The contents of this article do not represent the views of the Department of Veterans Affairs or the United States Government.
See editorial on pages 702-3, this issue.
Most estimates of short- and long-term survival after hepatic resection of colorectal cancer metastases are derived from surgical case series. For the current report, the authors used Medicare data to investigate operative mortality and long-term survival in a national sample and examined the factors associated with survival.
Data were analyzed from Medicare enrollees (age ≥65 years) who were admitted to hospital between January 1, 2000 and December 31, 2004 with a primary diagnosis of colorectal cancer with resection. The sample was restricted to patients who subsequently underwent hepatic resection for liver metastases. The Medicare Denominator File was used to determine operative mortality and long-term survival and the factors that were associated with those outcomes.
Of the 306,061 Medicare beneficiaries who were diagnosed with colorectal cancer, 3957 patients were identified who underwent hepatic resection for liver metastases. The crude 30-day and 90-day mortality rates were 4% and 8.2%, respectively, and the 5-year survival rate was 25.5%. Advancing age (hazards ratio [HR], 1.83; 95% confidence interval [95% CI], 1.32-2.53 for age ≥80 years vs ages 65-69 years), comorbid disease (HR, 1.40; 95% CI, 1.06-1.85 for Charlson ≥5 vs Charlson 0), and synchronous colon/hepatic resection (HR, 2.46; 95% CI, 1.89-3.20 for synchronous vs metachronous resection) were associated with worse 90-day mortality. Similarly, long-term mortality was associated with age (HR, 1.36; 95% CI, 1.18-1.56), comorbid disease (HR, 1.51; 95% CI, 1.36-1.69), and synchronous colon/hepatic resection (HR, 1.37; 95% CI, 1.24-1.51 for synchronous vs metachronous resection).
Colorectal cancer is the third most common cancer among men and women in the US.1 If it is diagnosed and resected at an early stage, then the prognosis is excellent. However, once disease has extended outside the bowel, the prognosis worsens substantially. The liver is the most common site of metastases. If left untreated, the 5-year survival rate is essentially nil.2 Therefore, medical and surgical strategies have been developed to treat advanced disease. Studies evaluating chemotherapy for metastatic colon cancer generally have been disappointing. For example, the median survival was <1.5 years in 2 studies of aggressive chemotherapy regimens for patients with advanced disease.3, 4 In contrast, for those with limited metastatic disease, reports from surgical series are much more favorable. A synthetic review comprising 19 case series that evaluated hepatic resection for colorectal metastases reported estimated 5-year survival rates between 21% and 44%.5 In what to our knowledge are the 2 largest US studies of this surgery, the 5-year survival rate was between 33%6 and 38%.7 Furthermore, the surgical mortality rate from hepatectomy generally is reported to be <5%.5, 8
Studies of surgical resection for hepatic metastases have several limitations. The data evaluating its effectiveness are derived almost exclusively from case series and usually are drawn from large academic centers, in which patient selection or surgical expertise may be superior to what is available in many communities. Many of these series are single-center studies that report on a small number of patients (averaging 283 patients in 10 US series).6, 7, 9-16 Finally, follow-up in these studies frequently is limited or incomplete. For example, data regarding mortality may be obtained exclusively through chart reviews rather than supplemented with the use of death registries. All of these factors may lead to an overly optimistic estimate of the benefits derived from hepatic resection for metastatic colon cancer.
In fact, a recently published study using administrative data (from the Surveillance, Epidemiology, and End Results [SEER] Program and Medicare) did investigate hepatic resection in patients with colorectal cancer metastases, and the authors reported that outcomes achieved nationally may not be as good as those observed in surgical case series.17 However, that study was limited in several ways. First, many of the patients included in the study were diagnosed with cancer over a decade ago (1991-2001), and therefore, the results may not reflect more current practice. Second, the study population was limited to those in the combined SEER-Medicare dataset (n = 833 patients) and, therefore, the study of factors associated with early and late mortality was limited.
Therefore, we assessed survival after hepatic resection of colorectal cancer metastases in a large population-based national sample using recent Medicare data. In addition to describing short- and long-term mortality in this population, we also examined important risk factors for these outcomes.
