Scolaris Content Display Scolaris Content Display

Mesh versus non‐mesh for inguinal and femoral hernia repair

Collapse all Expand all

Background

This is an update of a Cochrane Review first published in 2001.

Hernias are protrusions of all or part of an organ through the body wall that normally contains it. Groin hernias include inguinal (96%) and femoral (4%) hernias, and are often symptomatic with discomfort. They are extremely common, with an estimated lifetime risk in men of 27%. Occasionally they may present as emergencies with complications such as bowel incarceration, obstruction and strangulation. The definitive treatment of all hernias is surgical repair, inguinal hernia repair being one of the most common surgical procedures performed. Mesh (hernioplasty) and the traditional non‐mesh repairs (herniorrhaphy) are commonly used, with an increasing preference towards mesh repairs in high‐income countries.

Objectives

To evaluate the benefits and harms of different inguinal and femoral hernia repair techniques in adults, specifically comparing closure with mesh versus without mesh. Outcomes include hernia recurrence, complications (including neurovascular or visceral injury, haematoma, seroma, testicular injury, infection, postoperative pain), mortality, duration of operation, postoperative hospital stay and time to return to activities of daily living.

Search methods

We searched the following databases on 9 May 2018: Cochrane Colorectal Cancer Group Specialized Register, Cochrane Central Register of Controlled Trials (Issue 1), Ovid MEDLINE (from 1950), Ovid Embase (from 1974) and Web of Science (from 1900). Furthermore, we checked the WHO International Clinical Trials Registry Platform (ICTRP) and ClinicalTrials.gov for trials. We applied no language or publication restrictions. We also searched the reference lists of included trials and review articles.

Selection criteria

We included randomised controlled trials of mesh compared to non‐mesh inguinal or femoral hernia repairs in adults over the age of 18 years.

Data collection and analysis

We used standard methodological procedures expected by Cochrane. Where available, we collected information on adverse effects. We presented dichotomous data as risk ratios, and where possible we calculated the number needed to treat for an additional beneficial outcome (NNTB). We presented continuous data as mean difference. Analysis of missing data was based on intention‐to‐treat principles, and we assessed heterogeneity using an evaluation of clinical and methodological diversity, Chi2 test and I2 statistic. We used GRADE to assess the quality of evidence for each outcome.

Main results

We included 25 studies (6293 participants) in this review. All included studies specified inguinal hernias, and two studies reported that femoral hernias were included.

Mesh repair probably reduces the risk of hernia recurrence compared to non‐mesh repair (21 studies, 5575 participants; RR 0.46, 95% CI 0.26 to 0.80, I2 = 44%, moderate‐quality evidence). In absolute numbers, one hernia recurrence was prevented for every 46 mesh repairs compared with non‐mesh repairs. Twenty‐four studies (6293 participants) assessed a wide range of complications with varying follow‐up times. Neurovascular and visceral injuries were more common in non‐mesh repair groups (RR 0.61, 95% CI 0.49 to 0.76, I2 = 0%, NNTB = 22, high‐quality evidence). Wound infection was found slightly more commonly in the mesh group (20 studies, 4540 participants; RR 1.29, 95% CI 0.89 to 1.86, I2 = 0%, NNTB = 200, low‐quality evidence). Mesh repair reduced the risk of haematoma compared to non‐mesh repair (15 studies, 3773 participants; RR 0.88, 95% CI 0.68 to 1.13, I2 = 0%, NNTB = 143, low‐quality evidence). Seromas probably occur more frequently with mesh repair than with non‐mesh repair (14 studies, 2640 participants; RR 1.63, 95% CI 1.03 to 2.59, I2 = 0%, NNTB = 72, moderate‐quality evidence), as does wound swelling (two studies, 388 participants; RR 4.56, 95% CI 1.02 to 20.48, I2 = 33%, NNTB = 72, moderate‐quality evidence). The comparative effect on wound dehiscence is uncertain due to wide confidence intervals (two studies, 329 participants; RR 0.55, 95% CI 0.12 to 2.48, I2 = 37% NNTB = 77, low‐quality evidence). Testicular complications showed nearly equivocal results; they probably occurred slightly more often in the mesh group however the confidence interval around the effect was wide (14 studies, 3741 participants; RR 1.06, 95% CI 0.63 to 1.76, I2 = 0%, NNTB = 2000, low‐quality evidence). Mesh reduced the risk of postoperative urinary retention compared to non‐mesh (eight studies, 1539 participants; RR 0.53, 95% CI 0.38 to 0.73, I2 = 56%, NNTB = 16, moderate‐quality evidence).

Postoperative and chronic pain could not be compared due to variations in measurement methods and follow‐up time (low‐quality evidence).

No deaths occurred during the follow‐up periods reported in the seven studies (2546 participants) reporting this outcome (high‐quality evidence).

The average operating time was longer for non‐mesh repairs by a mean of 4 minutes 22 seconds, despite wide variation across the studies regarding size and direction of effect, thus this result is uncertain (20 studies, 4148 participants; 95% CI ‐6.85 to ‐1.60, I2= 97%, very low‐quality evidence). Hospital stay may be shorter with mesh repair, by 0.6 days (12 studies, 2966 participants; 95% CI ‐0.86 to ‐0.34, I2 = 98%, low‐quality evidence), and participants undergoing mesh repairs may return to normal activities of daily living a mean of 2.87 days sooner than those with non‐mesh repair (10 studies, 3183 participants; 95% CI ‐4.42 to ‐1.32, I2 = 96%, low‐quality evidence), although the results of both these outcomes are also limited by wide variation in the size and direction of effect across the studies.

Authors' conclusions

Mesh and non‐mesh repairs are effective surgical approaches in treating hernias, each demonstrating benefits in different areas. Compared to non‐mesh repairs, mesh repairs probably reduce the rate of hernia recurrence, and reduce visceral or neurovascular injuries, making mesh repair a common repair approach. Mesh repairs may result in a reduced length of hospital stay and time to return to activities of daily living, but these results are uncertain due to variation in the results of the studies. Non‐mesh repair is less likely to cause seroma formation and has been favoured in low‐income countries due to low cost and reduced availability of mesh materials. Risk of bias in the included studies was low to moderate and generally handled well by study authors, with attention to details of allocation, blinding, attrition and reporting.

PICOs

Population
Intervention
Comparison
Outcome

The PICO model is widely used and taught in evidence-based health care as a strategy for formulating questions and search strategies and for characterizing clinical studies or meta-analyses. PICO stands for four different potential components of a clinical question: Patient, Population or Problem; Intervention; Comparison; Outcome.

See more on using PICO in the Cochrane Handbook.

Comparing surgical groin hernia repair performed with or without mesh

Review question

This review assessed the difference in outcomes between surgical hernia repair with and without mesh.

Background

Hernias are out‐pouchings of an organ through the body wall that normally contains it; in this review, we refer to the bowel or its surrounding fatty tissues protruding through the abdominal wall in the groin region. This is a very common medical problem, affecting 27 out of every 100 men. These hernias can cause significant discomfort, and can occasionally become so tightly stuck that the blood supply can be cut off (strangulation), requiring emergency surgery. The curative treatment of hernias is surgical repair, which can be closed with sutured techniques (non‐mesh repair) or with a fine mesh to promote tissue growth to strengthen the previously weak area (mesh repair). Mesh repair is becoming increasingly popular in many countries, particularly in conjunction with laparoscopic (key‐hole) surgery.

Search date

We searched a number of databases for studies; this search was last updated on 9 May 2018.

Study characteristics

In this update of a review originally published in 2001, we included a total of 25 studies (with a total of 6293 people) undertaken in a number of different countries. A variety of outcomes were assessed, including return of the hernia after initial repair (hernia recurrence), a variety of complications including pain, duration of surgery, hospital stay and time before going back to normal activities.

Key results

One hernia recurrence is prevented for every 46 mesh repairs performed rather than non‐mesh repairs. Compared to non‐mesh repairs, mesh repairs are more likely to develop collections of fluid next to the surgical wound, but are less likely to result in difficulty urinating following the operation, or injury to nerves, blood vessels or other organs. Postoperative pain could not be clearly compared between studies due to differences in measurement methods and time frames, but overall the studies appeared to indicate that participants who had mesh repairs had less pain. The length of the surgical operation was slightly shorter for mesh repairs. Participants who had a mesh repair were more likely to have a shorter hospital stay and had a shorter average recovery time before returning to their normal activities.

Quality of the evidence

The studies included in this review used good‐quality methods, considered potential factors which could affect the results, and addressed their proposed outcomes clearly. In our assessment of the quality of evidence, we marked down some outcomes to 'moderate' quality, particularly due to variability within results.

Conclusions

Overall, hernia repairs with and without mesh both proved effective in the treatment of hernias, although mesh repairs demonstrated fewer hernia recurrences, a shorter operation time and faster return to normal activities. Non‐mesh repairs are still widely used, often due to the cost and poor availability of the mesh product itself.

Authors' conclusions

available in

Implications for practice

This review found that mesh inguinal hernia repairs may be associated with better outcomes than non‐mesh repairs in the elective setting (including for outcomes of recurrence, surgical complications, duration of surgery, postoperative stay, time to return to activities of daily living), however non‐mesh repairs remain a viable alternative depending on health service circumstances. The average effect on operative time, hospital stay and return to daily activities is uncertain due to wide variation between the results of the studies. Risk of bias in the included studies was low to moderate and generally handled well by study authors, with attention given to details of allocation, blinding, attrition and reporting.

The additional cost of a mesh in surgical repair (whilst possibly nullified by the shorter operative times and associated costs in the high‐income country) remains a significant and sometimes prohibitive cost in low‐income countries. There was not enough information to provide a conclusion regarding postoperative pain and mortality outcomes.

Implications for research

A number of gaps exist in the evidence available to date. As demonstrated in the analysis above, there were no statistically significant differences between mesh and non‐mesh repairs for complications including haematoma, wound dehiscence, wound infection, and testicular complications. Additional studies examining these outcomes would help differentiate the two interventions in this respect.

Assessment of postoperative pain was inconsistent amongst included studies and made meaningful analysis impossible. This was primarily due to the plethora of pain assessment techniques used in clinical practice worldwide. Selection of an international standard for pain assessment postoperatively for hernia repair would eliminate this barrier to analysis.

Whilst the included studies were largely sound in their methodology, the various inconsistencies and sources of bias outlined above limited the quality of the subsequent analysis. The inability to blind all investigators — specifically the surgical team — creates a consistent, significant source of bias in these studies. Due to the nature of the intervention, double‐blinding is an unrealistic expectation; however some modifications could be made to attempt to safeguard the blinding process, including having blinded assessors for postoperative reviews and ensuring that all participants at least are blinded to the intervention. Follow‐up time, randomisation techniques, outcome reporting, and complication reporting are all sources of inconsistency between trials. Future studies should focus on rigorous methodological implementation, and extensive reporting of outcomes and complications to improve the quality and comparability of studies. Additionally, studies are inevitably inconsistent when it comes to quantification of qualitative outcomes and assessment of their clinical significance. Identification of the most suitable quantification technique by comparison to previous studies is recommended.

This review does not have the scope to analyse cost reliably, as an economic modelling review would, and costs of the different hernia repairs — including further investigation into cheaper materials — would be useful.

Summary of findings

Open in table viewer
Summary of findings for the main comparison. Mesh compared to non‐mesh repair for inguinal and femoral hernia repair

Mesh compared to non‐mesh repair for inguinal and femoral hernia repair

Patient or population: adults undergoing inguinal and femoral hernia repair
Setting: multiple hospitals from small to large tertiary centres contributed results
Intervention: mesh repair
Comparison: non‐mesh repair

Outcomes

Anticipated absolute effects* (95% CI)

Relative effect
(95% CI)

№ of participants
(studies)

Certainty of the evidence
(GRADE)

Comments

Risk with non‐mesh repair

Risk with mesh

Hernia recurrence

Study population

RR 0.46
(0.26 to 0.80)

5575
(21 RCTs)

⊕⊕⊕⊝
MODERATE 1

In the context of surgical intervention, double blinding is difficult to achieve but most studies at least attempted single blinding.

Follow‐up: up to 5 years

4 per 100

2 per 100
(1 to 3)

Surgical complications ‐ neurovascular or visceral injury

Study population

RR 0.61
(0.49 to 0.76)

6293
(24 RCTs)

⊕⊕⊕⊕
HIGH

Follow‐up: up to 4.3 years

6 per 100

4 per 100

(3 to 5)

Surgical complications ‐ wound infection

Study population

RR 1.29
(0.89 to 1.86)

4540
(21 RCTs)

⊕⊕⊝⊝
LOW 2

Follow‐up: up to 5 years

2 per 100

3 per 100

(2 to 4)

Surgical complications ‐ wound dehiscence

Study population

RR 0.55
(0.12 to 2.48)

329
(2 RCTs)

⊕⊕⊝⊝
LOW 2

Follow‐up: up to 3 years

2 per 100

1 per 100

(0 to 6)

Mortality (within 30 days post‐surgery)

There were no reported events of mortality within 30 days so this outcome could not be compared.

However it can be concluded that both groups have very low rates of postoperative mortality.

Follow‐up on mortality: up to 5 years

0 per 100

0 per 100
(0 to 0)

Duration of surgery

(minutes)

The mean duration of surgery ranged from 10 to 94 minutes

MD 4.22 minutes lower
(6.85 lower to 1.6 lower)

4148
(20 RCTs)

⊕⊝⊝⊝
VERY LOW 3

The large degree of heterogeneity is likely to be related to variation in surgeon skill and familiarity with the intervention.

Duration of postoperative stay

(days)

The mean duration of postoperative stay ranged from 0.27 to 7.6 days

MD 0.6 days lower
(0.86 lower to 0.34 lower)

2966
(12 RCTs)

⊕⊕⊝⊝
LOW 4

Time to return to full ADLs (days)

The mean time to return to full ADLs ranged from 2.06 to 26 days

MD 2.87 days lower
(4.42 lower to 1.32 lower)

3183
(10 RCTs)

⊕⊕⊝⊝
LOW 4

Conversion from laparoscopic to open approach

No studies reported any conversion from laparoscopic to open technique where laparoscopic technique was used for mesh repair.

1680
(5 RCTs)

⊕⊕⊝⊝
LOW 5

The only laparoscopic techniques used for comparison were mesh repairs, and none of these studies reported conversion to open repair so this could not be compared.

*The risk in the intervention group (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI).

