Quadriceps muscle dysfunction is common in COPD. Determining, and, if possible, predicting quadriceps phenotype in COPD is important for patient stratification for therapeutic trials.
Quadriceps muscle dysfunction is common in COPD. Determining, and, if possible, predicting quadriceps phenotype in COPD is important for patient stratification for therapeutic trials.
In biopsies from 114 COPD patients and 30 controls, we measured fiber size and proportion and assessed the relationship with quadriceps function (strength and endurance), clinical phenotype (lung function, physical activity, fat-free mass) and exercise performance. In a subset (n = 40) we measured muscle mid-thigh cross-sectional area by computed tomography.
Normal ranges for fiber proportions and fiber cross-sectional area were defined from controls; we found isolated fiber shift in 31% of patients, isolated fiber (predominantly type II) atrophy in 20%, both shift and atrophy in 25%, and normal fiber parameters in 24%. Clinical parameters related poorly to muscle biopsy appearances.
Quadriceps morphology is heterogeneous in COPD and cannot be predicted without biopsy, underlining the need for biomarkers. Muscle Nerve 48: 488–497, 2013
chronic obstructive pulmonary disease
forced expiratory volume in 1 s
fat mass index
fat-free mass index
oxidative type I to glycolytic type II fiber type shift
muscle mid-thigh cross-sectional area
maximal voluntary contraction
carbon monoxide diffusion capacity
time to fatigue to 80% of initial pressure-product
unpotentiated quadriceps twitch force
Chronic obstructive pulmonary disease (COPD) is the most common respiratory disease in adults and the only common disease in the western world in which the morbidity and mortality continue to increase. Weakness and wasting of the quadriceps are associated with reduced quality of life and survival independent of lung function in COPD. Because pulmonary rehabilitation improves exercise performance and quality of life,[5, 6] most likely through improvement of lower limb muscle function,[7, 8] there has been increasing interest in the development of novel drugs which might act as an adjunct or alternative (for those unable to participate) to rehabilitation.
Based on pioneering, but relatively small studies, for example those of Bernard et al. or Jakobsson et al. it was historically considered that skeletal muscle weakness in COPD was a phenomenon which was related to severity of airflow limitation or hypoxia, and, therefore, identification of patients who might benefit from anabolic approaches would be straightforward. However, recent and substantially larger studies have shown clearly that both quadriceps weakness and wasting are present in patients with mild disease, yet it can also be absent in patients with the most severe disease.[11, 12] Thus, by the study of large cohorts, a revised view has formed in which it is now clear that quadriceps dysfunction is substantially independent of lung function.
For quadriceps biopsies it is generally agreed that a key feature of COPD is a change in fiber composition. In healthy 60- to 70-year-olds, the quadriceps is composed of approximately 50% type I (oxidative) fibers and 50% type II fibers (type IIx glycolytic and type IIa intermediate oxidative/glycolytic fibers). A minimal percentage of hybrid fibers are in transition between fiber types and, therefore, express 2 types of myosin simultaneously (e.g., I/IIa and IIa/IIx fibers). In the quadriceps of patients with COPD, there is a preponderance of type II fibers and an increase in hybrid fibers, particularly type IIa/IIx fibers, which we term (a type I to II) fiber shift (FS). Fiber atrophy (FA) is also reported[14, 15] although fewer data on the topic exist; macroscopic muscle atrophy, detected by bioelectrical impedance, DEXA, thigh CT, or ultrasound, is commonly reported. Thus the question arises as to how heterogeneous the muscle phenotype is in patients with COPD and whether the phenotype has any relation to either muscle function or clinical phenotype. The issue is particularly pertinent with the prospect of drugs able, at least in animals, to influence skeletal muscle characteristics through different modalities. For example an AMP Kinase activator, 5-amino-1-β-D-ribofuranosyl-imidazole-4-carboxamide (AICAR), has been reported to cause a type II to type I fiber type shift, while activin IIb receptor blockade results in fiber hypertrophy. Further discussion of this problem in relation to COPD is found in Steiner et al.
