We thank the correspondents for their interest in our recent review [1], and appreciate the opportunity to reply.

Clinical trumps statistical – synthesise the evidence to inform practice

A review of a clinical intervention can include both (i) quantitative (e.g. meta-analyses of effect sizes) and (ii) qualitative (e.g. subjective scoring and weighing of the available evidence according to stated criteria) aspects. This blog speaks to the claim that having included a non-randomised trial in our meta-analysis invalidates the review.

As suggested in the Cochrane Handbook [2] and stated in our methods, we included non-randomised trials if they were reasonably resistant to biases, relatively homogeneous, controlled for confounders, and were prospective in nature. Since the primary aim of this review was informing the clinician, for whom ESWT may be a treatment option, the main body of the text emphasised the GRADE recommendations; other objective data were placed in an online appendix which we felt was unlikely to ever be read by a non-clinician, and therefore unlikely to ever be misinterpreted and used to make a clinical decision on an actual patient.

Stated plainly, only university researchers, not practicing clinicians, would likely wade through the 134 pages of appendices so we felt there was no danger of harm to actual patients through misinterpretation of a forest plot buried in there. These data are simply included so that any academics wishing to conduct future research will have this as a resource, not to inform daily clinical practice.

We have reviewed our recommendation for ESWT in GTPS (Low to very low level evidence for 13 outcome measures, and “Not applicable” rating for the remaining 6) and are comfortable with this rating. Upon reflection, we see no need to change these recommendations.

Small trials and/or missing data

Further to this point, the correspondents suggest that very small trials should be included in the meta-analyses. As an academic exercise, the correspondents could readily evaluate the tables and forest plots provided in the appendices in terms of inclusion of the small trial [3], in a matter of minutes – indeed it is for this express purpose we included these data as online supplements. Nevertheless, we have reviewed characteristics of the studies under discussion and we strongly suggest that the addition of 20 more patients (less than 10 subjects per group after drop outs) in a pooled sample of 114 would not have shifted significantly (statistically or clinically) the presented results.

However, we apologise for the confusion here as we believe the correspondents feel that we may have excluded the small trial of Taunton et al [3] only because of the small sample size. This study would have been excluded irrespective of the sample size issue for technical reasons – no standard deviations for the outcome measures were available (despite contacting the authors) which prevented pooling these data[a]. This information was not reported as we were already substantially over the word count limit for the journal.

A single good RCT can provide a moderate level of evidence – maybe more, maybe less

Next, the correspondents suggest that we were too generous in grading as “moderate” the level of evidence for the research on proximal hamstring tendinopathy. Here the evidence is from one well-conducted randomized controlled trial [4]; we are unsure why the correspondents are referring to “observational studies”. When rating evidence quality, RCTs start as high quality evidence while observational studies start as low quality, and can then be modified according to influencing factors [5]. In our review we arbitrarily set a more strict criteria for high quality evidence and stated a priori that “an outcome with only one trial (will be graded) as low quality”. We subsequently upgraded the strength by one step to “moderate” due to the large effect size reported [5]. Again, upon reflection we believe that this requires no revision to the paper.

Caution is needed when reading papers, not just appendices

Finally, we agree with the correspondents that the conclusions need to be treated with caution, as we also concluded:

 “Caution needs to be exercised however in the clinical interpretation of these findings as it is likely that patient-specific individual confounding factors may have a significant effect on ultimate treatment outcomes, and these confounding aspects have not been fully examined.”

Korakakis Vasileios & Whiteley Rodney

Doha, Qatar

References:

1. Korakakis, V., et al., The effectiveness of extracorporeal shockwave therapy in common lower limb conditions: a systematic review including quantification of patient-rated pain reduction. British Journal of Sports Medicine, 2017: p. Online First: 27 September 2017. doi: 10.1136/bjsports-2016-097347.
2. Higgins, J.P.T. and S. Green, Cochrane Handbook for Systematic Reviews of Interventions. Version 5.1.0 ed. 2011: The Cochrane Collaboration.
3. Taunton, J.E., K.M. Taunton, and K.K. M., Treatment of patellar tendinopathy with extracorporeal shock wave therapy. B C Med J, 2003. 45(10): p. 500-507.
4. Cacchio, A., et al., Shockwave therapy for the treatment of chronic proximal hamstring tendinopathy in professional athletes. American Journal of Sports Medicine, 2011. 39(1): p. 146-153.
5. Guyatt, G., et al., GRADE guidelines: 1. Introduction – GRADE evidence profiles and summary of findings tables. Journal of Clinical Epidemiology, 2011. 64(4): p. 383-394.

[a]It may be possible to pool the treatment outcome data simply as “successful/not successful” however, from the source material, we find it difficult to support this. Specifically the text[3] is: “After treatment, anecdotal reports from the treatment group revealed five out of nine subjects had decreased pain and increased function; two subjects still experienced pain with stairs but felt they had improved overall, and two subjects reported still feeling pain with activity. In the control group, six of seven subjects reported no improvement or change. One of the control subjects experienced some pain relief after the five treatment sessions, but this subject also reported decreased frequency of activity during the study.”