Results of the search

From our searches, we identified 46 potential studies, of which 6 were from the Cochrane Neuromuscular Specialised Register, 8 were from CENTRAL,16 were from MEDLINE and 16 from Embase. After we removed duplicate records, 24 articles remained and we obtained them for assessment of eligibility. See Figure 1 for a flow chart illustrating the study selection process.

Included studies

See Characteristics of included studies.

We included five studies for meta‐analysis (Erel 2001; Hansen 2009; Kharwadkar 2005; Menovsky 2004; Theopold 2012).

The two‐group parallel randomised trial by Erel 2001 recruited 64 adults aged 18 to 80 years with carpal tunnel syndrome (CTS), undergoing open carpal tunnel decompression (OCTD). One group received a single deep subcutaneous absorbable 4‐0 Vicryl suture centrally and a completion continuous 4‐0 Vicryl intradermal suture for wound closure. The other group received interrupted 5‐0 Prolene sutures. The research took place in Hull, UK, over a six‐month period, although we do not know exactly when. The diagnostic criteria for CTS were not stated. The trial authors excluded those prone to poor wound healing (people taking corticosteroids, known to have "wound healing problems") and pregnant women. No hypothesis or primary outcome was stated. Outcomes were assessed at 10 days and 6 weeks postoperatively. The authors collected data on patient‐reported pain using a Likert scale (one to five) and the dichotomous outcomes of dehiscence, swelling, inflammation, infection and haematoma. No attrition was described. No funding source, conflicts of interest, or ethical review was described.

Hansen 2009 conducted a two‐group parallel, randomised trial of 58 hands (belonging to 50 participants) undergoing endoscopic carpal tunnel decompression (ECTD) for CTS confirmed by electrical studies. In one group the incision was closed with two interrupted 5‐0 non‐absorbable Novafil sutures and the other with a continuous intradermal 4‐0 monofilament Caprosyn suture. The study took place in Holstebro, Denmark, during 2006 to 2007. Participants with bilateral CTS having bilateral surgery underwent decompression on separate occasions and so were treated as independent cases. Inclusion criteria were not stated. The trialists appropriately excluded those with potential confounders for local pain and impaired healing (diabetes mellitus, inflammatory arthropathy and obesity), those ineligible for ECTD (due to prior wrist fracture or revisional surgery), and pregnant women. We assumed that the hand was the unit of randomisation but this is not defined. The power calculation lacked integral statistics (the desired alpha, beta and effect size) which precludes verification. Participants recorded their pain on a visual analogue scale (VAS) every day until follow‐up at days 10 to 14. Participants recorded scar satisfaction in private, three months postoperatively on a 5‐point ordinal scale ("very ugly", "ugly", "tolerable", "nice" or "very nice"). One participant withdrew from the trial on the day of surgery owing to symptom resolution, and two were excluded as the operation was converted to an open approach—we do not know if they were excluded pre‐ or post‐randomisation. One participant was lost to follow‐up and their data were excluded. The trial had ethical approval. The report does not state authors' conflicts of interest, predilection for suture material or any funding for the trial.

Kharwadkar 2005 was an ethically‐approved, two‐group, parallel randomised trial on 40 hands (belonging to 33 participants) with idiopathic CTS, treated with OCTD. In one group the incision was closed with 3‐0 Vicryl and the other with 3‐0 Prolene as subcuticular sutures. The trial took place in Scunthorpe, UK, in 2003. The diagnostic criteria for CTS were: 1) pain, paraesthesia or hypoaesthesia (or both) in the hand in the area innervated by the median nerve; and 2) electrophysiological confirmation of the diagnosis. The study included adults aged 18 to 75 years with CTS according to the above criteria who could also complete written questionnaires. The trialists excluded secondary CTS, those with a potential "double‐crush" (e.g. cervical radiculopathy or polyneuropathy) and those with "wound healing problems". The exclusion criteria were prior wrist surgery or trauma, secondary CTS or peripheral neuropathy, steroid use or "wound healing problems". No power calculation was provided. Participants completed validated questionnaires regarding their symptoms and function, the (Boston) Carpal Tunnel Syndrome Symptom Severity Scale (CTS‐SSS) and Functional Status Scale (CTS‐FSS) and independent assessments of outcomes were also undertaken at 2, 6 and 12 weeks postoperatively. There are many issues with Boston questionnaire, the foremost being that it is ostensibly not designed to measure the function of the hand; it was developed to quantify the symptoms of carpal tunnel syndrome (Levine 1993). Further: it is ambiguous with respect to which hand is being measured (or whether it measures both); it fails to adjust for unbalanced sided symptoms; and the difference between categories is subjective and may not be meaningful or externally valid. Presentation of these data as a quantification of hand function may be misleading. Further, as these outcome data are categorical, the authors should not present means (with standard deviations (SDs)). The trialists' methods section explains that rank‐based methods were used to analyse differences; we assume that the correct rank‐test was used for paired and unpaired data, although this is not stated. The direction of potential statistical bias cannot be confidently concluded. The trial report does not describe conflicts of interest, funding or pre‐existing suture preferences.

