Following neoadjuvant chemoradiotherapy (CRT) and interval proctectomy, 15–20 per cent of patients are found to have a pathological complete response (pCR) to combined multimodal therapy, but controversy persists about whether this yields a survival benefit. This systematic review evaluated current evidence regarding long-term oncological outcomes in patients found to have a pCR to neoadjuvant CRT.
Three major databases (PubMed, MEDLINE and the Cochrane Library) were searched. The systematic review included all original articles reporting long-term outcomes in patients with rectal cancer who had a pCR to neoadjuvant CRT, published in English, from January 1950 to March 2011.
A total of 724 studies were identified for screening. After applying inclusion and exclusion criteria, 16 studies involving 3363 patients (1263 with pCR and 2100 without) were included (mean age 60 years, 65·0 per cent men). Some 73·4 per cent had a sphincter-saving procedure. Mean follow-up was 55·5 (range 40–87) months. For patients with a pCR, the weighted mean local recurrence rate was 0·7 (range 0–2·6) per cent. Distant failure was observed in 8·7 per cent. Five-year overall and disease-free survival rates were 90·2 and 87·0 per cent respectively. Compared with non-responders, a pCR was associated with fewer local recurrences (odds ratio (OR) 0·25; P = 0·002) and less frequent distant failure (OR 0·23; P < 0·001), with a greater likelihood of being alive (OR 3·28; P = 0·001) and disease-free (OR 4·33, P < 0·001) at 5 years.
Over the past decade neoadjuvant chemoradiotherapy (CRT) followed by interval proctectomy has become the standard of care for locally advanced rectal cancer, resulting in improved local control without affecting long-term survival1–4. Many studies have shown that neoadjuvant CRT is associated with significant pathological downstaging of rectal cancers, with up to 20 per cent of patients having complete tumour sterilization (defined as absence of adenocarcinoma cells in the surgical resection specimen: pathological stage after CRT (yp) T0 N0 M0)5–8.
Currently, there is no level 1 evidence that a pathological complete response (pCR) confers a survival advantage and conflicting data exist, but are limited by small study numbers and short duration of follow-up. That a pCR represents a good prognosis for patients with rectal cancer remains to be shown definitively. Despite an apparent complete luminal and mural tumour response, up to 17 per cent of ypT0 tumours harbour disease in the mesorectal lymph nodes (ypN+)9. Similarly, approximately 8 per cent of patients with an apparent incomplete clinical response have a pCR10. The challenge remains to identify those patients with a clinical complete response (cCR) who have a true pCR; it is likely that combined radiological, biochemical and molecular biological markers will be required to predict accurately the ypN status of ypT0 tumours. For selected patients with a cCR, transanal excision of the residual scar to confirm the ypT0 status may suffice, followed by an expectant observational approach, obviating the need for radical surgery and possible permanent colostomy. Whether patients with a pCR should undergo adjuvant chemotherapy also remains the subject of debate.
The aim of this systematic review was to evaluate the oncological outcomes of patients with a pCR after neoadjuvant therapy followed by interval proctectomy. The primary outcomes of interest were rates of local recurrence, distant recurrence, 5-year overall survival and 5-year disease-free survival (DFS). These data were then compared with those of patients having an incomplete or no response to neoadjuvant therapy.
This review was conducted according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA)11.
Search strategy for identification of studies
A systematic literature search of the Cochrane Library, PubMed and MEDLINE databases was performed. The Cochrane database search was performed by combining the following search terms using the Boolean AND/OR operators: ‘rectal cancer’, ‘resection’ and ‘surgery’. For the PubMed and MEDLINE database searches, these same keywords (and variants) were used as both text words and Medical Search Headings (MeSH) terms, and were combined by using Boolean operators as follows: (‘rectal cancer*’ OR ‘rectal tumor*’ OR ‘rectal tumour*’) AND (‘chemoradiotherapy’ OR ‘neoadjuvant therapy*’ OR ‘pre-operative chemoradiotherapy’) AND (‘pathologic complete response*’ OR ‘tumour regression grade*’). The search was limited to original papers based on human studies, and there was no restriction on the date of publication.
