Keizer 2002.

Methods	Randomised controlled trial. Pre‐specified sample size calculation not reported. Overall validity of results: high risk of bias.
Participants	40 participants, 19 male and 21 female, mean age 42.8 years (range 25 to 72 years), mean duration of symptoms 10.5 months (range 6 to 48 months). Inclusion criteria: lateral elbow pain, pain at the lateral epicondyle on resisted dorsiflexion of the wrist with elbow extended, pain for at least 6 months, and failure of conservative treatment. Exclusion criteria: previous surgery; nerve entrapment; pregnancy; systemic neuromuscular disorders such as myasthenia gravis, neuropathy of the elbow (diagnosed by electromyography).
Interventions	Group 1 (20 participants), 'Surgery group': surgical wrist extensor release (Hohmann operation), under Biers block anaesthesia, as an outpatient; 4cm curved incision over tip of lateral epicondyle and distal over the extensor carpi radius brevis to expose the extensor origin, the origin of the ECRB was incised transversely, just ventral of the lateral epicondyle, and the release continued to the synovium of the radiohumeral joint, the synovium was incised and the joint inspected for intraarticular lesions, then the wound closed, and covered with a compression bandage and sling for 2 weeks, then the sutures removed; following surgery participants were instructed to extend the elbow 3 times daily. Group 2 (20 participants), 'Botulinum toxin group': 30 to 40 units of botulinum toxin were injected into the ECRB; if insufficient paresis of the third and fourth fingers by 6 week follow up, a second injection was given.
Outcomes	Assessed at 6 weeks, 3 months, 6 months, 1 year and 2 years. 1) Pain: participant subjective self‐assessment of presence of pain 4 point scale; no, occasional, regular, always. 2) Subjective loss of grip strength: participant subjective self‐assessment of loss of grip strength on 4 point scale; no, slight, moderate, severe. 3) Decrease in pain on VAS scale: 4 categories; 100%, 80‐100%, 50‐80%, <50%. 4) Satisfaction: participant subjective self‐assessment on 3 point scale; satisfied, moderately satisfied, not satisfied. 5) Sick leave on 3 point ordinal scale; full‐time, part‐time, unable to work. 6) Overall result: outcome assessor rated as 'Excellent, Good, Fair, Poor' using the modified Verhaar Scoring system (combination of pain and patient satisfaction, e.g., for 'Excellent': patient had no pain on lateral epicondyle, 100% decrease on VAS pain score, no pain on palpation, no pain on resisted extension of middle finger, and was satisfied; 'Good': occasional pain on lateral epicondyle, 80% to 100% decrease on VAS, slight palpation pain, slight pain provoked by resisted extension of middle finger, patient satisfied; 'Fair': moderate pain on lateral epicondyle, 50% to 80% decrease in VAS, moderate palpation pain, moderate pain provoked by resisted extension of middle finger, patient moderately satisfied; 'Poor': patient had pain on the lateral epicondyle, <50% decrease on VAS, severe palpation pain, severe pain on resisted extension of middle finger, patient dissatisfied). 7) Physical examination: includes range of motion, dichotomized as normal or limited (>5º); mean grip strength (with a dynamometer, elbow fully extended); local tenderness, resisted extension of wrist and resisted extension of middle finger categorised on 4 point scale (no, slight, moderate, severe).
Sources of funding	Not reported
Notes	VAS, visual analogue scale The review authors made an ad hoc decision not to extract any outcomes, as data were not in a form easily amenable to meta‐analysis in RevMan (e.g., 4 point ordinal scales used for 'Overall result', and 3 point scale used to measure patient satisfaction not able to be dichotomized without losing information; mean but no variance reported for grip strength, range of motion only reported as dichotomized data); and it is unlikely that other (future) studies will use the same outcome measures (e.g., overall result categorised in the same way). Acknowledgements: JPH Reulen, Department of Clinical Neurophysiology, University Hospital, Maastricht, the Netherlands, for statistical advice.
Risk of bias
Bias	Authors' judgement	Support for judgement
Adequate sequence generation?	Unclear risk	Comment: sequence generation not described
Allocation concealment?	Unclear risk	Comment: not reported
Blinding? All outcomes	High risk	Comment: not reported, but the intervention‐providers and participants were most likely unblinded due to the nature of the interventions
Incomplete outcome data addressed? All outcomes	High risk	Comment: At 12 months 2/20 missing from the surgery group (refused assessment due to poor outcome); the authors did not impute data from these 2 lost participants; at 24 months, 1/20 from the botulinum toxin group was lost to follow up, the authors used 'last outcome carried forward', LOCF (from 12 month assessment for the 24 months evaluation of overall result (scored 'excellent'). This is likely to overestimate the treatment effect for the botulinum toxin group. Sick leave: numbers missing vary across time points, no reasons given
Free of selective reporting?	High risk	Comment: The trial authors measured pain on a continuous VAS scale, but report data only as decrease in pain on a 4 point ordinal scale; adverse events not reported (unclear if measured); range of motion, although measured on a continuous scale is only reported as a proportion of participants with normal or limited range
Free of other bias?	Low risk