The Problem

Humeral head osteonecrosis is a rare but significant source of shoulder pain, which, in many cases, responds poorly to non-operative management. The humeral head receives its blood supply from intricate anastomoses originally branching off the axillary artery, which likely explains the relatively low rate of osteonecrosis compared to other articulations, such as the hip. Branching off the axillary artery in a variable fashion, the anterior and posterior circumflex arteries are the main suppliers of vascularity to the humeral head.

The anterior circumflex artery travels laterally underneath the long head of the biceps and gives off an ascending branch, terminating as the arcuate artery which enters the proximal humerus at the upper end of the bicipital groove, supplying both the lesser and greater tuberosities. The posterior circumflex artery, typically not as adherent to the humeral head as the anterior artery, travels about the posterior humeral neck and terminates in small perforating branches.

Prior studies have previously identified the anterolateral branch of the anterior humeral circumflex artery as the principal blood supply to the humeral head. However, recent enhances in technology have provided substantial evidence that the posterior humeral circumflex artery may play a much larger role. A recent study estimated that the posterior humeral circumflex artery serves as the main source of blood supply to three of the four quadrants of the head (lateral, superior, and inferior), and overall, supplies 64% of the blood supply, as compared with 36% from the anterior circumflex. This may explain why significant, comminuted fractures of the humeral head have a relatively low rate of osteonecrosis.

Continue Reading

Clinical Presentation

As the shoulder is not a significant weight-bearing joint and the glenoid is not as confining as the pelvic acetabulum, patients with osteonecrosis may be relatively asymptomatic. Symptomatic patients with humeral head osteonecrosis typically present with vague shoulder girdle pain, especially at night. In addition, patients with symptomatic osteonecrosis are typically younger than those with standard glenohumeral osteoarthritis, most likely related to other factors such as comorbidities or trauma.

The diagnosis may initially be thought to be soft tissue-related, as symptomatic stage I lesions will not manifest on plain radiographic imaging. This magnifies the importance of a thorough historical assessment, including screening for possible risk factors. As the lesion progresses, patients will subsequently develop pain with range of motion. As the lesions generally begin superiorly, the pain may typically initiate at about 60 degrees of humeral elevation. It is important to realize that scapulothoracic motion may also compensate such that patients begin feeling pain at higher degrees of forward flexion. With further joint destruction, patients may begin to feel significant crepitus as well as mechanical symptoms that may be associated with the formation of intra-articular loose bodies. With the development of bipolar osteoarthritis, range of motion deficits become more common. Once a diagnosis of osteonecrosis is made, it is critical to assess other joints, especially the hip, as there is a strong association between shoulder and hip osteonecrosis.

As humeral head osteonecrosis has clear associations with other comorbidities, those must be assessed. Humeral head osteonecrosis can be caused by either traumatic or atraumatic means. Traumatic causes of osteonecrosis may include proximal humerus fractures as well as iatrogenic injury from surgical fixation or arthroscopy. Atraumatic causes may result from either external factors (medication use, disease) or inherited disorders and blood dyscrasias. In a classic study, Hattrup and Cofield, in a study of 200 shoulders, found corticosteroids (56%) and trauma (18.5%) to be the most common potential causes of humeral head necrosis.


Three and four-part fractures of the proximal humerus may pose an up to 15-30% rate of osteonecrosis, respectively. Disruption of the branches of the anterior humeral circumflex artery may occur in up to 805 of these cases, reflecting the importance of the extensive anastomoses about the humeral head. Internal fixation poses iatrogenic danger to the vascularity secondary to extensive dissection and fixation as well as manipulation of fracture fragments or subsequent thromboses which may ensue. The approach used for fracture fixation may also be important, with relatively low rates of osteonecrosis being reported after limited anterolateral approaches. In addition, similar to femoral neck fractures, the quality of reduction may also contribute to the development of osteonecrosis.


The exact mechanism of osteonecrosis resulting from corticosteroid use is not entirely understood. Current hypotheses postulate that corticosteroids induce bone marrow fatty hypertrophy, which ultimately increases the intraosseous pressure and restricts arterial blood flow to the humeral head. Corticocosteroid-induced hyperlipidemia may also cause an embolic phenomenon. Presentation and diagnosis is most commonly made 6-24 months following induction of steroid use. In asymptomatic patients, the stage at initial visit, presence of pain, and continuation of peak doses of corticosteroids may predict progression of the disease. In symptomatic shoulders, Hernigou et al. showed that the extent and location of the lesion are the prime risk factors for disease progression.


