No evidence for stratified exercise therapy being cost-effective compared to usual exercise therapy in patients with knee osteoarthritis: Economic evaluation alongside cluster randomized controlled trial.

The trial was stopped early because of the COVID-19 pandemic, with 335 patients instead of the 408 planned, so it was underpowered and the cost estimates carry very wide confidence intervals. The cost saving only appeared from a societal perspective, which counts lost work productivity. When only healthcare costs were counted, the difference went the other way, by €30 per patient. Half the participants were already retired (53% in the experimental arm and 48% in the control arm), and those still working already had high work ability and low sick leave, so there was little room for work-related costs to differ. The contrast between the two arms was smaller than intended: the control group had fewer physiotherapy sessions than expected (9.6), and in the obesity subgroup the combined exercise and diet intervention was not delivered as recommended. Costs came from patient-reported questionnaires, which depend on people remembering what they spent and used. Because the pandemic cancelled or postponed primary and especially secondary care for periods, healthcare use in this trial may be lower than usual and the healthcare cost figures may not generalise. People over 85, under 40, with pain of 9 or more out of 10, or with severe conditions that make exercise unsafe were excluded, so the findings do not apply to them. Because of the length of treatment, the trial could not show possible long-term savings such as fewer knee replacements.

All authors declare that they have no competing interests. The article text provided does not state who funded the trial.

Do not read this as evidence that stratified exercise therapy saves money, even though the paper describes the approach as superior on cost. The saving came almost entirely from work-related costs and vanished when only healthcare costs were counted, where the difference was €30 per patient in favour of usual care. The confidence interval on total costs ran from €1728 saved to €918 extra, the quality of life difference was 0.006 on a scale from 0 to 1, and the trial stopped early with fewer participants than planned.