The effectiveness of this follow-up has been difficult to measure and thus has been questioned. Steele [2] concluded that these follow-up strategies were empirical and probably could not be justified on the basis of cost-effectiveness or benefit to patients. He recommended that patients be told to return to their physicians if they developed signs or symptoms of recurrence.

Most reports of salvage surgery for specific sites of recurrence or metastasis (such as the liver or the lungs) have come from academic practices, and the frequency and outcomes of salvage surgery in community settings are unknown. Furthermore, most reported experiences with salvage surgery date from the 1970s. Since that time, technical advances have been made in imaging, colonoscopy, hepatic surgery, thoracic surgery, and adjuvant therapy programs.

This report is based on retrospective data collected from a national trial of adjuvant therapy after surgery for colon cancer [3, 4]. Most patients in this trial were from community-based oncology practices and were accrued in the mid-1980s. Salvage surgery was done during the past decade. Our objectives were to determine the frequency of salvage surgery, describe factors motivating salvage surgery, and assess outcomes and surgical mortality.

Methods

Top Methods Results Discussion Author & Article Info References

We reviewed the records of surgically staged patients with stage II (Dukes B2) or stage III (Dukes C) colon cancer who entered this trial of adjuvant therapy after surgery for colon cancer, designated Intergroup 0035 by the National Cancer Institute. Patients with stage II cancer were randomly assigned to receive observation or 1 year of fluorouracil plus levamisole therapy followed by observation. Patients with stage III cancer could also be randomly assigned to receive levamisole alone. The methods used to measure the original trial's primary end points-recurrence and survival-are described elsewhere [3, 4]. Statistical analyses were done by using Statistical Analysis System procedures [5], and survival curves were generated by using Kaplan-Meier methods [6]. The effects of individual prognostic variables on disease-free survival were assessed by using log-rank statistics [5].

Results

Top Methods Results Discussion Author & Article Info References

A total of 1247 eligible patients were followed for a median of 7 years (range, 5 to 9.5 years); patient characteristics and study end points are documented elsewhere [3, 4]. Of these 1247 patients, 26% had stage II disease and 74% had stage III disease. Four hundred seventy-four patients were assigned to observation alone, 310 were assigned to receive levamisole, and 463 were assigned to receive levamisole plus fluorouracil.

Follow-up

Observation required visits every 3 months during year 1, every 4 months during year 2, and every 6 months thereafter for 5 years. Follow-up continued until year 8, with the type of test, frequency of testing, and schedule of physician office visits determined at the discretion of the treating physicians. At each visit, a history was taken and physical examination, hemography, liver panels, and chest radiography were done. Proctoscopy-barium enema or colonoscopy was done at 24 and 48 weeks and then annually. Results of optional CEA testing, CT, magnetic resonance imaging (MRI), abdominal ultrasonography, and all pertinent procedures were recorded, as were surgical and pathology reports. At least one CEA level was documented in 82% of patients.

Motivating Factors for Salvage Surgery

Cancer recurred in 548 of 1247 patients (44%). Of these 548 patients, 222 (41%) had salvage surgery (this excludes "open biopsies" and surgery done to palliate patients who were known before surgery to have unresectable recurrent disease). In addition, 11 patients had exploratory surgery (8 for elevated CEA levels and 3 for findings on CEA-directed imaging) but no documented recurrence. Salvage surgery was done a median of 16 months (range, 1 to 80 months) after primary surgery. Most procedures were performed in year 1 (n = 79) or year 2 (n = 85).

We reviewed each chart to identify the factor that led to salvage surgery (Table 1). If this information could not be derived from a chart, physicians were queried. Any patient who reported symptoms that could have been due to recurrent disease was classified as manifesting a clinical presentation. In asymptomatic patients, results of mandated or discretionary follow-up studies that led to the identification of recurrence were recorded. Patients with abnormal results on screening tests and elevated CEA levels were classified as having been identified by screening tests. Asymptomatic patients with normal results on screening tests and elevated CEA levels that led to additional confirmatory clinical studies were classified as having been identified by CEA screening. Patients whose physicians, at their discretion, regularly performed scanning that discovered recurrence were classified as having been identified by routine imaging. In several patients, the motivation for salvage surgery could not be identified despite queries directed to the treating physicians.

