Neural respiratory drive in healthy subjects and in COPD

The aim of the present study was to use the diaphragm electromyogram (EMGdi) to compare levels of neural respiratory drive (NRD) in a cohort of healthy subjects and chronic obstructive pulmonary disease (COPD) patients, and to investigate the relationship between NRD and pulmonary function in COPD. EMGdi was recorded at rest and normalised to peak EMGdi recorded during maximum inspiratory manoeuvres (EMGdi % max) in 100 healthy subjects and 30 patients with COPD, using a multipair oesophageal electrode. EMGdi was normalised to the amplitude of the diaphragm compound muscle action potential (CMAPdi,MS) in 64 healthy subjects. The mean±sd EMGdi % max was 9.0±3.4% in healthy subjects and 27.9±9.9% in COPD patients, and correlated with percentage predicted forced expiratory volume in one second, vital capacity and inspiratory capacity in patients. EMGdi % max was higher in healthy subjects aged 51–80 yrs than in those aged 18–50 yrs (11.4±3.4 versus 8.2±2.9%, respectively). Observations in the healthy group were similar when peak EMGdi or CMAPdi,MS were used to normalise EMGdi. Levels of neural respiratory drive were higher in chronic obstructive pulmonary disease patients than healthy subjects, and related to disease severity. Diaphragm compound muscle action potential could be used to normalise diaphragm electromyogram if volitional inspiratory manoeuvres could not be performed, allowing translation of the technique to critically ill and ventilated patients.

O bjective markers of disease severity that reflect the physiological load on the respiratory system in chronic obstructive pulmonary disease (COPD) are currently lacking. Although COPD severity is categorised in terms of forced expiratory volume in one second (FEV1) in management guidelines [1], correlations between FEV1 and breathlessness [2] or quality of life are modest [3], and reported relationships between FEV1 and prognosis are inconsistent [4][5][6]. Two small studies confirm that neural respiratory drive (NRD) is increased in COPD [7] and relates to symptoms [8], but the use of measurements of NRD to assess disease severity in COPD has not been fully investigated, in part because there are no data to define ranges of NRD within the healthy population.
In COPD, mechanical abnormalities including airflow obstruction, static and dynamic hyperinflation and intrinsic positive end-expiratory pressure increase the load on the respiratory muscles. The translation of inspiratory muscle contraction into negative intrathoracic pressure, and of pressure changes to ventilation, is impaired as a consequence of muscle shortening, increased velocity of contraction, alteration in geometry and reduced compliance of the respiratory system. This results in high NRD in COPD, and disproportionate increases whenever airways obstruction worsens (and hyperinflation increases) or ventilatory requirements increase. The neural output of the brainstem respiratory centre cannot easily be measured directly in humans, but NRD can be assessed indirectly by quantifying the electromyogram (EMG) of the respiratory muscles, which provides a method of assessing the level and pattern of their activation [9]. The EMG of the diaphragm (EMGdi), the major inspiratory muscle during resting tidal breathing in healthy individuals, can be recorded specifically using oesophageal electrodes positioned at the diaphragm crus [10]. Multipair oesophageal electrode catheters have been developed and used to measure NRD [11][12][13][14]. Using these electrodes, SINDERBY et al. [7] have shown that the amplitude of EMGdi is higher in patients with significant respiratory disease than in healthy subjects, when normalised to each subject's volitional maximum (EMGdi % max). However, the number of subjects in the study was small (five healthy males, five COPD patients and five post-polio infection), and it was not designed to explore the use of EMGdi % max as a marker of disease severity.
The main aims of the present study were to use the EMGdi to measure NRD in a large cohort of healthy subjects for comparison with levels of NRD in COPD patients, and to investigate the relationship between NRD and pulmonary function in COPD. It was hypothesised that resting EMGdi % max would be significantly higher in COPD patients than in healthy subjects, and that resting EMGdi % max would be highest in COPD patients with the most severe disease. A further aim was to test the hypothesis that there is a positive correlation between the amplitude of the diaphragm compound muscle action potential recorded following bilateral anterolateral magnetic stimulation (CMAPdi,MS) and the peak root mean square of spontaneous EMGdi activity (RMS-EMGdi,peak) recorded during maximal volitional inspiratory manoeuvres. If so, normalising EMGdi to CMAPdi,MS could avoid the difficulties associated with the use of volitional tests of maximum diaphragm activation for normalisation in clinical situations where it is impossible to perform the necessary inspiratory manoeuvres. Assessing the load on the respiratory system by quantifying EMGdi nonvolitionally in this way could be particularly valuable in the critical care population, particularly in the assessment of the need for ventilatory support, where EMGdi % max values could provide an index of ventilatory reserve. Indeed, calibration of the level of ventilatory support in response to levels of neural respiratory drive is the basis of novel neurally adjusted ventilatory assist (NAVA) technology currently in development [15].
Instrumentation and signal processing EMGdi recordings were made from the crural diaphragm using multipair oesophageal electrode catheters, as previously described [13]. Further details of the electrode design, positioning and signal processing are given in the online supplementary material.
EMGdi recordings at rest and during maximal inspiratory manoeuvres Recordings were made sitting upright in a chair, with a noseclip in place for all measurements except sniff nasal pressure.
To record EMGdi during resting breathing, subjects sat quietly in a relaxed posture for o5 min, until o2 min of stable, consistent EMGdi signals had been recorded. Airflow was measured through a mouthpiece connected in series to a pneumotachograph. EMGdi was then recorded during four inspiratory manoeuvres: 1) maximal inspiration to total lung capacity (TLC); 2) maximal static inspiratory effort at functional residual capacity (FRC) against a closed valve [16]; 3) maximal sniff from FRC; and 4) maximum voluntary ventilation for 15 s (''sprint MVV''). Manoeuvres 1-3 were repeated at least three times, until the investigator was satisfied that a truly maximum effort had been performed. The sprint MVV was performed once only.