Menovsky 2004 conducted a three‐group parallel, randomised trial on 61 people with electromyographically proven idiopathic CTS undergoing unilateral OCTD. In one group the skin was closed with 4‐0 Ethilon, one with 4‐0 Vicryl and the other with 4‐0 stainless steel based cable sutures; all in interrupted format. No inclusion or exclusion criteria were described. They "randomised" participants to one of three groups The outcomes of pain (as measured on a 100‐point VAS) and complications were assessed at 10 days and six weeks postoperatively by the first author (TM) although the diagnostic criteria for "superficial wound infection" or "posttraumatic dystrophy" were not stated. Additionally, hands were photographed postoperatively for independent cosmetic analysis (comprising scar redness, granuloma formation and hypertrophy measured on a three‐point scale of "none", "mild" and "severe") but we do not know when the hands were photographed, the conditions under which the images were acquired or who assessed them. There was no description of a funding source, conflicts of interest, pre‐existing suture preferences or ethical review.

Theopold 2012 was a two‐group parallel, randomised trial of 47 patients with CTS seeking open decompression at Cork University Teaching Hospital, Ireland. In one group the skin was closed with 4‐0 Vircyl Rapide and the other with 4‐0 Novofil as interrupted sutures. We are not told when the trial took place. The inclusion criteria and diagnostic criteria for CTS are not described. The trial excluded patients undergoing revision surgery, those with an allergy to suture materials, those prone to hypertrophic or keloid scarring and people who were immunosuppressed. On the fourth postoperative day, participants were assessed for wound infection, with the criteria being "any patient that was prescribed antibiotics for their wound", but we do not know who reviewed the wound, made the diagnosis or if this was a post hoc assessment. On the 14th postoperative day, participants with non‐absorbable sutures had the material removed and the pain of this procedure was scored on a VAS (ranging from 0 to 10) but again, we do not know who recorded this score or the circumstances under which it was measured. Six weeks postoperatively, scars were assessed using a validated scale (Patient and Observer Scar Assessment Scale (POSAS)) by a surgeon blind to grouping, although again we do not know the identity of this assessor. The POSAS, available at www.posas.org, is an ordinal scale for various outcomes spanning 1 to 10. However, the authors report means and standard deviations, and compare these outcome measures with "Students [sic] t‐test" which is not entirely appropriate. Ten categories cannot be reliably approximated to the normal; a mean score is translatable back to the category and we assume that such data is skewed, so violating t‐distribution methods. Rank‐based methods would have been more appropriate, with the presentation of median differences and their spread. This raises concerns over the validity of the statistical methods used because the method may influence the effect estimate. A power calculation justified the trial authors' sample size. Nine participants (23.7%) were lost to follow‐up but we do not know the group‐specific rates. The trial does not describe funding, pre‐existing suture preferences, conflicts of interest or ethical review.

Excluded studies

See Characteristics of excluded studies. As literature on this topic is very limited, we listed all papers assessed for inclusion, rather than follow usual practice, which is to restrict Excluded studies to reports of potentially randomised trials.

We excluded Freshwater 2012 and Dellon 2001 because these were letters to the editor in response to two included trials (Theopold 2012 and Erel 2001, respectively). We also excluded the case reports by Acioly 2013, Hung 2008, Kokkalis 2015, Lu 2017 and Zingale 2003 as they were not applicable to our review. The prospective studies by Dosani 2013 and MacFarlane 2014 reported the costs, healing profiles and clinical outcomes of absorbable versus non‐absorbable sutures for skin closure following open carpal tunnel decompression; however, neither were randomised trials and so we excluded them. Although a potentially eligible study, we had to exclude Kundra 2010 because it included patients undergoing numerous different hand operations (trigger finger release, Dupuytren's excision and carpal tunnel surgery) which were randomised to compare the aesthetic outcomes of absorbable and non‐absorbable sutures only; they did not record any outcomes of relevance to this review. The investigators reported scar outcomes for the entire sample; therefore, we wrote to the authors requesting data specific to carpal tunnel decompression surgery and any other outcomes which may have been recorded (no protocol is provided to cross‐check) and although the first author replied, the original data were not available and so we were obliged to exclude the trial. We excluded the review articles by Scholten 2012 and Jerosch‐Herold 2006 because they contained no primary outcome data. We excluded the case series by Nassar 2014 and Kokkalis 2016 as they described the use of a synthetic biofilm after revision OCTD. Ramos‐Zúñiga 2017 reported the outcomes a series of patients undergoing ECTD and so was not applicable. We excluded Lattré 2016 as this described their experience with OCTD and the hypothenar fat pad flap. We excluded Magalhães 2017 as this systematic review had similar aims to our review; we cross‐checked their references and found that we had screened all the same citations except Bolster 2013, which we excluded as this trial compared different formats of suture (interrupted versus vertical mattress) after OTCD. The case series by Tosti 2017 was excluded as this considered catheter‐associated radial artery pseudoaneurysms.