When considering the nature of FS in COPD, a particular concern relates to the sample size of published reports. At the time of our previous meta-analysis in 2007 the largest COPD cohort for which FS data were available was 32, and this remained the highest number available from a single study although the data from a cohort of 54 patients appeared after we completed sample collection for this study. We, therefore, undertook a substantially larger prospective cross-sectional study of COPD patients with a range of disease severity (judged spirometrically) to examine the heterogeneity of muscle FS and FA and associations between FS and FA as well as the muscle phenotype and clinical characteristics.
The study was approved by the Royal Brompton & Harefield NHS Trust and Ealing and West London Mental Health Trust Research Ethics Committees (06/Q0404/35 and 06/Q0410/54), and participants gave written informed consent.
Participation was requested from a convenience sample of GOLD Stage I to IV COPD patients seen at the COPD clinic at the Royal Brompton Hospital; 114 patients consented. Thirty healthy age- and gender-matched controls recruited by advertisement, were studied. Exclusion criteria were: diagnoses of heart, renal or liver failure, systemic inflammatory, metabolic or neuromuscular disorders, warfarin (bleeding risk from biopsy), or a moderate/severe exacerbation (i.e., requiring intervention) within the preceding 4 weeks.
Postbronchodilator spirometry, lung volumes (plethysmography), carbon monoxide diffusion capacity,[24-26] and resting arterialized capillary earlobe blood gas tensions were measured. Fat-free mass index (FFMI) was calculated from bioelectrical impedance measurements (Bodystat 1500, UK) taken after participants had rested supine for 20 min by use of a disease-specific regression equation for calculation of FFMI. A low FFMI was defined as an FFMI less than 15 kg/m2 for women and less than 16 kg/m2 in men. Muscle mid-thigh cross-sectional area (MTCSA, the bulk of muscle at the mid-point of the thigh) was measured by CT in a subset of 31 patients and 10 controls, as a low MTCSA indicates muscle atrophy and is a marker of increased mortality in COPD. Physical activity over 12 h on 2 days was measured (Dynaport accelerometer, McRoberts BV, Netherlands) as previously described in patients with COPD.
Quadriceps strength was measured as supine isometric Maximal Voluntary Contraction (MVC). Quadriceps endurance (T80) timed by force decline to 80% of initial force during repetitive magnetic femoral nerve stimulation in the dominant leg was measured as we have described.[29, 30] Exercise performance was assessed by a 6-min walk test according to ATS guidelines and by symptom-limited incremental cycle ergometry with metabolic testing, as described previously.
Percutaneous biopsy of the vastus lateralis was performed using the Bergstrom technique on an occasion separate from strength and cycle tests. Muscle samples were mounted on cork in OCT (Tissue Tek compound), frozen in precooled isopentane, then liquid nitrogen, and stored at −80°C until cryosectioning.
Immunohistochemistry using anti-type I and IIa myosin and laminin antibodies was performed on transverse muscle cross-sections. Epifluorescence signal was recorded using a Nikon Eclipse 800 microscope with a DXM 1200 camera (Nikon Instruments Europe BV, the Netherlands) under a ×10 objective using 3 filters: Texas Red (395 to 410 nm), FITC (490 to 505 nm) and DAPI UV (395 to 410 nm). Type I fibers were visualized as red, type IIa as green, type I/IIa (staining both red and green), and type IIx as black (null staining), and laminin staining of the fiber borders (blue) allowed calculation of individual fiber CSA. From this, type I, IIa (IIa and IIa/IIx), IIx, and I/IIa fiber proportions and the median fiber CSA for each fiber type could be calculated from at least 100 fibers per participant, as done previously.[34, 35] Data from 2 biopsies from 1 leg were pooled for 26 COPD patients and 25 controls to ensure data from at least 100 fibers in transverse section; the minimum figure of 100 fibers was chosen to ensure comparability with Whittom and co-workers. In 42 patients taking part in another study requiring bilateral biopsies, biopsies from both legs were available, and for these participants data from biopsies from each leg were pooled.
