Where the evidence stood a year ago
Low-dose radiotherapy for degenerative joint disease has been standard in German-speaking Europe for decades, with observational series reporting response rates of roughly 60 to 90 percent. The conventional schedule is 3 Gy in six fractions of 0.5 Gy.
Two sham-controlled randomized trials from Sint Maartenskliniek in Nijmegen tested it and found nothing. Patients with knee osteoarthritis and patients with hand osteoarthritis received six fractions of 1 Gy or six sham sessions. In the hand trial, 29 percent of irradiated patients and 36 percent of sham patients met responder criteria at three months. The knee trial, published by Mahler and colleagues in Annals of the Rheumatic Diseases in 2019, was likewise negative. Reporting six and twelve month outcomes in Lancet Rheumatology, van den Ende and colleagues found no delayed effect and concluded that the large effects reported in clinical practice can probably be explained by regression to the mean and response to placebo.
Advocates raised a specific objection: 1 Gy per fraction is double the standard German fraction size, and the proposed anti-inflammatory mechanism is thought to have a non-monotonic dose response with an optimum near 0.5 Gy. On that reading, the Dutch trials tested a dose outside the effective window.
This was an easy objection to dismiss as special pleading. It turned out to be right.
LoRD-KNeA changed the picture
At ASTRO 2025, Byoung Hyuck Kim presented the Korean LoRD-KNeA trial. It is the strongest evidence this field has ever produced, and its design addresses the dose objection directly.
The trial randomized 114 patients across three Korean sites, with primary knee osteoarthritis of Kellgren-Lawrence grade 2 to 3 and a baseline walking pain score of 50 to 90 out of 100, to one of three arms: sham irradiation, 0.3 Gy in six fractions, or 3 Gy in six fractions of 0.5 Gy. Concomitant analgesics other than a rescue drug were restricted through the four-month primary endpoint, and reirradiation was not permitted. All participants completed treatment.
At four months, the OMERACT-OARSI responder rate was 70.3 percent with 3 Gy and 41.7 percent with sham, with a p-value of 0.014. The 0.3 Gy arm, at 58.3 percent, did not differ significantly from sham. Clinically meaningful improvement was reported in 56.8 percent with 3 Gy against 30.6 percent with sham. No treatment-related toxicity was observed.
Three features make this hard to dismiss. It was sham-controlled and the sham response of 41.7 percent confirms how large the placebo effect in this condition is. The fraction size was 0.5 Gy, the schedule the objection to the Dutch trials specified. And the dose-response across three arms is internally coherent: an ineffective dose behaved like sham, and the standard dose separated.
Anyone who has argued, as we have, that the randomized evidence in this indication is negative needs to update. It no longer is.
The problem inside the positive trial
The primary endpoint separated. Several of the secondary endpoints did not.
LoRD-KNeA reported no significant differences between arms in change from baseline on the visual analogue scale, in patient global assessment, in serum inflammatory markers, or in rescue drug consumption.
This deserves attention rather than dismissal. OMERACT-OARSI response is a dichotomized composite built from pain, physical function and patient global assessment. When a responder analysis separates while the continuous measures feeding into it do not, the most common explanation is the dichotomization itself: converting a continuous variable into a threshold classification can produce statistical separation from a distribution shift too small to register on the continuous scale. The result is not wrong, but it is more fragile than the headline figure suggests, and rescue drug consumption not differing is a meaningful omission in an analgesic claim.
The absence of any change in serum inflammatory markers is a separate issue. The entire mechanistic rationale for low-dose radiotherapy rests on modulation of the inflammatory response. A clinical effect without any detectable biological correlate is possible, since serum markers are insensitive to localized joint inflammation, but it leaves the mechanism unsupported by the trial that best demonstrates the effect.
The long-term and structural data, and what they can bear
The most extensive long-term randomized dataset comes from Arkhangelsk. Between 2012 and 2014, 292 patients with Kellgren-Lawrence stage 0 to 2 knee osteoarthritis were randomly assigned to standard therapy with glucosamine and chondroitin plus NSAIDs, or the same therapy plus orthovoltage radiotherapy delivering 4.5 Gy in ten fractions of 0.45 Gy.
At approximately a decade, disability attributable to knee osteoarthritis had been established in 9.5 percent of the radiotherapy group and 17.8 percent of the control group, with a crude hazard ratio of 0.49 and a confidence interval of 0.26 to 0.95, strengthening to 0.24 after adjustment. A separate nine-year analysis of knee arthroplasty reported an incidence density ratio of 0.60 with a confidence interval of 0.18 to 1.88 and a p-value of 0.340. Ten-year radiographic analyses from the same cohort have reported less structural progression in the irradiated group.
These are the only long-term randomized data in this indication and they should not be ignored. But four features constrain what they can support.