Only articles published in the English language between January 1950 and March 2011 were included for review. Included were studies of patients with rectal tumours who underwent neoadjuvant CRT and radical rectal resection, which reported oncological outcomes and survival endpoints. Patients who had transanal excision of tumours and those who did not proceed to surgery were excluded from analysis. Only studies that included patients with pathologically verified rectal cancer were included. A minimum mean (or median) follow-up of 40 months and a minimum cohort of 40 patients was required for all included studies. Studies including recurrent lesions or lesions that had already been treated surgically were excluded. Other excluded studies were those reporting on the technique or feasibility only. Multiple publications involving the same series of patients (or duplicate patient populations) were identified and grouped together; only the most recent or parent study was included in this review to avoid double-counting of patients. Where there was uncertainty about duplicate patient groups (same group or institution reporting outcomes for a similar period, without clear indication that the smaller report was a substudy or interim results), a consensus was reached among the review authors regarding its inclusion or exclusion.
Assessment of methodological quality of studies
All eligible studies were assessed for methodological quality by two independent reviewers. The Newcastle–Ottawa Quality Assessment Scale for non-randomized cohort studies was applied by two reviewers to determine the overall quality of each eligible study. Points were awarded for patient selection (maximum 4 points), comparability of cohorts (maximum 2 points) and outcome assessment (maximum 3 points), and summed to give an overall quality rating with a maximum of 9 points12. Details of the scoring system are shown Table S1 (supporting information).
Data collection and statistical analysis
Data extracted from selected studies included: year of publication, authors' institution, number of patients and baseline characteristics, preoperative stage of disease, details of neoadjuvant and adjuvant therapy, distance of tumour from the anal verge, interval to surgery, type of surgery, pCR rate, duration of follow-up, survival statistics (5-year overall survival and DFS), local recurrence and distant recurrence rates. Statistical analyses were performed only on the extracted data from the selected studies.
Basic descriptive statistics (percentages and weighted means) were used to summarize the patient, study and outcomes data. Weighted means were calculated for oncological outcomes across all studies. Where studies reported results by sex, age or disease stage, the overall results were calculated using the proportional mix of sexes (or other variable) and individual scores. In studies with a control group (for example comparative studies comparing outcomes in those with a pCR versus patients with an incomplete or no response to neoadjuvant CRT), study characteristics and patient outcomes were obtained for all patients in the cohort.
Where possible, pooled analyses were performed using the DerSimonian–Laird random-effects model to compute odds ratios (ORs) with 95 per cent confidence intervals (c.i.). The decision to use a random-effects model was made in advance as a degree of heterogeneity between studies was anticipated. The P value for overall effect was calculated by means of the Z test and significance was set at P < 0·050. The pooled ORs and 95 per cent c.i. are presented graphically as forest plots. Each study is represented as a square on the chart area, the area of which is proportional to the weight of the study (the inverse of the study variance). The summary measure is represented by a diamond, the width of which corresponds to the 95 per cent c.i. The degree of heterogeneity is presented as the Q and I2 statistics. The I2 index of heterogeneity represents the percentage of the total variation, due to variation between studies. An I2 value of 0 per cent indicates no heterogeneity, 25 per cent low heterogeneity, 50 per cent moderate heterogeneity and 75 per cent high heterogeneity between studies13. The software package BioSTAT™ Comprehensive Meta-Analysis version 2.0 (BioSTAT, Englewood, New Jersey, USA) was used to perform the meta-analysis and generate forest plots.
Search yields and data retrieval
The initial literature search yielded 724 citations, of which 674 studies were excluded after initial screening of titles and abstracts. Of the 50 papers reviewed in full, 28 were rejected because they did not fulfil the inclusion criteria. On thorough review of the remaining 22 publications, a further eight eligible studies were identified from bibliographies and related citations. Of these 30 primary studies, 16 contained sufficient detail or met the criteria to merit inclusion in this review (Fig.1).
The data set consisted of 16 original studies5, 6, 8, 14–26. The majority were from Europe (6) or the USA (5); three were from Asia and one each from Canada and Australia. Regarding study design, the majority (14) were retrospective comparative or case studies; only two were documented as prospective cohort studies. These studies are summarized in Table1 and study quality is outlined in Table S1 (supporting information).
