|Home | About | Journals | Submit | Contact Us | Français|
The selection of patients for resective therapy, of hepatic colorectal metastases remains controversial. A number of clinical and pathologic prognostic risk factor have been variably reported to influence survival.
Between January 1981 and December 1991, 204 patients underwent curative hepatic resection for metastatic colorectal cancer. Fourteen clinical and pathologic determinants previously reported to influence outcome were examined retrospectively. This led to a proposed TNM staging system for metastatic colorectal cancer (mTNM).
No operative deaths occurred (death within 1 month). Overall 1-, 3-, and 5-year survivals were 91%, 43%, and 32%, respectively. Gender, Dukes’ classification, site of primary colorectal cancer, histologic differentiation, size of metastatic tumor, and intraoperative blood transfusion requirement were not statistically significant prognostic factors (p > 0.05). Age of 60 years or more, interval of 24 months or less between colorectal and hepatic resection, four or more gross tumors, bilobar involvement, positive resection margin, Lymph node involvement, and direct invasion to adjacent organs were significant poor prognostic factors (p < 0.05). In the absence of nodal disease or direct invasion, patients with unilobar solitary tumor of any size, or unilobar multiple tumors of 2 cm or smaller (stages I and II) had the highest survival rates of 93% at 1 year, 68% at 3 years, and 61% at 5 years. Unilobar disease with multiple lesions greater than 2 cm (stage III) resulted in 1-, 3-, and 5-year survivals of 98%, 45%, and 28%, respectively. Patients with bilobar involvement (multiple tumors, any size, or a single large metastasis) (stage IVA) had survival rates of 88% at 1 year, 28% at 3 years, and 20% at 5 years (p < 0.00001). Patients with nodal involvement or extrahepatic disease (stage IVB) experienced the poorest outcome with 1-, 3-, and 5-year survivals of 80%, 12%, and 0%, respectively (p < 0.00001).
The proposed m TNM staging system appears to be useful in predicting the outcomes after hepatic resection of metastatic colorectal tumors.
Although it is clear that hepatic resection for metastatic colorectal cancer can be performed quite safely, there is still controversy regarding patient selection.1–3 Previous studies reported that 5-year survival rates (rarely observed without resection) range from 16% to 52%.1–13 Many have addressed the in influence of various clinical and pathologic parameters on the outcome or resective therapy, but the results vary considerably from study to study.
We have reexamined our 204 consecutive patients who underwent hepatic resection for colorectal metastases during the last 11 years to identify clinical and pathologic prognosticators.
Metastatic tumor characteristics were further examined by utilizing a TNM staging system modified from the International Union Against Cancer (UICC) and The American Joint Committee on Cancer (AJCC) recommendations for primary hepatobiliary tumors.14 The results of this analysis will hopefully provide additional guidance identifying the patients most likely to benefit from surgical intervention and also help identify those at high risk for recurrence.
At the University of Pittsburgh Medical Center 204 patients underwent hepatic resection for metastatic colorectal carcinoma during an 11-year period from January 1981 to December 1991. These were all attempts at curative resection. The 130 men and 74 women ranged in age from 28 to 79 years with a mean age of 58.5 years (median, 60 years). Patient follow-up as of December 1993 ranged from 24 to 130 months (median, 69 months).
The interval between resection of the colorectal primary tumor and hepatic resection ranged from −6 months (primary not discovered until after resection) to 228 months with a median of 16 months. Two patients had Dukes’ A primary tumors, and 53 had Dukes’ B. Dukes’ C tumors comprised the largest group with 93 patients, and 56 patients had synchronous hepatic metastases. Most patients with Dukes’ D tumors (synchronous metastases) underwent hepatic resection within the first 3 months; however, nine patients were not referred or evaluated until after this interval.
