Fax: (216) 983-0347.
Survival after hepatic resection in metastatic colorectal cancer
A population-based study
Article first published online: 19 JAN 2007
Copyright © 2007 American Cancer Society
Volume 109, Issue 4, pages 718–726, 15 February 2007
How to Cite
Cummings, L. C., Payes, J. D. and Cooper, G. S. (2007), Survival after hepatic resection in metastatic colorectal cancer. Cancer, 109: 718–726. doi: 10.1002/cncr.22448
- Issue published online: 2 FEB 2007
- Article first published online: 19 JAN 2007
- Manuscript Accepted: 13 NOV 2006
- Manuscript Revised: 8 NOV 2006
- Manuscript Received: 6 SEP 2006
- Postdoctoral Training Grant in Digestive Diseases. Grant Number: T32-DK61917
- American Cancer Society. Grant Number: RSGT-01-072-03-CPHPC
- Established Investigator in Cancer Prevention, Control and Population Science. Grant Number: (K05-CA90677)
- and End Results Program;
- colorectal neoplasm;
- neoplasm metastasis;
- surgical procedures, operative;
- outcome assessment
Hepatectomy is the standard of care for patients with colorectal cancer who have isolated hepatic metastases; however, the long-term survival benefits of hepatectomy in this population have not been characterized well outside of case series. For the current study, a population-based database was used to compare the survival of patients with liver metastases from colorectal cancer who did and did not undergo hepatectomy.
Patients aged ≥65 years with incident colorectal cancer who were diagnosed from 1991 to 2001 were identified from the linked Surveillance, Epidemiology, and End Results (SEER)-Medicare database. Liver metastasis diagnoses, colorectal resections, and hepatectomies were identified from hospital, outpatient, and physician-supplier claims. Patients who did not undergo colorectal resection were excluded. Five-year survival from the time of cancer diagnosis was determined by the Kaplan–Meier method. Cox proportional hazards models were used to evaluate survival.
Among 13,599 patients who were identified with incident colorectal cancer and liver metastases, 7673 patients (56.4%) presented with stage IV disease, and the remaining patients presented with earlier stage disease and developed subsequent metastases. Only 833 patients (6.1%) in the cohort underwent hepatic resection, and their 30-day mortality rate was 4.3%. The 5-year survival was 32.8% among patients who underwent hepatic resection, compared with 10.5% among patients who did not undergo hepatic resection (P < .0001), and better survival was observed in the subset of patients who presented initially with disease in stages I through III. In a Cox model, which was controlled for age, sex, race, comorbidities, and stage at presentation, lack of hepatic resection was associated with a 2.78-fold increased risk of death.
Although hepatectomy rates among patients with colorectal cancer were low, hepatic resection was associated with improved survival. Cancer 2007 © 2007 American Cancer Society.
The liver is the most common site for colorectal cancer metastases, and approximately 40% of patients with metastatic colorectal cancer have spread confined to the liver.1 Surgical resection is the current standard of care for patients with isolated hepatic metastases from colorectal cancer, because the natural history of patients with unresected disease is almost uniformly fatal, with median survival ranging from 14 months to 21 months.2 However, hepatic resection has been associated with complication rates of 20% to 50% and with perioperative mortality rates of 2% to 4%.3 Because the majority of patients who undergo hepatic resection develop recurrent disease,4 this approach is not curative.
The use of hepatic resection for metastatic colorectal cancer is based on case series from large surgical centers that have demonstrated 5-year survival rates that range from 32% to 38%.5–7 However, as suggested previously, such studies may overestimate survival because of reporting bias.8 It is noteworthy that perioperative mortality from hepatic resection is lower at higher volume hospitals compared with lower volume hospitals.9, 10 Moreover, tertiary care centers may be subject to selection bias because of referral patterns, limiting the generalizability of these studies. Because the only absolute contraindications for hepatic resection in patients with stage IV disease are poor overall health, widely disseminated disease, or inability to resect all of the affected liver, less favorable results may occur because of differing practice patterns.
