Scolaris Content Display Scolaris Content Display

Cochrane Database of Systematic Reviews Protocol - Intervention

Local versus radical surgery for early rectal cancer with or without neoadjuvant or adjuvant therapy

Authors' declarations of interest

Version published: 11 May 2011 Version history

https://doi.org/10.1002/14651858.CD002198.pub2

This is not the most recent version

view the current version 13 June 2023
Collapse all Expand all

Abstract

Objectives

This is a protocol for a Cochrane Review (intervention). The objectives are as follows:

To assess the oncologic outcomes and operative harms of modern endoscopic local surgical excision compared to radical surgical resection in people with stage I rectal cancer.

Background

Description of the condition

Colorectal cancer is a public health concern worldwide. It is the third most frequent cancer overall with more than 1.8 million new cases globally each year (Vogel 2017; GLOBOCAN 2018). It is the second leading cause of cancer deaths and caused 880,792 deaths in 2018 (GLOBOCAN 2018). The overall incidence of and mortality rate for colorectal cancer have declined in recent decades. However, for people under 55, the incidence is increasing at an annual rate of 2.4% and the mortality rate at 1% (American Cancer Society 2018). Rectal cancer accounts for 35% of colorectal cancer cases and it is expected to increase in both genders (Glynne‐Jones 2017).

These concerning figures have prompted major advances in screening programs and diagnostics. As a result, approximately one third of rectal cancer patients present with early and localised stage I cancer, i.e. a tumour (T) confined to superficial layers (submucosa (T1) or muscularis propia (T2)) of the rectal wall without any evidence of local lymph node spread (N0) or distant metastasis (M0)(AJCC Cancer Staging Manual 2017). Stage I rectal cancer thus makes up a significant percentage of rectal cancers and has a 5‐year survival rate of about 90% (American Cancer Society 2014; National Cancer Institute 2017).

This favourable survival statistic, however, is derived from populations undergoing standard 'radical resection' of rectal cancer, in which the tumour and regional lymph nodes are removed. Since stage I rectal cancer is generally characterised by the absence of any local spread, this radical approach may be an over‐treatment for this subset of patients. However, Mou 2013 and Saraste 2013 report that even clinical T1‐2N0M0 rectal cancers can have undetected lymph node involvement rates of 6% to as high as 65%, and 11% to 78%, for T1 and T2 tumours respectively, depending on the presence of high‐risk pathologic features in the tumour. Radical resection removes rectal lymph nodes and minimises the risk of leaving any residual cancer (Chen 2006).

Description of the intervention

Radical resection of the rectum is performed according to principles of total mesorectal excision (TME), by which the tumour and all lymph nodes around the rectum are precisely removed as a whole within a mesorectal fascial envelope (Heald 1982). Principles of TME are strictly followed in all radical surgery methods including Hartmann’s resection, low anterior resection, or abdominoperineal resection, which may be performed through the abdomen using open, laparoscopic, hand‐assisted laparoscopic, or robotic techniques. More recently, TME has also been performed by a two‐phase transabdominal‐transanal approach, transanal TME (TaTME) (de Lacy 2018; Ma 2016; Sylla 2010).

Radical surgery emerged as the standard of care since it significantly reduces local recurrence and improves the oncologic results for rectal cancer patients (Heald 1982). It is, however, associated with considerable morbidity, not only from the operation itself, but also from the resulting functional impairment. Transabdominal rectal surgery involves incising through the abdominal wall necessitating hospital admission for pain control, intravenous fluids and recovery. It also requires an anastomosis that is challenging to perform transabdominally, especially for low‐lying rectal tumours, and carries a risk of leakage and sepsis. Furthermore, resection of the rectum and consequent loss of rectal reservoir function may result in abnormal bowel habits (low anterior resection syndrome) and faecal incontinence. Lastly, injury to autonomic nerves during pelvic surgery may cause urinary incontinence and sexual dysfunction in a significant number of patients (Andersson 2014; Hendren 2005). While utilisation of radical surgery is justified in patients with invasive rectal cancer with high probability of local spread, its use in early rectal cancers that may not require an extensive lymph node removal is questionable due to the high associated postoperative morbidity.