Table 1. Baseline characteristics of studies included in the systematic review
The 16 studies included 3363 patients who had undergone neoadjuvant CRT followed by interval proctectomy for rectal cancer. The mean age of included patients was 60 years and 65·0 per cent were men. Some 1263 had a pCR, and 2100 had an incomplete or no response. One study reported data only from 536 patients who had a pCR to neoadjuvant CRT15. Excluding this study from the pooled estimates and meta-analysis, the mean pCR rate for the entire cohort was 24·4 (range 10·3–91·3) per cent.
All patients had biopsy-proven rectal adenocarcinoma before starting therapy. The most consistent method of reporting disease stage was the American Joint Committee on Cancer tumour node metastasis (TNM) staging system. The stage was clearly documented in 11 of 16 studies; three studies reported that all patients had stage II/III disease, without subclassification. Among studies that described TNM stage, 34·1 per cent of patients had clinical stage II and 64·8 per cent had clinical stage III disease. Clinical staging was achieved by a combination of pelvic magnetic resonance imaging, endoanal ultrasonography and computed tomography. Mean distance of tumours from the anal verge, measured at proctoscopy, was 53·6 mm.
All patients received neoadjuvant CRT consisting of long-course radiotherapy (45 Gy in 7 studies and 50–50·4 Gy in 9). A variety of chemotherapeutic regimens were used to radiosensitize, with 5-fluorouracil (5-FU) being used in all 16 studies (Table1). A restorative sphincter-saving procedure was performed in 73·4 per cent of patients, 22·7 per cent had an abdominoperineal resection and 3·9 per cent an alternative procedure (low Hartmann's, pelvic exenteration, total proctocolectomy, anterior resection or transanal excision). Information on adjuvant chemotherapy was available for 13 of 16 studies; 61·4 per cent of patients in the pCR cohort received adjuvant chemotherapy.
Mean follow-up was 55·5 months and seven studies reported follow-up of 5 years or more. Oncological outcomes are summarized in Table2. Tumour recurrence in the pelvis (presacral, pelvic side wall or anastomotic) was considered a local recurrence, and diagnosed by a combination of clinical, histopathological, biochemical and radiological abnormalities. Recurrence of disease outside the pelvis was considered a distant recurrence. Outcomes of patients who had a pCR are compared with those of incomplete responders in Table S2 (supporting information). Twelve studies reported a 5-year local recurrence rate of 0 per cent among patients who had a pCR. The overall weighted mean local recurrence rate among all studies was 0·7 (range 0–2·6) per cent. Those with a pCR were almost four times less likely to develop local failure compared with incomplete responders (OR 0·25, 0·10 to 0·59; P = 0·002) (Fig.2a). Distant failure or metastasis was noted in 8·7 per cent of patients who had a pCR, at median follow-up of 55·5 months. A pCR was associated with a greater than fourfold decrease in the likelihood of developing distant failure (OR 0·23, 0·11 to 0·47; P < 0·001) (Fig.2b).
Table 2. Outcomes in 1263 patients with a pathological complete response, determined at a mean follow-up of 55·5 months
Interval to surgery (weeks)
Adjuvant chemotherapy (%)
Distant metastasis (%)
Local recurrence (%)
Overall survival (%)
Disease-free survival (%)
Three of 165 patients had metastases at surgery and did not undergo resection. Patients who had transanal excision (TAE) or no resection were excluded from the weighted means analysis. (L)AR, (low) anterior resection; APR, abdominoperineal resection; (T)PC, (total) proctocolectomy; CAA, coloanal anastomosis; IPAA, ileal pouch–anal anastomosis.
The 5-year overall survival rate for patients with a pCR was 90·2 per cent; these patients had a 3·3-fold overall survival advantage compared with incomplete responders (OR 3·28, 1·66 to 6·51; P = 0·001) (Fig.3a). A 5-year DFS rate of 87·0 per cent was noted in those with a pCR. This group was 4·3 times more likely to be disease-free at 5 years than non-responders (OR 4·33, 2·31 to 8·09; P < 0·001) (Fig.3b).