Given the morbidity and mortality associated with hepatic resection and the limitations of prior case series, further research that includes procedures performed at smaller medical centers is important to improve health care quality and to clarify whether the benefits of hepatic resection outweigh the potential risks of mortality, morbidity, and adverse effects on quality of life. A recent observational study that used a nationally representative database demonstrated in-hospital mortality rates for hepatic metastasectomy from various primary malignancies that were similar to those noted in prior case series of hepatic resection for metastatic colorectal cancer.10 However, that study did not evaluate long-term survival, and malignancy diagnoses were not confirmed.
For the current study, we used the linked Surveillance, Epidemiology, and End Results (SEER)-Medicare database to analyze the use of hepatic resection in metastatic colorectal cancer at a population level, including evaluation of long-term survival. The specific objectives of this study were to determine rates of hepatic resection for colorectal metastases; to evaluate postoperative complications from hepatic resection, including perioperative mortality; and to compare long-term survival in colorectal cancer patients who did and did not undergo hepatic resection for liver metastases.
MATERIALS AND METHODS
Patients and Data Sources
This study was approved by the Institutional Review Board at University Hospitals Case Medical Center. The SEER tumor registry provides a reliable population-based source of information regarding patients with colorectal cancer. This tumor registry is a program of the National Cancer Institute and tracks cancer incidence in the United States. The SEER Program began collecting data in 1973 and currently collects patient-specific information about cancer cases from >12 cancer registries encompassing approximately 25% of the United States population.11 Patients in the SEER database who are also Medicare beneficiaries can be identified by linking the SEER and Medicare databases, resulting in a joint database that consists of patients aged ≥65 years who were diagnosed with cancer while living in one of the SEER areas. The Patient Entitlement and Diagnosis Summary File (PEDSF) includes all individuals in the SEER database who have been matched successfully with Medicare enrollment records. SEER data include demographic information, such as age, race, and sex; primary site; stage at presentation; histology; and site-specific surgery.
Procedures in the linked SEER-Medicare database can be identified from procedure codes (International Classification of Diseases, ninth revision-clinical modification [ICD-9-CM] and Current Procedural Terminology, fourth edition [CPT-4]) from Medicare hospital inpatient claims (MedPAR), physician-supplier claims, outpatient standard analytical file (SAF) claims, and the SEER database. It is noteworthy that the SEER database includes information regarding treatment of each incident cancer for only the first 4 months after diagnosis. Therefore, depending on the time elapsed from diagnosis to surgery, some surgical procedures identified from Medicare claims may not be corroborated in the SEER database.
The linked SEER-Medicare database was used to identify patients aged ≥65 years who were diagnosed between 1991 and 2001 with incident cases of invasive colorectal cancer and who also had a diagnosis of hepatic metastases based on MedPAR, physician-supplier, or outpatient SAF claims dated between 1991 and 2003. Because of the high likelihood of incomplete claims, patients who were enrolled in a Medicare Health Maintenance Organization (HMO) within 24 months after their cancer diagnosis or who were not enrolled in Medicare Part B for atleast 24 months after diagnosis or until death were excluded. Patients with a prior history of malignancy, unstaged disease, or carcinoma in situ were excluded. Patients with any stage colorectal cancer at presentation who did not undergo resection of their primary tumor were excluded because of the paucity of hepatic resections in this group. Patients without Medicare claims data were excluded. Patients who presented with stage I through III colorectal cancer but had a claims diagnosis of hepatic metastases within 6 months of diagnosis were excluded initially to minimize the inclusion of patients with misstaged disease; however, in a supplemental analysis, those patients were incorporated and reclassified with distant (stage IV) disease.
Two main groups of patients with liver metastases were identified. The first group, which we designated the localized/regional (LR) group, included patients with localized or regional disease (stages I through III) who underwent colorectal resection and later developed hepatic metastases. The second group, which we designated the distant-colorectal resection (DCR) group, included patients with distant-stage disease (stage IV) at presentation who underwent resection of their primary tumor. In a supplemental analysis, patients who presented with localized or regional disease (stages I through III) but had a claims diagnosis of hepatic metastases within 6 months of their cancer diagnosis were restaged with distant disease (stage IV) and were added to the DCR group to form the DCR-modified group. Each of these groups was subdivided further into 2 comparison groups based on whether patients underwent liver resection.
Patient-level demographic and clinical variables that were identified from the PEDSF file included age, sex, race, anatomic site, stage of disease at diagnosis (historic SEER stage), and Part B/HMO enrollment information. SEER stage was dichotomized as either localized or regional (stages I through III) or distant (stage IV).