MATERIALS AND METHODS
By using Medicare's 100% national MedPAR files, all nonhealth maintenance organization (HMO) patients who were admitted to an acute care hospital between January 1, 2000 and December 31, 2004 with a first-time primary diagnosis of colorectal cancer with resection (International Classification of Diseases, 9th Revision [ICD-9] diagnosis codes 153 and 154 and Current Procedure Terminology codes 48.4-48.69 or 45.7-45.8) were identified. We identified patients with simultaneously diagnosed or subsequently developed liver metastases (ICD-9 code 197) who were admitted for hepatic resection (ICD-9 codes 50.22 and 50.3) between January 1, 2000 through December 31, 2004. The hepatic resection cohort consisted of patients with both the diagnostic code for liver metastases and 1 of the 2 surgical codes for hepatic resection.
Measures and Outcomes
Information on patient demographics (age, sex, and race) was obtained from the national Medicare Denominator File supplied by the Centers for Medicare and Medicaid Services. A modified (colon cancer removed) Charlson score was calculated using the comorbidities coded at the time of the initial colonic resection. Type of resection (partial hepatectomy vs lobectomy) was drawn directly from ICD-9 coding. Timing of the colonic and hepatic resections was defined as either synchronous or metachronous. Resections were considered synchronous if both surgeries (ie, colonic and hepatic resection) were performed during the initial hospitalization and metachronous if they were performed during separate hospitalizations. Metachronous resections were subclassified further between those performed within 1 year of initial primary resection (but not during the initial hospitalization) and those performed more than 1 year after the initial primary resection. The main outcome was survival after hepatic resection using the Medicare Denominator File to identify deaths through December 31, 2005.
Crude mortality at 30 days and 90 days was computed overall, by age group and according to the timing of resection. Long-term mortality was estimated using Kaplan-Meier survival curves overall and by age group to adjust for censoring because of mortality.18 For all survival analysis, we defined the index date as the admission date for hepatectomy. Cox proportional hazards regression was used to estimate the effects of patient characteristics (age, sex, race), comorbidity, and timing and type of surgery on 90-day and long-term mortality.19 Age, established at index date, was categorized into 4 groups (ages 65-69 years, 70-74 years, 75-79 years, and ≥80 years); race was categorized as black versus nonblack; and Charlson comorbidity was categorized as 0, 1, 2, 3 or 4, and ≥5 comorbidities. All analyses were performed using the SAS statistical software (version 9.1; SAS Institute Inc., Cary, NC). All statistical tests were performed at the 5% level of significance and were 2-sided.
We identified 306,061 Medicare beneficiaries who were diagnosed with and treated surgically for colorectal cancer. From this group, we identified 3957 individuals who underwent resection of the liver for metastatic disease. The characteristics of this cohort are reported in Table 1. Of those who underwent hepatic resection, 40% were aged ≥75 years, 45% were women, and 7.1% were black. Approximately 32% underwent synchronous resection (ie, same hospitalization) of their colon cancer and metastatic liver lesion, and the rest underwent metachronous resection (ie, during a subsequent or later hospitalization). Among the 2668 patients who underwent metachronous resection, 40% occurred within the first year after the colon resection, and 60% occurred ≥1 year after the colon resection.
Table 1. Baseline Characteristics of the Hepatic Resection Cohort
No. of Patients (%)
Crude Mortality at 90 Days
Charlson comorbidity score
Type of resection
Timing of hepatic resection
Metachronous at 1-12 mo
Metachronous at ≥12 mo
The crude 30-day mortality rate in this cohort was 4%; by 90 days, the mortality rate had doubled to 8.2%. Table 1 reports the crude 90-day mortality according to important risk factors. For patients aged ≥80 years, the 90-day mortality rate was 15.8% compared with 5.4% in patients ages 65 to 69 years. In univariate analysis, the timing of resection also was an important factor for predicting early mortality. At 90 days, the mortality rate for patients who underwent synchronous resection (13.3%) was more than twice that for the patients who underwent metachronous resection (5.8%). Because some patients with synchronous lesions (ie, detected at the time of their primary colorectal cancer) may have undergone delayed resection (after their first hospitalization with primary resection), we subclassified metachronous resection into those that occurred >1 year or <1 year after their primary colon resection. The 90-day mortality rate for patients who underwent synchronous resection was significantly greater than that for patients who underwent metachronous resection either during the first year after primary resection (4.8%) or 1 year after primary resection (6.5%).
In multivariate analysis (Table 2), advanced age and the timing of the hepatic resection remained significant predictors of early (90-day) mortality. In addition, hepatic lobectomy was associated with higher 90-day mortality compared with partial hepatectomy (hazards ratio [HR], 1.96; 95% confidence interval [95% CI], 1.54-2.49). Severe comorbid illness also independently predicted 90-day mortality. Specifically, patients who had the highest Charlson score were 40% more likely to have died at 90 days compared with patients who had the lowest Charlson score (HR, 1.40; 95% CI, 1.06-1.85).
