J Hepatobiliary Pancreat Sci (2014) 21:881–888 DOI: 10.1002/jhbp.149
ORIGINAL ARTICLE
Indication for neoadjuvant chemotherapy in patients with colorectal liver metastases based on a nomogram that predicts disease-free survival Masayuki Okuno · Etsuro Hatano · Satoru Seo · Kojiro Taura · Kentaro Yasuchika · Akio Nakajima · Takefumi Yazawa · Hiroaki Furuyama · Hiroshi Kawamoto · Shintaro Yagi · Ryuta Nishitai · Takahisa Fujikawa · Akira Arimoto · Masazumi Zaima · Tsunehiro Yoshimura · Hiroaki Terajima · Satoshi Kaihara · Dai Manaka · Akira Tanaka · Shinji Uemoto
Published online: 25 August 2014 © 2014 Japanese Society of Hepato-Biliary-Pancreatic Surgery
Abstract Background The purpose of this study was to validate the Beppu nomogram, which predicts disease-free survival M. Okuno · E. Hatano (*) · S. Seo · K. Taura · K. Yasuchika · S. Uemoto Department of Surgery, Graduate School of Medicine, Kyoto University, 54 Kawahara-cho, Syogoin, Sakyo-ku, Kyoto 606-8507, Japan e-mail: [email protected] A. Nakajima · A. Arimoto Department of Surgery, Osaka Red Cross Hospital, Osaka, Japan T. Yazawa · M. Zaima Department of Surgery, Shiga Medical Center for Adults, Moriyama, Shiga, Japan H. Furuyama · T. Yoshimura Department of Surgery, Tenri Yorozu Hospital, Tenri, Nara, Japan H. Kawamoto · H. Terajima Department of Surgery, Kitano Hospital, Osaka, Japan S. Yagi · S. Kaihara Department of Surgery, Kobe City Medical Center General Hospital, Kobe, Hyogo, Japan R. Nishitai · D. Manaka Department of Surgery, Kyoto Katsura Hospital, Kyoto, Japan T. Fujikawa · A. Tanaka Department of Surgery, Kokura Memorial Hospital, Fukuoka, Japan
(DFS) after resection of colorectal liver metastases, and to investigate the efficacy of neoadjuvant chemotherapy based on the nomogram-predicted recurrence risk. Methods We retrospectively analyzed 234 patients with colorectal liver metastases who underwent a hepatic resection at eight hospitals between 2005 and 2010. Results The nomogram c-index of all the patients was 0.59. The observed and the predicted 3-year DFS showed good agreement. When the patients were divided into two groups who received or did not receive pre-hepatectomy chemotherapy (PHC), the c-index of the patients who received PHC was inferior to that of the patients who did not (0.56 and 0.61, respectively). In patients who received PHC, DFS among the quintiles clustered by the nomogram score indicated no significant differences (P = 0.25), unlike in patients who did not receive PHC (P < 0.0001). Surprisingly, in patients with no risk factors for recurrence, neoadjuvant chemotherapy provided significantly lower DFS than no neoadjuvant chemotherapy (3-year DFS: 42.9% vs. 80.0%, P = 0.03). Conclusions The nomogram validation was shown to be moderately predictive. PHC decreased the performance of the nomogram and might produce no DFS benefit in patients with low recurrent risk. Keywords Colorectal liver metastases · Hepatic resection · Neoadjuvant chemotherapy · Nomogram
882
Introduction The liver is the most frequent site of metastasis from colorectal cancer [1, 2]. Hepatic resection is considered the only therapy that leads to a cure for colorectal liver metastases (CLM) [3, 4], and the overall survival (OS) following hepatic resection exceeds 30% [5–7]; however, some of the patients exhibit early recurrence, and the overall recurrence rate ranges from 56.7 to 75% after curative hepatic resection [8, 9]. Although a previous report indicated that perioperative chemotherapy for patients with CLM is efficacious [10], the significance of neoadjuvant chemotherapy and the type of patient who should receive it before hepatic resection is still unclear. A nomogram is a statistical tool that provides the overall probability of a specific outcome for an individual patient in more detail than the tumor staging system [11]. Beppu et al. developed a nomogram that predicts disease-free survival (DFS) in patients with CLM after hepatic resection from 727 patients at the 11 institutions (includes Kyoto University Hospital) of a Project Study for Hepatic Surgery of the Japanese Society of Hepato-Biliary-Pancreatic Surgery [12]. The Beppu nomogram can estimate the median, 1-, and 3-year DFS after curative hepatic resection based on six preoperative variables: timing of liver metastases (metachronous/synchronous), primary tumor LN status (negative/positive), number of tumors (1/2–4/≥5), largest tumor diameter (≤5 cm/>5 cm), extrahepatic metastatic disease (no/yes), and CA19-9 level (≤100 U/ml/>100 U/ml) (Fig. 1).
