Abstract
Background: There is a lack of outcome data beyond local recurrence rates after primary treatment in rectal cancer, despite more information being necessary for clinical decision-making. We sought to determine patient selection, therapeutic modalities and outcomes of locally recurrent rectal cancer treated with curative intent.
Methods: We searched MEDLINE (1990–2010) using the medical subject headings “rectal neoplasms” and “neoplasm recurrence, local.” Selection of cohort studies was based on the primary intention of treatment and availability of at least 1 outcome variable.
Results: We included 55 cohort studies comprising 3767 patients; 8 studies provided data on the rate of intentionally curative treatment from an unselected consecutive cohort of patients (481 of 1188 patients; 40%). Patients were symptomatic with pain in 50% (796 of 1607) of cases. Overall, 3088 of 3767 patients underwent resection. The R0 resection rate was 56% (1484 of 2637 patients). The rate of external beam radiotherapy was 100% in 9 studies, 0% in 5 studies, and ranged from 12% to 97% in 37 studies. Overall postoperative mortality was 2.2% (57 of 2515 patients). Five-year survival was at least 25%, with an upper limit of 41% in 11 of 18 studies including at least 50 resections. We found a significant increase in reported survival rates over time (r2 = 0.214, p = 0.007).
Conclusion: More uniformity in treatment protocols and reporting on outcomes for locally recurrent rectal cancer is warranted. The observed improvement of reported survival rates in time is probably related to better patient selection and optimized multimodality treatment in specialized centres.
Substantial progress has been made in local control of rectal cancer in the past decades. First, anatomic consideration of the mesorectal fascia has led to the development of total mesorectal excision (TME). This surgical technique encompasses sharp dissection under direct vision resulting in resection of the rectum and mesorectum, ideally with an intact visceral pelvic fascia covering the resection specimen. The TME technique has become widely accepted, as local recurrence rates declined from greater than 20% with conventional blunt dissection to 5%–10% with TME.1 Second, the use of neoadjuvant therapy further reduced local recurrence rates, as shown in a meta-analysis of randomized controlled trials.2
The local recurrence rate is the most relevant end point for interventional studies on neoadjuvant treatment in patients with rectal cancer, because survival is either not or only marginally improved.2 However, there is a lack of adequate data beyond local recurrence rates, despite more information being necessary for clinical decision-making. Therefore, the aim of this systematic review was to analyze the recent literature on intentionally curative treatment of locally recurrent rectal cancer.
Methods
We searched MEDLINE using the following medical subject headings: “rectal neoplasms” and “neoplasm recurrence, local.” The search was carried out for the period 1990–2010 using the limits “humans” and “all adult: 19+ years.” We reviewed all retrieved articles and systematically screened the reference lists of selected articles for additional studies of interest. The date of the most recent search was Oct. 1, 2010.
Selection of cohort studies in patients with locally recurrent rectal cancer was based on the primary intention of treatment and availability of at least 1 outcome variable, either local control or survival. Multiple studies from the same institution were included only if there were different inclusion criteria based on patient characteristics or type of treatment and if there was no substantial overlap.
We extracted the following data from the included papers (when provided): study period, selection criteria, treatment characteristics of primary rectal cancer, characteristics of local recurrence (symptom score S0–2), synchronous metastases, time interval from primary tumour to local recurrence, completeness of resection (R status), percentage of sacral resection, application of external beam radiotherapy (EBRT) for local recurrence (timing, dose, reirradiation, concurrent chemotherapy), use of intraoperative radiotherapy (IORT), operative mortality, follow-up, crude distant metastasis and local control rate, actuarial local control and survival and median survival.
Statistical analysis
We used descriptive statistics to determine overall results for evaluable studies. Analyses were performed using SPSS for Windows version 16.0.2. (SPSS Inc.).
Results
Description of selected cohorts
We selected 55 cohort studies comprising 3767 patients for detailed analysis (Fig. 1 and Table 1). Twenty-four studies provided data on patient selection leading to the final cohort (Table 1). The authors used several selection criteria based on type of treatment for primary rectal cancer, characteristics of local recurrence, type of treatment for local recurrence and intraoperative findings. In 8 studies,9,13,16,20,42–44,54 the total number of patients undergoing treatment with curative intent from an unselected consecutive cohort of patients could be determined: 481 of 1188 patients (40%). The median period of patient inclusion was 13.3 (range 3.9–34) years in 51 studies (Table 1).