CI: Confidence interval; RR: Risk ratio; OR: Odds ratio; ADLs: activities of daily living; RCT: randomised controlled trial; MD: mean difference

GRADE Working Group grades of evidence
High certainty: We are very confident that the true effect lies close to that of the estimate of the effect
Moderate certainty: We are moderately confident in the effect estimate: The true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different
Low certainty: Our confidence in the effect estimate is limited: The true effect may be substantially different from the estimate of the effect
Very low certainty: We have very little confidence in the effect estimate: The true effect is likely to be substantially different from the estimate of effect

1 Downgraded one level for inconsistency (moderate heterogeneity)

2 Downgraded two levels for imprecision (wide confidence interval overlapping no effect) and inconsistency (substantial heterogeneity)

3 Downgraded three levels for inconsistency (considerable heterogeneity) and imprecision (wide confidence interval overlapping no effect)

4 Downgraded two levels for inconsistency (considerable heterogeneity)

5 Downgraded two levels for risk of bias and imprecision due to low event rate (not reported)

Background

available in

Description of the condition

A hernia is defined as a protrusion of an organ or part of an organ through the body wall that normally contains it (Brooks 2014a). Abdominal wall hernias are common, with a prevalence in the general population of 4% for those aged over 45 years (Jenkins 2008). Inguinal and femoral hernias are known collectively as groin hernias (Brooks 2014a). Of all groin hernias, 96% are inguinal and 4% are femoral (Rutkow 1993). Men are eight times more likely to develop a groin hernia than women and 20 times more likely to require a groin hernia repair (Brooks 2014a).

It is a very common problem, with an estimated lifetime risk of groin hernia of 27% in men (3% in women) (Fitzgibbons 2015). There is still very limited evidence regarding prevalence, particularly in low‐income countries, but hernia repair is an extremely common general surgical procedure.

Groin hernias may present as a heaviness or discomfort in the groin region, or a visible or palpable bulge. Discomfort is usually most pronounced when intra‐abdominal pressure is increased, for example with heavy lifting, straining or prolonged standing. Risk factors for groin hernias include history of hernia or prior hernia repair, older age, male sex, chronic cough, chronic constipation, abdominal wall injury, smoking and family history of hernia (Brooks 2014a). The current literature seems to support the view that obesity may be a protective factor for groin hernias (Liem 1997; Rosemar 2008; Ruhl 2007).

Inguinal hernias are the most common type of hernia in both genders, accounting for 75% of all abdominal wall hernias, with a lifetime risk of 27% in men and 3% in women (Jenkins 2008). Inguinal hernias are classified as congenital or acquired (Brooks 2014a). Congenital inguinal hernias are caused by a failure of the processus vaginalis (invagination of the parietal peritoneum that precedes the migration and descent of the testes in males) to close. The portion of the processus vaginalis within the inguinal canal is called the 'canal of Nuck' in females and usually obliterates by the eighth foetal month of life (Brooks 2014a). This difference in development, in addition to the protective position of the round ligament, accounts for the far lower rates of inguinal (particularly indirect) hernias in women compared to men. In contrast, acquired hernias are due to the weakening or disruption of the fibromuscular tissues of the abdominal wall, allowing the protrusion of intra‐abdominal contents through the acquired defect (Brooks 2014a). This may be facilitated by inherent connective tissue abnormalities, chronic abdominal wall injury (including any chronic increased intra‐abdominal pressure) and possibly adverse effects of drugs such as glucocorticoids (thinning of skin and weakening of soft tissues) or smoking (Brooks 2014a; Cannon 1981; Sorenson 2002). Acquired hernias can present acutely and may require emergency surgical intervention.

Inguinal hernias are further classified as indirect or direct. Indirect inguinal hernias protrude through the internal inguinal ring, which is the site where the spermatic cord in males and the round ligament in females exits the abdomen (Brooks 2014a). Direct inguinal hernias protrude medial to the inferior epigastric vessels within Hesselbach's triangle (formed by the inguinal ligament inferiorly, the inferior epigastric vessels laterally and the rectus abdominis muscle medially) (Brooks 2014a). Femoral hernias are located inferior to the inguinal ligament and protrude through the femoral ring, medial to the femoral sheath containing the femoral artery and vein. Femoral hernias are acquired; the femoral ring can widen with age or injury (Brooks 2014a). Femoral hernias represent 20% to 31% of repairs in women compared to only 1% in men (Brooks 2014a).

It is important to differentiate femoral hernias from inguinal hernias given that femoral hernias are more likely to strangulate. As women are more likely to have femoral hernias than men, a relatively high proportion of women who present acutely with a symptomatic acquired hernia will require emergency management compared to their male counterparts. (Dahlstrand 2009; Koch 2005; Rosemar 2010). Classically, femoral hernias present as mildly painful non‐reducible groin lumps located inferolateral to the pubic tubercle; inguinal hernias are generally found superolaterally to the pubic tubercle (Whalen 2011). However, femoral hernias tend to move above the inguinal ligament, where they may be mistaken for an inguinal hernia. Differentiation on clinical grounds is notoriously unreliable, and unrelated to the experience of the examining practitioner (Whalen 2011). Ultrasonography, computed tomography or even diagnostic laparoscopy may have a role in further investigation of hernia type or occult hernia (Brooks 2014a; Whalen 2011).

Groin hernias may present as emergencies, with complications such as bowel incarceration and obstruction or strangulation (Brooks 2014a). Incarceration refers to the irreducible trapping of hernia contents within the hernia sac. Reduced venous and lymphatic flow leads to swelling of the incarcerated tissue, which can lead to impediment of arterial supply resulting in ischaemia and necrosis of the hernia contents (strangulation). The overall risk of incarceration and strangulation is low, between 0.3% and 3% per year (Brooks 2014a; Fitzgibbons 2006; Gallegos 1991).

Description of the intervention

The definitive treatment of all hernias is surgical repair, regardless of hernia origin or type. Repair of inguinal hernias is one of the most common general surgical procedures performed (McCormack 2003; Rutkow 1993). Urgent surgical repair is indicated for patients who develop complications. If this is undertaken within approximately four to six hours from onset of symptoms, an emergency surgical repair may prevent loss of bowel from prolonged strangulation (Brooks 2014b). However, for uncomplicated hernias, the optimal timing of repair and aspects of repair technique remain controversial. Currently it is recommended that patients with symptomatic hernias should undergo an elective hernia repair. For patients who are asymptomatic but have risk factors for groin hernia incarceration or strangulation, a hernia repair is generally undertaken as soon as is feasible (Brooks 2014b). For male patients with minimal symptoms, where a 'watchful waiting' approach is taken, the cumulative probability of developing problems such as increasing pain, incarceration or strangulation is 2.8% at three months, 4.5% to 23% at two years, and 31% at four years (Fitzgibbons 2006; Gallegos 1991; O'Dwyer 2006).

The aim of hernia repair surgery is not only to fix the current hernia defect, but also to reduce the risk of recurrence. Recurrence rates for primary hernia repair range from 0.5% to 15% depending on the hernia site, type of repair and clinical circumstances (Brooks 2014b).

Groin hernia repairs can involve the use of a mesh (otherwise known as a hernioplasty) or no mesh (that is, herniorrhaphy). The mesh used in hernia repair is typically made from a synthetic polymer, usually polypropylene, which is inert and does not cause abnormal inflammation. The mesh is lightweight and flexible, and designed to avoid impediment of local structures or positional movement. Meshes may be held in place using partially dissolvable sutures or a fibrin glue (or both), of which the glue may produce a more effective seal (Brooks 2014b). A mesh repair involves covering the hernial defect by placing the mesh on one of the layers of the abdominal wall either using an open approach or a minimal access laparoscopic technique (McCormack 2003). The approach to repair depends on a number of factors in each individual case, including the type of hernial defect, patient factors and the surgeon's preference. With the open approach, the repair is generally anterior to the hernial defect, whereas laparoscopic repair is approached from a posterior aspect. Prosthetic mesh is being increasingly incorporated into hernia surgery (either open or laparoscopic) as a component of tension‐free repair (Brooks 2014b).

Open techniques for inguinal hernia repairs include tension‐free mesh repairs such as the Lichtenstein, plug and patch, and Kugel (preperitoneal) repairs, and non‐mesh primary tissue approximation repairs such as the Shouldice, Bassini and McVay repairs (Brooks 2014b). In the tension‐free mesh repair category, the mesh is placed in front of the transversalis fascia, such as with the Lichtenstein tension‐free hernioplasty, or behind the transversalis fascia in the preperitoneal space, for example, the Kugel procedure (Amid 2005). With the tissue approximation repairs, which do not involve mesh placement, the Shouldice technique is generally the preferred suture‐based repair, which involves a four‐layer reconstruction of the fascia transversalis. Alternatives to the Shouldice technique include the original Bassini method, in which the edges of the defect are simply sewn back together with tension and, less commonly, the McVay method. The McVay method involves reinforcement of the inguinal canal by approximating the transversus abdominis aponeurosis and transversalis fascia to the pectineal ligament, thus restoring the canal floor by bringing together the femoral sheath and the inguinal ligament. It is important to note that the McVay style of repair is also typically used in open femoral hernia repairs, with possible approaches from an infra‐inguinal (Lockwood), trans‐inguinal (Lotheissen) or supra‐inguinal (McEvedy) aspect (Amid 2005).

The two main laparoscopic groin hernia repairs are the totally extraperitoneal (TEP) and transabdominal preperitoneal patch (TAPP) repairs (Bittner 2011), both requiring the use of a mesh. TEP repair is performed by gaining access to the preperitoneal space (that is, the space between the peritoneum and the anterior abdominal wall) using an anterior approach, without ever actually entering the abdomen (Ferzli 1998; McKernan 1993). A TAPP repair, on the other hand, requires the surgeon to enter the peritoneal (abdominal) cavity to access the preperitoneal space. Some of the more significant disadvantages of this approach include potential injury to adjacent organs and, long‐term, adhesions resulting in bowel obstruction (Wake 2005; Vader 1997).

Common early complications of hernia repair surgery include wound seroma or haematoma, urinary retention, bladder injury and superficial wound infection. Complications that may occur later following hernia repair surgery include persistent groin pain and post‐herniorrhaphy neuralgia, testicular complications, deep wound or mesh infection, recurrent hernia and mesh migration or erosion (Brooks 2014c). The incidence of post‐surgical complications is more common following emergent repairs and recurrent hernia repairs (Brooks 2014c).

A Cochrane Review published in 2008 showed that whilst laparoscopic repairs were associated with quicker recovery times and less persistent pain, the procedure itself usually takes longer and has higher rates of bladder and vascular injuries. Hernia recurrence after laparoscopic mesh repair was less common compared to open non‐mesh repair, with the main indicator of recurrence related to the use of a mesh rather than the approach itself (McCormack 2003).

How the intervention might work

Prior to 1958, abdominal wall hernias were closed with primary suture repair. Tension on the weak fascia was thought to be one of the main contributing factors to repair failure. Then, in 1958, Dr Francis Usher published his 'tension‐free' technique using a permanently implanted polypropylene mesh (Read 1999). This led to the Lichtenstein repair some 30 years later, which popularised mesh for hernia repair. The logic that Usher used to explain his use of polypropylene mesh was that the mesh was a material that could be used to close over the hernial defect and provide ongoing reinforcement to the attenuated fascia of the abdominal wall by encouraging growth of connective tissue (scar tissue) around and through the mesh fibres (Doctor 2006). It was expected that the best meshes would be those made of very strong material and able to induce the most fibrosis. Unfortunately, this fibrotic reaction led to pain and movement restriction and it soon became clear that this needed to be reduced. In order to do this, the surface area (and therefore strength) of the mesh had to be reduced. Calculations of intra‐abdominal pressures proved that this would be possible without compromising mesh function. In fact, the tensile strength of a mesh required to withstand the maximum abdominal pressure is only a tenth of that of most meshes (Brown 2010). This realisation led to the concept of lightweight meshes.

Lightweight and heavyweight meshes have been used in the repair of hernias. Compared to their heavyweight counterparts, lightweight meshes have large pores (normally 3 mm to 5 mm) and a small surface area. They stimulate a reduced inflammatory reaction and, therefore, have greater elasticity and flexibility (Klinge 2008). They also shrink less and have been shown to cause less pain compared to heavier meshes after Lichtenstein inguinal hernia repairs. Unfortunately, despite these improvements, patients continue to have complications such as recurrence, infection and adhesion formation (Brown 2010). Lightweight mesh may not increase the risk of inguinal hernia recurrence and seems to be associated with the reduced risk of developing chronic pain complications, however, these outcomes continue to be researched in studies with longer follow‐up times (Sajid 2013). Thus, the search for an ideal mesh continued.

The difficulty of finding a single, 'ideal' mesh is acknowledged by the development of composite meshes. These combine more than one material and are the basis of most new mesh designs. The main advantage of the composite meshes is that they can be used in the intraperitoneal space with minimal adhesion formation. Despite the vast selection of brands available, nearly all these meshes continue to use one of the three basic materials: polypropylene, polyester and expanded polytetraflouroethylene (ePTFE). These are used in combination with each other or with a range of additional materials such as titanium, omega 3, poliglecaprone 25 (Monocryl), polyvinylidene difluoride (PVDF) and hyaluronate. However, as might be expected, none of these synthetic materials are without disadvantages (O'Dwyer 2005).

The problems encountered with synthetic materials led to the development of bio‐materials, which are currently the most physiologically based implants. These consist of an acellular collagen matrix derived from human dermis or porcine small intestine submucosa. The matrix allows soft tissue to infiltrate the mesh which eventually becomes integrated into the body by a process of remodelling. Unfortunately, this process also appears to lead to a rapid reduction in their mechanical strength, and concerns regarding this have restricted their use to infected environments (where one would normally use an absorbable synthetic material such as polyglactin 910 (Vicryl)) (Brown 2010).

Why it is important to do this review

Currently, about one million meshes are used per year for hernia repairs globally (Klinge 2002). In 2002, the EU Hernia Trialists Collaboration analysed 58 randomised controlled trials and found that the use of mesh was superior to other techniques; in particular, the meta‐analyses noted fewer recurrences and less postoperative pain with mesh repair compared to all other techniques (EU Hernia Trialists 2002). Despite the favourable results of mesh repair and its adoption as common practice in high‐income countries, it has yet to be integrated as standard practice by all surgeons (Nixon 2009). Non‐mesh repairs are still commonly performed worldwide, particularly in low‐income countries; for example, in some African countries, when surgical treatment is provided (65% to 75% as an emergency procedure rather than elective), fewer than 5% of hernias are repaired using implanted mesh (Yang 2011). This is likely to be related to the increased costs involved in mesh (and also laparoscopic) repair which can be unaffordable in countries where the typical per capita government health expenditure (USD 28 in Ghana, USD 7 in Uganda) is usually less than the price of a single‐use package of commercial mesh (USD 40 to USD 100) (Yang 2011). The biological meshes are available at an even higher cost (Klinge 2008). Research into low‐cost mesh alternatives (such as polyethylene mesh, normally used in mosquito‐netting) and innovative construction of standard commercial meshes are being undertaken and show promise (Lofgren 2016; Yang 2011).