In all 222 patients, preoperative evaluations suggested that curative-intent surgery was feasible. Table 1 shows the numbers of patients who had curative-intent surgery and who were alive and disease-free after curative-intent surgery, according to the motivating factor for this surgery. We defined curative-intent surgery as complete resection of all known cancer with tumor-free margins; this was accomplished in 75% of patients when surgery was motivated by follow-up. Surgery motivated by CEA levels (usually performed for hepatic metastasis seen on CT) led to curative-intent surgery in 63% of patients. Only 30% of patients who presented with clinical symptoms had curative-intent surgery.

Curative-intent surgery was thought to be successful in 109 patients. Of these patients, 27 (25%) had clinical manifestations of bowel obstruction, abdominal pain, or gastrointestinal bleeding leading to surgery. In 36 of the 109 patients (33%), the initial motivation for surgery was a finding on follow-up, most often an abdominal mass, nodules on chest film, or local recurrence on endoscopy or colography. The most common motivation, elevated CEA levels, led to 38% of curative-intent procedures (41 of 109). In these cases, surgical indications were provided by subsequent imaging results, usually hepatic metastases on CT.

Record review identified 19 patients whose recurrences seemed to identify them as good candidates for surgery (such as isolated pulmonary or hepatic lesions in accessible areas) but in whom surgery was not considered. These patients were observed or received chemotherapy. They survived for a median of 26 months after relapse, and 2 were alive at last analysis. Site-specific median survival times after relapse were 20 months in 15 patients with hepatic relapse and 27 months in 4 patients with pulmonary relapse.

Salvage Surgery for Specific Organ Sites

Table 2 shows the numbers of patients who had salvage surgery and curative-intent salvage surgery, according to site of recurrence. The most frequent sites of recurrence were the liver, the local site, the abdomen, and the lungs. Curative-intent surgery was attempted in most patients who had local recurrence or metastatic disease confined to the liver, lungs, ovary, or abdominal wall. Curative-intent surgery was accomplished in one third of patients with intra-abdominal or retroperitoneal metastasis and 11% of patients with multiple organ involvement.

Results of Salvage Surgery

The median follow-up period after curative-intent surgery for the 109 patients in whom this surgery was thought to be successful was 72 months (range, 20 to 103 months). Sixty-five patients died of cancer, 11 survived with known recurrence, 5 died without evident recurrence, and 28 survived disease-free. The initial motivating factors for surgery in these 28 patients are shown in Table 1, and the sites of recurrence are shown in Table 2. Among the disease-free patients, the median follow-up period was 60 months (range, 14 to 94 months). Kaplan-Meier estimates showed that the median disease-free survival time after curative-intent surgery was 19 months and that 23% of patients were disease-free at 5 years. The median survival time for the 109 patients was 33 months: Forty-five percent were still alive at 3 years, and 35% were still alive at 5 years. In contrast, for the 113 patients who had incomplete resections during salvage surgery, the median survival time was 11 months and the 3-year survival rate was 6%.

Curative-Intent Surgery for Specific Organ Sites

Hepatic Metastasis

Hepatic resection was attempted in 49 patients; 2 of these patients (4%) died after surgery. In 28 of these 49 patients (57%), curative-intent resection was done and the median interval from initial surgery to salvage surgery was 17 months (range, 2 to 61 months). All but 3 patients had surgery within 6 weeks after recurrence was recognized. Seventeen patients had solitary lesions, and 11 patients had at least 2 lesions. The median diameter of the longest tumor was 3.7 cm (range, 0.8 to 11 cm). Depending on disease location and extent, wedge resection, segmentectomy, or lobectomy was done. Ten of 28 patients (36%) who had curative-intent surgery were alive and disease-free at a median of 61 months (range, 25 to 90 months) after salvage surgery. As Figure 1 shows, the estimated median disease-free survival time of these patients is 22 months, and their 5-year recurrence-free survival rate is 32%.

