Criteria for considering studies for this review
Types of studies
We will include all randomised controlled trials (RCTs) and quasi-RCTs. We will include published and unpublished reports and abstracts if they meet our selection criteria. We will not exclude cross-over studies. We will not apply any language restrictions. Non-randomised evidence is likely to be important and we will include such evidence in the Discussion. If we find no RCTs or quasi-RCTs, we will consider all other types of publications for a narrative approach to the current literature, but will make no attempt at a meta-analysis of non-randomised studies.
Types of participants
We will include people with a clinical diagnosis of tibial or deep peroneal nerve compression or entrapment as well as TTS in the review. We will not require participants to have had electrophysiological testing and we will accept the trial authors' definition of what constituted these conditions. We will include people of either sex, pregnant women and all age groups including children.
We will report results for tibial and peroneal nerves separately and perform separate analyses where appropriate.
Types of interventions
We will include studies comparing the following surgical interventions:
We will include any of the above surgical intervention(s) in the following comparisons:
surgical intervention(s) versus conservative management or no intervention;
surgical intervention(s) versus surgical intervention(s), comparing the various forms of surgical intervention including decompression, nerve grafting or neurolysis;
surgical intervention(s) versus medical or other active management.
Interventions in the active arm of the trial other than the intervention(s) under study must be identical in the control arm.
Types of outcome measures
We do not plan to use outcomes as criteria for including studies in the review.
Our primary outcome will be pain and other 'positive' sensory symptoms, including paraesthesia, dysaesthesias, hyperaesthesia, allodynia, hyperalgesia and causalgia.
We will measure this outcome as either dichotomous, that is as the presence or absence of symptoms, or as continuous, which may involve the use of a visual analogue scale (VAS).
We will assess these symptoms at approximately one year or longer after surgery.
'Negative' symptoms including numbness, hypoalgesia or anaesthesia. We will measure this outcome as either dichotomous, that is the presence or absence of symptoms, or as continuous, which may involve the use of a VAS.
Function, including time to return to work or resumption of normal activities of daily living (ADL).
Quality of life scales. The measurement of this outcome might involve the use of established scales which can be adapted, such as the SF-36 (Rothwell 1997) or the Nottingham Health Profile (NHP).
Overall improvement of TTS and ATTS. Studies may measure this outcome dichotomously, from the patients' perspective, by inquiring whether there has been an overall improvement in symptoms.
Adverse events as a result of surgery.
We will assess the secondary outcomes one year or longer after surgery.
We plan to present the findings of the review in a 'Summary of findings' table created using GRADEPro software (GRADEpro). We will use the GRADE approach (Schünemann 2011) to assess the quality of the evidence. We will rate the evidence for each outcome as of high, moderate, low, or very low quality and provide the rationale for these decisions. We will consider downgrading evidence for the following reasons: risk of bias; imprecision; indirectness; unexplained heterogeneity or publication bias. Reasons for upgrading evidence will be: a large magnitude of effect; if all plausible confounding tends to underestimate an apparent intervention effect; and a dose-response gradient. The following outcomes will be included: 'positive' sensory symptoms, 'negative' sensory symptoms, overall symptom improvement and adverse events. There will be a table for each comparison provided that data are available for at least one of these outcomes.
Search methods for identification of studies
With the assistance of the Cochrane Neuromuscular Diseases Group (CNMDG) we will search the Cochrane Neuromuscular Disease Group Specialized Register, CENTRAL (The Cochrane Library current issue), MEDLINE (1966 to present), EMBASE (1980 to present), CINAHL Plus (1937 to present) and AMED (1985 to present) to identify all the RCTs available evaluating the efficacy of surgical interventions in tibial and deep peroneal nerve entrapment and compression syndromes.
The search strategy for MEDLINE is in Appendix 1.
We will undertake a final search of MEDLINE, EMBASE, CINAHL Plus and AMED within six months prior to publication. We will also explore free Internet search engines such as Google Scholar. There will be no language restrictions and where possible we will translate potentially eligible foreign language papers.