An updated meta‐analysis of the current literature is needed regarding the use of mesh in inguinal and femoral hernia repairs. In clinical practice, laparoscopy is increasingly used as the operative approach of choice. To address this, this review will include an analysis of trials exploring the use of mesh in the context of all operative approaches. The previous Cochrane Review considering mesh versus non‐mesh repairs specified only open repair as an inclusion criterion, and did not include the increasingly favoured laparoscopic approach.

In this meta‐analysis, assessment of the overall quality of the evidence and risk of bias for each outcome will make the available evidence more transparent and allow clinicians to make better informed decisions.

Objectives

available in

To evaluate the benefits and harms of different inguinal and femoral hernia repair techniques in adults, specifically comparing closure with mesh versus without mesh. Outcomes include hernia recurrence, complications (including neurovascular or visceral injury, haematoma, seroma, testicular injury, infection, postoperative pain), mortality, duration of operation, postoperative hospital stay, time to return to activities of daily living and conversion from laparoscopic to open approach.

Methods

available in

Criteria for considering studies for this review

Types of studies

All individual parallel randomised controlled trials (RCTs) and cluster‐RCTs investigating mesh compared to non‐mesh techniques for open or laparoscopic repair of inguinal or femoral hernias were eligible for inclusion.

Types of participants

We included persons aged 18 years or older with a clinically diagnosed inguinal or femoral hernia, or both, where surgical management was indicated (in accordance with the preceding Cochrane Review).

Types of interventions

Concerning inguinal hernias, we accepted any of the following mainstream surgical techniques.

  • Mesh repairs

    • Open, including the Lichtenstein approach

    • Laparoscopic, including transabdominal preperitoneal (TAPP) and totally extraperitoneal (TEP) approaches

    • Any type of commercially marketed non‐absorbable mesh or absorbable biomesh may be used; including 'plug‐and‐patch' kits (that is, absorbable plugs/tacks with (non‐)absorbable patch/mesh).

  • Suture repairs

    • Tension, including Bassini, McVay and Shouldice approaches

    • Tension‐free, including Desarda and Guarnieri approaches

    • Any type of commercially marketed non‐absorbable or absorbable sutures may be used.

Concerning femoral hernias, we accepted any of the following mainstream surgical techniques.

  • Mesh repairs

    • Open mesh/mesh plug repair

    • Laparoscopic; including TAPP or TEP approaches

    • Any type of commercially marketed non‐absorbable mesh or absorbable biomesh may be used; including 'plug‐and‐patch' kits (that is, absorbable plugs/tacks with (non‐)absorbable patch/mesh).

  • Suture repairs

    • Open McVay suture repair, including Lockwood's infra‐inguinal, Lotheissen's trans‐inguinal and McEvedy's high approaches

    • Any type of commercially marketed non‐absorbable or absorbable sutures may be used.

Types of outcome measures

Primary outcomes

  • Recurrence of the same hernia (this excludes formation of a hernia at a new site not previously repaired or reinforced)

  • Surgical complications

    • Neurovascular or visceral injury

    • Wound infection at single or multiple sites, including deep and superficial wound infections

    • Haematoma

    • Seroma

    • Postoperative wound swelling

    • Wound dehiscence

    • Testicular injury or complications, including testicular swelling and atrophy

    • Urinary retention postoperatively

    • Postoperative pain, including acute and chronic pain

  • Mortality (number of associated deaths within 30 days of the operation during the study trial period)

Secondary outcomes

  • Duration of surgical operation (minutes)

  • Duration of postoperative hospital stay (days)

  • Time required to return to full activities of daily living including work and exercise (days)

  • Number of operations where conversion from laparoscopic to open approach was required

We will also present a brief narrative account of cost‐effectiveness of repair, as this is considered a clinically significant factor, particularly in low‐income countries, where this plays a large role in access to healthcare (see discussion).

Search methods for identification of studies

Electronic searches

We searched the following electronic databases, with no restrictions on language or date of publication.

  • Cochrane Central Register of Controlled Trials (CENTRAL Issue 4, May 9th, 2018) (Appendix 1),

  • Ovid MEDLINE (1950 to May 9th, 2018) (Appendix 2),

  • Ovid Embase (1974 to May 9th, 2018) (Appendix 3)

  • Web of Science (1900 to May 9th, 2018) (Appendix 4).

The search was run separately on each database and results were checked against each other.

Searching other resources

We searched relevant clinical trials registers such as the World Health Organization (WHO) International Clinical Trials Registry Platform (ICTRP) (http://www.who.int/ictrp/en/) and ClinicalTrials.gov (http://clinicaltrials.gov/) May 9th 2018 for completed and ongoing trials.

In addition to this, we searched reference lists of included trials and review articles, books related to surgical hernia repairs, abstracts from general surgical conferences concerning hernias and mesh repair, and sent written enquiries to the authors of major relevant studies and experts in the field.

Data collection and analysis

Selection of studies

Three review authors (KL, ET, ST) assessed titles and abstracts retrieved from the search to determine their relevance concerning the objectives of this review. We managed disagreements through discussion, a deciding arbiter (MD), or both. We entered all search results into Review Manager 5.3 (RevMan 2014).

Data extraction and management

Two review authors (KL, ET) designed a data extraction sheet for study reports, which was pilot tested using sample studies and revised by the other authors (MD, MVD). Onto this report, two authors (KL, ST) independently extracted and recorded key features of each study including details of the following.

  • Methods: study design, total duration of study and run in, number of study centres and location, study setting, withdrawals, date of study

  • Participants: number, mean age, age range, gender, severity of condition, diagnostic criteria, baseline lung function, smoking history, inclusion criteria, exclusion criteria

  • Interventions: intervention, comparison, concomitant medications, excluded medications

  • Outcomes: primary and secondary outcomes specified and collected, time points reported. A brief narrative account of cost effectiveness included in discussion.

  • Notes: funding for trial, notable conflicts of interest of trial authors

We managed disagreements through discussion, a deciding arbiter (MVD) or both. We entered and presented the data for each included study into a table in Review Manager 5.3 (RevMan 2014).

Assessment of risk of bias in included studies

Two review authors (KL, ST) independently analysed each study in conjunction with the Cochrane tool for assessing risk of bias (Higgins 2011) (Appendix 5). This approach uses a domain‐based evaluation that aims to address main potential areas of bias in studies, where a judgement (low, high or unclear risk of bias) is assigned for each of the following domains.

  • Random sequence generation (low risk if true random sequence generation was described)

  • Allocation concealment (low risk if sealed, opaque, numbered envelopes were used, or central allocation after registration)

  • Blinding of participants and personnel (risk of performance bias; low risk of both participant and personnel were blinded to intervention)

  • Blinding of outcome assessment (risk of detection bias; low risk if both the the assessors were blinded to intervention)

  • Incomplete or selective outcome data reporting (low risk if more than 80% of those randomly assigned were assessed)

  • Any other potential sources of bias (e.g. study stopped early because of a data‐dependent process, notable baseline imbalance, surgeon competence or experience).

A high risk of bias indicates that the study design has not met the criteria for a low risk classification as noted above for each of the respective domains. Similarly, an unclear risk of bias denotes that the study has not declared sufficient information regarding their study design to make a judgement. We managed any disagreements through discussion, a deciding arbiter (MVD), or both. We presented our assessment of risk of bias for the included studies in the 'Risk of bias' summary tables and graphs as generated through input into Review Manager 5.3 (RevMan 2014).

Measures of treatment effect

We presented dichotomous (binary) data as a measure of risk by using a risk ratio (RR) with 95% confidence intervals (CIs). Where possible, we calculated the absolute risk reduction (ARR) and number needed to treat to benefit (NNTB) or number needed to treat to harm (NNTH) for comparison against other treatments or non‐treatment.

We presented continuous data as a mean difference (MD) if the same scale was used. Alternatively, a standardised mean difference (SMD) was calculated (that is, an average of the combined standard deviations) in the event that each study used a different scale measuring the same concept. In this case, we assessed the impact of using the highest verses the lowest of the available standard deviations (SDs) on the overall estimate of effect. If SDs were not reported we estimated the SD based on similar studies and used this in the meta‐analysis (Higgins 2011)

The treatment effect was considered statistically significant if the P value was less than 0.05.

Unit of analysis issues

The participant is the unit of analysis in our review. Nonetheless:

  • if the unit of analysis was not the same as the unit of randomisation, such as in cluster‐randomised trials, we adjusted for clustering by using the guidance given in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011); and

  • if there were multiple measurements for the same participant (for example, multiple hernias in the same person), we analysed the data as in a cluster‐randomised trial.

Dealing with missing data

We contacted the trial authors of the original studies if further data or information was required. We performed analyses based on intention‐to‐treat (ITT) principles, whereby the missing data for randomised participants was assumed to be treatment failures in this review. However, this approach of ITT analysis (that is, assuming dropouts as failures) may underestimate the effect of the intervention, therefore we performed both ITT and on‐treatment (that is, non‐ITT) analyses to explore the impact of missing data on the overall outcome (Higgins 2011). Furthermore, for continuous data we assessed the impact of missing data on the overall estimate of effect by imputing missing data in the following ways: best‐case scenario, where the missing data were considered 2 SDs greater in the intervention arm than in the control arm; and worst‐case scenario, where the missing data were considered 2 SDs less than in the control arm.

Assessment of heterogeneity

We assessed the included studies for heterogeneity through three successive steps to determine if they should be pooled with the rest of the included studies or considered separately.

  • Two review authors (KL, JN) independently analysed the included studies for their heterogeneity, including the extent of clinical diversity (participants, interventions and outcomes), and methodological diversity (study design and risk of bias).

  • We then assessed the included studies for statistical heterogeneity using the Chi2 test, with a P value of less than 0.10 being statistically significant.

  • We then calculated the I2 statistic as instructed in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011), where 0% to 40% is likely to indicate minimal heterogeneity, 30% to 60% may represent moderate heterogeneity, 60% to 90% may represent substantial heterogeneity, and 90% to 100% may represent considerably significant heterogeneity. The importance of the observed value of I2 does depend on the magnitude and direction of the treatment effects, and strength of evidence for heterogeneity (that is, the P value from the Chi2 test or the confidence interval for I2).

Assessment of reporting biases

Given that there was a sufficient number of studies pooled (more than 10), we performed a funnel plot to visually assess the risk of publication bias, where more pronounced asymmetry of the funnel plot may indicate a substantial overestimation of the intervention effect, as recommended in Chapter 10 of the Cochrane Handbook for Systematic Reviews of Interventions (Sterne 2011).

Data synthesis

We used the fixed‐effect model in the absence of statistical heterogeneity, according to the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011), where the analysis produced an estimate of the true effect.

In the event of statistical heterogeneity we used the random‐effects model for pooling the study data, using the Mantel‐Haenszel method, according to the Cochrane Handbook for Systematic Review of Interventions (Higgins 2011). We used an I2 of 40% or higher as a cut‐off for heterogeneity, i.e. we used a random effects model if the I2 was 40% or greater. In the event that there was an insufficient number of studies (less than two) to produce an average effect in a random‐effects model, a fixed‐effect model was used.

Where cluster‐RCTs were included, we used the generic inverse variance method (Higgins 2011).

We planned to use the Peto one‐step odds ratio method for meta‐analysis of rare events (event rates below 1%), according to the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011).

Subgroup analysis and investigation of heterogeneity

If sufficient data were available, we explored further the treatment effect in specific subgroups, including:

  • inguinal versus femoral hernia;

  • direct versus indirect inguinal hernia;

  • male versus female participants;

  • participants meeting the American Society of Anaesthesiology (ASA) criteria of 1 to 2 versus participants meeting the ASA criteria of 3 to 4*;

  • Body Mass Index (BMI) greater than or equal to 30 versus BMI less than 30**;

  • elective versus emergency surgery;

  • different types of mesh, for example, biological versus composite synthetics.

*American Society of Anaesthesiology (ASA) physical status classification of perioperative patients: 1 = healthy person; 2 = mild systemic disease; 3 = severe systemic disease; 4 = severe systemic disease that is a constant threat to life (Skalad 1941).

**Body Mass Index (BMI): those greater than or equal to 30 are classified as obese in accordance to WHO 2006.

Sensitivity analysis

If sufficient data were available, we performed the following sensitivity analyses.

  • In order to determine the impact of risk of bias on the overall effect estimate, we removed high risk of bias studies from the pooled analysis and compared the results. High risk of bias studies were assessed to be 'high risk' in at least one domain (including selection bias).

  • In order to determine the impact of heterogeneity on the overall estimate of effect, we removed studies that contributed to heterogeneity from the analyses and compared the results.

We used two different methods of pooling to test sensitivity: we pooled all studies together and then removed studies from the meta‐analysis one by one, noting if there was any significant change in the overall results; and simultaneously, we compared the use of a fixed‐effect versus random‐effects model for the pooling analysis as we excluded each study one by one.

GRADE and 'Summary of findings' table

We presented a 'Summary of findings' table including the following outcomes: hernia recurrence, complications, mortality, conversion from laparoscopic to open approach, duration of surgery, duration of postoperative stay and time to return to full activities of daily living.
We used the five GRADE considerations (study limitations, consistency of effect, imprecision, indirectness and publication bias) to assess the quality of evidence as it relates to the studies which contribute data to the meta‐analyses for these outcomes (Atkins 2004). We used methods and recommendations described in Section 8.5 and Chapter 12 of the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011) using GRADEpro GDT software (GRADEpro GDT 2014). We justified all decisions to down‐ or up‐grade the quality of studies using footnotes, and we made comments to aid the reader's understanding of the review where necessary. Please see Table 1 for more information regarding criteria for quality of evidence grading (according to the GRADE recommendations, Schünemann 2011).

Open in table viewer
Table 1. Quality of Evidence, GRADE definitions

Grade

Definition

High

We are very confident that the true effect lies close to that of the estimate of the effect.

Moderate

We are moderately confident in the effect estimate: The true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different.

Low

Our confidence in the effect estimate is limited: The true effect may be substantially different from the estimate of the effect.

Very Low

We have very little confidence in the effect estimate: The true effect is likely to be substantially different from the estimate of effect.

Results

Description of studies

Results of the search

We obtained 2006 records (of which 1076 duplicates were removed, leaving 930 records) from our searches of the respective databases, as described in Search methods for identification of studies. As noted in Selection of studies, we subsequently screened the titles and abstracts of 930 studies and retained 37 publications of randomised controlled trials (RCTs) which were considered for inclusion, of which 29 were initially included in this systematic review (see Figure 1 for study flow diagram). Of these 29 reports, we found that some presented results from the same studies at different time points or emphasised a different outcome. Therefore, we have included 25 studies in this review (all data available from each paper was included in our analysis).


