Scand J Med Sci Sports. 2018;28:1467–1468. wileyonlinelibrary.com/journal/sms
© 2017 John Wiley & Sons A/S.
Published by John Wiley & Sons Ltd
LETTER TO THE EDITOR
Changes in muscle size via MRI and ultrasound: Are they
We read with great interest the recent study by Franchi et al
which concluded that changes in muscle thickness measured
via ultrasound tracked well with changes in anatomical cross-
sectional area measured via magnetic resonance imaging
(MRI). While this is a very important study due to the wide-
spread use of ultrasound in training studies, there may be a
slightly more advantageous way to analyze the results. A cor-
relation between the percentage change in muscle thickness
and the percentage change in anatomical cross- sectional area
was computed, yielding a significant correlation (r = .69).
This correlation appeared to be primarily driven by four
limbs (likely from two individuals) detailing a major limi-
tation with correlational analyses as they can be skewed by
outliers and are reliant on sufficient variability in the data.
We used a graph digitizer to estimate the values provided
in the figure (Figure 4A of Franchi et al
) and computed the
exact same correlation (r = .69), which was negated when
excluding the four limbs that responded to a much greater
extent than the rest (r = .34; P = .236).
An alternative analysis exists through equivalency testing
to assess whether the two measurements were not too dif-
ferent from one another.
While a traditional t test examines
whether a 95% confidence interval crosses zero when test-
ing the difference between two groups, an equivalency test
assesses whether the mean difference and 90% confidence
interval lie entirely within the established boundaries (using a
90% confidence interval yields a 0.05 α for significance test-
). These boundaries must be established by the researcher
prior to data collection, but a common method is to use half
of the minimal difference.
Importantly, this method allows
for the comparison of values within the same individual and
is not biased toward outliers, nor is it impacted by the level of
variability in the data, as is the case for correlational analyses.
Using the data set we obtained from Figure 4A of Franchi
we ran an equivalency trial using 2.3% as the upper and
lower boundaries as this corresponds to half of the minimal
difference presented for muscle thickness (4.6%/2 = 2.3%).
The results are detailed in Figure 1 and illustrate that ultra-
sound measured muscle thickness and MRI measured ana-
tomical cross- sectional area cannot be deemed equivalent.
This appears to be primarily driven by the tendency of muscle
growth to be greater when using ultrasound measured muscle
thickness measurements, and this is evident given the mean
difference (−2.1%; 90% confidence interval: −4.0, −0.3).
Had there been no mean difference these measures would
have been considered equivalent as the 90% confidence in-
terval would lie entirely within the established boundaries
(mean: 0; 90% confidence interval: −1.9, 1.8). Therefore,
it appears that ultrasound measured muscle thickness and
MRI measured cross- sectional area track similarly; how-
ever, the ultrasound produces higher magnitudes of growth
when expressed as a percentage. This may be an important
Results of the equivalency
test. The limits of agreement (dotted lines)
are set at one half of the reported minimal
difference for muscle thickness (2.3%).
The equivalency trial using the reported
data was not significant illustrating that
the two measures are not close enough
to be considered equivalent. However,
had the ultrasound and MRI produced
similar magnitudes of muscle growth the
measurements would have yielded fairly
similar results, albeit reporting higher values
for muscle thickness.