|Home | About | Journals | Submit | Contact Us | Français|
Local excision (LE) has been used in an attempt to preserve anal function in T1-2 rectal carcinoma. The current study compares LE to radical resection (RR), each with or without radiation therapy (RT).
Patients reported to the SEER registry of the National Cancer Institute from 1988 to 2003 who had T1-2N0M0 rectal carcinoma were identified. A retrospective analysis of survival was performed using the Kaplan-Meier method. Comparative risks of mortality were evaluated using multivariate adjusted Cox regression models.
Of 4,320 patients, 13% underwent LE alone, 7% underwent LE plus RT, 70% underwent RR alone, and 10% underwent RR plus RT. On multivariate analysis, patients who underwent LE without RT had inferior overall survival compared to patients who underwent RR (P < .05). Patients who underwent LE with or without RT had inferior cause-specific survival compared to patients who underwent RR (P < .05).
RR without RT was associated with superior overall survival compared to LE without RT, and RR without RT was associated with superior cause-specific survival compared to LE with or without RT. Randomized trials are necessary to determine if LE with or without RT can offer equivalent survival compared to RR in early stage rectal carcinoma.
Radical resection (RR) with lymph node dissection offers excellent outcomes in early stage rectal cancer and represents the gold standard in terms of oncologic control. RR, however, is associated with increased morbidity and mortality compared to local excision (LE).1–5 Furthermore, RR of low-lying rectal cancers often requires colostomy. LE offers the possibility of minimal morbidity and sphincter preservation.
The study reported herein used the National Cancer Institute’s Surveillance, Epidemiology, and End Results (SEER) database to assess overall survival (OS) and cause-specific survival (CSS) in early stage rectal cancer treated with RR vs. LE, each with or without radiation therapy (RT).
Patients with primary invasive carcinoma of the rectum diagnosed from 1988 to 2003 were identified in the SEER 17-Registries 1973–2003 data set, November 2005 submission.6 Patients with distant metastases and/or lymph node metastases and patients with tumor size > 40 mm were excluded. The analysis was limited to patients with one primary only, or rectal cancer as the first of 2 or more primary malignancies, as the presence of a prior malignancy could influence treatment and survival. Tumor (T) staging is according to the American Joint Committee on Cancer (AJCC) Staging Manual 5th edition and was limited to T1 or T2.7 Patients were required to be pathologic N0 (if lymph nodes were assessed pathologically) or clinical stage N0 (if lymph nodes were not assessed pathologically).
Cancer directed surgery was performed in all patients. SEER database definition of cancer directed surgery excludes “surgical procedures performed solely for the purpose of establishing a diagnosis/ stage or for the relief of symptoms.” (SEER program code manual, 1988).8 Patients were classified as undergoing LE or RR as described in Table 1.
At least one lymph node was evaluated in 91% of patients treated with RR (median 6, mean 7.4 lymph nodes). In the LE group, no lymph nodes were evaluated in 91% of patients (median 0, mean 0.3 lymph nodes).
Rates of CSS and OS were estimated using the Kaplan-Meier method. Multivariate Cox proportional hazards regression analyses were conducted to identify predictors of CSS and OS.9 First, we fitted a preliminary model containing all the explanatory variables —T stage, grade, race, gender, age at diagnosis, year of diagnosis, histologic type, size of primary tumor, and the main variable of interest, treatment group. The functional form of the continuous variables was obtained by examining the martingale residuals of the null model; the effect of both size and year at diagnosis were well described by a linear term. Age, however, seemed to be better described with a quadratic term, especially in the case of overall survival.
We conducted an interaction test and also evaluated the underlying assumption of proportional hazards by examining the residuals and conducting a formal test. We found (for both CSS and OS) that no interaction term needed to be added to the model. The proportional hazards assumptions were met for all variables in the model.
A total of 4,320 patients were identified. Median follow-up was 47 months. Patient characteristics are described in Table 2.
Table 3 shows that patients who under-went RR were more likely to be male and to have a younger age at diagnosis, earlier year of diagnosis, histologic type adenocarcinoma (NOS), higher grade disease, higher T stage, and larger size of tumor compared to patients who underwent LE (P < .05). Table 4 shows that patients who received RT were more likely to be male and to have a younger age at diagnosis, later year of diagnosis, histologic type adenocarcinoma (NOS), higher grade disease, higher T stage, and larger size of tumor compared to patients who did not receive RT (P < .05).
The majority of patients who underwent LE +RT (95%) received postoperative RT. Therefore, radiation sequence with surgery was not evaluated as a prognostic variable. Radiation was delivered prior to surgery in 52% of patients who underwent RR. In these patients, it is possible that T and N staging are inaccurate, as RT can down-stage disease, and SEER coding does not differentiate clinical vs. pathologic staging. Therefore, OS and CSS is likely biased against those who underwent resection and RT.