View larger version (17K): [in this window] [in a new window]   Figure 1. Disease-free survival after curative-intent salvage surgery by site of relapse.

Two adverse prognostic factors were 1) an interval of less than 1 year between primary surgery and salvage surgery and 2) multiple metastases. Only 1 of 10 patients who had initial and salvage surgery within 1 year survived without recurrence. All 7 patients with more than one nodule who had curative-intent resection died of colon cancer.

Local Recurrence

Forty-four patients had salvage surgery for local recurrence; 1 died after surgery. The most frequent motivation for surgery was routine bowel imaging results (20 patients), but 15 patients were symptomatic, with bleeding, pain, or changed bowel habits. Twenty-four patients had curative-intent surgery, and the rest had palliative resections or bypass procedures. Curative-intent surgery was done promptly; it was delayed beyond 6 weeks in only 3 patients. The median interval from resection of the primary lesion to curative-intent surgery was 14 months (range, 2 to 71 months). Abdominal perineal or low anterior resection was done in 10 patients, segmental resection was done in 11, hemicolectomy was done in 1, and subtotal colectomy was done in 2. Seven of the 24 patients who had curative-intent surgery (29%) are alive and disease-free (median follow-up, 67 months [range, 20 to 95 months]). One patient died of an unrelated cause 9 months after salvage surgery. The estimated median disease-free survival interval was 26 months, and the 5-year recurrence-free survival rate is 27% (Figure 1). No prognostic determinants were identified in these patients.

Pulmonary Metastasis

All 20 patients who had salvage surgery for pulmonary metastasis had curative-intent surgery; none died after surgery. All had recurrence identified by routine chest radiography. The median interval from primary surgery to salvage surgery was 35 months (range, 10 to 65 months). Sixteen of 20 patients who underwent resection had solitary lesions, 4 had multiple lesions, and 4 had hilar node metastasis at surgery. Seven patients had wedge or segmental resections, and the rest had lobectomy. In contrast to patients with local and hepatic recurrences, 45% of patients with pulmonary recurrences had surgery delayed 2 to 12 months after recognition of recurrence. Among the 20 patients who had curative-intent resections, 6 (30%) are currently alive and disease-free (median follow-up period, 62 months [range, 41 to 75 months]). Three patients died of unrelated causes without relapse at 9, 11, and 22 months after salvage surgery. For all patients who had a potentially curative procedure after pulmonary relapse, the estimated median disease-free survival time was 22 months and the estimated 5-year recurrence-free survival rate was 27%. No prognostic determinants were evident in these patients, although the 2 patients who had initial recurrence within 1 year have had recurrence, as have all 4 patients with more than one pulmonary lesion at initial recurrence.

Other Metastatic Sites

Attempts at salvage surgery for other metastatic sites were disappointing. Of 63 patients with metastases to the abdomen, ovaries, retroperitoneal nodes, abdominal wall, and rare sites, 32 (51%) had curative-intent surgery and only 5 are currently alive and disease-free (median interval from salvage surgery to last contact, 95 months [range, 47 to 101 months]). Isolated metastatic lesions were seen in the small bowel serosa in 2 survivors, the bladder in 1 survivor, an adrenal gland in 1 survivor, and an ovary in 1 survivor. The 5-year estimated recurrence-free survival rate for these 32 patients was 16%.

Multiple Metastatic Sites

Surgery for metastases involving more than one organ system was nonproductive. Five of 47 patients had curative-intent surgery, and all died of recurrence.

Results of Salvage Surgery by Treatment Group

The standard of care for patients with stage III colon cancer has shifted from observation to use of fluorouracil plus levamisole, leucovorin, or both; examination of the results of salvage surgery by treatment group is therefore relevant (Table 3). Although fewer patients receiving fluorouracil plus levamisole had recurrence, the percentages of patients with recurrence in whom salvage surgery was attempted and curative-intent surgery was performed were similar across the treatment groups. However, the 5-year disease-free survival rate was lower in patients who underwent curative-intent surgery after receiving fluorouracil plus levamisole than in patients who underwent curative-intent surgery after observation alone (18.1% compared with 29.5%).

