The Prognostic Value of Cardiopulmonary Exercise Testing in Idiopathic Pulmonary Fibrosis

Abstract

Rationale: Idiopathic pulmonary fibrosis (IPF) is characterized by progressive dyspnea, impaired gas exchange, and ultimate mortality.

Objectives: To test the hypothesis that maximal oxygen uptake during cardiopulmonary exercise testing at baseline and with short-term longitudinal measures would predict mortality in patients with idiopathic pulmonary fibrosis.

Methods: Data from 117 patients with IPF and longitudinal cardiopulmonary exercise tests were examined retrospectively. Survival was calculated from the date of the first cardiopulmonary exercise test.

Measurements and Main Results: Patients with baseline maximal oxygen uptake less than 8.3 ml/kg/min had an increased risk of death (n = 8; hazard ratio, 3.24; 95% confidence interval, 1.10–9.56; P = 0.03) after adjusting for age, gender, smoking status, baseline forced vital capacity, and baseline diffusion capacity for carbon monoxide. We were unable to define a unit change in maximal oxygen uptake that predicted survival in our cohort.

Conclusions: We conclude that a threshold maximal oxygen uptake of 8.3 ml/kg/min during cardiopulmonary exercise testing at baseline adds prognostic information for patients with IPF.

AT A GLANCE COMMENTARY

Scientific Knowledge on the Subject

Prognosis in idiopathic pulmonary fibrosis (IPF) is poor. Predicting prognosis by examining exercise performance with the six-minute-walk test has led to conflicting results.

What This Study Adds to the Field

This study shows that o₂max predicts mortality in patients with IPF. Patients with o₂max less than 8.3 ml/kg/min at baseline had an increased risk of death.

Idiopathic pulmonary fibrosis (IPF) is a disease of unknown etiology characterized by progressive dyspnea and ultimate mortality (1). Mean survival from time of diagnosis to death is 3 years (1). However, the disease course is variable: some patients progress rapidly and others remain stable for many years (2). There is no effective treatment and many patients, if eligible, are referred for lung transplantation. Identification of surrogate short-term measures of mortality is critical to the management and study of patients with IPF.

Several factors have been identified that predict poor survival in patients with IPF, including age, sex, smoking history, diffusion capacity for carbon monoxide (Dl_CO), FVC, degree of fibrosis on high-resolution computerized tomography of the chest, and number of fibroblastic foci on histopathology (3–11). Longitudinal changes in FVC or Dl_CO have been found to have important prognostic value. A decrease in FVC of at least 10% or Dl_CO of at least 15% over 6 or 12 months is associated with decreased survival (10–15).

Gas exchange worsens with exercise in IPF (1, 16, 17). Several studies have examined this feature using either cardiopulmonary exercise tests (CPET) or the 6-minute-walk test (6MWT). A decrease in Pa_O2 during CPET in patients with IPF contributes up to 10.5% of the total clinical, radiographic, and physiologic (CRP) score (5) used to estimate prognosis in IPF. Desaturation during CPET has been shown to predict mortality in some (16) but not all (18, 19) studies. Desaturation below 88% during 6MWT is a more consistent marker of increased risk for mortality (7, 9, 10), whereas shorter walk distance is less predictive (15). Longitudinal change in 6MWT data predicts mortality in patients who do not desaturate less than 88% at baseline (15).

Patients with IPF have impaired ventilatory and cardiovascular responses to exercise (20) due to multiple abnormalities, including low tidal volume; a failure to decrease ventilatory dead space; a rapid, shallow breathing pattern; impaired gas exchange due to interstitial fibrosis; pulmonary hypertension; ventilation/perfusion mismatching; and low mixed venous O₂.

o₂max is an integrated measure of cardiovascular, respiratory, and neuromuscular function (21). In prior studies of patients with interstitial lung disease, o₂max correlated poorly with measures of lung volume, suggesting that it more accurately reflects derangements in hemodynamics as well as ventilation during exercise (20). Although change in FVC is a good surrogate for subsequent mortality, it is imperfect as some patients die without a 10% decline in FVC, whereas others can live for prolonged periods even after a 10% decline in FVC (2, 22). Therefore we chose to examine longitudinal change in o₂max a priori because it is an integrated measure of cardiovascular, respiratory, and neuromuscular function (21). We tested the hypothesis that a decrease in o₂max during baseline and short-term longitudinal CPETs predicts mortality in patients with IPF.

