management of liver metastases from colorectal cancer. Ann Surg 26; 244: 254 259. 2 Leporrier J, Maurel J, Chiche L, Bara S, Segol P, Launoy G. A population-based study of the incidence, management and prognosis of hepatic metastases from colorectal cancer. Br J Surg 26; 93: 465 474. 3 Adam R. Chemotherapy and surgery: new perspectives on the treatment of unresectable liver metastases. Ann Oncol 23; 14(Suppl 2): ii13 ii16. 4 Adam R, Delvart V, Pascal G, Valeanu A, Castaing D, Azoulay D et al. Rescue surgery for unresectable colorectal liver metastases downstaged by chemotherapy: a model to predict long-term survival. Ann Surg 24; 24: 644 657. 5 Scheele J, Stangl R, Altendorf-Hofmann A. Hepatic metastases from colorectal carcinoma: impact of surgical resection on the natural history. Br J Surg 199; 77: 1241 1246. 6 Bismuth H, Adam R, Levi F, Farabos C, Waechter F, Castaing D et al. Resection of nonresectable liver metastases from colorectal cancer after neoadjuvant chemotherapy. Ann Surg 1996; 224: 59 52. 7 Adam R, Wicherts DA, de Haas RJ, Ciacio O, Lévi F, Paule B et al. Patients with initially unresectable colorectal liver metastases: is there a possibility of cure? J Clin Oncol 29; 27: 1829 1835. 8 Masi G, Loupakis F, Pollina L, Vasile E, Cupini S, Ricci S et al. Long-term outcome of initially unresectable metastatic colorectal cancer patients treated with 5-fluorouracil/ leucovorin, oxaliplatin, and irinotecan (FOLFOXIRI) followed by radical surgery of metastases. Ann Surg 29; 249: 42 425. 9 Goere D, Deshaies I, de Baere T, Boige V, Malka D, Dumont F et al. Prolonged survival of initially unresectable hepatic colorectal cancer patients treated with hepatic arterial infusion of oxaliplatin followed by radical surgery of metastases. Ann Surg 21; 251: 686 691. 1 Capussotti L, Muratore A, Mulas MM, Massucco P, Aglietta M. Neoadjuvant chemotherapy and resection for initially irresectable colorectal liver metastases. Br J Surg 26; 93: 11 16. 11 Lam VW, Spiro C, Laurence JM, Johnston E, Hollands MJ, Pleass HC et al. A systematic review of clinical response and survival outcomes of downsizing systemic chemotherapy and rescue liver surgery in patients with initially unresectable colorectal liver metastases. Ann Surg Oncol 212; 19: 1292 131. 12 Takahashi S, Konishi M, Kinoshita T, Gotohda N, Kato Y, Saito N et al. Predictors for early recurrence after hepatectomy for initially unresectable colorectal liver metastasis. J Gastrointest Surg 213; 17: 939 948. 13 Kattan MW, Gonen M, Jarnagin WR, DeMatteo R, D Angelica M, Weiser M et al. A nomogram for predicting disease-specific survival after hepatic resection for metastatic colorectal cancer. Ann Surg 28; 247: 282 287. 14 Kanemitsu Y, Kato T. Prognostic models for predicting death after hepatectomy in individuals with hepatic metastases from colorectal cancer. World J Surg 28; 32: 197 117. 15 Rees M, Tekkis PP, Welsh FK, O Rourke T, John TG. Evaluation of long-term survival after hepatic resection for metastatic colorectal cancer: a multifactorial model of 929 patients. Ann Surg 28; 247: 125 135. 16 Beppu T, Sakamoto Y, Hasegawa K, Honda G, Tanaka K, Kotera Y et al. A nomogram predicting disease-free survival in patients with colorectal liver metastases treated with hepatic resection: multicenter data collection as a Project Study for Hepatic Surgery of the Japanese Society of Hepato-Biliary-Pancreatic Surgery. J Hepatobiliary Pancreat Sci 212; 19: 72 84. 