Fig. 1 A nomogram that predicts disease-free survival (DFS) in patients with colorectal liver metastases treated with hepatic resection. Each variable was calculated, the total score was determined and a horizontal line was drawn from the total point to the right-hand axis to determine the patient’s probability of 3-year and 5-year DFS (Beppu et al., J Hepatobiliary Pancreat Sci. 2012). CA19-9 carbohydrate antigen 19-9
J Hepatobiliary Pancreat Sci (2014) 21:881–888
An external validation is required when the nomogram is adapted to other patients with characteristics that differ from the nomogram development cohort [13]. In particular, this nomogram cohort was composed of patients who underwent hepatectomies before 2004. Only one patient who received irinotecan-based pre-hepatectomy chemotherapy (PHC) was included [12]. The increase in the number of patients who received PHC may affect the current performance of this nomogram. If the nomogram currently predicts recurrence risk well, we could determine the treatment strategy for deciding the indication for PHC based on the nomogram. The aim of this multi-center study was to evaluate the external validity of the Beppu nomogram based on the patients who underwent a hepatic resection after 2005, when oxaliplatin or irinotecan-based chemotherapy was approved for use in Japan, and to investigate the indication for PHC in patients with resectable CLM based on the nomogrampredicted recurrence risk.
Methods Patients We retrospectively collected data from CLM patients who underwent a primary curative hepatic resection at eight hospitals between January 2005 and October 2010. Curative resection was defined as no tumor remaining both macroscopically and microscopically. All of the patients who underwent the concomitant use of radiofrequency ablation
J Hepatobiliary Pancreat Sci (2014) 21:881–888
883
or microwave coagulation were excluded. We collected the patient data regarding the component of the nomogram (timing of liver metastases, primary tumor lymph nodes status, number of tumors, largest tumor diameter, extra hepatic metastatic disease, and CA19-9 level) at the time of the hepatic resection and before PHC. For each patient, a total nomogram score was calculated based on these data, and predicted DFS, 1-year DFS rate and 3-year DFS rate were estimated from the nomogram. In addition to the nomogram required data, we noted the patient’s age, sex, primary cancer status, and the PHC content regarding CLM, DFS and OS. Patients were followed up according to each institution’s follow-up program. Statistical analysis The characteristics of the patients were analyzed using Wilcoxon test for continuous variables and χ2 test or Fisher’s exact test for categorical variables. DFS and OS were estimated using the Kaplan–Meier method. Nomogram validation included calibration and discrimination. Calibration was performed using two approaches. Firstly, the patients were divided into quintiles according to their nomogram derived risk so that the same proportion of patients were assigned to each quintile group; then, the predicted probability of 3-year DFS was compared to the observed 3-year DFS. Thus, a calibration plot was constructed. Secondly, patients were clustered into quintiles according to their nomogram score (i.e., 0, 1–5, 6–10, 11–15, and 16-), and the difference in the 3-year DFS between each quintile was calculated using the Kaplan– Meier method. Discrimination was quantified using the con-
cordance index (c-index), which is a modification of the area under the receiver operating characteristic curve (AUC) but was appropriate for censored data [14]. The c-index shows the probability that given two randomly selected patients, the patient with lower predicted survival will exhibit lower survival. A c-index = 0.50 represents agreement by chance; a c-index = 1.0 represents perfect discrimination [15]. JMP 10.0 (SAS Institute, Cary, NC, USA) was used for the statistical analysis. All tests were two-tailed, and a P < 0.05 was considered significant. Results A total of 234 patients were included in this study. Median follow-up was 43.1 months (range, 0.3–99.9 months), and the median follow-up of the survivors was 51.8 months (range, 