Similar to femoral head osteonecrosis, sickle cell disease has also been implicated in cases of humeral head osteonecrosis. Arterial occlusions from altered red blood cell structure causes repeated microinfarcts, with subsequent local ischemia and bone necrosis. The rate of osteonecrosis in these patients may reach 6%, and patients with higher hemoglobin levels may have an increased risk secondary to an increased concentration of red blood cells. In patients with symptomatic humeral head osteonecrosis, almost 90% of patients may be expected to have collapse at long-term follow up, with the mean interval between the onset of pain and collapse to be about 6 years. Patients with prior known hip necrosis and patients with sickle-cell/beta-thalassemia may be at higher risk for developing osteonecrosis.

Alcohol use

Clear associations have been made with femoral head osteonecrosis and alcohol use. Similar associations have been made regarding the humeral head, however the exact incidence and risk is not known. Similar to corticosteroids, it has been hypothesized that fatty emboli to the subchondral vasculature ultimately lead to stasis and necrosis.

Other causes

Other rarer causes of humeral head osteonecrosis include Caisson’s disease (dysbarism), lysosomal storage diseases such as Gaucher’s disease, irradiation, chemotherapy, chronic dialysis, and systemic diseases such as systemic lupus erythematosus, rheumatoid arthritis, and Cushing’s syndrome. Diseases such as lupus may cause a vasculitis, which in turn produces local bony ischemia. However, it is not clear to what effect the actual pathology contributes to osteonecrosis, as many of these patients are concomitantly prescribed corticosteroids for disease control.

Diagnostic Workup

If necessary, infectious or neoplastic etiologies can be ruled out with the appropriate laboratory workup. Standard radiographs, including internal and external rotation anteroposterior (AP), scapular Y, and axillary radiographs, should all be obtained in patients with suspected osteonecrosis.

The Cruess classification is likely the most widely used classification scheme for humeral head necrosis, and is an adaptation of the system developed by Ficat and Arlet for femoral head lesions. In a stage I lesion, there are histologic changes in the head but no obvious pathology identified on routine radiographs. Advanced imaging such as bone scan or magnetic resonance imaging (MRI) is necessary for the diagnosis at this stage. Stage II lesions are consistent with cell death without resorption, and usually manifest as an area of sclerosis most pronounced at the superior humeral head (Figure 1). Stage III lesions are characterized by subchondral collapse, evidenced by the radiographic “crescent sign”, typically best seen on the external rotation view. The round contour of the humeral head is now compromised; however the overall sphericity is not significantly deformed (Figure 2). Stage IV lesions manifest as significant collapse and destruction of the trabecular pattern with resulting arthrosis of the head. The glenoid is typically preserved. In a stage V lesion, however, the glenoid is involved with bipolar osteoarthritis and a resulting incongruous joint.

Figure 1.

Humeral head osteonecrosis in a 29 year old female.

Figure 2.

Stage III humeral head osteonecrosis manifested with subchondral collapse and the “crescent sign”.

In those patients with no X-ray findings and where symptomatic stage I disease is suspected based on the presence of associated risk factors or known presence of hip osteonecrosis, we routinely perform non-contrast MRI to detect early lesions. Magnetic resonance imaging clearly establishes the transition between dead bone and living bone, with dead bone having a low-intensity signal on both T1 and T2-weighted images.

In stage I disease, it may manifest as a series of low-signal bands, which progressively enlarge to a uniform loss of signal in stage II disease. In addition, recent work has found that the extent of necrosis found in early stage humeral head osteonecrosis helps predict subsequent collapse, with a larger extent of necrosis predicting a higher collapse rate. Bone scans may also be performed if there are contraindications to traditional magnetic resonance imaging, such as the presence of a pacemaker or significant claustrophobia.

In patients in whom total shoulder arthroplasty is considered, we routinely obtain a pre-operative computed tomography (CT) scan to assess for the amount and location of glenoid bone loss, especially if considerable retroversion is present, as this information is critical in determining whether or not eccentric reaming or bone grafting will be necessary.

