Types of studies

We included randomised controlled trials (RCTs), irrespective of their use of blinding, language of publication, publication status, date of publication, study setting or sample sizes. We planned to include cluster randomised clinical trials if the effect estimate was available after adjusting for clustering effect. We excluded other study designs.

Types of participants

All adults and children undergoing non‐obstetric abdominal closure irrespective of the reason for surgery, whether the surgery was performed as an elective or emergency procedure, or the size of the incision. We have excluded obstetric operations because the closure of the peritoneum in those has been assessed in another review (Bamigboye 2003), and the anatomy of closure is different, as it involves the visceral peritoneum that lines the abdominal organs. Obstetric operations are generally considered to generate 'clean' wounds compared to non‐obstetric operations, which can have 'clean', 'clean‐contaminated', 'contaminated' or 'dirty' wounds (Garner 1986) (Appendix 1). The incidence of postoperative wound complications is lower in clean operations (Garner 1986). We included trials in which the closure or non‐closure of the parietal peritoneum was the only systematic difference between treatment groups.

Types of interventions

Parietal peritoneal closure (either as part of an abdominal mass closure or as a separate layer) compared with no parietal peritoneal closure.

Types of outcome measures

Electronic searches

In February 2013 we searched the following electronic databases for potentially relevant trials:

1. The Cochrane Wounds Group Specialised Register (searched 14 February 2013);

2. The Cochrane Central Register of Controlled Trials (CENTRAL) (2013, Issue 1);

3. The Database of Abstracts of Reviews of Effects (DARE) (2013, Issue 1);

4. Ovid MEDLINE (1946 to February Week 1, 2013);

5. Ovid MEDLINE (In‐Process & Other Non‐Indexed Citations, February 13, 2013);

6. Ovid EMBASE (1974 to 2013 Week 06);

7. EBSCO CINAHL (1982 to 8 February 2013).

We used the following search strategy in CENTRAL and DARE:

#1 MeSH descriptor: [Sutures] explode all trees 662
#2 MeSH descriptor: [Wound Closure Techniques] explode all trees 1533
#3 (closure or close or closing or sutur*):ti,ab,kw 13372
#4 #1 or #2 or #3 13581
#5 MeSH descriptor: [Peritoneum] explode all trees 213
#6 peritone*:ti,ab,kw 2501
#7 #5 or #6 2501
#8 MeSH descriptor: [Abdomen] explode all trees and with qualifiers: [Surgery ‐ SU] 1298
#9 MeSH descriptor: [Laparotomy] explode all trees 664
#10 MeSH descriptor: [Abdominal Injuries] explode all trees 127
#11 MeSH descriptor: [Abdominal Cavity] explode all trees 472
#12 MeSH descriptor: [Abdominal Wall] explode all trees 73
#13 (abdom* near/5 surg*):ti,ab,kw 3763
#14 laparotom*:ti,ab,kw 1512
#15 #8 or #9 or #10 or #11 or #12 or #13 or #14 5711
#16 #4 and #7 and #15 119

The search strategies for Ovid MEDLINE, Ovid EMBASE and EBSCO CINAHL can be found in Appendix 2. We combined the Ovid MEDLINE search with the Cochrane Highly Sensitive Search Strategy for identifying randomised trials in MEDLINE: sensitivity‐ and precision‐maximising version (2008 revision) (Lefebvre 2011). We combined the EMBASE search with the Ovid EMBASE filter developed by the UK Cochrane Centre (Lefebvre 2011). We combined the CINAHL searches with the trial filters developed by the Scottish Intercollegiate Guidelines Network (SIGN 2011). We did not restrict studies with respect to language, date of publication or study setting.

We searched the metaRegister of Controlled Trials (mRCT) (http://www.controlled‐trials.com/mrct/) and ICTRP (International Clinical Trials Registry Platform) (http://apps.who.int/trialsearch/). The metaRegister includes the ISRCTN Register and NIH ClinicalTrials.gov Register among other registers. The ICTRP portal includes these trial registers along with trial registry data from a number of countries.

Searching other resources

We searched the references of included trials to identify further relevant trials. We contacted Ethicon and Dolphin sutures to enquire about any trials of which they were aware.

Selection of studies

Two review authors (KSG and ECD) identified the trials for inclusion independently by screening titles and abstracts and retrieving full text copies of those that potentially met the inclusion criteria. We have listed the excluded studies with reasons for their exclusion (Characteristics of excluded studies). We resolved any differences in opinion through discussion amongst the review authors.

Data extraction and management

Both review authors independently extracted the following data:

1. Year and language of publication.

2. Country of conduct of the trial.

3. Year of conduct of the trial.

4. Inclusion and exclusion criteria.

5. Sample size.

6. Details of suture material used and surgical technique used if the peritoneum was closed.

7. Outcomes (as described above).

8. Risk of bias (as described below).

We sought further information from authors when information was not available. Had multiple reports been identified for a trial, we planned to examine all the reports for information. We planned to seek clarification for any unclear or missing information by contacting the authors of individual trials. If there had been any doubt about whether the trials shared the same participants (by identifying common authors and centres), we planned to contact the study authors to check whether the trial report had been duplicated. We resolved any differences in opinion through discussion amongst the review authors.

