Response to Neoadjuvant Chemotherapy Does Not Predict Overall Survival for Patients With Synchronous Colorectal Hepatic Metastases
- First Online:
- Cite this article as:
- Gallagher, D.J., Zheng, J., Capanu, M. et al. Ann Surg Oncol (2009) 16: 1844. doi:10.1245/s10434-009-0348-1
- 212 Views
We investigated the relation between response to neoadjuvant chemotherapy and overall survival (OS) in patients with colorectal liver metastases (CLM).
It has previously been reported that patients with synchronous CLM whose disease progresses while receiving neoadjuvant chemotherapy or who do not receive neoadjuvant chemotherapy experience worse survival than patients whose disease responds to neoadjuvant chemotherapy.
By means of a prospectively maintained surgical database, between 1995 and 2003, we identified 111 patients with a synchronous CLM who received neoadjuvant chemotherapy before hepatic resection. The disease of all 111 patients was deemed resectable, and patients underwent hepatic resection with curative intent.
The median OS after liver resection was 62 months, with a median follow-up of 63 months. Median OS was similar between the three study groups classified by response to neoadjuvant chemotherapy (complete or partial response, 58 months; stable disease, 65 months; and disease progression, 61 months; P = .98). By univariate analysis, carcinoembryonic antigen level after liver resection of <5 ng/dL, size of metastatic lesion of ≤5 cm, lymph node–negative primary tumor, and disease-negative margins were associated with improved survival. Patients in the disease progression group had more positive margins and metastases >5 cm in size than patients in the complete or partial response group and the stable disease group. Patients whose tumor progressed but who received postoperative hepatic arterial infusion had a trend toward improved survival compared with those who did not receive hepatic arterial infusion (70% vs. 50% at 3 years, permutation log rank test P = .12).
Response to neoadjuvant chemotherapy did not correlate with OS even after controlling for margins, stage of primary tumor, and postoperative carcinoembryonic antigen level. Postoperative salvage treatment may have helped the survival of some patients.
Approximately 1 million people will be diagnosed with colorectal cancer in 2009. Fifty percent of those with metastatic disease have hepatic metastases. However, at the time of presentation, only 10% to 20% of these patients have resectable disease. The incidence of synchronous liver metastases is reported to be between 23% and 46.8% among all patients with colorectal liver metastases (CLM).1–6 Untreated, the median survival of patients with synchronous CLM is 4.5 months (range, 0–36 months).7
Although surgical resection has never been tested in a prospective, randomized study, several series have demonstrated long-term survival and long-term cure after complete resection, which supports this approach.8 For patients with resectable disease, complete surgical resection is associated with 30% survival at 5 years, although more recent studies show 5-year survival rates after hepatectomy approaching 40% to 50%.9,10 In contrast, few unresected patients survive beyond 5 years.11
Synchronous metastases are associated with a less favorable outcome than metachronous disease.1,12–17 Tsai et al. suggested that these different presentations represent different biologies and that synchronous disease is more aggressive, with a greater tendency to disseminate.18 However, as is the case for patients with other poor baseline characteristics, surgical resection still offers a potential for long-term survival.1
The definition of resectability varies from center to center, but metastases are generally considered resectable if an R0 resection can be achieved and the liver remnant after hepatectomy is ≥20% to 30% of the original liver volume. For patients with initially unresectable disease, preoperative chemotherapy is often provided to reduce the size of metastases to possibly allow future resection. Some studies suggest a correlation between the response rate to chemotherapy and the resection rate of liver metastases.19 Adam et al. showed that patients with four or more liver metastases who achieve stable disease or response with chemotherapy have improved 5-year survival after hepatectomy compared with those whose disease progresses (37% and 30% vs. 8%, respectively, P < .0001).20 This approach is also supported by a previous publication from our group, in which the survival of 46 patients with synchronous liver metastases treated with neoadjuvant chemotherapy before resection was compared with the survival of similar patients who did not receive neoadjuvant chemotherapy. Those who had disease stabilization or regression while receiving chemotherapy had a survival benefit over those who did not receive chemotherapy (87% vs. 38% 5-year specific survival, respectively, P = .03).21 Many oncologists have embraced this approach, using it to select patients for treatment and also to treat initially resectable patients with neoadjuvant chemotherapy to potentially facilitate less radical surgery, treat undetectable micrometastases, and guide the selection of patients for resection. However, this approach may unfairly deny liver resection to a group of patients who still have resectable disease and who, despite their disease’s poor response to chemotherapy, may still achieve long-term survival.
We conducted a retrospective study of patients with resectable disease treated with preoperative chemotherapy followed by surgical resection of liver metastases to determine whether response to neoadjuvant chemotherapy correlated with survival in resectable patients with synchronous CLM. We include updated survival data on 46 patients previously reported by Allen et al., and 65 additional patients who have been treated with neoadjuvant chemotherapy and surgery for resectable synchronous CLM.