Neoadjuvant CRT produces a pCR in 15–20 per cent of patients with locally advanced rectal cancer5–8. Data from this review show that local recurrence rates (0·7 per cent), 5-year overall survival (90·2 per cent) and DFS (87·0 per cent) are comparable with those after R0 proctectomy for stage I rectal cancer27.
The effect of pCR on local recurrence after surgical excision is dramatic. With the advent of neoadjuvant CRT and optimization of surgical technique, including widespread adoption of total mesorectal excision, local recurrence rates have fallen dramatically after R0 resection of locally advanced rectal cancer1, 28–31. With surgery alone, local failure after rectal cancer surgery tends to occur before distant failure and predominantly within the first 2 years32–35. However, data exist to suggest that neoadjuvant CRT before radical surgery delays the development of pelvic recurrence36, 37. Even with close surveillance and early detection of local recurrence, many patients are not amenable to curative resection. For those who are, survival rates after radical salvage surgery remain suboptimal38–40.
Capirci and colleagues15 have published the largest experience of patients with a pCR following neoadjuvant CRT; after a median follow-up of 46 months, they reported a local failure rate of 0·9 per cent with distant failure occurring in 8·9 per cent of those undergoing radical surgery after neoadjuvant CRT. Median time to detection of pelvic recurrence was 26 months. Maas and co-workers41 recently published a comparative study addressing oncological outcomes for patients with and without a pCR to neoadjuvant CRT. The 5-year crude DFS rate in patients with a pCR was 83·3 per cent compared with 65·6 per cent for incomplete responders. Five-year local recurrence rates were 2·8 versus 9·7 per cent, and the rate of distant metastasis was 11·2 versus 25·2 per cent. These data are similar to the findings reported here. Many of the studies included in the Maas article were not included in the present review as Maas and co-workers included updated data and unpublished personal contributions from authors of published previously studies; these data were unavailable to the present authors.
The present finding of a local recurrence rate of 0·7 per cent at median follow-up of 55·5 months suggests that a pCR after neoadjuvant CRT practically eradicates the risk of local recurrence following surgery. Many data have emerged to support the contention that a pCR also improves oncological outcomes in terms of distant failure. Of the studies included in this analysis, all but one19 showed a survival advantage for patients who had a pCR compared with those who had an incomplete or no response to neoadjuvant CRT5, 6, 8, 14–18, 20–26. In an analysis of 106 patients treated with neoadjuvant CRT before radical surgery, Pucciarelli and colleagues19 did not find a survival benefit associated with a pCR. Their data suggested that clinical T category and neoadjuvant chemotherapeutic regimens were the most accurate predictors of outcome following neoadjuvant therapy. However, a small number of patients in that study did have a pCR (19, 17·9 per cent) and follow-up was relatively short at 40 months. Onaitis and colleagues42 similarly did not demonstrate a local recurrence or survival advantage for 30 patients who had a pCR, compared with 110 who had an incomplete or absent response; however, interpretation of these data is tempered by limited follow-up of 27 months. The study by Onaitis et al. was not included in the present review as it did not meet all the inclusion criteria.
Although local recurrence appears to be eradicated almost completely by a pCR, distant failure is not. In this review 8·7 per cent of patients developed metastatic disease despite achieving a pCR. This rate of distant failure, despite the apparent disappearance of the primary tumour, may be an indication that adjuvant chemotherapy is warranted in these patients. The primary role of 5-FU as part of a neoadjuvant regimen is to radiosensitize the primary tumour. In itself, neoadjuvant 5-FU provides very little chemotherapeutic benefit. Although the primary tumour may be sterilized completely following CRT, viable distant micrometastatic cells may not respond and could provoke distant failure. Thus, the challenge remains to identify patients with a pCR who can potentially benefit from adjuvant chemotherapy. Many groups are now exploring the possibility of increasing the pCR rate following neoadjuvant CRT. Administration of additional radiotherapy (‘boost dose’), concomitant non-radiosensitizing infusional chemotherapy43, chemotherapy before neoadjuvant CRT44 and use of alternative chemotherapeutic agents for radiosensitization24, 45 have all been shown to have moderate benefit.
