J Appl Physiol 98: 274 –281, 2005.
First published September 10, 2004; doi:10.1152/japplphysiol.01380.2003.
Diaphragm electromyogram root mean square response to hypercapnia and its
intersubject and day-to-day variation
Bhajan Singh,1–3 Janine A. Panizza,1,2 and Kevin E. Finucane1,2
1
Department of Pulmonary Physiology, Sir Charles Gairdner Hospital, 2West Australian Sleep Disorders Research Institute,
QEII Medical Centre, and 3Department of Physiology, University of Western Australia, Nedlands, Western Australia, Australia
Submitted 22 December 2003; accepted in final form 31 August 2004
IN HUMANS, THE VENTILATORY control system is conventionally
studied by relating a mechanical response of the respiratory
system such as minute ventilation (V̇E), mean inspiratory flow
rate (VT/TI), or airway pressure during the first 100 ms after
airway occlusion (36) to variation in a single stimulus such as
the arterial PCO2. However, these responses may underestimate
neural drive when respiratory muscle function is impaired by
muscle weakness or changes in muscle length because of
hyperinflation and, in the case of V̇E and VT/TI, when respiratory system resistance or elastance is increased. Diaphragm
electromyogram (EMGdi) is a more direct measure of respiratory muscle activation. In animals, direct relationships have
been found between rectified integrated EMGdi and end-tidal
PCO2 (PETCO2) (1), the phrenic neurogram (22), and diaphragm
O2 consumption (28, 34). In humans, the diaphragm accounts
for a substantial fraction of inspiratory work (23, 24), and the
average rate of rise of inspiratory EMGdi moving time average
increases linearly with PETCO2, transdiaphragmatic pressure
(Pdi), VT/TI, and occlusion pressure (25). However, EMGdi is
rarely used to assess ventilatory control in clinical practice
because of difficulties in detection and quantification. Beck and
Sinderby and colleagues (4, 5, 31, 33) have recently developed
methods to improve quantification of EMGdi in humans by
detecting activity using a multielectrode esophageal catheter,
evaluating signal quality using spectral analysis, and measuring its root mean square to quantify signal strength. The root
mean square reflects the number and firing rate of motor units
recruited (3) and is regarded as a more valid measure of EMG
power than full-wave rectification and integration (2). Beck et
al. (3) found that diaphragm root mean square (RMSdi) was a
reliable index of global diaphragm activation up to but not
exceeding ⬃75% of maximal activation.
The use of RMSdi as a measure of diaphragm activation has
a number of limitations. First, because RMSdi varies with the
square root of activation (3), the relationship between RMSdi
and a respiratory stimulus such as PCO2 may not be linear.
Second, absolute values of RMSdi vary widely between subjects (3). To allow comparisons between subjects, RMSdi has
been normalized to values obtained at rest (RMSdirest) (3) or
voluntary maximal inspiration (RMSdimax) (30). However,
these approaches may be suboptimal because RMSdirest is
likely to have a low signal-to-noise ratio, RMSdimax underestimates activation (3), and both are dependent on effort. Finally, the variation in RMSdi in a subject from one occasion of
measurement to another has not been defined.
There are several potential sources of intersubject variation
in RMSdi. First, anatomical differences in the relationship of
the crural diaphragm to the gastroesophageal junction could
influence the position or orientation of the electrode relative to
diaphragm muscle. Second, RMSdi may vary between subjects
because of differences in diaphragm anatomy such as muscle
fiber diameter, which affects the propagation velocity and thus
shape and size of the muscle fiber action potential (13), and the
number of fibers in a motor unit (20). These differences may be
reflected in diaphragm thickness and, although there are no
established methods to measure the thickness of the crural
diaphragm in situ, thickness of the costal diaphragm can be
measured by ultrasound (8). Third, there may be differences
Address for reprint requests and other correspondence: B. Singh, Dept. of
Pulmonary Physiology, Sir Charles Gairdner Hospital, Hospital Ave., Nedlands, WA 6009, Australia (E-mail: [email protected]).
The costs of publication of this article were defrayed in part by the payment
of page charges. The article must therefore be hereby marked “advertisement”
in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
diaphragm activation; esophageal catheter; normalization; electrocardiogram
274
8750-7587/05 $8.00 Copyright © 2005 the American Physiological Society
http://www. jap.org
Downloaded from http://jap.physiology.org/ by 10.220.33.5 on June 15, 2017
Singh, Bhajan, Janine A. Panizza, and Kevin E. Finucane.
Diaphragm electromyogram root mean square response to hypercapnia and its intersubject and day-to-day variation. J Appl Physiol 98:
274 –281, 2005. First published September 10, 2004; doi:10.1152/
japplphysiol.01380.2003.—Diaphragm activation can be quantified
by measuring the root mean square of crural EMG (RMSdi) (Beck J,
Sinderby C, Lindstrom L, and Grassino A, J Appl Physiol 85:
1123–1134, 1998). To examine intersubject and day-to-day variation
in the RMSdi-PCO2 relationship, end-tidal PCO2, minute ventilation
(V̇E), respiratory frequency (fB), and RMSdi were measured in seven
healthy subjects on two occasions during steady-state ventilation at
seven levels of inspired O2 fraction (FICO2) from 0 to 0.08 in random
order. RMSdi was measured with a multielectrode esophageal catheter
and controlled for signal contamination and diaphragm position.
