Types of studies

We considered randomised controlled trials (RCTs) evaluating any intervention for treating post‐extraction bleeding (PEB). We excluded quasi‐RCTs, cross‐over trials and preventive trials. We had planned to include split‐mouth studies, provided there was no possibility of contamination. Split‐mouth design is one of the self‐controlled study designs that is unique to dentistry. The design is characterised by subdividing the mouth of the subject into homogeneous within‐patient experimental units such as quadrants, sextants, contralateral or ipsilateral quadrants or sextants, or a symmetrical combination of these.

Types of participants

We considered trials with participants of any age and either gender, who reported PEB, regardless of the type of teeth (anterior or posterior, mandibular or maxillary).

Types of interventions

We considered any surgical or non‐surgical technique or material used for the treatment of PEB. We had planned to make the following comparisons.

1. Direct comparisons of different interventions

2. Intervention versus placebo or no treatment

Types of outcome measures

Electronic searches

Cochrane Oral Health’s Information Specialist conducted systematic searches in the following databases for randomised controlled trials and controlled clinical trials without language or publication status restrictions:

• Cochrane Oral Health’s Trials Register (searched 24 January 2018) (Appendix 1);

• Cochrane Central Register of Controlled Trials (CENTRAL; 2017, Issue 12) in the Cochrane Library (searched 24 January 2018) (Appendix 2);

• MEDLINE Ovid (1946 to 24 January 2018) (Appendix 3);

• Embase Ovid (1 May 2015 to 24 January 2018) (Appendix 4);

• CINAHL EBSCO (Cumulative Index to Nursing and Allied Health Literature; 1937 to 24 January 2018) (Appendix 5).

Subject strategies were modelled on the search strategy designed for MEDLINE Ovid. Where appropriate, they were combined with subject strategy adaptations of the highly sensitive search strategy designed by Cochrane for identifying randomised controlled trials and controlled clinical trials as described in the Cochrane Handbook for Systematic Reviews of Interventions Chapter 6 (Lefebvre 2011).

Due to the Cochrane Centralised Search Project to identify all clinical trials in the database and add them to CENTRAL, only most recent months of the Embase database were searched at the review's inception, and this search was updated for this version of the review. See the searching page on the Cochrane Oral Health website for more information. No other restrictions were placed on the date of publication when searching the electronic databases.

Searching other resources

The following trial registries were searched for ongoing studies:

• US National Institutes of Health Ongoing Trials Register ClinicalTrials.gov (clinicaltrials.gov; searched 24 January 2018) (Appendix 6);

• World Health Organization International Clinical Trials Registry Platform (apps.who.int/trialsearch; searched 24 January 2018) (Appendix 7).

We searched the reference lists of relevant systematic reviews for further studies.

We did not perform a separate search for adverse effects of interventions.

Selection of studies

Three pairs of review authors (Ashok L (AL) and Prashanti Eachempati (PE), Himanshi Aggarwal (HA) and Kiran Kumar (KK), and Muthu MS (MMS) and Haszelini Hassan (HH)), independently and in duplicate, screened the titles and abstracts from the electronic searches to identify potentially eligible studies. The search was designed to be sensitive and include controlled clinical trials; these were filtered out early in the selection process if they were not randomised. We obtained full‐text copies of all eligible and potentially eligible studies, which were independently evaluated by two authors (Sumanth Kumbargere Nagraj (SKN) and PE) . We resolved disagreements by discussion. When resolution was not possible, we consulted an arbiter (Adinegara Lutfi Abas). We assessed articles in languages other than English after the abstracts had been translated (Table 1). We did not find any trials that fulfilled the inclusion criteria.

Data extraction and management

We had planned that two review authors (SKN, PE) would independently extract the data; and would not have been blinded to the authors. We would have resolved any disagreements by discussion between the two review authors, if necessary, consulting a third review author (PE) in order to reach consensus. We had planned to extract data using a customised data extraction form. All the items in the data extraction form were designed following guidance from the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011). We would have entered study details into the 'Characteristics of included studies' table in Review Manager (RevMan) software (RevMan 2014), recording the following details for each included trial:

• publication details such as year of publication, language;

• demographic details of the report;

• inclusion and exclusion criteria;

• type of trial, sample size, method of randomisation, allocation concealment, blinding, method of assessing the outcomes and drop‐outs, if any;

• type of intervention;

• details of the outcomes reported;

• duration of follow‐up;

• results of the intervention;

• funding details.

We had planned to write, email or telephone the contact author of included studies when clarification of details or additional data were required.

Assessment of risk of bias in included studies

Two review authors (SKN and PE) had planned to independently assess the risk of bias in the included trials in seven domains:

• random sequence generation (selection bias);

• allocation concealment (selection bias);

• blinding of participants and personnel (performance bias);

• blinding of outcome assessment (detection bias);

• incomplete outcome data (attrition bias);

• selective outcome reporting (reporting bias);

• other biases.