Major hepatic resections were performed in 171 patients (84%). Of the 115 patients having right or left lobectomies. 96 patients underwent standard lobectomies. 15 had wedge resections in addition to lobectomy, three patients had “extended” lobectomies, and one patient had multiple wedge resections with a right lobectomy. Fifty-five trisegmentectomies were performed with 47 right and 8 left.8
Ninety-one patients had solitary lesions. 15 of which had minor resections and 76 required either lobectomy (n = 54) or trisegmentectomy (n = 22). Multiple metastases were resected in the remaining 113 patients with 40 having four or more lesions (range, 4 to 11 lesions).
A retrospective review of all available inpatient and outpatient records including operative and surgical pathology reports was performed. Patient follow-up and outcome were documented by clinical visits, telephone interview, or correspondence.
The following clinical and pathologic risk factors were examined for prognostic influence: gender, age, interval between colorectal and hepatic resection, Duke’s stage, site of colorectal primary tumor, histologic differentiation of the secondary tumor, number of metastases, tumor size (largest nodule diameter), unilobar or bilobar disease, type of hepatic resection, margin of resection, involvement of lymph nodes or contiguous structures, and intraoperative blood transfusion.
Survival time was calculated from the date of hepatic resection until death and disease-free survival from the date of hepatic resection until the time of tumor recurrence. Survival curves were generated by using the Kaplan-Meier (product-limit) method and were compared by using the generalized Wilcoxon (Breslow) test. Univariate Cox’s proportional hazards model was used to calculate the relative risk (RR) of mortality and tumor recurrence and 95% confidence intervals. Patients alive as of December 31, 1993, were right censored at time of follow-up. A multivariate stepwise Cox’s regression analysis (backward elimination method) was performed to identify factors that were independently associated with tumor recurrence and mortality. A p value <0.05 was considered statistically significant.
The overall survival was 91 % at 1 year, 43% at 3 years, and 32% at 5 years (Fig. 1). Median survival was 33 ± 2.22·(± SE) months. There were no deaths within the first 30 days of operation and only two deaths (1 %) within the first 120 days. As of December 1993, 72 of the 204 patients were alive and 50 patients were disease free at a mean of 65.1 ± 32.4 months (median, 62.5 months) after hepatic resection. Twenty-two patients were alive with recurrent disease at a mean of 20.7 ± 12.9 months (median, 18.5 months). Death with recurrent disease occurred in 124 (61 %) patients. Eight (4%) patients died free of disease at a mean of 27.9 ± 30.9 months (median, 13.5 months) after hepatic resection.
The actuarial overall and disease-free survival rates, stratified according to patient and primary tumor characteristics, are shown in Table I. Gender, incremental disease-free interval, site of primary tumor, and Dukes’ stage were not correlated with patient survival. Patients older than 60 years of age and patients with a less than 24-month interval between colorectal and hepatic resection experienced a poorer outcome (p < 0.05).
When disease-free survival rates were examined, men appeared to have longer recurrence-free survival. Early primary tumor stage (Dukes’ A or B) and more than 24-month interval between colorectal and hepatic resection were also found to be associated with longer disease-free survival. The remaining factors were not found to be significant predictors of disease-free survival when subjected to univariate analysis (Table I).
A number of hepatic metastasis features were found to affect patient and disease-free survival (Table II). The size (largest diameter) and differentiation of the metastatic tumor did not influence outcome when subjected to univariate analysis.
Patients with unilobar disease experienced superior patient and disease-free survival when compared with patients with bilobar disease (p < 0.0001). Of the 80 patients with bilobar disease, 52 (65%) had lesions requiring trisegmentectomy. The remainder underwent extended lobectomy (3), lobectomy plus wedge resection (16), or multiple wedge resections (9).
A positive resection margin was associated with an extremely poor outcome with a patient survival rate of 88% and 12% at 1 and 3 years, respectively. Most patients with positive margins had major hepatic resections (seven lobectomies, eight trisegmentectomies), and 13 (76%) of 17 had bilobar disease. Patients with resection margins of 1 mm to 1 cm and greater than 1 cm had similar 5-year patient and disease-free survival rates.
