This systematic review and meta-analysis was aimed at investigating the effect and safety of acupuncture for the treatment of chronic spinal pain.

The authors included 22 randomized controlled trials (RCTs) involving patients with chronic spinal pain treated by acupuncture versus sham acupuncture, no treatment, or another treatment were included. Chronic spinal pain was defined as:

  • chronic neck pain,
  • chronic low back pain,
  • or sciatica for more than 3 months.

Fourteen studies had a high risk of bias, 5 studies had a low risk of bias, and 5 studies had an unclear risk of bias. Pooled analysis revealed that:

  • acupuncture can reduce chronic spinal pain compared to sham acupuncture (weighted mean difference [WMD]  -12.05, 95% confidence interval [CI] -15.86 to -8.24),
  • acupuncture can reduce chronic spinal pain compared to mediation control (WMD -18.27, 95% CI -28.18 to -8.37),
  • acupuncture can reduce chronic spinal pain compared to usual care control (WMD -9.57, 95% CI -13.48 to -9.44),
  • acupuncture can reduce chronic spinal pain compared to no treatment control (WMD -17.10, 95% CI -24.83 to -9.37).

In terms of functional disability, acupuncture can improve physical function at

  • immediate-term follow-up (standardized mean difference [SMD] -1.74, 95% CI -2.04 to -1.44),
  • short-term follow-up (SMD -0.89, 95% CI -1.15 to -0.62),
  • long-term follow-up (SMD -1.25, 95% CI -1.48 to -1.03).

Trials assessed as having a high risk of bias (WMD −13.45, 95% CI −17.23 to −9.66, I 2 96.2%, moderate-quality evidence, including 14 studies and 1379 patients) found greater effects of acupuncture treatment than trials assessed as having a low risk of bias (WMD −11.99, 95% CI −13.94 to −10.03, I 2 44.6%, high-quality evidence, including 4 studies and 432 patients), but smaller effects than trials assessed as having an unclear risk of bias (WMD −14.51, 95% CI −17.25 to −11.78, I 2 0%, high-quality evidence, including 3 studies and 190 patients).

Only 6 trials provided information on adverse events. No trial reported data on serious adverse events during acupuncture treatment. The most frequent adverse events were temporarily worsened pain and needle pain at the acupuncture site, which can decrease quickly after a short period of rest.

The authors concluded that compared to no treatment, sham acupuncture, or conventional therapy such as medication, massage, and physical exercise, acupuncture has a significantly superior effect on the reduction in chronic spinal pain and function improvement. Acupuncture might be an effective treatment for patients with chronic spinal pain and it is a safe therapy.

I think this is a thorough review which produced interesting findings. I agree with most of what the authors report, except with their conclusions which I find too optimistic. In view of the facts that

  • only 5 RCTs had a low risk of bias,
  • collectively, the rigorous trials reported smaller effect sizes,
  • the majority of trials failed to mention adverse effects which, in my view, casts considerable doubt on their quality and ethical standard,

I would have phrased the conclusion differently: compared to no treatment, sham acupuncture, or conventional therapies, acupuncture seems to have a significantly superior effect on pain and function. Due to the lack rigour of most studies, these effects are less certain than one would have wished. Many trials fail to report adverse effects which reflects poorly on their quality and ethics and prevents conclusions about the safety of acupuncture. In essence, this means that the effectiveness and safety of acupuncture as a treatment of chronic spinal pain remains uncertain.

4 Responses to Acupuncture for spinal pain: findings from a new meta-analysis

  • A number of years back, I lived in S. Korea. I was having a terrible fibromyalgia flare up and I was talking to my boss about it. She suggested acupuncture. I was so desperate, I tried it. It did nothing for the pain. Not surprising. Laying on a warm bed with calm music playing relaxed me. But still had the pain. I also got some sort of TCM like tea. It was disgusting. I refused to drink it. The only thing interesting about the visit to the “hospital” was looking at the little museum there. Needless to say, I didn’t go back and never said anything to my boss.

  • For me this systematic review and meta-analysis inspires far less confidence. I think there are unfortunate methodological errors in the analysis. Let’s take a look at the forest plot for figure 3A, I think it’s fair to say this is the “main results figure”, which shows the most data for the outcomes the authors were interested in. We can see “a” and “b” entries entries for Giles 2003, Cho 2013, Seo 2017 and Leibing 2002. In the case of Seo 2017 the “a” and “b” refer to separate studies which is fine. The remaining trials with parts “a” and “b” have been included in this meta-analysis erroneously.