For each of these components, we had planned to assign a judgment regarding the risk of bias as either high, low, or unclear based on guidance in Higgins 2011. We had planned to contact the trial authors if details were missing or unclear, and resolve disagreements through consensus. We had planned to record our judgements and justifications in 'Risk of bias' tables for each included study and generate a 'Risk of bias' summary graph and figure. We would have used these judgements to grade the overall quality of evidence for each comparison and outcome in the 'Summary of findings' tables.

We had planned to summarise the risk of bias according to Higgins 2011; see Table 2.

Measures of treatment effect

We had expected our primary outcome to be expressed in dichotomous data. We would have expressed the effect estimate as a risk ratio (RR) with 95% confidence intervals (CI). We had expected our secondary outcomes to be presented as dichotomous data, continuous data, or ordinal scales. We had planned to use means and standard deviations to calculate mean differences and 95% CI for continuous data measured with the same scales and standardised mean differences if studies used different scales to measure the same outcome.

If data were expressed in ordinal scales, we had planned to explore the possibility of converting them to dichotomous outcomes. If outcomes were reported at baseline, trial endpoints, or follow‐up, we had planned to extract the mean change and standard deviation from baseline for each treatment group. We had intended to pool either end scores or change scores, but preferred end scores when available; we would have combined end and change scores where necessary.

We had planned to analyse data expressed as counts (number of bleeding incidents) in the same way as continuous data.

Unit of analysis issues

We expected two types of non‐standard study designs in this review:

• multiple treatment groups;

• split‐mouth design

We were expecting data related to repeated observation on participants for our secondary outcome of time required for the control of bleeding. In this case, we had planned to follow the method described in section 9.3.4 of the Cochrane Handbook (Higgins 2011).

In trials where adverse effects were described as counts, we wanted to follow the method described in section 9.2.5 of the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011). In the case of dropouts, we had planned to use what the paper reports and deal with it in the 'Risk of bias' assessment. We were not expecting to find any cluster‐randomised trials for this condition.

Dealing with missing data

We had planned to contact study authors to obtain missing data. We would have used the methods outlined in section 16.1.2 of the Cochrane Handbook for Systematic Reviews of Interventions to calculate missing standard deviations. If it had not been possible to calculate the SDs, we would have described the outcomes qualitatively (Higgins 2011).

Assessment of heterogeneity

If meta‐analyses were performed, we would have assessed heterogeneity using a Chi² test, where a P value < 0.1 indicated statistically significant heterogeneity. We would have quantified heterogeneity using the I² statistic as follows:

• 0% to 40% implies slight heterogeneity;

• 30% to 60% moderate heterogeneity;

• 50% to 90% substantial heterogeneity;

• 75% to 100% considerable heterogeneity.

If there was considerable heterogeneity (I² > 75%), which could not be explained by the subgroup analyses, we planned not to conduct meta‐analysis. We had planned to interpret I² values between 0% to 40% as possibly insignificant, 30% to 60% as possibly significant, 50% to 90% as possibly substantial, and 75% to 100% as possibly very substantial ('considerable'); depending on whether the inconsistency in results was due to differences in the direction of effect estimates between trials rather than due to differences in the magnitude of effect estimates favouring an intervention (Deeks 2011).

Assessment of reporting biases

If there were more than 10 studies included in a meta‐analysis, we had planned to assess the possible presence of reporting bias by testing for asymmetry in a funnel plot. If present, we would have carried out statistical analyses using the methods described by Egger 1997.

Data synthesis

We had planned to analyse the data using RevMan software (RevMan 2014). If the data available from the studies had similar comparisons and outcomes, we would have undertaken a meta‐analysis. Our general approach would have been to use a random‐effects model. With this approach, the CIs for the average intervention effect are wider than those obtained using a fixed‐effect approach, leading to a more conservative interpretation. We wanted to use all end scores or all change scores when available, but would have combined end and change scores where necessary, using the criteria in section 9.4.5.2 of the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011). We had planned to report the results from studies not included in a meta‐analysis in additional tables.

Subgroup analysis and investigation of heterogeneity

If there was significant heterogeneity, we had planned to explore the reasons by performing the following subgroup analyses based on different groups of patients. The subgroups were to be divided based on:

• type of PEB (primary, reactionary, or secondary);

• type of extraction (surgical or forceps);

• type of underlying bleeding or clotting disorder (deficiency of factors, qualitative disorders, vessel disorders);

• drug history (anticoagulant, antiplatelet, or combination);

• type of teeth (deciduous or permanent, mobile or firm, maxillary or mandibular, anterior or posterior).

Sensitivity analysis

If there were sufficient included studies, we would have undertaken sensitivity analysis based on risk of bias, including only studies at low risk of bias.