In contrast to radical surgery, local excision avoids an abdominal incision, risk of an anastomotic leak and functional problems of a pelvic resection, by precisely targeting and removing only the tumour through the anus. This less invasive approach is thus associated with fewer postoperative complications compared to radical surgery: 5.6% versus 14.6%, respectively (You 2007). However, traditional 'open' transanal excision was associated with an unacceptably high local recurrence rate of up to 30% (Parks 1968), compared to a low recurrence rate of 2% to 4% with radical surgery (Garcia‐Aguilar 2000; Madbouly 2005; Mellgren 2000; Ptok 2007). Improved visualisation could be achieved using transsphincteric and transsacral approaches but these were associated with high rates of fistula formation (Harvey 2004). Technological advances have offered improved visualisation for transanal local excision through insufflating the rectum and magnifying the lesion. The first technical advance was the introduction of transanal endoscopic microsurgery (TEM). Using instruments manufactured by Wolf, Professor Buess in Germany pioneered endoscopic transanal removal of rectal lesions (Buess 1983). This advance incorporates a rigid proctoscope with optics enabling clear visualisation and magnification of the lesion with three‐dimensional depth perception. At present, several other platforms also provide equipment for transanal excision using the same concept, including transanal endoscopic operation (TEO), transanal minimally invasive surgery (TAMIS), and transanal single‐port microsurgery (TSPM). TEM and TEO use rigid proctoscopes, while TAMIS and TSPM use flexible anal ports with laparoscopic instruments (Atallah 2010; Khoo 2010; Lorenz 2010). While these transanal procedures cannot achieve total mesorectal excision from a technical standpoint, their improved visualisation and higher accuracy decrease positive resection margins and hence, local recurrence rates, approaching radical total mesorectal excision surgery results (Bach 2009; Hompes 2011; Serra‐Aracil 2008).

How the intervention might work

Since local excision does not treat potential metastases to regional lymph nodes, a chance of local spread remains. Neoadjuvant or adjuvant chemoradiotherapy may be used to treat regional spread and improve local control and survival after local excisions (Doornebosch 2009; Lezoche 2008; Lezoche 2012; Morino 2011; Nair 2008; Sasaki 2017; Winde 1996). Chemoradiotherapy may be rationally applied for T1 and T2 cancers with higher risk pathological features including poor differentiation, lymphovascular invasion and tumour budding (Nash 2009).

Why it is important to do this review

While a number of studies and reviews have sporadically evaluated some of the surgical techniques for early rectal cancer (e.g. Shaikh 2015), we aim to summarise systematically the evidence on safety and efficacy of the radical versus local approach. Given that local excision is increasingly being adopted worldwide, it should be evaluated accurately and regularly using the high‐quality Cochrane methodology to ensure it is at least of equal oncological effectiveness to radical resection, in addition to its major advantage of being less invasive. With the addition of neoadjuvant or adjuvant chemoradiotherapy to treatment protocols, it is also important to determine whether chemoradiotherapy enables the use of local excision techniques for higher‐grade tumours (T2) or those with high‐risk features. Equal efficacy of local excision and radical resection would potentially translate into safer surgery, fewer sphincter and genitourinary complications, and better quality of life; all important outcomes from the patients’ perspective (Wrenn 2018). Given the rapid advances in technology and surgical techniques, it is necessary and timely to assess and summarise the short‐ and long‐term outcomes from recent trials to support the continued use of local excision as an equivalent or superior option to radical surgery for patients with early rectal cancer.

Objectives

To assess the oncologic outcomes and operative harms of modern endoscopic local surgical excision compared to radical surgical resection in people with stage I rectal cancer.

Methods

Criteria for considering studies for this review

Types of studies

We will include randomised controlled trials (RCTs) that compare local excision and radical resection for early rectal cancer. Published and unpublished studies, including non‐English language studies, are eligible for inclusion.

We will exclude cluster‐RCTs, which are an unsuitable and unlikely design for this type of intervention, and studies that do not report separate data for benign and malignant lesions. We will consider for inclusion studies that include only a subset of relevant participants based on the availability of separate data for those participants in the article or upon request from the authors.

