When your X-ray looks fine but your knee still hurts

Seeing a GP, being shown your knee X-ray, and being told "it's not too bad yet" is a familiar experience for people living with significant pain. The X-ray probably showed some narrowing of the joint space or a few small bony spurs — the features that make up a Kellgren–Lawrence (KL) grade, the standard system used to classify knee osteoarthritis on plain film. A KL grade of 1 or 2 is routinely described as mild or moderate. For many patients, the pain is anything but.

The reason this framing falls short is structural: KL grading measures only what a plain radiograph can show — cartilage-space narrowing and osteophytes. It cannot detect damage occurring within the bone itself, below the cartilage surface. A 2025 cohort study of 204 knees with early-to-moderate OA (KL grades 0–2) tested this directly: total bone marrow lesion (BML) burden — detectable only on MRI — explained 5.7% of the variation in WOMAC pain scores, while the extent of cartilage damage explained just 1.5–2.3%. BMLs drove pain at roughly two to three times the magnitude that cartilage loss did in this group — a difference that a normal-looking X-ray cannot register.

That gap between X-ray grade and lived experience has direct consequences for treatment selection: patients stratified by KL grade alone may be systematically misidentified as mild cases when the true driver of their symptoms lies in the subchondral bone. What the X-ray misses is happening inside the bone itself.

What bone marrow lesions actually are

Bone marrow lesions sit in the subchondral bone — the dense layer of bone that lies directly beneath the articular cartilage. On MRI, using STIR or fat-suppressed sequences, they appear as bright patches of increased signal, reflecting elevated water content in bone tissue that is normally relatively dry. The name is slightly misleading: a BML is not a single, discrete pathological entity but a zone where several processes overlap — localised bone-marrow oedema, trabecular microdamage, fibrosis, and abnormal remodelling can all be present within the same lesion.

Think of it loosely as a bruise within the bone — one that a plain radiograph cannot capture because radiographs record mineralised structure, not marrow fluid or micro-architectural disruption.

BMLs are common in symptomatic knee OA: roughly two-thirds of patients have at least one detectable lesion, and both the number and size of lesions tend to increase as the disease advances. They are not confined to late-stage disease, however — lesions can appear in early or mild OA, which partly explains why symptom burden does not always follow cartilage-loss grade.

Critically, BML burden can be measured and tracked. Validated scoring systems — WORMS and BLOKS are the most widely used — assign a numeric score to lesion size within each joint compartment, making BMLs a quantifiable biomarker that can be compared across time-points or between treatment arms in clinical trials.

Which pains BMLs cause — and which they don't

The pattern many patients describe — knee pain that surges on the stairs or during a walk but quietens when sitting down — has a structural explanation that BML location data helps clarify.

A longitudinal analysis of 1,412 varus knees from the Osteoarthritis Initiative followed participants over 24 months and mapped BML location against specific pain activities. Medial femorotibial BMLs were significantly associated with weight-bearing pain during walking and standing. Lateral patellofemoral BMLs were associated with stair-climbing pain. Neither region was significantly linked to non-weight-bearing pain after adjustment — meaning BMLs appear to drive mechanically loaded discomfort, not persistent background pain at rest.

The implication is that load amplifies the pain signal from an already-damaged subchondral environment. Remove the load, and the signal drops; apply it again, and it returns.

This spatial specificity also has direct value for tracking treatment response. Over the same 24-month follow-up, changes in medial femorotibial BML score tracked changes in weight-bearing pain with a statistically significant association (B=0.10, p=0.01). When BML burden reduced, walking pain tended to reduce alongside it; when it increased, pain worsened. That kind of dynamic, compartment-sensitive signal makes BML score a meaningful marker for evaluating whether a treatment is actually working — something a slowly-shifting structural grade cannot provide.

Why BMLs generate pain X-rays cannot account for

Three converging biological pathways help explain why BMLs are so closely tied to pain.