    For the Giles 2003 study the authors have made multiple comparisons of the same treatment group to separate control groups and entered both comparisons into the meta-analysis separately, thus “double-counting” the acupuncture group. In this case the acupuncture group was compared to “manipulation” and the same acupuncture group was compared to “medication”. The appropriate course of action would be to combine the two control groups or perhaps select one appropriate control group to compare the acupuncture group to.

    For the Leibing 2002 trial a similar mistake was made by comparing the same acupuncture treatment group to the “sham group” and the “control group” and then entering results from both comparisons into the meta-analysis “double-counting” the acupuncture group again.

    For the Cho 2013 study the authors have made separate entries of the same comparison at two different time points; “real acupuncture” compared to “sham acupuncture” at “end of treatments” and then “real acupuncture” compared to “sham acupuncture” at “primary endpoint” are entered as separate comparisons under “Cho 2013a” and “Cho 2013b”. The correct course of action would be to choose a single appropriate time point to enter into their meta-analysis.

    The first two instances of “pseudo-replication” are critical (but common) errors when dealing with trials that have multiple treatment groups. The Chocrane handbook for systematic reviews (Version 5.1.0) makes a couple of comments on this; Section 16.5.4: “One approach that must be avoided is simply to enter several comparisons into the meta-analysis when these have one or more intervention groups in common. This ‘double-counts’ the participants in the ‘shared’ intervention group(s) and creates a unit-of-analysis error due to the unaddressed correlation between the estimated intervention effects from multiple comparisons.” Also refer to section 9.3.9: “A serious unit-of-analysis problem arises if the same group of participants is included twice in the same meta-analysis (for example, if ‘Dose 1 vs Placebo’ and ‘Dose 2 vs Placebo’ are both included in the same meta-analysis, with the same placebo patients in both comparisons).”

    The error with Leibing 2002a/b seems far less forgivable. Due to these mistakes the forest plots in figure 3 have several more entries than they deserve generating an impression of an abundance of trial data that doesn’t really exist. It is tempting to correct and replot their analyses to see if it makes any difference to the results but I generally don’t bother with that process any more. It is my concern that if the authors have made critical errors at the data analysis stage it’s fair to suspect the entire process. You can go back and check the data extraction but if the literature search and selection wasn’t done correctly this is all a waste of time.

    Actually I have a whole list of other issues with this paper but I don’t want to spend too long on this comment which is already spiralling out of control… Briefly; I don’t think it’s appropriate to plot an overall estimate from back pain, neck pain and sciatica estimates when the results are derived from (some of) the same patients.

    Oh, and I also disagree with the authors choice to include studies with such diverse treatments all under the banner of “acupuncture”, is “bee venom acupuncture” really comparable to just using needles?? There are also some trials from where the data has been pulled that I think are somewhat… interesting? In particular the Zaringhalam 2010 study I would like to have a closer look at.

    I’d be interested to hear anyone’s thoughts, perhaps I’ve made some mistake in my thought process here. I might cross-post a revision of this comment to pubpeer at some point, with permission.

    • thanks!
      that’s very helpful.
      I must admit I did not bother to look at the primary studies.
      very good to see someone who is more thorough!!
      [pemission granted]

  • following on from the above – acupuncture studies nearly tend to be far from rigorous and amalgamating a number of such studies in this way merely allows for obfuscation.

    In addition I believe that there are the serious underlying issues that should always give pause for thought. The underlying premise is highly implausible. Meridians and Qi have never been shown to exist, There is zero agreement on the “correct” placement points of acupuncture needles. The myriad variations of acupuncture types including auricular etc all of which are claimed to work and even acupressure and dry needling.
    The fact that sham seems to work as well as “real.” That toothpicks work as well as needles. That acupuncture in a rubber fake hand has been shown as effective as in the real thing. That “ancient” acupuncture has been shown to be a myth.
    That all they ever get are studies showing weak evidence.
    That it is all really nothing more than a theatrical placebo.

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.

Subscribe via email

Enter your email address to receive notifications of new blog posts by email.

Recent Comments

Note that comments can be edited for up to five minutes after they are first submitted but you must tick the box: “Save my name, email, and website in this browser for the next time I comment.”

The most recent comments from all posts can be seen here.