Searching other resources
We will scan relevant conference abstracts from meetings of the Peripheral Nerve Society (PNS) and the British Peripheral Nerve Society (BPNS). We will handsearch reference lists from trials and any relevant systematic reviews selected by electronic searching to identify any other relevant trials. We will contact authors to identify any additional published or unpublished data.
In order to minimise publication bias, we will perform a search of the grey literature. We will search the King's Fund health charity database, OpenSIGLE and the National Technical Information Service (NTIS) for relevant grey literature.
We will search for unpublished and ongoing trials. We will identify ongoing trials mainly from the following databases:
We will search other national and international trial registers if appropriate and the Clinical Trial Results web site (www.clinicaltrialresults.org/) for slide presentations reporting the results of relevant clinical trials.
We will search the National Institute for Health Research DARE (Database of Abstracts and Reviews of Effects) and HTA (Health Technology Assessments) databases to identify systematic reviews and assessments. We will search the NHS Economic Evaluation Database (NHS EED) for cost information concerning interventions, for inclusion in the Discussion.
Data collection and analysis
Selection of studies
Two review authors (NP and DL) will assess identified studies independently on the basis of title and abstract. They will resolve disagreements by discussion and where necessary contact the remaining authors for further assistance.
We will obtain the full text of potentially relevant studies. Two review authors (NP and DL) will independently assess the study reports according to the eligibility criteria. In cases of disagreement, we will put the study to all the other authors for consultation and discussion, who will come to a majority decision regarding inclusion. We will give particular care to the issue of multiple publications of the same data. We will report the study selection process in a PRISMA flow chart.
Data extraction and management
We will design a standardised data collection form which we will first pilot. The data extraction form will include methods, participants, interventions, comparators, outcomes, and note other trial details (for example, location, funding source, conflicts of interest).
Bilateral leg involvement of tibial and deep peroneal nerve palsy may be present in studies and where appropriate we will extract these data, with details of randomisation techniques as well as any statistical methods used to adjust for bilateral cases.
Once the review authors have finalised and agreed the data extraction form, two authors (CSH and DW) will independently collect data from included studies. The two review authors will compare the extracted data and will resolve differences by discussion. One review author will enter the data into the Cochrane software Review Manager (RevMan) 5 (RevMan 2012) and another author will check the data entry. We will resolve differences by discussion, with additional participation from other co-authors if necessary.
If necessary we will contact trial authors to obtain more detail about study characteristics.
Assessment of risk of bias in included studies
Two review authors (NP and DL) will independently assess the risk of bias in each study; with the participation of the remaining review authors if disagreements arise.
We will assess the methods, design and conduct of the included studies for any risk of bias. We will use the Cochrane 'Risk of bias' tool (Higgins 2011). We will consider :the following items: random sequence generation, allocation concealment, blinding of participants and personal, blinding of outcome assessors, incomplete outcome data (attrition bias) (including use of intention-to-treat analysis), selective reporting, and other bias.
The review authors' 'Risk of bias' assessments will be "High risk of bias", "Low risk of bias" or "Unclear risk of bias", where "Unclear risk of bias" means that there is insufficient information to form a judgement or that the risk of bias is uncertain. We will provide an explanation for each evaluation. We will also generate a 'Risk of bias' summary figure using RevMan 5 (RevMan 2012).
Measures of treatment effect
For dichotomous outcomes, such as presence or absence of pain, positive sensory symptoms and negative sensory symptoms, we will report the risk ratio (RR). In the case of rare dichotomous events the Peto method will be the most appropriate method of analysis and this involves calculation of the odds ratio (OR). For continuous outcomes, such as the time to resolution of activities of daily living, we will calculate the mean difference (MD). In cases where different scales are used for the measurement of continuous data we will use the standardised mean difference (SMD). We will report summary statistics with 95% confidence intervals (CIs) in all cases.
