Avascular Necrosis of the Femoral Head

If you have deep groin pain that started for no reason, and you have ever taken a serious course of steroid tablets, had a transplant, have sickle cell disease, or drink heavily, book an appointment this week and ask specifically about an MRI.

A normal X-ray does not rule this out. In early disease the X-ray is normal by design, and a screening scan found this condition in 6 out of 100 people who had no symptoms at all. If you develop a fever, cannot put any weight on the leg at all, or feel systemically unwell, that is a different emergency and needs same-day care.

Takes one phone call. No equipment needed.

The Verdict

Deep groin pain plus a normal X-ray plus a steroid history is an MRI, not a stretching plan.

The ball of your hip is fed by essentially one artery with no backup, like a room wired with a single power cable. Cut that cable and the bone inside quietly dies, but nothing looks different from the outside for months, because the shape is still intact. It only fails when the dead bone underneath finally fatigues and the joint surface caves in, the way a roof holds fine until the beams beneath it rot through, and then goes all at once.

What this actually is: the bone under the surface of your hip joint has lost its blood supply and died, and dead bone cannot repair itself.
What most people get wrong: a normal X-ray feels like the all-clear, and for this one condition it means almost nothing, because the X-ray only changes once the damage is already done.
Start here: ask for an MRI, and ask for the size and location of the affected patch, because a small patch on the inner side of the joint behaves completely differently from a large one on the outer side.

Best for

Anyone with new unexplained deep hip or groin pain who has taken significant steroids, had a transplant or chemotherapy, has sickle cell disease, drinks heavily, or has previously broken or dislocated that hip.

Skip if

Your hip pain followed an obvious injury with no risk factors, or you have already had a normal MRI of the hip.

Want the full evidence? Keep scrolling

What Works

Ranked by evidence strength. Read the scope note first: the top two tiers contain almost nothing a physical therapist owns, and that is an accurate picture of this evidence base rather than an omission.

Cinematic anatomical rendering of the hip joint and femoral head

Tier 1 · Strong evidence

Timely MRI, with lesion size and location recorded. HIGH 59% of 664 untreated hips with no symptoms progressed to pain or collapse, and small inner-side patches collapsed under 10% of the time (Mont 2010, 16 studies). Screening MRI found the condition in 6 of 100 people with no symptoms at all (95% CI 2.2 to 12.6, Tervonen 1992). This is the highest-value act available here, and any first-contact clinician can do it.
Urgent hospital reduction of a dislocated hip. MODERATE Osteonecrosis rose from 4.34% to 8.69% to 21.73% as time to reduction went from 6 to 12 hours, to 12 to 24 hours, to beyond 24 hours (p=0.030, N=44, Milenkovic 2022). Outside physical therapy scope, listed because recognising it and getting the patient to hospital is not.

Exercise Prescription

There is no research showing that any exercise, stretch, or amount of walking changes what happens to the bone in this condition. A dedicated literature search for protected weight-bearing returned zero papers, and the only weight-bearing report in 67 papers is three adolescents from 1985. Every dose below is general clinical practice, not evidence. Your specialist's instructions override this page. This is a plumbing problem in the blood supply to a bone, not a weak-muscle problem.

Gentle hip rotations 2 × 10 each side · daily general practice Lie on your back, knees bent, feet flat. Let one knee roll slowly outward and back to the middle. Should feel like a stretch, never sharp.

Heel slides 2 × 10 each side · daily general practice Lie on your back. Slide one heel slowly up towards your bottom, then back down. Move only as far as is comfortable.

Seated marching 2 × 10 each side · daily general practice Sit tall in a chair. Lift one knee a few inches, lower it, then the other. Slow and controlled.

Upper body and other-leg strength work as tolerated · 2 to 3 × weekly general practice Keep training everything that does not load the painful hip. Normal training rules apply to the parts that are not affected.

Pool walking, if available 10 to 20 minutes · 2 to 3 × weekly general practice Chest-deep water takes most of the load off the joint while letting you move normally.

Tier 2 and Tier 3 interventions

Tier 2, moderate evidence.

Core decompression for stage I disease. MODERATE 84% success at stage I, falling to 63% at stage II and 29% at stage III, and better than conservative care at stage I only (Castro 2000). Non-randomised pooled comparison.
BMAC augmentation of core decompression. MODERATE 120-month survival free of hip replacement 69.4% versus 48.5% overall (p<0.0005), and BMAC was the only orthobiologic whose benefit survived direct comparison with matched controls (Zaffagnini 2025, 106 studies, N=4,505). The stem-cell advantage did not hold beyond 5 years (Li 2023), so the honest framing is delay rather than prevention.
Total hip replacement after collapse. MODERATE The endpoint every survival curve here is measured against. About a third of osteotomies convert to hip replacement over roughly 7 years (268 of 852, Quaranta 2021).