Therapy for primary rectal cancer
The surgical procedure for the initial tumour was described in 44 studies (Table 1). The median abdominoperineal resection (APR) rate was 27% (range 0%–62%) in these studies. Local excision was specified as the primary surgical treatment in 13 studies; 2 of them26,35 exclusively included patients who had undergone local excision, and the median local excision rate was 10% (range 6%–30%) in the remaining 11 studies.4,7,17,20,25,29,34,40,44,45,47 Included patients had undergone EBRT as part of the primary treatment in a median of 28% (range 0%–100%) of cases in 43 studies (Table 1). Previous radiotherapy dose was specified in 10 studies,4,7,9,12,19,27,28,41,55,57 with a total dose ranging from 10 to 110 Gy. The median percentage of patients who underwent chemotherapy as part of the treatment for the primary tumour was 37% (range 0%–100%) in 21 evaluable studies (Table 1).
Clinical presentation of local recurrence
Patients were asymptomatic (S0) when presenting with locally recurrent rectal cancer in 29% (571 of 2000) of cases in 27 studies (Table 1). Symptoms were classified according to presence of pain in 23 studies (Table 1), and 50% (796 of 1607) of patients were symptomatic with pain (S2). Five studies4,14,31,33,39 reporting on symptoms had a clinically relevant patient selection (sacral resection or intraluminal recurrence only). After exclusion of these studies, the rates of S0 and S2 clinical presentation were 30% (563 of 1884) and 47% (700 of 1491), respectively. The median interval between primary tumour and local recurrence ranged from 12 to 44 months in 40 studies (Table 1), with an overall maximum interval of 354 months, as reported by Das and colleagues.57
Data on synchronous distant metastases were provided in 48 studies (Table 1). In 18 of these, there was no clinical sign of distant dissemination at the time of treatment for local recurrence. The proportion of patients in whom distant metastases were identified ranged from 2% to 41% in the other 30 studies, with a reported percentage below 10% in 14 of these studies (Table 1). All distant metastases were resected with curative intent in 11 of these 30 studies (3%–16% of included patients), while selected patients received surgical treatment for distant disease in another 8 studies (2%–26% of included patients; Table 1).
Surgery for local recurrence
All patients from 30 studies underwent resection of the pelvic recurrence (Table 2). In the remaining 25 studies, the percentage of resection varied from 17% to 99%, with a resection rate above 70% in 15 of these 25 studies (Table 2). Differences in resection rates were mainly explained by the fact that patients with unresectable disease or intraoperatively detected distant disease were not excluded from the initial study population in some studies. The overall number of patients who underwent resection was 3088 of 3767 in the 55 selected studies. In the 51 studies specifying the duration of the study period, the median number of resections performed per year was 3.6 (range 0.8–19.6; Table 2). After excluding multicentre studies, only 12 centres performed at least 5 resections per year. The median reported R0 resection rate of 46 studies was 59% (range 14%–100%; Table 2). The overall calculated R0 resection rate was 56% (1484 of 2637 patients). A macroscopically complete resection (R0/R1) could be achieved in a median of 85% (range 31%–100%) of patients in 38 studies (Table 2). Part of the sacrum was included in the resection in all patients in 5 studies12,14,33,40,51 and in no patients in another 17 studies (Table 2). In the remaining 20 studies reporting on sacral resection, the median rate was 13% (range 2%–73%; Table 2).
Perioperative treatment for local recurrence
External beam radiotherapy was uniformly performed either in the preoperative or in the postoperative setting for all patients in 9 studies, whereas no patients in 5 of the studies underwent EBRT. The rate of EBRT reported in 37 other studies ranged from 12% to 97% (Table 2). External beam radiotherapy was given preoperatively in 37 studies; in 20 of these, radiotherapy was uniformly applied in the preoperative setting (Table 2). Postoperative EBRT was described in 25 studies, and irradiation occurred exclusively in the postoperative setting in 8 of these 25 studies (Table 2). Radiotherapy dose was specified in 28 studies.3–9,11,18,19, 28,29, 30,33–37,41,44,46,47,50–53,56,57 Without prior EBRT, the total dose was 40–60 Gy in most studies, with an overall range of 15–80 Gy. Conventional fractionation (1.8–2.0 Gy per fraction) was uniformly applied in 13 studies,3,6,7,9,11,18,19,28,30,36,46,50,52 and selective use of a hypofractionated schedule (fractions of 5 Gy) was described in 2 studies.34,44 In 3 studies,27,41,57 all patients had undergone EBRT for the primary tumour; the median (and range) rates were 50.4 (30–55) Gy, 50.4 (30–74) Gy and 47 (25–70) Gy, respectively. Reirradiation schedules were as follows: 40.8 Gy in daily fractions of 1.2 Gy, median 34.8 (range 15–49.2) Gy in daily fractions of 1.2 or 1.8 Gy, and 39 Gy in fractions of 1.5 Gy twice daily, respectively.27,41,57 An additional 8 studies3,19,28,40,42,47,52,56 described the use of re-EBRT, 7 of which specified the rate.3,19,28,40,42,47,52 The radiotherapy schedule was specified in 3 of these studies,28,52,56 with a total dose ranging from 23.4 to 30.6 Gy in daily fractions of 1.2 or 1.8 Gy. External beam radiotherapy was part of a chemoradiotherapy regimen, mostly fluorouracil-based, in 27 studies (Table 2); routine use of chemoradiotherapy was reported in only 3 studies.18,27,41
Intraoperative radiotherapy (IORT) was routinely used as part of the treatment protocol in 7 studies; in another 16 studies, the rate of IORT varied from 8% to 91% (Table 2). Thirteen studies3,6,7,11,15,19,25,28,35,36,47,52,57 specified the total dose of IORT, which mostly ranged from 10 to 20 Gy.