Study flow diagram.

Study flow diagram.

Included studies

Twenty‐five RCTs met the inclusion criteria for this study, reviewing the effectiveness of mesh compared to non‐mesh hernia repairs. A total of 6293 participants (3289 given mesh repair and 3004 given non‐mesh repair) were included in the analysis, from a host of high and low‐income countries including: France, Lebanon, Iceland, Germany, India, Egypt, Belgium, Turkey, Uganda, the USA, Japan, Sweeden, Mexico, the UK, China, Netherlands, and Poland. The most common mesh repair used was Lichtenstein repair (Abd El Maksoud 2014; Barth 1998; Butters 2007; Chakraborty 2007; Elsebae 2008; Kaynak 2007; Kucuk 2010; Kux 1994; Manyilirah 2012; McGillicuddy 1998; Naveen 2014; Nordin 2002; Panda 2012; Prior 1998; van Veen 2007), whilst the most common non‐mesh repair was the Shouldice technique (Barth 1998; Berndsen 2007; Butters 2007; Hauters 1996; Kux 1994; Leibl 1995; Lermite 2012; McGillicuddy 1998; Nordin 2002; Prieto‐Diaz‐Chavez 2009; Schmitz 1997; van Veen 2007; Zieren 1998). Other mesh techniques noted were transabdominal preperitoneal patch (TAPP) repair (Berndsen 2007; Butters 2007; Hauters 1996; Leibl 1995; Zieren 1998), non‐specified 'tension‐free' mesh/plug repair (Lermite 2012; Prieto‐Diaz‐Chavez 2009; Schmitz 1997; Shi 2010), Prolene Hernia System mesh repair (Nakagawa 2013), Shulmann repair (van Veen 2007). Other non‐mesh techniques utilised included modified darn (Abd El Maksoud 2014), Abrahamson's darn (Chakraborty 2007), Bassini (Elsebae 2008; Naveen 2014; Panda 2012; Prior 1998; van Veen 2007; Witkowski 2000), Moloney darn (Kaynak 2007; Kucuk 2010), Desarda (Manyilirah 2012), Marcy (Nakagawa 2013) and McVay repair (van Veen 2007). The number of participants in each study ranged from 40 (Panda 2012), to 1183 (Berndsen 2007). Of the 25 studies, three were published in German (Kux 1994; Leibl 1995; Schmitz 1997), and one in French (Hauters 1996); all others were published in English. The most common outcomes studied included hernia recurrence, complications and postoperative pain. Most studies targeted participants with reducible, unilateral, primary hernias requiring elective repair.

Excluded studies

We excluded five studies; two did not meet the inclusion criteria of being a prospective randomised controlled trial (Bay‐Nielsen 2004; Chan 2008), and one did not meet inclusion criteria of being a randomised controlled trial (Suradom 2011). We excluded the remaining two studies as they compared only mesh versus mesh techniques (Abu‐Own 2000; Aigner 2014).

Risk of bias in included studies

Allocation

In the assessment of the risk of selection bias, for a study to qualify as low‐risk it had to describe an adequate system of random sequence generation, either via computer‐generated software or handling through an independent centre. Conversely, studies deemed high‐risk typically used allocation methods that were more transparent, with blinding risk. Overall, we assessed 13 studies as low‐risk (Abd El Maksoud 2014; Berndsen 2007; Chakraborty 2007; Manyilirah 2012; McGillicuddy 1998; Nakagawa 2013; Naveen 2014; Prieto‐Diaz‐Chavez 2009; Prior 1998; Shi 2010; van Veen 2007; Witkowski 2000; Zieren 1998) and Elsebae 2008 as high risk (participants were given a registration number upon enrolment in the trial and were assigned according to even/odd number), with the remaining studies classified as unclear.

Additionally, allocation concealment methods were assessed; acceptable methods included central randomisation or allocation with the use of sequentially numbered, opaque and sealed envelopes. Overall, we assessed ten studies as low‐risk (Abd El Maksoud 2014; Berndsen 2007; Chakraborty 2007; Hauters 1996; Manyilirah 2012; Nakagawa 2013; Nordin 2002; Prior 1998; van Veen 2007; Witkowski 2000), with the remaining studies classified as unclear.

Blinding

In the assessment of performance bias (blinding of participants and personnel) and detection bias (blinding of outcome assessment), it is recognised that given the nature of the intervention as a surgical procedure, true double‐blinding may be difficult to achieve. But given the subjective nature of certain outcomes — e.g. postoperative pain reported by participants and, to a lesser extent, clinical complications detected in follow‐up — there is potential for bias to occur if participants/personnel and outcome assessors are not blinded. Subsequently, studies assessed as having low risk of performance bias had to describe a sufficient method of blinding the participant; and to be assessed as having low risk of detection bias they had to have adequate blinding of the outcome assessor. When no method of blinding was outlined, we assigned a judgement of unclear risk of bias, and if blinding was unlikely to be achieved given the description of methods, these studies were assessed as high risk of bias.

Six of the included studies were described as low‐risk for performance bias (Barth 1998; Lermite 2012; Manyilirah 2012; Nakagawa 2013; Nordin 2002; Prior 1998) and one was found to be high risk, described as an open label study (Hauters 1996), with the remaining classified as unclear risk given lack of description regarding blinding methods.

Four studies were found to be at high risk of detection bias since the outcomes were assessed by surgeons who had performed the procedure, or were otherwise not blinded at the time of assessment (Abd El Maksoud 2014; Hauters 1996; Nakagawa 2013; Prieto‐Diaz‐Chavez 2009). We assessed five studies as being low risk of detection bias (Manyilirah 2012; Nordin 2002; Panda 2012; Prior 1998; van Veen 2007), and the remaining studies had an unclear risk of bias, as no explicit description of a blinding process could be interpreted from the text.

Incomplete outcome data

In the assessment of attrition bias, a study determined to be low‐risk had to demonstrate no missing outcome data, or if there were incomplete data with loss to follow‐up, it had to account for the missing data or perform an intention‐to‐treat analysis as applicable. Sixteen studies handled loss to follow‐up clearly, with an intention‐to‐treat model, and were determined to be at low risk of attrition bias (Barth 1998; Berndsen 2007; Butters 2007; Chakraborty 2007; Kaynak 2007; Leibl 1995; Lermite 2012; Manyilirah 2012; McGillicuddy 1998; Nakagawa 2013; Naveen 2014; Nordin 2002; Schmitz 1997; van Veen 2007; Witkowski 2000; Zieren 1998). A high‐risk study had to demonstrate missing outcome data or withdrawals/exclusions from the study, that may be directly related to an outcome, however we identified no studies in this risk group. We judged the remaining studies as having unclear risks of attrition bias, often due to a lack of reporting regarding dropouts and loss to follow‐up (Abd El Maksoud 2014; Elsebae 2008; Hauters 1996; Kaynak 2007; Kucuk 2010; Kux 1994; Panda 2012; Prieto‐Diaz‐Chavez 2009; Prior 1998; Shi 2010).

Selective reporting

In the assessment of reporting bias, a study deemed low‐risk was required to sufficiently report and discuss the primary or secondary outcomes explicitly outlined in their methodology or abstract. We assessed 22 studies as being low‐risk (Abd El Maksoud 2014; Barth 1998; Berndsen 2007; Butters 2007; Elsebae 2008; Kaynak 2007; Kucuk 2010; Kux 1994; Leibl 1995; Lermite 2012; Manyilirah 2012; McGillicuddy 1998; Nakagawa 2013; Naveen 2014; Nordin 2002; Panda 2012; Prieto‐Diaz‐Chavez 2009; Prior 1998; Schmitz 1997; Shi 2010; Witkowski 2000; Zieren 1998). In studies that were assessed as high‐risk, there may have been evidence through earlier versions or protocols, that outcomes initially intended to be analysed, were excluded in newer publications; or if there was suspicion that outcomes that were deemed to be particularly relevant, were being omitted. In one study identified as high‐risk (Chakraborty 2007), it was recognised that the study had been published in a low‐income nation and while the study portrayed one intervention as being more costly than another, cost had not been measured as an outcome, underlying a possible agenda with the potential to introduce bias. A second high‐risk study (van Veen 2007), intended to measure quality of life as an outcome in an earlier publications and protocols (van Veen 2007), but was not discussed in the final publication. Two other studies were at unclear risk of bias (Hauters 1996; Leibl 1995).

We created funnel plots for each outcome involving 10 or more studies, in order to assess for publication bias and consider other contributing factors of asymmetry. None of the studies measuring dichotomous outcomes clearly identified evidence of publication bias, resembling the classic symmetrical inverted funnel (see Figure 2 and Figure 3). The complications were assessed in funnel plots individually and none that included more than 10 studies demonstrated significant asymmetry. However, duration of surgery demonstrated a very wide spread, without any studies in the bottom half of the graph, which is potentially suggestive of publication bias (excluding smaller studies without statistically significant effects). This outcome also had significant heterogeneity which likely contributed to this appearance. The funnel plot for duration of postoperative stay was fairly symmetrical and resembled the classic shape, however the funnel plot for duration to return to full activities of daily living was similarly was 'top‐heavy' in appearance though much less wide‐spread. This is again in the context of a large amount of heterogeneity.


Funnel plot of comparison: mesh vs non‐mesh repair, primary outcome: hernia recurrence.

Funnel plot of comparison: mesh vs non‐mesh repair, primary outcome: hernia recurrence.


Funnel plot of comparison: mesh vs non‐mesh repair, primary outcome: complications.

Funnel plot of comparison: mesh vs non‐mesh repair, primary outcome: complications.

Other potential sources of bias

One of the included studies declared a commercial source of funding: Berndsen 2007 declared financial support from Ethicon Endosurgery, Johnson and Johnson companies, however in another publication of this study by Arvidsson and colleagues, the authors went on to clarify that Ethicon did not have any involvement in the design or conduct of the study, or data analysis. Another study, Manyilirah 2012, declared financial support from Makere University and Mulago National Referral and Teaching Hospital. All other papers declared no conflict of interest and did not describe any sources of funding.

Please refer to Figure 4 and Figure 5 for summaries of our 'Risk of bias' analysis of included studies.


Risk of bias graph: review authors' judgements about each 'Risk of bias' item presented as percentages across all included studies.

Risk of bias graph: review authors' judgements about each 'Risk of bias' item presented as percentages across all included studies.


Risk of bias summary: review authors' judgements about each 'Risk of bias' item for each included study.

Risk of bias summary: review authors' judgements about each 'Risk of bias' item for each included study.

Effects of interventions

See: Summary of findings for the main comparison Mesh compared to non‐mesh repair for inguinal and femoral hernia repair

Primary outcomes

Hernia recurrence

Twenty‐one studies were eligible for evaluation of hernia recurrence as a primary outcome (Abd El Maksoud 2014; Barth 1998; Berndsen 2007; Butters 2007; Chakraborty 2007; Leibl 1995; Elsebae 2008; Kux 1994; Hauters 1996; Kaynak 2007; Kucuk 2010; McGillicuddy 1998; Nakagawa 2013; Naveen 2014; Nordin 2002; Panda 2012; Prior 1998; Shi 2010; van Veen 2007; Witkowski 2000; Zieren 1998). Eight of these studies did not report recurrence in either mesh or non‐mesh group (Barth 1998; Chakraborty 2007; Leibl 1995; Kucuk 2010; Nakagawa 2013; Panda 2012; Prior 1998; Zieren 1998). On review of the remaining studies, in total there were 52/2834 events of hernia recurrence in the mesh repair group, compared to 110/2741 events of hernia recurrence in the non‐mesh repair group. We performed the analysis using a random‐effects model due to there being moderate heterogeneity (I2 = 44%). Meta‐analysis demonstrated a statistically significant difference in recurrence between the mesh and non‐mesh groups, with hernia recurrence occurring more frequently in the non‐mesh group (risk ratio (RR) 0.46, 95% confidence interval (CI) 0.26 to 0.80, moderate‐quality evidence). See Analysis 1.1 and Figure 6. The NNTB is 46; that is, one hernia recurrence was prevented for every 46 mesh repairs performed rather than non‐mesh repairs.


Forest plot of comparison: 1 Comparison: Mesh vs Non‐Mesh repair, outcome: 1.1 Primary Outcome: Hernia Recurrence.

Forest plot of comparison: 1 Comparison: Mesh vs Non‐Mesh repair, outcome: 1.1 Primary Outcome: Hernia Recurrence.

Using the GRADE working group grades of evidence (Schünemann 2011), we downgraded our assessment of the quality of the evidence by one level, to moderate quality, due to inconsistency (moderate heterogeneity); these results are outlined in the summary of findings Table for the main comparison. Included studies did not provide adequate data to enable a subgroup analysis of femoral and inguinal hernias (or did not include femoral hernias).

Sensitivity analysis

We conducted a sensitivity analysis to assess the impact of risk of bias on the overall outcome; this did not significantly change the overall estimate of effect (RR 0.54, 95% CI 0.32 to 0.93). After the studies demonstrating high risk of bias were removed, we deemed the quality of the evidence according to GRADE to have increased from moderate to high due to a reduction in heterogeneity (I2 = 39%). On further investigation we noted that one study, Berndsen 2007, contributed largely to the heterogeneity for this outcome; once this study was removed, the I2 reduced to 7%, although the RR remained consistent (RR 0.38, 95% CI 0.24 to 0.60).

Surgical complications

Twenty‐four studies reported complications as a study outcome (Abd El Maksoud 2014; Barth 1998; Berndsen 2007; Butters 2007; Chakraborty 2007; Elsebae 2008; Hauters 1996; Kaynak 2007; Kucuk 2010; Kux 1994; Leibl 1995; Lermite 2012; Manyilirah 2012; McGillicuddy 1998; Nakagawa 2013; Naveen 2014; Nordin 2002; Panda 2012; Prior 1998; Schmitz 1997; Shi 2010; van Veen 2007; Witkowski 2000; Zieren 1998). Serious operative complications were reported by all papers to be rare. Some of these included deep wound infection, neurovisceral damage and testicular damage/atrophy. Two studies (Lermite 2012; Prior 1998) did not specify which group sustained some specific listed complications, so part of their data was not included as it could not be assigned to a group. Paraesthesia, if assessed as a stand‐alone outcome, was included in the neurovascular injury group. Only one study, Kucuk 2010, addressed mesh rejection as an outcome (reported three events of rejection in the mesh group), and no subgroup analysis could be conducted as there was no available comparison. An overall analysis of complications was not performed in favour of subgroup analysis due to the risk of overlapping totals. Please refer to Figure 7, Analysis 1.2, Table 2 and Table 3 for details of analysis.