Resection of primary tumor
Patients who underwent resection of the primary tumor were identified based on documentation of resection from the PEDSF file or from MedPAR, physician-supplier, or outpatient SAF files. The following site-specific surgery codes from PEDSF were included: local tumor excision; partial colectomy, but less than hemicolectomy; hemicolectomy; total colectomy; total proctocolectomy; colectomy or coloproctectomy with an en bloc resection of other organs; and colectomy, unspecified. The following colon resection ICD-9-CM codes from MedPAR and outpatient SAF claims were used: partial excision of large intestine (45.7), cecectomy (45.72), right hemicolectomy (45.73), resection of transverse colon (45.74), left hemicolectomy (45.75), sigmoidectomy (45.76), other partial excision of large intestine (45.79), total intra-abdominal colectomy (excision of cecum, colon, and sigmoid; 45.8), abdominoperineal resection of rectum (48.5), trans-sacral rectosigmoidectomy (48.61), anterior resection of rectum with synchronous colostomy (48.62), other anterior resection of rectum (48.63), and partial proctectomy (48.69). The following CPT-4 procedure codes for colon resection from physician-supplier files were included: partial colectomy (44140, 44141, 44143, 44144, 44145, 44146, and 44147), total abdominal colectomy (44150, 44151, 44152, 44153, 44155, and 44156), colectomy with removal of terminal ileum and ileocolostomy (44160), proctectomy (45110, 45111, 45112, 45113, 45114, 45116, 45119, and 45123), and pelvic exenteration (45126).
Patients with hepatic metastases were identified from Medicare claims that listed the ICD-9-CM diagnosis code for malignant neoplasm of the liver that was specified as secondary (197.7). For the small number of patients who underwent hepatic resection after their cancer diagnosis for whom this code was missing, a diagnosis of hepatic metastases was imputed. The following ICD-9-CM procedure codes were used to identify hepatic resection from MedPAR and outpatient SAF claims: partial hepatectomy/wedge resection (50.22), lobectomy of liver (50.3), and total hepatectomy (50.4). The CPT-4 hepatectomy codes for partial lobectomy (47120), trisegmentectomy (47122), total left lobectomy (47125), and total right lobectomy (47130) also were used to identify patients who underwent hepatic resection from physician-supplier claims. The surgical procedure date corresponding to the hepatic resection procedure code was taken as the date of hepatic resection from MedPAR claims. For the few MedPAR claims on which the hepatic resection date was missing, the midpoint of hospitalization was imputed. For hepatic resection procedures that were identified from physician-supplier claims, the procedure date usually was provided as a range of dates (from date of service through to date of service). For these procedures, the corresponding MedPAR claim was identified, and either the from date of service or the admission date was used as the hepatic resection date.
ICD-9-CM codes from MedPAR and physician-supplier claims within 1 year prior to cancer diagnosis were used to identify comorbid conditions. A modified comorbidity index based on the adaptation by Deyo et al. of the Charlson index was calculated.12 To qualify as a comorbid condition, we required conditions to be listed more than once in outpatient claims, as suggested previously.13 Malignancy and metastatic cancer were excluded from the comorbidity index calculation.
To analyze complications from hepatic resection, the MedPAR claim for hospitalization for hepatic resection was identified. For those patients whose hepatic resections were identified from physician-supplier claims but had no ICD-9-CM procedure code for hepatic resection on their MedPAR claims, when possible, the corresponding MedPAR claim that had dates of hospitalization encompassing either the from date of service or to date of service from the physician-supplier claim for hepatic resection was identified. For patients who underwent ≥1 hepatic resection, analysis of complications was limited to those that occurred during hospitalization for the first hepatic resection. The following ICD-9-CM diagnosis codes were used to identify hepatic resection-related complications from MedPAR claims: accidental laceration (998.2 and E870.0), postoperative hemorrhage (998.1), posthemorrhagic anemia (285.1), anesthetic reaction (995.4, E938, and E945.2), wound dehiscence (998.3), liver abscess (572.0), peritonitis (567.2), ileus (560.1), gastrointestinal hemorrhage (578, 578.0, 578.1, and 578.9), gastrointestinal complications (997.4), biliary fistula (576.4), intestinal fistula (569.81), stomach or duodenal fistula (537.4), and postoperative infection (998.5).