Outcome after treatment with curative intent
Postoperative mortality was 2.2% (57 of 2515 patients), across 46 studies (Table 3), most of which defined postoperative mortality as in-hospital mortality or 30-day mortality; 2 studies48,52 defined it as 60-d and 90-d mortality, respectively. The median follow-up was less than 36 months in 24 of 37 studies with available data (Table 3). The observed crude rate of distant metastases during follow-up varied between 9% and 68%, with a median rate of 41% across 29 studies (Table 3). The crude local control rate in patients who underwent R0 resection ranged between 29% and 100%; the rate was above 75% in 9 of 14 studies (Table 3). The actuarial local control rate for R0 resection was more than 55% in 7 of 8 studies (Table 3). The most consistently reported outcome parameter was 5-year survival, which ranged from 11% to 51%. Five-year survival was at least 25%, with an upper limit of 41%, in 11 of 18 studies including at least 50 resections (Table 3). Figure 2 displays the reported 5-year survival rates of series without selection for type of surgical resection, by year of publication. The curve fit model demonstrates a significant increase in reported survival rates over time (r2 = 0.214, p = 0.007).
Topographical differences
We grouped studies into 4 categories according to the geographic areas from which they originated: Europe, the United States, Japan and the remaining countries. As shown in Table 4, the highest rate of prior EBRT was found in the United States (50%), and EBRT was most frequently used to treat local recurrence in Europe. Japan had the lowest rates of EBRT and IORT, but a more aggressive surgical approach was suggested, with 39% of resections including part of the sacrum.
Discussion
In general, local recurrence is the reflection of an aggressive biological behaviour of the primary tumour as it is accompanied by synchronous distant disease in a high percentage of patients. In the Dutch TME trial,58 83 of 129 patients (63%) with local recurrence also had distant metastases. The Swedish Rectal Cancer Trial and Stockholm I trial reported distant disease in 66 of 143 patients (46%) and 86 of 156 patients (55%) with local recurrence, respectively.59,60 Our systematic review suggests that 40% of unselected consecutive patients with locally recurrent rectal cancer are candidates for intentionally curative treatment. Curative treatment of both local and distant recurrence is achievable in only a small subgroup of patients (Table 1). Probably such treatment should only be considered in patients with indolent tumour behaviour based on a long disease-free interval (at least 2 yr) from primary treatment.
Overall, half of the patients were classified as having S2 clinical presentation (symptomatic with pain). Although these were selected patients who were candidates for intentionally curative treatment, this finding is concordant with an unselected cohort of 156 patients with locally recurrent rectal cancer from the Stockholm Rectal Cancer Study: the rates of S0, S1 and S2 clinical presentation were 13%, 33% and 54%, respectively.60 Intractable pain associated with a pelvic recurrence is an awful clinical condition. While distant disease is the determining factor for prognosis in most of these patients, local recurrence will generally affect quality of life.61,62 Complete or partial initial relief of pain after radiotherapy alone or after multimodality treatment, including surgery, is reported in up to 83% of patients, although the rate of long-term pain-free survival is about 30%.27,41,63
The 2 most important predictors of radical resection of local recurrence are previous anterior resection instead of APR and the absence of pain at the time of recurrence.43,64,65 Intraluminal recurrences, especially after initial local excision, are separated from the bony pelvis and sacral nerves by remaining soft tissue, thereby not resulting in pain and enabling resection with adequate margins in almost all patients.26,35,39 Given the worse outcome for local recurrence after prior APR, optimal primary treatment of distal cancers is of utmost importance. The pelvis becomes narrower at the level of the levator ani. When following this natural curve, the surgeon will end up with a so-called coning resection, thereby increasing the risk of tumour-positive margins. The extralevatoric APR with en bloc removal of the levator muscle in combination with downstaging by neoadjuvant therapy will improve local control in distal rectal cancers.66,67
We found a wide variety in treatment protocols with regard to perioperative radiotherapy for locally recurrent rectal cancer reported in the studies we reviewed (Tables 2 and 4). Some institutes did not include EBRT or IORT in their protocols, while the indication for radiotherapy ranged from highly selective to routine use at other centres. In addition, EBRT had been applied either in the preoperative or in the postoperative setting. Chemotherapy was not always added as radiosensitizer, especially in series published before 2005. The indication for EBRT did not depend only on prior EBRT; comparing data on EBRT from Tables 1 and 2 revealed that some patients received no irradiation during the entire treatment, neither for the primary tumour, nor for the local recurrence. On the other hand, there is a tendency toward reirradiation in patients with locally recurrent rectal cancer after prior pelvic radiotherapy.