Forest plot of comparison: 1 Comparison 1: Mesh vs Non‐Mesh repair, outcome: 1.2 Primary Outcome: Complications.

Forest plot of comparison: 1 Comparison 1: Mesh vs Non‐Mesh repair, outcome: 1.2 Primary Outcome: Complications.

Open in table viewer
Table 2. Overview of complications reported in primary studies

Study

Group (Total number)

Neurovascular injury (including paraesthesia) or visceral injury

Wound infection

Haematoma

Seroma

Postoperative wound swelling

Wound dehiscence

Testicular injury or complications

Urinary retention

Total complications

Abd El Maksoud 2014

Mesh (119)

Not reported

8

Not reported

7

Not reported

Not reported

0

5

20

Non‐mesh (108)

Not reported

2

Not reported

4

Not reported

Not reported

1

6

13

Barth 1998

Mesh (54)

Not reported

Not reported

2

Not reported

Not reported

Not reported

Not reported

2

4

Non‐mesh (51)

Not reported

Not reported

1

Not reported

Not reported

Not reported

Not reported

0

1

Berndsen 2007

Mesh (538)

55

Not reported

49

Not reported

Not reported

Not reported

Not reported

Not reported

104

Non‐mesh (530)

107

Not reported

60

Not reported

Not reported

Not reported

Not reported

Not reported

167

Butters 2007

Mesh (187)

12

Not reported

Not reported

Not reported

Not reported

Not reported

0

Not reported

12

Non‐mesh (93)

9

Not reported

Not reported

Not reported

Not reported

Not reported

0

Not reported

9

Chakraborty 2007

Mesh (120)

0

10

Not reported

Not reported

Not reported

Not reported

5

Not reported

15

Non‐mesh (120)

0

5

Not reported

Not reported

Not reported

Not reported

5

Not reported

10

Elsebae 2008

Mesh (27)

Not reported

1

Not reported

1

Not reported

Not reported

Not reported

Not reported

2

Non‐mesh (27)

Not reported

3

Not reported

0

Not reported

Not reported

Not reported

Not reported

3

Hauters 1996

Mesh (35)

Not reported

0

Not reported

3

Not reported

Not reported

4

Not reported

7

Non‐mesh (35)

Not reported

1

Not reported

0

Not reported

Not reported

2

Not reported

3

Kaynak 2007

Mesh (354)

Not reported

5

7

8

Not reported

Not reported

0

Not reported

20

Non‐mesh (297)

Not reported

3

8

6

Not reported

Not reported

0

Not reported

17

Kucuk 2010

Mesh (130)

1

7

0

3

Not reported

Not reported

Not reported

Not reported

11

Non‐mesh (176)

1

9

2

3

Not reported

Not reported

Not reported

Not reported

15

Kux 1994

Mesh (102)

Not reported

2

2

Not reported

Not reported

Not reported

5

2

11

Non‐mesh (107)

Not reported

0

2

Not reported

Not reported

Not reported

3

34

39

Leibl 1995

Mesh (54)

Not reported

0

2

Not reported

Not reported

Not reported

0

2

4

Non‐mesh (48)

Not reported

0

1

Not reported

Not reported

Not reported

2

1

4

Lermite 2012

Mesh (156)

9

Group not specified

Group not specified

Group not specified

Not reported

Not reported

Not reported

Group not specified

9

Non‐mesh (144)

9

Group not specified

Group not specified

Group not specified

Not reported

Not reported

Not reported

Group not specified

9

Manyilirah 2012

Mesh (51)

1

0

2

1

Not reported

Not reported

4

Not reported

8

Non‐mesh (50)

2

0

1

0

Not reported

Not reported

4

Not reported

7

McGillicuddy 1998

Mesh (380)

37

3

Not reported

Not reported

Not reported

Not reported

0

Not reported

40

Non‐mesh (337)

45

2

Not reported

Not reported

Not reported

Not reported

2

Not reported

47

Nakagawa 2013

Mesh (45)

Not reported

0

1

1

8

Not reported

2

Not reported

12

Non‐mesh (46)

Not reported

1

2

1

1

Not reported

0

Not reported

5

Naveen 2014

Mesh (35)

Not reported

2

1

8

Not reported

Not reported

Not reported

2

13

Non‐mesh (35)

Not reported

0

2

3

Not reported

Not reported

Not reported

3

8

Nordin 2002

Mesh (149)

Not reported

6

4

1

1

Not reported

2

32

46

Non‐mesh (148)

Not reported

3

2

0

1

Not reported

0

42

48

Panda 2012

Mesh (20)

1

3

0

1

Not reported

0

Not reported

Not reported

5

Non‐mesh (20)

4

5

0

6

Not reported

0

Not reported

Not reported

15

Prior 1998

Mesh (42)

9

7

Group not specified

Group not specified

Not reported

Not reported

Not reported

Not reported

16

Non‐mesh (38)

8

4

Group not specified

Group not specified

Not reported

Not reported

Not reported

Not reported

12

Schmitz 1997

Mesh (32)

Not reported

0

6

0

Not reported

Not reported

0

Not reported

6

Non‐mesh (32)

Not reported

0

4

0

Not reported

Not reported

1

Not reported

5

Shi 2010

Mesh (283)

2

Not reported

Not reported

Not reported

Not reported

Not reported

3

Not reported

5

Non‐mesh (269)

1

Not reported

Not reported

Not reported

Not reported

Not reported

2

Not reported

3

van Veen 2007

Mesh (146)

Not reported

1

15

4

Not reported

1

Not reported

0

21

Non‐mesh (143)

Not reported

0

17

0

Not reported

0

Not reported

1

18

Witkowski 2000

Mesh (70)

Not reported

2

2

2

Not reported

Not reported

1

Not reported

7

Non‐mesh (70)

Not reported

5

5

1

Not reported

Not reported

2

Not reported

13

Zieren 1998

Mesh (160)

Not reported

2

11

6

Not reported

Not reported

Not reported

3

22

Non‐mesh (80)

Not reported

2

4

1

Not reported

Not reported

Not reported

2

9

Open in table viewer
Table 3. Mesh compared to non‐mesh repair for inguinal and femoral hernia repair, complications subgroups

Mesh compared to non‐mesh repair for inguinal and femoral hernia repair; complications subgroups

Patient or population: adults undergoing inguinal and femoral hernia repair
Setting: multiple hospitals from small to large tertiary centres contributed results
Intervention: mesh
Comparison: non‐mesh repair

Outcomes

Anticipated absolute effects* (95% CI)

Relative effect
(95% CI)

№ of participants
(studies)

Certainty of the evidence
(GRADE)

Comments

Risk with non‐mesh repair

risk with mesh

Complications ‐ neurovascular or visceral injury

Study population

RR 0.61
(0.49 to 0.76)

6293
(24 RCTs)

⊕⊕⊕⊕
HIGH

Follow‐up: up to 4.3 years

6 per 100

4 per 100
(3 to 5)

Complications ‐ wound infection

Study population

RR 1.29
(0.89 to 1.86)

4540
(21 RCTs)

⊕⊕⊝⊝
LOW 1

Follow‐up: up to 5 years

2 per 100

3 per 100
(2 to 4)

Complications ‐ haematoma

Study population

RR 0.88
(0.68 to 1.13)

3773
(16 RCTs)

⊕⊕⊝⊝
LOW 1

Follow‐up: up to 5 years

6 per 100

5 per 100
(4 to 7)

Complications ‐ seroma

Study population

RR 1.63
(1.03 to 2.59)

2640
(14 RCTs)

⊕⊕⊕⊝
MODERATE 2

Follow‐up: up to 4 years

2 per 100

3 per 100
(2 to 5)

Complications ‐ postoperative wound swelling

Study population

RR 4.56
(1.02 to 20.48)

388
(2 RCTs)

⊕⊕⊕⊝
MODERATE 2

Follow‐up: up to 5 years

1 per 100

5 per 100
(1 to 21)

Complications ‐ wound dehiscence

Study population

RR 0.55
(0.12 to 2.48)

329
(2 RCTs)

⊕⊕⊝⊝
LOW 1

Follow‐up: up to 3 years

2 per 100

1 per 100
(0 to 6)

Complications ‐ testicular injury or complications

Study population

RR 1.06
(0.63 to 1.76)

3741
(14 RCTs)

⊕⊕⊝⊝
LOW 1

Follow‐up: up to 4 years

1 per 100

1 per 100
(1 to 2)

Complications ‐ urinary retention

Study population

RR 0.53
(0.38 to 0.73)

1539
(8 RCTs)

⊕⊕⊕⊝
MODERATE 3

The degree of heterogeneity may be related to differing definitions or measurement of urinary retention

Follow‐up: up to 18 months

12 per 100

7 per 100
(5 to 9)

Complications ‐ pain

No clear conclusion could be reached regarding post‐operative and chronic pain in mesh compared to non‐mesh hernia repair, as the studies used different methods and grading scores to determine severity of pain, as well as many different time intervals chosen for analysis.

4999
(22 RCTs)

⊕⊝⊝⊝
VERY LOW 4

No meaningful meta‐analysis was able to be performed due to inconsistent methods/lack of comparable endpoints.

Follow‐up: up to 5 years

*The risk in the intervention group (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI).

CI: Confidence interval; RR: Risk ratio; OR: Odds ratio;

GRADE Working Group grades of evidence
High certainty: We are very confident that the true effect lies close to that of the estimate of the effect
Moderate certainty: We are moderately confident in the effect estimate: The true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different
Low certainty: Our confidence in the effect estimate is limited: The true effect may be substantially different from the estimate of the effect
Very low certainty: We have very little confidence in the effect estimate: The true effect is likely to be substantially different from the estimate of effect

1 Downgraded two levels for imprecision (wide confidence interval overlapping no effect) and inconsistency (substantial heterogeneity)

2 Downgraded one level for imprecision (wide confidence interval, relatively small population)

3 Downgraded one level for inconsistency (substantial heterogeneity)

4 Downgraded three levels for risk of bias (subjective nature of outcome and various methods of measurement), inconsistency, imprecision and indirectness (various measures, including indirectly with analgesia use)

Neurovascular or visceral injury

There was high‐quality evidence showing that neurovascular and visceral injuries were more common in the non‐mesh repair group (RR 0.61, 95% CI 0.49 to 0.76, I2 = 0%, NNTB = 22). This was assessed in 24 of 25 studies, representing an event rate of 185/3004, compared to 125/3289 in the mesh group.

Wound infection

Wound infection was found more commonly in the mesh group, (in 20 of 25 studies) with an event rate of 61/2354 (compared to 46/2186 in the non‐mesh group), with unconvincing effect size (RR 1.29, 95% CI 0.89 to 1.86, I2 = 0%, NNTB = 200, low‐quality evidence).

Haematoma

Haematomas were found more often in the non‐mesh group (111/1833) than the mesh group (104/1940), in 15 of 25 studies (RR 0.88, 95% CI 0.68 to 1.13, I2 =0%, NNTB = 143, low‐quality evidence).

Seroma

The mesh repair groups demonstrated a higher risk of seroma in 14 of 25 studies, with an event rate of 46/1373 compared to 25/1267 events in the non‐mesh group (RR 1.63, 95% CI 1.03 to 2.59, I2 = 0%, NNTB = 72, moderate‐quality evidence).

Postoperative wound swelling

The mesh repair groups demonstrated a higher risk of postoperative wound swelling: 9/194 events compared to 2/194 events in the non‐mesh group (2 of 25 studies) (RR 4.56, 95% CI 1.02 to 20.48, I2 = 33% , NNTB = 72, moderate‐quality evidence).

Wound dehiscence

Wound dehiscence events were found more often in the non‐mesh group (4/163 compared to 2/166, 2 of 25 studies), but this difference had a wide confidence interval with an unconvincing effect size (RR 0.55, 95% CI 0.12 to 2.48, I2 = 37%, NNTB = 77, low‐quality evidence).

Testicular injury or complications, including testicular swelling and atrophy

Testicular complications showed nearly equivocal results in the 14 of 25 studies that contributed data. However, a slight, non‐statistically significant increase was evident in the mesh group, with an event rate of 26/1981 (compared to 24/1760 in the non‐mesh group) (RR 1.06, CI 0.63 to 1.76, I2 = 0%, NNTB = 2000, low‐quality evidence).

Urinary retention postoperatively

Non‐mesh groups had more postoperative urinary retention (8 of 25 studies): 89/720 events compared to 48/819 events in the mesh group (RR 0.53, 95% CI 0.38 to 0.73, I2 = 56%, NNTB = 16, moderate‐quality evidence).

Postoperative pain, including acute and chronic pain

Regarding postoperative and chronic pain, 12 studies used the visual analogue scale to measure this outcome (Abd El Maksoud 2014; Butters 2007; Berndsen 2007; Shi 2010; Lermite 2012; Manyilirah 2012; Nakagawa 2013; Nordin 2002; Prieto‐Diaz‐Chavez 2009; Prior 1998; Witkowski 2000; Zieren 1998). Four studies quantified pain as a dichotomous outcome (reporting pain as present or not present) (Kucuk 2010; McGillicuddy 1998; Panda 2012; van Veen 2007); others used a numerical rating scale and one study measured pain indirectly through analgesia requirement (Barth 1998; Chakraborty 2007; Naveen 2014). Due to the many different methods of data collection and different time frames for analysis, we were unable to adequately compare these results for meta‐analysis (GRADE rating = very low quality), however, generally the studies indicated that postoperative pain is greater in the groups with non‐mesh repairs.

Sensitivity analysis

Sensitivity analysis of included studies did not demonstrate any significant changes in overall effect for most complications. Notably, none of the studies with high risk of bias contributed to neurovascular/visceral injury analysis (Analysis 1.2.1), so there was no change in estimate. However for both the postoperative wound swelling and wound dehiscence outcomes, we removed one of two studies; this resulted in RR 0.99, 95% CI 0.06 to 15.73 and RR 0.25, 95% CI 0.03 to 2.05, respectively, and a resulting drop in the quality of evidence to very low, according to GRADE criteria. When checking for heterogeneity‐inducing studies in the urinary retention outcome, we noted that removal of one study, Kux 1994 reduced heterogeneity (I2 dropped from 56% to 0%), however the overall result still supported the finding of fewer events in the mesh group (RR 0.79, 95% CI 0.56 to 1.13).

Mortality

Seven studies outlined and discussed mortality as an outcome (Berndsen 2007; Butters 2007; Elsebae 2008; Lermite 2012; McGillicuddy 1998; Nordin 2002; Prior 1998), but none of them reported any deaths. Mortality was not discussed in the remaining 18 included studies.