The extent of hepatic resection was associated with disease-free survival (p = 0.014); however, it was only marginally associated with patient survival (p = 0.058). Most patients underwent right or left lobectomy (n = 115). The second largest group (n = 54) underwent either right or left trisegmentectomies. These two groups of major hepatic resections had 5-year survival rates of 39% and 21 %, respectively. The bilobar tumor distribution of 16 patients required wedge resections in addition to lobectomy. The median survival in these patients was 18 months, which was considerably worse than the 43-month median survival rate for those requiring lobectomy alone (p < 0.0001). Patients with minor resections (wedge resection and left lateral segmentectomy) had expectedly higher 5-year patient and disease-free survival rates of 46% and 49%, respectively. Those who required multiple wedge (or segmental) resection (n = 9) fared significantly worse with a 3-year survival rate of only 33% and none at 5 years (p < 0.05). All of these patients had bilobar disease, and one third had positive margins or involvement of adjacent structures.
The number of metastatic tumors in the liver was found to be associated with both disease-free and patient survival. Patients with solitary lesions fared best with a 5-year survival of 45%. Survival rates for patients with multiple tumors, especially those with four or more, were significantly less (p < 0.003). Of the 40 patients with four or more lesions there were four patients with positive margins and six patients with nodal involvement or extrahepatic disease. The majority (29 of 40, 73%) had bilobar disease, and 37 required major hepatic resections (22 lobectomies and 15 trisegmentectomies).
Data on intraoperative transfusion were available for 182 of the 204 patients. The mean transfusion requirement for hepatic resection in this series was 3.68 units (± 4.1 units) of packed red blood cells with a median of 3 units. Forty patients (20%) did not receive a transfusion. Most of these patients underwent wedge resection, left lateral segmentectomy, right lobectomy, or left lobectomy (Table III). Almost all patients requiring blood (1 or more units) underwent major hepatic resections (78 right or left lobectomies, 45 trisegmentectomies). When subjected to univariate analysis, there was no significant difference in survival (p = 0.776). The 5-year survival rates of patients receiving 0, 1 to 5, and 6 to 10 units of blood were similar at 46%, 35%, and 41 %, respectively.
The interrelationship between tumor distribution, number of metastases, tumor size, and disease not confined to the liver was examined in accordance with a proposed TNM staging system for metastatic colorectal cancer (mTNM). (Table IV). Unilobar disease is confined to stages I through III, with bilobar disease and disease outside the liver comprising stages IVA and IVB, respectively. The actuarial patient and disease-free survivals correlated well with the proposed mTNM staging system (Fig. 2). Patients with stage I and II disease (n = 67) had the best patient and disease-free survival at 5 years of 61 % and 40%, respectively (p < 0.0001) (Table II). Patients with extrahepatic lymph node involvement or invasion of contiguous structures experienced expectedly poorer survival rates, and recurrent (or perhaps persistent) disease was ubiquitous.
Tumor size, unilobar or bilobar disease, and lymph node involvement were excluded from the multivariate analysis because the combination of these factors defines mTNM staging. They were excluded to avoid problems related to multicollinearity of risk factors.15 The results of the multivariate analysis of mortality and tumor recurrence are shown in Tables V and andVI,VI, respectively. The following factors were found to be independently associated with tumor recurrence: number of tumors (four or more), involved margins, and mTNM stages IVA and IVB. Similar results were obtained when analyzing mortality. Involved margins and stages IVA and IVB disease were found to be independent prognostic factors. Stage III disease was also found to be significant, whereas the remaining factors previously examined were not found to be significant in a multivariate context.