Prognostic Factors for Recurrence-Free Survival after Curative-Intent Surgery

We evaluated factors other than specific organ involvement to determine their relation to recurrence-free survival after curative-intent surgery (Table 4). The time from primary surgery to initial recurrence had some influence: Patients who had recurrence in less than 1 year had a poor outlook, and those who had recurrence after 3 years had much more favorable results (Figure 2). Most striking was the influence of the number of resected metastatic lesions (Figure 3). After we adjusted for the other covariates, only the presence of multiple as opposed to single lesions was a significant independent predictor of subsequent disease-free survival (relative risk, 2.2 [95% CI, 1.4 to 3.5]; P = 0.001).

View larger version (15K): [in this window] [in a new window]   Figure 2. Disease-free survival after curative-intent salvage surgery by relapse interval.

View larger version (11K): [in this window] [in a new window]   Figure 3. Disease-free survival after curative-intent salvage surgery by number of lesions.

Second Primary Colorectal Cancers

Twenty-four patients (1.9%) developed second colon carcinomas. In 9 of these patients, advanced disease (presumably originating from the first cancer) was evident. The other 15 patients had curative-intent surgery. A second colon carcinoma was discovered because of symptoms in 2 patients and anemia in 2 patients, was found incidentally at appendiceal surgery in 1 patient, and was discovered by positive results on fecal occult blood testing in 1 patient and by routine bowel surveillance in 9 patients. The median interval from resection of the first primary cancer to recurrence was 21 months (range, 6 to 69 months). Ten patients had stage I disease, 4 had stage II disease, and 1 had stage III disease. Two of 15 patients developed metastatic disease, 1 died of an unrelated cause, and 12 are currently alive and disease-free (median interval, 68 months [range, 39 to 96 months]). The estimated 5-year recurrence-free survival rate was 80%. None of the 9 patients in whom second lesions were detected by routine bowel surveillance have had recurrence.

Discussion

Top Methods Results Discussion Author & Article Info References

In the years after they had primary surgical resection of stage II and stage III colon cancer, 20% of our patients (41% of patients with recurrence) had laparotomy or thoracotomy for resection of recurrent disease. Salvage procedures were more common than we expected. Postoperative monitoring and surgery cost thousands of dollars per patient and may be done for 100 000 new patients annually in the United States. These efforts will cure only a small percentage of patients; thus, there is some debate about whether this practice is justifiable. Nelson [7] concluded that "although universally performed, postoperative diagnostic screening in patients with colorectal cancer has not been demonstrated to diminish overall mortality. The burden of proof is on those who propose benefit." Despite this argument and those of Steele [2], routine follow-up is prevalent. Our results suggest that this follow-up has a tangible clinical benefit in patients in whom recurrent disease is detected early and salvage surgery with curative intent can be attempted. Because most of our disease-free patients have been followed for more than 4 years after curative-intent surgery, it is reasonable to conclude that they have been cured.

The 5-year disease-free survival rate of 32% seen in patients with hepatic metastasis in our study is similar to the 25% rate seen in 607 patients in the multi-institutional retrospective study by Hughes and coworkers [8], the 31% rate seen in 555 patients collected by Ballantyne and Quin [9], and the rates in 13 series cataloguing outcomes in almost 2600 patients, reviewed by Fong and associates [10]. In patients with pulmonary metastasis, our disease-free survival rate of 27% is similar to the rate of 31% in a collected series of 301 patients reported by Turk and Wanebo [11] and the rate of 30.5% in 139 patients reported by McAfee and colleagues [12]. No reports of patients with curative-intent resection of locally recurrent colon cancer exist for comparison with our 5-year disease-free survival rate of 27%. The postoperative mortality rate of less than 2% is encouraging.

