PROFHER Explained: What the Trial Really Says About Proximal Humerus Fractures

Should a displaced proximal humerus fracture be fixed? Most clinicians answer with the PROFHER trial, the UK randomised trial published in JAMA in 2015, and many quote it for fractures it never studied. This post sets out what PROFHER randomised, what it measured, what the five-year follow-up added and where its conclusions stop. It builds on our proximal humerus fractures topic page. The short answer: for adults with a displaced fracture involving the surgical neck, surgery as practised in the trial gave no better Oxford Shoulder Score than a sling and physiotherapy at two or five years, and it cost more.

What question did PROFHER ask?

PROFHER (PROximal Fracture of the Humerus: Evaluation by Randomisation) was set up to test a trend. Surgery for these fractures was increasing, partly because locking plates had become available, but there was little reliable trial evidence that it helped. The question was whether surgery improved patient-reported outcome compared with non-surgical treatment in adults with a displaced fracture of the proximal humerus involving the surgical neck.

It was a pragmatic, multicentre, parallel-group randomised trial funded by the NIHR Health Technology Assessment programme. The word pragmatic matters. PROFHER tested a treatment policy as delivered in ordinary NHS hospitals, not one operation performed by one expert. It sits with DRAFFT, FLOW and the other trials on our landmark trials page.

Who was in the trial?

250 patients were recruited from 32 NHS hospitals between September 2008 and April 2011, all within three weeks of injury. Mean age was 66 years (range 24 to 92), 192 (77%) were women and 249 (99.6%) were white. 108 patients (43%) were under 65, so this was not a trial of the frail elderly alone. Three-quarters of the injuries (186 of 250) were falls from standing height or less.

Head-splitting fractures were not on the exclusion list, but the investigators expected that these rare injuries would not be judged eligible.

What the fractures looked like. Two shoulder surgeons independently classified the baseline radiographs using Neer's displacement thresholds (more than 1 cm of translation or 45° of angulation). They found 18 one-part, 128 two-part and 104 three- or four-part fractures. The one-part fractures were surgical-neck fractures that fell short of the Neer thresholds but that the treating surgeon still considered for surgery. Neer is only moderately reliable between observers, with interobserver kappa of 0.48 to 0.52 in Sidor's study, so treat these counts as approximate.

The selection funnel. Of 1,250 patients identified with a suitable fracture, 563 were eligible and 250 consented. Another 117 otherwise eligible patients were not approached because their surgeon had a firm treatment preference. PROFHER therefore describes fractures where the treating surgeon was genuinely uncertain, which is most of them but not all.

What did each arm actually receive?

Surgery was left to the treating surgeon, who had to use a technique they were experienced with. Of the 125 patients allocated to surgery, 109 had an operation and 16 were treated non-surgically: 8 changed their minds, 6 became unfit for surgery and in 2 the surgeon's opinion differed. A plate and screws were used in 82.6% of operations (a PHILOS plate in 66) and a hemiarthroplasty in 10 (9.2%). Surgery took place a median of 9 days after injury. A reverse shoulder arthroplasty was used only once in the main trial, as secondary surgery for malunion.

Non-surgical care was a sling or collar and cuff, followed by active rehabilitation; 123 of 125 patients received it. Both arms were given the same information leaflet on sling care and the same physiotherapy protocol. In effect the trial compared surgery plus rehabilitation with rehabilitation alone.

What did PROFHER find?

The primary outcome was the Oxford Shoulder Score (OSS), scored 0 to 48 with higher scores better, collected at 6, 12 and 24 months and analysed over the two years. The sample size was based on a minimal clinically important difference of 5 points.

Neither prespecified subgroup changed the answer. The investigators expected surgery to help more in patients under 65 and in fractures involving one or both tuberosities, but neither interaction was significant (P = 0.27 for age, P = 0.95 for tuberosity involvement). Surgery cost on average £1,780.73 more per patient at two years and had only a 5% probability of being cost-effective at £20,000 per quality-adjusted life-year.

Is PROFHER a negative trial or just underpowered?

Examiners often test this distinction. A non-significant P value alone does not prove that two treatments are equivalent, because a small trial can miss a real effect. PROFHER's case rests on the confidence interval. The upper limit for the two-year difference, 2.84 points, sits below the 5-point threshold the trial was designed to detect, and every 95% CI excluded that prespecified difference. The data rule out a clinically important average benefit from surgery in this population. With 215 patients complete at two years against a target of about 200, the trial reached its planned power. Our statistical power and sample size page covers the reasoning.

What the interval does not rule out is a smaller benefit, or a benefit confined to a subgroup too small to detect. The six-month result hints that surgery may speed early recovery a little, but even that interval stopped short of 5 points, and the gap had gone by one year. If the confidence-interval logic is unfamiliar, read statistics for the surgeon before the viva.

Did the result hold at five years?

