Types of studies

We included all randomized controlled trials (RCTs), regardless of sample size, language, or publication status, which compared (1) drain use versus no drain use, (2) different types of drains, or (3) different schedules for drain removal in people undergoing pancreatic surgery. We excluded quasi‐randomized studies, in which the allocation was performed on the basis of a pseudo‐random sequence (e.g. odd or even hospital number or date of birth, alternation, and non‐randomized studies) because of the potential for bias (Reeves 2011).

Types of participants

We included people, regardless of age, sex, or race, who underwent elective pancreatic resections (open or laparoscopic) for any pancreatic or extra‐pancreatic disease.

Types of interventions

1. Drain use versus no drain use.

2. One type of drain versus another.

3. Early versus late drain removal (no more than four days versus more than four days).

Types of outcome measures

Electronic searches

For the last version of this review, we searched the Cochrane Central Register of Controlled Trials (CENTRAL; 2016, Issue 8) in the Cochrane Library, MEDLINE (1946 to 28 August 2016), Embase (1980 to 28 August 2016), Science Citation Index Expanded (1900 to 28 August 2016), and Chinese Biomedical Literature Database (CBM; 1978 to 28 August 2016; Cheng 2016b; Peng 2015). For this updated review, we searched the following electronic databases from 2016 to 15 November 2017, with no language or date of publication restrictions:

1. the Cochrane Central Register of Controlled Trials (CENTRAL; 2017, Issue 11) in the Cochrane Library (searched 15 November 2017; Appendix 2);

2. MEDLINE Ovid (2016 to 15 November 2017; Appendix 3);

3. Embase Ovid (2016 to 15 November 2017; Appendix 4);

4. Science Citation Index Expanded (Web of Science; 2016 to 15 November 2017; Appendix 5); and 

5. Chinese Biomedical Literature Database (CBM; 2016 to 15 November 2017; Appendix 6).

Searching other resources

We checked reference lists of all primary studies and relevant review articles that were identified during the search for RCTs for additional references. We contacted authors of identified studies and asked them to identify other published and unpublished studies.

We searched PubMed for errata or retractions of eligible studies, and reported the date this was done in the results section (www.ncbi.nlm.nih.gov/pubmed). We also searched the meeting abstracts via the Society of American Gastrointestinal and Endoscopic Surgeons (SAGES; www.sages.org/; accessed 15 November 2017) and Conference Proceedings Citation Index to explore further relevant clinical studies.

Selection of studies

Two review authors (WZ, SH) independently screened the titles and abstracts of all the reports we identified as a result of the search, and coded them as 'retrieve' (eligible, potentially eligible, or unclear) or 'do not retrieve'. We retrieved the full‐text study reports and publications of identified reports, and two review authors (WZ, SH) independently screened the full text, identified studies for inclusion, and identified and recorded reasons for exclusion of the ineligible studies. We resolved any disagreements through discussion, or if required, we consulted a third review author (ZL). We identified and excluded duplicates and collated multiple reports of the same study, so that each study, rather than each report, was the unit of interest in the review. We recorded the selection process in sufficient detail to complete a PRISMA flow diagram and a 'Characteristics of excluded studies' table (Moher 2009).

Data extraction and management

We used a standard data collection form for study characteristics and outcome data, which had been piloted on at least one study in the review. Two review authors (JX, ML) extracted the following study characteristics from included studies:

1. methods: study design, total duration of study and run in, number of study centers and location, study setting, withdrawals, and date of study;

2. participants: number (N), mean age, age range, gender, severity of condition, diagnostic criteria, inclusion criteria, and exclusion criteria;

3. interventions: intervention, comparison;

4. outcomes: primary and secondary outcomes specified and collected, time points reported;

5. notes: funding for trial, notable conflicts of interest of trial authors.

