The result for worse knee pain vanished on a second look
Among 354 people starting with moderate-to-severe pain the pair beat placebo, 79% against 54%. A trial enrolling that exact group, 164 people, found no advantage.
What the study found
The one subgroup that responded did not respond when it was retested
Among the 354 people in the big trial who started with moderate-to-severe pain, the combination did beat placebo, 79% responding against 54%. Its own authors called that exploratory. A Spanish trial then enrolled that exact group, 164 people for six months, and found no superiority.
The caveat · The retest trial was small
The subgroup was one of two specified in advance, but was never powered as a test of its own and no independent trial has repeated it. The Spanish trial is small, at 164 people, and its placebo group's larger improvement appears only when everyone randomised is counted, vanishing among completers.