Table 2. Results of Multivariate Regression Model on Short-term (90-Day) and Long-term Mortality
Overall, 5 years after hepatic resection, 25.47% of patients (standard error [SE], ±0.01%) remained alive (Fig. 1). Figure 2 shows the long-term survival of the hepatic resection cohort by age group. The 5-year survival rate ranged from 29.3% (SE, ±0.02) in the group ages 65 to 69 years to 17.1% (SE, ±0.02) in the group aged ≥80 years. Survival differed significantly across these age strata (P < .001).
The results of the multivariate long-term survival analysis are shown in Table 2. The factors that were most predictive of long-term mortality were increasing age (HR, 1.35; 95% CI, 1.18-1.55 for age ≥80 years vs ages 65-69 years), comorbid disease (HR, 1.51; 95% CI, 1.36-1.69 for highest vs lowest Charlson score), and synchronous hepatic resection (HR, 1.37; 95% CI, 1.24-1.51 for synchronous vs metachronous resection). Although the type of resection (hepatic lobectomy vs partial hepatectomy) was associated with short-term survival, it was not associated with long-term survival (HR, 1.09; 95% CI, 0.99-1.20).
The natural history of patients who develop hepatic metastases from colorectal cancer has been well described. Wagner et al2 reported the history of 252 patients with unresected disease. Although the patients in that study who had solitary lesions fared better than those who had widespread disease, essentially all died within 5 years. To improve mortality, surgeons have attempted resection of limited hepatic metastases. Early surgical reports of long-term survival after this procedure were met with some skepticism.20 However, case series demonstrating long-term survival with this intervention have continued to emerge, and hepatic resection currently is standard therapy for patients who have isolated liver metastases. In fact, US studies focusing on the use of this technique in experienced centers limited to the last 15 years suggest that 5-year survival rates approaching 60% can be achieved.21, 22 Some groups even have reported long-term ‘cure’ with this approach.23
The results from our population-based study suggest that the benefit from resection may be less than previously described by others. A 5-year survival rate of 25.5% in Medicare beneficiaries is the lowest reported of any US study to date and is significantly lower than the results reported in the 2 largest US studies. The larger of the 2 studies was a multicenter trial that included 607 patients and reported a 5-year survival rate of 33%.6 A single-center report that evaluated outcomes in 466 patients resulted in a 5-year survival rate of 38%.7 Some of these differences may be explained by the finding that many surgical series report long-term results only for patients who achieve negative margins at the time of resection. It has been demonstrated that surgical margin status is an important factor in predicting long-term survival.22 In this study, we have reported survival after surgery irrespective of postoperative margin status.
Another important finding of our study is that the early mortality risk associated with hepatic resection may be somewhat greater than was believed previously. The 30-day mortality rate of 4.2% is higher than the 0% to 3.8% rates reported in many US studies.7, 11-15 By 90 days, overall mortality nearly doubles to 8.3%. Although these deaths may not be attributable directly to the resection (ie, ‘surgical mortality’), such an estimate is useful to a patient who is considering the risks involved in this surgery. In fact, some other recent studies of hepatic resection also used the 90-day outcome measure because of the delayed nature of some of the complications (eg, progressive liver failure) that occur directly related to this surgery.24, 25 Although our results demonstrate a somewhat higher ‘early’ mortality than what has been reported in most other US studies, they are quite consistent with another, older US study that used administrative data. Wade et al9 used the Veterans Affairs (VA) patient treatment file and identified 161 Veterans (all men) who underwent hepatic resection for colorectal cancer metastases. In that study, the 30- and 90-day mortality rates were 4% and 8%, respectively, nearly identical to ours. A 4.3% 30-day mortality rate also was observed by Cummings et al in a cohort of patients who underwent hepatic resection derived from SEER-Medicare data.17
The current analysis highlights the higher risk of this surgery in patients of advancing age and in those who have metastatic disease identified at the time of their initial colorectal cancer diagnosis. These findings have not been confirmed uniformly by other studies. In at least 5 US studies7, 14, 16, 26, 27 no significant association was observed between age and outcome. Only 1 recent US study that focused exclusively on patients who underwent hepatic resection for colorectal metastases15 reported an effect similar to ours. Another very large surgical series of patients (n = 1059) who underwent hepatic resection for various indications (nearly all malignant) also has been reported.24 The indication for hepatic resection in that study was colorectal cancer metastases in 50% of patients. Age also was an independent predictor of mortality in that study. It is important to note that, when considering our analysis of age, our cohort was limited to those who were eligible for Medicare (aged ≥65 years). However, the average age in the 5 studies that reported no association of mortality with age was >60 years and, thus, did not differ substantially from our cohort. It is possible that the outcomes of older patients are better in these centers, because the criteria for selecting patients for surgery may be more stringent than that used more commonly in the community.