METHODS

Patient Selection

This is a retrospective analysis of 117 patients in the University of Michigan Specialized Center of in the Pathobiology of Fibrotic Lung Disease Database. Patients in this database were referred for enrollment in study protocols for suspected IPF based on typical symptoms, physiologic findings, and radiographic findings (1). Patients with a high-resolution computerized tomography scan showing a definite pattern of usual interstitial pneumonitis (23) were not required to undergo surgical lung biopsy (n = 42) (24, 25). Patients with underlying connective tissue disease, occupational or environmental exposures, or histopathologic pattern on surgical lung biopsy other than usual interstitial pneumonitis were excluded. Patients were treated with varied regimens, including no therapy, immunosuppression (prednisone ± azathioprine or cyclophosphamide), colchicine, and experimental protocols. The lack of a standardized treatment regimen prevented an analysis of the data based on therapy. Approval for the use of these data was provided by our Institutional Review Board. Subgroups of these patients have been previously described (7, 15, 26).

Pulmonary Function, 6-Minute Walk, and Cardiopulmonary Exercise Tests

Pulmonary function and exercise tests, including FVC, Dl_CO, 6MWT, and CPET, were performed as described (7, 18). Desaturation during a 6MWT was defined a priori as less than 88% based on published data (7).

Statistical Analysis

The date of the first CPET was used as the start date for survival analysis. Death date was supplemented by searching the Social Security Death Index (27); patients not listed in this index were censored 3 months prior to the analysis date to account for potential lags in reporting. Multivariate Cox proportional hazard models (28) adjusted for age, sex, baseline forced vital capacity percent predicted, baseline Dl_CO percent predicted, and smoking status were constructed for potential o₂max thresholds ranging from 4.4 ml/kg/min to 25.1 ml/kg/min, in increments of 0.3 ml/kg/min. Resulting hazard ratios were plotted against o₂max to determine if a threshold o₂max could be identified that correlated with increased risk of mortality. Similar secondary analyses were performed on resting Pa_O2. Baseline characteristics between patients with o₂max above and below the thresholds were compared using t tests (29) for continuous measures and chi-square (30) tests for categorical measures. Data are expressed as mean ± standard deviation (SD) or frequency (%). In secondary analyses, index of concordance (31) was used compare the o₂max threshold, desaturation less than 88% during a 6MWT, and resting Pa_O2 to determine which was the strongest predictor of survival. Survival between patients with baseline o₂max above and below thresholds was examined with unadjusted Kaplan Meier survival curves (32) and log-rank tests (33, 34). Multivariate Cox proportional hazard models studied the predictive value of the o₂max threshold adjusted for age, gender, smoking status, baseline FVC percent predicted (FVC%), and baseline Dl_CO percent predicted (Dl_CO%). Statistical significance was set at P = 0.05. Statistical analysis was performed with R software (http://www.r-project.org/index.html) and SPSS (version 14.0; SPSS Inc., Chicago, IL).

RESULTS

Data from 117 patients with at least two cardiopulmonary exercise tests were analyzed. Seventy-five (64%) patients were diagnosed with a surgical lung biopsy. Patients who did not undergo a lung biopsy were significantly older (66.73 ± 8.50 vs. 63.07 ± 8.23 y, P = 0.024) but otherwise not different than those with a lung biopsy (data not shown).

We explored the predictive value of o₂max on survival. o₂max did not predict survival when examined as a continuous variable (hazard ratio [HR], 0.969; 95% confidence interval [CI], 0.88–1.07; P = 0.55). Exploratory analyses revealed a threshold o₂max of 8.3 ml/min/kg that was associated with an increased risk of subsequent mortality (Figure 1). This threshold effectively captures patients with a higher risk of mortality and discriminates these patients from those with a lower risk of mortality.

Figure 1.

Determination of a o₂max threshold of 8.3 ml/kg/min. Multivariate Cox proportional hazard models adjusted for age, sex, baseline FVC percent predicted, baseline diffusion capacity of carbon monoxide percent predicted, and smoking status were constructed for potential o₂max thresholds ranging from 4.4 ml/kg/min to 25.1 ml/kg/min, in increments of 0.3 ml/kg/min. The vertical axis gives hazard ratios comparing risk of patients above and below the corresponding threshold along the horizontal axis. A threshold baseline o₂max of 8.3 ml/min/kg was determined (dashed line).

Demographic data and baseline physiologic data for patients with an initial o₂max above or below 8.3 ml/min/kg are presented in Table 1. Patients whose baseline o₂max was below threshold were more often female (P = 0.04), had significantly lower FVC% (P = 0.01) and Dl_CO% (P = 0.006), shorter 6-minute walk distance (P = 0.002), and significantly lower exercise capacity during baseline CPET compared with those with baseline o₂max above threshold (Table 1). Surprisingly, not all patients with o₂max below threshold at baseline desaturated during a baseline 6MWT.