18.9–99.9 months). The median DFS was 18.9 months, and the 3-year DFS rate was 36.9% in all patients. One patient died within 30 days after hepatic resection but was still included in this validation. Patient characteristics for the component of the nomogram in the external validation cohort and the nomogram development cohort are shown in Table 1. Although the number of patients with low CA19-9 levels (≤100 ng/ml) were significantly less in the external validation cohort than in the nomogram development cohort (P > 0.01), the two cohorts were similar in other variables that were required for the nomogram. During the follow-up period, 65.0% of the patients (152 patients) experienced recurrence. The comparison of the variables related to the nomogram and the nomogram score between the recurrent group (n = 152) and the no-recurrent group (n = 82) are shown in Table 2. The mean nomogram
Table 1 Clinical characteristics of the patient cohorts from the nomogram development set and the external validation set
Timing of metastasis Primary tumor LN status Number of tumors
Largest tumor diameter Extrahepatic metastatic disease (at hepatectomy) CA19-9 level (before hepatectomy) Values in brackets represent the percentage CA19-9 carbohydrate antigen 19-9, LN lymph node
Metachronous Synchronous Negative Positive 1 2–4 5≤ ≤5 cm >5 cm No Yes ≤100 >100
Nomogram development set n = 727
External validation set n = 234
P
359 (55.0) 294 (45.0) 210 (32.7) 432 (67.3) 321 (49.5) 234 (36.1) 94 (14.5) 485 (75.9) 154 (24.1) 595 (90.7) 61 (9.3) 478 (75.6) 154 (24.4)
126 (53.8) 108 (46.2) 88 (37.6) 146 (62.4) 118 (50.4) 89 (38.0) 27 (11.5) 188 (80.3) 46 (19.7) 208 (88.9) 26 (11.1) 199 (85.0) 35 (15.0)
0.77 0.18 0.52
0.17 0.42 0.003
884
J Hepatobiliary Pancreat Sci (2014) 21:881–888
Table 2 Comparison of clinical characteristics between the recurrent group and the non-recurrent group
Timing of metastasis Primary tumor LN status Number of tumors
Largest tumor diameter Extrahepatic metastatic disease (at hepatectomy) CA19-9 level (before hepatectomy)
Metachronous Synchronous Negative Positive 1 2–4 5≤ ≤5 cm >5 cm No Yes ≤100 >100
Nomogram score
Recurrent group n = 152
No-recurrent group n = 82
78 (51.3) 74 (48.7) 45 (29.6) 107 (70.4) 67 (44.1) 64 (42.1) 21 (13.8) 116 (76.3) 36 (23.7) 131 (86.2) 21 (13.8) 127 (83.5) 25 (16.5) 8.18 ± 5.08
48 (58.5) 34 (41.5) 43 (52.4) 39 (47.6) 51 (62.2) 25 (30.5) 6 (7.3) 72 (87.8) 10 (12.2) 77 (93.9) 5 (6.1) 72 (87.8) 10 (12.2) 5.52 ± 4.40
P
0.29 0.001 0.026
0.035 0.084 0.38
ORIGINAL ARTICLE
Indication for neoadjuvant chemotherapy in patients with colorectal liver metastases based on a nomogram that predicts disease-free survival Masayuki Okuno · Etsuro Hatano · Satoru Seo · Kojiro Taura · Kentaro Yasuchika · Akio Nakajima · Takefumi Yazawa · Hiroaki Furuyama · Hiroshi Kawamoto · Shintaro Yagi · Ryuta Nishitai · Takahisa Fujikawa · Akira Arimoto · Masazumi Zaima · Tsunehiro Yoshimura · Hiroaki Terajima · Satoshi Kaihara · Dai Manaka · Akira Tanaka · Shinji Uemoto
Published online: 25 August 2014 © 2014 Japanese Society of Hepato-Biliary-Pancreatic Surgery
Abstract Background The purpose of this study was to validate the Beppu nomogram, which predicts disease-free survival M. Okuno · E. Hatano (*) · S. Seo · K. Taura · K. Yasuchika · S. Uemoto Department of Surgery, Graduate School of Medicine, Kyoto University, 54 Kawahara-cho, Syogoin, Sakyo-ku, Kyoto 606-8507, Japan e-mail: [email protected] A. Nakajima · A. Arimoto Department of Surgery, Osaka Red Cross Hospital, Osaka, Japan T. Yazawa · M. Zaima Department of Surgery, Shiga Medical Center for Adults, Moriyama, Shiga, Japan H. Furuyama · T. Yoshimura Department of Surgery, Tenri Yorozu Hospital, Tenri, Nara, Japan H. Kawamoto · H. Terajima Department of Surgery, Kitano Hospital, Osaka, Japan S. Yagi · S. Kaihara Department of Surgery, Kobe City Medical Center General Hospital, Kobe, Hyogo, Japan R. Nishitai · D. Manaka Department of Surgery, Kyoto Katsura Hospital, Kyoto, Japan T. Fujikawa · A. Tanaka Department of Surgery, Kokura Memorial Hospital, Fukuoka, Japan