Non–Operative Management

We believe that non-operative management is indicated for the patient with minimal pain and deformity, as well as those patients with more advanced disease whose medical comorbidities preclude surgical intervention. Initial modalities, such as non-steroidal anti-inflammatories (NSAIDs) as well as physical therapy designed to maintain range of motion and glenohumeral stabilization can be instituted in cases where there is relatively insignificant deformity. Patients are typically advised to avoid activities and angles of flexion which maximize contact between the diseased head and the glenoid. In addition, corticosteroid injections into the glenohumeral joint may provide symptomatic relief; however, this must be weighed against the possible risk of advancing the osteonecrotic process especially in patients with steroid-induced osteonecrosis.

In addition, offending factors must be minimized. If possible, the use of corticosteroids should be discontinued or minimized. Alcohol and tobacco intake should be stopped. In addition, extra attention should be paid in patients with known diseases (sickle-cell, etc.) to minimize potential ischemic crises.

We believe that as the disease becomes more pronounced (stage III and onward), there is a decreased role for non-operative management. By this time, there is more advanced, irreversible deformity of the humeral head which will likely respond poorly to non-operative measures.

Indications for Surgery

We typically begin all patients with humeral head osteonecrosis in a focused, supervised course of non-operative management. However, with increased time and head collapse, non-operative management commonly fails and patients are indicated for operative intervention.

Surgical Technique

Patient positioning

In accordance with American Academy of Orthopaedic Surgeons (AAOS) recommendations, we always initial the surgical site pre-operatively and perform a well-executed, team-oriented time-out. We typically perform these procedures with the patient in the beach chair position and the head of the bed elevated as to surgeon preference. For an arthroscopic procedure, we typically place the bed angled to about 70-80 degrees. For arthroplasty procedures, we tend to place patients more supine, with the bed angled at about 30-40 degrees. We find this position allows for easier extension and manipulation of the shoulder joint during steps crucial to the procedure.

Either the use of regional (interscalene) or general anesthesia may be administered. We routinely place the patient in the beach chair position with the support of a padded Mayo stand to support the arm during portions of the case, and it is easily removable. In addition, one could consider a hydraulic arm suspension system if lacking assistants. The head of the bed is elevated approximately 30 degrees with the knees flexed and supported with pillows. The head is placed in a padded support taking care to keep the neck in neutral. It is critical to prep the entire extremity free such that adequate manipulation of the shoulder can be achieved. The arm should be able to achieve maximum adduction, external rotation, and extension to facilitate exposure.

Core decompression

The goal of this procedure is to decrease intraosseous pressure in the humeral head, while promoting revascularization of the subchondral bone. We reserve this procedure for those patients with symptomatic osteonecrosis with no evidence of bony collapse.

The procedure is begun by performing a diagnostic arthroscopy and assessing for other possible causes of shoulder pain. In addition, it allows us excellent visualization of the glenohumeral joint cartilage. After the arthroscopy is performed, we begin by making a 1 cm lateral incision, beginning about 2-3 cms below the lateral acromion. We prefer this incision for core decompression as this obviates the risk of potentially injuring the anterior circumflex artery, thus preserving more blood flow to the humeral head. The deltoid fascia is split vertically in line with its fibers. The axillary nerve typically lies 5-7 cms distal to the lateral acromion; therefore we routinely do not look for the nerve in these cases however we remain cognizant of its proximity. Under fluoroscopic guidance, we place a guidewire into the lesion and then drill over this with a reamer (Figure 3). A small curette may then be used to remove any remaining necrotic bone from the lesion. Although vascularized bone grafting has been reported in the literature (scapula, acromion), we routinely do not use this technique in our practice.

Figure 3.

Core decompression for humeral head osteonecrosis. This can also be done arthroscopically with the assistance of an anterior cruciate ligament targeting guide.

Alternatively, one can perform an arthroscopically-assisted decompression. An anterior cruciate ligament intra-articular guide can be placed over the lesion and by using a lateral stab incision, the guide pin can be inserted under both fluoroscopic and arthroscopic guidance to ensure no inadvertent penetration of the articular cartilage. The guide pin may then be over-reamed with a 6-7 mm reamer.