Increasing the interval to surgery has also been postulated to increase the pCR rate by facilitating prolonged tumour necrosis. Tulchinsky and colleagues et al.46 found that increasing the interval to surgery to more than 7 weeks improved the pCR rate to 35 per cent, compared with 17 per cent in those undergoing surgery within 7 weeks; interval to surgery was prognostic of overall survival and DFS on multivariable analysis. Similarly, Kalady and co-workers47 found that increasing the interval to more than 8 weeks was associated with a higher rate of pCR, and improved local control and overall survival. A National Institutes of Health study is currently investigating the optimal delay to surgery after neoadjuvant CRT, with intervals ranging from 6 to 24 weeks7.
The major implication of a pCR to neoadjuvant CRT, in terms of oncological outcomes, is that the patient may become eligible for less radical surgical resection or an expectant ‘watch and wait’ approach. In selected patients who appear to have a complete luminal/mural response, many observers recommend transanal excision of the tumour scar to ensure ypT0 status with observation of the mesorectal lymph nodes48–54. Habr-Gama and colleagues10 advocated an expectant policy for patients with a cCR after neoadjuvant CRT. In a series of 265 patients, 71 (26·8 per cent) had a cCR; of these, only 2·8 per cent had an endoluminal recurrence, which was treated successfully with transanal excision or brachytherapy. No pelvic recurrences were observed, but 4·2 per cent developed distant metastases. Ten-year overall survival and DFS rates in patients who achieved a cCR were 100 and 86 per cent10. However, patients displaying an apparent cCR after neoadjuvant CRT may still have disease in the mesorectal lymph nodes. It has been reported that 2–27 per cent of ypT0 tumours are ypN+9, 55–59.
The ability to identify patients with a cCR who are also likely to have a pCR would have major clinical implications. If such information were available and accurate, it could obviate the need for radical surgery and possibly a permanent stoma in selected patients. It is hoped that advances in radiological imaging, molecular biology and biochemical markers may facilitate this. Outcomes following salvage surgery for local endoluminal recurrence of disease are comparable with those of primary surgery after neoadjuvant CRT. However, when an expectant observational approach fails, survival is significantly compromised by systemic failure60.
This review has several limitations. There is concern regarding outcome reporting bias given that not every study reported all four oncological outcomes (local recurrence, distant recurrence, overall survival and DFS). Similarly, with regard to publication bias, a sensitivity analysis suggested that there might be three missing studies reporting data on overall survival and DFS, but imputing data from these studies would not have altered the point estimates significantly (according to a ‘trim and fill’ analysis). All patients received long-course radiotherapy and, although a variety of radiosensitizing chemotherapeutic regimens were employed, 5-FU was used in all patients. Data on adjuvant chemotherapy were incomplete; however, in the 12 studies reporting its use, 61·4 per cent of patients received adjuvant therapy. It would be of interest to know what proportion of patients experiencing local and distant failure received adjuvant chemotherapy, but these data were not reported.
The challenge remains to identify those with a cCR and a ‘true pCR’ who may be spared radical surgery and, similarly, patients with a pCR who may benefit from adjuvant chemotherapy to minimize the risk of subsequent local or distant failure. It is concluded that a pCR is associated with excellent long-term oncological outcomes, almost eradicating local failure and achieving low rates of distant failure with high rates of overall survival and DFS.
Additional supporting information may be found in the online version of this article:
Table S1 Newcastle–Ottawa Quality Assessment Scale score for studies included in the review (Word document)
Table S2 Oncological outcomes of patients who had a pathological complete response to neoadjuvant chemoradiotherapy versus those who did not (mean follow-up 55·5 months) (Word document)
Please note: John Wiley & Sons Ltd is not responsible for the functionality of any supporting materials supplied by the authors. Any queries (other than missing material) should be directed to the corresponding author for the article.
The authors thank Jeff Hammel for statistical guidance in conducting the meta-analysis.
Disclosure: The authors declare no conflict of interest.