In assessing the outcomes of curative-intent surgery, it is tempting to compare the 48 patients who had hepatic or pulmonary surgery for relapse with the 19 patients who had recurrences that seemed resectable but who were managed nonsurgically. Among patients with hepatic relapse, median survival after relapse was 33 months in 28 patients after curative-intent surgery and 20 months in 15 patients who did not have surgery; for patients with pulmonary relapse, median survival after relapse was 49 months in 20 patients after curative-intent surgery and 27 months in 4 patients who did not have surgery. It is impossible to determine whether survival differences resulted from surgery or selection bias. Bias is inherent in this type of comparison: The 48 patients who had surgery had to survive long enough to have a second operation. Although more than 50% of these patients had surgery within 2 to 3 weeks of relapse, 6 had surgery 4 to 22 months after diagnosis. Another source of bias is uncertainty about whether curative-intent surgery was actually possible in all 19 patients who were managed without surgery.

The good survival rates seen in the 109 patients who had curative-intent surgery might be attributable to the selection of patients with indolent disease and good prognosis, regardless of intervention. Nevertheless, recurrence-free survival curves appear to plateau after 3 years, indicating that curative-intent surgery can enhance cure rates. The long follow-up period for the 28 disease-free patients (21 survived for more than 3 years and 16 survived for more than 5 years) would speak strongly for cure.

Expense is a significant challenge in follow-up for colorectal cancer. In our study, the maximum yield of additional cures that could be attributed to postoperative follow-up was 28 patients with recurrence and 14 patients with second primary cancers (overall, 3.4% of the total patient group).

One approach to improving postoperative follow-up for colorectal cancer would be to increase effectiveness and reduce costs. Our data suggest several practice changes that might be beneficial. First, surgical resection during year 1 after surgery rarely results in cure. Progressive disease identified during that year requires a change in therapy, but aggressive surgical intervention may not be justified. Of 250 patients who had recurrence during year 1, only 28 had curative-intent surgery, and only 5 of the 28 came from the cohort that received fluorouracil plus levamisole. Of these 28 patients, 7 were alive and disease-free 2 years after salvage surgery (Figure 2). Only 1 patient who had recurrence detected by standard follow-up procedures during year 1 remains disease-free.

With regard to specific follow-up tools, chest radiography is a low-cost method for diagnosing pulmonary metastasis. In our study, half of patients with newly diagnosed solitary pulmonary metastasis had surgical treatment delayed from 2 to 12 months after the first abnormal chest radiograph. More expedient resection may have enhanced outcomes.

Diagnosis of local recurrence or second primary bowel cancer requires imaging or inspection of the bowel. Unfortunately, fecal occult blood testing cannot be used as a surrogate for these procedures. Ahlquist and colleagues [13] found that hemoccult testing recognized only 21% of proven intraluminal recurrences and only 33% of new colon tumors.

It was disappointing that few patients were disease-free 5 years after curative-intent surgery for hepatic metastasis. Even though the frequency of initial recurrence in the liver exceeded that of initial recurrence in either the lung or the local site of disease by several-fold in this trial [3], the total number of curative-intent procedures was only slightly higher in the hepatic metastasis group. The success rate in the hepatic metastasis group when curative-intent surgery could be done was similar to that in the pulmonary or local recurrence groups. Perhaps the disproportionate overall yield in cures of patients with hepatic metastasis reflects the aggressive nature of hepatic recurrence.

The possibility that our tools for recognizing hepatic recurrences that may respond to salvage surgery are inadequate or are used suboptimally must be considered. We show that hepatic metastasis (for which curative-intent surgery is feasible) was diagnosed almost entirely by monitoring of CEA levels. Routine liver imaging was optional in this trial. In a previous review, however, CEA testing was found to lack sensitivity and specificity and to be associated with false-positive and frequent false-negative results [14].

Overall surgical salvage cure rates were similar in patients with elevated CEA levels, patients with normal CEA levels, and patients who were not monitored at all. In addition, 50% of patients who had hepatic resection after CEA level elevation had large lesions (range, 4 to 10 cm in diameter). It seems that increasing cure rates for patients with hepatic metastasis will require more effective diagnostic approaches in addition to better surgical techniques and methods to minimize postoperative mortality.