Hemiarthroplasty/Total shoulder arthroplasty

In those cases of osteonecrosis with significant bipolar lesions, we consider total shoulder arthroplasty. Otherwise, we feel that the majority of patients with humeral-based lesions can be successfully treated with hemiarthroplasty or resurfacing (Figure 4). Pre-operative planning is essential in assessing the amount and location of humeral deformity and glenoid wear. In addition, if considerable rotator cuff pathology is present and irreparable, a reverse total shoulder arthroplasty may be considered.

Figure 4.

Humeral head osteonecrosis treated with humeral head resurfacing.

We use a routine deltopectoral approach beginning just lateral to the coracoid and extending to the level of the deltoid insertion. Dissection is carried to the deep fascia and the fat stripe marking the deltopectoral interval where the cephalic vein is localized. The interval can also be localized by externally rotating the arm, putting the more horizontally-oriented pectoralis fibers under tension, making it easier to separate from the deltoid. If it is still difficult to establish the interval, the coracoid should be palpated as this marks the superior rotator interval and tracing this distally should lead one to fall directly into the deltopectoral interval.

We routinely retract the cephalic vein laterally, maintaining its tributaries. Medial retraction would require cauterization of the multiple branches draining the deltoid muscle. We place a self-retaining retractor in the interval, exposing the clavipectoral fascia. The fascia is incised just lateral to the watershed area of the conjoined tendon muscles. The self-retaining retractor is then redirected, retracting the deltoid lateral and the conjoined tendon medially. A very important step is to use a Darrach retractor (or other instrument) to release subdeltoid and subacromial adhesions; this will greatly aid the overall exposure. We then localize the biceps tendon just proximal to the pectoralis tendon; we tenotomize it at this location and tenodese it to the pectoralis major tendon. In addition, for added exposure, one can tenotomize the upper 1 cm of the pectoralis major tendon as it may allow for more complete posterior retraction of the humeral head during glenoid preparation.

The bursa is then cleaned off the subscapularis tendon and the arm is brought into external rotation to increase distance from the exiting axillary nerve inferomedially. We recommend palpation of the nerve as it exits beneath the subscapularis medially for localization purposes. Branches of the anterior circumflex artery and its venae comitantes (three sisters) are cauterized. We prefer a subscapularis tenotomy performed 1 cm medial to the lesser tuberosity unless the patient has less than 20 degrees of external rotation; at which point either a lengthening procedure or lesser tuberosity osteotomy may be performed. Stay sutures are placed into the subscapularis and capsular layer as it is being released.

Two additional mobilization procedures are then performed. The rotator interval is released superiorly, following the biceps tendon, essentially freeing the superior subscapularis.

Additionally, with the arm externally rotated, an inferior capsular release is performed directly off the humeral neck, taking care to stay directly on the bone as this minimizes iatrogenic damage to the axillary nerve. The humeral head should now be dislocated with the arm in adduction, extension, and external rotation. Osteophytes at the neck should be removed to fully appreciate the head-neck junction in preparation for the humeral resection.

When a hemiarthroplasty is being performed, the humeral cut is then made at approximately 30 degrees of retroversion based on the system used. If a resurfacing is performed, appropriate sizing of the humeral head should be performed first, followed by sequential reaming based upon the system used. We only utilize resurfacing when there is no need to address the glenoid – i.e., the glenoid is not involved. The canal is then sequentially reamed, broached, and trialled. If a hemiarthroplasty is the planned procedure, a trial head is then placed and translation in the glenoid and range of motion are assessed with the trials in place, and the final implant placed.

If proceeding with a total shoulder arthroplasty, glenoid exposure is then performed. With the arm placed at 30 degrees of abduction and neutral rotation, a spiked glenoid retractor is placed at the posterior lip of the glenoid, retracting the humeral head posteriorly. Another retractor is placed along the anterior glenoid rim and a small Hohmann retractor at the superior glenoid. The biceps remnant is excised as well as the glenoid labrum. A 360 degree release of capsular attachments from the subscapularis is performed for optimal mobilization. The anterior capsular attachments to the glenoid are then released. A small inferior capsular release from the glenoid can then be performed with a flat elevator or with electrocautery for approximately 1 cm, as it approaches the insertion of the long head of the triceps. Care must be taken not to excessively abduct the arm during this portion as it may bring the axillary nerve closer to the capsular attachments. Any glenoid osteophytes can be carefully removed.