Statistical Analysis and Definitions
Categorical variables were compared by using the Pearson chi-square test or the Fisher exact test, as appropriate. P values ≤.05 were considered significant. Perioperative mortality was defined as death within 30 days from the date of hepatic resection. Data were analyzed with SAS software (version 9.1 for Windows; SAS Institute Inc., Cary, NC) and S-PLUS (version 7.0 for Windows; Insightful Corporation, Seattle, WA).
Survival was calculated as the number of months from cancer diagnosis to death. Median survival was calculated in months from the month of cancer diagnosis. Survival was censored at 5 years or at the latest date for which reliable information regarding vital status was reported (December 31, 2003), whichever came earlier. Survival >5 years after cancer diagnosis was not determined due concerns that longer survival may not necessarily reflect death from colorectal cancer in this older cohort of patients. Survivor functions were estimated by using the Kaplan–Meier method. The log-rank test was used to compare survival curves. Variables that were associated on univariate analysis with significant differences in 5-year survival were entered into a multivariate Cox proportional hazards model. If the variables were no longer statistically significant in the multivariate model based on the partial likelihood ratio test, then they were removed in a stepwise fashion from the model unless we believed that they were meaningful clinically. Continuous variables were then examined for linearity in the logit, categorized as appropriate, and the model was refit with the recoded variables.
From the linked SEER-Medicare database, 13,599 patients who met the inclusion and exclusion criteria for incident colorectal cancer with liver metastases at some point in their disease course were identified. Only patients who had undergone resection of the primary tumor were included in the cohort. In the entire cohort, 56.4% (n = 7673) presented initially with stage IV cancer, 31.3% (n = 4263) presented initially with stage II or III disease, and 12.2% (n = 1663) presented initially with stage I disease. The vast majority of patients were white (84.5%; n = 11,497), African Americans comprised 9.2% of the sample (n = 1253), and patients of other or unknown race comprised the remaining 6.2% of the sample (n = 849). Almost half of the patients (48.8%; n = 6642) were men. Only 13.4% of patients (n = 1825) had ≥1 comorbidity. The rate of hepatic resection in the entire cohort was low at 6.1% (n = 833). However, hepatic resection rates were higher in patients with no comorbidities (6.4%; n = 754) than in patients with ≥1 comorbidity (4.3%; n = 79), and this difference was statistically significant (P < .001).
Patients were grouped further on the basis of their presenting stage (Fig. 1), as discussed above. Baseline characteristics of the LR and DCR groups are shown in Table 1. Hepatic resection rates were low at 5.6% (n = 431) in the DCR group and 6.8% (n = 402) in the LR group. In general, patients who underwent hepatic resection tended to be younger and had fewer comorbid illnesses than patients who did not undergo hepatectomy. Among patients in the LR group, there was a significant difference in rates of hepatic resection by race (P = .02), with whites more likely than African Americans to undergo hepatic resection. This difference was not apparent in the DCR group. Patients in the DCR group were more likely to undergo hepatic resection if they had no comorbidities (P = .003).
|Characteristic||LR group, n = 5926||DCR group, n = 7673|
|HR, n = 402||No HR, n = 5524||HR, n = 431||No HR, n = 7242|
|Mean age ± SD, y||71.6 ± 5.1||75.6 ± 6.8||73.8 ± 6.4||75.8 ± 6.8|
|≥1 Comorbidities, %||10.5||13.8||8.6||13.6|
Complications for individuals who underwent their first hepatic resection were analyzed (Table 2). Of 833 individuals who underwent hepatic resection, only 824 could be analyzed for complications because of hepatic resection claims that could not be matched with MedPAR claims. In total, 384 complications among 263 individuals were identified, including 11.6% of patients (n = 96) who experienced ≥2 complications and 31.9% of patients who experienced ≥1complication. The most common complication was posthemorrhagic anemia in 12.4% of patients (n = 102), followed by gastrointestinal complications in 10.7% of patients (n = 88). None of the patients had a diagnosis code that indicated anesthetic reaction, biliary fistula, or stomach or duodenal fistula.
