Types of studies

Randomised controlled trials (RCTs) only.

Types of participants

We included all ages of participants in the review. There was no restriction on the setting in which the treatment was delivered. We considered studies including both new and recurrent pilonidal sinus.

Types of interventions

Fibrin glue (any brand containing a mixture of fibrinogen and thrombin, human or bovine) compared with other surgical techniques such as primary closure, closure through secondary intention and off‐midline closure techniques. We also included trials if they used fibrin glue as an adjunct to surgery (compared with surgery alone).

Types of outcome measures

Electronic searches

We searched the following electronic databases to identify relevant RCTs:

• the Cochrane Wounds Specialised Register (searched 13 December 2016);

• The Cochrane Central Register of Controlled Trials (CENTRAL; 2016, issue 12) in the Cochrane Library;

• Ovid MEDLINE (including Epub Ahead of Print, In‐Process & Other Non‐Indexed Citations, Ovid MEDLINE Daily) (1946 to 13 December 2016);

• Ovid Embase (1974 to 13 December 2016);

• EBSCO CINAHL Plus (1937 to 13 December 2016).

The search strategies for CENTRAL, Ovid MEDLINE, Ovid Embase and EBSCO CINAHL Plus can be found in Appendix 1. We did not combine the topic‐specific searches with any RCT filters as the number of search results were low. There were no restrictions with respect to language, date of publication or study setting.

We also searched the following clinical trials registries (searched December 2016):

• ClinicalTrials.gov;

• World Health Organization (WHO) International Clinical Trials Registry Platform (ICTRP)

• EU Clinical Trials Register.

Searching other resources

There was no language restriction in the search and, where necessary, we translated non‐English language papers using Google Translate. We sought grey literature from OpenSIGLE. We searched conference proceedings from the last three years (Association of Coloproctology of Great Britain and Ireland, European Society of Coloproctology, Royal Society of Medicine Coloproctology Section and American Society of Colon and Rectal Surgeons). We handsearched reference lists of identified studies for other suitable papers. We utilised Google Scholar to identify papers that had cited identified studies.

Selection of studies

Two review authors independently matched studies to the inclusion and exclusion criteria (BD and JPW) and, if any disagreement existed, a third review author was consulted and we reached agreement by consensus. We extracted the titles and abstracts of potentially relevant studies onto an electronic database and these underwent full‐text review with duplicates being removed. We used full‐text reports to match against the inclusion and exclusion criteria. We used the kappa statistic to measure agreement during the selection process. We identified duplicate publications of the same studies by using author names, study location, intervention details and cohort dates. We planned to link any duplicate publications. As one review author had authored one of the included studies (JL), we performed selection of studies using two different authors to reduce bias.

Data extraction and management

Two review authors independently collected and extracted data into an electronic database (BD and ST). As one review author had authored one of the included studies, we performed data extraction using two different authors to reduce bias. We then compared the data and resolved discrepancies by checking the original study. We stored data on a password‐protected computer. If any information was not reported, we contacted the study authors for further information.

Where possible we extracted the following data:

• bibliographic data including date of completion/publication;

• country of origin;

• publication status of study;

• source of funding for trial;

• trial design;

• care setting;

• number of participants randomised to each trial arm and number included in final analysis;

• eligibility criteria and key baseline participant data including category(s) and location(s) of sinus;

• details of treatment regimen received by each group;

• duration of treatment;

• details of any co‐interventions;

• primary and secondary outcome(s) (with definitions and, where applicable, time points);

• outcome data for primary and secondary outcomes (by group);

• duration of follow‐up;

• number of withdrawals (by group) and number of withdrawals (by group) due to adverse events;

• adverse events.

Assessment of risk of bias in included studies

Two study authors (BD and ST) assessed risk of bias independently and a third author resolved any disagreements (JPW). We used the Cochrane tool for assessing risk of bias (Appendix 2; Higgins 2011a). We recorded sequence generation, allocation concealment, blinding of participants, personnel and outcome assessors, incomplete outcome data, selective outcome reporting and other sources of bias. We assigned each study low, high or unclear risk of bias using published criteria (Higgins 2011a). To assess selective outcome reporting, we searched clinical trials databases for the original study registration and compared this with the published study. In terms of blinding, we recognised that this would be difficult to implement fully as in other surgical studies. However, blinding of outcome assessment was deemed possible in studies comparing fibrin glue as an adjunct to surgery. We presented risk of bias data in a summary table and 'Risk of bias' graph.

Measures of treatment effect

We presented dichotomous outcomes as risk ratios (RR). We presented continuous outcomes as mean differences (MD) if scales were comparable or standardised mean differences (SMD) if non‐comparable scales were used. We would have reported time‐to‐event data as hazard ratios (HR). If HR were not reported, we would have estimated these using the events in each group and derived the P value from the log rank test (Tierney 2007). Where time to healing was wrongly reported as a continuous outcome (as opposed to a time‐to‐event outcome) and we were not able to derive hazard ratios we endeavoured to find out whether all participants had healed and we planned to exclude such data from meta‐analyses.

Unit of analysis issues

As the person was the unit of analysis and they are likely to have only one pilonidal sinus, there were no unit of analysis issues (Higgins 2011b).

Dealing with missing data

If studies with missing data were identified, we contacted the study authors. If we received no response, we extracted data from published graphs if appropriate. If standard deviations were not reported, we estimated these from other studies in the review. We used the most clinically similar study to the one with missing values for these estimates. We did not use methods of imputation for missing participant data.

Assessment of heterogeneity

We assessed clinical heterogeneity during the data extraction process, especially the trial methods and baseline demographics. If significant clinical heterogeneity was present, we did not pool studies and discussed them in the narrative review. For statistical heterogeneity, we used the I² statistic (Higgins 2003). We pooled data using the fixed‐effect model in the first instance. If substantial heterogeneity was present, we pooled data using the random‐effects model. We used the following cut‐off values for the I² statistic (Deeks 2011):

• 0% to 40%: might not be important;

• 30% to 60%: may represent moderate heterogeneity;

• 50% to 90%: may represent substantial heterogeneity;

• 75% to 100%: considerable heterogeneity.

Assessment of reporting biases

We assessed selective reporting bias as part of the Cochrane 'Risk of bias' tool. We planned to assess publication bias qualitatively using funnel plots and quantitatively using Egger's regression test if 10 studies or more were included in the analysis (Sterne 2011). However, the low number of included studies precluded this.

Data synthesis

We presented meta‐analysis as forest plots if methods were similar enough that data synthesis was deemed appropriate. If not appropriate, we performed qualitative synthesis. We did not estimate missing data on measures of central tendency from reported values (mean estimated from median). We contacted study authors to provide any missing values. We combined continuous outcomes using generic inverse variance from the standard errors. We combined risk ratios using the Mantel‐Haenszel method. We used a fixed‐effect model if there was no heterogeneity. If heterogeneity was present, we used the random‐effects model (Deeks 2011).

Subgroup analysis and investigation of heterogeneity

We planned to conduct subgroup analysis and meta‐regression. However, the small number of included studies precluded this.

Sensitivity analysis

We planned to conduct a sensitivity analysis excluding studies that we judged high risk of bias for any of the 'Risk of bias' assessments and those sponsored by pharmaceutical companies. If sensitivity analysis results differed substantially, we would have carefully interpreted the final results.