Calculation of resting EMGdi
The raw signal was converted to root mean square (RMS; Powerlab Chart v5.4 software, ADInstruments, Chalgrove, UK), using a time constant of 50 ms and a moving window. The maximum RMS-EMGdi value during 100-ms subdivisions of each breath was then determined, manually selecting EMGdi signals falling between QRS complexes of the ECG artefact. The mean maximum RMS-EMGdi per breath over two representative 30-s subdivisions of the whole recording was then calculated.

Calculation of EMGdi % max
EMGdi signals recorded during each of the maximum inspiratory manoeuvres were converted to RMS. The largest RMS-EMGdi value calculated by analysis of these recordings was labelled ''RMS-EMGdi,peak''. EMGdi % max for each subject was then calculated as the mean maximum RMS-EMGdi per breath as a percentage of RMS-EMGdi,peak.
Assessment of intra-and interobserver reproducibility of EMGdi % max In total, 10 healthy subjects were studied on two occasions .24 h apart, at the same time of day. The intraobserver reproducibility of EMGdi % max measurements was assessed by comparing the results of a single investigator's (C.J. Jolley) analysis of measurements made on two separate days. The interobserver reproducibility of EMGdi % max measurements was assessed by comparing the results of two investigators' (C.J. Jolley and C. Reilly) analysis of a single set of measurements in five of these subjects.
Bilateral anterolateral magnetic phrenic nerve stimulation Bilateral anterolateral magnetic phrenic nerve stimulation (BAMPS) was performed using two double circular 43-mm coils (P/N 9784-00; Magstim Co.,Whitland, UK) placed anterolaterally over the left and right phrenic nerves, as previously described [17]. The coils were powered by a Magstim 200 stimulator (Magstim Co.). During the study, subjects were seated upright in a chair with a nose-clip in place. Stimulation was performed at end-expiration with the abdomen unbound. BAMPS was performed at 80, 85, 90, 95 and 100% maximum stimulator output (MSO), to determine supramaximality. The amplitude of the CMAPdi,MS was measured as the peak-trough amplitude, as previously described [18]. The interoccasion coefficient of variation (CV) in the previous study was 8.6% [18].