Tier 3, weak or expert opinion only.

Bisphosphonates (alendronate). LOW Positive at 10 years, but against historical controls with no concurrent comparison (Agarwala 2011); the supporting review grades its own conclusion at level 3a (Luo 2014).
Long-term anticoagulation in inherited clotting disorders. LOW 0 of 13 hips collapsed over a mean 12 years, in 9 patients selected for a specific inherited trait (Glueck 2020). Mechanistically coherent, and it says nothing about unselected disease.
Pulsed electromagnetic fields. LOW Best responses in early stages across 10 studies; direct current and capacitive coupling ambiguous (Al-Jabri 2017).
Platelet-rich plasma. LOW 17 studies reviewed narratively with no pooled effect estimate (Han 2020), and not confirmed against controls in the survival analysis.
Acupuncture and moxibustion. LOW 9 randomised trials, N=630, improved scores as an add-on to conventional care (Jin 2021). Unblindable, composite subjective endpoints, no collapse or surgery endpoint reported.
Symptom-directed physical therapy. LOW Range maintenance, gait aids for comfort, general conditioning. No trial exists. Prescribed for comfort and function, explicitly not for disease modification.

What Doesn't Work

Treating a normal X-ray as a negative result. The commonest way this condition is missed, and the X-ray is normal in early disease by design.
Protected weight-bearing prescribed as disease modification. A dedicated search returned zero papers. Crutches may help pain. Nobody has shown they save a femoral head, and presenting them as treatment is how the specialist referral gets delayed.
A hip strengthening and mobility programme for presumed impingement in a patient with a risk factor and a normal X-ray. Not harmful in itself. It consumes the window.
Reassuring a patient about their steroid course with a population number. Dose sets population risk and genetics decides individual susceptibility, so no individual prediction exists.
Using a general hip questionnaire to find cases. Oxford Hip Score and EQ-5D did not separate a group containing 3 cases of osteonecrosis from a group containing none.
Carrying "stage" without lesion size and location. It discards the variable doing the prognostic work.

Red Flags

Refer, do not rehabilitate. There is a clock on this one.

Dark cinematic rendering of the hip and pelvis

New unexplained deep groin, buttock or front-of-thigh pain in anyone with a risk factor (steroids, sickle cell disease, transplant, chemotherapy, clotting disorder, heavy alcohol use, HIV, prior hip fracture or dislocation). In stage II hips, all became painful within 2 years, all collapsed, and the mean gap between pain starting and the joint caving in was 11 months (Hernigou 2006).
A normal plain X-ray does not exclude this. A screening scan found disease in 6% of a group with no symptoms at all. A normal X-ray in a symptomatic at-risk patient is not a negative result.
Sudden change in pain, new catching or giving way, or a new limp in a known lesion. Possible collapse. Re-image, do not progress the programme.
Hip pain at any interval after a previous hip dislocation or fracture, including years later. The surveillance window is long.
The other hip needs checking. Bilateral involvement is common, and the entire natural-history literature is built on symptom-free second hips.
Fever, feeling systemically unwell, unable to weight-bear at all. That is a suspected joint infection until proven otherwise and it is a separate emergency.

Refer to: orthopaedics, ideally a hip-preservation service, urgently for suspected pre-collapse disease. The referral must specifically request MRI and must name the risk factor, because the request is what gets the right scan.

Return to Training

Conventional checkpoints, not validated criteria. No return-to-activity criterion has ever been tested in this condition, and any timeline offered here would be invented.

Specialist has confirmed the diagnosis, the stage, and the lesion size and location
Specialist has given explicit clearance to load the leg, and has specified any weight-bearing limit
The other hip has been imaged and its status is known
Full pain-free active hip range, or a stable range the specialist has accepted
Can walk 30 minutes without pain and without a limp
Can do bodyweight squats and step-ups through comfortable range, with no pain in the following 24 hours
No catching, clicking or giving way
Symptom-stable for 4 consecutive weeks
Both patient and clinician understand that clearance is a specialist decision that can be revised on the next scan

Conviction

Moderate overall, and the spread across endpoints is wide enough that the single label undersells one half and oversells the other.

HIGH Untreated symptom-free osteonecrosis progresses in a substantial proportion. Lesion size and location stratify prognosis. Displaced femoral neck fracture carries substantially higher risk than undisplaced (20.7% versus 4.7%).