Because RT is a treatment evaluated in this study and because RT typically requires 5–6 weeks to deliver, patients surviving less than one month (n = 22) were excluded from the survival analyses.
Prognostic variables by multivariate analysis for OS and CSS are depicted in Tables 5 and and6,6, respectively. OS and CSS curves are depicted in Figures 1 and and2,2, respectively (see page 110). Five- and 10-year OS and CSS rates are described in Tables 7 and and8,8, respectively. Compared to RR-alone, LE alone was associated with inferior OS (P < .0001). Compared to RR alone groups, OS rates in LE +RT and RR + RT groups were not statistically different (P = .09 for each). Compared to RR alone, LE ±RT was associated with inferior CSS (P < .01). Compared to RR only, RR +RT was associated with inferior CSS, but this difference did not reach statistical significance (P = .08).
OS curves for patients with T1 and T2 disease by treatment type are depicted in Figures 3 and and4,4, respectively. Compared to RR alone, LE alone was associated with inferior OS in both T1 and T2 patients (P = .0002 for T1 and P < .0001 for T2). OS in patients under-going RR + RT was not statistically different compared to RR alone (P = .14 for T1 and P = .09 for T2). OS in patients undergoing LE +RT was not statistically different compared to RR alone (P = .74 for T1 and P = .09 for T2).
CSS curves for patients with T1 and T2 disease by treatment type are depicted in Figures 5 and and6,6, respectively. Compared to RR only, LE ± RT was associated with inferior CSS in the T1 subset (P < .01). CSS in patients undergoing RR compared to RR +RT was not statistically different in the T1 subset (P = .29). In the T2 subset, CSS did not differ based on treatment.
On multivariate analysis, patients who were male, had T2 compared to T1 disease, adenocarcinoma in a villous adenoma compared to adenocarcinoma NOS, or were treated with LE compared to RR were more likely to die of causes other than rectal cancer (Table 9).
Prior to 1998, information on the type of LE is not available in SEER, though SEER does track whether or not pathologic specimens were sent for analysis. Following 1998, type of LE was recorded in some but not all patients. Given the small numbers of patients in subgroups by type of LE, meaningful statistical analysis could not be performed based on each individual type of LE. Patients undergoing polypectomy, excisional biopsy, or LE with biopsy (includes polypectomy, snare, or laser surgery) were evaluated. On multivariate analysis, CSS did not differ among these groups. OS was superior in patients undergoing excisional biopsy compared to polypectomy (Table 10, Figures 7 and and88).
Due to the morbidity associated with radical resection of early stage rectal cancer, in particular the need for colostomy in low-lying tumors, local excision has been used. The equivalence of LE compared to RR in terms of cancer control has not, however, been proven.
Lee et al reported a complication rate of 4% in patients undergoing transanal endoscopic microsurgery (TEM) vs. 48% in patients undergoing RR.3 Abdominoperineal resection (APR) is associated with risk of urinary dysfunction in up to 20% of patients and risk of erectile dysfunction and retrograde ejaculate in up to 50% of men.10 Nerve-sparing techniques, however, may reduce risk of serious urinary dysfunction to < 5% and reduce risk of sexual dysfunction in men to 13%–25%.11
Despite the potential morbidities associated with RR, patients are often able to adapt.13 Furthermore, quality of life with colostomy may be superior to quality of life with poor sphincter function following sphincter preservation.14
In a prospective study, Cancer and Leukemia Group B (CALGB) reported a 5-year disease-free survival (DFS) rate of 83% in T1 patients treated with full thickness LE and 71% in T2 patients treated with full thickness LE and chemoradiotherapy. While some studies have demonstrated similar 5-year DFS rates with RR in T1 disease, others have shown superior results.3,15–17 In T2 patients, historical 5- year DFS rates using RR are generally superior to that reported by CALGB.3,18
Local recurrence rates for T1-2 rectal cancer are higher following LE compared to APR in a majority of studies (Table 11).1,3,12,15–19 If these recurrences can be salvaged, then OS will not be compromised, despite differences in DFS. Unfortunately, salvage rates range from 38%–75%, and 35%–39% of patients will have synchronous distant recurrence.1,13,15,18,20
In the current study, LE alone is associated with inferior OS for both T1 and T2 patients compared to RR alone. A possible explanation for this observation is that patients with significant medical comorbidities or shorter life expectancy are more likely to be treated with LE rather than RR. Indeed, causes of death other than rectal cancer are higher in patients undergoing LE alone compared to RR alone. However, CSS is also inferior with LE alone compared to RR alone. In other words, patients undergoing LE are more likely to die of their cancer. Therefore, we believe that the most likely explanation for these results is that LE does not provide equivalent cancer control compared to RR, resulting in increased death from cancer.