After the glenoid is adequately exposed, the glenoid is sequentially reamed and trialled in the standard fashion according to the pre-operative plan and the type of instrumentation used. Closure is performed in standard fashion using #2 non-absorbable sutures for the subscapularis tenotomy. The rotator interval can be closed depending on surgeon preference. The deltopectoral interval is gently re-approximated and the tissues closed in a layered fashion.

Pearls and Pitfalls of Technique

  • In core decompression, stay superior with the incision to avoid the axillary nerve, which typically courses around the humerus between 5 and 7 cms distal to the lateral acromion.

  • Patient preparation and setup should allow full shoulder range of motion in order to properly expose the humeral head.

  • In total shoulder arthroplasty/hemiarthroplasty, take care to release subdeltoid/subacromial adhesions, make an appropriate humeral head cut close to the supraspinatus insertion, and do a capsular release to the 6 o’clock location on the humeral neck.

  • Poor access to the shoulder secondary to positioning.

  • Failure to do an adequate capsular release during arthroplasty procedures could hinder exposure.

  • Erring medial with subscapularis tenotomy too close to musculotendinous junction.

  • Failure to perform 360 degree release of subscapularis in total shoulder arthroplasty.

  • Failure to respect proximity of the axillary nerve.

  • Poor exposure of the glenoid, possibly secondary to:

    High humeral head cut

    Minimal capsular release

    Failure to clear subdeltoid/subacromial adhesions

Potential Complications

Complications secondary to these procedures are similar to those for surgery in general, which include infection, bleeding, stiffness, persistent pain, or damage to neurovascular structures. The axillary nerve must be respected in any approach to the shoulder. In addition, when an arthroplasty is performed, patients should be counseled about possible complications such as fracture, loosening, scapular notching (if performing reverse total shoulder arthroplasty), and instability.

Post–operative Rehabilitation

Rehabilitation after core decompression begins early, when post-operative pain permits range of motion of the shoulder. Patients are given a sling for comfort for 3-5 days and then are advised to discontinue the sling and perform active range of motion as tolerated. Rehabilitation after total shoulder arthroplasty, resurfacing, or hemiarthroplasty is typically dependent on the subscapularis mobilization technique. We typically tenotomize the subscapularis, therefore this repair needs to be protected post-operatively.

For the first 6 weeks, patients are protected in a sling. The sling is removed to perform passive range of motion exercises for the shoulder. Forward elevation in the plane of the scapula can proceed as tolerated. External rotation limits are set based upon the subscapularis repair and are determined intra-operatively. Internal rotation is limited to the chest wall. Patients can perform active range of motion of the elbow, wrist, and hand. At 6 weeks, the sling is discontinued and an active range of motion program is initiated for forward elevation, external rotation, and internal rotation behind the back. We also begin isometric deltoid and rotator cuff strengthening. Gentle stretching continues to increase the range of shoulder motion. Resistive strengthening (theraband, weights) begins at about 12 weeks post-operatively.

Outcomes/Evidence in the Literature

LaPorte, DM, Mont, MA, Mohan, V, Pierre-Jacques, H, Jones, LC, Hungerford, DS. ” Osteonecrosis of the humeral head treated by core decompression”. Clin Orthop Relat Res. 1998. pp. 254-260. (Authors examined 63 shoulders treated by core decompression. They found the best results in patients with grade I AVN according to the classification of Ficat and Arlet, with 94% successful outcomes. Grade II success rate was 88%, followed by 70% at grade III, and a very poor success rate of grade IV changes with an overall success rate of 14%. They conclude that this procedure should be reserved for patients with grades I-III changes.)

Harreld, KL, Marulanda, GA, Ulrich, SD, Marker, DR, Seyler, TM, Mont, MA. “Small-diameter percutaneous decompression for osteonecrosis of the shoulder”. Am J Orthop (Belle Mead NJ). vol. 38. 2009. pp. 348-354. (Small-diameter percutaneous perforations/core decompressions were performed for patients with early stage disease. At a mean follow up of 32 months, they noted 25 out of 26 successful clinical and functional outcomes. They compared their results to a historical control group and found this group to have a 48% progression to arthroplasty at 2-4.5 years of follow up.)