Bone nerves, not just joint-capsule nerves. Electrophysiological studies in a rat OA model show that, in late-stage disease, pain signalling shifts from the joint capsule to afferent neurons innervating the subchondral bone itself. Those bone-afferent nerves become activated and sensitised — firing more readily under load — which early laboratory evidence suggests underlies why BML-associated pain is worse during weight-bearing than at rest.

Osteoclast-driven bone erosion. A second pathway involves the cells responsible for breaking down bone. Serum TRAcP5b, an osteoclast activity marker, correlates with both baseline knee pain severity and three-year pain change in people with knee OA; TRAcP-positive osteoclasts are also more abundant in symptomatic joints than in post-mortem controls. The association is correlational, but it supports a direct role for subchondral bone resorption in pain generation and points to anti-resorptive therapy as a rational therapeutic direction.

An inflamed, senescent microenvironment. Single-cell RNA sequencing of BML tissue (2025) identified a population of senescent non-classical monocytes producing TNF, a pro-inflammatory signal that appears to amplify damage in both bone and overlying cartilage. Early evidence suggests this operates through a senescence-associated secretory phenotype (SASP) — a self-reinforcing inflammatory state — with TCF7L2 acting as a shared molecular switch across immune cells and chondrocytes. The mechanism is still being characterised and the data remain largely preclinical.

Taken together, neurological sensitisation, subchondral bone resorption, and immune senescence each point to treatment targets located within the bone itself — independent of cartilage grade.

When BML presence alone is not enough to predict pain

Not every MRI that reports a BML fully explains a patient's pain. A 2025 cross-sectional analysis from the Bunkyo Health Study (n=1,145, mean age 72.9 years) found that BML presence alone was not significantly associated with knee pain (OR 1.32, 95% CI 0.95–1.83). The signal became meaningful only when BMLs coexisted with other subchondral changes: BML alongside subchondral bone attrition raised pain odds to OR 2.22, and BML alongside subchondral bone cysts to OR 1.79. Importantly, both associations held regardless of whether there was cartilage damage directly above the lesion.

The pain-generating unit, in other words, appears to be a composite subchondral phenotype — bone marrow change combined with the structural deterioration surrounding it — rather than the BML in isolation. This helps explain why some patients with MRI-reported BMLs report surprisingly little pain: co-pathology, not BML presence alone, seems to be the decisive variable.

A single BML score threshold applied without assessing the broader subchondral picture will therefore oversimplify the clinical signal. This is why treatment decisions in knee OA increasingly depend on the full MRI phenotype, not just any one finding read in isolation.

How BML findings change treatment decisions

Identifying a BML-positive subchondral phenotype on MRI shifts the treatment decision in a concrete direction: therapies that target bone directly, rather than the joint space alone, become relevant.

BML score as a trial endpoint. Establishing that BML burden can change with treatment mattered first in drug-trial design. A strontium ranelate OA study found BML score fell significantly — but only in patients who had BMLs at entry — alongside reductions in WOMAC pain and the cartilage-turnover marker CTX-II. That selective response validated BML score change as a meaningful structural endpoint in disease-modifying trials, separate from joint-space width.

Subchondroplasty. Available in specialist centres rather than as a standard-pathway procedure, this technique injects a calcium-phosphate bone-substitute material directly into the BML region. In a 50-patient series, 54% of participants achieved a clinically meaningful pain reduction (≥4 NRS points) at six months, with WOMAC and function scores improving through twelve months. Outcomes were not universal: 22% converted to total knee replacement within the 26-month follow-up.

Core decompression with intraosseous biologics. In a small series of patients who had failed conventional care, combining surgical core decompression with intraosseous autologous conditioned plasma reduced pain scores (NPRS) from 8.3 to 1.5 within one week, with KOOS improving by 44.6 points at twelve months. This is not yet standard of care; the evidence base remains limited to early-phase series.

Anti-resorptive therapy. Zoledronic acid, a bisphosphonate that suppresses osteoclast activity, is currently being evaluated in randomised controlled trials for its capacity to reduce knee pain and delay total knee replacement in patients without severe joint-space narrowing — a population likely enriched for BML-dominant disease. Results are awaited.