MODERATE Symptoms precede collapse and the gap is short in advanced pre-collapse lesions. Cumulative steroid dose in a compressed window drives risk while a single 24-hour megadose produced zero cases in 59 patients. Reduction delay matters after dislocation. BMAC delays hip replacement.

LOW Core decompression over conservative care at stage I. Bisphosphonates, anticoagulation, pulsed electromagnetic fields, platelet-rich plasma, acupuncture. And low-and-negative for the idea that staging systems are reliable enough to carry stage-specific thresholds: the best of five scored AUC 0.65.

NO EVIDENCE Any exercise, weight-bearing, loading or gait-aid prescription. Not thin evidence. None.

What would change my mind, per claim

On protected weight-bearing, the claim that matters most here.

A randomised trial in adults with pre-collapse disease confirmed on MRI, stratified by lesion size and location, N of at least 300, randomising protected weight-bearing against unrestricted weight-bearing, minimum 3-year follow-up, with femoral head collapse as the primary endpoint. That trial would, for the first time, tell a physical therapist whether the crutches they hand out change anything. Its absence is currently filled by convention.

On the staging systems.

Replication of the five-system reliability comparison at N of at least 200 with more raters. If the AUC of 0.65 holds at scale, this field is making stage-specific decisions on an instrument that cannot reliably assign stage, which reaches further than any single treatment result here.

What would NOT change my mind.

Another uncontrolled series of an adjunct therapy measured against historical natural history. Because natural history varies from under 10% to a clear majority by lesion size, that comparison is structurally uninformative: case mix alone can generate the entire reported effect.

The Full Picture — Anatomy, Diagnosis & Evidence

What's Actually Going On

Cinematic cross-section of the femoral head and its blood supply

The femoral head is fed by an end-arterial supply dominated by one vessel running along the femoral neck, with essentially no collateral backup. Interrupt it and the bone underneath the joint surface dies.

What follows is mechanical rather than inflammatory. Dead bone cannot remodel, so it fatigues under load, a crack forms beneath the surface, and then the surface collapses. Collapse is the point of no return, which is why every staging system in this field is organised around whether it has happened.

Two routes reach that point. Mechanical interruption, where a displaced fracture tears the vessels at the moment of injury, or a dislocation kinks and stretches them. And non-traumatic occlusion, where steroids, alcohol, sickle cell disease, inherited clotting disorders and high blood fats converge on blocked vessels and expanding marrow fat. The clotting account has direct support: anticoagulation was trialled specifically in carriers of two inherited clotting variants, and raised blood fats came out as an independent predictor in two entirely separate disease populations.

How to Identify It

Dark cinematic anatomical study of the hip joint

There is no clinical test. No physical examination test for this condition with any published accuracy appears in the 67 papers reviewed. It is an imaging diagnosis, and specifically an MRI diagnosis, because plain radiographs are normal in early disease by definition.

Any physical examination test Sn: data unavailable | Sp: data unavailable because none exists
MRI Sn: data unavailable | Sp: data unavailable and it is still the diagnosis
Screening MRI in an at-risk symptom-free group 6 of 100 positive (95% CI 2.2 to 12.6) Tervonen 1992

Sensitivity figures from general hip pain prediction rules were considered and deliberately not borrowed. They were not developed in this condition, and a borrowed number in a diagnostic slot is an invented number.

What to ask instead: have you ever taken steroid tablets, and how much over what period? Have you had a transplant or chemotherapy? Do you have sickle cell disease or a clotting disorder? Have you ever broken or dislocated this hip? How much do you drink? How is the other hip?

The Debate

No clinical practice guideline was identified for this condition through this evidence sweep as of 13 August 2026. Several are named in the preliminary landscape scan, none was retrieved as an indexed record, and where the scan could be checked its attributions were unreliable, so none is cited here.

Is symptom-free osteonecrosis benign?

Older position

A benign natural history, with treatment of the symptom-free hip regarded as controversial.

VS

Mont 2010, 664 hips · Hernigou 2006, 121 hips over 14 years

394 of 664 untreated symptom-free hips (59%) progressed. In sickle cell disease, 91% became painful and 77% collapsed.

Both sides were reading real data through different case mixes. Small, inner-side lesions collapse under 10% of the time; medium, large or outer-side lesions progress in a substantial proportion. Size and location decide, not the absence of symptoms.

Does speed of treatment prevent it after hip trauma?

Konarski 2022, 52 papers, N=5,930

After femoral neck fracture fixation, time from injury to surgery did not correlate with osteonecrosis (p=0.843).