Secondary outcomes

Duration of surgical operation

Twenty studies measured duration of surgical operation (Abd El Maksoud 2014; Barth 1998; Berndsen 2007; Chakraborty 2007; Schmitz 1997; Leibl 1995; Elsebae 2008; Hauters 1996; Kaynak 2007; Kucuk 2010; Lermite 2012; Manyilirah 2012; Nakagawa 2013; Naveen 2014; Nordin 2002; Panda 2012; Prieto‐Diaz‐Chavez 2009; Prior 1998; Witkowski 2000; Zieren 1998). Of these, three did not provide enough data in their publications to calculate standard deviation (Chakraborty 2007; Nordin 2002; Prior 1998). We were therefore unable to include these studies in the meta‐analysis. Overall, longer durations were noted in the non‐mesh group, with surgery taking a mean of four minutes and 22 seconds longer (with a CI ranging between ‐6.85 to ‐1.6; where negative numerals indicate non‐mesh repair duration being longer). Please refer to Figure 8 for further details. However, there was a large amount of intrinsic heterogeneity between studies, as may be expected (I2 = 97%), so we applied a random‐effects model, and overall this represented very low‐quality evidence. Sensitivity analysis demonstrated no significant difference to the overall effect and heterogeneity remained very high; the mean duration was 3.58 minutes lower in the mesh group, with a larger range, and the GRADE rating remained low.


Forest plot of comparison: 1 Outcomes, outcome: 1.10 Duration of surgery.

Forest plot of comparison: 1 Outcomes, outcome: 1.10 Duration of surgery.

Duration of postoperative hospital stay

Sixteen studies addressed postoperative stay in hospital as an outcome (Abd El Maksoud 2014; Chakraborty 2007; Elsebae 2008; Kaynak 2007; Kux 1994; Lermite 2012; Nakagawa 2013; Naveen 2014; Nordin 2002; Panda 2012; Prieto‐Diaz‐Chavez 2009; Prior 1998; Shi 2010; van Veen 2007; Witkowski 2000; Zieren 1998). However, four of these studies could not be included in the meta‐analysis, due to inadequate data in which the standard deviations were not reported, and otherwise could not be estimated (Chakraborty 2007; Kux 1994; Nakagawa 2013; Prior 1998).

As there was a large degree of heterogeneity between the studies (I2 = 98%), we used a random‐effects model was for analysis. Additionally, the use of a random‐effects model was seen to be more appropriate; under the fixed‐effect model the result was heavily skewed by a single study with a weighting of 99.4%, whereas under the random‐effects model this study had a weighting of 11.6% (Prieto‐Diaz‐Chavez 2009). The result was statistically significant (P < 0.00001), demonstrating mesh repair as the intervention associated with a shorter postoperative stay by 0.6 days (95% CI ‐0.86 to ‐0.34), though the quality of the evidence according to GRADE was low. Sensitivity analysis did not make a significant difference to the overall effect (the mean difference was 1.04 days sooner, 95% CI ‐1.80 to ‐0.28; and the GRADE rating remained low), with a high degree of heterogeneity remaining (I2 = 98%).

Time to return to activities of daily living

Thirteen studies addressed the time to return to activities of daily living (ADLs) as an outcome (Abd El Maksoud 2014; Barth 1998; Berndsen 2007; Chakraborty 2007; Kaynak 2007; Kux 1994; Leibl 1995; Lermite 2012; Manyilirah 2012; Nordin 2002; Prieto‐Diaz‐Chavez 2009; van Veen 2007; Zieren 1998). However, three studies were not included in the forest plot analysis, due to insufficient data (Chakraborty 2007; Kux 1994; Leibl 1995). The meta‐analysis revealed a mean difference of 2.87 days sooner to return to ADLs in the mesh group (95% CI ‐4.42 to ‐1.32); the quality of the evidence was low. We used a random‐effects model due to there being high heterogeneity (I2 = 96%) and a wide range (2 to 26 days admission). Sensitivity analysis did not make any significant difference to the overall effect (the mean time to return to ADLs was 2.86 days sooner in mesh group, and the evidence was of moderate quality).

Conversion from laparoscopic to open surgery

None of the included studies reported any conversion from laparoscopic to open technique. Only five of the included studies compared a laparoscopic technique (laparoscopic transabdominal preperitoneal (TAPP) repairs) (Berndsen 2007; Butters 2007; Hauters 1996; Leibl 1995; Zieren 1998). All other studies compared an open mesh and non‐mesh technique of hernia repair.

Conversions from non‐mesh to mesh technique were documented in two studies, either due to technical difficulty or because it was deemed necessary for best possible repair by the surgeon (Nordin 2002; van Veen 2007). One of these studies also reported seven conversions from mesh to non‐mesh repair (van Veen 2007). One study confirmed that all participants received their allocated interventions (Manyilirah 2012), but the rest of the studies reviewed did not discuss operation conversions at all.

Subgroup analysis

Hernia type

All included studies specified inguinal hernias; only two studies reported including femoral hernias in their analysis (Prieto‐Diaz‐Chavez 2009; Witkowski 2000). In each, the number of femoral hernias in the mesh group was proportionately half the number in the non‐mesh group (5.4% compared to 10.3% of total participants in Prieto‐Diaz‐Chavez 2009, and 5.5% compared to 11% in Witkowski 2000). The outcomes were discussed with all types of hernia included in each group as a whole, therefore inadequate data were accessible to perform subgroup analysis on inguinal compared to femoral hernias.

Nineteen studies included both direct and indirect hernias in each hernia repair group (Abd El Maksoud 2014; Barth 1998; Berndsen 2007; Chakraborty 2007; Elsebae 2008; Hauters 1996; Kaynak 2007; Kucuk 2010; Leibl 1995; Manyilirah 2012; Nakagawa 2013; Naveen 2014; Nordin 2002; Prior 1998; Schmitz 1997; Shi 2010; van Veen 2007; Witkowski 2000; Zieren 1998). Seventeen of these trials provided ratios in each group (Barth 1998; Chakraborty 2007; Elsebae 2008; Hauters 1996; Kaynak 2007; Kucuk 2010; Leibl 1995; Manyilirah 2012; Nakagawa 2013; Naveen 2014; Nordin 2002; Prior 1998; Schmitz 1997; Shi 2010; van Veen 2007; Witkowski 2000; Zieren 1998). No comment was made in six studies whether hernias were found to be direct or indirect at the time of repair (Butters 2007; Kux 1994; Lermite 2012; McGillicuddy 1998; Panda 2012; Prieto‐Diaz‐Chavez 2009). As results were never reported separately for direct and indirect hernias, we were unable to perform subgroup analysis.

Gender

Eight studies included only male participants, either by inclusion criteria or incidentally (Abd El Maksoud 2014; Berndsen 2007; Butters 2007; Chakraborty 2007; Elsebae 2008; Lermite 2012; McGillicuddy 1998; Nordin 2002). Those that included both genders found women to be the minority in each group, without exception (Hauters 1996; Kaynak 2007; Kucuk 2010; Leibl 1995; Manyilirah 2012; Nakagawa 2013; Prieto‐Diaz‐Chavez 2009; Schmitz 1997; Shi 2010; Witkowski 2000; Zieren 1998). Two studies included both male and female participants but no further detail was provided (Naveen 2014; Panda 2012) and four studies did not report or discuss gender (Barth 1998; Kux 1994; Prior 1998; van Veen 2007). Again, because only general statements were made for each group and not linked to specific outcome results, we could not perform the planned subgroup analysis.

Comparison of comorbidities

American Society of Anaesthesiology (ASA) physical status classification of perioperative participants at baseline was reported (or reported to be measured without providing detail) in five studies (Berndsen 2007; Elsebae 2008; Lermite 2012; Prior 1998; Witkowski 2000).

Thirteen studies reported body mass index (BMI) or participant weight (Abd El Maksoud 2014; Berndsen 2007; Butters 2007; Hauters 1996; Kux 1994; Leibl 1995; Lermite 2012; Manyilirah 2012; McGillicuddy 1998; Nakagawa 2013; Prieto‐Diaz‐Chavez 2009; van Veen 2007; Zieren 1998). Of these, nine studies specifically compared median or mean BMI (Abd El Maksoud 2014; Butters 2007; Hauters 1996; Lermite 2012; Manyilirah 2012; Nakagawa 2013; Prieto‐Diaz‐Chavez 2009; van Veen 2007; Zieren 1998), which was found to be comparable between groups in each study.

Emergent and elective repairs

Only one study clearly selected participants presenting with strangulated inguinal hernias (Elsebae 2008). Another, Panda 2012, included participants with 'obstructed' inguinal hernias, although the clinical circumstances were not discussed in detail and it was deemed to be unclear regarding emergent or elective repair status, so a subgroup analysis could not be performed. All other studies examined hernia repair in an elective setting.

Discussion

available in

Summary of main results

We included 25 studies in this systematic review, with a total of 6293 participants in a diversity of settings in high and low‐income countries. Mesh intervention provided a statistically significant reduction in hernia recurrence, but reliability of this finding was downgraded to moderate quality of evidence due to lack of blinding. This is in accordance with the findings of the EU Hernia Trialists (EU Hernia Trialists 2002). The moderate heterogeneity detected in the analysis of hernia recurrence was attributed, in part, to the variety of mesh versus non‐mesh techniques and open versus laparoscopic techniques used across the selected studies. It may also have been affected by the temporal differences in follow‐up. Additionally, the diversity in operator experience, procedures and anaesthetic management would have contributed to this heterogeneity. Neurovascular and visceral injuries and postoperative urinary retention were less likely in the mesh group. Non‐mesh interventions were found to have a lower rate of seroma and postoperative wound swelling, indicative of a more significant local reaction to the prosthesis. They were found to have slightly lower rates of wound infection and haematoma. Wound infection particularly has been raised as a concern of potentially having an association with mesh repairs over non‐mesh repairs, although the evidence in the literature has remained controversial (Falagas 2005). A number of large studies have shown similar rates of wound infection in both groups (Falagas 2005), as we have found in this review. There may be a number of factors contributing to infection risk, including possible differences in technique or location of repair. Non‐mesh repairs had a slightly higher rate of testicular complications (including a range of severity from postoperative scrotal swelling to testicular atrophy) and wound dehiscence. All of these postoperative complications had a moderate to high GRADE of evidence. Nonetheless, apart from neurovascular and visceral injury and urinary retention, most had quite wide confidence intervals and poor effect size, leading us to conclude that these differences are of minimal clinical significance.

No clear conclusion was reached regarding postoperative and chronic pain in mesh compared to non‐mesh inguinal hernia repair, although most studies favoured mesh repair as having less associated postoperative pain. As pain is a symptom with often multi‐faceted pathophysiology and affected by many personal factors, it is difficult to measure objectively, and a number of different scales and methods of assessment were used in these studies (e.g. pain at rest/with movement, at different intervals postoperatively, with grading scales, with/without the use of analgesia and quantifying pain by analysing analgesic requirement). Moreover, several studies in this review failed to include the number of participants who reported significant pain but instead, provided an average score of pain at certain points in time postoperatively. Multiple studies utilised the Visual Analogue Scale (VAS) (a psychometric response scale used in questionnaires as a measurement instrument for subjective characteristics or attitudes that cannot be directly measured). Even with a common gauge for pain, it was difficult to interpret the data; different authors used different values to denote “significant pain”; whereas some studies identify this to be 200 mm (Berndsen 2007), a more recent study by (Boonstra 2014) recognises that mild pain is represented by values 0.1 cm to 3.8 cm, moderate pain by 3.9 cm to 5.7 cm, and severe pain 5.8 cm to 10 cm. We downgraded our GRADE assessment of the quality of the evidence for this outcome by three levels (very low quality) due to inconsistency and variability in scoring/measurement methods.

No analysis was performed regarding surgery‐related deaths as no events were reported in any studies.

Mesh repairs had a shorter operative time however we downgraded our GRADE assessment by three levels, to very low quality, due to imprecision and a considerable amount of heterogeneity for this outcome. It should be noted that different methods of mesh and non‐mesh surgeries may have contributed to this heterogeneity, including some studies using laparoscopic methods in one or both groups (specifically described in Berndsen 2007; Leibl 1995; Hauters 1996; Nakagawa 2013; Zieren 1998), compared to others who used only open surgical methods. Among these, one study specified only open techniques based on expected optimal management in the setting of emergent surgery for incarcerated hernias (specifically, Elsebae 2008), but generally incarcerated/strangulated hernias met the exclusion criteria for most studies analysed in this review. As laparoscopic methods become more widespread, this may impact on the duration of surgical hernia repairs. Different surgeons had different average operating times; skill and experience of the surgeon will impact largely on the outcome, as well as his/her familiarity with the techniques used. Postoperative hospital stay was reduced in the mesh group by only 0.6 days (moderate GRADE). An additional half‐day of hospital stay is unlikely to be clinically significant, and this result would be very system‐ and clinician‐dependent. Between the studies, the difference in length of stay ranged from 0 (van Veen 2007) to 3.1 days (Shi 2010). And of the 12 studies included in our meta‐analysis of this outcome (Abd El Maksoud 2014; Elsebae 2008; Kaynak 2007; Lermite 2012; Naveen 2014; Nordin 2002; Panda 2012; Prieto‐Diaz‐Chavez 2009; Shi 2010; van Veen 2007; Witkowski 2000; Zieren 1998), only three (Abd El Maksoud 2014; Elsebae 2008; Nordin 2002) found that participants in the mesh repair group had a longer postoperative stay. The source of the significant heterogeneity may be in part attributed to the differences in the patient populations between the studies, i.e. comorbidities, selection of emergent or elective hernia repairs and also differences in postoperative management and hospital policy. A similar high degree of heterogeneity was evident in time to return to ADLs (resulting in downgrading to moderate quality evidence), returning to daily activities a mean of 2.87 days earlier in the mesh repair group.