All data were analyzed with help of a medical statistician using SPSS 15 (IBM) software.
Fiber proportions and fiber CSA from our healthy controls were not normally distributed and did not log transform to a normal distribution. Therefore, as is conventional for nonparametric data, reference ranges were set between the 2.5th and 97.5th percentile. Because there are gender influences in fiber characteristics, we calculated reference ranges for fiber characteristics for elderly control men and women separately.
Patient data were compared with the gender-appropriate reference interval calculated from our own controls. Individuals were deemed to have FS if their type I fiber proportions fell below the lower end of the reference range for type I fiber proportions, or if the type IIx or type II (sum of type IIa and IIx) fiber proportions were above the upper end of the relevant reference range. This approach (but using normally distributed data and, therefore, a reference range 2 standard deviations on either side of the mean) has been used previously to define FS. The same approach was used to define FA, i.e., a reduction in CSA of individual fibers; fiber CSA (at least for type IIx fibers) is closely correlated with FFM in patients with COPD. FA was designated if the CSA of any of the pure fiber types (I, IIa, or IIx) fell below the lower limit of the corresponding reference range. FA was not defined on the basis of hybrid fibers, because in some individuals there were none or too few in a specimen to give a representative result.
Group differences in normally and nonnormally distributed variables were analyzed with an unpaired t-test and the Mann Whitney U-test respectively and with the Fisher exact test for categorical variables. Pearson and Spearman rank correlation coefficients were used to analyze relationships between normally and nonnormally distributed variables respectively. Linear regression determined whether relationships between normally distributed variables were independent.
A total of 114 COPD patients and 30 healthy controls completed the study. Patients and controls were age- and gender-matched. Patients were divided evenly between GOLD Grade II (moderate), III (severe), and IV (very severe) COPD, and there were 4 GOLD Grade I (mild) patients. As expected, patients had significantly worse lung function, arterial blood oxygen tensions, FFMI, MTCSA, quadriceps strength and endurance, physical activity, and exercise performance. A median of 223 (166, 333) fibers was analyzed from each subject. Considered as a group, we observed, in line with prior findings, a lower proportion of type I, a higher proportion of type II fibers, and a reduced type IIx fiber CSA compared with controls. The physiological and fiber characteristics of both groups are shown in Table 1.
|COPD patients (n = 114)||Healthy controls (n = 30)||P-value|
|Age (years)||65 (61,73)||67 (63,72)||0.55|
|Gender (% men)||67||53||0.18|
|Smoking history (pack-years)||45 (34,61)||3 (0,19)||<0.0001|
|FEV1 (% predicted)||41 (27,57)||109 (100,113)||<0.0001|
|TLCO (% predicted)||44 (28,55)||87 (80,96)||<0.0001|
|PaO2 (kPa)||9.2 (8.4,10.2)||11.1 (10.1,11.7)||<0.0001|
|Fat-free mass index (kg/m2)||15.7 (14.5,17.2)||16.5 (15.3,20)||0.01|
|Muscle mid-thigh CSA (cm2)||94 (27) (n = 31)||115 (24) (n = 10)||0.03|
|Quadriceps MVC (kg)||28 (21,38)||34 (28,43)||0.01|
|Quadriceps twitch force (kg)||7.6 (6.1,9.2)||8.5 (6.9,10.9)||0.09|
|Quadriceps endurance (T80;s)||80 (65,105)||110 (81,184)||0.002|
|Locomotion time (min/ 12 hours)||41 (23,67)||95 (59,138)||<0.0001|
|6MW distance (% predicted)||76 (24)||121 (14)||<0.0001|
|Peak VO2 (% predicted)||46 (34,57)||97 (79,106)||<0.0001|
|Proportion of Type I fibers (%)||30 (22,37)||52 (41,62)||<0.0001|
|Proportion of Type I/IIa fibers (%)||3 (1,7)||2 (0,6)||0.04|
|Proportion of Type IIa fibers (%)||60 (53,68)||41 (35,48)||<0.0001|
|Proportion of Type IIx fibers (%)||4 (1,9)||1 (0,4)||0.005|
|CSA of Type I fibers (μm2)||5133 (4054,6114)||5510 (4618,6178)||0.15|
|CSA of Type I/IIa fibers (μm2)||5160 (3855,6048)||5437 (4536,6595)||0.31|
|CSA of Type IIa fibers (μm2)||4111 (3130,4840)||4255 (3028,5584)||0.43|
|CSA of Type IIx fibers (μm2)||2952 (1951,3772)||4793 (3808,6811)||<0.0001|
Twenty-four percent of our cohort had histologically normal muscle. Isolated FS was evident in 31% of patients, while 20% had isolated FA (predominantly type IIx and IIa atrophy). Therefore, only 25% of patients displayed evidence of both processes. Individual data are presented in Figure 1, confirming that a substantial proportion of patients had similar fiber proportions and areas to that observed in the normal population.