(DFS) after resection of colorectal liver metastases, and to investigate the efficacy of neoadjuvant chemotherapy based on the nomogram-predicted recurrence risk. Methods We retrospectively analyzed 234 patients with colorectal liver metastases who underwent a hepatic resection at eight hospitals between 2005 and 2010. Results The nomogram c-index of all the patients was 0.59. The observed and the predicted 3-year DFS showed good agreement. When the patients were divided into two groups who received or did not receive pre-hepatectomy chemotherapy (PHC), the c-index of the patients who received PHC was inferior to that of the patients who did not (0.56 and 0.61, respectively). In patients who received PHC, DFS among the quintiles clustered by the nomogram score indicated no significant differences (P = 0.25), unlike in patients who did not receive PHC (P < 0.0001). Surprisingly, in patients with no risk factors for recurrence, neoadjuvant chemotherapy provided significantly lower DFS than no neoadjuvant chemotherapy (3-year DFS: 42.9% vs. 80.0%, P = 0.03). Conclusions The nomogram validation was shown to be moderately predictive. PHC decreased the performance of the nomogram and might produce no DFS benefit in patients with low recurrent risk. Keywords Colorectal liver metastases · Hepatic resection · Neoadjuvant chemotherapy · Nomogram
882
Introduction The liver is the most frequent site of metastasis from colorectal cancer [1, 2]. Hepatic resection is considered the only therapy that leads to a cure for colorectal liver metastases (CLM) [3, 4], and the overall survival (OS) following hepatic resection exceeds 30% [5–7]; however, some of the patients exhibit early recurrence, and the overall recurrence rate ranges from 56.7 to 75% after curative hepatic resection [8, 9]. Although a previous report indicated that perioperative chemotherapy for patients with CLM is efficacious [10], the significance of neoadjuvant chemotherapy and the type of patient who should receive it before hepatic resection is still unclear. A nomogram is a statistical tool that provides the overall probability of a specific outcome for an individual patient in more detail than the tumor staging system [11]. Beppu et al. developed a nomogram that predicts disease-free survival (DFS) in patients with CLM after hepatic resection from 727 patients at the 11 institutions (includes Kyoto University Hospital) of a Project Study for Hepatic Surgery of the Japanese Society of Hepato-Biliary-Pancreatic Surgery [12]. The Beppu nomogram can estimate the median, 1-, and 3-year DFS after curative hepatic resection based on six preoperative variables: timing of liver metastases (metachronous/synchronous), primary tumor LN status (negative/positive), number of tumors (1/2–4/≥5), largest tumor diameter (≤5 cm/>5 cm), extrahepatic metastatic disease (no/yes), and CA19-9 level (≤100 U/ml/>100 U/ml) (Fig. 1).
Fig. 1 A nomogram that predicts disease-free survival (DFS) in patients with colorectal liver metastases treated with hepatic resection. Each variable was calculated, the total score was determined and a horizontal line was drawn from the total point to the right-hand axis to determine the patient’s probability of 3-year and 5-year DFS (Beppu et al., J Hepatobiliary Pancreat Sci. 2012). CA19-9 carbohydrate antigen 19-9
J Hepatobiliary Pancreat Sci (2014) 21:881–888
An external validation is required when the nomogram is adapted to other patients with characteristics that differ from the nomogram development cohort [13]. In particular, this nomogram cohort was composed of patients who underwent hepatectomies before 2004. Only one patient who received irinotecan-based pre-hepatectomy chemotherapy (PHC) was included [12]. The increase in the number of patients who received PHC may affect the current performance of this nomogram. If the nomogram currently predicts recurrence risk well, we could determine the treatment strategy for deciding the indication for PHC based on the nomogram. The aim of this multi-center study was to evaluate the external validity of the Beppu nomogram based on the patients who underwent a hepatic resection after 2005, when oxaliplatin or irinotecan-based chemotherapy was approved for use in Japan, and to investigate the indication for PHC in patients with resectable CLM based on the nomogrampredicted recurrence risk.
