Many of my blogs have a recurring theme: an intervention that sounds great (biologically plausible), has great early results, but on rigorous scientific testing is shown to be ineffective. Prolotherapy does not fit that mould, because I don’t even buy the biological argument in the first place. Prolotherapy involves the injection of irritant solutions in order to strengthen tissues and relieve pain. That makes about as much sense as homeopathy, or being struck on the head to relieve a headache.
Searching prolotherapy in the medical literature doesn’t get you far. There is a Cochrane review from 2010 that shows it to be ineffective on its own for back pain. When I searched for individual research articles, I found 75 articles in Medline. Of these, there were 8 randomised trials. Some of these compared prolotherapy to corticosteroid injections, giving it the illusion of effectiveness; here’s how. For each of these studies, the two treatments showed similar treatment effects. The conclusion was that prolotherapy was as good as corticosteroid injections. While technically correct, the illusion lies in the implication, or underlying assumption, that both treatments are effective. The conclusion does not tell us whether either treatment is more effective than placebo. Most of the placebo studies (here, here and here) show prolotherapy to be ineffective or barely effective. The placebo studies for corticosteroid injections aren’t much better and will be the subject of a future blog post.
Comparing one treatment to another is commonly done, and often leads to this illusion of effectiveness, because both treatments make use of the placebo effect. We conclude that both treatments are equally effective, without considering the equally valid conclusion that both treatments are equally ineffective. This is the problem with so called Comparative Effectiveness Research: even the title is misleading, because it assumes an effect, and by excluding comparisons with an inactive control (placebo), comparative studies often ask the wrong question.