For each RCT, the following data were extracted and tabulated into Microsoft Excel: a) sample size of each treatment group, b) baseline mean and standard deviation for each outcome measure of interest, and c) what the study authors defined as MID for their power calculation where that was performed. Follow up study data were ignored as they were irrelevant for the purposes of our study.
The MIDs of each outcome measure of interest were calculated as follows:

for all outcome measures (single-item and multi-item patient-reported) the MID was assumed to be equal to half the pooled SD at baseline based on the rule of half SD [5 (link)]; where the reliability of the instrument is very high (which is generally the case for single-item outcome measures, such as pain VAS), using the one SEM rule would result in a very small value of MID and therefore the half SD threshold is a more stringent criterion [10 (link)].

for multi-item patient-reported outcome measures (e.g. Victorian Institute of Sport Assessment, VISA), the one SEM rule was also used to calculate the MID to see how that compares to the corresponding value deriving from the rule of half SD [10 (link)]. Cronbach’s alpha was used as an internal consistency measure in the equation.

The SEM is a statistic used to estimate how reliably an instrument measures an individual’s “true score”. The Cronbach’s alpha, which measures the internal consistency of an instrument, is the degree of the inter-relatedness among the items of the instrument [14 (link)]. In contrast to test–retest statistics (e.g. intraclass correlation coefficient, ICC), using Cronbach’s alpha is especially useful when the assessed parameters can change over short periods of time [15 (link)].
For each tendinopathy separately, pooled SDs of VAS at the most commonly reported specific settings (e.g. at rest, with activity etc.) were also computed and presented separately where there was a sufficiently large pooled population size (n > 200 patients). We standardised pain VAS scores by using an 11-point scale (0–10) as this was the most commonly used scale; where a 0–100 scale was used by studies, reported values were converted to their equivalent on a 0–10 scale by dividing them by “10”.
Free full text: Click here