Abstract
We thank the authors for their letter and interest in our article. Their important conflicts of interest have been noted before, and we have answered many of their queries in previous correspondence.3,4 We are grateful for the opportunity to correct many of their misunderstandings and, as they do, to compare and contrast the START: REACTS and SPACE trials.
The authors correctly identify the review of Staibano et al, 9 but substantially misquote it with regard to our study. Our paper was complimented for having a low risk of bias. While they make the generic comment that poorly run sample size re-estimation trials can have statistical errors, these were not directed at our study. Our study was a group-sequential design, not a sample size re-estimation (this is clearly described in the 1-year paper and referenced prior statistical papers). A helpful summary of the differences between such designs is given by Bhatt and Mehta 1 in their 2016 New England Journal of Medicine review. The robustness of the design has been confirmed in subsequent analyses across multiple trials.1,5-7
The authors reference our figures for pain-free range of motion at baseline, a necessary measure used for the Constant score. This is typically a lower range than for active range of motion, which they quote. We are therefore unable to compare the two. Given this misunderstanding (and in the absence of an available measurement protocol in SPACE, which we have provided for START:REACTS), we question the validity of Constant scores in SPACE. 10 We are also unable to compare opioid use as this is not presented in Verma et al. 10
We only included upper border subscapularis tears of <1 cm, unlike the SPACE study, which included repairable tears, despite repair failure being common. Nonetheless, to reassure readers, a post hoc analysis of START:REACTS removing participants with subscapularis tears gives an identical result (1-year Oxford Shoulder Score: MD, −4.2; 95% CI, −0.5 to −8.1; P = .026; 2-year Oxford Shoulder Score: MD, −3.9; 95% CI, 1.1 to −9.0; P = .122). We can therefore confidently refute this criticism as not relevant to outcomes.
Our physiotherapy protocol has been publicly available on our study website since publication (https://warwick.ac.uk/fac/sci/med/research/ctu/trials/startreacts/). This was given to all participants in a consistent way, via physical therapy and patient booklets, as described clearly in both papers. It aligned with recommended company guidance on use of a sling and rehabilitation. This has been communicated in writing to the authors previously. 4 We have not been able to identify a published physical therapy protocol for SPACE.
We have not considered our main findings an underpowered exploratory analysis. This text is clearly related to subgroup analyses only in both papers and is a misrepresentation of our text. That we use such analyses to identify and explore differences between males and females is important in our view. Sex differences were noted in SPACE, but data were not provided. Worse results in females are important and should be clearly reported.
Our primary analyses are definitive. Again, this contrasts with the SPACE study, for which the paper described the primary outcome variable as the change in American Shoulder and Elbow Surgeons (ASES) score, which is not consistent with the registration on ClinicalTrials.gov. Essentially, the SPACE trial failed to show a benefit from the balloon on the primary endpoint. It would have been preferable to present both Constant scores and Oxford Shoulder Scores in START:REACTS, but this was rendered impossible due to the pandemic, and they are known to be well correlated. The change in focus to the change in ASES score is not explained in SPACE.
The authors present multiple incorrect and flawed assertions about our trial, many of which are clearly documented in the papers and prior correspondence. Their assertion that the two populations differ is not supported in data. Age, sex, and baseline strength were very similar between trials. Other data cannot be compared, in part due to limited reporting in SPACE. The authors fail to note clear flaws in their own trial. Constant scores are reported, but there is no evidence that pain-free range of motion was recorded and no published measurement protocol (ours is available on our website). EQ-5D scores are presented in the incorrect format; they should be presented as utilities, typically –0.59 to1, or frequencies and proportions by domain. The reporting of their paper differs substantially from their statistical analysis plan. Blinding was less consistently used compared with our study. But most importantly, even using partial repair as a comparator, a treatment that has had poor results as previously reported in the literature, they failed to show a meaningful benefit.2,8
We agree that not all randomized trials are the same. The authors should be reflective that all trials have flaws, including their own. Our conclusion, based on both trials, remains that the balloon should not be recommended for people with massive or irreparable tears of the rotator cuff.
Footnotes
The authors declared that they have no conflicts of interest in the authorship and publication of this contribution. AOSSM checks author disclosures against the Open Payments Database (OPD). AOSSM has not conducted an independent investigation on the OPD and disclaims any liability or responsibility relating thereto.