Uribe, JW, Botto-van Bemden, A. “Partial humeral head resurfacing for osteonecrosis”. J Shoulder Elbow Surg. vol. 18. 2009. pp. 711-716. (The authors studied 12 shoulders with advanced stage osteonecrosis measuring less than 40 mm in diameter treated with the HemiCAP prosthesis. Statistically significant improvements in visual analog scale, functional outcomes, and range of motion were noted, with no implant complications at mean 30 month follow up.)

Hettrich, CM, Boraiah, S, Dyke, JP, Neviaser, A, Helfet, DL, Lorich, DG. “Quantitative assessment of the vascularity of the proximal part of the humerus”. J Bone Joint Surg Am. vol. 92. 2010. pp. 943-948. (In a gadolinium-based MRI and anatomic study the authors found that, contrary to prior hypotheses, the posterior humeral circumflex artery provided 64% of the blood supply to the humeral head overall, with the anterior humeral circumflex artery supplying 36%. The posterior humeral circumflex artery also provided significantly more of the blood supply in three of the four quadrants of the humeral head.)

Hernigou, P, Flouzat-Lachaniette, CH, Roussignol, X, Poignard, A. “The natural progression of shoulder osteonecrosis related to corticosteroid treatment”. Clin Orthop Relat Res. vol. 468. 2010. pp. 1809-1816. (The authors studied 215 shoulders with humeral AVN and assessed delay from steroid treatment to development of symptoms. They found the mean delay was approximately 15 months. In addition, stage of disease (according to Cruess and modified by Steinberg) was a significant predictor of progression, as all shoulders with stage II disease progressed to symptoms while only 54%of stage I shoulders became symptomatic.)

Poignard, A, Flouzat-Lachaniette, CH, Amzallag, J, Galacteros, F, Hernigou, P. “The natural progression of symptomatic humeral head osteonecrosis in adults with sickle cell disease”. J Bone Joint Surg Am. vol. 94. 2012. pp. 156-162. (The authors found that untreated symptomatic shoulder osteonecrosis related to sickle cell disease has a high likelihood of progressing to humeral head collapse, and the natural evolution in the long term requires surgical treatment for many of these patients. At an average 20 years of follow up, 86% of shoulders progressed to collapse, with no regression of the lesions on imaging.)

Hattrup, SJ, Cofield, RH. “Osteonecrosis of the humeral head: results of replacement”. J Shoulder Elbow Surg. vol. 9. 2000. pp. 177-182. (The authors assess their results of prosthetic replacement at an average of 8.9 years follow-up, and found that inferior results in American Shoulder and Elbow Surgeons (ASES) scores and range of motion were noted in post-traumatic osteonecrosis, with superior results noted in steroid-induced osteonecrosis. Little difference was found between hemiarthroplasty and total shoulder arthroplasty. The most common post-operative complication was rotator cuff tearing, found in approximately 18% of shoulders.)

Feeley, BT, Fealy, S, Dines, DM, Warren, RF, Craig, EV. “Hemiarthroplasty and total shoulder arthroplasty for avascular necrosis of the humeral head”. J Shoulder Elbow Surg. vol. 17. 2008. pp. 689-694. (The authors of the study assessed patients at an average 4.8 years of follow up, and noted decreased range of motion in patients with post-traumatic osteonecrosis, with no differences in overall outcomes in patients with total shoulder arthroplasty versus hemiarthroplasty. However, patients with total shoulder arthroplasty had a higher complication rate (22 versus 8%), postulating that total shoulder arthroplasty should only be reserved for bipolar arthritic changes.)

Raiss, P, Kasten, P, Baumann, F, Moser, M, Rickert, M, Loew, M. “Treatment of osteonecrosis of the humeral head with cementless surface replacement arthroplasty”. J Bone Joint Surg Am. vol. 91. 2009. pp. 340-349. (The authors evaluated the results of cementless humeral surface replacement arthroplasty in those patients with bone necrosis of less than 31% of the overall humeral head. At 3 years of follow-up, patients had statistically improved range of motion and Constant scores, with no implant loosening or revisions necessary.)


Osteonecrosis of the shoulder is a rare but potentially debilitating entity. Detection of risk factors could lead to a more prompt and accurate diagnosis. Once diagnosed, non-operative management may be commenced however patients should be counseled on risk of progression. Early stages may be treated with less-invasive techniques such as core decompression. Later stages have been treated successfully with a variety of procedures, ranging from resurfacing to total shoulder arthroplasty.