Statistical analysis
Ranges of EMGdi % max were expressed as 95% confidence intervals of the mean. Comparisons between healthy and COPD subjects were made using independent sample t-tests except comparisons of sex distributions, which were made using Fisher's exact test. Values of p were considered to be significant at ,0.05 level. Relationships between EMGdi and anthropometric or lung function variables were investigated by regression analysis. Intra-and interobserver reproducibility was assessed by calculating the CV and by Bland-Altman analysis [19].

RESULTS
Data are presented as mean¡SD. Anthropometric and lung function data for both healthy subjects and COPD patients are summarised in table 1.
Representative traces at rest and during maximum voluntary ventilation in a healthy subject and a COPD patient are shown in figure 1. Comparisons of peak RMS-EMGdi values during the different manoeuvres are given in the online supplementary material.
Correlations between EMGdi % max and age, height, weight and BMI were similar in the white European and Chinese subgroups. Data comparing EMGdi % max values in white European and Chinese ethnic groups are provided in the online supplementary material.

CMAPdi,MS
CMAPdi,MS was assessed in 64 subjects. Supramaximality was judged to have been achieved when the mean CMAPdi,MS amplitude at 100% MSO was less than 5% greater than the highest mean CMAPdi,MS amplitude achieved at the lower stimulator outputs. Using these criteria, supramaximality was achieved in 92.8% of the subjects. The CMAPdi,MS amplitude achieved at 100% MSO values was recorded if supramaximality was not achieved.
Representative traces recorded during BAMPS are shown in the online supplementary material. The mean¡SD CMAPdi,MS amplitude was 2.4¡0.7 mV. The phrenic nerve conduction time (PNCT), defined as the time from the stimulation artefact to the onset of the CMAP, was 6.9¡0.7 ms.
All patients completed FEV1 and VC measurements, and IC was measured in 20 patients. Significant correlations, best described by curve regression functions, were observed between EMGdi % max and FEV1 % pred, VC % pred and IC % pred, and between (VT % VCpred)/(EMGdi % max) and FEV1 % pred, VC % pred   within the same subject). In the healthy group, significant differences were observed between all manoeuvres except sniff and PI,max (p50.62).

Intrasubject reproducibility
The coefficient of repeatability between EMGdi % max measurements made and analysed by the same investigator on two separate days, in 10 subjects, was 0.94 [19]. The mean¡SD CV was 0.09¡0.05.

Interobserver reproducibility
The mean¡SD CV of EMGdi % max between measurements made in five subjects on the same day, comparing the results of analysis by two investigators, was 0.10¡0.08. The intraclass correlation coefficient was 0.71. Tables and Bland-Altman plots of these data are presented in the online supplementary material.

DISCUSSION
The present study is the first to define normal ranges of NRD as EMGdi % max in a large population of healthy adults; the largest previous study included 15 participants [7]. In healthy subjects, levels of EMGdi % max increased slightly with age, but there were no significant correlations between EMGdi % max and sex, height, weight or BMI in this largely non-obese population.
EMGdi % max was found to be significantly higher in the COPD group than in matched healthy subjects. The average EMGdi % max in the current cohort of healthy subjects was 9.0¡3.4% overall, which is of the same order as levels of EMGdi % max described previously in a smaller study by SINDERBY et al. [7], who demonstrated resting values of 8.4¡2.5% and 43.4¡22.1% in five healthy and five severe COPD patients, respectively, using similar methods. The lower average EMGdi % max found in the present COPD patients is likely to reflect the range of disease severity. By showing that there is a correlation between EMGdi % max and the degree of airflow obstruction and hyperinflation, and that this is a reproducible measure, the present study builds on earlier observations and demonstrates the potential value of EMGdi % max as an objective marker of disease severity in COPD.