|No. of complications, any|
|Bleeding (posthemorrhagic anemia or postoperative hemorrhage)||130 (15.8)|
|Posthemorrhagic anemia||102 (12.4)|
|Gastrointestinal complications||88 (10.7)|
|Paralytic ileus||73 (8.9)|
|Postoperative hemorrhage||28 (3.4)|
|Postoperative infection||27 (3.3)|
|Accidental laceration||24 (2.9)|
|Gastrointestinal hemorrhage||15 (1.8)|
Of 833 patients who underwent hepatic resection, 4.3% (n = 36) died within 30 days of surgery. Of these, 16 patients presented initially with disease in stages I through III, and 20 patients presented initially with stage IV disease. The 30-day mortality rate in the LR group was 4.0%, compared with 4.6% among patients who presented initially with stage IV disease. There was no significant difference in 30-day mortality between the patients who presented initially with disease in stages I through III and patients who presented initially with stage IV disease.
The 5-year survival rate in the overall cohort was 32.8% among patients who underwent hepatic resection compared with 10.5% among patients who did not undergo hepatic resection (P < .0001; log-rank test). The median survival, which was calculated from the time of cancer diagnosis, was 45 months among patients who underwent hepatic resection versus 17 months among patients who did not undergo hepatic resection (Fig. 2A). Hepatic resection was associated with improved survival in the LR and DCR groups (Fig. 2B,C). We also examined the impact of stage at presentation on long-term survival among the subset of patients who underwent hepatic resection (Fig. 3). Not surprisingly, the 5-year survival rate was higher among patients who presented initially with disease in stages I through III (43.3% vs 22.9%; P < .0001; log-rank test).
Long-term survival in the overall cohort stratified by race, sex, and comorbidities also was evaluated. The 5-year survival rate was 11.4% among whites, 10.5% among African Americans, and 18.9% among patients of other or unknown race. There was a significant difference in survival among the 3 racial groups (P < .0001; log-rank test). The median survival was 18 months (95% confidence interval [95% CI], 18–19 months) in whites compared with 18 months (95% CI, 17–19 months) in African Americans and 23 months (95% CI, 21–25 months) in the other/unknown race group. The 5-year survival rate was 11.3% among women and 12.3% among men. Among patients with no comorbidities, the 5-year survival rate was 12.4% compared with 8.0% among patients with ≥1 comorbidity (P < .0001; log-rank test). The median survival in patients with no comorbidities was 19 months compared with 16 months in patients with ≥1 comorbidities.
Using Cox proportional hazards models, age at diagnosis, sex, race, hepatic resection, stage at presentation, and presence of comorbidities were examined as univariate predictors of death. In the overall cohort, on univariate analysis, age was associated with an increased risk of death (Table 3). Each increase in age by 1 year increased the risk of death by 3%. Male gender decreased the risk of death by 10%. White race or other/unknown race was associated with a decreased risk of death compared with African-American race. Patients who did not undergo hepatic resection had a 2.4-fold increase in the risk of death. The presence of ≥1 comorbidities increased the risk of death by 20%. Patients who presented initially with stage IV disease had a 3.3-fold increased risk of death compared with patients who presented initially with disease in stages I though III. Within the LR and DCR groups, age, male gender, lack of hepatic resection, and the presence of ≥1 comorbidities all significantly affected the risk of death. African-American race was associated with an increased risk of death in the LR and DCR groups.
|Overall cohort, N = 13,599|
|Age at Dx||1.03||1.03–1.04||≤.0001|
|No hepatic resection||2.42||2.21–2.66||≤.0001|
|Distant stage at Dx||3.29||3.16–3.42||≤.0001|
|LR group, n = 5926|
|Age at Dx||1.04||1.04–1.05||≤.0001|
|No hepatic resection||2.23||1.93–2.58||≤.0001|
|DCR group, n = 7673|
|Age at Dx||1.03||1.03–1.03||≤.0001|
|No hepatic resection||2.94||2.62–3.31||≤.0001|
Separate multivariate Cox models were developed for the overall cohort and for the LR and DCR comparison groups. The final model for each group is presented in Table 4. In the overall cohort, not undergoing hepatic resection was associated with a 2.8-fold increased risk of death. Male gender was not associated with a significantly decreased risk of death in the LR group but was retained in the model, because we believed it was significant clinically. Within the DCR group, age ≥80 years at diagnosis was associated with a 38% increased risk of death, and white or of other/unknown race was associated with a significantly decreased risk of death compared with African-American race (P < .01; partial likelihood ratio test). Within the LR and DCR groups, not undergoing hepatic resection was associated with at least a 1.9-fold increase in the risk of death.