Allocation

Both Kharwadkar 2005 and Theopold 2012 used sealed opaque envelopes to conceal allocation (presumably from participants, although not explicitly stated, because blinding the surgeon would not be possible), and so we judged these studies to have low risk of allocation bias. The risk of bias from the method of random sequence generation was low in Kharwadkar 2005, but unclear in Theopold 2012, as the report did not provide details.

We judged Hansen 2009 to have unclear risk of selection bias, as methods of randomisation and allocation concealment were inadequately described. We judged Erel 2001 and Menovsky 2004 as having unclear risk of bias from the method of randomisation and high risk of bias from unconcealed allocation.

Blinding

We assumed that blinding of participants and interviewers did not take place in Erel 2001, since these procedures were not described. Also Erel 2001 did not describe how outcomes were assessed, e.g. who determined the presence or absence of residual inflammation. This is particularly concerning given that "data were collected by telephone interview where possible". Also, the paper does not detail how the outcome of residual pain was measured, i.e. a mildly sore hand after substantial activity in the absence of analgesia is not the same as significant rest pain in the hand despite the use of opioid analgesia. We considered the trial at high risk of performance and detection bias.

Whilst Hansen 2009 did not qualify how they assessed for the presence of infection postoperatively, it is reasonable to deduce that an appropriately qualified healthcare professional with experience in hand surgery could reliably identify this adverse outcome. More importantly, we do not know if the same nurse removed the sutures and measured the outcome of infection. Similarly, this study states that the questionnaire regarding cosmesis was completed and placed into a closed envelope, but it is again unclear whether participants or researchers were blinded to the collected data and whether statistical analysis was undertaken with knowledge of the grouping. Therefore, we have judged the risk of performance bias as unclear (given the lack of information on participant blinding) and the risk of detection bias as unclear (for the same reason).

In Kharwadkar 2005, participants completed validated questionnaires regarding their hand function postoperatively and independent assessments of outcomes were also undertaken at 2, 6, and 12 weeks postoperatively. However, it is not clear whether the assessors were blinded to the intervention. The authors did not declare any conflicts of interest.

In the three‐group trial by Menovsky 2004, performance bias was likely as blinding was unclear and detection bias was possible for pain and numbness outcomes, given that the lead author and operating surgeon undertook the postoperative evaluation of these symptoms. Conversely, an independent group of four unnamed persons performed assessments of aesthetic outcomes based on photographs. This mixed picture led us to assign a judgement of 'unclear' risk of detection bias.

The data from Theopold 2012 are at high risk of performance and detection bias. Participants were not blinded and it is unclear who obtained the patient‐reported pain data on the 14th postoperative day; however, it is reasonable to assume that the methods were consistent and patient‐reported data from the 14th postoperative day is at high risk of bias. Data obtained six weeks postoperatively was via assessors blind to grouping. Overall we judged there to be a high risk of performance and detection bias given the majority of data (and those we have used) were obtained from unblinded participants and assessors.

Incomplete outcome data

Erel 2001 had no responses from six participants (9.4%) at 10 days for their primary outcome of pain, and had at least four dropouts (6.3%) at six weeks. We assessed the trial as having unclear risk of attrition bias. There was a high dropout rate (19%) in Theopold 2012 and so we assigned a judgement of 'unclear' risk of bias in this domain. Four people dropped out for reasons unrelated to the interventions in Hansen 2009. We considered this to represent a low risk of bias. Other studies reported complete data.

Selective reporting

The risk of selective outcome reporting by trial participants was high in Erel 2001 as numerical results were not reported for some outcomes, and unclear in Hansen 2009. We also assessed the other three trials as 'unclear' risk because there were no published protocols to which we could compare the published articles (Kharwadkar 2005; Menovsky 2004; Theopold 2012).

Other potential sources of bias

The contact author of Hansen 2009 confirmed in communication with review authors that the trialists had no conflicts of interest and that the trial was unfunded. He also declared a preference for absorbable sutures.

