What a bone marrow lesion actually is

Receiving an MRI report that mentions 'bone marrow oedema' or a 'bone marrow lesion' can feel alarming — the language sounds serious, and it is not always explained clearly. In practice, this is one of the more common incidental and symptomatic findings on knee MRI, and understanding what it actually represents helps put the result in context.

On MRI, a bone marrow lesion (BML) appears as a bright area on fluid-sensitive sequences — specifically T2-weighted or STIR images — within the subchondral bone: the dense layer of bone sitting just beneath the cartilage surface of the knee joint. The brightness tells the scanner that something is abnormal in that region of bone, but it does not, on its own, tell you why.

The term 'oedema' implies swelling or fluid, which is partly misleading. Histological studies — examinations of actual tissue samples — show that BMLs are structurally more complex than a simple fluid collection. The tissue within a lesion typically contains a combination of microfractures of the fine trabecular bone scaffold, replacement of healthy marrow with fibrovascular scar-like tissue, abnormal new blood vessel growth, and elevated bone turnover. This composite picture explains why BMLs can be a significant source of pain, not merely a passive fluid signal.

Several distinct conditions produce this MRI appearance: osteoarthritis is the most common clinical context, but BMLs also arise after acute bone bruising (for example, following an ACL injury), from stress or insufficiency fractures, in transient bone marrow oedema syndrome, and in osteonecrosis. Each of these has a different MRI pattern, a different natural history, and different implications for treatment. The imaging finding alone does not establish the cause — that requires clinical history, examination findings, and specialist review working together.

Why bone marrow lesions cause pain

The pain associated with a BML does not come simply from mechanical pressure or inflammation in the joint above. Laboratory analysis of tissue taken directly from BMLs reveals elevated concentrations of tumour necrosis factor-alpha (TNF-α), matrix metalloproteinases (MMPs), and substance P — molecules that are simultaneously pro-inflammatory and nociceptive, meaning they directly sensitise pain receptors embedded within the bone itself.

Substance P, in particular, is a well-characterised pain-signalling neuropeptide. Its presence inside the subchondral bone helps explain why patients often describe a deep, poorly localised ache that feels distinct from the surface discomfort associated with cartilage wear or synovial irritation.

Alongside this chemical environment, BMLs appear to promote pathological nerve ingrowth. Healthy mature bone contains relatively few free nerve endings; within BML tissue, new nociceptive fibres extend into regions not normally innervated. This structural change creates persistent pain pathways that can continue firing even when loading is reduced or the joint surface is treated — which is why some patients report that pain outlasts what might be expected from a purely mechanical or inflammatory process.

Taken together, the biochemical and structural changes mean that BML pain originates at bone level, below the cartilage and joint cavity. Any intervention that acts only on the synovial environment is therefore operating upstream of where much of that pain signal is actually being generated.

How strongly BMLs predict knee pain

Among all the structural findings that appear on a knee MRI, BMLs have the strongest statistical association with pain in knee osteoarthritis. Patients whose scans show a BML are two to five times more likely to report knee pain than those whose scans do not — a relationship that holds up across multiple epidemiological studies and makes BML status one of the more clinically meaningful things a radiologist can report.

The contrast is particularly striking when looked at in prevalence terms. Data cited from Felson and colleagues (Ann Intern Med, 2001) show BMLs present in approximately 77.5% of patients with painful knee OA, compared with roughly 30% in patients whose OA was present on imaging but not causing symptoms. That near three-fold difference in prevalence between symptomatic and asymptomatic knees is why specialists treat this finding as a clinical signal rather than incidental background noise.

Lesion size adds further nuance. Larger BMLs are found almost exclusively in patients who are in pain; smaller lesions, or those that have begun to resolve, may carry less clinical weight. This matters because not every BML is permanent. Post-traumatic bone bruising — the kind that follows an ACL injury, for example — often resolves spontaneously over weeks to months as the bone heals. OA-associated lesions behave quite differently: without any change in the underlying mechanical or inflammatory environment, they tend to persist or enlarge over time.

None of this means that a BML on a scan automatically explains a patient's symptoms. The imaging-is-not-a-verdict principle applies here as elsewhere: size, location, and the patient's actual clinical picture all have to be weighed together.