When stratified by stage, the association of superior OS with RR alone compared to LE alone remains significant in both T1 and T2 patients. However, CSS remains significant only in the T1 subgroup. The lack of statistical significance in the T2 subgroup is likely due to a lack of power to detect a difference.
In the CALGB study of LE, a plateau for disease-free survival was not reached. Chakravarti et al also noted late recurrences (>5 years) in patients treated with LE + RT.20 Therefore, median follow-up times of 5 years or less, typical in many studies (Table 12), may not be adequate to determine the equivalence of RR to LE. In the current study, CSS decreased by 3%–4% between 5 and 10 years in patients undergoing RR±RT or LE +RT, but decreased 10% in patients undergoing LE alone. These results show that patients continue to die of their disease after 5 years, and these late deaths may be more prominent in patients undergoing LE alone.
Patient selection influences the success of LE. Lymphovascular invasion, poorly differentiated disease, higher T stage, larger size, tumor fragmentation, and involved or uncertain margins all portend higher risk of local recurrence.21
A major criticism of LE is that lymph nodes are not pathologically evaluated. Risk of lymph node metastases is 6%–18% for T1 tumors and 12%–28% for T2 tumors.15,18,22 Consideration of risk factors for lymph node involvement, including lymphovascular invasion, high grade, and involvement of the lower third of the submucosa, may aid in identifying patients at lower risk for nodal disease.19 Some authors report 0% lymph node involvement for T1 tumors without lymphovascular invasion.19
We chose to compare patients who were pathologic N0 (in the RR group) to patients who were clinical N0 (in the LE group) because in practice, treatment decisions regarding local excision are made based on clinical nodal stage. The superior survival in the RR group observed in this study could be due in part to inaccuracy of clinical nodal staging, suggesting a role for pathologic lymph node evaluation even in early stage disease.
Endoscopic ultrasound (EUS) evaluation of lymph nodes is superior to CT and MRI, though MRI with endorectal coil may yield similar accuracy to EUS.23,24 Land-mann et al report that in pathologic T1 and T2 rectal cancer, EUS missed nodal disease in 33% and 25% of patients, respectively, suggesting better predictors of lymph node involvement are necessary.23
Nodal regional lymph node spread in T1-2 disease is often in the mesorectum, which is removed in total mesorectal excision (TME).25 At the time that the CALGB study was reported, patients treated with LE were compared to historical controls using blunt dissection. TME has advanced the surgical care of rectal cancer, and has further reduced local recurrence compared to blunt dissection.26 The use of TME was not tracked by SEER during the time frame of this study. However, it is likely that a majority of RR patients did not undergo TME, as use of the technique did not blossom until the 1990s. In the modern surgical era, it is possible that TME offers a more substantial advantage compared to LE.
Techniques for LE have also improved. Transanal endoscopic microsurgery (TEM) offers full thickness excision of the tumor. View is expanded, which enhances the surgeon’s ability to visualize the tumor and resect it with adequate margins.3,27 Despite improvements compared to prior LE techniques, TEM does not remove mesorectum, and therefore does not address the potential for lymph node involvement in that region.
The goal of LE is full thickness excision with negative margins, but this ideal is difficult to achieve uniformly. In the CALGB study, of 161 patients enrolled, only 110 met eligibility criteria. The most common reasons for ineligibility were related to quality of excision and included lack of full thickness excision and involved or undocumented margin status.
A major limitation of the current study is inability to accurately define the type and quality of surgery. We do not have information regarding margin status, whether excision was full thickness, or if TEM was used. We did observe that OS but not CSS was superior in the subset of patients treated with excisional biopsy compared to polypectomy, which could reflect a tendency for less healthy patients to undergo polypectomy rather than excisional biopsy.
The addition of RT to LE has been employed in an attempt to decrease local-regional recurrence. Retrospective studies of LE ± RT are shown in Table 12.20,28–30 While the addition of RT to LE in the setting of T2 disease is generally accepted, the addition of RT to LE in T1 disease has largely been limited to patients with high-risk features, including positive margins, lymphovascular invasion, and high grade. However, retrospective series evaluating the use of RT in T1 disease have shown promise.
In the current study, RR alone is associated with superior CSS compared to LE +RT, but differences in OS do not reach statistical significance. This difference is likely due in part to a bias for patients with high-risk features not tracked by SEER (such as close/positive margins or lymphovascular invasion) to undergo RT in addition to LE. The other possible explanation is that the addition of RT to LE does not provide equivalent cancer control compared to RR. OS may be equivalent because the study is insufficiently powered to show a statistically significant difference. We attempted to limit biases related to perioperative mortality by excluding patients living less than 1 month from the time of diagnosis. Nonetheless, ill patients who are slow to recover from surgery are less likely to receive RT, and a cut-off time of 1 month survival does not eliminate this bias.
