Setting

All studies took place in the operating department. One was a multi‐centred US study (Towfigh 2008), one an international multi‐centred study (von Eckardstein 2011). The remainder took place in single hospitals in France (Dromzee 2012), Turkey (Hanedan 2014), Australia (Iyer 2011), or the Netherlands (Vierhout 2014). The setting of one study was not made explicit (Falk‐Brynhildsen 2014).

Participants

Dromzee 2012 included 56 participants undergoing posterior spinal correction for scoliosis. Participants had a mean age of 15 years. Falk‐Brynhildsen 2014 recruited 140 patients undergoing elective CABG with harvesting of the saphenous vein for at least two coronary arterial grafts: 110 participants were males and the mean age of all participants was 67 years. The venous harvest wound on the leg was used as the site of interest. Hanedan 2014 included 96 people undergoing cardiac surgery (specifically CABG, valve repair or repair of congenital heart disease), the mean age was 51 and 63 of the participants were male. The sternal wound site was assessed for infection. Iyer 2011 included 47 participants with a mean age of 67 years, 39 of whom were male, undergoing coronary artery bypass grafting (CABG). Randomisation was by leg, rather than by patient. The long saphenous vein site on both legs was chosen as the site of interest. Recruitment was halted at 47 participants after further ethical committee review. Towfigh 2008 included 177 participants, 170 of whom were male, undergoing an elective hernia repair in the operating theatre. The mean age was 53 years. The trial design had been informed by a sample size calculation based on the results from a porcine study, and it was estimated that 206 participants were required to demonstrate a 23% difference in wound contamination between groups, with 90% power. An interim analysis, however, was conducted after 104 participants had been recruited, and the sample size was revised to 742 participants required to detect a 10% difference in wound contamination with 80% power. The trial, however, was stopped prematurely having recruited only 177 participants, when the FDA granted regulatory approval for the product as a class II investigational device. The Vierhout 2014 study recruited 50 participants undergoing femoro‐popliteal, femoro‐crural or femoro‐femoral crossover bypass graft insertion; of these 28 participants were male and the mean age was 71. For this study an initial sample size calculation worked on the assumption of a 12% incidence of surgical site infection, meaning 180 patients would be required in both groups. However an interim analysis recorded an SSI incidence at 6% and as a result subsequent power calculations required the recruitment of 748 patients to both groups. At this point the review board giving ethical approval deemed this number of participants too large and the trial was stopped. von Eckardstein 2011 randomised 293 participants for CABG. One‐hundred and forty‐nine were male. The mean age of patients in this trial was 63 years. Both sternal and leg wounds were used as surgical sites for the study.

Experimental intervention

Two studies used the InteguSeal IS 200 applicator to administer the sealant (Dromzee 2012; Vierhout 2014). This applicator was used following standard preoperative skin preparation, namely 5% alcoholic povidone‐iodine solution in Dromzee 2012 and 0.5% chlorhexidine in 70% isopropyl alcohol in Vierhout 2014.

In Falk‐Brynhildsen 2014, an operating room nurse used the InteguSeal IS100 applicator to apply the microbial skin sealant on the patient’s leg (at the venous harvest site) just before the surgical incision but after preoperative skin preparation (which consisted of 0.5% chlorhexidine in 70% ethanol). Iyer 2011 used the InteguSeal IS 100 applicator to apply microbial sealant to saphenous vein harvest sites on one of the legs on a random basis (the other leg acted as the participant's control). This followed application of alcoholic povidone‐iodine solution. Both substances were allowed to dry after application.

In the Towfigh 2008 trial, a microbial, film‐forming, liquid sealant was applied via a disposable sponge, to the surgical sites of participants undergoing elective hernia repair. Application was immediately prior to the surgical incision and followed the standard preoperative skin preparation procedure of an application of 10% povidone‐iodine. In Hanedan 2014 cyanoacrylate microbial (InteguSeal) was applied after standard skin preparation (10% povidone‐iodine solution).

In the von Eckardstein 2011 trial, a microbial skin sealant (InteguSeal) was applied following standard surgical preparation. The sealant was considered dry when a film formed on the skin.

Control intervention

In all studies the standard preoperative skin preparation was used in the experimental and control sites: that is, 10% povidone‐iodine (Hanedan 2014; Towfigh 2008), alcoholic povidone‐iodine solution (Iyer 2011), 0.5% chlorhexidine solution in 70% ethanol (Falk‐Brynhildsen 2014), chlorhexidine 0.5% in 70% isopropyl alcohol (Vierhout 2014) or povidone iodine or 0.7% available iodine in isopropyl alcohol 74% weight/weight (w/w) (von Eckardstein 2011). In Dromzee 2012 5% povidone‐iodine was used for standard surgical skin preparation followed by 3M Steri‐Drapes in the control group. Drapes were also used in the control group in the Hanedan 2014 trial. In the text of Hanedan 2014 the phrase "control group" is at one point used to describe the use of sealants and vice versa ‐ correspondence with the author clarifies that the control group was indeed the group receiving drapes and skin preparation but no sealants, and that cyanoacrylate microbial sealants were used only in the intervention group.