Physiological measures of quadriceps strength and endurance were not reliable markers of FA or FS, although some differences were apparent between groups. Patients with isolated FS were younger than patients with normal muscle (P = 0.007), and there were more women with FA than with normal muscle (P = 0.001). There was a higher proportion of patients on long-term oral prednisolone with isolated FS than isolated FA (P = 0.01), but the proportion of current smokers or smoking pack-year history was not different between COPD patients with different muscle phenotypes. The main physiological characteristics of each subgroup are shown in Table 2.
|Healthy controls (n = 30)||Normal muscle (n = 27)||Fiber shift (n = 35)||Atrophy (n = 23)||Fiber shift and atrophy (n = 29)|
|Age (years)||67 (63,72)||69 (64,75)||64 (59,68)b||63 (59,73)||67 (61,73)|
|Gender (% men)||53||89||66a||44b||66a|
|FEV1 (% pred)||109 (100,113)||37 (24,56)||32 (25,51)||43 (29,67)||46 (28,58)|
|TLCO (% pred)||87 (80,96)||49 (33,58)||33 (25,56)||47 (39,52)||34 (26,52)a|
|PaO2 (kPa)||11.1 (10.1,11.7)||9.4 (8.5,10.8)||9.3 (8.3,10.4)||9.5 (8.5,10.3)||8.9 (8.2,10.1)|
|BMI (kg/m2)||24.9 (23.8,28.3)||24.2 (21.4,26.4)||23.6 (20.8,29.5)||24.5 (22.0,26.3)||23.7 (20.7,26.9)|
|% with reduced FFMI||13||33||54||52||48|
|MTCSA (cm2)||110 (106,144)||77 (72,116) (n = 11)||94 (73,120) (n = 15)||97 (n = 1)||90 (73,128) (n = 4)|
|Qs MVC (kg)||34 (28,43)||30 (24,42)||28 (19,38)||27 (19,34)||28 (21,37)|
|Qs MVC (% pred)||82 (66,91)||71 (62,93)||66 (53,77)||73 (52,80)||70 (58,87)|
|TwQ (kg)||8.5 (6.9,10.9)||7.2 (5.9,9.2)||7.3 (6.2,9.2)||7.7 (5.7,8.6)||8.3 (6.9,9.9)|
|T80, (s)||110 (81,184)||88 (71,109)||75 (65,94)||83 (65,118)||80 (63,90)|
|Walking time (min)||95 (59,138)||38 (27,70)||38 (19,58)||62 (31,68)||47 (22,66)|
|Movement intensity (Mi, m/s2)||2.28 (1.73,2.79)||1.65 (1.41,2.05)||1.68 (1.32,1.82)||1.67 (1.46,1.83)||1.78 (1.54,1.97)|
|6MW (m)||600 (86)||381 (303,513)||351 (240,446)a||480 (411,508)c||396 (303,473)d|
|6MW (% pred)||121 (14)||77 (63,95)||66 (46,82)a||91 (70,103)c||79 (63,89)c|
|Peak VO2 (ml/kg/min)||23.0 (17.2,27.0)||12.6 (10.9,15.9)||10.4 (8.30,14.8)a||12.4 (10.6,14.2)||10.2 (9.10,14.0)a|
|Peak VO2 (% pred)||97 (79,106)||50 (38,58)||41 (28,48)b||49 (39,62)c||45 (34,58)|
The relationships between FS and FA and exercise performance are shown in Figure 2. Patients with FS had impaired exercise performance (both absolute and as % predicted for height, weight, gender, and FEV1) compared both with patients with normal muscle histology (e.g., P = 0.008 and 0.025, respectively for peak VO2 and 6MW, both as % predicted) and those with combined FS and atrophy (P = 0.01 for 6MW % predicted) as shown in Table 2. Patients with isolated FA did not have reduced exercise tolerance compared with patients with normal muscle, and, therefore, they had significantly greater exercise capacity than patients with isolated FS (P = 0.001 and 0.006, respectively, for 6MW and peak VO2, both as % predicted). Patients with combined muscle FA and FS