Biologic Versus Synthetic Mesh in Ventral Hernia Repair: Participant-Level Analysis of Two Randomized Controlled Trials at Twenty-Four to Thirty-Six Months

Search method

PubMed, Embase, Scopus, Cochrane Central Register of Controlled Trials, and clinicaltrials.gov were searched to identify all published and ongoing RCTs comparing biologic mesh versus synthetic mesh. The search term used was (“biologic mesh” OR “matrix”) AND (“ventral hernia” OR “incisional hernia” OR “abdominal wall reconstruction”) AND (“randomized controlled trial” or “RCT”). This was adapted to the different databases. After deduplication, two authors, (N.H.D. and O.A.O.) independently screened titles, abstracts, and articles to identify eligible studies. Discrepancies were discussed and resolved with one of the two senior authors (M.K.L. or H.H.). The references of all retrieved abstracts were crosschecked for additional citations not captured by the search strategy.

Study selection and data extraction

Inclusion criteria were any RCT evaluating biologic mesh versus synthetic mesh for the repair of open ventral hernia repair. Exclusion criteria were inability to access patient level data, study design other than RCT, articles pertaining to hernia prevention rather than hernia treatment.

All authors were contacted to share the participant level data along with data dictionary. Only variables with common definitions were used such as variables reported in the National Surgical Quality Improvement Project, ¹⁰ definitions by the U.S. Centers for Disease Control and Prevention (CDC) ¹¹ for infections, and standardized definitions widely accepted among hernia surgeons. The following data was requested: age, gender, race/ethnicity, body mass index (BMI), smoking history, diabetes mellitus (DM), chronic obstructive pulmonary disease (COPD), immunosuppression, American Society of Anesthesiologists (ASA) score, past surgical history, prior ventral hernia repairs, hernia defect size, mesh type (randomization arm), mesh location, component separation (type and side), surgeon experience/expertise, and outcomes data, as below.

Primary outcome was a composite of major complications defined as major infection, re-operation, hernia recurrence, or death. Major infection included deep infection that encompasses mesh infection, organ/space infection, or enterocutaneous fistula. Re-operation was defined as any procedure involving the fascia, mesh, or peritoneal cavity. Hernia recurrence was based upon clinical examination and on demand imaging, which is current care. This outcome was chosen as it encompasses all major risks and potential benefits of each treatment.

Secondary outcomes included the individual outcomes of the primary outcome, SSI, surgical site occurrence (SSO), hospital re-admission, and hospital length of stay. Surgical site infection was defined by the CDC. ¹¹ Surgical site occurrence included wound complications such as seromas, hematomas, and wound dehiscence.

Patient-level analysis

Binary and discrete outcomes were evaluated with a logistic and negative binomial generalized linear model, respectively, under both frequentist and Bayesian frameworks.^12–14 All models were adjusted for location of study, ASA score, hernia width, mesh location, and wound class. For frequentist analyses, a relative risk (RR), 95% confidence interval, and p value were reported. A p value <0.05 was considered statistically significant. Bayesian models were performed using three different priors: a neutral prior centered at an RR of 1.0 (i.e., a priori no effect on the outcome) and a 95% credible interval (CrI) of 0.3–3.7; a skeptical prior (i.e., non-superiority) centered at an RR of 1.0 and a 95% CrI of 0.5–1.9; and an optimistic prior centered at an RR of 0.5 and a 95% CrI of 0.2–3.0 was used. These priors allow us to explore how the study results vary depending on doubt or confidence regarding treatment benefit with biologic mesh, using synthetic mesh as the reference, and therefore act as a sensitivity analysis. ¹⁵

For Bayesian analyses, we reported point estimates, 95% credible interval of treatment effect and the probability of benefit. The probability of treatment benefit was determined by the posterior probability that a RR is <1. In Bayesian analysis, a probability of superiority (or benefit) is reported when a probability of approaching 100% supports superiority, 50% suggests no clear superiority, and 0% suggests inferiority. Subgroup analyses of the primary outcome, planned a priori, were performed for the following variables one at a time: institution, ASA score, hernia width, mesh location, and wound class. Variables were chosen in consideration of how the included trials performed their respective randomization. All analyses were conducted in R software version 3.6.2 using the lme4 and brm package. ¹⁶

The protocol was registered in the PROSPERO international prospective register of systematic reviews (registration number CRD42020191138) on July 12, 2020.

Patient-level analysis

A total of 252 patients were included; 126 patients were randomized to the intervention of biologic mesh, and 126 patients were randomized to the control of synthetic mesh (Fig. 2). Patients randomized to biologic mesh received porcine acellular dermal matrix mesh and patients randomized to synthetic mesh received polypropylene mesh. Baseline demographics and hernia characteristics were similar between the two groups. Most patients were females and at least one-third of patients had one or more comorbid condition, including obesity. More than half of patients presented with a recurrent ventral hernia, and most hernias were either medium or large in size (Table 1). Slightly less than half of cases were contaminated (44%) with the majority of mesh placed in a retromuscular position. Primary fascial closure was performed in most patients (Table 2).

OPEN IN VIEWER

FIG. 2.

Flowsheet of included patients.

Patients were followed for a median of 2.4 years (interquartile range [IQR], 1.9–3.2) with 202 patients (80%) completing follow-up. Major complications occurred in 33 (33%) patients receiving biologic mesh and in 39 (38%) patients receiving synthetic mesh (RR, 0.91; 95% CI, 0.63–1.31; p value 0.600; Table 3). On Bayesian logistic regression, using a neutral prior, there was an 59% probability of benefit of fewer major complications with biologic mesh relative to synthetic mesh; using a skeptical prior, there was a 58% probability, and with an optimistic prior, there was a 63% probability of benefit with biologic mesh versus synthetic mesh (Table 4).

Mesh infection occurred in 14 (14%) patients assigned to biologic mesh versus 21 (20%) patients in the synthetic arm (RR, 0.77; 95% CI, 0.41–1.40; p value 0.386). Sixteen (16%) patients in the biologic arm underwent re-operation whereas 18 (17%) patients in the synthetic arm required re-operation (RR, 0.88; 95% CI, 0.49–1.61; p-value 0.696). There were 10 total re-operations for mesh infection, four (4%) in the biologic group and six (6%) in the synthetic group. Sixteen (16%) patients receiving biologic mesh suffered hernia recurrence as opposed to 11 (11%) patients receiving synthetic mesh (RR, 1.58; 95% CI, 0.81–3.09; p value 0.183). Three deaths occurred, one (1%) in the biologic arm and two (2%) in the synthetic arm. In the biologic arm, the single death occurred as a result of bacteremia in a patient with hepatocellular carcinoma. In the synthetic arm, one death was the result of fungal sepsis and one was reported as cardiopulmonary arrest at home. The remainder of secondary outcomes are reported in Table 3 and Table 4.

Subgroup analyses are reported in Table 5. In clean cases utilizing biologic mesh, 19% resulted in a major complication versus 28% of cases using synthetic mesh (RR, 0.60; 95% CI, 0.25–1.44; p value 0.450). Conversely, 52% of contaminated cases using biologic mesh resulted in a major complication as opposed to 44% of contaminated cases with synthetic mesh (RR, 1.39; 95% CI, 0.59–3.26; p value 0.450).