Outcome measures

In Dromzee 2012 the primary outcome measure was surgical site infection, though it is unclear from the study text how this has been defined. Secondary outcomes in Dromzee 2012 were intraoperative blood loss, intraoperative time and number of vertebral levels fused.

The primary outcome measure reported by four trials was bacterial contamination (Falk‐Brynhildsen 2014; Hanedan 2014; Towfigh 2008; von Eckardstein 2011). Falk‐Brynhildsen 2014 measured postoperative skin contamination as a primary outcome measure; secondary end‐points were time to recolonisation and surgical site infection, defined as participants being given a prescription by a doctor for antibiotics to treat wound infection. In Hanedan 2014 primary outcome measures were bacterial contamination of wound site and postoperative white blood cell (WBC) count; rates of SSI (defined as any evidence of sternal wound infection at follow‐up) were a secondary outcome measure, and adverse events and mortality were also reported. Towfigh 2008 measured secondary outcomes that included: prevalence of antibiotic‐resistant Staphylococcus aureus in the wound during surgery; any difference in postoperative SSI rates; and safety outcomes associated with microbial sealant, including adverse events. In von Eckardstein 2011 secondary outcomes were comparison of bacterial count pre‐ and post‐CABG and the proportion of SSIs in both groups according to the Centers for Disease Control Prevention National Nosocomial Infections Surveillance (NNIS) criteria.

In two studies (Iyer 2011; Vierhout 2014) the primary outcome measure was surgical wound infection graded according to the Southamption wound grading system (Bailey 1992). Secondary outcomes in Iyer 2011 were irritation and allergies. Secondary outcomes in Vierhout 2014 were not described in the methodology section but all‐cause mortality and serious complications were reported.

Two studies were funded by Kimberly‐Clark Health Care, the manufacturers of the microbial sealant (Towfigh 2008; von Eckardstein 2011). In the von Eckardstein 2011 study two authors disclosed that they had a financial relationship with Kimberly‐Clark Corporation. Kimberly‐Clark Health Care provided applicators for experimental use in one study (Falk‐Brynhildsen 2014). Other studies disclosed no relationship with a manufacturer (Dromzee 2012; Hanedan 2014; Iyer 2011; Vierhout 2014).

Sequence generation

The method of generating the randomisation sequence was reported in six studies (Dromzee 2012; Falk‐Brynhildsen 2014; Hanedan 2014; Towfigh 2008; Vierhout 2014; von Eckardstein 2011 ) but not reported in Iyer 2011.

Allocation concealment

The method of allocation concealment was not stated in four studies (Dromzee 2012; Falk‐Brynhildsen 2014; Hanedan 2014; Iyer 2011). In Towfigh 2008 the sealed envelopes were not confirmed as opaque and sequentially‐numbered. In Vierhout 2014 envelopes were "blinded" but opacity and sequential numbering was not confirmed. von Eckardstein 2011 conducted allocation concealment and reported the method.

Blinding

Participant blinding was unclear in one study (Iyer 2011), and in three studies participants were blinded (Falk‐Brynhildsen 2014; Hanedan 2014; Towfigh 2008). In Vierhout 2014 patients were blinded to their allocation until 48 hours postoperatively, at which point dressings were taken down and the use of sealants inevitably revealed to the patient. No report of blinding of participants was made in Dromzee 2012. The von Eckardstein 2011 trial was described as an "open label" and therefore participants were not blinded. In five studies outcome assessors were blinded to the intervention (Falk‐Brynhildsen 2014; Hanedan 2014; Iyer 2011; Towfigh 2008; Vierhout 2014). Dromzee 2012 does not describe blinding of care providers or outcome assessors. von Eckardstein 2011 was described as an open label trial therefore blinding of care providers or outcome assessors did not take place. The nature of the intervention may make blinding of care providers (viz., surgeons) quite difficult. This is acknowledged by both Dromzee 2012 and Towfigh 2008.