Determinants of EMGdi % max
In general, levels of NRD increase when the load on the respiratory muscles increases relative to their capacity, i.e. if the load increases, the capacity of the muscles decreases, or a combination of these two changes. Levels of EMGdi % max can therefore be explained in terms of ventilatory mechanics, and the pathophysiological changes in ventilatory mechanics that occur with disease.

Healthy subjects
An average EMGdi % max of 9.0% in normal subjects, in whom it can be assumed that there is no neuromechanical dissociation, is consistent with the high levels of ventilatory reserve that are known to exist in healthy individuals. The slightly increased EMGdi % max observed in the older (51-80 yrs of age) cohort compared with that of subjects aged ,50 yrs is likely to reflect the known ''normal'' changes in ventilatory mechanics occurring with increased age. Declines in FEV1 [20], VC [21], respiratory muscle strength [22] and chest wall compliance [23] observed during healthy ageing all increase the load:capacity ratio of the respiratory muscle pump, reducing ventilatory reserve, and would explain the tendency to higher levels of EMGdi % max in the older age group. The findings of significant negative correlations between EMGdi % max and absolute FEV1  and VC are consistent with this. The current observation that the correlation of SNIP and PI,max with EMGdi % max is weak suggests that altered lung and chest wall mechanics are more important contributors to increased drive than reduced diaphragm contractility in the healthy older cohort.

COPD
The results of the present study confirm the hypotheses that EMGdi % max would be higher in COPD patients than healthy subjects, and that the levels of EMGdi % max would be highest in patients with the most severe disease. High levels of EMGdi % max indicate that there is a relative increase in the RMS of EMGdi in COPD compared with healthy subjects, i.e. recruitment of larger numbers of diaphragm motor units and/or an increase in diaphragm motor unit firing rate in COPD. It is, in fact, well known that the firing frequency of motor neurons supplying both the diaphragm [24] and nondiaphragmatic muscles [25] is increased in COPD. However, DE TROYER et al. [24] and GANDEVIA et al. [25] used needle electrodes, which is clearly not feasible in general clinical practice.
The increase in the RMS of EMGdi in COPD is likely to be the result of three main factors. First, the diaphragm must generate more pressure to achieve a given VT, compared with healthy subjects. Increased airways resistance in COPD results in significant expiratory airflow limitation at rest, leading to gas trapping, which increases intrathoracic end-expiratory pressure. A positive end-expiratory pressure imposes a threshold load that must be overcome before inspiratory airflow can be generated. A reduction in chest wall compliance, as hyperinflation progresses, also contributes to the mechanical load associated with inspiration in severe disease. Hyperinflation is also due to a loss of elastic recoil in emphysema. Secondly, the maximum pressuregenerating capacity of the diaphragm is reduced. POLKEY et al.
[26] demonstrated a linear negative correlation of twitch transdiaphragmatic pressure with increasing lung volume of 3.5 cmH 2 O?L -1 . The ability of the diaphragm to generate transdiaphragmatic and oesophageal pressure is, therefore, reduced in COPD, and these changes are exaggerated with acute-on-chronic hyperinflation. Thirdly, patients with COPD need to generate increased absolute levels of ventilation to overcome ventilation/perfusion (V9/Q9) mismatch [27].
Although RMS-EMGdi,peak in COPD patients was 83% lower than in healthy subjects, a lower denominator is unlikely to explain the higher EMGdi % max in COPD, as correcting for this gives an average EMGdi % max in COPD of 23.5¡8.2%, still significantly higher than EMGdi % max in the healthy group (p,0.001).
Significance of raised EMGdi % max and reduced (VT % VCpred)/(EMGdi % max) in COPD The current finding that EMGdi % max is raised, and is negatively correlated with FEV1, VC and the degree of hyperinflation at rest (described in terms of IC % pred) in COPD, reinforces the contention that, by providing a composite measure of ventilatory load and capacity, EMGdi % max could provide an alternative method of assessing COPD disease severity. This could be the focus of future studies in larger numbers of COPD patients, including a more detailed investigation of the relationship between EMGdi % max and other physiological measures, including hypoxaemia, hypercapnia and V9/Q9 mismatch, than were carried out in the present study.
(VT % VCpred)/(EMGdi % max) was lower in COPD patients than healthy subjects, reflecting neuromechanical uncoupling in COPD, and this correlated with disease severity. Neuromechanical dissociation has previously been demonstrated during exercise in COPD using EMGdi measurements [28], but the relationship with disease severity has not previously been documented. This observation also emphasises the value of EMGdi % max over other commonly used indirect measures of ventilatory drive, such as mouth occlusion pressure at 100 ms (P0.1), or the amplitude of tidal oesophageal pressure swings, which will underestimate levels of NRD in COPD patients with the most neuromechanical dissociation.