A number of staging systems have been described for metastatic colorectal cancer to the liver.16–19 Each has supportive data that have promulgated their use as significant prognosticators by their respective authors. Unfortunately, no system has been widely adopted to allow comparisons among studies. The application of Fortner’s revised staging system to our series would result in the vast majority of patients being categorized as stage I because of the small number with regional or extrahepatic disease (stages II and III).17 Our present analysis supports the further staging of disease confined to the liver (with complete extirpation) and its prognostic significance. Gennari et al.18 and Doci et al.19 suggested the use of a system more akin to TNM staging that took into account multiplicity and distribution, but the extent of liver involvement (percentage of parenchymal disease) was used rather than tumor size. We believed that tumor size, as measured in centimeters (maximum nodule diameter), and unilobar or bilobar involvement were more readily reproducible and applicable. After we analyzed the commonly reported potential risk factors influencing survival and recurrence, we applied a proposed mTNM staging system for hepatic metastatic tumors (Table IV). In our series of 204 patients the correlations to both overall and disease-free survival were shown to be highly statistically significant. In addition, the variables used in the staging system were all significant prognosticators when subjected to univariate analysis. The interrelationship among tumor size, multiplicity, and distribution is well illustrated by the proposed mTNM staging system, and it in turn reflects the influence of tumor burden on patient survival.
A number of studies have suggested that the presence of four or more metastases is a particularly adverse prognostic factor, but this has not been a consistent observation.5, 7, 13, 16, 20–22 In our previous report5 we had only seven patients with four or more lesions, and none survived more than 3 years. However, the results of this current analysis are statistically more compelling. The observed 20% 5-year survival rate of these patients argues against considering this an absolute contraindication to resective therapy. In contrast to the findings of Scheele et al., 4 we found that bilobar distribution of metastases had a significant deleterious effect on overall and disease-free survival. Interestingly, 29 of 40 patients with four or more lesions had bilobar disease as did the majority of patients with large metastases (10 to 20 cm). Consideration of tumor size alone did not reveal prognostic significance, but the consequences of size with respect to tumor distribution and requirement for more extensive (bilobar) resections were observed to be important.
A multiinstitutional retrospective review from the Registry of Hepatic Metastases found that resection margins of less than or equal to 1 cm had a negative effect on long-term survival. It is notable that resection margins were not available for most patients in their series and patients with positive margins were grouped with those having margin widths of 1 cm or less.7 When grouped separately we found no significant difference in survival rates unless the margin exhibited residual tumor. A positive resection margin was a powerful predictor of patient survival and recurrence in both a univariate and multivariate context. These observations were not surprising because residual tumor constitutes surgical treatment failure from the outset and outcome is expectedly poor.
Some reports have observed lower survival rates in older patients, but they did not approach statistical significance.5, 7 Although we found that patients who were older than 60 years had somewhat poorer survival rates, the risk of recurrence was similar to those younger than 60 years of age.
The metastasis-free interval after resection of the colorectal primary tumor appears to be an important prognostic factor. The Registry of Hepatic Metastases study by Hughes et al.7 found that a disease-free interval of less than 1 year was associated with inferior survival rates. Our analysis of incremental increases in disease-free interval revealed a trend for improved survival with increasing time. Patients with disease-free intervals of more than 24 months had superior overall and disease-free survival rates. It would appear that a longer metastasis-free interval connotes more favorable tumor biology and outcome.
In contrast to other previously reported studies we did not find metastatic tumor size, Dukes’ stage, tumor differentiation, or amount of blood transfused to be significant prognostic factors.6, 7, 11, 16, 19, 20, 23
The data from this updated series of hepatic resections for metastatic colorectal cancer affirm our resolve to continue an aggressive surgical approach to this disease. In addition to examining the influence of traditionally reported prognostic factors we classified our patients with a simple mTNM staging system. The resultant analysis revealed this staging system to be a significant prognosticator for both survival and recurrent disease. Its application would more readily allow comparisons among studies and allow for prospective evaluation of the staging criteria and the role of adjuvant treatment strategies. Further refinements of staging criteria could evolve to include biologic, molecular, or genetic factors in addition to the anatomic extent of disease, which is the primary basis for TNM staging at present.
Presented at the Fifty-first .annual meeting of the Central Surgical Association, Chicago, Ill., March 3–5, 1994.