Two review authors (JX, ML) independently extracted outcome data from included studies. We noted in the 'Characteristics of included studies' table if outcome data were not reported in a usable way. We resolved disagreements by consensus, or by involving a third review author (ZL). One review author (SH) copied the data from the data collection form into Review Manager 5 (RevMan 2014). We double‐checked that the data were entered correctly by comparing the study reports with how the data were presented in the systematic review. A second review author spot‐checked study characteristics for accuracy against the trial report.

Assessment of risk of bias in included studies

Two review authors (WZ, SH) independently assessed risk of bias for each study, using the criteria outlined in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011). We resolved any disagreements by discussion, or by involving a third review author (ZL). We assessed the risk of bias according to the following domains:

1. random sequence generation;

2. allocation concealment;

3. blinding of outcome assessment;

4. incomplete outcome data;

5. selective outcome reporting;

6. other bias.

We graded each potential source of bias as high, low, or unclear risk, and provided a quote from the study report together with a justification for our judgement in the 'Risk of bias' table. We summarized the 'Risk of bias' judgements across different studies for each of the domains listed. We considered blinding separately for different key outcomes where necessary (e.g. for unblinded outcome assessment, risk of bias for all‐cause mortality may be different than for a participant‐reported pain scale). Where information on risk of bias related to unpublished data or correspondence with a trialist, we noted this in the 'Risk of bias' table.

When considering treatment effects, we took into account the risk of bias for the studies that contributed to that outcome.

Measures of treatment effect

We analyzed dichotomous data as risk ratio (RR) and continuous data as mean difference (MD) or standardized mean difference (SMD) with 95% confidence intervals (CI). We ensured that higher scores for continuous outcomes had the same meaning for the particular outcome, explained the direction to the reader, and reported where the directions were reversed, if this was necessary.

We undertook meta‐analyses only where this was meaningful, that is, if the treatments, participants, and underlying clinical question were similar enough for pooling to make sense.

A common way that trialists indicate when they have skewed data is by reporting medians and interquartile ranges. When we encountered this, we noted that the data were skewed and considered the implication of this.

Where multiple trial arms were reported in a single trial, we included only the relevant arms. If two comparisons (e.g. drug A versus placebo and drug B versus placebo) were entered into the same meta‐analysis, we halved the control group to avoid double‐counting.

Unit of analysis issues

The unit of analysis was the individual participant. We did not find any cross‐over or cluster‐randomized studies.

Dealing with missing data

We contacted investigators or study sponsors in order to verify key study characteristics and obtained missing numerical outcome data when needed (e.g. when a study was identified as abstract only). We did not get a response. Thus, we only used the available data in the analyses.

Assessment of heterogeneity

We used the I² statistic to measure heterogeneity among the studies in each analysis (Higgins 2003). When we identified substantial heterogeneity (greater than 50%), we explored it by prespecified subgroup analysis, and we interpreted summary effect measures with caution.

Assessment of reporting biases

We did not perform funnel plots to assess reporting biases because there were fewer than 10 included studies (Sterne 2011).

Data synthesis

We performed the meta‐analyses using Review Manager 5 software (RevMan 2014). For all analyses, we used a random‐effects model.

Subgroup analysis and investigation of heterogeneity

We had intended to perform the following subgroup analyses, but were unable to because of limited data:

1. RCTs with low risk of bias versus RCTs with high risk of bias;

2. different etiologies (pancreatic cancer, chronic pancreatitis, and others);

3. type of operation (proximal, distal, and central pancreatectomy).

Sensitivity analysis

We performed sensitivity analyses, defined a priori, to assess the robustness of our conclusions. This involved:

1. changing between a fixed‐effect model and a random‐effects model;

2. changing between RR, risk differences (RD), and odds ratios (OR) for dichotomous outcomes;

3. changing between MD and SMD for continuous outcomes;

4. changing between worst‐case and best‐case scenario analyses for missing data.

If the results did not change, they were considered to have low sensitivity. If the results changed, they were considered to have high sensitivity.