What BMLs predict for OA progression

Beyond their role as a pain signal, BMLs carry structural implications that clinicians weigh carefully when planning any intervention — and those implications go further than symptom management alone.

Research indicates that BMLs are associated with radiographic progression of OA specifically within the compartment where the lesion sits, a spatial relationship that points to bone-level stress as a driver of joint deterioration rather than simply a consequence of it. More strikingly, Singh et al. (2019) found that the presence of a BML increases the likelihood of eventual total knee arthroplasty by up to nine-fold. That figure is not a forecast for any individual patient; it is the reason specialists treat lesion size and location as clinically significant variables on a pre-treatment MRI, rather than a secondary finding to be noted and set aside.

Baseline lesion size also predicts future incident and worsening pain — which is the basis for using it as a risk-stratification marker when deciding the timing and type of intervention. A patient whose scan shows a large or persistent BML in a deteriorating compartment occupies a materially different clinical position from one with a smaller, self-limiting lesion following a discrete injury.

For all these reasons, BMLs identified on a pre-injection MRI should prompt a structured conversation about intervention options and timing, rather than a default 'watch and wait' approach. Their prognostic weight supports earlier specialist assessment, particularly when lesion size or compartmental load suggests the risk of progressive structural change.

What BMLs mean for injection planning

Identifying a BML changes the fundamental question about injection treatment: not just whether to inject, but where the therapeutic agent needs to reach to have any effect.

Standard intra-articular injections — corticosteroids, hyaluronic acid, and most conventional platelet-rich plasma protocols — are delivered into the joint space and act primarily on the synovial environment. They do not penetrate the subchondral bone in any therapeutically meaningful concentration. Where the dominant pain source sits beneath the cartilage rather than within the joint cavity above it, this creates a mismatch between delivery point and pathology target. Patients whose symptoms are driven mainly by active BMLs may therefore gain limited benefit from IA-only approaches — not because the injection has failed, but because the target was never reached.

Intraosseous (IO) delivery addresses this directly. IO injections of bone marrow aspirate concentrate (BMAC) place regenerative agents into the subchondral bone itself; early studies have reported safety and initial efficacy signals for this approach in knee BMLs (Dragoo, 2024; Centeno et al., 2021). Ultrasound-guided trans-cortical IO PRP injection has also been reported as technically feasible for tibial plateau lesions. Both approaches remain in the early-phase evidence tier — promising, but without the robust head-to-head trial data needed to confirm superiority over optimised IA protocols.

There is also an emerging signal from a published study suggesting that certain intra-articular viscosupplements may reduce patellofemoral BMLs, possibly through joint offloading and changes in fluid dynamics — a finding that warrants further investigation but should not yet be treated as established practice.

The choice between IA, IO, or a combined approach depends on lesion size, cause, compartment, and overall OA stage. BML status should form part of any pre-injection planning discussion, not be treated as incidental context.

Assessment, open questions, and when to seek specialist input

A BML on an MRI report is not a surgical verdict. The spectrum runs from self-resolving post-traumatic lesions to high-risk OA-associated lesions, and meaningful non-surgical pathways exist at each stage — but the appropriate path is not the same for every patient.

Load modification, physiotherapy-guided strengthening, and weight optimisation remain the first-line foundation regardless of lesion size or severity. These measures reduce the mechanical stress on subchondral bone that sustains and enlarges lesions over time, and they are not steps to be bypassed on the way to injection.

Specialist assessment is warranted when a lesion is large, when symptoms persist beyond the expected natural-history resolution window, or when a patient has not responded to conservative care or a standard intra-articular injection course. At that point, deciding between IA and IO delivery — and which biologic agent is appropriate — requires the full clinical picture: imaging, examination, symptom trajectory, and cause of the lesion, not the MRI finding in isolation.

Two evidence gaps are worth naming honestly. Whether lesion size should formally gate the choice between IA and IO delivery is not yet standardised. Which biologic formulation best supports long-term subchondral repair remains under investigation. In practice, clinicians navigate these gaps by using lesion grade, underlying cause, compartment load, and prior treatment response as a composite signal — rather than waiting for a threshold rule that does not yet exist.

For patients at this decision point, an individual assessment is the most reliable way to determine which pathway best fits their specific findings and clinical history. The AMSK suitability assessment at amsk.co.uk offers a structured starting point for that conversation.