TABLE 1.

BASELINE CHARACTERISTICS OF PATIENTS WITH BASELINE o₂max ABOVE OR BELOW THE 8.3 ml/kg/min THRESHOLD

	Above Threshold	Below Threshold	P Value
n	109	8
Age, yr	64.0 ± 8.4	69.1 ± 8.6	0.10
Sex			0.044
Male	78 (78.8%)	3 (37.5%)
Female	31 (31.3%)	5 (62.5%)
Diagnosed with lung biopsy	70 (70.7%)	5 (62.5%)	0.92
Smokers			0.30
Never smokers	28 (28.3%)	4 (50%)
Former smokers	76 (76.7%)	4 (50%)
Current smokers	5 (5.0%)	0
Pack-years	39.2 ± 24.9	42.0 ± 44.4	0.84
Spirometry at baseline
FVC, L	2.8 ± 0.9	1.8 ± 0.6	0.002
FVC%	69.0 ± 17.8	53.0 ± 9.6	0.01
DlCO, ml/min/mm Hg	12.3 ± 4.4	7.1 ± 1.3	0.002
DlCO%	47.8 ± 14.9	31.9 ± 7.7	0.006
Exercise at baseline
Exercise PaO2	67.4 ± 15.0	59.4 ± 12.6	0.25
Exercise Aa gradient	41.0 ± 14.7	42.2 ± 13.3	0.85
Exercise saturation	89.5 ± 4.9	88.0 ± 5.7	0.41
Minutes of exercise	7.4 ± 1.9	4.2 ± 1.5	<0.001
Work, W	95.2 ± 35.0	40.0 ± 11.6	<0.001
o2max, ml/kg/min	14.0 ± 4.2	6.9 ± 1.4	<0.001
Exercise gas exchange score	12.4 ± 9.8	20.0 ± 14.2	0.04
6MWT at baseline
Desaturation < 88%	38 (38.4%)	5 (62.5%)	0.12
Distance, ft	1,116.6 ± 456.0	582.2 ± 386.1	0.002

Definition of abbreviations: Aa gradient = alveolar-arterial gradient; Dl_CO% = diffusion capacity for carbon monoxide percent predicted; FVC% = forced vital capacity percent predicted; o₂max = maximal oxygen consumption during cardiopulmonary exercise testing; 6MWT = 6-min-walk test.

Data are shown as mean ± standard deviation (SD) or frequency (%).

Multivariate relationships between o₂max and baseline demographic and pulmonary function variables were explored (Table 2). In multivariate linear regression models, age, male gender, history of smoking, and baseline FVC were predictors of o₂max. A linear predictor incorporating these variables significantly predicted whether a patient's baseline o₂max would be below threshold (Table 3). With this model, patients who are younger, male, never smokers, with higher FVC% and Dl_CO% are more likely to have a o₂max above 8.3 ml/min/kg, and thus are predicted to have overall improved survival than older, female, ever smokers with lower FVC% and Dl_CO%.

TABLE 2.

MULTIVARIATE LINEAR REGRESSION MODEL DETECTING ASSOCIATIONS WITH o₂max

	β	SE	P Value
Age, yr	−0.23	0.04	<0.001
Male sex	2.53	0.79	0.002
Ever smoker	−1.61	0.72	0.028
Baseline FVC%	0.82	0.28	0.004
Baseline DlCO%	0.30	0.31	0.332

Definition of abbreviations: Dl_CO% = diffusion capacity for carbon monoxide percent predicted; FVC% = forced vital capacity percent predicted; SE = standard error; o₂max = maximal oxygen consumption during cardiopulmonary exercise testing.

n = 103.

TABLE 3.

LOGISTIC REGRESSION MODEL OF PREDICTORS OF o₂max < 8.3 ml/kg/min

	OR	95% CI	P Value
Linear predictor	0.57	0.37–0.87	0.01

Linear predictor = −0.232(age) + 2.53(male gender) − 1.6(ever smoker) + 0.82(FVC%) + 0.30(Dl_CO%).

Definition of abbreviations: CI = confidence interval; Dl_CO% = baseline diffusion capacity for carbon monoxide percent predicted; FVC% = baseline forced vital capacity percent predicted; OR = odds ratio; o₂max = maximal oxygen consumption during cardiopulmonary exercise testing.

n = 103.

Survival in patients with baseline o₂max below the 8.3 ml/kg/min threshold was lower than that of patients above the threshold over time (log rank P < 0.001, Figure 2). This difference was maintained in multivariate Cox proportional hazard survival models adjusting for patient age, smoking status, male sex, baseline FVC%, and baseline Dl_CO% (HR for being below threshold, 3.24; 95% CI, 1.10–9.56; P = 0.03) (Figure 3 and Table 4).