We used the prospectively maintained hepatobiliary surgical database at Memorial Sloan-Kettering Cancer Center (MSKCC) to identify patients who presented with synchronous CLM between 1995 and 2003, and who were treated with neoadjuvant chemotherapy followed by hepatic resection. Patients had histologically confirmed colorectal adenocarcinoma and were reviewed at a multidisciplinary conference at MSKCC, where they were considered to have resectable disease. They were then referred for preoperative systemic chemotherapy. Most patients were treated at MSKCC, but some received chemotherapy at the institution from which they were referred. While receiving treatment, patients underwent imaging at intervals decided by the treating physician. Patients whose disease responded to therapy were treated to best response or to a point where reduction in size of the metastases facilitated less radical surgery. If disease achieved a complete response, resection was performed to remove areas of previously documented disease on computed tomographic or magnetic resonance imaging scan. Patients with nonresponding disease were operated on at the first sign of radiological progression. All hepatic surgery was performed at MSKCC. Most patients received neoadjuvant chemotherapy between colectomy and resection of hepatic metastases. Four patients underwent a combined procedure with resection of both the colorectal primary tumor and the hepatic metastases, preceded by neoadjuvant chemotherapy. Radiofrequency ablation and cryoablation were occasionally used as adjuncts to hepatic resection.
Some patients had agreed to be on hepatic arterial infusion (HAI) and systemic protocols before resection, and a pump was placed at the time of resection. Patients not on HAI protocols were offered systemic chemotherapy at the discretion of their physician on the basis of performance status, tolerance of and response to neoadjuvant chemotherapy, and predicted risk of recurrence. The risk of recurrence was determined by the MSKCC clinical risk score and completeness of surgical resection.1 The number of cycles and type of postoperative chemotherapy was decided by the patient’s medical oncologist. Most patients were treated postoperatively at MSKCC, but some returned to their local institution.
Fisher’s exact test was performed to evaluate the association of response to neoadjuvant chemotherapy before hepatic resection (complete or partial response, CR/PR; stable disease, S; and disease progression, P) with a number of clinically important variables, including carcinoembryonic antigen (CEA) levels before liver resection (<30 or ≥30 ng/dL), number of metastases, size of largest metastasis, existence of extrahepatic disease, adjuvant systemic chemotherapy, adjuvant HAI, distribution (unilobar, bilobar), stage of primary tumor, margins (positive, negative) and post–liver resection CEA of <5 or ≥5 ng/dL. Overall survival (OS) was calculated from the date of hepatic resection to date of death or date of last follow-up and estimated by the Kaplan-Meier method. The log rank test or the permutation log rank test were used to test for association of survival with clinically important risk factors.22 Variables significant at the .05 level were included in a multivariable Cox proportional regression model, which controlled for the response to neoadjuvant therapy.
Between 1995 and 2003, a total of 111 patients underwent hepatectomy, preceded by neoadjuvant chemotherapy, for resectable, synchronous CLM. Seventy-six men and 35 women were operated on; their median age was 61 years (range, 27–85 years). Eighty-four patients had primary disease in the colon, and 27 had primary disease in the rectum. Seventy-six patients (67%) had disease-positive lymph nodes. The median CEA before resection of liver metastases was 9.5 ng/dL, and the median CEA after liver resection was 3.1 ng/dL.
Staged resection of the primary tumor followed by resection of hepatic metastases was performed in 107 patients, and 4 patients underwent synchronous resection. The median time between colon and liver resection was 7.0 months (range, 0–39 months). In addition to the primary liver resection, cryoablation was performed in two patients and radiofrequency ablation in one patient. Twenty-seven patients had >4 metastases resected, and the median number of lesions removed was 2 (range, 1–10 lesions). The median size of resected lesions was 2.5 cm, and 22 patients had metastases of >5 cm in size. The median MSKCC clinical risk score was 3. Eight patients had histologically confirmed extrahepatic disease resected at the time of hepatic surgery (one pulmonary wedge resection, one oophorectomy, and one intraperitoneal lesion resected; and five portal lymph node resections). Second resections were performed in 11 patients for recurrent disease after liver resection (7 pulmonary resections; 2 repeat liver resections; 1 peritoneal metastasis resection; 1 liver and lung resection).