did not have impaired exercise capacity compared with patients with normal muscle (P = 0.53 and 0.33, respectively for 6MW and peak VO2 both as % predicted). Therefore, isolated FS was associated with reduced exercise tolerance, whereas FA appeared to be associated with preserved exercise tolerance, even on a background of FS (see Table 2). There was no significant difference in proportion of patients who cited leg discomfort as a cause for stopping exercise in patients with normal quadriceps fiber characteristics (28%) and patients with abnormal fiber characteristics (shift (31%), atrophy (13%), or both (27%)).
There were only weak correlations between proportion of type I fibers in the quadriceps muscle and diffusion capacity of the lung (r = 0.29, P = 0.002), arterial blood oxygen tensions at rest (r = 0.22, P = 0.02), and locomotion time (r = 0.21, P = 0.04) in the patients. In multiple regression analysis, TLCO % predicted was the only independent predictor of type I fiber proportion (r = 0.31, r2 = 0.09, P = 0.003) and type IIx fiber CSA (r = 0.36, r2 = 0.13, P = 0.001). However, examination of the individual data (Fig. 3) again confirms substantial heterogeneity of these parameters even when patients are split as a function of FS.
Noninvasive measures of muscle mass and function were not good markers of quadriceps fiber CSA and proportions. A low FFMI was not a robust predictor of FA. In fact type I and II fiber CSA were not different statistically between patients with a low FFMI and a normal FFMI [5099(3950,5958)μm2 versus 5204(4108,6307)μm2, P = 0.54, Fig. 4A, 3671(2964,4581)μm2 versus 4103(3300,4815)μm2, P = 0.10, Fig. 4B].
In patients, quadriceps MVC and twitch force were correlated with type II fiber CSA (r = 0.42, P < 0.0001, Fig. 4C and r = 0.29, P = 0.003) but not type I fiber CSA (r = 0.06, P = 0.50 and r = 0.004, P = 0.97). T80, as a marker of quadriceps endurance, was a weak correlate of type I fiber proportion (r = 0.23, P = 0.02, Fig. 4D) and, therefore, inversely correlated with type II fiber proportion (r = −0.28, P = 0.005) in patients. Moreover there was no significant correlation between MTCSA and either type I or type II fiber CSA (r = −0.09, P = 0.62 and r = 0.26, P = 0.16, respectively) in the subset of 31 patients considered alone, or when patients and controls were pooled together (n = 40).
In our COPD cohort, we confirmed a preponderance of type II fibers in the quadriceps muscle. However, while our cohort had a lower muscle mass than controls and manifestations of reduced muscle mass (e.g., weakness) we did not find evidence of FA except in the small number of type IIx fibers. Only weak relationships existed with phenotypic measurements of either lung or muscle function indicating that prediction of the biopsy appearances without undertaking a biopsy is impractical. Lastly, we note that 24% of patients had entirely normal fiber proportions and size even though the range of FEV1 of this group indicated that they had severe, mostly GOLD III, disease.
