Do patient characteristics affect acupuncture treatment outcomes?

Inspired by Witt et al Clin J Pain 2019.[1]

Photo by MC, taken at Kew Gardens Temperate House.

This paper is clearly important to consider, despite the limited conclusion of the abstract, since it comes from the ATC (Acupuncture Trialists Collaboration) and involves further analysis of the largest dataset available in the field of acupuncture for pain.

I cannot help hearing the voice of a good friend and colleague whenever I discuss this big dataset:

“It’s like stirring mud” he says.

TU

What does he mean by that? My friend likes very tight experimental and clinical research which gives definitive answers to well defined questions. That is quite different from: “Is acupuncture efficacious in chronic pain?” Clinical uncertainty and variability on chronic pain conditions in practice translates to increased variability in measures used in sample populations of trials and this means it is hard to see subtle effects because they disappear in the variations (the statistical noise), hence the reference to ‘stirring mud’.

Systematic reviewers can be divided into either ‘lumpers’ or ‘splitters’.

MC

Systematic reviewers can be divided into either ‘lumpers’ or ‘splitters’. The lumpers go for looser selection criteria for the trials they combine and consequently larger overall data sets. They hope that more data will mean clearer results – they sacrifice homogeneity for statistical power. The splitters do the opposite – they place homogeneity over power, and often have too little statistical power to draw conclusions. When we consider the ATC dataset, we must consider this to be towards the lumping end of this spectrum, and hence the term ‘stirring mud’, although it is a huge pot that is being stirred. So it is hard to see anything no matter how hard we stir, but if we are lucky, a large gold nugget might come to the surface. The small nuggets (of clinical wisdom) never will surface by this method I’m afraid, but big population level factors should become apparent.

So what did Claudia (Witt) and her colleagues look for?

  • Age
  • Gender
  • Pain duration
  • Baseline pain severity
  • Baseline psychological distress

Their data set included 25 trials with sham acupuncture controls and 7097 patients, and 25 trials with no acupuncture controls and 16 041 patients. Certainly impressive numbers (large pot), and the most powerful statistical methods were used rather than relying on those from the original trials (the smoothest and thinnest mud possible, given where it came from).

Nothing was found for age or pain duration. So in practice we should not be worried about either factor. Having said that, I think the latter (pain duration) is relevant in neuropathic pain conditions because of the structural changes that occur with time in the spinal cord, in particular the loss of inhibitory interneurons. This dataset did not include neuropathic pain.

Baseline psychological distress appeared to have a very small effect, and this may simply represent an association with the key characteristic that determined acupuncture treatment outcome – baseline pain severity. The most severe baseline pain is likely to be correlated with psychological distress.

It is useful to have the association of baseline pain severity with bigger effects of acupuncture treatment confirmed. This confirms both our clinical impression as well as being in line with mechanistic ideas of reduced central amplification of nociception via augmentation of descending inhibition. The same quantity of inhibiting neuromodulators will have a bigger effect via a membrane that is more sensitized.

In terms of choosing patients, of course, this does not really help us narrow our population, unless perhaps we are recruiting for an RCT, where we often see a minimum baseline pain requirement for inclusion.

“…the results of sex were inconsistent”

[1]

I have yet to comment on gender! Claudia says: “…the results of sex were inconsistent”, which prompted me to look at the numbers. We see highly significant but small effects in opposite directions depending on the control used. So there were bigger effects for male patients in sham controlled trials and bigger effects for female patients in trials with no acupuncture controls. This is intriguing is it not?

If I might be allowed to speculate… well who is going to stop me? This is a blog after all :D. My guess would be that women respond to context effects of treatment more effectively by being more aware of their environment, whereas men tend to focus on specifics and be less aware of more peripheral aspects. This would translate to a slightly bigger effect of sham techniques in women, and a bigger effect overall in clinical practice. If we want to measure the biggest effects for acupuncture then, we should use only men in sham controlled trials and only women in open trials. I am being deliberately black and white here to make the point. In reality there are many arguments against such an approach.

Before I sign off, this discussion of different effects in men and women reminds me of a question I posed at a BMAS meeting in the Barber-Surgeons’ Hall in 2010. The question was addressed to Nadine Foster, who was presenting her data on acupuncture for knee osteoarthritis, which included a formal cost-effectiveness analysis.[2,3] I suggested, rather provocatively, that we should only be using acupuncture in treating women with OA knee, since Claudia’s data (from the ARC OA knee trial [4]) on cost utility showed that the cost per additional QALY (quality adjusted life year) in men was much higher, and outside the threshold for provision in state-funded systems (such as the NHS).[5] The figures were quite dramatic, with the cost for men being in the range €100-250k and for women €10-20k per additional QALY. Of course Nadine immediately dismissed this suggestion of clinical segregation by gender, but should we revisit this?

Is it right for us to discard the nuggets that surface when we stir the mud just because they are inconvenient or appear to be politically unpalatable?

References

1         Witt CM, Vertosick EA, Foster NE, et al. The Effect of Patient Characteristics on Acupuncture Treatment Outcomes. Clin J Pain 2019;35:428–34. doi:10.1097/AJP.0000000000000691

2         Foster NE, Thomas E, Barlas P, et al. Acupuncture as an adjunct to exercise based physiotherapy for osteoarthritis of the knee: randomised controlled trial. BMJ 2007;335:436. doi:10.1136/bmj.39280.509803.BE

3         Whitehurst DGT, Bryan S, Hay EM, et al. Cost-effectiveness of acupuncture care as an adjunct to exercise-based physical therapy for osteoarthritis of the knee. Phys Ther 2011;91:630–41. doi:10.2522/ptj.20100239

4         Witt CM, Jena S, Brinkhaus B, et al. Acupuncture in patients with osteoarthritis of the knee or hip: a randomized, controlled trial with an additional nonrandomized arm. Arthritis Rheum 2006;54:3485–93. doi:10.1002/art.22154

5         Reinhold T, Witt CM, Jena S, et al. Quality of life and cost-effectiveness of acupuncture treatment in patients with osteoarthritis pain. Eur J Heal Econ 2008;9:209–19. doi:10.1007/s10198-007-0062-5


Declaration of interests MC