Neoadjuvant chemotherapy regimens used were infusional 5-fluorouracil/leucovorin plus irinotecan (FOLFIRI) (n = 54 patients), 5-fluorouracil/leucovorin (5-FU/LV) (n = 37), infusional 5-fluorouracil/leucovorin plus oxaliplatin (FOLFOX) (n = 13), and other regimens (n = 8); the median number of cycles administered was 4. Neoadjuvant chemotherapy resulted in a complete response in 6 and a partial response in 41 patients (CR/PR, 37%), stable disease in 52 patients (S, 47%), and progressive disease in 18 patients (P, 16%). The median duration of treatment in these 18 patients was 5 months (range, 1–14 months). The median duration of postoperative chemotherapy was 5 months. Twenty patients did not receive adjuvant therapy, 43 received both adjuvant HAI and systemic treatment, and 48 received adjuvant systemic chemotherapy alone.
Response to Neoadjuvant Therapy
Frequency of clinically important variables for each response group
(N = 41)
(N = 52)
(N = 18)
(N = 111)
No. of metastases ≥ 4, n (%)
Size of largest metastasis >5 cm, n (%)
Extrahepatic disease found, n (%)
Adjuvant systemic chemotherapy only, n (%)
Adjuvant HAI, n (%)
Male sex, n (%)
Unilobar disease, n (%)
Lymph node positive–primary tumor, n (%)
Disease-positive margins, n (%)
Age ≥ 65 y, n (%)
Prehepatic resection CEA
≥30 ng/dL, n (%)
Posthepatic resection CEA
≥ 5 ng/dL, n (%)
OS from hepatic resection stratified by clinically important risk factors
Median OS (95% CI)
5-y Survival (95% CI)
Log rank P value
HAI + systemic
CEA before liver resection:
CEA after liver resection:
No. of metastases
Size of largest metastases
Lymph node–positive primary tumor
Multivariable Cox regression model for hepatectomy overall survivala
Multivariable model factor
Hazard ratio (95% CI)
Primary lymph node status
Posthepatic resection CEA
In this study, response to neoadjuvant chemotherapy was not related to OS after hepatic resection for patients with resectable synchronous CLM. This finding is different from that of Adam et al., who showed that patients with ≥4 hepatic metastases who had stable disease or whose disease responded to chemotherapy have improved 5-year survival compared with those whose disease progressed while the patient was receiving neoadjuvant chemotherapy (37% and 30% vs. 8%, respectively, P < .0001).23 However, in the study of Adam et al., only 74% had synchronous disease; in our study, all patients had synchronous disease. For patients with synchronous resectable disease, Allen et al. reported a survival benefit for patients whose disease did not progress while receiving neoadjuvant chemotherapy, compared with those who did not receive chemotherapy (87% vs. 38% 5-year specific survival, respectively, P = .03).21 Our study includes the 46 patients who received neoadjuvant chemotherapy and could be assessed for response in the study of Allen et al., as well as an additional 65 patients who also received neoadjuvant chemotherapy. Our analysis compares the survival of patients whose disease progressed while the patient was receiving neoadjuvant therapy versus patients whose disease responded to neoadjuvant therapy, and therefore, it is a different comparison than that addressed by Allen et al., in which the survival of patients whose disease responded to neoadjuvant therapy is compared with that of patients who did not receive neoadjuvant therapy. Note also that we had a longer follow-up for the 46 patients shared by the two studies (median follow-up now 63 months, vs. 30 months in the earlier study).
In many studies, patients with synchronous CLM have a worse prognosis than those with metachronous CLM.1,12–16,18,24 Nonetheless, in this study, median OS from the time of liver resection was 62 months with 52% alive at 5 years, which compares favorably with other reports. The 5-year survival from colon resection was 60% and was not different for the three study groups. Almost all patients (107 of 111 patients) in this study underwent a staged resection because many were only referred to us after colectomy for definitive management of hepatic metastases. Response to neoadjuvant chemotherapy has been proposed as a selection criterion for hepatic resection, with the hope of avoiding operative morbidity and delay in other systemic treatment in patients who would not benefit from surgery. The absence of a correlation between response to neoadjuvant chemotherapy and OS in our study suggests that this approach would deny some patients a chance of long-term survival.
Independent prognosticators of survival in this study were similar to those previously identified. Several prognostic scoring systems for patients with CLM have been developed.1,5,25–27 We compared the commonly used risk factors across response groups and tested them in univariate and multivariate analyses. By univariate analysis, node-positive primary tumor, size of metastatic lesion of >5 cm, positive resection margins, and CEA of >5 ng/dL after liver resection were statistically significantly associated with poorer survival. Only positive margins, node-positive primary tumor, and CEA of ≥5 ng/mL after liver resection were statistically significant in a multivariate model.
Differences in known prognostic indicators among the three response groups did not explain our findings. Size of metastatic lesion of >5 cm and the presence of positive margins differed between each group. Importantly, the patients whose disease progressed while they were receiving treatment had larger lesions and more disease-positive margins. This group should have worse survival, but survival was the same in each group, further supporting the concept that disease progression while receiving chemotherapy should not contraindicate hepatic resection.