In Erel 2001 the Chi² test was incorrectly used for some comparisons as they violated the assumptions that < 20% of expected counts are < 5. Ideally, they would have used exact methods (or resampling, also known as bootstrapping) as a wrongly used test of proportional difference is more likely to find a significant difference where none exists. We are unsure how the presumed continuous variable of 'weeks off work' was analysed but this is not an outcome in this review.

We assessed the risk of 'other bias' as unclear for Theopold 2012. The study reported no conflicts of interest and no specific funding, however, trialists did not provide information on pre‐existing suture preference.

Kharwadkar 2005 and Menovsky 2004 did not describe conflicts of interest or individual preferences for suture type and these omissions weaken the reliability of the results; we therefore assessed them as at unclear risk of bias. We feel that it is important for surgeons to state their prior biases, especially when outcome assessment is unblinded. This is because some surgeons may prefer one suture material over another and this could bias the assessment of outcomes.

Absorbable versus non‐absorbable sutures after open carpal tunnel decompression (OCTD)

Primary outcome: postoperative pain

Postoperative pain at 10 days

This outcome was reported in Erel 2001, Kharwadkar 2005 and Menovsky 2004.

Ten days after OCTD, the standardised mean difference (SMD) in pain in absorbable versus non‐absorbable suture groups was 0.03 (95% confidence interval (CI) ‐0.43 to 0.48; 3 randomised controlled trials (RCTs), number of participants (N) = 137; I² = 43%), as shown in Analysis 1.1 and Figure 3. As a rule of thumb, an SMD of less than 0.2 indicates no clear difference (Higgins 2011). The I² suggests moderate heterogeneity, probably arising from variation in surgical techniques and outcome measures. Analyses using fixed‐effect instead of the random‐effects model had little effect on the findings. We judged the quality of the evidence to be very low according to GRADE criteria, as shown in summary of findings Table for the main comparison.

Figure 3

---

---

Forest plot of comparison: 1 Absorbable versus non‐absorbable sutures: open and endoscopic carpal tunnel decompression, outcome: 1.1 Postoperative pain (10 days) after CTD.

Postoperative pain at 6 weeks

This outcome was reported in four trials of wound closure following OCTD (Erel 2001; Kharwadkar 2005; Menovsky 2004; Theopold 2012). The SMD in pain in absorbable versus non‐absorbable suture groups indicated little or no difference between suture types at six weeks (SMD 0.06, 95% CI ‐0.72 to 0.84; 4 trials, N = 175; I² = 84%; Analysis 1.2, Figure 4). The I² suggests substantial heterogeneity, probably arising from variation in surgical techniques and outcome measures. Analyses using fixed‐effect versus random‐effects models had little effect on the findings. The quality of evidence was very low (summary of findings Table for the main comparison).

Figure 4

---

---

Forest plot of comparison: 1 Absorbable versus non‐absorbable sutures: open and endoscopic carpal tunnel decompression, outcome: 1.2 Postoperative pain (6 weeks) after open CTD.

Secondary outcomes

Postoperative hand function

Kharwadkar 2005 (N = 36) was the only study to report hand function, which was measured using the Carpal Tunnel Syndome Functional Status Scale (CTS‐FSS). The report provided mean scores and standard deviations (SDs) for each group pre‐operatively, at 2 weeks (MD ‐0.10, 95% CI ‐0.53 to 0.33), 6 weeks (MD 0.00, 95% CI ‐0.39 to 0.39), and 12 weeks (0.00, 95% CI ‐0.37 to 0.37), suggesting no differences between the groups (Analysis 1.3). The evidence is very low quality (summary of findings Table for the main comparison).

Scar satisfaction and wound inflammation

Data for this outcome were provided in Kharwadkar 2005 and Erel 2001.

The risk ratio (RR) for wound inflammation with use of absorbable versus non‐absorbable sutures after OCTD was 2.28 (95% CI 0.24 to 21.91; 2 RCTs, N = 95; I² = 90%), favouring non‐absorbable sutures (Analysis 1.4; Figure 5). The I² suggests substantial heterogeneity, probably arising from variation in surgical techniques and outcome measures. The quality of evidence was very low (summary of findings Table for the main comparison).

Figure 5

---

---

Forest plot of comparison: 1 Absorbable versus non‐absorbable sutures: open and endoscopic carpal tunnel decompression, outcome: 1.3 Wound inflammation.

Scar satisfaction was not measured.

Postoperative adverse events

We were unable to perform a narrative review or pooled analysis of postoperative and adverse events as these outcomes were not reported in the included studies.

Absorbable versus non‐absorbable sutures after endoscopic carpal tunnel decompression