The lack of convergence of findings from randomised evaluations of EWST for plantar heel pain has resulted in clinical uncertainty about its effectiveness. Within this systematic review, we have been able to evaluate the effectiveness of ESWT in a meta-analysis and used the pooled data to arrive at more precise conclusions about its usefulness in clinical practice.
The meta-analysis shows a statistically significant benefit with ESWT on plantar heel pain from outcomes of 897 patients' VAS scores of morning (first-step) pain assessed at or around 12 weeks but we do not consider this clinically significant since the observed benefit equates to less than one half centimeter on a 10 cm VAS. The 95% confidence interval is compatible with a mean treatment benefit of at most 0.83 cm. A sensitivity analysis including only those higher quality trials did not produce evidence of a statistically significant benefit. Only one trial included in the review discussed what might constitute a clinically meaningful reduction in plantar heel pain: Buchbinder et al [10
], suggest that 0.7 cm reduction of heel pain may not be clinically relevant.
We included one trial in the meta-analysis which used sub-clinical doses as controls [10
] and combined these patient outcomes with those from trials which used sham treatments as controls [11
]. All six trials [10
] also used different doses of ESWT but, despite the differences in the use of control interventions and doses, no evidence of heterogeneity in the patient outcomes was detected in the pooled estimate (figure ). Nor does there appear to be a dose-response relationship for ESWT; trials using both high and low doses have reported similar effects as is evident from the estimates from the trials by Haake et al [11
] and Abt et al [21
] (Table , figure ).
We were grateful to the authors of trials included in this review who provided supplementary data in response to our correspondence [10
] but disappointed that data from all 11 trials were not available to us. Five trials were not included in the meta-analysis either because adequate data were not provided [22
] the timing of the outcomes differed greatly from the other trials [31
] or the outcomes were clinically irrelevant [30
]. Consequently, information about the effects of ESWT in 310 patients with heel pain was effectively lost to re-analysis. Any future reporting of patient outcomes should include means of pain scores with measures of variance in order that new trials can be included in meta-analyses and weighted mean differences and confidence intervals calculated [36
Rompe et al conducted a small trial (n = 40) which evaluated the benefits of ESWT in running athletes [31
] and reported a mean difference of 2.60 (95% confidence interval 1.37 to 3.83) for morning pain at 6 months. This effect size is statistically significantly different from the combined outcomes presented in Figure but not statistically different from the mean difference in outcomes reported in the small trial by Abt et al [21
] 2.00 (95%confidence interval 0.47 to 3.53) at 19 weeks (n = 37). That the two smallest trials included in the review should produce between-group comparisons of pain in the morning that reach statistical significance when estimates from larger studies do not is surprising. Sample size is an important factor in experimental bias in clinical trials as effect size estimates from small studies can be highly variable [37
]. The effect sizes from these small studies may be due to ESWT being beneficial in certain sub groups within the population (e.g. runners), or may be as a result of a failure to blind the participants successfully to their treatment allocation, as previously reported by one of the authors [30
]. Alternatively, these data may be aberrant values that are more likely to occur by chance in small studies than larger ones [38
ESWT was not considered a suitable therapy for the first-line management of heel pain by the majority of the investigators. This may be because of limited access to this relatively new and expensive equipment or, more likely, because of the favourable natural history of this condition.
In the absence of a validated heel pain specific outcome measure, our a priori
choice of morning pain as the primary outcome measure was vindicated by eight of the of the eleven included trials collecting morning pain or first step/start up pain outcomes. One trialist [10
] used a problem elicitation technique
which confirmed "walking after getting out of bed in the morning" as the most frequently reported problem by patients with heel pain. We had planned to pool additional secondary outcome measures, such as walking pain, but this was not possible because of the diversity of the outcome measures used and differences in the data collected. Some of the outcomes that have been used to assess the effects of treatments were clinically irrelevant in our opinion [30
]. Night pain and resting pain are not symptoms that we commonly encounter in patients seeking treatment for plantar heel pain. Three trials [11
] incorporated the Roles Maudsley scale and one trial [10
] used the Maryland Foot Score as measures of disability. It is commendable that two of the investigators [10
] used generic health outcomes, SF36 and SF 12 respectively. Future trials should include outcomes of disability as well as the impact on health related quality of life and not just pain when assessing the effect of interventions for heel pain.
Of the eight outcomes listed in Table , only "pain at rest" is distinct with four of the five trials [11
] favouring ESWT compared with placebo or reduced dose. As previously discussed, this outcome measure is not a key feature of plantar heel pain. All other outcome measures are equivocal.
Minimal side effects were reported by Abt et al [21
] and Buchbinder et al [10
]. The most frequently reported adverse event from the use of ESWT is pain [11
] which appeared to affect some patients both during and after the procedure.
The quality of reporting varied amongst trials. The three most recent trials [10
] all received above average quality scores for trial reporting. This is an encouraging development for those interested in improving the outcomes for patients who have heel pain and may reflect both the use of checklists such as the CONSORT statement [36
] for trial reports now demanded by many journal editors as well as a greater awareness of good trial reporting practice by trialists themselves. There was however, a contrast in the results obtained from the four better quality trials, scoring three or above, when meta-analyzed separately from the two poorer quality trials. Better quality trials did not favour ESWT whilst the poorer quality ones did.
At least two of the trials included in our meta-analysis, received some form of sponsorship from a company manufacturing ESWT [27
] although this has not been made explicit within the published papers. Both these trials reported significant benefit from ESWT. One further trial Haake et al [11
] declared being supplied with the ESWT equipment and reported no statistically significant effects between the two groups. Six of the trials [21
] have not made it clear whether there is any conflict of interest or not. In a systematic review to investigate whether the funding of drug studies by the pharmaceutical industry is associated with bias, Lexchin et al [39
] concluded that industry sponsorship was more likely to produce results favouring the sponsors' product than studies funded from other sources.
In view of concerns about publication bias, it is encouraging that three large, negative trials have been published in high impact journals. We were unable to recognize the existence of small, unpublished studies showing no statistically significant benefits. However, the existence of any such trials would only serve to endorse the findings of the meta-analysis in this systematic review.