The fact that most patients received postoperative treatment in this study may account for the good long-term survival in each response category: 82% received adjuvant systemic chemotherapy, 40% of whom received HAI and systemic therapy. A randomized controlled trial of 173 patients treated with adjuvant fluorouracil and folinic acid has demonstrated a marked improvement in disease-free survival over surgery alone (odds ratio, .66; 95% CI, .46–.96; P = .028), and a nonsignificant trend toward increased OS (.73; 95% CI, .48–1.10; P = .13).28 A recent retrospective review of 792 patients supports an improvement in OS with adjuvant chemotherapy compared with surgery alone (median OS, 47 vs. 36 months, P = .007).29 One study reported improved survival for patients with synchronous CLM treated with postoperative 5-FU- or HAI floxuridine-based therapy (hazard ratio [HR] = .62, 95% CI, .50–.78; HR = .51, 95% CI, .28–.97, respectively). Adjuvant chemotherapy also improves survival in another group of poor-risk patients (four or more bilobar metastases) compared with those treated with surgery alone (51.5 vs. 23 months, P < .01).30
It therefore seems that adjuvant chemotherapy is of benefit after hepatic resection, but does the addition of neoadjuvant treatment help further? The EORTC 40983 prospective study randomized 364 patients with resectable CLM to six cycles of neoadjuvant FOLFOX4 plus six cycles of adjuvant FOLFOX4 or surgery alone.31 An 8.1% improvement was reported in disease-free survival (P = .04), but this study does not help answer the question of whether both neoadjuvant and adjuvant chemotherapy are necessary. Further support for adjuvant therapy was provided by a recent pooled analysis of two phase 3 adjuvant trials that closed prematurely as a result of poor accrual. There was a trend toward an improvement in median progression-free survival (HR = 1.32, 95% CI, 1.00–1.76, P = .058), and in a multivariate analysis, adjuvant chemotherapy was statistically significantly associated with a better OS (HR = 1.39; 95% CI, 1.00–1.93, P = .046).32,33 However, in our univariate analysis, the use of adjuvant chemotherapy did not correlate with improved survival, irrespective of whether systemic plus HAI (P = .48) or systemic therapy alone (P = .78) was used, although only 20 patients received no adjuvant treatment. For the group whose disease progressed while receiving neoadjuvant chemotherapy, there was a trend toward improved 5-year OS compared with those who did not receive adjuvant HAI (70% vs. 50%, permutation log rank P = .12). It is possible that salvage postoperative HAI helped this group of patients.
Interpretation of the results of this study is limited for a number of reasons. First, it is a retrospective review. Although we did not identify any differences in known clinical prognostic indicators between the groups that would explain why patients whose disease progressed did better than expected, only a prospective randomized study could account for such a bias. Furthermore, all our patients underwent resection, and therefore, the number of patients whose disease progressed while they were receiving preoperative chemotherapy and whose disease could not be resected is unknown. Occasionally the disease of patients receiving neoadjuvant chemotherapy progresses to a point where it is unresectable—an occurrence that further argues against the use of neoadjuvant chemotherapy.
The small number of patients in the study limits its power. Because of the small numbers, we were unable to examine the effect of other factors on outcome, such as whether novel agents like bevacizumab and cetuximab improved response and survival. This approach is feasible and is being asked by the current Eastern Cooperative Oncology Group 40051 study.31,34 We identified no difference in median OS or 5-year survival by neoadjuvant chemotherapy used (5-FU, FOLFIRI, FOLFOX, or other, P = .8), although the small numbers preclude a more meaningful analysis. Regimens incorporating biological agents and three-drug regimens such as FOLFOXIRI have been associated with improved response rates.35–37 It is possible that the neoadjuvant use of these treatments would produce different results, although improved response did not correlate with better survival in this study.
In summary, this study indicates that patients whose disease progresses while receiving neoadjuvant chemotherapy but that remains resectable can proceed to hepatic resection. It argues against using progression while receiving preoperative treatment as a reason to not offer hepatic resection. If patients are treated with neoadjuvant chemotherapy, they must be monitored closely and scanned frequently while receiving treatment. At the first sign of disease progression, chemotherapy should be stopped and hepatic resection performed if still possible. Up-front resection followed by adjuvant chemotherapy remains reasonable in certain cases. Some patients can be helped by salvage treatment despite progression of disease while receiving neoadjuvant chemotherapy, possibly as a result of appropriate adjuvant treatment (systemic chemotherapy with or without HAI) after liver resection.
We thank Dr. Rebecca White and Tracy Mak for their help in gathering data for this study.