RMSdi was normalized for values obtained during quiet breathing at
functional residual capacity, at FICO2 of 0.04, and during an inspiratory capacity maneuver (RMSdi%max) as well as ECG R-wave
amplitude at functional residual capacity (RMSdi/ECGR), fB, and
thickness of the costal diaphragm measured by ultrasound. RMSdi
increased linearly with PCO2 (mean r2 ⫽ 0.83 ⫾ 0.10); at the highest
FICO2, RMSdi%max was 40.2 ⫾ 11.6%. Relative to the intersubject
variation in the V̇E-PCO2 relationship, intersubject variations in the
slopes and intercepts of the RMSdi-PCO2 relationships were 1.7 and
1.8 times, respectively, and RMSdi%max-PCO2 relationships 0.9 and
1.3 times, respectively, and were unrelated to fB and diaphragm
thickness. Relative to the day-to-day variation in the V̇E-PCO2 relationship, day-to-day variation in the slopes and intercepts of the
RMSdi-PCO2 relationships were 2.8 and 4.4 times, respectively, and
RMSdi/ECGR-PCO2 relationships 1.3 and 2.2 times, respectively. It
was concluded that the RMSdi-PCO2 relationship measures chemosensitivity and is best compared between subjects via RMSdi%max
and on separate occasions in the same subject via RMSdi/ECGR.
DIAPHRAGM EMG ROOT MEAN SQUARE RESPONSE TO CO2
METHODS
Subjects. Seven healthy subjects, five men and two women, age
(mean ⫾ SD) 40.7 ⫾ 11.1 yr, body mass index 25.2 ⫾ 2.1 kg/m2, and
forced expiratory volume in 1 s of 3.59 ⫾ 0.70 liters, participated in
the study. Informed consent was obtained from each subject, and
ethical approval was granted by the Committee for Human Rights,
University of Western Australia.
Measurements. In each subject, the response to hypercapnia was
examined at seven fractional concentrations of inspired CO2 (FICO2),
i.e., 0, 0.02, 0.04, 0.05, 0.06, 0.07, and 0.08, on two occasions at an
interval of between 6 and 98 days (mean 34 ⫾ 34 days). The fractional
concentration of CO2 in expired gas was measured with a rapid gas
analyzer (Morgan Benchmark, model 503, P. R. Morgan, Kent, UK)
and used to calculate the PETCO2. A pneumotachograph was used to
measure inspiratory flow and was integrated over time to obtain
inspired volume.
EMGdi signals were recorded and processed by using methods
described by Beck and Sinderby and colleagues (4, 5, 31, 33). A
J Appl Physiol • VOL
purpose-built Silastic esophageal catheter, 2.5 mm in diameter with an
array of eight electrode rings at 1-cm intervals at its distal end
(Dentsleeve, Wayville, South Australia, Australia), was used. Each
electrode ring was 2 mm wide and consisted of fine stainless steel wire
wrapped around the catheter. Six bipolar electrode pairs were created
by wiring the electrodes in an overlapping array, i.e., 1 vs. 3, 2 vs. 4,
3 vs. 5, 4 vs. 6, 5 vs. 7 and 6 vs. 8 (Fig. 1). EMG signals were collected
at a rate of 2 kHz, amplified (Grass Wideband A.C. amplifier, model
17P3C, Quincy, MA), band-pass filtered between 10 and 1,000 Hz,
digitized, and, via purpose-written software (Labview Version 6),
displayed online on a personal computer and stored for subsequent
analysis. The EMG signals were later reviewed and processed in the
time and frequency domains as described by Sinderby et al. (31, 33)
with purpose-written software in Labview (version 6) to calculate and
display RMSdi, CFdi, and signal-contamination criteria of selected
EMG segments from each electrode pair. The catheter was positioned
across the diaphragm so that the electrical center of the diaphragm
was in the center of the electrode span. The electrical center of the
diaphragm was identified by a change in polarity of the EMG signal
on either side of the center. For each breath, a single RMSdi value was
obtained from a segment of diaphragm EMG activity during inspiration that was free of cardiac activity, i.e., between the T and P waves
of successive ECG complexes. The RMSdi was measured by subtracting the signal below from the signal above the electrical center, to
improve the signal-to-noise ratio (33). The duration of each such
segment was between 300 and 400 ms, and RMSdi represents the
mean deviation of the signal from zero over the selection. The
electrical center of the diaphragm remained within the span of the
electrode array in all tests. RMSdi and CFdi were only accepted if the
frequency spectra fulfilled criteria for an uncontaminated signal recommended by Sinderby et al. (31).
Protocol. Subjects were seated and inspired from a 50-liter bag
containing FICO2 of between 0 and 0.08 in random order, inspired O2
fraction 0.21, and the remainder nitrogen (Fig. 1). Subjects were
blinded to the level of FICO2. To equilibrate with the gas mixture, the
subject took two vital capacity inspirations and 20 tidal breaths from
the bag before collection of 30 breaths for analysis. For each condition
of measurement, the RMSdi value for the last 10 breaths was averaged, because the data showed that the coefficient of variation (CV)
increased when fewer than 10 breaths were analyzed (Fig. 2). There
was no change in PETCO2 during the last 10 breaths of each collection.
EMG data could not be retrieved for one condition of measurement of
one study in two subjects, i.e., FICO2 0 in the first study in subject 5
and FICO2 0.08 in the second study in subject 4. In all subjects, the
thickness of the costal diaphragm in the midaxillary line was measured in a seated posture during quiet breathing by an experienced
radiologist using a two-dimensional ultrasound (Toshiba PowerVision
6000, model SSA-370A) and methods described by Cohn et al. (8).
Normalization. RMSdirest and RMSdisubmax were the RMSdi values at FICO2 0 and 0.04, respectively. RMSdimax was measured at the
beginning and end of each study as the mean of three slow inspirations
from functional residual capacity to total lung capacity. There was no
difference between RMSdimax measured at the beginning and end of
each study (0.347 ⫾ 0.111 vs. 0.340 ⫾ 0.093 arbitrary units), and
RMSdimax at the beginning of each study was used in normalizations.