Figure 2.

Unadjusted Kaplan-Meier survival curves for patients whose baseline o₂max was greater than 8.3 ml/min/kg (Above Threshold, n = 99) or less than 8.3 ml/min/kg (Below Threshold, n = 8). Survival time was calculated from the date of the first cardiopulmonary exercise test.

Figure 3.

Cox proportional hazard survival curves for patients with baseline o₂max greater than 8.3 ml/kg/min (solid line) and less than 8.3 ml/kg/min (dashed line) adjusted for age, sex, smoking status, baseline FVC percent predicted, and baseline Dl_CO percent predicted. Survival time was calculated from the date of the first cardiopulmonary exercise test.

TABLE 4.

MULTIVARIATE COX PROPORTIONAL HAZARD SURVIVAL MODEL ASSESSING THE PREDICTIVE VALUE OF o₂max < 8.3 ml/min/kg

	HR	95%CI	P Value
o2max below threshold	3.24	1.10–9.56	0.03
Age, yr	1.04	1.00–1.08	0.08
Male sex	1.29	0.62–2.69	0.50
Ever smoker	1.33	0.71–2.48	0.37
Baseline FVC%	0.93	0.72–1.21	0.58
Baseline DlCO%	0.66	0.49–0.90	0.01

Definition of abbreviations: CI = confidence interval; Dl_CO% = diffusion capacity for carbon monoxide percent predicted; FVC% = forced vital capacity percent predicted; HR = hazard ratio; o₂max = maximal oxygen consumption during cardiopulmonary exercise testing.

n = 103.

In the majority of patients studied, o₂max did not change between baseline and 6 months (n = 99). Of the 109 patients with o₂max above threshold at baseline, 5 had a further decline in o₂max below the threshold with longitudinal measures. However, we were unable to define a unit change in o₂max that predicted survival in our cohort (data not shown).

Prior studies have shown that resting Pa_O2 predicts survival in IPF. In secondary analyses, resting Pa_O2 was a significant predictor of mortality (HR, 0.934; 95% CI, 0.88–0.99; P = 0.02), when adjusted for age, sex, smoking status, and baseline FVC% and Dl_CO%. A clear threshold for resting Pa_O2 could not be determined (data not shown). An adjusted multivariate Cox model with the o₂max threshold (HR, 2.66; 95% CI, 0.74–9.5; P = 0.13) and resting Pa_O2 (HR, 0.95; 95% CI, 0.89–1.00; P = 0.09) did not show significance of either predictor. When the o₂max threshold, resting Pa_O2, and desaturation below 88% during a 6MWT were included in an adjusted Cox model, the o₂max threshold was a significant predictor of mortality (HR, 3.48; 95% CI, 1.16–10.43; P = 0.03), whereas desaturation less than 88% was not (HR, 1.49; 95% CI, 0.617–3.62; P = 0.37). Index of concordance analysis demonstrated that the o₂max threshold is a more robust predictor of survival than resting Pa_O2 or desaturation less than 88% during a 6MWT (Table 5).

TABLE 5.

INDEX OF CONCORDANCE ANALYSIS COMPARING THE STRENGTH OF o₂max < 8.3 ml/kg/min, DESATURATION < 88% DURING A 6-MINUTE WALK TEST, AND RESTING Pa_O2 AS PREDICTORS OF SURVIVAL IN IDIOPATHIC PULMONARY FIBROSIS

Model	Ratio	Concordance
o2max < 8.3 ml/kg/min	0.716	2342.5
Desaturation < 88%	0.708	2316.5
Resting PaO2	0.702	2322.7

Definition of abbreviations: Dl_CO% = diffusion capacity for carbon monoxide percent predicted; FVC% = forced vital capacity percent predicted; Pa_O2 = arterial partial pressure of oxygen; o₂max = maximal oxygen consumption during cardiopulmonary exercise testing; 6MWT = 6-min-walk test.

Covariates in each model were: age, male sex, smoking status, FVC%, and Dl_CO%. Number of paired observations: 3,271.

DISCUSSION

In this study, we examined the relationship between maximal oxygen uptake during cardiopulmonary exercise testing and mortality. We hypothesized that o₂max measured during cardiopulmonary exercise testing predicts mortality in patients with IPF. We found that o₂max examined as a continuous variable does not predict mortality in IPF. However, baseline threshold o₂max of 8.3 ml/kg/min predicts mortality in these patients. This threshold is a robust predictor of survival when compared with desaturation less than 88% during a 6MWT and resting Pa_O2. Demographic and pulmonary function data can be used to estimate whether o₂max is above or below the 8.3 ml/kg/min threshold.