Overall completeness and applicability of evidence

We conducted a comprehensive electronic search of several databases (see Search methods for identification of studies for details) to obtain the included studies. Completed and ongoing trials were searched for in relevant clinical trials registers such as the World Health Organization International Clinical Trials Registry Platform and ClinicalTrials.gov for completeness. We also reviewed reference lists of included trials and review articles, books related to surgical hernia repairs, abstracts from general surgical conferences concerning hernias and mesh repair. We imposed no restrictions on publication or language, in order to minimise the risk of missing relevant studies. Studies from 15 different countries were gathered (United States, United Kingdom, Uganda, Turkey, The Netherlands, Sweden, Poland, Mexico, Lebanon, Japan, India, Germany, France, Egypt and China), representing different regions and different socioeconomic settings. The commonality of hernias worldwide (Brooks 2014b) was reflected in this review by including robust multi‐centre studies with large cohort sizes from different countries; the consistency of the results of our meta‐analysis, which supports the mainstream use of mesh repairs, are therefore relevant to all patients requiring surgical hernia repair. In saying that, the specific surgical techniques used and emphasis on cost were often informed by the location of the study and economic status. Overall, this is unlikely to have an impact on the applicability of the results, although it should be noted that routine open repairs in a non‐emergent setting may no longer be relevant in many high‐income countries, just as in the resource‐poor setting mesh repair may not always be feasible. The most common mesh repair used was Lichtenstein repair (Abd El Maksoud 2014; Barth 1998; Butters 2007; Chakraborty 2007; Elsebae 2008; Kaynak 2007; Kucuk 2010; Kux 1994; Manyilirah 2012; McGillicuddy 1998; Naveen 2014; Nordin 2002; Panda 2012; Prior 1998; van Veen 2007), whilst the most common non‐mesh repair was the Shouldice technique (Barth 1998; Berndsen 2007; Butters 2007; Hauters 1996; Kux 1994; Leibl 1995; Lermite 2012; McGillicuddy 1998; Nordin 2002; Prieto‐Diaz‐Chavez 2009; Schmitz 1997; van Veen 2007; Zieren 1998), which is consistent with the commonly performed repair techniques outlined in Description of the intervention) and reflective of practice.

Cost of surgery and hospital admission

Of the included studies, six included cost‐effectiveness as an outcome or directly commented on the comparative cost (Barth 1998; Chakraborty 2007; Kaynak 2007; Prieto‐Diaz‐Chavez 2009; Shi 2010; Zieren 1998). A number of others, mainly originating in low‐income countries, included a comment on cost as part of an introductory or discussion statement, generally stating that non‐mesh repairs are more cost‐effective (mainly due to cost of mesh materials) (Abd El Maksoud 2014; Manyilirah 2012; Naveen 2014). Due to the scope of this review and the different currencies used to calculate cost and limited information regarding what was included in cost analysis, we did not perform a meta‐analysis. If only material costs were considered, mesh repairs were more expensive than non‐mesh repairs. One study reported the difference in surgical material cost was INR 250 to 350 for the non‐mesh repairs compared to INR 1900 to 2100 for mesh repairs (Chakraborty 2007). Another reported the average material cost of mesh repair was CNY 6221.3 compared to CNY 4518 (Shi 2010). Another reported material costs that were significantly higher for the laparoscopic mesh procedure (USD 1211) compared to the open mesh (USD 124) and open non‐mesh (USD 69) procedures (Zieren 1998). A further study did not provide details but stated that the cost of the non‐mesh repair was 'significantly less' than that of mesh repair (Kaynak 2007). However, if other factors and costs were considered, mesh repairs may be more cost‐effective. For example, in Barth 1998, it was reported that a mesh was USD 46; but with the longer duration of surgery in a non‐mesh repair (while there was no additional cost for a mesh), the extra operative time on average cost USD 180. One study provided a detailed analysis of cost of mesh compared with non‐mesh repair and concluded that mesh repair overall costs were on average USD 837.66 compared to non‐mesh costs of USD 885.15 (Prieto‐Diaz‐Chavez 2009). In this analysis they considered a number of factors, including operative time, hospital stay, other healthcare costs and disability‐adjusted life years, as well as the material costs of the surgery.

Cost is a clinically significant factor, particularly in low‐income countries, where this plays a large role in access to healthcare. However, cheaper materials, such as mosquito mesh, have been proposed as an affordable alternative in theses areas. If effective, this may nullify the concerns regarding cost for mesh repair. So far results have been promising; a recent study describes low‐cost meshes worth less than USD 1 (compared to USD 125) commercial meshes in Uganda, with very comparable clinical outcomes and complication rates (Lofgren 2016).

Quality of the evidence

Overall, the methodological quality of the included studies was good, considering the nature of the intervention measured. The number of studies contributed a large number of participants to the meta‐analysis and most studies demonstrated clear methods and addressed potential biases, providing results applicable to an international audience where mesh and non‐mesh methods of hernia repair are available.

Whilst the studies demonstrated a good methodological approach, we downgraded our assessment of the quality of evidence contributing to many outcomes (hernia recurrence, wound infection, haematoma, seroma, wound dehiscence, wound swelling, urinary retention, duration of postoperative stay, time to return to ADLs), due to there being wide confidence intervals, heterogeneity and inconsistency. There were no outliers that appeared to significantly skew results. Wound swelling and dehiscence were sparsely reported, with only two studies assessing these outcomes (Nakagawa 2013 and Nordin 2002 for wound swelling, and Panda 2012 and van Veen 2007 for wound dehiscence). In general, the studies addressed their proposed outcomes well. Due to the nature of the intervention, blinding is difficult to achieve; as a result the method/attempt for blinding was not described or was unclear in most studies, with the exception of six studies which at least confirmed single, if not double, blinding (Manyilirah 2012; Nakagawa 2013; Nordin 2002; Panda 2012; Prior 1998; van Veen 2007).

For the outcome neurovascular/visceral injury, we deemed the evidence to be of high quality. For all other complications apart from pain, we deemed the evidence to be of moderate quality. The evidence on postoperative pain was judged to be of very low quality, largely due to inconsistency of measurement methods and interval of follow‐up, and therefore incomparability. Duration of surgery demonstrated significant heterogeneity of results, with an overall finding that non‐mesh repairs take longer than mesh by about four minutes; this heterogeneity is likely to be due to variation in operative experience and skill between surgeons. We accounted for this statistical heterogeneity by using a random‐effects model and downgrading the evidence a further level. Overall sensitivity analysis did not make any significant difference to estimate effects.

Please refer to summary of findings Table for the main comparison and Table 3 for further detail regarding quality of evidence.

Potential biases in the review process

Throughout the review process, at a number of stages the review authors undertook tasks independently to reduce the risk of potential bias. Following an initial search of the electronic databases, numerous studies and their respective abstracts were reviewed by three study authors independently to include/exclude them, based on the protocol criteria. Similarly the data extraction and assessment of the risk of bias was undertaken independently by two of the study authors, where differences were resolved through discussion. However, several of the studies that were included in the review were in foreign languages, which led to deviation from our initial methodology. Due to costs of translating the full article to English, the data extraction and assessment of bias was not performed by one of the review authors but rather by a translator employed by Cochrane. To avoid bias relating to heterogeneity, we used a random‐effects model when heterogeneity was found to be moderate or greater.

Agreements and disagreements with other studies or reviews

Through our review of other literature relating to the topic of mesh versus non‐mesh repair of hernias, we noted a fairly consistent opinion that mesh repair should be preferential to other non‐mesh techniques.

The European Hernia Society (EHS) guidelines provide a grade A recommendation that all adult male patients with inguinal hernia should be operated on using a mesh technique, either with open Lichtenstein or endoscopic inguinal hernia techniques (Simons 2009). Similarly the Danish Surgical Society, informed by the Daniesh Hernia database, states that in any male patient with a groin hernia a mesh repair should be undertaken, with no recommendation for non‐mesh Shouldice, Bassini or McVay techniques (Rosenberg 2011).

The EHS justify their recommendations based on Grade IA evidence of reduced hernia recurrence with mesh repair (EU Hernia Trialists Collaboration 2000; EU Hernia Trialists Collaboration 2002). In the previous version of our Cochrane Review, performed in 2001, the reduced risk of recurrence with mesh was measured to be 50% to 75% (Scott 2001). Also, a more recent multi‐centre trial reports a low cumulative recurrence rate within five years, with totally extraperitoneal (TEP) and Lichtenstein techniques ranging from 1.2% to 3.5% across the treatment groups (Eklund 2009). Subsequently, considering the outcome of hernia recurrence, it appears well‐accepted that a mesh technique is superior to that of non‐mesh; this is equally reflected in our latest analyses (risk ratio 0.46).

When considering the other outcomes of complications or postoperative pain, the advantage of a mesh repair is equivocal. When directly comparing mesh to non‐mesh techniques, the EHS measured the incidence of complications to be similar between treatment groups (EU Hernia Trialists Collaboration 2000; Simons 2009). However, with endoscopic repairs as a subgroup of the mesh group, the EHS has demonstrated with Grade IIB evidence that endoscopic repairs have been associated with higher rates of rare but serious complications such as visceral and vascular injury (Simons 2009). Amongst the included studies of this review, while the recorded rate of complications was 4% to 5% between mesh and non‐mesh — which is comparable — the burden of complications was distributed differently, with greater risk of seroma in the mesh group and more neurovascular or visceral injury in non‐mesh group. Furthermore, the EHS guidelines showed with Grade IIB evidence that mesh repair reduces the risk of chronic pain (Simons 2009); however, in our review we could not find any statistically significant difference between mesh and non‐mesh with regards to postoperative or chronic pain.

Study flow diagram.
Figures and Tables -
Figure 1

Study flow diagram.

Funnel plot of comparison: mesh vs non‐mesh repair, primary outcome: hernia recurrence.
Figures and Tables -
Figure 2

Funnel plot of comparison: mesh vs non‐mesh repair, primary outcome: hernia recurrence.

Funnel plot of comparison: mesh vs non‐mesh repair, primary outcome: complications.
Figures and Tables -
Figure 3

Funnel plot of comparison: mesh vs non‐mesh repair, primary outcome: complications.

Risk of bias graph: review authors' judgements about each 'Risk of bias' item presented as percentages across all included studies.
Figures and Tables -
Figure 4

Risk of bias graph: review authors' judgements about each 'Risk of bias' item presented as percentages across all included studies.

Risk of bias summary: review authors' judgements about each 'Risk of bias' item for each included study.
Figures and Tables -
Figure 5

Risk of bias summary: review authors' judgements about each 'Risk of bias' item for each included study.

Forest plot of comparison: 1 Comparison: Mesh vs Non‐Mesh repair, outcome: 1.1 Primary Outcome: Hernia Recurrence.
Figures and Tables -
Figure 6

Forest plot of comparison: 1 Comparison: Mesh vs Non‐Mesh repair, outcome: 1.1 Primary Outcome: Hernia Recurrence.

Forest plot of comparison: 1 Comparison 1: Mesh vs Non‐Mesh repair, outcome: 1.2 Primary Outcome: Complications.
Figures and Tables -
Figure 7

Forest plot of comparison: 1 Comparison 1: Mesh vs Non‐Mesh repair, outcome: 1.2 Primary Outcome: Complications.

Forest plot of comparison: 1 Outcomes, outcome: 1.10 Duration of surgery.
Figures and Tables -
Figure 8

Forest plot of comparison: 1 Outcomes, outcome: 1.10 Duration of surgery.

Comparison 1 Comparison: mesh versus non‐mesh repair, Outcome 1 Primary Outcome: Hernia Recurrence.
Figures and Tables -
Analysis 1.1

Comparison 1 Comparison: mesh versus non‐mesh repair, Outcome 1 Primary Outcome: Hernia Recurrence.

Comparison 1 Comparison: mesh versus non‐mesh repair, Outcome 2 Primary Outcome: Complications.
Figures and Tables -
Analysis 1.2

Comparison 1 Comparison: mesh versus non‐mesh repair, Outcome 2 Primary Outcome: Complications.

Comparison 1 Comparison: mesh versus non‐mesh repair, Outcome 3 Primary outcome: Mortality, 30 days post‐operation.
Figures and Tables -
Analysis 1.3

Comparison 1 Comparison: mesh versus non‐mesh repair, Outcome 3 Primary outcome: Mortality, 30 days post‐operation.

Comparison 1 Comparison: mesh versus non‐mesh repair, Outcome 4 Duration of surgical operation.
Figures and Tables -
Analysis 1.4

Comparison 1 Comparison: mesh versus non‐mesh repair, Outcome 4 Duration of surgical operation.

Comparison 1 Comparison: mesh versus non‐mesh repair, Outcome 5 Duration of Postoperative Stay.
Figures and Tables -
Analysis 1.5

Comparison 1 Comparison: mesh versus non‐mesh repair, Outcome 5 Duration of Postoperative Stay.

Comparison 1 Comparison: mesh versus non‐mesh repair, Outcome 6 Time to return to full ADLs.
Figures and Tables -
Analysis 1.6

Comparison 1 Comparison: mesh versus non‐mesh repair, Outcome 6 Time to return to full ADLs.

Summary of findings for the main comparison. Mesh compared to non‐mesh repair for inguinal and femoral hernia repair

Mesh compared to non‐mesh repair for inguinal and femoral hernia repair

Patient or population: adults undergoing inguinal and femoral hernia repair
Setting: multiple hospitals from small to large tertiary centres contributed results
Intervention: mesh repair
Comparison: non‐mesh repair

Outcomes

Anticipated absolute effects* (95% CI)

Relative effect
(95% CI)

№ of participants
(studies)

Certainty of the evidence
(GRADE)

Comments

Risk with non‐mesh repair

Risk with mesh

Hernia recurrence

Study population

RR 0.46
(0.26 to 0.80)

5575
(21 RCTs)

⊕⊕⊕⊝
MODERATE 1

In the context of surgical intervention, double blinding is difficult to achieve but most studies at least attempted single blinding.

Follow‐up: up to 5 years

4 per 100

2 per 100
(1 to 3)

Surgical complications ‐ neurovascular or visceral injury

Study population

RR 0.61
(0.49 to 0.76)

6293
(24 RCTs)

⊕⊕⊕⊕
HIGH

Follow‐up: up to 4.3 years

6 per 100

4 per 100

(3 to 5)

Surgical complications ‐ wound infection

Study population

RR 1.29
(0.89 to 1.86)

4540
(21 RCTs)

⊕⊕⊝⊝
LOW 2

Follow‐up: up to 5 years

2 per 100

3 per 100

(2 to 4)

Surgical complications ‐ wound dehiscence

Study population

RR 0.55
(0.12 to 2.48)

329
(2 RCTs)

⊕⊕⊝⊝
LOW 2

Follow‐up: up to 3 years

2 per 100

1 per 100

(0 to 6)

Mortality (within 30 days post‐surgery)

There were no reported events of mortality within 30 days so this outcome could not be compared.

However it can be concluded that both groups have very low rates of postoperative mortality.

Follow‐up on mortality: up to 5 years

0 per 100

0 per 100
(0 to 0)

Duration of surgery

(minutes)

The mean duration of surgery ranged from 10 to 94 minutes

MD 4.22 minutes lower
(6.85 lower to 1.6 lower)

4148
(20 RCTs)

⊕⊝⊝⊝
VERY LOW 3

The large degree of heterogeneity is likely to be related to variation in surgeon skill and familiarity with the intervention.

