Determinants of exercise capacity in cystic fibrosis patients with mild-to-moderate lung disease

A total of 102 adult patients (sex ratio M:F 0.52) with CF were enrolled at four CF centers in France: Lille (75 patients), Rouen (15 patients), Dijon (5 patients), and Grenoble (7 patients). Written informed consent for participation in the study was obtained from participants. Use of the patient data was approved by the local ethics committee, and the study was considered observational and approved as such by the Institutional Review Board of the French Learned Society for Pulmonology (Société de Pneumologie de Langue Française, CEPRO 2012 009).

Clinical, nutritional, biological, PFT, and CPET data were obtained on the same day, either at diagnosis or at the routine annual evaluation, and were retrospectively collected. When patients were seen at several follow-up visits, only the data from the first visit were recorded. Hypoxemic patients did not perform CPET and were excluded from the analysis. A diagnosis of CF was obtained by sweat chloride > 60 mmol/L and the presence of CFTR gene mutations by molecular analysis. Additional characteristics recorded at the time of testing were diseases usually associated with CF, bacterial colonization, treatments, and nutritional status, including height and weight measurements and impedance analysis.

Forced vital capacity (FVC), forced expiratory volume in 1 s (FEV₁), FEV₁ to FVC ratio, total lung capacity (TLC), and residual volume (RV) were measured by plethysmography (Jaeger-Masterlab® cabin). Diffusing capacity of the lung for carbon monoxide (DLCO: mL CO/min/mm Hg) was adjusted for hemoglobin concentration in g/dL according to Cotes’ equation: corrected (Hb) DLCO = DLCO × (10.2 + Hb)/(1.7 × Hb). Following ATS/ERS 2005 guidelines, the lower limits of normal were set at the 5th percentile (or predicted minus 1.64 SD) of each reference population. The results are expressed as percentages of the predicted values. Predicted normal values were derived from standard equations [20‐22].

PFTs and CPET were performed in an air-conditioned laboratory (22°C constant temperature), using a standardized protocol as previously described [23, 24]. The CPET protocol was the same at each center. Each patient underwent a symptom-limited incremental exercise test on an ergometric bicycle (Ergoline-Ergometrics 800®). The protocol included a warm-up period of 3 min at 20 W followed by a progressively increasing work rate (WR) in a ramp fashion and then 3 min recovery. The ramped WR increment was individualized (range, 8–30 W/min). During exercise, heart rate (HR) was monitored continuously by 12-lead ECG, and arterial oxygen saturation (SpO₂) was measured by pulse oximetry (Nellcor N-395). The expired gases were analyzed with an Ergocard®, focusing on oxygen consumption (VO₂), carbon dioxide production (VCO₂), minute ventilation (VE), and tidal volume (VT). The oxygen pulse (VO₂/HR) was calculated. Measurements of PaO₂ and PaCO₂ were performed on room air at rest and at peak exercise. Normal values for PaO₂ were derived from [25]. Lactatemia was determined at maximal exercise. Breathing reserve (BR) was calculated as BR = (predicted maximum minute ventilation [MMV] – V_E peak)/MMV, with MMV estimated from MMV = FEV₁ × 40. HR peak was expressed as a percentage of maximum predicted HR, calculated as HR max = 210 – (0.65 × age). Dead space (V_D/V_T) was calculated according to Bohr’s equation corrected for the additional instrument dead space: V_D/V_T = (PaCO₂ – PECO₂ mean)/PaCO₂ – (V_D [machine]/V_T) where PECO₂ is the partial pressure of expired CO₂. Predicted values for VO₂ max were calculated from reference equations [26]. Poor motivation appeared not to be an interfering factor in our analysis, as all patients had at least one of the following: BR < 15%, peak HR > 90% of predicted, peak lactate > 7 mmol/L, peak exercise PaO₂ < 55 mm Hg, and peak V_E/VO₂ > 35 or peak RER > 1.15 [23]. Immediately after exercise, subjects were asked to score their sense of breathlessness and muscle fatigue at peak exercise using Borg scales.

