Summary of findings for the main comparison. Surgical versus conservative interventions for treating fractures of the middle third of the clavicle.

Surgical versus conservative interventions for treating fractures of the middle third of the clavicle
Patient or population: adults with acute fracture of the middle third of the clavicle Settings: hospital (tertiary care) Intervention: surgery (e.g. plate fixation or intramedullary fixation) Comparison: conservative intervention (e.g. arm sling or figure‐of‐eight bandage)
Outcomes	Illustrative comparative risks* (95% CI)		Relative effect (95% CI)	No of participants (studies)	Quality of the evidence (GRADE)	Comments
	Assumed risk	Corresponding risk
	Conservative treatment	Surgical treatment
Function (overall) Various toolsa; higher values mean better function (0 to 100 scale, 100 is best function) Follow‐up: 12 months	Mean (SD) population Constant score 89 (7)b	The standardised mean difference in function (overall) in the surgery groups was 0.33 standard deviations higher (0.02 lower to 0.67 higher)		838 (10 studies)	⊕⊕⊝⊝ lowc,d	The SMD resultb equates to a mean improvement of 2.3 points (0.14 points worse to 4.69 points better) in favour of surgery on the Constant scale (0 to 100 scale; higher scores mean better function). This does not represent a clinically important difference. The MCID of Constant scale ranges from 5.7 to 10.4 points.
Pain VAS: 0 to 100 mm (0 is no pain and 100 mm is the worst score) Follow‐up: 12 months	The mean pain in the control groups was 7 mm	The mean pain in the surgery groups was reduced by 0.6 mm (‐3.51 lower to 2.31 higher)		277 (3 studies)	⊕⊕⊝⊝ lowc,e	This is not a clinically important effect. The MCID of VAS for shoulder injury is 14 mm.
Treatment failuref Indication for non‐routine secondary surgery Follow‐up: 12 months	132 per 1000g	43 per 1000 (27 to 66)	RR 0.32 (0.20 to 0.50)	1197 (12 studies)	⊕⊕⊝⊝ lowc,e	The main source of treatment failure in the conservative treatment group was symptomatic non‐union (61 cases/527 (11.6%); 10 studies) and in the surgical group it was mechanical failure (16/468 (3.4%); 7 studies). Overall treatment failure in the conservative group is dominated by the results of 2 studiesh (289 participants): 33/77 = 43%. 2 studies (118 participants) reported that they had no treatment failures in either group.
Cosmetic problemi Follow‐up: 12 months	218 per 1000g	120 per 1000 (68 to 214)	RR 0.55 (0.31 to 0.98)	1130 (11 studies)	⊕⊝⊝⊝ very lowc,e,j	The derivation of this outcome (number of people with ≥ 1 cosmetic problems) varied across the studies. Cosmetic deformity and/or asymmetry was greater in the conservative group and hardware prominence not requiring implant removal and scarring occurred only in the surgical group.k
Total of participants with adverse event outcomel Follow‐up: 12 months	107 per 1000g	144 per 1000 (73 to 282)	RR 1.34 (0.68 to 2.64)	1317 (14 studies)	⊕⊝⊝⊝ very lowc,e,m	The derivation of this outcome (number of people with ≥ 1 adverse events) varied across the studies. Hardware, removed likely due to discomfort, was most frequent in the surgical group (52/508 (10.2%); 9 studies), and symptomatic malunion (45/397 (11.3%); 9 studies), in the conservative group. Infection and/or dehiscence occurred only in the surgical group (22/686 (3.2%); 11 studies).n
Health‐related quality of life: SF‐36 or SF‐12 physical component score at 1 year (0 to 100 points, 100 is best)	The mean health‐related quality of life: SF‐36 or SF‐12 physical component score at 1 year was 54.7	MD 0.30 higher (1.95 lower to 2.56 higher)		321 (2 studies)	⊕⊕⊝⊝ lowc,e	This does not represent a clinically important difference. The MCID of SF‐36 ranges from 2.0 to 7.8 points (scale 0 to 100).
*The basis for the assumed risk for dichotomous outcomes is provided in footnotes. The corresponding risk (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; MCID: minimal clinical important difference; MD: mean difference; RR: risk ratio; SD: standard deviation; SMD: standardised mean difference; VAS: visual analogue scale
GRADE Working Group grades of evidence: High quality: we are very confident that the true effect lies close to that of the estimate of the effect. Moderate quality: 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 quality: our confidence in the effect estimate is limited: the true effect may be substantially different from the estimate of the effect. Very low quality: we have very little confidence in the effect estimate: the true effect is likely to be substantially different from the estimate of effect.

^aPatient‐reported functional scores were Constant score (7 studies), the Disability of the Arm, Shoulder, and Hand questionnaire (DASH; 1 study), the UCLA score (1 study) and Single Assessment Numeric Evaluation (SANE) score (1 study).
 ^bThis is based on the mean (SD) Constant score in a healthy population as reported in Yian 2005; mean improvement in Constant score back‐translated from SMD using the SD from the healthy population.
 ^cAll studies had methodological flaws; only two adequately concealed treatment allocation, and all were susceptible to detection bias in the measurement of self‐reported outcomes due to lack of blinding of the participant. All outcomes were downgraded one level for serious risk of bias.
 ^dThere was considerable heterogeneity (I² = 83%), and thus we downgraded the quality of the evidence by one level for serious inconsistency.
 ^eTotal number of events and/or number of participants were small, and thus we downgraded the quality of the evidence by one level for serious imprecision.
 ^fTreatment failure, measured by the number of participants who had undergone or were being considered for a non‐routine secondary surgical intervention for symptomatic non‐union, malunion or other complications (e.g. mechanical failure defined as a condition in which an implant loses its capacity to carry a load). This does not include plate removal for hardware irritation.
 ^gThe assumed risk was the mean risk in the control group.
 ^hThese two studies had 13 of the 16 symptomatic malunions: 9 were from one study. This is an unusually high number: most malunions are asymptomatic or with mild symptoms with no need for other types of treatment.
 ⁱThe number of people with one or more cosmetic problems, primarily deformity, asymmetry and/or shoulder droop, hardware prominence not requiring removal, unsightly scar and bump at fracture site. These data were often provided for individual outcomes and we applied a rule to avoid potential unit of analysis issues.
 ^jThere was considerable heterogeneity (I² = 63%) with results from different studies spread on both sides of the line of no effect; thus we downgraded the evidence by one level for serious inconsistency.
 ^kCosmetic deformity and/or asymmetry was greater in the conservative group (21/398 versus 93/361; RR 0.16, 95% CI 0.05 to 0.46). Hardware irritation not requiring implant removal (28/308), and unsightly but otherwise not problematic scars (14/273) occurred only in the surgical group.
 ^lTotal of participants with adverse event outcome: one or more of local infection, dehiscence, discomfort leading to implant removal, skin and nerve problems, stiffness, refracture, symptomatic malunion and other adverse events (e.g. thrombosis, shoulder impingement, cardiovascular event and transient plexus irritation).
 ^mThere was considerable heterogeneity (I² = 72%) with results from different studies spread on both sides of the line of no effect; thus the evidence was downgraded one level for serious inconsistency.
 ⁿSkin and nerve problems were also more common after surgical treatment (75/338 versus 17/310; RR 4.86, 95% CI 1.85 to 12.76; 6 studies).
 ^oHealth‐rerelated quality‐of‐life score was the physical component of SF‐36 and SF‐12; the two studies did not report the full score.