Types of participants

Participants aged 18 years or older who had radical resection or local excision for rectal cancer ‐ defined as a tumour with distal extension less than or equal to 15 cm from the anal verge (measured by sigmoidoscopy) – at an early stage will be included:

  1. Stage I (T1‐2N0M0, TNM staging system) (AJCC Cancer Staging Manual 2017), or

  2. Stage A (Dukes staging system) (Astler 1954)

We will exclude participants with in situ tumours (TisN0M0, stage 0) rectal cancer, as well as participants with locally advanced cancer (stage II and higher), synchronous lesions, recurrent cancer, or tumour types other than adenocarcinoma.

Types of interventions

Intervention arm: local excision

Local excision implies full‐thickness excision of the lesion with free margins and with curative intent using any of the aforementioned local excision techniques: TEM, TEO, TAMIS or TSPM.

Descriptions of local excision techniques

  1. TEM (transanal endoscopic microsurgery): surgical removal of the tumour through a 4cm rigid rectoscope by establishing pneumorectum and using an oblique‐angled stereoscopic endoscope with modified surgical instruments (Buess 1983).

  2. TEO (transanal endoscopic operation): surgical removal of the tumour using a similar platform to TEM but with regular laparoscopic instruments (insufflator, graspers, thermal energy devices and needle drivers) through the rigid rectoscope.

  3. TAMIS (transanal minimally invasive surgery) and TSPM (transanal single port microsurgery): removal of the tumour through a single‐incision laparoscopic surgery port (SILS port, or GelPOINT path transanal access platform) to establish pneumorectum, by using ordinary laparoscopic instruments, including graspers, thermal energy devices and needle drivers (Atallah 2010; Khoo 2010; Lorenz 2010).

Control arm: radical resection

Radical resection techniques serve as the standard control interventions. Procedures such as Hartmann’s resection, low anterior resection, and abdominoperineal resection are included. These procedures accomplish complete resection of the rectum with regional lymph nodes within the mesorectal fascia (i.e. TME) using the open, laparoscopic, robotic or TaTME approach.

Descriptions of radical operations to achieve TME

  1. Hartmann's resection (proctosigmoidectomy): surgical resection of the rectosigmoid colon with closure of the anorectal stump and formation of an end‐colostomy temporarily or permanently.

  2. Low anterior resection: surgical removal of the rectum harbouring the tumour and subsequent anastomosis of the colon to the remaining rectum.

  3. Abdominoperineal resection: surgical removal of a low‐lying rectal tumour involving the anal sphincter with creation of a permanent colostomy.

  4. TaTME (transanal total mesorectal excision): surgical removal of a rectal tumour through a combined transabdominal‐transanal minimally invasive approach, where the proximal dissection is performed through the abdomen laparoscopically and distal dissection through a transanal approach.

Comparisons

The use of chemoradiotherapy will be permitted in one or both arms of the study, i.e. not required to be applied equally. This means that three comparisons need to be made:

  1. local excision versus radical resection;

  2. chemoradiotherapy plus local excision versus radical resection; and

  3. chemoradiotherapy plus local excision versus chemoradiotherapy plus radical resection.

Types of outcome measures

The following outcomes of interest will be sought.

Primary outcomes

  1. Disease‐free survival (D‐FS), defined as the chance that a patient is without local recurrence or metastasis at a given time point

  2. Sphincter function, as measured by standardised scales such as the Wexner scale (Jorge 1993) and low anterior resection syndrome (LARS) score (Emmertsen 2012)

Secondary outcomes
1. Oncologic outcomes

  1. Cancer‐related survival (C‐RS), defined as the chance that a patient has not died of reasons related to cancer at a given time point

  2. Local recurrence‐free survival (LR‐FS), defined as the chance that a patient is without local recurrence of cancer at a given time point

  3. Metastasis‐free survival (M‐FS), defined as the chance that a patient is without metastasis of cancer at a given time point

2. Surgical outcomes

  1. 30‐day postoperative mortality

  2. Major and minor surgical complications (using the Clavien‐Dindo classification (Dindo 2004), grades I and II as minor, and grades III or higher as major)

  3. Length of hospital stay (LoS, days)

  4. Conversion rate

3. Functional outcomes

  1. Quality of life, as measured by standardised scales such as the Short Form Health Survey (SF‐36, Ware 1992) and Euro Quality of Life‐5D scales (EQ‐5D, The EuroQol Group 1990)

  2. Genitourinary function, as measured by standardised scales such as Quality of Life Questionnaire CR‐29 (QLQ CR29, Gujral 2007))

Search methods for identification of studies

Electronic searches

We will search the following electronic databases for primary studies from 1983 onwards, without any restrictions as to language or publication status.

