Steinberg (University of Pennsylvania) Classification of Osteonecrosis

View larger

View larger

Steinberg and colleagues at the University of Pennsylvania published their system in 1995 to address the limitations of the Ficat and Arlet classification — principally that Ficat grouped all pre-collapse MRI disease into a single stage and did not quantify lesion size. The Steinberg system therefore separates seven radiographic stages (0 through VI) and adds an alphabetic modifier (A, B, or C) based on the proportion of the femoral head involved on MRI.

Steinberg staging relies on MRI from Stage I onward, so the examiner expects you to know the MRI signs that actually make the diagnosis. MRI is the most sensitive and specific modality (both around 99 percent) and detects disease long before radiographs change.

• Double-line sign (the characteristic sign): on T2-weighted images, an inner bright line (hypervascular granulation tissue at the repair interface) running parallel to an outer low-signal line (sclerotic reactive bone) at the margin of the necrotic segment. It is the single most characteristic MRI feature of osteonecrosis.
• Geographic lesion with a reactive interface: a well-demarcated, usually anterosuperior subchondral segment ringed by a serpiginous low-signal rim on T1 (the interface between dead and living bone). The central necrotic marrow often retains fat signal early in the disease.
• Bone marrow oedema: diffuse low T1 / high T2 signal extending into the head and neck. It is non-specific, but in established osteonecrosis it correlates with pain and impending collapse and is a poor prognostic sign.
• Subchondral fracture line: a low-signal subchondral line, the MRI correlate of the radiographic crescent sign, marking the transition to Steinberg Stage III.

The Steinberg system adds an alphabetic modifier to every stage (except 0 and VI, where quantification does not change management). The modifier is determined by measuring the proportion of the femoral head involved by the osteonecrotic lesion on MRI — specifically on the coronal T1-weighted images where the low-signal lesion is most clearly delineated against the bright fatty marrow.

View larger

The Steinberg classification guides treatment by combining the stage (structural integrity of the head) with the modifier (size of the necrotic lesion). Two variables matter: is the head still intact, and how much of it is dead?

The ARCO system in the comparison table above is the original 1993 version. An up-to-date examiner may ask about the 2019 revised ARCO classification, which was simplified for reproducibility and is now widely used in research:

The key changes from 1993 were: Stage 0 was removed (a normal-imaging stage is unverifiable and clinically unused), each stage was tied to a clear radiograph/MRI step, and Stage III was subdivided by the degree of collapse using a 2 mm femoral-head-depression threshold that carries prognostic and surgical weight. The 2019 consensus deliberately kept the staging simple and recommended that lesion size and location be reported separately rather than embedded as sub-stages, because the older quantitative sub-classes had poor inter-observer reproducibility — the modern counterpoint to Steinberg's embedded A, B, C size modifiers.

• MRI quantification is operator-dependent. Estimating the percentage of head involvement from two-dimensional slices has moderate inter-observer variability, particularly for irregular or multifocal lesions. Three-dimensional volumetric MRI measurements are more reproducible but not yet part of routine staging.
• The A, B, C thresholds (15 percent and 30 percent) are somewhat arbitrary. They were chosen to create three roughly equal prognostic groups, but the real relationship between lesion size and collapse risk is continuous — a lesion at 29 percent is not meaningfully different from one at 31 percent, yet they fall into different modifier groups.
• The system does not incorporate the lesion location. A small lateral lesion (poor prognosis) and a large medial lesion (better prognosis) may receive the same modifier. This is why some centres combine Steinberg with the JIC location-based typing for the most complete assessment.
• Stage 0 is almost never used clinically. It requires histological confirmation in a patient with normal imaging, which is rare outside research or contralateral hip biopsy during surgery for the other side. Do not expect to encounter it in practice.
• The post-collapse stages (IV–VI) are less useful than the pre-collapse stages. Once the head has collapsed, the classification refines how badly it is destroyed, but the treatment decision (arthroplasty) is largely the same. The real value of Steinberg lies in Stages 0–III, where it guides whether to operate, what to do, and how urgently.
• Bisphosphonates, extracorporeal shockwave therapy, and newer biological adjuncts may alter the natural history at any given stage. The classification does not account for treatment received when assigning prognosis.

Multiple classification systems exist for osteonecrosis. The examiner may ask you to compare them. The key differences are whether MRI is incorporated, whether lesion size is quantified, and whether the post-collapse stages are subdivided.