We found a threshold baseline o₂max less than 8.3 ml/kg/min predicts mortality in patients with IPF. A similar threshold value has not been reported in fibrotic lung disease. In a study of 86 patients with primary pulmonary hypertension, Wensel and colleagues (35) found a o₂max threshold of 10.4 ml/kg/min predicts mortality. Various measures of exercise capacity have been examined for their ability to predict survival in IPF. A CRP score was developed by Watters and colleagues (36) as a tool to assess and follow patients' clinical impairment from IPF. The score used an exercise gas exchange score, which assigned points based on change in saturation during exercise, change in Vo₂, and predicted o₂max. The maximum points attributable to exercise gas exchange in the CRP score was 30, greater than the radiological, symptoms, and pulmonary function components of the score, reflecting the importance of exercise gas exchange in clinical impairment in IPF. More recently, Miki and colleagues (37) calculated the change in Pa_O2 per change in Vo₂ during CPET (ΔPa_O2/ΔVo₂ or Pa_O2 slope) and found this relationship to predict mortality in patients with IPF. However, not all studies show that CPET measurements of gas exchange predict survival (18, 19). Studies that have used multistep scores (CRP score) or slope calculations (Pa_O2 slope) to define the risk of mortality attributable to exercise gas exchange in IPF may be too cumbersome for routine use outside of clinical trials. Our data suggest that a simple threshold for o₂max of 8.3 ml/kg/min predicts mortality in patients with IPF, without the need for lengthy calculations.

Several authors have examined the prognostic value of the 6MWT or other walk tests in IPF. Desaturation during 6MWT (7, 9), distance walked (9), and progressive impairment in longitudinal 6MWTs (15) have been found to predict mortality in IPF. One criticism of the 6MWT is that it is a patient-driven, symptom- and effort-limited test. This may explain the controversy in the literature about whether distance walked or desaturation is a better predictor of mortality. It may also explain why not all patients in this study with a baseline o₂max below threshold had desaturation during 6MWT: these patients might not have walked sufficiently fast or far enough to produce desaturation.

Using index of concordance techniques, we compared the ability of desaturation less than 88% during a 6MWT and the o₂max threshold to predict mortality in IPF. Despite the small number of patients with o₂max below threshold at baseline, this variable was a stronger predictor than desaturation less than 88% in our cohort in multivariate Cox models. The o₂max threshold is also more robust than resting Pa_O2 and desaturation less than 88% during a 6MWT in concordance analyses.

We were unable to identify a unit change in o₂max per time that predicts survival in short-term follow-up. This may be due to the small number of patients who had a decline in o₂max during the follow-up period. This could suggest that o₂max is more stable over time compared with other measures, such as FVC or Dl_CO. It could also reflect a selection bias in that as patients became sicker they may have declined exercise testing but still been able to perform pulmonary function testing. Further prospectively collected data are needed to explore the changes in o₂max over time and their impact on survival.

There are several limitations to this study. Patients in this study were not evaluated for the presence of pulmonary hypertension. Pulmonary hypertension has been shown to be an important predictor of mortality in IPF (38–40), although its presence does not universally portend a poor outcome (40). Patients included in this study were enrolled in a number of treatment protocols; the varied nature of the protocols prevents an analysis of the data based on therapy. However, none of the therapies in use in these protocols has been found to slow, halt, or reverse pulmonary fibrosis in this population, and prior analyses of the effects of therapy on this population have shown no benefit (7, 18, 41). In our cohort of 117 patients, 8 had a baseline o₂max below threshold and 5 had a decrease in o₂max to below threshold over the course of follow-up, yet 46% of the patients died during follow-up. Other measures that demonstrate significant change over time, such as serial measures of FVC%, may be more sensitive predictors of mortality in this population. Alternately, deaths could have been due to acute exacerbations of IPF or other acute events, which we did not measure.

In this study, we examined the prognostic value of CPET in IPF. A baseline o₂max less than 8.3 ml/kg/min threshold was identified, below which the risk of death was greatly increased. This study provides an easy-to-use threshold for o₂max for patients with IPF that predicts an increased risk of death. An unexpected finding was that not all patients with o₂max below threshold desaturated during a 6MWT. Direct comparison of baseline o₂max less than 8.3 ml/kg/min during CPET and desaturation less than 88% during a 6MWT shows that the o₂max threshold is a better predictor of survival in IPF. CPET and the 8.3 ml/kg/min threshold provide additional information to clinicians and their patients to help guide therapeutic decisions in IPF.