Assessment of risk of bias in included studies

We followed the instructions given in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011b). According to empirical evidence (Kjaergard 2001; Moher 1998; Schulz 1995; Wood 2008), we planned to assess the risk of bias of included trials based on the following risk of bias domains:

Measures of treatment effect

For dichotomous variables, we calculated the risk ratio (RR) with 95% confidence interval (CI). The dichotomous outcomes include short‐term mortality. If the event is very rare (less than 1% occurrence), we planned to use the Peto odds ratio and 95% CI for dichotomous variables (Higgins 2011d). We planned to calculate the rate ratio (RaR) with 95% CI for serious adverse events (since more than one serious adverse event can develop in the same patient), such as incidence of incisional hernia, incidence of adhesions and the incidence of requirement for re‐operation if the annual incidence was available, but none of these were available. So, we have calculated the risk ratio for incisional hernia, adhesions and requirement for re‐operation, ignoring the difference in time between the different trials, clearly aware that this may be a source of heterogeneity. For continuous variables, we calculated the mean difference (MD) with 95% CI for outcomes that can be quantified, such as hospital stay, and planned to calculate the standardised mean difference (SMD) with 95% CI for quality of life (where different assessment scales might be used).

Unit of analysis issues

The unit of analysis was individual patients undergoing laparotomy or laparoscopy.

Dealing with missing data

We performed an intention‐to‐treat analysis whenever possible (Newell 1992). We imputed data for binary outcomes using various scenarios such as best‐best scenario, best‐worst scenario, worst‐best scenario, and worst‐worst scenario (Gurusamy 2009). In the best‐best scenario, the outcomes of patients with missing data in both groups were assumed to be good. In the best‐worst scenario, the outcomes of patients with missing data in the intervention group were assumed to be good while the outcomes were assumed to be bad for those with missing data in the control group. The worst‐best scenario was the opposite of the best‐worst scenario, i.e. the outcomes of patients with missing data in the intervention group were assumed to be bad while they were assumed to be good for those in the control group with missing data. In the worst‐worst scenario, the outcomes of patients with missing data in both groups were assumed to be bad.

For continuous outcomes, we used available‐case analysis. We have imputed the standard deviation from P values according to the instructions in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011c), and we planned to use the median for the meta‐analysis when the mean was not available. If it was not possible to calculate the standard deviation from the P value or the confidence intervals, we imputed the standard deviation as the highest standard deviation in the other trials included under that outcome, fully recognising that this form of imputation will decrease the weight of the study for calculation of mean differences and bias the effect estimate to no effect in case of standardised mean difference (Higgins 2011d).

Assessment of heterogeneity

We explored heterogeneity using the Chi² test with significance set at P value 0.10, and measured the quantity of heterogeneity using the I² statistic (Higgins 2002).

Thresholds for the interpretation of the I² statistic can be misleading. A rough guide to interpretation is as follows (Deeks 2011).

1. 0% to 40%: might not be important.

2. 30% to 60%: may represent moderate heterogeneity.

3. 50% to 90%: may represent substantial heterogeneity.

4. 75% to 100%: represents considerable heterogeneity.

We took factors such as clinical and methodological heterogeneity into account while interpreting the I² statistic, particularly where there was overlapping of ranges.

Assessment of reporting biases

We planned to explore reporting bias using visual asymmetry on the funnel plot (Egger 1997; Macaskill 2001). We planned to perform linear regression as described by Egger 1997 to determine funnel plot asymmetry. We explored reporting bias from selective outcome reporting of the trials.

Data synthesis

We performed the meta‐analysis using RevMan 5 software (RevMan 2011), and following the recommendations of The Cochrane Collaboration (Deeks 2011). We used both a random‐effects model (DerSimonian 1986), which provides an average of the intervention effect, and a fixed‐effect model (DeMets 1987), which provides the average intervention effect, for meta‐analyses. Where there was discrepancy between the two models identified from the pooled estimates and their confidence intervals resulting in a change in interpretation of information, we reported both results; otherwise we reported the results of the fixed‐effect model. With regard to dichotomous outcomes, risk ratio calculations do not include trials in which no events occurred in either group in the meta‐analysis, whereas risk difference calculations do. We planned to report the risk difference (RD) if the results using this association measure were different from risk ratio in terms of statistical significance. However, risk ratio is the measure that we used to arrive at conclusions, since risk ratios perform better when there are differences in the control event rate (proportion of patients who develop the event in the control).

Subgroup analysis and investigation of heterogeneity

We planned to perform the following subgroup analyses.

1. Peritoneal closure with different types of suture material.

2. Peritoneal closure using continuous sutures versus interrupted sutures.

3. Trials with adults compared to trials with children.

4. Trials with clean wounds versus clean‐contaminated wounds versus contaminated wounds versus dirty wounds (Garner 1986) (Appendix 1).

5. Laparotomy versus laparoscopy.

6. Vertical versus transverse incisions.

We planned to use a P value of less than 0.05 for the Chi² test to identify the differences (heterogeneity in the effect estimates) between subgroups.

Sensitivity analysis

We planned to perform a sensitivity analysis by excluding trials at high risk of bias. We planned to perform sensitivity analysis by imputing data for dichotomous outcomes using various scenarios including best‐best analysis, best‐worst analysis, worst‐best analysis, and worst‐worst analysis scenarios (Gurusamy 2009). In each of these analyses, the first component provides information on the way we planned to impute the outcome for intervention group and the second component provides information on the way we planned to impute the outcome for the control group. The word 'best' indicates that the participant with the missing outcome had a favourable event and the word 'worst' indicates that the participant with the missing outcome had an unfavourable event. We also planned to perform sensitivity analysis by excluding the trials in which the mean and the standard deviation were imputed.