For all studies, the mean RMSdimax was 90.9 ⫾ 4.5% of the highest
RMSdimax. The ECGR was the mean amplitude of five R waves of the
ECG at end expiration at FICO2 ⫽ 0 detected by the electrode pair at
the center of electrical activity of the diaphragm. The thickness of the
costal diaphragm was the mean of five measurements at the midaxillary line at functional residual capacity.
Data analysis and statistics. Maximum voluntary ventilation was
estimated as 35 times forced expiratory volume in 1 s. The relationships between responses (RMSdi, V̇E, and VT/TI) and PETCO2 were
examined via linear and polynomial regressions. To determine the
factors associated with intersubject variability in RMSdi, at each level
of F I CO 2 , the relationships between RMSdi and RMSdi rest ,
98 • JANUARY 2005 •
www.jap.org
Downloaded from http://jap.physiology.org/ by 10.220.33.5 on June 15, 2017
between subjects in the filtering properties of the electrode or
surrounding tissue. These factors are likely to also affect the
EMG frequency spectrum. Finally, studies of diaphragm activation in chronic airflow limitation and after polio by Sinderby
et al. (30, 32) compared RMSdi values per breath without
considering differences in pattern of breathing between subjects. As the ventilatory response to a stimulus consists of an
increase in both tidal volume and respiratory frequency (fB), it
may be more appropriate to express RMSdi as a function of
time rather than per breath (21).
Day-to-day variation in RMSdi may also result from alterations in the position or orientation of the electrode relative to
muscle and/or filtering properties of the electrode or surrounding tissues. An equivalent variation may be seen in the amplitude of other electrical signals detected by the esophageal
electrode. The electrocardiogram (ECG) unavoidably contaminates esophageal recordings of diaphragm EMG and could
provide an effort-independent electrical signal to control for
these day-to-day variations. Because the pericardium is adherent to the left hemidiaphragm, at a constant lung volume, the
position and electrical axis of the heart may have a relatively
constant relationship to the gastroesophageal junction and
crural diaphragm.
The aims of this study were to define 1) the response of
RMSdi to hypercapnia in healthy subjects, 2) the intersubject
and day-to-day variations in the relationships between RMSdi
and PCO2, and 3) the effect of normalization for RMSdirest, root
mean square at a submaximal effort (RMSdisubmax), RMSdimax,
diaphragm EMG center frequency (CFdi), diaphragm thickness, fB, and ECG R-wave amplitude (ECGR) on intersubject
and day-to-day variations in the relationships between RMSdi
and PCO2. It was hypothesized that the relationship between
RMSdi and PCO2 was near linear at moderate diaphragm
activation, intersubject and day-to-day variations in the relationship between RMSdi and PCO2 were substantial, and these
variations could be minimized by normalization. It was found
that the relationship between RMSdi and PCO2 was linear at
moderate diaphragm activation, intersubject and day-to-day
variations in the relationship between RMSdi and PCO2 were
higher than that of the relationship between V̇E and PCO2,
intersubject variation in RMSdi was reduced best by normalization for RMSdimax, and day-by-day variation in RMSdi was
reduced best by normalization for RMSdirest or ECGR.
275
276
DIAPHRAGM EMG ROOT MEAN SQUARE RESPONSE TO CO2
RMSdisubmax, RMSdimax, ECGR, CFdi,, fB, and diaphragm thickness
were examined by linear regressions. Intersubject variability in the
relationship between RMSdi and PETCO2 and the effect of normalization with RMSdirest (RMSdi%rest), RMSdisubmax (RMSdi%submax),
RMSdimax (RMSdi%max), ECGR (RMSdi/ECGR), fB (RMSdi 䡠 fB),
and diaphragm thickness (RMSdi/Tdi) were examined by using the
CV of the slopes and intercepts of the linear regressions of the first
study in each subject. Day-to-day variability in the relationship
between RMSdi and PETCO2 and the effect of various normalizations,
as above, were examined by using the CV of the difference in slopes
and intercepts of the linear regressions between the two studies in
each subject. The variation in CFdi between subjects and at different
Fig. 2. Breath-to-breath variation in diaphragm EMG RMS (RMSdi). Upper
limit of 95% confidence interval of coefficients of variation of RMSdi at FICO2
0.08 in both studies is shown in all subjects (n ⫽ 13), according to number of
breaths analyzed, starting with the last 3 breaths.
J Appl Physiol • VOL
levels of FICO2 was analyzed by using a two-way ANOVA. The
difference between RMSdimax at the beginning and end of each test
was compared via a paired t-test. Mean slopes and intercepts are
expressed as means ⫾ SE, and all other data are expressed as means ⫾
SD. Significance was defined as P ⬍ 0.05.
RESULTS
Response to hypercapnia. At FICO2 0.08, the subjects
achieved 33.9 ⫾ 9.0% of predicted maximum voluntary ventilation (range 20.0 – 47.8%), a mean inspiratory flow rate of
1.57 ⫾ 0.42 l/s (range 0.86 –2.37 l/s), and RMSdi 40.2 ⫾
11.6% of RMSdimax (range 19.2– 62.3%). CFdi varied between
subjects (P ⬍ 0.001) but was not influenced by FICO2 (P ⫽
0.126). In all tests, RMSdi, V̇E, and VT/TI were linearly related
to PETCO2 [mean coefficient of determination (r2) ⫽ 0.83 ⫾
0.10, 0.96 ⫾ 0.02, and 0.96 ⫾ 0.02, respectively] (Table 1). A
second-order polynomial did not significantly increase the r2
value of the RMSdi-PETCO2 relationship. In most subjects, the
r2 value of the linear regression of RMSdi against PETCO2
relationship increased when fB was considered (RMSdi䡠fBPETCO2, mean r2 ⫽ 0.88 ⫾ 0.07). The relationship between
mean RMSdi%max and mean PETCO2 at each FICO2 for all
subjects in the first study is shown in Fig. 3.
