Cell-Based Versus Corticosteroid Injections for Knee Pain in Osteoarthritis: A Randomized Phase 3 Trial

Major Points and Findings:

The MILES study is the largest randomized trial of cell injections for knee osteoarthritis published so far, with 480 patients at five US centres and FDA oversight. Its result was negative. One year after a single injection, none of three cell products, including cultured umbilical cord tissue mesenchymal stromal cells, reduced knee pain more than an ordinary corticosteroid injection. Pain improved by a clinically meaningful amount in all four groups, and MRI scores did not change in any of them.

The study is relevant to anyone considering cell therapy for an arthritic knee. It gives no indication that the cells are unsafe, and it tested one fairly low dose given once. In a patient-blinded comparison, the improvement people felt after a cell injection was matched by a standard steroid injection.

Aim:

To find out which source of cells works best for knee osteoarthritis, and whether any of them beats corticosteroid. The authors expected cell therapies to be superior at one year. That hypothesis was rejected.

Methods:

MILES was a phase 2/3, multicentre, single-blind, randomized controlled trial (NCT03818737) at Emory, Duke, two Sanford Health sites and the Andrews Institute, funded by the Marcus Foundation. Between March 2019 and June 2021, 570 people were screened and 480 randomized.

Participants were adults aged 40 to 70 with X-ray confirmed tibiofemoral knee osteoarthritis of Kellgren-Lawrence (KL) grade II, III or IV and continued pain despite conservative treatment. Exclusion criteria included a tense knee effusion, varus or valgus deformity of more than 10 degrees, hyaluronic acid within 6 months or PRP or stem cells within 1 year in that knee, and any corticosteroid injection within 3 months. Daily opioid use, malignancy within 5 years, autoimmune disease needing immunosuppression, uncontrolled diabetes and known osteoporosis were also excluded. Average age was 58.3 years, average BMI 30.8, and 55% were women. KL grades were spread evenly, with 30% grade II, 40% grade III and 30% grade IV.

There were four treatments, each given as one ultrasound-guided injection after aspirating any joint fluid:

• BMAC: autologous bone marrow aspirate concentrate. 61 mL of marrow was drawn from six sites on the back of the pelvis and spun down to 7 mL.
• SVF: autologous stromal vascular fraction. At least 100 mL of fat was removed by syringe liposuction, digested with an enzyme system for 30 minutes and washed. Of this, 5 mL was injected.
• UCT: allogeneic umbilical cord tissue MSCs, made at Duke’s GMP laboratory from cords donated after elective caesarean delivery, culture-expanded to passage 2, frozen in 10% DMSO, shipped, thawed in the clinic and diluted in saline to 7 mL. The release criterion was between 10 and 20 million nucleated cells.
• CSI: corticosteroid injection, 1 mL of Depo-Medrol (methylprednisolone) in 6 mL of saline.

Blinding: Within each of three arms, patients were randomized 3:1 to cells or steroid. Those assigned to steroid in the BMAC and SVF arms underwent a sham marrow or fat harvest, and syringes were wrapped in opaque tape. The treating doctors were not blinded.

Outcomes: The co-primary outcomes were change in pain on a 100 mm visual analogue scale (VAS) and in the KOOS pain score from baseline to 12 months, with visits at 1, 3, 6, 9 and 12 months. Secondary outcomes were EQ-5D, PROMIS-29, and a whole-knee MRI score (0 to 69) at baseline, 6 and 12 months. Analysis was by intention to treat.

Results:

Five products failed release testing (one BMAC, four SVF, all on endotoxin) and were not injected. In total 440 people received treatment.

Pain at 12 months: Every group improved, and the groups could not be separated.

• VAS change: BMAC -24.3, SVF -19.4, UCT -20.1, corticosteroid -20.9. Differences versus corticosteroid: BMAC -3.4 (P = 0.19), SVF 1.5 (P = 0.56), UCT 0.8 (P = 0.76).
• KOOS pain change: BMAC 19.1, SVF 17.2, UCT 16.2, corticosteroid 17.7. Differences versus corticosteroid: 1.4 (P = 0.49), -0.5 (P = 0.82) and -1.5 (P = 0.44).
• All sensitivity analyses gave the same answer.

In subgroup analyses, KL grade did not predict who responded. For VAS only, treatment interacted with age group (P = 0.02) and sex (P = 0.01). KOOS pain showed no interaction.

EQ-5D and PROMIS-29 followed similar courses in the four groups. Only PROMIS physical function differed in pattern over time (P = 0.048), with no clinically important difference apparent. MRI scores changed by less than one point in every group (BMAC 0.53, SVF -0.40, UCT -0.26, corticosteroid 0.30). Cell injections did not visibly rebuild the joint in a year.

Safety: There were no procedure-related serious adverse events, allergic reactions or joint infections. Nearly all events had settled within a week. The pattern differed by product:

• Joint swelling occurred in 24.1% after cord tissue MSCs against 7.4% after corticosteroid (P = 0.01), and in 17.8% after BMAC and 14.7% after SVF.
• Joint pain afterwards in about a quarter of every group (25.9% UCT, 25.9% corticosteroid).
• The harvests carried their own burden: bruising in 38.6% of the fat-harvest arm and 12.2% of the marrow-harvest arm, haematoma in 12.4% and 2.9%, and procedural pain in 33.8% and 29.5%, against 0.6% in the arm with no harvest.

Conclusions:

At one year there was no cell therapy more effective than a corticosteroid injection for knee osteoarthritis, and all the procedures were safe. The authors point out that the placebo response to knee injections is large, citing response rates of up to 50% and effects lasting 6 months or longer. There was no saline arm, so the trial cannot say how much of the improvement in any group was due to the injection itself. Doctors were unblinded, and one year may be too short for structural change on MRI. A 2 to 3 year extension is planned.

Background Information:

The cord MSC arm received 10 to 20 million cells once. Several smaller trials that report a benefit (see the 2025 meta-analysis by Cao and colleagues, also summarised on this site) used 50 to 100 million cells, so MILES does not rule out an effect at higher or repeated doses. Those trials, though, rarely had 120 patients per group.

On the product itself, the authors state that most commercial birth tissue products that have been tested contain no live MSCs. The product used here consisted of culture-expanded, GMP-manufactured cells, checked for count and viability at the bedside. Even so, it did no better than the steroid. The paper also recalls an earlier rigorous trial in which bone marrow concentrate in one knee and saline in the other produced equal pain relief.

Several authors work for Sanford Health, which has a financial interest in the maker of the SVF system. A correction of 22 December 2023 added two baseline rows to Table 1 and changed no results.