VS

Milenkovic 2022, N=44

After traumatic posterior dislocation, reduction delay was the main factor: 4.34%, 8.69%, 21.73% (p=0.030).

Not a conflict, and teaching them as one urgent-hip rule loses the distinction. A dislocated head is kinked off a supply that reduction restores, so the clock genuinely runs. In a fracture the vessels tore at the moment of injury, and surgical timing cannot undo that. Note the evidence asymmetry too: 5,930 pooled patients against 44, with single-digit event counts per delay stratum.

Do the newer biological treatments work?

Zaffagnini 2025, headline curve

120-month survival free of hip replacement 69.4% with orthobiologics versus 48.5% without (p<0.0005), across 106 studies and 4,505 patients.

VS

The same paper's controlled analysis · Li 2023

Against matched controls, only BMAC survived. Stem-cell augmentation lost its advantage beyond 5 years.

Follow the controlled analysis, not the headline curve. The same review that produces the number produces the deflation, which is the correct way to read it. The honest framing is delay of hip replacement rather than prevention of it.

Honest Limitations

1. The whole field is staged by an instrument with modest reliability

The research finding: five staging systems compared head to head. The best performer scored AUC 0.65 against MRI progression, and one produced significant disagreement between raters (Vezirhüyük 2025).

The real-world gap: every treatment threshold here is stage-keyed. When two reports disagree about stage I versus stage II, the disagreement may be in the instrument rather than in the hip. This is one study at N=20 with three raters, so it is a warning rather than a verdict.

Clinical adjustment: do not treat a stage label as a hard fact. Ask for lesion size and location, which is what actually predicts.

2. Every conservative and adjunct claim has a comparator problem

The research finding: alendronate is measured against historical natural history. Anticoagulation has no control arm and required a specific inherited trait for entry (9 patients). Orthobiologic survival is pooled across non-randomised series.

The real-world gap: natural history varies from under 10% collapse to well over half depending on lesion size. Any uncontrolled series that happened to enrol small inner-side lesions will report an excellent outcome the treatment did not cause.

Clinical adjustment: treat every uncontrolled positive result here as hypothesis-generating, and weight the controlled analyses inside the same papers over their headline curves.

3. The physical therapy evidence base does not exist, and the absence has a specific shape

The research finding: a dedicated literature query on protected weight-bearing and collapse prevention returned zero papers. A dedicated rehabilitation query returned two: a 1997 bracing study in Perthes' disease, a different paediatric condition, and a 1985 report of three adolescents with sickle cell disease treated with crutches and range-of-motion work.

The real-world gap: protected weight-bearing is what actually gets prescribed. It has never been tested against collapse in adults. Handing a patient crutches feels like treatment and can substitute for the referral that is the real intervention.

Clinical adjustment: prescribe symptom relief honestly as symptom relief. Do not tell a patient that offloading will save the joint, because nobody knows whether it does.

The Nuance

Cinematic rendering of hip joint anatomy in deep shadow

Conservative management success rate: 86%, 61%, 59% and 25% by stage, from 8 pooled conservative series (Castro 2000). The stage list in that published abstract is printed verbatim as "stage 0, I, III, and III", four rates against a list containing III twice and no II. The near-certain intent is 0, I, II, III. It is reported here as printed rather than silently renumbered, because this page keys decisions to stage and a quietly corrected stage is exactly the sort of invented precision that survives review.

Surgical success rate: core decompression 84%, 63% and 29% for stages I, II and III, statistically better than conservative care at stage I only (Castro 2000). Orthobiologic augmentation raises 120-month survival free of hip replacement to 69.4% against 48.5% (Zaffagnini 2025). Osteotomy converts to hip replacement in 31.5% over roughly 7 years (Quaranta 2021).

When surgery is indicated: pre-collapse disease with a medium, large or outer-side lesion. Symptomatic pre-collapse disease of any size, given that symptoms preceded collapse in every case observed in the longest cohort available. Established collapse with intractable pain, where joint replacement is the answer.

When conservative care is sufficient: small, inner-side, symptom-free lesions, which collapse under 10% of the time. That is the one genuinely reassuring category here, and it is defined by size and location rather than by absence of symptoms. Also patients unfit for surgery, where management becomes symptom control.

The honest truth. This is not a condition where conservative management is a credible alternative to surgery for a large or outer-side lesion, and that runs opposite to most conditions covered on this site. In most of them a majority do well without an operation and the honest thing is to say so. Here, the majority of untreated symptom-free hips progressed, and in the longest cohort three quarters of them collapsed. What the evidence does support is that prognosis is stratified, so a small inner-side lesion is genuinely low risk, and that no treatment of any kind has been shown in a randomised, stage-stratified trial to prevent collapse. That is why the specialist conversation is a conversation and not an algorithm, and the physical therapist's contribution is getting the patient into it while the joint is still round.