Intersubject variation. Intersubject differences in RMSdi at
equivalent FICO2 were related to RMSdirest, RMSdisubmax, and
RMSdimax (Table 2) but not to CFdi, ECGR, fB, or diaphragm
thickness. Diaphragm thickness at functional residual capacity
ranged from 15 to 43 mm and was closely related to body mass
index (r2 ⫽ 0.94) and height (r2 ⫽ 0.73). Intersubject differences in RMSdi%max at equivalent FICO2 were not related
to fB.
98 • JANUARY 2005 •
www.jap.org
Downloaded from http://jap.physiology.org/ by 10.220.33.5 on June 15, 2017
Fig. 1. Measurement of the diaphragm
EMG response to CO2. Subjects inhaled
from a 50-liter bag containing a predetermined inspired CO2 fraction (FICO2) ranging
from 0 to 0.08, inspired O2 fraction (FIO2)
0.21, and the remainder nitrogen. Expired
CO2 (FECO2) was measured with a rapid gas
analyzer and flow with a pneumotachograph.
Diaphragm EMG was detected using a multielectrode esophageal catheter, then amplified, band-pass filtered, and displayed on a
personal computer. CF, center frequency;
RMS, root mean square. See text for details.
277
DIAPHRAGM EMG ROOT MEAN SQUARE RESPONSE TO CO2
Table 1. Linear regressions of RMSdi against PETCO2 (n ⫽ 7) at each study
First Study
Subject
1
2
3
4
5
6
7
Mean
SE
SD
CV, %
Slope, AU/Torr ⫻
2.62
6.39
6.10
8.30
2.73
2.15
2.98
4.47
0.91
10⫺3
Second Study
Intercept, AU ⫻
10⫺1
⫺0.71
⫺2.43
⫺2.06
⫺2.94
⫺0.65
⫺0.44
⫺1.04
⫺1.47
0.38
Slope AU/Torr ⫻
r2
0.87
0.88
0.91
0.98
0.63
0.76
0.75
0.83
10⫺3
Intercept AU/Torr ⫻ 10⫺2
r2
⫺0.73
⫺2.33
⫺1.70
⫺2.33
⫺0.59
⫺0.86
⫺3.14
⫺1.67
0.37
0.78
0.79
0.82
0.99
0.79
0.89
0.84
0.84
2.40
6.42
4.95
7.34
3.43
2.58
7.91
5.00
0.86
0.12
20.5
0.08
25.6
17.2
22.1
RMSdi, diaphragm electromyogram root mean square; PETCO2, end-tidal partial pressure of CO2; AU, arbitrary units; CV, coefficient of variation.
Table 3 and Fig. 4 show the mean day-to-day variation in the
slopes and intercepts of the relationships between measured
responses and PETCO2. The day-to-day variation was least for
the VT/TI-PETCO2 relationship (Table 3 and Fig. 4). Relative to
the V̇E-PETCO2 relationship, the day-to-day variation in the
slope and intercept of the RMSdi-PETCO2 relationship was 2.8
and 4.4 times greater, respectively (Table 3). Normalization for
RMSdirest and ECGR and, to a lesser degree, RMSdimax reduced the day-to-day variation in the slopes and intercepts of
the RMSdi-PETCO2 relationships (Table 3 and Fig. 4). Day-today variations in slopes were similar for the RMSdi/ECGRPETCO2 and V̇E-PETCO2 relationships (Table 3 and Fig. 4).
DISCUSSION
This study found that diaphragm activation, quantified by
measuring RMSdi, increased linearly in response to hypercapnia up to moderate levels of diaphragm activation, implying
that the RMSdi response to hypercapnia measures chemosensitivity. At the highest FICO2 used, RMSdi was ⬍65% of
RMSdimax in all subjects. The relationships between V̇E and
PETCO2 indicated substantial intersubject and day-to-day variation in chemosensitivity. Intersubject variation in the RMSdiPETCO2 relationship was almost twice the variation in the
V̇E-PETCO2 relationship in the same subjects and can be attributed, first, to individual differences in chemosensitivity and,
second, to differences in anatomy and muscle-electrode relationship because the effect of the second source of intersubject
variation was virtually removed by normalizing RMSdi for
Table 2. Coefficients of determination (r2) for linear
regressions of RMSdi against RMSdirest, RMSdisubmax,
and RMSdimax at each FICO2 in the first study.
Fig. 3. Relationship between mean ⫾ SD RMSdi values normalized with
RMSdi values during an inspiratory capacity maneuver (RMSdi%max) and
mean ⫾ SD end-tidal PCO2 (PETCO2) (n ⫽ 7) at each FICO2 in the first study.
J Appl Physiol • VOL
FICO2
Mean PETCO2, Torr
n
RMSdirest
0
0.02
0.04
0.05
0.06
0.07
0.08
43.1⫾3.4
44.9⫾2.3
48.0⫾1.9
50.8⫾1.8
53.3⫾2.0
57.1⫾1.7
61.9⫾1.6
6
7
7
7
7
7
7
0.57
0.69*
0.58
0.59
0.82*
0.87†
RMSdisubmax
RMSdimax
0.69*
0.69*
0.60
0.63*
0.33
0.65*
0.66*
0.83†
0.86†
0.82*
0.79*
0.63*
0.60*
FICO2, fraction of CO2 in inspired gas; RMSdirest, RMSdi at FICO2 0;
RMSdisubmax, RMSdi at FICO2 0.04; RMSdimax, RMSdi during slow inspiratory
capacity maneuver. Significant relationships: *P ⬍ 0.05 and †P ⬍ 0.01.
