A five-item score predicted osteonecrosis better than the Garden classification

Design

retrospective multicentre comparative study, LASSO-penalised then multivariable logistic regression, internally derived score, Level III

Population

1,012 adults aged 18 to 65 with femoral neck fracture fixed with three cannulated screws at three Chinese centres, minimum 2-year follow-up

Primary outcome

osteonecrosis of the femoral head

Effect

4% osteonecrosis at scores 0-3 vs 42% at 6-8; AUC 0.80 (95% CI 0.77-0.83) vs Garden 0.70 (0.67-0.73)

Seven surgeons at three Chinese centres fixed 1,288 femoral neck fractures in adults aged 18 to 65 with three cannulated screws between 2012 and 2023; 1,012 had complete imaging and at least two years of follow-up. Osteonecrosis of the femoral head developed in 21%. Fifteen candidate variables were shrunk by LASSO-penalised regression, and the five that survived were weighted into an eight-point score.

Displacement dominates it: displaced fracture carried an odds ratio of 7.9 (95% CI 5.1-12.1) and three of the eight points. A superior subcapital fracture line scored two (OR 3.5), and an inferior subcapital line (OR 3.0), comminution (OR 2.4) and age 45 or over (OR 2.1) one each. Osteonecrosis occurred in 4% of patients scoring 0 to 3 and 42% of those scoring 6 to 8. Discrimination was better than the Garden classification, AUC 0.80 (0.77-0.83) against 0.70 (0.67-0.73), and among displaced fractures a score of 6 or more still separated risk (OR 5.5, 3.2-9.5).

The thresholds were chosen for different jobs and behave differently. At 4 or more the score is a rule-out — 95% sensitivity, 96% negative predictive value, but 34% specificity, so it labels most patients as at risk. At 6 or more it is a rule-in of sorts, 75% sensitivity and 74% specificity, but positive predictive value is only 42%.

That makes this a counselling and follow-up tool rather than a decision rule. It has not been externally validated, the authors say so, and it was derived in a population selected for head preservation. But it gives numbers to a conversation usually conducted in adjectives: a young patient with an undisplaced fracture and a low score can be told 4%, and one scoring 6 or more can be told that two in five go on to osteonecrosis and what the arthroplasty conversation would look like.

• Score the five variables preoperatively — displacement, superior and inferior subcapital fracture line, comminution, age 45 or over
• Use the score for counselling and follow-up planning, not to decide between fixation and arthroplasty
• Quote 4% at a score of 0 to 3 and about 42% at 6 to 8, and say the score is not yet externally validated
• Plan longer imaging follow-up for high scorers — osteonecrosis can present well beyond the first year
• Do not replace the Garden classification with it; the score complements the classification rather than superseding it

Why it matters

The osteonecrosis conversation before screw fixation in a working-age adult has until now been conducted without a number.

Don't overread it

This is an internally derived score from retrospective single-country data with no external validation — it should not yet drive the choice of operation.

The statistics, in plain English

An area under the curve of 0.80 against 0.70 means the score ranks two randomly chosen patients correctly 80% of the time rather than 70% — better, but far from certainty. The two thresholds trade off in opposite directions: at 4 points a negative predictive value of 96% makes a low score genuinely reassuring, while specificity of 34% means most patients score above it. At 6 points the positive predictive value of 42% means fewer than half of those flagged will develop osteonecrosis. Because the score was built and tested in the same dataset, its real-world performance will almost certainly be worse than these figures until someone validates it elsewhere.

Read the rest in the app

You have read your two free briefings this month. The app carries all 27 specialties, every morning, free — and this finding is waiting in it.

Get Daily Doctor — free

Scan to keep reading on your phone. No account needed to start.