17 Iasonos A, Schrag D, Raj GV, Panageas KS. How to build and interpret a nomogram for cancer prognosis. J Clin Oncol 28; 26: 1364 137. 18 Sternberg CN. Are nomograms better than currently available stage groupings for bladder cancer? J Clin Oncol 26; 24: 3819 382. 19 Kattan MW. Nomograms are superior to staging and risk grouping systems for identifying high-risk patients: preoperative application in prostate cancer. Curr Opin Urol 23; 13: 111 116. 2 Therasse P, Arbuck SG, Eisenhauer EA, Wanders J, Kaplan RS, Rubinstein L 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 2; 92: 25 216. 21 Azoulay D, Castaing D, Smail A, Adam R, Cailliez V, Laurent A et al. Resection of nonresectable liver metastases from colorectal cancer after percutaneous portal vein embolization. Ann Surg 2; 231: 48 486. 22 Adam R, Miller R, Pitombo M, Wicherts DA, de Haas RJ, Bitsakou G et al. Two-stage hepatectomy approach for initially unresectable colorectal hepatic metastases. Surg Oncol Clin N Am 27; 16: 525 536, viii. 23 Wicherts DA, Miller R, de Haas RJ, Bitsakou G, Vibert E, Veilhan LA et al. Long-term results of two-stage hepatectomy for irresectable colorectal cancer liver metastases. Ann Surg 28; 248: 994 15. 24 Dindo D, Demartines N, Clavien PA. Classification of surgical complications: a new proposal with evaluation in a cohort of 6336 patients and results of a survey. Ann Surg 24; 24: 25 213. 25 Fong Y, Fortner J, Sun RL, Brennan MF, Blumgart LH. Clinical score for predicting recurrence after hepatic

Nomogram for initially unresectable colorectal liver metastases 599 resection for metastatic colorectal cancer: analysis of 11 consecutive cases. Ann Surg 1999; 23: 39 318. 26 Nordlinger B, Guiguet M, Vaillant JC, Balladur P, Boudjema K, Bachellier P et al. Surgical resection of colorectal carcinoma metastases to the liver. A prognostic scoring system to improve case selection, based on 1568 patients. Association Française de Chirurgie. Cancer 1996; 77: 1254 1262. 27 Adam R, Pascal G, Castaing D, Azoulay D, Delvart V, Paule B et al. Tumor progression while on chemotherapy: a contraindication to liver resection for multiple colorectal metastases? Ann Surg 24; 24: 152 161. 28 Kornprat P, Jarnagin WR, Gonen M, DeMatteo RP, Fong Y, Blumgart LH et al. Outcome after hepatectomy for multiple (four or more) colorectal metastases in the era of effective chemotherapy. Ann Surg Oncol 27; 14: 1151 116. 29 Viganò L, Capussotti L, Barroso E, Nuzzo G, Laurent C, Ijzermans JN et al. Progression while receiving preoperative chemotherapy should not be an absolute contraindication to liver resection for colorectal metastases. Ann Surg Oncol 212; 19: 2786 2796. 3 Adam R, De Gramont A, Figueras J, Guthrie A, Kokudo N, Kunstlinger F et al. The oncosurgery approach to managing liver metastases from colorectal cancer: a multidisciplinary international consensus. Oncologist 212; 17: 1225 1239. 31 Cauchy F, Faivre S, Belghiti J. Surgical results after downstaging of initially marginal or non-resectable liver metastases. Dig Dis 212; 3(Suppl 2): 143 149. Supporting information Additional supporting information may be found in the online version of this article: Table S1 Detailed information on patients with concomitant extrahepatic disease (Word document) Fig. S1 Overall survival after diagnosis of liver metastasis (Word document) Fig. S2 Overall survival after diagnosis of liver metastasis according to the potential global curability of surgery (Word document)