98 • JANUARY 2005 •
www.jap.org
Downloaded from http://jap.physiology.org/ by 10.220.33.5 on June 15, 2017
Table 3 and Fig. 4 show the means and variations of the
slopes and intercepts for the relationships between measured
responses and PETCO2 at the first study. Variation in slopes and
intercepts between subjects was least for the VT/TI-PETCO2
relationship (Table 3 and Fig. 4). The slope of the V̇E-PETCO2
relationship varied from 0.57 to 2.65 l䡠min⫺1 䡠Torr⫺1 (mean
1.68 ⫾ 0.21 l䡠min⫺1 䡠Torr⫺1). Relative to the V̇E-PETCO2 relationship, intersubject variation in slopes and intercepts for the
RMSdi-PETCO2 relationship was 1.7 and 1.8 times greater,
respectively (Table 3). Normalization for RMSdimax and, to a
lesser degree, RMSdirest, RMSdisubmax, and ECGR reduced the
intersubject variation in the slopes and intercepts of the
RMSdi-PETCO2 relationships (Table 3 and Fig. 4). Intersubject
variation in slopes of the RMSdi%max-PETCO2 and V̇E-PETCO2
relationships were similar (Table 3 and Fig. 4).
Day-to-day variation. At the second study, intersubject differences in RMSdi at equivalent FICO2 were related to
RMSdirest and RMSdisubmax, but not to RMSdimax. Relative to
the first study, the slopes and intercepts of the VT/TI-PETCO2
relationship decreased, the intercepts of the V̇E-PETCO2 relationships decreased, and the slopes and intercepts of the
RMSdi-PETCO2 relationship were similar (Table 1). Relative to
the first study, intersubject variations in the slopes and intercepts of the RMSdi-PETCO2 relationship were reduced in the
second study (Table 1) and were not further reduced by
normalization.
278
DIAPHRAGM EMG ROOT MEAN SQUARE RESPONSE TO CO2
Table 3. Intersubject and day-to-day variations in the linear regressions of measured responses against PETCO2
First Study
Response
n
V̇E
7
VT/TI
7
RMSdi
7
RMSdi%rest
6
RMSdi%submax
7
RMSdi%max
7
RMSdi/ECGR
7
Slope, l䡠min 䡠Torr
Intercept, l/min
Slope, l䡠s⫺1䡠Torr⫺1
Intercept, l/s
Slope, AU/mmHg
Intercept, AU
Slope, %/Torr
Intercept, %
Slope, %/Torr
Intercept, %
Slope, %/Torr
Intercept, %
Slope, AU
Intercept, AU
⫺1
⫺1
Difference Between First and Second Study
Mean⫾SE
CV, %
Mean⫾SE
CV, %
1.68⫾0.21
⫺63.7⫾9.1
5.79⫾0.56⫻10⫺2
⫺2.12⫾0.23
4.47⫾0.91⫻10⫺3
⫺0.15⫾0.04
9.82⫾1.63
⫺327.3⫾74.1
6.51⫾0.93
211.0⫾42.0
1.21⫾0.14
⫺38.9⫾7.2
1.55⫾0.24
⫺50.3⫾10.4
12.4
14.3
9.7
10.8
20.5
25.6
16.6
22.6
14.3
19.9
11.6
18.6
15.3
20.7
⫺0.22⫾0.11
9.8⫾3.7*
⫺1.27⫾0.38⫻10⫺2*
0.61⫾0.12†
⫺0.54⫾0.78⫻10⫺3
0.02⫾0.03
⫺3.31⫾2.30
136.6⫾89.3
⫺0.58⫾1.56
32.3⫾66.5
⫺0.17⫾0.17
6.82⫾7.63
⫺0.53⫾0.37
16.1⫾13.6
51.7
37.9
29.8
20.3
144.4
168.1
69.5
65.4
271.7
206.0
102.3
111.9
68.9
84.8
RMSdimax. Day-to-day variation in the slope of the RMSdiPETCO2 relationship was ⬃2.8 times the day-to-day variation in
the V̇E-PETCO2 relationship in the same subjects. The day-today variation in the RMSdi-PETCO2 relationship in a subject
was also attributable in part to day-to-day variation in chemosensitivity; the residual variation in the RMSdi-PETCO2 could be
minimized by normalizing RMSdi for RMSdirest or ECGR,
implying that the variation was mainly due to differences in the
muscle-electrode relationship. Before discussing these findings
and conclusions, we consider the extent to which the methods
used allow accurate estimates of diaphragm activation in response to hypercapnia and influence interpretation of the results.
Limitations. The conclusions of this study are based on
estimates of diaphragm EMG signal strength measured using
methods proposed by Beck and Sinderby and colleagues (4, 5,
31, 33). To maintain constant muscle-to-electrode distance,
EMG activity was continuously recorded from six bipolar
electrode pairs spanning 7 cm (Fig. 1). To improve signal-tonoise ratios, the signals above and below the electrical center
of the diaphragm were subtracted; this required the electrical
center of the diaphragm to remain within the middle 5 cm of
the array. Although diaphragm excursion during the inspiratory
capacity maneuver used to measure RMSdimax is likely to have
exceeded 5 cm (7, 35), the electrical center remained within the
span of electrode pairs at all times, supporting the notion that
the electrodes move with the diaphragm. The progressive
increase in RMSdi with hypercapnia supports the findings of
Beck et al. (5) that this method enables the electrode to remain
in a relatively constant spatial relationship with the crural
diaphragm during respiration.