The continuous variables are reported as mean ± SD. Normal distribution of quantitative variables was tested by the Shapiro–Wilk test. Differences in FEV₁ between the groups were determined with the Student’s t-test or Mann–Whitney test. Bivariate analyses were performed to study the relationships between each explanatory variable and the VO2 peak. Pearson’s or Spearman’s correlation coefficient was used for quantitative variables, and the Student’s t-test or Mann–Whitney test for qualitative variables. In addition, a multivariable linear regression with a stepwise selection at the level 0.2 was performed to identify a subset of the most important explanatory variables for the relationship with VO2 peak. In order to avoid the problem of multicolinearity which happens when the explanatory variables are highly correlated, and to obtain a parsimonious model, we adopt the following strategy: first, variables with p < 0.2 were selected and included in a Principal Component Analysis (PCA) in order to study their correlations. Then, the variables included in the multivariable regression model were selected by the results of PCA (graphic correlation circle) on the basis of their clinical pertinence. The stability of the model was assessed by a bootstrap method [27]. The bootstrap resampling method was based on 1000 replicates of the initial dataset. Multivariable regression with a stepwise selection at the level 0.2 was performed on each of these replicates. The inclusion of the variable in the final model was confirmed if this candidate variable was selected in at least 70% of these 1000 analyses. In the final model, for each variable we computed the partial R-square, coefficient, 95% confidence intervals and adjusted p-value. Final variables from the multivariate analysis were applied to each group of FEV₁. All analyses were achieved with SAS software version 9.2 (SAS Institute Inc., Cary, NC). All tests were performed at the significant level at 0.05.

The demographic and clinical characteristics are shown in Table 1, and the resting PFT results are presented in Table 2. The cohort consisted of 102 CF patients with a mean age of 28 ± 11 years (range 17–67). The time from diagnosis to evaluation was 16 ± 10 years. For data analysis, the cohort was divided into two groups according to their FEV₁: group 1 patients with severe lung disease (FEV₁ < 50%, 48 patients) and group 2 patients with mild-to-moderate lung disease (FEV₁ ≥ 50%, 54 patients). Group 1 had a significantly higher frequency of homozygosity for the CFTR ∆F 508 mutation, pancreatic insufficiency, bronchial colonization with Pseudomonas aeruginosa, and biological inflammatory syndrome, and significantly lower body mass index (BMI) and longer disease duration (data not shown).

The patients had a range of disease severities and were recruited from four centers in France. Patients from different centers all had characteristics consistent with the French CF Registry 2009 Report [28] and showed equivalent frequencies of key CF characteristics (∆F508 mutation, exocrine pancreatic insufficiency, and colonization with P. aeruginosa), supporting the reproducibility of the results and the potential for extrapolation to other patient populations.

Resting PFT values (Table 2) were significantly more altered in group 1 than in group 2 (Table 2). As expected, three major functional abnormalities were found: obstructive syndrome (FEV₁/FVC = 65 ± 15% of predicted), altered distension (RV = 176 ± 65% of predicted), and altered DLCO (68 ± 18% of predicted).

Exercise cessation was mainly due to leg fatigue in combination with dyspnea (62%), whereas leg fatigue alone or dyspnea alone was observed in 17% and 21% of patients, respectively. The VO₂ peak value (weight-adjusted VO₂) was decreased to < 84% of predicted in 83/102 (82%) of patients (48/48 [100%] in group 1 and 35/54 [65%] in group 2) and was significantly lower in group 1 than in group 2 (Table 3).

Analysis of the ventilatory response (V_E peak, BR, respiratory rate (RR), V_T/FVC peak) highlighted the differences according to FEV₁ impairment (Table 3). Group 1 had a lower absolute value of V_E at peak exercise, and a depletion of BR. Hyperventilation was due to simultaneous increases in RR and tidal volume. Impairment in pulmonary gas exchange was more severe in group 1, as shown by higher values of P(A-a)O₂, V_D/V_T peak, and PaCO₂, and lower values of PaO₂. Cardiocirculatory responses were normal in group 2, but patients in group 1 showed low VO₂/HR values and a significant decrease in peak HR. Four patients experienced ECG abnormalities but continued with the exercise test.

Significant correlations were observed between VO₂ peak and nutritional status (BMI, lean body mass), inflammation markers (C-reactive protein [CRP], leukocytosis), resting PFT (FVC, FEV₁, RV, DLCO, P(A-a)O₂), and quantifiable parameters of CPET (V_E peak, V_E/VO₂ peak, V_E/VCO₂ peak, BR, V_D/V_T peak, PaO₂ peak, P(A-a)O₂ peak, and HR (Table 4 and Figure 1).

The results of the stepwise multiple regression analysis for determinants of exercise capacity are shown in Table 5. Of the variables entered into the model (BMI, FEV₁, FVC, DLCO, PaO₂, V_E/VCO₂ peak, BR, V_D/V_T peak, PaO₂ peak, PaCO₂ peak, P(A-a)O₂ peak, and lactatemia peak), only FEV_1, V_E/VCO₂ peak, and BR were found to be independent predictors of exercise capacity (r² = 0.67). Analysis of these three variables showed that, for group 1, 31% of the VO₂ peak was explained by FEV₁, whereas the major determinants of the VO₂ peak in group 2 were BR , FEV₁ and V_E/VCO₂ peak (Table 5).

Separate analysis in the cohort of Lille (75 out of the 102 patients) showed the same results: FEV₁, BR and V_E/CO₂ were independent predictors of exercise capacity (r² = 0.65) (data not shown).