|Overall cohort, N = 13,599|
|Age at Dx ≥80 y||1.47||1.41–1.53||≤.0001|
|No hepatic resection||2.78||2.53–3.04||≤.0001|
|Distant stage at Dx||3.52||3.38–3.66||≤.0001|
|LR group, n = 5926|
|Age at Dx||1.04||1.04–1.05||≤.0001|
|No hepatic resection||1.92||1.65–2.22||≤.0001|
|DCR group, n = 7673|
|Age at Dx ≥80 y||1.38||1.32–1.46||≤.0001|
|No hepatic resection||2.84||2.53–3.19||≤.0001|
In a supplemental analysis, the 1824 patients who presented with localized or regional stage but who had a claims diagnosis of liver metastases within 6 months of their cancer diagnosis were restaged as distant and were added to the DCR group to form the DCR-modified group. The rate of hepatic resection in the DCR-modified group was 5.8% (n = 547). As in the DCR group, patients in the DCR-modified group who underwent hepatic resection tended to be younger and had fewer comorbidities. Complication and perioperative mortality rates were similar in the DCR-modified group and the DCR group. The 5-year survival rate in patients who underwent hepatic resection was 34.2% compared with 12.2% in patients who did not undergo hepatic resection (P < .0001; log-rank test). The results of Cox proportional hazards modeling were similar in the DCR-modified group and DCR group, except that being restaged was associated with a 55% decreased risk of death in the DCR-modified group (hazard ratio, 0.45; 95% CI, 0.42–0.48), even after controlling for age, sex, race, hepatic resection status, and comorbidities. Because these restaged patients were excluded from the initial DCR group, there were no restaged patients in the nonmodified DCR group. Consistent with our other results, not undergoing hepatic resection was associated with an increased risk of death in the DCR-modified group (hazard ratio, 2.60; 95% CI, 2.34–2.90).
The current study was conducted to determine whether the complication rates and survival benefit observed for patients who undergo hepatectomy for metastatic colorectal cancer described in case series from tertiary referral centers also would be observed at a population level. The SEER-Medicare database provided a large sample of colorectal cancer patients aged ≥65 years with information regarding staging and surgical procedures.
The actual rate of hepatic resection in the overall cohort was surprisingly low at 6.1%. Among patients in the DCR group, the rate of hepatic resection was 5.6% compared with 6.8% in the LR group. This rate was relatively low but probably was not the result of underreporting, because SEER and Medicare data have a high level of agreement for identifying colorectal cancer patients who have not undergone surgery.14 This hepatectomy rate was consistent with the results published in 2004 by Temple et al., who used SEER-Medicare data to evaluate surgical treatment for patients with stage IV colorectal cancer.15 In that study, the researchers observed that only 4.8% of patients who underwent cancer-directed surgery underwent hepatectomy within 4 months of diagnosis. Among patients who did not undergo cancer-directed surgery within 4 months of diagnosis, only 0.4% underwent hepatectomy.
The relatively low rates of hepatectomy observed in the current study may reflect the perceived morbidity of this procedure in an elderly population, although prior studies have produced conflicting results regarding the impact of older age on long-term outcome.6, 16 In addition, even the LR group, which presumably may have been more likely to include patients with isolated hepatic metastases (because these patients presented initially with disease in stages I through III), had a hepatic resection rate of only 6.8%. Our cohort also had a higher 30-day mortality rate (4.3%) than the rates observed in previous studies, which reported 30-day mortality rates from 0% to 2.8%,5, 6, 17 and this difference may reflect the inclusion of patients who underwent hepatic resection at lower volume centers. However, the complication rate calculated for the current study (31.9%) was comparable to that reported in other series (19%–31%).6, 17, 18
In the current study, hepatic resection was associated with improved survival in the overall cohort, with a 32.8% 5-year survival rate after cancer diagnosis in hepatectomy patients and a 10.5% 5-year survival rate in nonhepatectomy patients. In comparison, 5-year survival rates of 37% and 40% were reported in prior case series from large surgical centers in which survival was calculated from the time of liver resection rather than from the time of diagnosis.6, 17 Median survival from the time of liver resection was 42 months and 46 months, respectively, for those 2 case series. Despite differences in time points from which 5-year survival was calculated between these case series and the current study, it is apparent that the patients in both case series had a longer median survival. This discrepancy may have been caused by selection bias resulting from the experience at a single tertiary referral center. It is noteworthy that a European multicenter case series of 1568 patients who underwent hepatectomy for colorectal metastases demonstrated a 5-year survival rate of only 28%, although perioperative deaths and patients with extrahepatic disease were excluded from the survival analysis.18 Taken together, these results suggest that long-term survival rates after hepatic resection may not be as optimistic as was reported previously from single centers. Nonetheless, hepatic resection is associated with improved survival compared with no resection.