  1. The Cochrane Central Register of Controlled Trials (CENTRAL) (Appendix 1)

  2. Ovid MEDLINE (Appendix 2)

  3. Ovid Embase (Appendix 3)

  4. Science Citation Index Expanded via the Web of Science (Appendix 4)

Searching other resources

Grey literature

We will search the grey literature database, OpenGrey (www.opengrey.eu), ProQuest dissertations and theses database (www.proquest.com), and ISI Conference abstracts and proceedings database (www.proceedings.com).

We have compiled a list of the most relevant conferences to include in the grey literature search by seeking the expert opinion of a group of colorectal surgeons. We have selected the following peer‐reviewed conferences based on impact within the rectal cancer literature (from 1983 to present): European Colorectal Congress; American Society of Colon and Rectal Surgeons Annual Meeting; International Society of Laparoscopic Colorectal Surgeons Congress; and Congress of Asia Pacific Federation of Coloproctology.

We will also search for relevant documents from key organisations such as the American Society of Clinical Oncology (ASCO), the American Society of Colon and Rectal Surgeons (ASCRS), the European Society of Coloproctology (ESCP), and European Society for Medical Oncology (ESMO) from 1983 onwards.

Handsearching

We will handsearch reference lists of all retrieved and relevant publications identified through the above strategies. We will identify reviews, meta‐analyses, and guidelines on the topic of our review and look for further relevant studies.

Unpublished and ongoing studies

We will search the following trial registries for unpublished and ongoing studies, and contact authors to ask for their data, if relevant (search strategies for ongoing trials are presented in Appendix 5).

  1. ClinicalTrials.gov by United States National Library of Medicine (www.clinicaltrials.gov)

  2. International Standard Randomised Controlled Trial Number Registry (www.isrctn.com)

  3. The International Clinical Trials Registry Platform by World Health Organization (www.apps.who.int/trialsearch)

  4. National Cancer Institute (NCI) Clinical Trials Registry (www.cancer.gov/clinicaltrials)

We will update search results no later than three months prior to submission of the review. We will evaluate and flag relevant new articles as 'awaiting assessment'.

Data collection and analysis

Selection of studies

Two review authors (MAM and NM) will independently screen the title and abstracts of studies identified from the electronic, grey literature and hand searches, and determine whether a study is eligible for inclusion (see Criteria for considering studies for this review). We will provide a study flow diagram to document the process of including or excluding studies (PRISMA flow chart,Moher 2009). We will retrieve the full texts of potentially eligible references for definitive assessment of eligibility. At this stage, we will only exclude those studies classified by both review authors as ‘exclude’. Disagreements will be resolved through discussion with a third review author (TP). We will try to obtain further information about any trial published only as an abstract. If a full report is not available or there is insufficient information to determine the eligibility of a study, we will label them as 'awaiting assessment' and contact the study authors. We will summarise in a table all the relevant studies labelled as ‘excluded’ after the assessment of the full text with the main reason for their exclusion. We will perform the review process, including citation screening, study selection, risk of bias assessment and data extraction, using the Covidence software (Covidence 2018).

Data extraction and management

Two reviewers (MAM and NM) will independently extract data from included trials, using a piloted data extraction form based on the CONSORT statement for non‐pharmacological interventions for RCTs (Boutron 2008). For each included study, we will attempt to obtain the protocol of the study and extract the following characteristics to present in the 'Characteristics of included studies' table.