Bilateral CMAPdi,MS values
To the best of the current authors' knowledge, there have been no previous studies that have compared CMAPdi,MS and RMS-EMGdi,peak in healthy subjects. Normal ranges of bilateral    [29]. P0.1 would, however, underestimate NRD in patients with disordered ventilatory mechanics, such as in COPD, where neuromechanical dissociation progresses exponentially as airflow obstruction and hyperinflation worsen [28]. This approach would also allow the level of EMGdi activity at which neural-assist ventilators such as NAVA [15] are triggered to be defined as EMGdi % max, hence defining this threshold in terms of ventilatory reserve. Since ventilatory failure is the outcome of a critically low ventilatory reserve, this could prove to be a more appropriate approach than increasing NAVA support in response to changes from baseline EMGdi activity.

Potential clinical applications of EMGdi measurements to quantify NRD
The current study technique could be usefully applied to measure disease severity, progression and responses to treatment, in any disorder characterised by increased ventilatory load (e.g. airflow obstruction in asthma and COPD, reduced lung compliance in pulmonary fibrosis, or cardiac failure), reduced ventilatory capacity (in neuromuscular disease), or where there is a combination of both factors, such as in COPD, as herein discussed. The value of the method over other objective physiological measurements of disease severity, such as spirometry, or measurement of lung volumes, is that recording EMGdi gives a breath-by-breath measure of the load on the respiratory system, and can be used to provide measurements continuously during sleep without waking the patient [30], during exercise [28], and, as mentioned in the aforegoing discussion, could in addition be measured nonvolitionally in ventilated patients. The main factor limiting the translation of the technique to clinical practice is the acceptability of the oesophageal catheters to patients. However, in the current authors' experience of the use of these and similar catheters to assess intrathoracic pressure in clinical practice, the catheters are acceptable in .95% of patients and are usually well tolerated.
In conclusion, the present study has demonstrated, in a large cohort of healthy subjects and patients with chronic obstructive pulmonary disease, that levels of neural respiratory drive, measured as diaphragm electromyogram as a percentage of maximum, are higher in patients with chronic obstructive pulmonary disease than in healthy subjects, and highest in patients with the most severe airflow obstruction and hyperinflation. Diaphragm electromyogram as a percentage of maximum therefore provides a composite measure of ventilatory load and capacity, and could provide a method of assessing chronic obstructive pulmonary disease severity. Normal ranges of diaphragm electromyogram as a percentage of maximum have also been established, which may be used for comparative data in future studies in patients with chronic obstructive pulmonary disease, or indeed any other cardiorespiratory disease, to further understanding of the pathophysiology of ventilatory failure. The current findings also demonstrate that nonvolitional activation of the diaphragm is of potential use in the assessment of diaphragm electromyogram as a percentage of maximum in patients who are unable to perform maximal volitional inspiratory manoeuvres.