Sources

14 of the 34 sources behind this page. All are abstract-resolved and identifier-anchored; the full set sits in the protocol card.

Mont MA, Zywiel MG, Marker DR, et al. (2010). The natural history of untreated asymptomatic osteonecrosis of the femoral head. JBJS Am. PMID 20844158 · systematic review, 16 studies, 664 hips · 59% progressed; small medial lesions under 10%
Hernigou P, Habibi A, Bachir D, et al. (2006). The natural history of asymptomatic osteonecrosis of the femoral head in adults with sickle cell disease. JBJS Am. PMID 17142405 · prospective cohort, 121 hips, mean 14-year follow-up · 91% symptomatic, 77% collapsed, 11-month mean pain-to-collapse interval at stage II
Konarski W, Poboży T, Kotela A, et al. (2022). The risk of avascular necrosis following the stabilization of femoral neck fractures. Int J Environ Res Public Health. PMID 36011686 · meta-analysis, 52 papers, N=5,930 · displaced 20.7% (95% CI 12.8 to 28.5) vs undisplaced 4.7% (3.4 to 6.0)
Milenkovic S, Mitkovic M, Mitkovic M (2022). Avascular necrosis of the femoral head after traumatic posterior hip dislocation. Eur J Trauma Emerg Surg. PMID 32929549 · retrospective cohort, N=44 · 4.34% / 8.69% / 21.73% by reduction delay (p=0.030)
Castro FP Jr, Barrack RL (2000). Core decompression and conservative treatment for avascular necrosis of the femoral head. Am J Orthop. PMID 10746469 · meta-analysis, 30 series · core decompression superior to conservative care at stage I only
Zaffagnini M, Boffa A, Andriolo L, et al. (2025). Orthobiologic therapies delay the need for hip arthroplasty. Knee Surg Sports Traumatol Arthrosc. PMID 39543728 · systematic review with survival analysis, 106 studies, N=4,505 · 69.4% vs 48.5% at 120 months; only BMAC confirmed against controls
Li M, Chen D, Ma Y, et al. (2023). Stem cell therapy combined with core decompression versus core decompression alone. J Orthop Surg Res. PMID 37533122 · meta-analysis, 11 RCTs plus 7 retrospective, 916 patients · advantage did not hold beyond 5 years
Vezirhüyük M, Celtik M, Şahin E, et al. (2025). Comparative analysis of reliability and validity of five staging classifications. BMC Musculoskelet Disord. PMID 39948599 · observational reliability study, N=20 · best system AUC 0.65 (p<0.01)
Tervonen O, Mueller DM, Matteson EL, et al. (1992). Clinically occult avascular necrosis of the hip. Radiology. PMID 1535906 · screening MRI, N=100 · 6% occult disease (95% CI 2.2 to 12.6)
Shibatani M, Fujioka M, Arai Y, et al. (2008). Degree of corticosteroid treatment within the first 2 months of renal transplantation. Acta Orthop. PMID 18839369 · prospective MRI surveillance, N=150 · dose-response for total steroid dose in the first 2 months
Wing PC, Nance P, Connell DG, et al. (1998). Risk of avascular necrosis following short term megadose methylprednisolone treatment. Spinal Cord. PMID 9773448 · prospective cohort with blinded MRI read · zero cases; true incidence bounded under 5%
Agarwala S, Shah SB (2011). Ten-year follow-up of avascular necrosis of femoral head treated with alendronate for 3 years. J Arthroplasty. PMID 21256699 · single-arm follow-up, 40 patients, 53 hips · historical controls only
Glueck CJ, Freiberg RA, Wang P (2020). Long-term anticoagulation prevents progression of stages I and II primary osteonecrosis of the hip in patients with familial thrombophilia. Orthopedics. PMID 32271930 · prospective single-arm cohort, 9 patients, 13 hips · 0 of 13 collapsed over a mean 12 years
Washington ER, Root L (1985). Conservative treatment of sickle cell avascular necrosis of the femoral head. J Pediatr Orthop. PMID 3988921 · case report, 3 adolescents · the only weight-bearing report retrieved in 67 papers

Go Deeper

A normal scan is not always an all-clear, and knowing which ones to trust is most of the job. The Verdict breaks down one condition a week, evidence-scored, free.

Join The Verdict, free