Our data suggest that follow-up after potentially curative surgery for colon cancer begins usually results in curative salvage surgery when recurrence is detected during years 2 and 3. This is particularly evident in the patients who received fluorouracil plus levamisole. Table 3 shows that only 57 of these patients (12%) had recurrence during year 1; only 1 of these 57 patients achieved 2-year disease-free survival after curative-intent surgery. Chest radiography done every 6 months for 12 to 48 months and prompt resection of newly identified solitary pulmonary nodules may be reasonable. For recognition of local recurrence and second primary cancers, periodic colonoscopy may be useful. Whereas barium enemas detect polyps greater than 1 cm with an efficacy similar to that of colonoscopy, radiography does not allow for inspection and biopsy of the anastomotic site [15, 16]. Our data justify annual examinations for 3 to 5 years. If a patient has three consecutive colonoscopic examinations with negative results, less frequent reexamination (as recommended by Winawer and coworkers [17] on the basis of data derived from the National Polyp Study) seems appropriate.

These recommendations overlap with but are more intensive than the recent clinical guidelines from Winawer and coworkers [18] for colorectal cancer screening in persons with resected colon cancer. In that study [18], initial colonoscopy within 1 year after colorectal cancer resection was suggested if the colon was inadequately assessed before surgery. The next colonoscopic examination was offered at 3 years. In our study, 68 of 1247 patients (5.4%) developed second primary or locally recurrent tumors, most within 3 years of resection of the initial primary cancer. The recommendation of Winawer and coworkers [18] decreases the possibility of discerning anastomotic recurrence or diagnosing polyps or tumors when surgical intervention may be most productive.

Clearly, the outcomes of patients with hepatic resection need to be improved. Wernecke and associates [19] identified 100% of lesions greater than 2 cm and 66% of lesions 1 to 2 cm in diameter on ultrasonography; these results are similar to those obtained with CT and MRI in the same patients. For lesions less than 1 cm in diameter, ultrasonography was distinctly inferior. Certainly, we could hope for greater sensitivity for CEA monitoring than was shown in this study, where only three lesions measuring 1 to 2 cm in diameter were detected. The specificity of ultrasonography should also be very high because little difficulty (except with small lesions) is expected in distinguishing the hypovascular colonic cancer metastasis from common benign cysts and hemangiomas.

Computed tomography has the advantage of ready interpretation by the clinician and is less operator-dependent. Routine use of the more costly MRI or positron-emission tomography as a screening tool has little justification. Although ultrasonography or CT of the liver is more expensive than CEA testing, the finding of an elevated CEA level must be supplemented by radiographs to identify lesions for salvage surgery. The aggregate cost of this strategy may well exceed the cost of monitoring with ultrasonography or CT.

Finally, it is important to understand the potential value of surgery in the postoperative care of patients with colorectal cancer. The generalist, the medical oncologist, and the gastroenterologist often coordinate patient follow-up and must appreciate the curative potential and low risk of salvage surgery. Because such surgery and the monitoring procedures associated with it play major and costly roles in the management of patients with colorectal cancer, further research is needed to identify the most cost-effective strategies for monitoring patients after colon cancer resection.

Dr. Fleming: Department of Biostatistics, University of Washington, F600 Health Sciences, SC-32, Seattle, WA 98195.

Dr. Tangen: Department of Biostatistics, University of Washington, 1501 Fourth Avenue, Suite 2105, Seattle, WA 98101.

Dr. Macdonald: The Saint Vincents Comprehensive Cancer Center, 111 Eighth Avenue, Suite 1510, New York, NY 10011.

Dr. Haller: University of Pennsylvania Cancer Center, 6 Penn Tower, 3400 Spruce Street, Philadelphia, PA 19104.

Dr. Laurie: Grand Forks Clinic, 1000 South Columbia Road, Grand Forks, ND 58201-4032.