Diaphragm EMG was only sampled between the T and P
waves of successive ECG complexes to avoid ECG artifact. In
most studies, this consisted of between 0.3 and 0.4 s of EMG
activity. Inability to sample from an identical part of each
inspiration led to some variability in RMSdi from breath to
breath; this variability was reduced by using the mean RMSdi
over 10 breaths during steady-state ventilation at each condition of measurement. The data in this study showed that, by
J Appl Physiol • VOL
using 10 breaths, there was a 95% likelihood that the CV in
each subject was ⬍10% at FICO2 0.08 (Fig. 2). RMSdi was not
corrected for alterations in action potential propagation velocity because CFdi did not change at different levels of FICO2.
Esophageal electrodes detect EMG activity arising from the
crural diaphragm. The costal and crural parts of the diaphragm
differ in embryological origin (16), innervation (17), proportions of muscle-fiber types (27), and actions on the lower rib
cage (11). There is conflicting evidence on the homogeneity of
activation and shortening of the costal and crural diaphragms.
Several studies report that the costal and crural diaphragms
contract simultaneously (6, 29) and that their electrical activity
increases similarly when an inspiratory load is added (22). In
intact awake lambs, Cooke et al. (9) found identical EMG
responses of the costal and crural diaphragms to various
hypoxic and hypercapnic gas mixtures, indicating that these
components of the diaphragm comprise a single functional
unit. Other studies show regional differences in activation of
the recruitment of the costal and crural diaphragms (12).
Studies in dogs show that regional EMG activity within intercostal muscles is directly related to regional mechanical advantage (10, 18, 19). In the diaphragm of dogs, regional
mechanical advantage and blood flow are heterogeneous (37,
15) but matched and similar between adjacent costal and crural
regions (15). The distribution of regional blood flow does not
change with exercise (15). Assuming that the distribution of
activation within the diaphragm is proportional to the distribution of mechanical advantage, these findings suggest a direct
relationship between crural and costal EMG that does not
change with exercise. Consistent with this, in humans, Beck et
al. (3) found that activation of the crural diaphragm, assessed
by the EMG technique we studied, was linearly related to a
measure of global activation of the diaphragm (ratio of Pdi to
maximal Pdi) independent of diaphragm length up to 75% of
maximal Pdi. Thus available evidence suggests that activation
of the crural diaphragm is representative of global activation of
the diaphragm.
The mean slopes and intercepts of the relationship between
V̇E and hypercapnia in this study are lower than those reported
98 • JANUARY 2005 •
www.jap.org
Downloaded from http://jap.physiology.org/ by 10.220.33.5 on June 15, 2017
V̇E, minute ventilation; VT/TI, mean inspiratory flow rate; RMSdi%rest, RMSdi expressed as a percentage of RMSdi value at FICO2 0; RMSdi%submax, RMSdi
expressed as a percentage of RMSdi at FICO2 0.04; RMSdi%max, RMSdi expressed as a percentage of RMSdi value during inspiratory capacity maneuver;
RMSdi/ECGR, RMSdi divided by amplitude of the ECG R wave at end-expiration at FICO2 0. Significant difference between first and second study: *P ⬍ 0.05
and †P ⬍ 0.01.
DIAPHRAGM EMG ROOT MEAN SQUARE RESPONSE TO CO2
279
by Read and Leigh (26) and Irsigler (14) and can be attributed
primarily to differences in chemosensitivity. In 126 healthy
subjects, Irsigler (14) found that the mean slope of the relationship between V̇E and PCO2 was 2.60 ⫾ 0.11
l䡠min⫺1 䡠Torr⫺1, but there was considerable variation so that
the range in slopes was 1.5 to 5.0 l䡠min⫺1 䡠Torr⫺1 in 80% of
the subjects. Most of the remaining subjects in that study had
slopes ⬍1.5 l䡠min⫺1 䡠Torr⫺1, and women were less responsive
than men (14). In the present study, one female subject (subject
6) had a slope below 1.5 l䡠min⫺1 䡠Torr⫺1. The possibility that
the difference in methods, i.e., steady-state ventilation at several levels of PCO2 in this study compared with progressive
hypercapnia induced by rebreathing used by Read and Leigh
(26) and Irsigler (14), also contributed to the observed differences cannot be excluded.
At the second study, there was less intersubject variation in
the slopes and intercepts of the RMSdi-PETCO2 relationship,
J Appl Physiol • VOL
there was no relationship between RMSdi and RMSdimax, and
intersubject variation in the RMSdi-PETCO2 relationship was
not reduced by normalization. These findings may represent a
type II statistical error due to the relatively small sample size.
The levels of diaphragm activation achieved in this study
were moderate, so that the results do not establish the limits of
linearity of the relationship between RMSdi and PETCO2 during
hypercapnia. The range of FICO2 used was adequate to establish
the ventilatory and diaphragmatic EMG responses to hypercapnia in healthy subjects; in practice, higher levels of response are unnecessary and unlikely to be tolerated in healthy
subjects because of the large increase in ventilation and the
dyspnea experienced. In chronic airflow limitation, the level of
diaphragm activation during quiet breathing (30) approximates
that found in this study at the highest PCO2 breathed. In such
subjects, the RMSdi responses to progressive hypercapnia
could approach the value during maximal inspiration and the
98 • JANUARY 2005 •
www.jap.org
Downloaded from http://jap.physiology.org/ by 10.220.33.5 on June 15, 2017
Fig. 4. Intersubject (A) and day-to-day (B) variations in linear
regression slopes (F) and intercepts (E) of responses to hypercapnia (n ⫽ 7) expressed as a ratio of variations in linear
regression slopes and intercepts of the RMSdi-PETCO2 relationship. V̇E, minute ventilation; VT/TI, mean inspiratory flow rate;
RMSdi%rest, RMSdi%submax, RMSdi%max, and RMSdi/
ECGR, effect of normalization with RMSdi values obtained at
rest, at FICO2 0.04, during inspiratory capacity maneuver, and
by ECG R-wave amplitude at functional residual capacity,
respectively.