Incomplete outcome data

There were no drop outs in three studies (Dromzee 2012; Hanedan 2014; Iyer 2011). In Towfigh 2008 29 participants were lost to follow‐up, and data were not analysed on an intention‐to‐treat basis. Data were analysed by intention‐to‐treat analysis in von Eckardstein 2011, but 31 participants were described as ineligible for the per protocol analysis. Ten patients were lost to follow‐up (Falk‐Brynhildsen 2014), eight due to postoperative death and two due to language difficulties ‐ it is not clear from the study text that data were analysed on an intention‐to‐treat basis. There was a single postoperative death and no other drop outs in Hanedan 2014. In Vierhout 2014 three patients died postoperatively and though in the study text authors describe calculating power "on the basis of an intention‐to‐treat principle", these early postoperative deaths were excluded from subsequent analysis.

Selective outcome reporting

The presence of selective outcome reporting was unclear for all studies. Prespecified outcomes as stated in the studies were reported, but we were only able to access the study protocol for one study, Falk‐Brynhildsen 2014, which described a somewhat broader inclusion criteria and listed calculation of an ASEPSIS score as an outcome measure, though this calculation did not appear in the published report and no account was given for these differences in the study text.

Other potential sources of bias

In one study there was a baseline imbalance with a significantly longer operation time in the control group (Hanedan 2014). In two studies there was baseline imbalance between the groups in relation to obesity, with no apparent adjustment for this (Towfigh 2008; von Eckardstein 2011). In both studies there were more obese participants in the intervention group. No baseline imbalance was described in Dromzee 2012, Towfigh 2008 or Vierhout 2014.

Three trials were stopped prematurely (Iyer 2011; Towfigh 2008; Vierhout 2014). In Iyer 2011 the study was halted after further ethical review. The Towfigh 2008 trial was stopped prematurely when the FDA granted regulatory approval for InteguSeal (microbial sealant) as a class II medical device. The trial reports this as 'additional information' in the published paper. Neither Iyer 2011 nor Towfigh 2008 mentioned the effect of this early stopping on either the findings (up to the point of stopping), or on the risk of bias, neither was this event discussed in the papers. On this basis we would suggest that the studies were at high risk of bias. In Vierhout 2014 an interim analysis of SSI rates lead to an upwards revision of the sample size needed to yield adequate power, at which point the study's ethics committee halted the trial. There is no discussion of the effects of early stopping in the text, though the halting of the trial and the reasons behind this were addressed in the paper.

All‐cause mortality

Mortality was reported in one study (von Eckardstein 2011), where there were four deaths (one in the intervention group and three in the control group) of cardiac/circulatory origin; none was considered to be related to the microbial sealant. Eight deaths were reported in another study (Falk‐Brynhildsen 2014) but the causes of these deaths were not discussed in the study text. Hanedan 2014 reported one postoperative mortality, though the reasons for this are not detailed in the trial report. There were two deaths due to postoperative cardiac complications and one due to bowel ischaemia reported in Vierhout 2014.

Adverse reactions

No adverse events were reported by Falk‐Brynhildsen 2014 or Iyer 2011. Towfigh 2008 reported adverse events in both groups but none were considered to be due to the microbial sealant. von Eckardstein 2011 reported adverse events (11 in the intervention group and 16 in the control group). Hanedan 2014 reported one case of sternal wound dehiscence (with no clinical or laboratory evidence of infection noted) in one patient in the microbial sealant group, requiring surgical management. Vierhout 2014 reported noninfectious lymphatic complications requiring fluid draining in two participants in the intervention group (n = 25) and two in the control group (n = 22), the authors do not state that they consider this related to the use or absence of cyanoacrylate sealants.

Other serious infection or infectious complications

In Towfigh 2008 one participant was given antibiotics for a deep infection with MRSA and was admitted to hospital for wound debridement, mesh removal and intravenous antibiotics. Four participants in Iyer 2011 required incision and drainage in the control group. In one study where wound contamination was being assessed, positive cultures were grown from samples from 38 patients, (27/68 from the microbial sealant group and 11/28 from the control group, in all of these cases there was no clinical evidence of infection (Hanedan 2014). One patient in the sealant group of the Vierhout 2014 study required further surgery at 22 days due to an infected graft following SSI.

Length of hospital stay, rates of hospital re‐admissions, costs

The above outcomes were not reported in any of the included studies.

Postoperative antibiotic use

In Falk‐Brynhildsen 2014 all patients received pre‐, intra‐ and post‐operative prophylactic antibiotics; in this trial the use of antibiotics in the two‐month postoperative period was used to define SSI. In Hanedan 2014 all patients received pre‐ and post‐operative prophylactic antibiotics. In Dromzee 2012, postoperative antibiotics and wound debridement were used to treat the six infection (event) cases. The two SSI cases amongst the control group in the Vierhout 2014 trial were treated with antibiotics (oral amoxicillin/clavulanic acid 625 mg for three days), which lead to healing of the SSI.