Severity of airflow obstruction was not a predictor of FS, in contrast to our meta-analysis that included only severe and very severe COPD patients. It may be that 114 subjects were insufficient to show this association, although it was sufficient to demonstrate a statistically but not clinically significant association for TLCO. In any event, as shown in Figure 3, this was insufficient at a practical level to distinguish patients with FS from those who did not. Our finding once again confirms, this time at a biopsy level, that COPD-related muscle dysfunction is not confined to those with the worst lung function.[11, 12, 39] Indeed FEV1 was not an independent predictor of FS. The absence of an association between physical activity and FS was also unexpected, because it is a distinguishing factor between the inactive quadriceps and continuously active diaphragm muscles that characteristically do and do not exhibit type I to type II FS, respectively in COPD.[40, 41] However, it is consistent with the observation that exercise training as part of pulmonary rehabilitation does not increase quadriceps type I fiber proportion in COPD despite a modest shift in type II fiber sub-types.
Our observations have significant implications for the design of early phase clinical trials; without a biopsy even a carefully characterized cohort of patients attending a specialist hospital will contain on average only 50% of patients who have FS or FA. This might not matter if a new medicine were exceptionally powerful, but one might expect intuitively only modest clinical benefit would be obtained by treating a patient who was already within the normal range. Thus early phase trials, which by their nature tend to be small, could run the risk of generating a false negative result if by chance patients with an inappropriate (for the postulated mode of action of a new medicine) phenotype are included. This risk could be mitigated in several ways. First a biopsy could be taken before study entry; however the delay and cost of analysis might make running the study difficult, although the use of minimally invasive biopsy techniques might make this more acceptable to patients. Second, biopsies could be taken before and after a study with an a priori provision for subset analysis of patients with a biopsy appearance deemed most likely to benefit based on a novel medicine's mode of action. Third, reliable biomarkers could be developed to address phenotype. P NMR spectroscopy might be such a candidate; another could be microRNA-1 (miR-1) which is related to fiber type in COPD and which can be measured in blood.
The finding that patients with FS had the poorest exercise performance of all the COPD subgroups, including those patients with both FS and FA, was interesting. One explanation may be that the slightly poorer lung function in this group compared with the others explains the worse exercise performance. However, we doubt this explanation, because over this range of FEV1 the relationship between 6MW distance and FEV1 is poor and the differences in FEV1 seen were not statistically significant in this cohort. Moreover, FS was inversely associated with exercise performance independent of lung function, as assessed by regression analysis (not shown). In contrast, FA was weakly but positively associated with 6MW distance % predicted and peak VO2 % predicted. The reason for the latter finding is unclear; we speculate that mild FA may be adaptive by increasing relative capillary density and improving local oxygen delivery. Alternatively, reduced muscle protein synthesis leading to FA means less demand on the scarce ATP resources in the muscle when patients exercise.
The main strength of the study, by design, was its size, making a type II error unlikely regarding our negative results and the comprehensive and rigorous clinical phenotyping and fiber typing. Analysis of 2 biopsies for many subjects reduced the effect of sampling bias, a precaution not previously used in clinical studies. It is noteworthy that our study failed to replicate neither the conclusions drawn from a meta-analysis of smaller studies addressing the same topic nor prior reports of FA, except in the case of type IIx. We suspect this represents a publication bias so that, especially in an emerging discipline, studies with statistically significant “positive” findings are more likely to be published than those with negative findings this phenomenon is well recognized in other fields of medicine.[50, 51]
Although the heterogeneity of FS and FA is clearly demonstrated by these data, the cut-offs for distinguishing normal from abnormal rests on data derived from normal subjects (Fig. 1; Table 1). Our normal values for type I fiber proportion are similar to those of our previous meta-analysis and to those of Torres et al., which is the current largest study in the literature, although Torres et al. reported a much higher proportion of type IIx fibers than this study or those of other investigators.