  1. Study title, authors, accrual dates, year of publication, journal citation

  2. Funding sources and conflicts of interest

  3. Study setting, design, whether single‐ or multi‐centre (and number of centres)

  4. Inclusion and exclusion criteria

  5. Number of participants

  6. Follow‐up duration

  7. Stage of rectal cancer and tumour characteristics

  8. Age, sex, and relevant baseline characteristics of participants (e.g. comorbid conditions)

  9. Number of participants randomised and assessed for each outcome

  10. Intervention and control

  11. Co‐interventions (e.g. neoadjuvant chemoradiotherapy, adjuvant chemoradiotherapy)

  12. Number of exclusions, withdrawals and missing data

  13. Outcome measures reported

  14. Treatment protocol and quality of intervention (Herbert 2005)

  15. Risk of bias domains

We will report the data according to the intention‐to‐treat principle. The review authors will compare results and resolve any disagreements by discussion until a consensus is reached. One review author (MAM) will populate the data automatically into Review Manager 5.3 software using the Covidence software, and a second author (NM) will check the data entry.

For outcomes relating to surgical complications in studies before 2004 or for those not using the Clavien‐Dindo classification, if re‐categorisation and pooling such data with Clavien‐Dindo grades is reasonable, we will attempt to do so. Otherwise, we will report them separately.

If relevant data are not directly reported (for example, hazard ratio (HR) and its 95% confidence interval (CI) for time‐to‐event data, or standard deviation (SD) for continuous data), we will attempt to extract the data required for their estimation using methods developed by Parmar and colleagues (Higgins 2011; Parmar 1998; Tierney 2007). For any studies not in English, Spanish or French (languages in which at least one of the authors is fluent), official translation by a professional translation service will be organised. 

Dealing with duplicate publications

We will exclude duplicate publications and report these in the 'Characteristics of excluded studies’ table. When we identify multiple reports of a particular study, the study will be the unit of interest for this review. We will give priority to the report with the most complete data and longest follow‐up, but we will examine all reports and collate them in order to ensure maximum data are extracted from the study. We will indicate the primary source for each study.

Assessment of risk of bias in included studies

Two reviewers (MAM and NM) will independently assess and judge the risk of bias across the following domains based on criteria used in the Cochrane 'Risk of bias tool' (Chapter 8.5.d, Higgins 2017, Appendix 6, Figure 3).

  1. Random sequence generation: was the allocation sequence adequately generated?

  2. Allocation concealment: was allocation adequately concealed?

  3. Blinding (masking) of participants, personnel, and outcome assessors: was knowledge of the allocated interventions adequately prevented during the study?

  4. Incomplete outcome data: were incomplete outcome data adequately addressed?

  5. Selective outcome reporting: are reports of the study free of suggestion of selective outcome reporting?

  6. Other problems in protocol execution or design that may affect individual outcomes reported such as:

    • Were the groups similar at baseline regarding the most important prognostic indicators?

    • Did both treatment and control arms receive treatment consistent with the standard of care for disease stage?

    • Were the co‐interventions (other than chemoradiotherapy) within treatment arms avoided or similar?

    • Was the study apparently free of early stopping?

    • Was the study apparently free of academic bias?

    • Was the study apparently free of any source of funding bias?

Each domain will be labelled as having a low, high, or unclear risk of bias. We will resolve disagreements through discussion to reach consensus and, if necessary, with the involvement of a third author (TP). We will consider domains 1, 2, 4, and 5 as 'key domains' when assessing the risk of bias for each study. For individual outcomes, other domains will also be used for assessing the risk of bias (Appendix 7). We will summarise and present the overall risk of bias for each outcome considering the assessments for its key domains in two different manners (Higgins 2017)(Appendix 8):

  • Within each study across domains: each outcome will be defined as having a ‘low risk of bias’ only if it meets all the key domains; as ‘high risk of bias’ if it demonstrates high risk of bias for one or more of them; or an ‘unclear risk of bias’ if it demonstrates unclear risk of bias for at least one key domain without any of them described as ‘high risk of bias’.

  • Across studies: each outcome will be defined as having a ‘low risk of bias’ if most information is from studies at low risk of bias; as ‘high risk of bias’ if the proportion of information from studies at high risk of bias is sufficient to affect the interpretation of the results; or an ‘unclear risk of bias’ if most information is from studies at low or unclear risk of bias.