Methods Patients We retrospectively collected data from CLM patients who underwent a primary curative hepatic resection at eight hospitals between January 2005 and October 2010. Curative resection was defined as no tumor remaining both macroscopically and microscopically. All of the patients who underwent the concomitant use of radiofrequency ablation
J Hepatobiliary Pancreat Sci (2014) 21:881–888
883
or microwave coagulation were excluded. We collected the patient data regarding the component of the nomogram (timing of liver metastases, primary tumor lymph nodes status, number of tumors, largest tumor diameter, extra hepatic metastatic disease, and CA19-9 level) at the time of the hepatic resection and before PHC. For each patient, a total nomogram score was calculated based on these data, and predicted DFS, 1-year DFS rate and 3-year DFS rate were estimated from the nomogram. In addition to the nomogram required data, we noted the patient’s age, sex, primary cancer status, and the PHC content regarding CLM, DFS and OS. Patients were followed up according to each institution’s follow-up program. Statistical analysis The characteristics of the patients were analyzed using Wilcoxon test for continuous variables and χ2 test or Fisher’s exact test for categorical variables. DFS and OS were estimated using the Kaplan–Meier method. Nomogram validation included calibration and discrimination. Calibration was performed using two approaches. Firstly, the patients were divided into quintiles according to their nomogram derived risk so that the same proportion of patients were assigned to each quintile group; then, the predicted probability of 3-year DFS was compared to the observed 3-year DFS. Thus, a calibration plot was constructed. Secondly, patients were clustered into quintiles according to their nomogram score (i.e., 0, 1–5, 6–10, 11–15, and 16-), and the difference in the 3-year DFS between each quintile was calculated using the Kaplan– Meier method. Discrimination was quantified using the con-
cordance index (c-index), which is a modification of the area under the receiver operating characteristic curve (AUC) but was appropriate for censored data [14]. The c-index shows the probability that given two randomly selected patients, the patient with lower predicted survival will exhibit lower survival. A c-index = 0.50 represents agreement by chance; a c-index = 1.0 represents perfect discrimination [15]. JMP 10.0 (SAS Institute, Cary, NC, USA) was used for the statistical analysis. All tests were two-tailed, and a P < 0.05 was considered significant. Results A total of 234 patients were included in this study. Median follow-up was 43.1 months (range, 0.3–99.9 months), and the median follow-up of the survivors was 51.8 months (range, 18.9–99.9 months). The median DFS was 18.9 months, and the 3-year DFS rate was 36.9% in all patients. One patient died within 30 days after hepatic resection but was still included in this validation. Patient characteristics for the component of the nomogram in the external validation cohort and the nomogram development cohort are shown in Table 1. Although the number of patients with low CA19-9 levels (≤100 ng/ml) were significantly less in the external validation cohort than in the nomogram development cohort (P > 0.01), the two cohorts were similar in other variables that were required for the nomogram. During the follow-up period, 65.0% of the patients (152 patients) experienced recurrence. The comparison of the variables related to the nomogram and the nomogram score between the recurrent group (n = 152) and the no-recurrent group (n = 82) are shown in Table 2. The mean nomogram
Table 1 Clinical characteristics of the patient cohorts from the nomogram development set and the external validation set
Timing of metastasis Primary tumor LN status Number of tumors
Largest tumor diameter Extrahepatic metastatic disease (at hepatectomy) CA19-9 level (before hepatectomy) Values in brackets represent the percentage CA19-9 carbohydrate antigen 19-9, LN lymph node
Metachronous Synchronous Negative Positive 1 2–4 5≤ ≤5 cm >5 cm No Yes ≤100 >100
Nomogram development set n = 727
External validation set n = 234
P
359 (55.0) 294 (45.0) 210 (32.7) 432 (67.3) 321 (49.5) 234 (36.1) 94 (14.5) 485 (75.9) 154 (24.1) 595 (90.7) 61 (9.3) 478 (75.6) 154 (24.4)
126 (53.8) 108 (46.2) 88 (37.6) 146 (62.4) 118 (50.4) 89 (38.0) 27 (11.5) 188 (80.3) 46 (19.7) 208 (88.9) 26 (11.1) 199 (85.0) 35 (15.0)
0.77 0.18 0.52
0.17 0.42 0.003
884
J Hepatobiliary Pancreat Sci (2014) 21:881–888
Table 2 Comparison of clinical characteristics between the recurrent group and the non-recurrent group
Timing of metastasis Primary tumor LN status Number of tumors
Largest tumor diameter Extrahepatic metastatic disease (at hepatectomy) CA19-9 level (before hepatectomy)
Metachronous Synchronous Negative Positive 1 2–4 5≤ ≤5 cm >5 cm No Yes ≤100 >100
Nomogram score
Recurrent group n = 152
No-recurrent group n = 82
78 (51.3) 74 (48.7) 45 (29.6) 107 (70.4) 67 (44.1) 64 (42.1) 21 (13.8) 116 (76.3) 36 (23.7) 131 (86.2) 21 (13.8) 127 (83.5) 25 (16.5) 8.18 ± 5.08
48 (58.5) 34 (41.5) 43 (52.4) 39 (47.6) 51 (62.2) 25 (30.5) 6 (7.3) 72 (87.8) 10 (12.2) 77 (93.9) 5 (6.1) 72 (87.8) 10 (12.2) 5.52 ± 4.40
P
0.29 0.001 0.026
0.035 0.084 0.38