280
DIAPHRAGM EMG ROOT MEAN SQUARE RESPONSE TO CO2
J Appl Physiol • VOL
measured from the electrode closest to the electrical center of
the diaphragm at end expiration at rest can be used to reduce
day-to-day variation in RMSdi and has the advantage over
RMSdirest or RMSdimax in being independent of effort.
These findings suggest that RMSdi measured via a multielectrode esophageal catheter and methods described by Beck
and Sinderby and colleagues (4, 5, 31, 33) can be used to
quantify moderate levels of diaphragm activation and the
neural drive to the diaphragm during progressive hypercapnia.
It is recommended that RMSdi is normalized for RMSdimax for
comparisons between subjects and normalized for the amplitude of the ECG R wave for comparisons of repeat measurements in the same subject.
ACKNOWLEDGMENTS
The authors thank W. J. Noffsinger, M. Lam, and D. Austin for technical
assistance and S. Dhaliwal for statistical advice.
GRANTS
This research was supported by a grant from the Medical Research Fund of
Western Australia and Sir Charles Gairdner Hospital Research Fund. B. Singh
is the recipient of an Australian Lung Foundation/Boehringer Ingelheim
Chronic Airflow Limitation Research Fellowship.
REFERENCES
1. Altose MD, Stanley NN, Cherniak NS, and Fishman AP. Effects of
mechanical loading and hypercapnia on inspiratory muscle EMG. J Appl
Physiol 38: 467– 473, 1975.
2. Basmajian JV and De Luca CJ. Muscles Alive: Their Functions Revealed by Electromyography. Baltimore, MD: Williams & Wilkins, 1985.
3. Beck J, Sinderby C, Lindstrom L, and Grassino A. Effects of lung
volume on diaphragm EMG signal strength during voluntary contractions.
J Appl Physiol 85: 1123–1134, 1998.
4. Beck J, Sinderby C, Lindstrom L, and Grassino A. Influence of bipolar
esophageal electrode positioning on measurements of human crural diaphragm electromyography. J Appl Physiol 81: 1434 –1449, 1996.
5. Beck J, Sinderby C, Weinberg J, and Grassino A. Effects of muscleto-electrode distance on the human diaphragm electromyogram. J Appl
Physiol 79: 975–985, 1995.
6. Boyd WH and Basmajian JV. Electromyography of the diaphragm in
rabbits. Am J Physiol 204: 943–948, 1963.
7. Braun NMT, Arora NS, and Rochester DF. Force-length relationship of
the normal human diaphragm. J Appl Physiol 53: 405– 412, 1982.
8. Cohn D, Benditt JO, Eveloff S, and McCool FD. Diaphragm thickness
during inspiration. J Appl Physiol 83: 291–296, 1997.
9. Cooke IRC, Soust M, and Berger PJ. Differential recruitment of inspiratory muscles in response to chemical drive. Respir Physiol 92:
167–181, 1993.
10. De Troyer A and Legrand A. Inhomogeneous activation of the parasternal intercostals during breathing. J Appl Physiol 79: 55– 62, 1995.
11. De Troyer A, Samson M, Sigrist S, and Macklem PT. The diaphragm:
two muscles. Science 213: 237–238, 1981.
12. Easton PA, Abe T, Smith J, Fitting JW, Guerraty A, and Grassino
AE. Activity of the costal and crural diaphragm during progressive
hypoxia and hypercapnia. J Appl Physiol 78: 1985–1992, 1995.
13. Hakansson CH. Conduction velocity and amplitude of action potential as
related to circumference in the isolated frog muscle. Acta Physiol Scand
37: 14 –34, 1956.
14. Irsigler GB. Carbon dioxide response lines in young adults: the limits of
the normal response. Am Rev Respir Dis 114: 529 –536, 1976.
15. Johnson RLJ, Hsia CCW, Takeda SI, Wait JL, and Glenny RW.
Efficient design of the diaphragm: distribution of blood flow relative to
mechanical advantage. J Appl Physiol 93: 925–930, 2002.
16. Langman J. Medical Embryology. Baltimore, MD: Williams & Wilkins,
1975.
17. Laskowski MB and Sanes JR. Topographic mapping of motor pools onto
skeletal muscles. J Neurosci 7: 252–260, 1987.
18. Legrand A, Brancatisano A, Decramer M, and De Troyer A. Rostrocaudal gradient of electrical activation in the parasternal intercostals
muscle of the dog. J Physiol 495: 247–254, 1996.
98 • JANUARY 2005 •
www.jap.org
Downloaded from http://jap.physiology.org/ by 10.220.33.5 on June 15, 2017
relationship between RMSdi and PETCO2 could become nonlinear.
Implications. The finding of a linear relationship between
PCO2 and RMSdi, up to moderate levels of diaphragm activation (Table 1 and Fig. 3), is consistent with the findings of Beck
et al. (3), who found a linear relationship between RMSdi and
the ratio of Pdi to maximal Pdi up to 75% of maximal Pdi. This
confirms that RMSdi accurately reflects moderate levels of
activation of the diaphragm, although it could underestimate
higher levels of activation because RMSdi varies with the
square root of activation (3). In most subjects, the r2 value for
the relationship between RMSdi and PCO2 increased when fB
was considered, implying that most subjects responded to
different levels of PCO2 by varying the frequency and amount
of diaphragm activation in slightly different proportions. The
linear relationship between RMSdi and PETCO2 up to 60%
RMSdimax suggests that RMSdi accurately reflects neural drive
to the diaphragm. As a measure of chemosensitivity, it provides an output that is independent of diaphragm length,
diaphragm weakness and threshold, resistive and elastic loads.