The extended follow-up, reported by Handoll and colleagues in 2017, sent postal questionnaires at three, four and five years to the 176 participants who consented. OSS data were available for 164, 155 and 149 of them. There was no statistically or clinically significant difference at any point. At five years the adjusted difference was −1.09 points (95% CI −3.41 to 1.23, P = 0.36), nominally favouring non-surgical care. Nobody had secondary shoulder surgery for a new complication, so secondary surgery stayed at 11 patients in each arm, and EQ-5D quality of life did not differ.

Two caveats apply. The extended study fell 11 patients short of its target of 160, so it did not reach its revised power, although the authors argue the balanced losses and narrow intervals make this unlikely to matter. More further fractures were reported in the non-surgical arm (50 against 31). The authors considered it unsafe to attribute this to treatment and suggested it was mainly chance, with more smokers in that group. Do not quote it as a harm of non-operative care.

What PROFHER does not show

The trial answers one question well. Most misuse comes from carrying it into fractures and patients it did not include.

Fracture-dislocations and open fractures. Both were excluded by design, so PROFHER provides no evidence about them. Our shoulder fracture-dislocations page sets out how those injuries are managed.

Head-splitting and articular fractures. These were not formally excluded but were expected to be ineligible, and the trial offers no usable evidence for them. See humeral head fractures.

Isolated tuberosity fractures. Inclusion required surgical-neck involvement, so an isolated greater tuberosity fracture is outside the trial. The topic page gives its much tighter fixation threshold: more than 5 mm of displacement, or more than 3 mm in an overhead athlete or manual worker.

Patients with a clear surgical indication. Nerve injury, severe soft-tissue compromise and multiple upper-limb injuries were excluded, as were the 117 patients whose surgeon had already decided. PROFHER does not cover every displaced fracture seen in clinic, only those where a reasonable surgeon was unsure.

The young high-energy patient. Younger adults were included and the age-by-treatment interaction was not significant, but most injuries were low-energy falls. The 2022 Cochrane review states that evidence is absent or insufficient for people under 60 and for high-energy trauma.

A specific operation. The surgical arm was mostly locking plates with a few hemiarthroplasties, chosen by the surgeon. A null result for surgery as a policy is not a null result for each operation, and it says nothing about primary reverse shoulder arthroplasty, which was not used.

Other health systems. The cohort was 99.6% white, treated in the NHS with NHS rehabilitation pathways. Where structured physiotherapy is less available, the non-surgical arm may not perform the same way.

What are the main criticisms of PROFHER?

A 2017 survey of UK surgeons found that critics raised selection bias, the exclusion of complex fractures, the absence of reverse arthroplasty and the pragmatic design. Ten of 265 respondents (4%) were highly critical. The table puts each criticism next to what the trial reports.

The investigators answered the reverse arthroplasty point in their report: reverse arthroplasty is more costly and more invasive, and its benefit for this population was unproven. In their view, the onus is on those who argue the results no longer apply to show the newer operations are better.

How did PROFHER change practice?

In the UK, the 2017 survey of 265 surgeons from the British Orthopaedic Association and the British Elbow and Shoulder Society found that 137 (about half) had changed practice to some extent because of PROFHER, by operating on fewer eligible fractures. A third of those who had not changed (43 of 128) were already treating these patients non-operatively.

In the wider evidence, the 2022 Cochrane review pooled 10 trials (717 participants) comparing surgery with non-surgical treatment for displaced fractures. It found high-certainty evidence of no clinically important difference in patient-reported shoulder function at one year (SMD 0.10, 95% CI −0.07 to 0.27) and two years. It also found low-certainty evidence of more additional surgery after operative treatment (RR 2.06, 95% CI 1.21 to 3.51), an illustrative 38 extra operations per 1,000 surgically treated patients.

In the United States, a database study of 2011 to 2019 found no significant change in the rate of operative treatment after 2015. Publishing a trial is not the same as changing practice.

What comes next. PROFHER-2, whose protocol was published in 2023, randomises adults over 65 with three- or four-part fractures, with or without dislocation, between reverse arthroplasty, hemiarthroplasty and non-surgical care, with a target of 380 patients and the OSS at 24 months as the primary outcome. Check whether it has reported before you quote it.

Who should still be considered for surgery?

PROFHER supports a default, not a rule. For the typical patient, an adult with a displaced surgical-neck fracture after a low-energy fall, the discussion should start with non-surgical care and present the trial results honestly. Outside that group, the topic page's algorithm still offers fixation with a plate or nail for a two-part surgical-neck fracture in a young or active patient. It also prefers reverse shoulder arthroplasty to hemiarthroplasty in many older patients with three- or four-part, head-split or unsalvageable fractures when surgery is chosen. Fracture-dislocations, open injuries, neurovascular injury and a displaced greater tuberosity fragment are decided on their own merits. When fixation is chosen, the locking plate operative guide covers the technique.

In the exam

Examiners want the population, the outcome and the limitation, not just the acronym. Our guide to structuring viva answers shows the general pattern.