Butt 1996b.

Study characteristics
Methods	Quasi‐RCT: allocation according to treatment preference of consultant on‐call at time of admission. 4 consultants involved in study, 2 opted for surgical treatment and 2 opted for non‐surgical treatment.
Participants	Setting: Russell's Hall Hospital, Dudley, UK Size: 42 participants; 20 in surgical group, 22 in non‐surgical group Recruitment period: January 1988 to March 1991 Baseline characteristics: surgical group: 20 participants, mean age 77.6 years; non‐surgical group: 22 participants, mean age 80.5 years; sex and fracture type not reported Inclusion criteria: aged > 60 years with displaced fractures of the distal femur Exclusion criteria: people who were physiologically unfit for surgery
Interventions	All participants received low‐dose warfarin as thromboprophylaxis. Participants were then allocated to either: surgical treatment: fracture was fixed with a DCS applied laterally, supplemented with bone graft if medial cortex was deficient. Continuous passive mobilisation prescribed for 48 hours postoperatively. Functional cast brace then applied when wound was healed, and person was mobilised with a walking frame. Intravenous antibiotics given preoperatively and 24 hours postoperatively; non‐surgical treatment: skeletal traction applied using a Denham pin and the limb placed in a Thomas splint with a Pearson knee flexion attachment. Splints removed at 3–4 weeks and Perkin's exercises were started. A functional cast brace applied at 6–8 weeks.
Outcomes	Schedule: not formally reported but until "union had been achieved". Outcomes: Schatzker and Lambert criteria (full extension, loss of flexion, valgus/varus/rotational deformity, pain, joint congruency), length of hospital stay, deep vein thrombosis, urinary tract infection, pneumonia, wound and pin tract infection, pressure sores, delayed union, malunion, loosening of implant or traction pin, death
Notes	Funding/sponsor/declarations of interest: authors reported no declaration of interest. No information provided on funding/sponsor. Attempt at personal communication unsuccessful
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	High risk	Quotes: "randomised controlled trial"; "prospective study"; "Each of the four consultants remained on call for a week at a time; two opted for operative treatment and the others treated all patients by traction". Comment: quasi‐randomised study.
Allocation concealment (selection bias)	High risk	Predictable allocation – allocation based on treatment preference of on‐call consultant.
Blinding of participants and personnel (performance bias) All outcomes	High risk	Study compared surgical vs non‐surgical treatment and so it was not possible to blind either participants or personnel. Each intervention carried out by a consultant surgeon: 2 consultants performed operative treatment and 2 consultants supervised non‐operative treatment. However, details of providers of the specific intervention not specified.
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	No mention of blinding of outcome assessment. Additionally, outcome assessment was poorly reported.
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	20 participants allocated to surgical treatment and 22 to non‐surgical treatment. 1 participant crossed over from surgical to non‐surgical group and 2 participants were excluded after allocation from the non‐surgical group. In the final per‐protocol analysis, 17 participants analysed in surgical group and 19 in non‐surgical group. Overall attrition 6/44, but the cross‐over gave rise to concern.
Selective reporting (reporting bias)	Unclear risk	No protocol available. Timing of outcome assessment not given.
Other bias	Low risk	We identified no other sources of bias.