Croake et al
priori significance level of α = .05 (
df
= 1,
F
= 5.71,
P
=
.054). This indicated that vocal intensity may play a role in
LEMG signal reliability. No interaction effects between
intensity and time factors were found.
To further evaluate the reliability of LEMG signals
across the 3 testing days, ICCs using a 2-factor mixed ran-
dom effect model were calculated. Average measures from
6 sample trials per participant were compared across the 3
testing sessions to evaluate reliability. The ICCs across
Sessions 1 through 3 were 0.56 for the 65 dB condition and
0.40 for the 75 dB condition. These between-session ICCs
were low, suggesting an interpretation of poor to, at best,
fair reliability across testing sessions. Although not stan-
dardized, typical calculated values for ICC interpretation
are as follows: less than 0.40 = poor reliability, 0.40 to
0.75 = fair to good reliability, and greater than 0.75 = excel-
lent reliability.
19
In addition, the SEM was calculated. The
SEM was then used to determine the MDC in microvolts
necessary to demonstrate a true difference if the muscle was
tested on multiple days with 95% confidence boundaries.
Results indicated that a change of 51µV would be necessary
to determine a true difference in LEMG activity between
testing sessions. Numerical results of ICC, SEM, and MDC
are presented in Table 5. Intraclass correlations for within-
session data revealed strong reliability among participants
ranging from 0.84 to 0.95 and from 0.88 to 0.98 for the 65
dB and 75 dB conditions, respectively. Comparisons of
within-session data are located in Table 6.
Discussion
The use of clinical in-office LEMG has been incorporated
into the diagnostic routine for the evaluation and treatment
of voice disorders in many practices across the country.
8
Although some evidence supports LEMG use in the
diagnosis and prognosis of certain neuromuscular disorders,
the general reliability of the LEMG signal in normal partici-
pants has not been carefully evaluated. This study measured
the reliability of the LEMG signal in normal, vocally healthy
participants over time with the central aim of determining if
significant LEMG signal variance occurred as a function of
multiple testing sessions. Our second aim was to determine
if vocalization intensity affected the LEMG signal. In limb
studies, control over the degree of muscle contraction is nec-
essary to achieve results that are comparable within and
across participants. Both maximal and submaximal contrac-
tions have been shown to demonstrate strong reliability in
limb muscle.
20
We used vocal intensity as a method to con-
trol laryngeal muscle contraction levels among participants.
Our results indicated that between-session LEMG reliability
was poor to fair and that control of vocal intensity may be an
important performance variable to help improve the reliabil-
ity of these measurements.
This study mirrored the University of Iowa head and
neck protocol for LEMG diagnostics.
21
In addition to this
basic protocol, we used vocal intensity control and a
Faraday booth to reduce ambient electrical noise to improve
the fidelity of the data and provide the optimal set of cir-
cumstances under which to perform our LEMG clinical
evaluation. The intent of this study was not to quantify
LEMG precisely but rather to use quantitative means to
measure LEMG in an ideal environment to test the hypoth-
esis that clinical LEMG data are variable across testing ses-
sions even with added control parameters in place.
Repeated measures analysis of variance indicated a non-
significant effect for time of testing, suggesting that LEMG
signals for pooled data did not vary significantly across test-
ing sessions. Intraclass correlation coefficient analysis for
within-session reliability was considered excellent for both
intensity conditions ranging from 0.84 to 0.95 and from
0.88 to 0.89 for the 65 dB and 75 dB conditions, respec-
tively. However, the between-session ICC revealed poor to
fair reliability for both intensity conditions. It should be
noted that the most qualitatively consistent data from our
study were collected when the participants vocalized at 65
dB, indicating a less reliable measure at greater loudness
levels. Data from Sessions 1 and 2 at 65 dB represented the
strongest reliability association with an
R
2
value of 0.048.
This indicated poor reliability even across the most consis-
tent recording sessions (see Figures 1 and 2).
Reliability debates concerning the clinical usefulness of
LEMG for diagnostic and prognostic applications have
been raised.
1
According to a recent evidence-based review
and clinical recommendations, LEMG data have been con-
sidered questionable for clinical uses such as diagnosing
paresis/paralysis from joint fixation, for accuracy diagnos-
ing diseases of the neuromuscular junction, and for provid-
ing accurate diagnostic information of neuropathic and
myopathic disorders.
1,8
To address these questions, it has
Table 5.
Intraclass Correlation Coefficients (ICCs) Between
Sessions, Standard Error of Measurement (SEM), and Minimum
Detectable Change (MDC) in µV Necessary to Detect True
Change Between Measurements.
ICC Sessions
1–3
SEM
Sessions 1–3
MDC in µV @
70% Confidence
MDC in µV @
95% Confidence
65 dB 0.56
18.57
26.26
51.47
75 dB 0.40
23.0
32.5
63.7
Table 6.
Intraclass Correlation Coefficients (ICCs) for Within-
Session Data.
Within-Session
ICC
Session 1
Session 2
Session 3
65 dB
0.95
0.93
0.84
75 dB
0.88
0.84
0.98
59