The authors acknowledge that certain outcomes are subjective and thus affected by lack of masking in a study (self‐reported or investigator‐reported outcomes, e.g. outcome 1.2 sphincter function). We will describe and tabulate subjective outcomes that are inconsistently reported and thus, excluded from statistical analysis. On the other hand, objective outcomes are less affected by lack of blinding and we will extract all data relating to these outcomes. However, we will flag for high risk of bias those studies not providing quantitative and validated measures of objective outcomes.

Measures of treatment effect

Measure of treatment effect depends on the types of data presented in individual studies.

  • For time‐to‐event data (e.g. D‐FS), we will use the hazard ratio (HR) and its 95% confidence interval (CI).

  • For dichotomous data (e.g. major postoperative complications), we will extract the number of participants in each treatment arm who experienced the outcome of interest and the number of participants assessed, in order to estimate the relative risk (RR) and its 95% CI.

  • For continuous data (e.g. LoS) we will calculate the mean difference (MD) with standard deviations (SD). If measures are reported using different scales (e.g. QoL), we will use the standardised mean difference (SMD). When medians and inter‐quartile range are the only data provided, the median will be used as a proxy measure of the mean and the difference between the first and third inter‐quartile as equivalent to 1.35 times SD.

We will use the Peto odds ratio (Peto OR) method where the event rate is less than 10% (Deeks 2017). We will report study results with their associated 95% confidence intervals (CI). In case of reporting P values for outcomes, the statistical significance will be considered at 0.05.

Unit of analysis issues

The unit of analysis in this review is the individual patient. We will not include cluster‐randomised trials or other forms of trial design that may introduce unit of analysis issues.

Dealing with missing data

In the case of missing data, we will attempt to contact the study authors to obtain the data, and we will document our attempts and their replies in the 'Characteristics of included studies' table. Where we are unsuccessful, we will analyse the data as reported and address the potential impact of the missing outcomes on the results of our review by performing a sensitivity analysis (best case/worst case analysis). We will also document the number of participants randomised and analysed to clarify the possibility of attrition bias.

In the surgical treatment of participants with early rectal cancer, exclusion of participants after randomisation is sometimes justifiable. For example, some participants admitted to the trial may have been found intra‐operatively to have metastatic disease, revealing the fact that the participant was not eligible for the trial (Fergusson 2002; Higgins 2011). We will consider this type of post‐randomisation exclusion as appropriate.

Assessment of heterogeneity

Where studies are considered similar enough (based on consideration of participants and interventions) to allow pooling of data using meta‐analysis, we will assess the degree of heterogeneity by visual inspection of forest plots and by examining the Chi2 test for heterogeneity. We will report our reasons for deciding that studies were similar enough to pool statistically. Heterogeneity will be quantified using the I2 statistic. An I2 value of 50% or more will be considered to represent substantial levels of heterogeneity (Deeks 2017), but this value will be interpreted in light of the size and direction of effects and the strength of the evidence for heterogeneity, based on the P value from the Chi2 test (Deeks 2017). Where heterogeneity is present in pooled effect estimates, we will explore possible reasons for variability by conducting subgroup analyses to examine if the population, interventions and outcomes (or the way in which they are measured) differ substantially. If few trials are included in the meta‐analysis, the Chi2 test has little power to detect heterogeneity. Therefore we will not necessarily interpret a non‐significant result as evidence of no heterogeneity but instead would interpret the findings with care.

Where we detect substantial clinical, methodological or statistical heterogeneity across included studies, we will not report pooled results from meta‐analysis but will instead use a narrative approach to data synthesis. In this event, we will clearly report our reasons for deciding that studies were too dissimilar to meta‐analyse. We will also attempt to explore possible clinical or methodological reasons for this variation by grouping studies that are similar in terms of surgical methods, co‐interventions, or other possible factors to explore differences in intervention effects.

Assessment of reporting biases

We will attempt to minimise reporting bias by undertaking an extensive and exhaustive search process and by including both published and unpublished studies, and critically appraising the latter to assure they are of sufficient quality. We will assess potential publication bias for the primary outcomes with the use of funnel plots if we identify more than 10 included studies, as recommended in chapter 10 in the Cochrane Handbook for Systematic Reviews of Interventions (Page 2020). Nevertheless, because there are several explanations for funnel plot asymmetry when assessing for small study effect, we will interpret the findings cautiously.