There is no concluding evidence to determine the most optimal treatment strategy for locally recurrent rectal cancer. This is also related to the heterogeneity of the disease, as shown in the present review, based on type of primary surgery, previous radiotherapy, extent of recurrent disease (i.e., fixation grade, extension to pelvic sidewall) and the presence of symptoms or distant metastases. The available data do not enable pooling of data from several subgroups to compare different treatment approaches among different disease entities. However, there is increasing consensus that EBRT should be given preoperatively with concurrent chemotherapy, as demonstrated by papers published since 2006 (Table 2). This recommendation is based on the need for optimal preoperative downsizing and downstaging to maximize the chance of an R0 resection, which is the most important predictor for survival after treatment for locally recurrent rectal cancer.68,69 The calculated overall R0 resection rate of 56% leads us to conclude that there is room for improvement. The use of IORT is still controversial, and data from randomized controlled trials are lacking.
There is a need for more complete and uniform reporting on outcome parameters (Table 3). Better comparison of data can be achieved by determining outcomes for similar groups of patients (e.g., those who undergo R0 or R0/R1 resection) using standardized parameters, such as 3- and 5-year local control and overall survival. Missing follow-up data and inappropriate length of follow-up in most of the remaining studies reflect the low quality of available data on the treatment of locally recurrent rectal cancer.
The use of EBRT for primary rectal cancer has increased during the past decades. Without previous radio-therapy, patients with locally recurrent rectal cancer can be optimally treated by full-dose (chemo)radiotherapy. In the Dutch TME trial, radiotherapy at a dose of 45 Gy or higher was applied in 42% of patients for local recurrence after TME alone, whereas the rate was only 4% for the preoperative radiotherapy group.69 A radiotherapy dose less than 45 Gy was associated with shorter survival after local recurrence in both univariable and multivariable analysis. Long-term follow-up of the Swedish and Dutch rectal cancer trials showed that time from local recurrence to death was significantly shorter in the irradiation group than in patients who underwent surgery alone.59,69 Thus, radiotherapy does not affect systemic dissemination and, therefore, local recurrences after radiotherapy are more often concomitant with distant disease, leading to a worse prognosis from time of local recurrence. From these data, it can be concluded that radiotherapy for primary resectable rectal cancer mostly prevents potentially curable local recurrence. More selective application of neoadjuvant therapy for primary resectable rectal cancer can minimize early and late adverse effects, and broadens therapeutic options if local recurrence develops.
Conclusion
If intentionally curative treatment in a patient with locally recurrent rectal cancer is considered by the multidisciplinary team, a standardized approach with optimal neoadjuvant treatment is indicated. Full-dose chemoradiotherapy or an adapted schedule depending on previous EBRT maximizes the chance of an R0 resection, which is the most important prognostic factor. Surgery should be performed in specialized centres by an experienced surgeon. The distant metastasis rate during follow-up supports the use of adjuvant chemotherapy, although there is no conclusive evidence. Systemic chemotherapy preceding or following (re)irradiation in a neoadjuvant setting is probably worthwhile to explore in future studies given the systemic nature of the disease. Finally, prospective data collection should ideally be performed in the setting of a trial and/or (inter)national cancer registry. Standardized reporting of actuarial local control and survival for predefined categories (i.e., completeness of resection) based on these data will improve available evidence.
Footnotes
Competing interests: None declared.
Contributors: All authors helped design the study, analyzed data and approved publication of the article. P.J. Tanis acquired the data and wrote the article. A. Doeksen and J.J.B. van Lanschot reviewed the article.
- Accepted March 5, 2012.