Intra-articular biomechanical treatment. One small study reported reductions in patellofemoral BMLs on MRI following a single intra-articular polyacrylamide hydrogel injection, alongside pain improvement. The direction is consistent with the hypothesis that offloading subchondral stress reduces BML burden, but the evidence remains preliminary.

Most of the above data come from smaller series or secondary analyses rather than large pre-specified trials; the field is actively moving toward BML-stratified study designs. For patients whose pain has not responded to standard management, specialist MRI assessment to characterise the full subchondral picture — BML burden, attrition, and cyst co-pathology together — is the step that determines which of these pathways is appropriate.

1. [1] Bone marrow lesion coexisted with subchondral bone attrition and/or subchondral bone cyst is associated with knee pain in knee osteoarthritis regardless of cartilage lesion: the Bunkyo health study. (2025). https://doi.org/10.1186/s13075-025-03644-2 https://doi.org/10.1186/s13075-025-03644-2
2. [2] Imaging of osteoarthritis-related bone marrow lesions in clinical trials. (2025). https://doi.org/10.1016/j.ostima.2025.100270 https://doi.org/10.1016/j.ostima.2025.100270
3. [3] Cartilage lesions are not the main factor influencing pain and functional impairment in early knee osteoarthritis. (2025). https://doi.org/10.1302/2633-1462.67.BJO-2025-0010.R1 https://doi.org/10.1302/2633-1462.67.BJO-2025-0010.R1
4. [4] Efficacy of Subchondroplasty in the Treatment of Pain Associated With Bone Marrow Lesions in the Osteoarthritic Knee. (2023). https://doi.org/10.1177/23259671231163528 https://doi.org/10.1177/23259671231163528
5. [5] Association of subchondral bone marrow lesion localisation with weight-bearing pain in people with knee osteoarthritis: data from the Osteoarthritis Initiative. (2020). https://doi.org/10.1186/s13075-021-02422-0 https://doi.org/10.1186/s13075-021-02422-0
6. [6] Pain prediction by serum biomarkers of bone turnover in people with knee osteoarthritis: TRAcP5b and cathepsin K in OA. (2017). https://doi.org/10.1016/j.joca.2017.01.002 https://doi.org/10.1016/j.joca.2017.01.002
7. [7] Zoledronic Acid in the Treatment of Primary Knee Osteoarthritis: A Randomized Clinical Trial. (2025). https://doi.org/10.7759/cureus.85087 https://doi.org/10.7759/cureus.85087
8. [8] Effect of Intravenous Zoledronic Acid on Total Knee Replacement in Patients With Symptomatic Knee Osteoarthritis and Without Severe Joint Space Narrowing. (2024). https://doi.org/10.1002/art.42831 https://doi.org/10.1002/art.42831
9. [9] Changes to the activity and sensitivity of nerves innervating subchondral bone contribute to pain in late-stage osteoarthritis. (2021). https://doi.org/10.1097/j.pain.0000000000002355 https://doi.org/10.1097/j.pain.0000000000002355
10. [10] Disruption of the microenvironmental ecosystem in subchondral bone marrow lesions: Roles in osteoarthritis pathophysiology, pain and progression. (2025). https://doi.org/10.1016/j.ostima.2025.100374 https://doi.org/10.1016/j.ostima.2025.100374
11. [11] Core Decompression Combined with Intraosseous Autologous Conditioned Plasma Injections Decreases Pain and Improves Function in Patients with Symptomatic Knee Bone Marrow Lesions. (2023). https://doi.org/10.3390/biomedicines11071799 https://doi.org/10.3390/biomedicines11071799
12. [12] Single-cell sequencing reveals a senescent immune landscape in bone marrow lesions inducing articular cartilage damage in osteoarthritis. (2025). https://doi.org/10.1038/s41413-025-00467-4 https://doi.org/10.1038/s41413-025-00467-4