Unit of analysis issues
As cross-over or cluster-randomised controlled trials are unlikely, we do not expect any unit of analysis issues arising from these designs. Similarly, due to the nature of the interventions we do not expect multiple treatments, or repeated outcome measures.
For parallel trials comparing more than two treatments (i.e. more than two intervention arms) we will use only one comparison per meta-analysis to avoid any unit of analysis issues.
Bilateral leg involvement of the tibial and deep peroneal nerve palsy and subsequent surgical treatment could be present in the studies recovered. We will identify the studies with bilateral cases and we will report whether the analyses used in those studies are valid, based on their methods of randomisation. We anticipate three distinct scenarios and randomisation methods:
Randomisation of subjects (with both legs given the same treatment), where the comparison of treatments is made entirely between subjects, but which must take into account the fact that observations on legs from the same subject are not independent.
Randomisation of legs, with each subject’s legs allocated to different treatments. The comparison of treatments will be made entirely within subjects.
Randomisation of legs, with no constraint that each subject’s legs be given different treatments. This permits comparison of treatments both within and between subjects.
We will perform meta-analysis using inflated variances and the generic inverse-variance methods using RevMan, if appropriate.
Dealing with missing data
We will consider incomplete data to have been addressed in a study if the final analysis includes 85% or more of the participants. If the final analysis includes fewer than 85% of participants but trial authors have taken appropriate steps to ensure that the missing data do not bias the results, we will deem this satisfactory.
If it is not clear whether the study meets these requirements for missing data, we intend to calculate the statistics on an intention-to-treat analysis (ITT) basis. Where summary statistics are not available we intend to calculate these summary statistics (e.g. standard deviations) from the available data.
We will make every effort to contact authors and retrieve missing data.
Assessment of heterogeneity
We will explore clinical and statistical heterogeneity across studies before considering a meta-analysis. If there is evidence of strong heterogeneity among studies, we will not carry out a meta-analysis. We will assess statistical heterogeneity using the Chi2 test and the I2 statistic (Higgins 2011). For the Chi2 test we will set the level of significance at 90% because of the low power of the test. For the I2 statistic, we will assume that a value > 50% represents significant heterogeneity and we will use a random-effects model for the analysis. For high levels of heterogeneity (I2 statistic > 75%), we will not carry out meta-analysis but present the results on a forest plot without the pooled estimate.
Assessment of reporting biases
We will explore potential reporting bias through examination of a funnel plot if the number of included studies permits it, that is if there are at least 10 included studies (Egger 1997).
We will use the Mantel-Haenszel method (Greenland 1985; Mantel 1959) for the meta-analysis. We will use the Peto OR method to pool relevant ORs together if the number of events is small (we anticipate this to be the case when considering adverse events, as their occurrence is rare).
For continuous data we will calculate the SMD. We expect that studies will use different quality of life instruments and hence we expect to use SMDs for the quality of life analysis.
We will carry out all analyses based on the intention-to-treat principle (DerSimonian 1986).
Subgroup analysis and investigation of heterogeneity
We will investigate potential sources of heterogeneity using subgroup analysis when sufficient studies are available. We will evaluate the changes in effect sizes and heterogeneity across the subgroups. The planned subgroups are the following.
Types of surgical interventions: surgical interventions of neurolysis, grafting or decompression.
Other concomitant conditions, where data are available. These include physiological or metabolic conditions (for example, pregnancy or hypothyroidism), neuropathies (for example, those associated with diabetes mellitus) or space occupying lesions (for example, nerve sheath tumours or ganglion).
We will investigate the influence of within-study risk of bias on the overall effect estimate by carrying out sensitivity analysis for studies at low-to-medium risk of bias versus studies at high risk of bias.
We will also perform a sensitivity analysis on studies in which the diagnosis has been confirmed by neurophysiological findings versus those without neurophysiological confirmation.