When stratified by stage, patients with T1 tumors demonstrate inferior CSS but equivalent OS with the use of LE+RT compared to RR. In the T2 patients, neither OS nor CSS are statistically different in patients undergoing LE +RT compared to RR. As discussed previously, the addition of RT to LE in T2 cancers is generally accepted as standard of care. However, the addition of RT to LE is not necessarily standard in the treatment of T1 disease in the absence of high risk features. Therefore, the most likely hypothesis for the inferior CSS in T1 patients treated with LE +RT is that patients who received RT in addition to LE have high-risk features that are not tracked by SEER.
In T2 patients, such biases are less likely to exist, as most LE patients will be treated with RT regardless of the presence of additional high-risk features. As a result, the RR vs. LE +RT groups might be better balanced in the T2 subset, and the addition of LE to RT does provide cancer control equivalent to RR. Alternatively, the study may simply be underpowered to detect differences in CSS in the T2 subgroup, or the RR patients may be more likely to have high-risk features not tracked by SEER compared to LE +RT patients.
Given concerns of decreases in OS and CSS associated with LE alone compared to RR in this and other studies, the addition of RT to treat T1 tumors warrants further investigation in prospective trials.
While the addition of RT to LE is attractive in terms of maximizing cancer control, side effects exist. In the acute setting, proctitis, dermatitis, diarrhea, and bladder/ urethral irritation can occur. Late/chronic side effects may include alteration in bowel habits, proctitis, rectal stricture, vaginal dryness, stenosis, or adhesions in women, fistulas, pelvic fractures, and secondary malignancies.31,32 In general, the risk of severe side effects is low.
In the adjuvant treatment of stage IIBIII rectal cancer, preoperative chemoradio-therapy improves local control compared to postoperative RT, enhances the rate of sphincter preservation, and reduces the rate of anastomotic stricture.33 The application of preoperative RT to early stage rectal cancers is therefore of interest. Several series have been reported, but follow-up is short and patient numbers are small.34
The addition of 5-fluorouracil (5-FU) chemotherapy to RT in the setting of locally advanced rectal cancer improves local control and survival.35 Phase I/II studies have reported higher pathologic complete response rates in locally advanced rectal cancer patients when newer chemothera-peutic agents are combined with 5-FU (or capecitabine) and RT.36,37 Biologic agents such as bevacizumab (an antiangiogenic agent) may also be of benefit.38 Therefore, advances in systemic therapy may potentially be combined with LE and RT in the future to improve outcomes.
Women were less likely to die of causes other than rectal cancer, possibly because they have fewer medical comorbidities. Another hypothis is that women, who were less likely to undergo RR or RT (P < .05), had a decrease in treatment-related morbidity and mortality. Female gender was associated with superior CSS, though this difference did not reach statistical significant (P = .0585).
Patients with T2 disease were more likely to die of causes other than rectal cancer compared to patients with T1 disease. Patients with T2 disease were also more likely to die of rectal cancer compared to patients with T1 disease. One potential reason for these findings is that patients with T2 disease have more serious medical comorbidities. This explanation, however, is unlikely, because T2 disease, like T1 disease, is often discovered at the time of routine screening colonoscopy and is not a hallmark of patients with poor access to or compliance with healthcare. Another possibility is that patients with T2 disease are more likely to undergo RR and/or RT (P < .05), which may result in higher treatment- related mortality.
Patients with adenocarcinoma in a villous adenoma were more likely to die of causes other than rectal cancer compared to other histologies. Adenocarcinoma arising in a villous adenoma was associated with inferior OS but not CSS compared to adenocarcinoma NOS. Villous adenoma of the rectum has been associated with an increased risk of subsequent colon cancer,39 and it is possible that the development of subsequent colon cancer resulted in inferior overall survival and an increase in death from causes other than rectal cancer.
While useful in the generation of hypotheses, SEER data cannot be used as proof of themerit of a particular therapy. SEER suffers from the inherent biases of retrospective data. It does not have information regarding high-risk features, including surgical margin status, lymphovascular invasion, and perineural invasion. Chemotherapy data are not available. Patient factors including comorbid disease, performance status, and weight loss are not included. Finally, the accuracy of entered data is not guaranteed. Because of these limitations, prospective data are needed.
In this retrospective study, RR alone is associated with superior OS compared to LE alone in early stage rectal cancer. A statistically significant difference in OS was not detected between RR alone vs. LE +RT. However, RR was associated with superior CSS compared to LE ±RT. These findings support RR as the gold standard treatment for T1-2N0M0 rectal carcinomas in terms of cancer control. It is possible that a subset of these early stage rectal carcinomas may be adequately treated with LE, but these data highlight a need to further characterize appropriate patient selection and treatment, and clinical trials should be encouraged to achieve this goal.
Disclosures of Potential Conflicts of Interest
The authors indicated no potential conflicts of interest.