Data synthesis

We will perform the analyses using Review Manager 5.3 provided by the Cochrane Collaboration (RevMan 2014), and using Chapter 9 of the Cochrane Handbook for Systematic Reviews of Interventions as a guide (Deeks 2017). We will combine the outcome measures from individual trials in a meta‐analysis to calculate a pooled effect estimate for each outcome only if the study participants, interventions and outcomes are clinically and methodologically comparable to provide a meaningful summary. Due to the anticipated variability in the interventions of included studies, we will use a random‐effects model for meta‐analysis (Deeks 2017; Mantel 1959). We will exclude all studies rated at a high risk of bias on sequence generation from the analysis, and we will conduct a sensitivity analysis to investigate the effects of this decision. If we are unable to pool data and perform meta‐analysis for an outcome, we will report the reasons for not doing so and will instead seek to interpret the data qualitatively for the outcome in question across studies. We will present a narrative summary of the outcome in tables. We will acknowledge and take into account the heterogeneity of the studies in our interpretations and results.

Subgroup analysis and investigation of heterogeneity

We will undertake the following subgroup analyses.

  • Patients with T1 versus T2 tumours

  • Receiving neoadjuvant versus adjuvant chemoradiotherapy (Akgun 2017)

We will also try to conduct subgroup analysis based on specific type of chemoradiotherapy, if possible. Any subgroup analysis will be followed by formal statistical testing and reported.

Sensitivity analysis

We will undertake sensitivity analyses for the risk of bias by: excluding studies deemed at high risk of bias for the key domains; changing from random‐effects to fixed‐effect models for meta‐analysis (Mantel 1959); excluding small or large studies from the analysis; and excluding studies with high rates of missing data.

Trial sequential analysis

We will perform a Trial Sequential Analysis (TSA) using TSA software (Thorlund 2011), to calculate the required information size, and to determine whether the cumulative Z‐curve of the trial sequential analysis boundaries for benefit, harm or futility were crossed for the main outcomes in the summary of findings table.

Summary of findings

We will apply the Grades of Recommendation, Assessment, Development, and Evaluation developed by the GRADE Working Group for grading the quality of evidence (GRADE Working Group 2004). The GRADE approach specifies four levels of quality (high, moderate, low and very low), which can be downgraded by one (serious concern) or two levels (very serious concern) for the following reasons: study limitations (risk of bias); inconsistency (unexplained heterogeneity, inconsistency of results); indirectness (indirect population, intervention, control, outcomes); imprecision (wide confidence intervals); or potential publication bias (Schünemann 2011). If meta‐analysis is not possible, we will present results in a narrative ‘Summary of findings’ table format.

We will use GRADEprofiler software (GRADEpro GDT) to construct our 'Summary of findings' (SoF) table. A second author (NM) will confirm the result and a third author (TP) will settle any disputes.

We will include the following outcomes in our SoF table.

  1. D‐FS

  2. Sphincter function

  3. CR‐S

  4. LR‐FS

  5. Minor postoperative complications (grade I‐II surgical complications) 

  6. Major postoperative complications (grade III‐IV surgical complications)

  7. Quality of life

We will report all outcomes in an additional, complete SoF table.

Summary of findings and assessment of the certainty of the evidence

We applied the Grades of Recommendation, Assessment, Development, and Evaluation developed by the GRADE Working Group for grading the quality of evidence (GRADE Working Group 2004). The GRADE approach specifies four levels of quality (high, moderate, low and very low), which can be downgraded by one (serious concern) or two levels (very serious concern) for the following reasons: study limitations (risk of bias), inconsistency (unexplained heterogeneity, inconsistency of results), indirectness (indirect population, intervention, control, outcomes), and imprecision (wide confidence intervals) or potential publication bias (Schünemann 2011).

We used the GRADEprofiler software (GRADEpro GDT) to construct our summary of findings (SoF) table and grade the quality of evidence. Results were confirmed by involving the second and adjudicating reviewers.

We predetermined the following outcomes for our main SoF table (Summary of findings table 1): D‐FS, sphincter function, C‐RS, LR‐FS, major postoperative complications, minor postoperative complications, and quality of life. A SoF table with all outcomes is also reported.