The trial was open-label with no sham. The primary long-term endpoint, formal establishment of disability, is an adjudicated determination influenced by patient-reported symptoms and by clinical documentation, which is precisely the type of endpoint most sensitive to unblinding.
The control arm was glucosamine and chondroitin. These are not established disease-modifying agents, so the comparison is against a treatment with no expected structural effect rather than against best available care.
The population included Kellgren-Lawrence stage 0, meaning patients with no radiographic osteoarthritis at all. Structural progression endpoints behave very differently in a population that starts without radiographic disease, and the radiation protection calculus in that group is unlike anything in oncologic practice.
Finally, the arthroplasty analysis was not statistically significant, with a confidence interval spanning a factor of ten, yet the published conclusion describes a two-thirds reduction in arthroplasty risk. That gap between the statistic and its summary is the kind of thing that should be flagged rather than repeated.
What the presentation needs to contain
The abstract is embargoed and the program listing does not describe the design, so these are the questions rather than the answers.
Which trial is this. The combination of patient-reported, structural and long-term clinical endpoints in a randomized design matches the Arkhangelsk cohort profile closely, but the identification is not established from the program listing and matters for how the results should be read.
Was there a sham arm. After LoRD-KNeA, an unblinded comparison adds very little to the symptomatic question. The value of the presentation lies almost entirely in whether the structural and long-term findings come from a blinded design or an open one.
How structural progression was assessed and by whom. Kellgren-Lawrence grading has substantial inter-reader variability, and progression scoring over a decade in a population starting at stage 0 to 2 is exactly where reader expectation operates. Blinded, independent, duplicate reading with reported agreement statistics is the minimum standard. Quantitative measures such as joint space width, cartilage morphology or bone marrow lesion volume would be stronger than categorical grading.
Whether structural and clinical outcomes agree. In osteoarthritis research, disease modification requires concordance between structural preservation and clinically meaningful benefit. Imaging endpoints alone are not accepted surrogates. Radiographic difference without symptomatic or functional benefit is uninterpretable, and symptomatic benefit without structural change supports a symptom-modifying treatment and nothing more.
Attrition over the long-term interval. After a decade, the number remaining under observation, the handling of missing data, and the treatments patients received in the intervening years all bear more weight than the effect estimate.
Safety, which is now a live question rather than a formality
Acute toxicity at these doses is negligible and LoRD-KNeA reported none. The issue is different.
The American Radium Society published Appropriate Use Criteria for low-dose radiotherapy in osteoarthritis in 2026, and with a credible positive randomized trial behind it, adoption in the United States will accelerate. The population is very large, delivery is inexpensive, and reimbursement exists.
The relevant question is who gets treated. For a 70-year-old with disabling symptoms and exhausted conservative options, a small theoretical carcinogenic risk is easily justified. For a patient in their thirties with Kellgren-Lawrence grade 0 to 1 disease, the same calculation is entirely different, and the Arkhangelsk population sits in exactly that range. If the structural data are used to argue for earlier intervention in less symptomatic patients, the risk-benefit analysis shifts in a direction the efficacy data do not address.
Assessment
We were wrong about the state of this literature a year ago, and it is worth saying so plainly. The symptomatic question is no longer open in the way it was. A sham-controlled trial with a coherent three-arm dose response is real evidence, and the objection that rescued low-dose radiotherapy from the Dutch trials turned out to be the correct one.
What remains open is everything the presentation title promises beyond pain. Structural modification in osteoarthritis is a claim that has defeated far better-resourced programs, and the existing long-term data supporting it come from an open-label trial with a subjective primary endpoint and a published conclusion that outruns its own statistics.
Expected findings
We expect the presentation to report durable patient-reported benefit and to report structural findings favoring radiotherapy. We expect those structural findings to come from radiographic grading rather than quantitative imaging, and we expect the long-term clinical endpoints to be positive.
Whether that matters depends almost entirely on blinding. If structural progression was scored by independent blinded readers with reported agreement, and it separates, this becomes a genuinely important presentation and the first serious signal that a treatment delivered in six small fractions might alter the natural history of osteoarthritis. That would warrant a large, blinded, structurally endpointed confirmatory trial, and the field should build one rather than treat the finding as settled.
If the structural assessment was unblinded, the finding is compatible with reader expectation and adds nothing to what LoRD-KNeA already established, which is that low-dose radiotherapy at 0.5 Gy per fraction relieves symptoms in a proportion of patients beyond placebo. That is a real and useful result. It is not disease modification, and the distinction should survive the applause.
We will report on the presentation from ASTRO 2026.
This is the fifth article in the RadOncToday series previewing the Clinical Trials and Plenary sessions at ASTRO 2026 over the three weeks before the meeting.