Although age was not an independent predictor of FS, patients with FS were slightly younger than patients without FS. Type I to type II fiber ratio is higher in healthy elderly people than younger subjects. This alone does not explain our finding, as patients with FS and controls were a similar age (Fig. 3D).
First, this was a cross-sectional study, and, therefore, changes in muscle morphology over time with progression of disease could not be demonstrated. Second, the COPD cohort was drawn from patients referred for hospital care and thus may not be typical of general practice, although care was taken not to study only those with severe lung function impairment. Third, there are some aspects of patient phenotype (for example oxygenation during sleep) and muscle phenotype (capillarization, markers of atrophy/hypertrophy signalling, or oxidative enzyme content) not measured. Fourth, we did not assess contractile fatigue in response to exercise. We did, however, use the effort-independent technique of repetitive magnetic stimulation that we believe to be more specifically reflective of the fiber type characteristics of the muscle. Last, genetic factors that could impact on quadriceps fiber characteristics were not measured here; genetic factors are known to account for approximately 45% of the variance in human skeletal muscle type I fiber proportion and also influence muscle mass.[53, 54] For example, angiotensin-converting enzyme and Vitamin D polymorphisms influence strength in COPD.[55-57]
In conclusion, when considered as a group, there was considerable heterogeneity in fiber type proportions and FS, with 24% of the patients having normal fiber characteristics. Importantly, FA and FS frequently did not co-exist, and FEV1 was not an independent predictor of muscle phenotype. Stratification of patients for trials of drugs aiming to change either FA or FS cannot presently be reliably done without a biopsy; trial design for novel medicines needs to address this.
This project was supported by the NIHR Respiratory Biomedical Research Unit at the Royal Brompton Hospital and Imperial College, London UK. We thank Dr. Winston Banya, in the Statistics Department of the Royal Brompton Hospital, for his help with the statistical analysis, Derek Cramer, Mark Unstead, and the lung function department at the Royal Brompton Hospital for performing the pulmonary function tests, Professor David Hansell and the CT department at the Royal Brompton Hospital for performing the thigh scans, and Kathleen Daenen at the University Hospital Maastricht for the cryosectioning. Samantha A. Natanek carried out the physiological testing, collection of muscle biopsies, the histological assays, the statistical analysis, with MIP conceived the idea for the manuscript, and drafted the manuscript. Harry R. Gosker gave assistance with the histological assays, statistical testing, interpretation of the data, and manuscript writing. Ilse G.M.S. Slot assisted with the histological assays and interpretation of the data. Gemma S. Marsh assisted SAN with collection of the muscle biopsies and physiological testing and interpretation of the data. Nicholas S. Hopkinson helped with physiological testing, data interpretation and manuscript writing. William D-C Man gave assistance with interpretation of the data and manuscript writing. Ruth Tal-Singer contributed to conception of the idea, helped with data interpretation and gave assistance with manuscript writing. John Moxham contributed to conception of the idea, gave assistance with interpretation of the data, and manuscript writing. Paul R. Kemp gave assistance with interpretation of the data and manuscript writing. Annemie M.W.J. Schols supervised collection of the laboratory analysis, helped with data interpretation, conception of the idea, and manuscript writing. Michael I. Polkey supervised collection of the physiological data, helped with data interpretation, conception of the idea and manuscript writing. Dr. Natanek (née Sathyapala) was funded by a Wellcome Clinical Fellowship (079686), Dr. Gosker by a Netherlands Asthma Foundation grant (NAF 3.4.09.068), and W.D-C. Man by a National Institute for Health Research Clinician Scientist Award (CS/7/2007) and a Medical Research Council New Investigator Research Grant (G1002113). Professor Polkey's salary is part-funded by the NIHR Respiratory Biomedical Research Unit at the Royal Brompton Hospital and Imperial College. Ruth Tal-Singer is an employee and shareholder of GSK; GSK funded Ms Marsh' salary. Drs. Schols and Polkey contributed equally to this work.