Intersubject variation in the relationship between RMSdi
and PETCO2 was partly attributable to intersubject variation
in chemosensitivity, evidenced by the variability in the
relationship between V̇E and PETCO2 (Table 3). The variability in the RMSdi-PCO2 relationship exceeded that of the
V̇E-PCO2 relationship (Table 3 and Fig. 4). This additional
variability could be due to differences in diaphragm anatomy, spatial relationship of electrode to muscle, possibly
because of anatomical differences at the gastroesophageal
junction, and/or filtering properties of the electrode or surrounding tissue. These factors are likely also to influence
RMSdirest, RMSdisubmax, and RMSdimax, which could be
used to reduce these intersubject differences, so that the
main source of intersubject variation in RMSdi%max-PCO2
relationship was chemosensitivity. These differences are
also likely to influence CFdi, which increases with action
potential propagation velocity and decreases as the muscleto-electrode distance increases (5). However, intersubject
differences in CFdi in the present study did not account for
intersubject variation in RMSdi. This could be due to
cancellation by the different sources of variation or the wide
range in CFdi obtained with esophageal electrodes. The fB
and thickness of the costal diaphragm did not account for
intersubject differences in the relationship between RMSdi
and hypercapnia. These findings validate the approach of
Sinderby et al. (30), who used the ratio of RMSdi to
RMSdimax to show increased diaphragm activation in subjects with severe chronic airflow limitation and restriction
due to previous poliomyelitis compared with healthy subjects.
In keeping with the findings of Irsigler (14), considerable
day-to-day variability in the V̇E and VT/TI responses to hypercapnia was found, and this accounted for part of the day-to-day
variation in RMSdi. Additional day-to-day variation in the
RMSdi response to hypercapnia is likely to be due to variations
in the muscle-to-electrode distance and orientation and perhaps
changes to the filtering properties of the electrode. Such differences are likely to be best corrected by normalization with
an effort-independent electrical signal of constant amplitude
that is detected by the same esophageal electrode used to detect
EMG activity. The data show that the R wave of the ECG
DIAPHRAGM EMG ROOT MEAN SQUARE RESPONSE TO CO2
J Appl Physiol • VOL
29. Sant’Ambrogio G, Frazier DT, Wilson MF, and Agostoni E. Motor innervation and pattern of activity of cat diaphragm. J Appl Physiol 18: 43–46, 1963.
30. Sinderby C, Beck J, Spahija J, Weinberg J, and Grassino A. Voluntary
activation of the human diaphragm in health and disease. J Appl Physiol
85: 2146 –2158, 1998.
31. Sinderby C, Lindstrom L, and Grassino AE. Automatic assessment of
electromyogram quality. J Appl Physiol 79: 1803–1815, 1995.
32. Sinderby C, Spahija J, Beck J, Kaminski D, Yan S, Comtois N, and
Sliwinski P. Diaphragm activation during exercise in chronic obstructive pulmonary disease. Am J Respir Crit Care Med 163: 1637–1641,
2001.
33. Sinderby CA, Beck JC, Lindstrom LH, and Grassino AE. Enhancement of signal quality in esophageal recordings of diaphragm EMG.
J Appl Physiol 82: 1370 –1377, 1997.
34. Soust M, Walker AM, and Berger PJ. Diaphragm V̇O2, diaphragm
EMG, pressure-time product and calculated ventilation in newborn lambs
during hypercapnic hyperpnoea. Respir Physiol 76: 107–117, 1989.
35. Wade OL. Movements of the thoracic cage and diaphragm in respiration.
J Physiol 124: 193–212, 1954.
36. Whitelaw WA, Derenne JP, and Milic-Emili J. Occlusion pressure as a
measure of respiratory centre output in conscious man. Respir Physiol 23:
181–199, 1975.
37. Wilson TA, Boriek AM, and Rodarte JR. Mechanical advantage of the
canine diaphragm. J Appl Physiol 85: 2284 –2290, 1998.
98 • JANUARY 2005 •
www.jap.org
Downloaded from http://jap.physiology.org/ by 10.220.33.5 on June 15, 2017
19. Legrand A and De Troyer A. Spatial distribution of external and internal
intercostals activity in dogs. J Physiol 518: 291–300, 1999.
20. Lindstrom L. A model describing the power spectrum of myoelectric
signals. I. Single fiber signal. Res Lab Med Electr 5: 73, 1973.
21. Lopata M, Evanich MJ, and Lourenco RV. Quantification of diaphragm
EMG response to CO2 rebreathing in humans. J Appl Physiol 43: 262–270,
1977.
22. Lourenco RV, Cherniack NS, Malm JR, and Fishman AP. Nervous
output from the respiratory center during obstructed breathing. J Appl
Physiol 21: 527–533, 1966.
23. Mead J and Loring SH. Analysis of volume displacement and length changes of
the diaphragm during breathing. J Appl Physiol 53: 750–755, 1982.
24. Mead J, Loring SH, and Smith JC. Volume displacements of the chest
wall and their mechanical significance. In: The Thorax (2nd ed.), edited by
Roussos C. New York: Dekker, 1995, p. 565–586.
25. Onal E, Lopata M, Ginzburg AS, and O’Connor TD. Diaphragmatic
EMG and transdiaphragmatic pressure measurements with a single catheter. Am Rev Respir Dis 124: 563–565, 1981.
26. Read DJ and Leigh J. Blood-brain tissue PCO2 relationships and ventilation during rebreathing. J Appl Physiol 23: 53–70, 1967.
27. Riley DA and Berger J. A regional histochemical and electromyographic
analysis of the cat respiratory diaphragm. Exp Neurol 66: 636 – 649, 1979.
28. Rochester DF and Bettini G. Diaphragmatic blood flow and energy
expenditure in the dog. Effects of inspiratory airflow resistance and
hypercapnia. J Clin Invest 57: 661– 672, 1976.
281