Duration of postoperative stay

(days)

The mean duration of postoperative stay ranged from 0.27 to 7.6 days

MD 0.6 days lower
(0.86 lower to 0.34 lower)

2966
(12 RCTs)

⊕⊕⊝⊝
LOW 4

Time to return to full ADLs (days)

The mean time to return to full ADLs ranged from 2.06 to 26 days

MD 2.87 days lower
(4.42 lower to 1.32 lower)

3183
(10 RCTs)

⊕⊕⊝⊝
LOW 4

Conversion from laparoscopic to open approach

No studies reported any conversion from laparoscopic to open technique where laparoscopic technique was used for mesh repair.

1680
(5 RCTs)

⊕⊕⊝⊝
LOW 5

The only laparoscopic techniques used for comparison were mesh repairs, and none of these studies reported conversion to open repair so this could not be compared.

*The risk in the intervention group (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI).

CI: Confidence interval; RR: Risk ratio; OR: Odds ratio; ADLs: activities of daily living; RCT: randomised controlled trial; MD: mean difference

GRADE Working Group grades of evidence
High certainty: We are very confident that the true effect lies close to that of the estimate of the effect
Moderate certainty: We are moderately confident in the effect estimate: The true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different
Low certainty: Our confidence in the effect estimate is limited: The true effect may be substantially different from the estimate of the effect
Very low certainty: We have very little confidence in the effect estimate: The true effect is likely to be substantially different from the estimate of effect

1 Downgraded one level for inconsistency (moderate heterogeneity)

2 Downgraded two levels for imprecision (wide confidence interval overlapping no effect) and inconsistency (substantial heterogeneity)

3 Downgraded three levels for inconsistency (considerable heterogeneity) and imprecision (wide confidence interval overlapping no effect)

4 Downgraded two levels for inconsistency (considerable heterogeneity)

5 Downgraded two levels for risk of bias and imprecision due to low event rate (not reported)

Figures and Tables -
Summary of findings for the main comparison. Mesh compared to non‐mesh repair for inguinal and femoral hernia repair
Table 1. Quality of Evidence, GRADE definitions

Grade

Definition

High

We are very confident that the true effect lies close to that of the estimate of the effect.

Moderate

We are moderately confident in the effect estimate: The true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different.

Low

Our confidence in the effect estimate is limited: The true effect may be substantially different from the estimate of the effect.

Very Low

We have very little confidence in the effect estimate: The true effect is likely to be substantially different from the estimate of effect.

Figures and Tables -
Table 1. Quality of Evidence, GRADE definitions
Table 2. Overview of complications reported in primary studies

Study

Group (Total number)

Neurovascular injury (including paraesthesia) or visceral injury

Wound infection

Haematoma

Seroma

Postoperative wound swelling

Wound dehiscence

Testicular injury or complications

Urinary retention

Total complications

Abd El Maksoud 2014

Mesh (119)

Not reported

8

Not reported

7

Not reported

Not reported

0

5

20

Non‐mesh (108)

Not reported

2

Not reported

4

Not reported

Not reported

1

6

13

Barth 1998

Mesh (54)

Not reported

Not reported

2

Not reported

Not reported

Not reported

Not reported

2

4

Non‐mesh (51)

Not reported

Not reported

1

Not reported

Not reported

Not reported

Not reported

0

1

Berndsen 2007

Mesh (538)

55

Not reported

49

Not reported

Not reported

Not reported

Not reported

Not reported

104

Non‐mesh (530)

107

Not reported

60

Not reported

Not reported

Not reported

Not reported

Not reported

167

Butters 2007

Mesh (187)

12

Not reported

Not reported

Not reported

Not reported

Not reported

0

Not reported

12

Non‐mesh (93)

9

Not reported

Not reported

Not reported

Not reported

Not reported

0

Not reported

9

Chakraborty 2007

Mesh (120)

0

10

Not reported

Not reported

Not reported

Not reported

5

Not reported

15

Non‐mesh (120)

0

5

Not reported

Not reported

Not reported

Not reported

5

Not reported

10

Elsebae 2008

Mesh (27)

Not reported

1

Not reported

1

Not reported

Not reported

Not reported

Not reported

2

Non‐mesh (27)

Not reported

3

Not reported

0

Not reported

Not reported

Not reported

Not reported

3

Hauters 1996

Mesh (35)

Not reported

0

Not reported

3

Not reported

Not reported

4

Not reported

7

Non‐mesh (35)

Not reported

1

Not reported

0

Not reported

Not reported

2

Not reported

3

Kaynak 2007

Mesh (354)

Not reported

5

7

8

Not reported

Not reported

0

Not reported

20

Non‐mesh (297)

Not reported

3

8

6

Not reported

Not reported

0

Not reported

17

Kucuk 2010

Mesh (130)

1

7

0

3

Not reported

Not reported

Not reported

Not reported

11

Non‐mesh (176)

1

9

2

3

Not reported

Not reported

Not reported

Not reported

15

Kux 1994

Mesh (102)

Not reported

2

2

Not reported

Not reported

Not reported

5

2

11

Non‐mesh (107)

Not reported

0

2

Not reported

Not reported

Not reported

3

34

39

Leibl 1995

Mesh (54)

Not reported

0

2

Not reported

Not reported

Not reported

0

2

4

Non‐mesh (48)

Not reported

0

1

Not reported

Not reported

Not reported

2

1

4

Lermite 2012

Mesh (156)

9

Group not specified

Group not specified

Group not specified

Not reported

Not reported

Not reported

Group not specified

9

Non‐mesh (144)

9

Group not specified

Group not specified

Group not specified

Not reported

Not reported

Not reported

Group not specified

9

Manyilirah 2012

Mesh (51)

1

0

2

1

Not reported

Not reported

4

Not reported

8

Non‐mesh (50)

2

0

1

0

Not reported

Not reported

4

Not reported

7

McGillicuddy 1998

Mesh (380)

37

3

Not reported

Not reported

Not reported

Not reported

0

Not reported

40

Non‐mesh (337)

45

2

Not reported

Not reported

Not reported

Not reported

2

Not reported

47

Nakagawa 2013

Mesh (45)

Not reported

0

1

1

8

Not reported

2

Not reported

12

Non‐mesh (46)

Not reported

1

2

1

1

Not reported

0

Not reported

5

Naveen 2014

Mesh (35)

Not reported

2

1

8

Not reported

Not reported

Not reported

2

13

Non‐mesh (35)

Not reported

0

2

3

Not reported

Not reported

Not reported

3

8

Nordin 2002

Mesh (149)

Not reported

6

4

1

1

Not reported

2

32

46

Non‐mesh (148)

Not reported

3

2

0

1

Not reported

0

42

48

Panda 2012

Mesh (20)

1

3

0

1

Not reported

0

Not reported

Not reported

5

Non‐mesh (20)

4

5

0

6

Not reported

0

Not reported

Not reported

15

Prior 1998

Mesh (42)

9

7

Group not specified

Group not specified

Not reported

Not reported

Not reported

Not reported

16

Non‐mesh (38)

8

4

Group not specified

Group not specified

Not reported

Not reported

Not reported

Not reported

12

Schmitz 1997

Mesh (32)

Not reported

0

6

0

Not reported

Not reported

0

Not reported

6

Non‐mesh (32)

Not reported

0

4

0

Not reported

Not reported

1

Not reported

5

Shi 2010

Mesh (283)

2

Not reported

Not reported

Not reported

Not reported

Not reported

3

Not reported

5

Non‐mesh (269)

1

Not reported

Not reported

Not reported

Not reported

Not reported

2

Not reported

3

van Veen 2007

Mesh (146)

Not reported

1

15

4

Not reported

1

Not reported

0

21

Non‐mesh (143)

Not reported

0

17

0

Not reported

0

Not reported

1

18

Witkowski 2000

Mesh (70)

Not reported

2

2

2

Not reported

Not reported

1

Not reported

7

Non‐mesh (70)

Not reported

5

5

1

Not reported

Not reported

2

Not reported

13

Zieren 1998

Mesh (160)

Not reported

2

11

6

Not reported

Not reported

Not reported

3

22

Non‐mesh (80)

Not reported

2

4

1

Not reported

Not reported

Not reported

2

9

Figures and Tables -
Table 2. Overview of complications reported in primary studies
Table 3. Mesh compared to non‐mesh repair for inguinal and femoral hernia repair, complications subgroups

Mesh compared to non‐mesh repair for inguinal and femoral hernia repair; complications subgroups

Patient or population: adults undergoing inguinal and femoral hernia repair
Setting: multiple hospitals from small to large tertiary centres contributed results
Intervention: mesh
Comparison: non‐mesh repair

Outcomes

Anticipated absolute effects* (95% CI)

Relative effect
(95% CI)

№ of participants
(studies)

Certainty of the evidence
(GRADE)

Comments

Risk with non‐mesh repair

risk with mesh

Complications ‐ neurovascular or visceral injury

Study population

RR 0.61
(0.49 to 0.76)

6293
(24 RCTs)

⊕⊕⊕⊕
HIGH

Follow‐up: up to 4.3 years

6 per 100

4 per 100
(3 to 5)

Complications ‐ wound infection

Study population

RR 1.29
(0.89 to 1.86)

4540
(21 RCTs)

⊕⊕⊝⊝
LOW 1

Follow‐up: up to 5 years

2 per 100

3 per 100
(2 to 4)

Complications ‐ haematoma

Study population

RR 0.88
(0.68 to 1.13)

3773
(16 RCTs)

⊕⊕⊝⊝
LOW 1

Follow‐up: up to 5 years

6 per 100

5 per 100
(4 to 7)

Complications ‐ seroma

Study population

RR 1.63
(1.03 to 2.59)

2640
(14 RCTs)

⊕⊕⊕⊝
MODERATE 2

Follow‐up: up to 4 years

2 per 100

3 per 100
(2 to 5)

Complications ‐ postoperative wound swelling

Study population

RR 4.56
(1.02 to 20.48)

388
(2 RCTs)

⊕⊕⊕⊝
MODERATE 2

Follow‐up: up to 5 years

1 per 100

5 per 100
(1 to 21)

Complications ‐ wound dehiscence

Study population

RR 0.55
(0.12 to 2.48)

329
(2 RCTs)

⊕⊕⊝⊝
LOW 1

Follow‐up: up to 3 years

2 per 100

1 per 100
(0 to 6)

Complications ‐ testicular injury or complications

Study population

RR 1.06
(0.63 to 1.76)

3741
(14 RCTs)

⊕⊕⊝⊝
LOW 1

Follow‐up: up to 4 years

1 per 100

1 per 100
(1 to 2)

Complications ‐ urinary retention

Study population

RR 0.53
(0.38 to 0.73)

1539
(8 RCTs)

⊕⊕⊕⊝
MODERATE 3

The degree of heterogeneity may be related to differing definitions or measurement of urinary retention

Follow‐up: up to 18 months

12 per 100

7 per 100
(5 to 9)

Complications ‐ pain

No clear conclusion could be reached regarding post‐operative and chronic pain in mesh compared to non‐mesh hernia repair, as the studies used different methods and grading scores to determine severity of pain, as well as many different time intervals chosen for analysis.

4999
(22 RCTs)

⊕⊝⊝⊝
VERY LOW 4

No meaningful meta‐analysis was able to be performed due to inconsistent methods/lack of comparable endpoints.

Follow‐up: up to 5 years

*The risk in the intervention group (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI).

CI: Confidence interval; RR: Risk ratio; OR: Odds ratio;

GRADE Working Group grades of evidence
High certainty: We are very confident that the true effect lies close to that of the estimate of the effect
Moderate certainty: We are moderately confident in the effect estimate: The true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different
Low certainty: Our confidence in the effect estimate is limited: The true effect may be substantially different from the estimate of the effect
Very low certainty: We have very little confidence in the effect estimate: The true effect is likely to be substantially different from the estimate of effect

1 Downgraded two levels for imprecision (wide confidence interval overlapping no effect) and inconsistency (substantial heterogeneity)

2 Downgraded one level for imprecision (wide confidence interval, relatively small population)

3 Downgraded one level for inconsistency (substantial heterogeneity)

4 Downgraded three levels for risk of bias (subjective nature of outcome and various methods of measurement), inconsistency, imprecision and indirectness (various measures, including indirectly with analgesia use)

Figures and Tables -
Table 3. Mesh compared to non‐mesh repair for inguinal and femoral hernia repair, complications subgroups
Comparison 1. Comparison: mesh versus non‐mesh repair

Outcome or subgroup title

No. of studies

No. of participants

Statistical method

Effect size

1 Primary Outcome: Hernia Recurrence Show forest plot

21

5575

Risk Ratio (M‐H, Random, 95% CI)

0.46 [0.26, 0.80]

2 Primary Outcome: Complications Show forest plot

24

Risk Ratio (M‐H, Fixed, 95% CI)

Subtotals only

2.1 Neurovascular or visceral injury

24

6293

Risk Ratio (M‐H, Fixed, 95% CI)

0.61 [0.49, 0.76]

2.2 Wound infection

20

4540

Risk Ratio (M‐H, Fixed, 95% CI)

1.29 [0.89, 1.86]

2.3 Haematoma

15

3773

Risk Ratio (M‐H, Fixed, 95% CI)

0.88 [0.68, 1.13]

2.4 Seroma

14

2640

Risk Ratio (M‐H, Fixed, 95% CI)

1.63 [1.03, 2.59]

2.5 Post‐operative wound swelling

2

388

Risk Ratio (M‐H, Fixed, 95% CI)

4.56 [1.02, 20.48]

2.6 Wound dehiscence

2

329

Risk Ratio (M‐H, Fixed, 95% CI)

0.55 [0.12, 2.48]

2.7 Testicular injury or complications

14

3741

Risk Ratio (M‐H, Fixed, 95% CI)

1.06 [0.63, 1.76]

2.8 Urinary retention

8

1539

Risk Ratio (M‐H, Fixed, 95% CI)

0.53 [0.38, 0.73]

3 Primary outcome: Mortality, 30 days post‐operation Show forest plot

7

2546

Risk Ratio (M‐H, Fixed, 95% CI)

0.0 [0.0, 0.0]

4 Duration of surgical operation Show forest plot

20

4148

Mean Difference (IV, Random, 95% CI)

‐4.22 [‐6.85, ‐1.60]

5 Duration of Postoperative Stay Show forest plot

12

2966

Mean Difference (IV, Random, 95% CI)

‐0.60 [‐0.86, ‐0.34]

6 Time to return to full ADLs Show forest plot

10

3183

Mean Difference (IV, Random, 95% CI)

‐2.87 [‐4.42, ‐1.32]

Figures and Tables -
Comparison 1. Comparison: mesh versus non‐mesh repair