We also observed a marked difference in outcome between patients who underwent surgery for synchronous as opposed to metachronous metastases. Results from older case series were inconsistent in this area. Some investigators reported increased risk for those with synchronous disease,7, 14, 26, 28, 29 and others did not.12, 15, 16, 27 Although controversy in this area remains, most recent work suggests that synchronous resection can be considered,30-34 but only in well selected patient populations. For example, age34 and extent of metastatic burden (and, thus, extent of hepatic resection required)30, 31, 33, 34 appear to be important factors that increase the risk of early morbidity and mortality from synchronous resection. Our work suggests that, if such careful selection is required for these patients, then it may not be occurring regularly on a national basis.
The main strength of this study is its size and national representativeness. To our knowledge, it is the largest study of hepatectomy performed to date and represents a complete national sample of individuals aged >65 years who underwent surgery between 2000 and 2004, excluding only those who were enrolled in HMOs, or underwent their operation at a VA medical center. Most case series have emanated from single centers. The experience at these centers may not reflect the broader national experience of patients who undergo resection.
The current study had several limitations. First, these results reflect only the experience of patients aged >65 years. Therefore, the ability to generalize our results to a younger cohort cannot be assessed. However, >66% of incident colorectal cancers occur in this age group.35 A second limitation is that we could only use information available in administrative data and could not control for other important clinical variables, such as positive tumor margins or hepatic tumor size.7 Because tumor margins cannot be predicted preoperatively, however, the current study may provide more accurate estimates of the average risk experience of patients who are facing a decision regarding whether to attempt resection. Another potential limitation when using administrative data relates to the accuracy of the coding. However, our analysis derives from the diagnostic codes for cancer and various surgical procedures. Prior work suggests that coding for these items is quite good with sensitivity and specificity both >90%.36 Third, although, in the current study, we reported the true risk associated with undergoing this surgery, we did not compare the effectiveness of this therapy with other approaches for metastatic colorectal cancer (eg, nonoperative). Given the lack of trials assessing the effectiveness of this intervention, this limitation applies to all studies in this area. Finally, we have described the results after hepatic resection for the ‘average’ patient undergoing the procedure on a national level. It is possible that certain centers, because of patient selection and surgical expertise, perform substantially better than the results shown in this analysis. In fact, given the well documented association between volume and outcome for colorectal cancer resection37, 38 and hepatic resection,39 this seems likely. However, our results may be generalized to the majority of patients who undergo this procedure and are not at the small number of centers with particular expertise in performing this operation.
In summary, estimates of survival after hepatic resection for colorectal metastases derived from single-center series may be overly optimistic. On the basis of our work, a more reasonable 5-year survival rate for all comers is 25%. The surgical risk of undergoing this intervention may be greater than estimated in previous case series. Specifically, conventional ‘30-day mortality’ rates may not accurately reflect the early mortality experienced by patients who undergo this surgery. By 90 days, the overall mortality rate was 8% and exceeded 10% in certain subgroups (aged >80 years; synchronous resection). Finally, advancing age and synchronous resection predicted worse long-term survival.
Some investigators have suggested that randomized trials would not be appropriate to test the efficacy of surgical resection of isolated metastatic disease.7 However, our work suggests that it may be possible to identify some groups that incur the most up-front risk from this surgery while deriving the least long-term benefit. For example, patients aged >80 years may be 1 such group. Trials that directly compare less invasive therapies with surgical resection of metastases in terms of effectiveness and quality of life in these populations might be considered. Finally, it appears that the results achieved with surgical resection of colorectal metastases nationally may not be as good as the results achieved at select surgical centers with expertise in this area. Further work to gain a better understanding of the major drivers of this observed disparity (eg, patient selection) and programs to improve outcomes nationally are warranted.
Conflict of Interest Disclosure
Financial support was provided by grants from the Robert Wood Johnson Foundation and the National Institutes of Health (Grant CA52192). The funding organizations had no role in the design, collection, analysis, interpretation of the data, or preparation of the article.
Dr Robertson is supported by a Department of Veterans Affairs Health Services Research and Development Career Development Award.