The addition of 1824 patients who were restaged from localized/regional disease to distant disease because of a claims diagnosis of liver metastases within 6 months of their cancer diagnosis had minimal impact on our conclusions. It is noteworthy that being restaged was associated with a hazard ratio of 0.45 in the final Cox proportional hazards model for the DCR-modified group, suggesting that restaged individuals experienced a different disease course from that of the rest of the DCR-modified group.
There were several limitations to this study. First, because of the administrative nature of these data, comorbidities and complications may have been undercoded,19, 20 because these diagnoses may not have had much impact on reimbursement. For example, no hepatic resections complicated by biliary fistulae were identified, even though this complication has been reported at a rate of 3% to 5% in case series.4 The ICD-9-CM code that was used to identify biliary fistula specifies a choledochoduodenal fistula and, presumably, would not apply to other diagnoses without specific diagnosis codes available, such as an intra-abdominal bile leak or biliocutaneous fistula. Second, the dataset was limited to patients aged ≥65 years. However, because colorectal cancer incidence increases with increasing age, this population has the highest incidence of colorectal cancer.21, 22 Third, the use of nonsurgical treatments, including radiofrequency ablation, cryotherapy, and chemotherapy, was not assessed. Radiofrequency ablation confers a minimal survival advantage over chemotherapy alone,23 and chemotherapy for metastatic colorectal cancer has become more common only in the last few years with the introduction of new chemotherapeutic agents, such as irinotecan, oxaliplatin, bevacizumab, and cetuximab; whereas the current cohort included patients who were diagnosed through December 2001. Although all patients in the cohort had liver metastases, it was not possible to distinguish among those with isolated, resectable hepatic metastases, those with unresectable hepatic metastases, and those with distant metastases in multiple organs. Finally, patients who underwent hepatic resection generally were younger and healthier than patients who did not undergo hepatectomy. However, not undergoing hepatic resection still was associated with a 2.78-fold increased risk of death after adjusting for age and comorbidities.
Data from the current study suggest that hepatectomy rates among patients with colorectal cancer were low, although these rates reflect a period when 5-fluorouracil was the primary chemotherapy modality for colorectal cancer. Current practice incorporating newer chemotherapeutic agents for metastatic colorectal cancer may result in higher hepatectomy rates. The current data also suggest that hepatectomy for metastatic colorectal cancer was associated with higher perioperative mortality than was reported previously. Patients who underwent hepatectomy appeared to have improved long-term survival compared with patients who do not undergo hepatectomy, but their 5-year survival was lower than reported previously in case series. Further research should investigate potential selection biases in patients who undergo resection.
We acknowledge the efforts of the Applied Research Program, National Cancer Institute; the Office of Research, Development, and Information, Centers for Medicare and Medicaid Services; Information Management Services, Inc.; and the SEER Program tumor registries in the creation of the SEER-Medicare database
- 3Surgical management of hepatic colorectal metastases. In: SaltzLB, ed. Colorectal Cancer: Multimodality Management. Totawa, NJ: Humana Press; 2002: 375–389., .
- 11Overview of the SEER-Medicare data: content, research applications, and generalizability to the United States elderly population. Med Care. 2002; 40( suppl): IV-3–IV-18., , , , .
- 13Assessing comorbidity using claims data: an overview. Med Care. 2002; 40(suppl): IV-26–IV-35., , .
- 14Use of SEER-Medicare data for measuring cancer surgery. Med Care. 2002; 40(suppl): IV-43–IV-48., , , , , .