Cardiopulmonary Exercise Testing: Another Tool in the Prognostication Tool Kit for Cystic Fibrosis

Survival for individuals with cystic fibrosis (CF) is improving over time, but progressive respiratory failure remains the number one cause of death for individuals with CF (1). Historically, FEV₁ <30% of the predicted value has prompted discussions in CF clinics about the potential need for lung transplantation (LTx) (2, 3). However, survival with advanced lung disease is increasing over time, with a recent estimate of median survival of 6.6 years after FEV₁ <30% in the United States (3–6). Despite the improved survival times for individuals with FEV₁ <30%, rates of death in the United States are approximately 10% per year after this lung function threshold is reached (6). Although FEV₁ has been shown to have a strong and consistent association with death or LTx in CF, there are other predictors as well, including malnutrition, hypoxemia, hypercarbia, pulmonary hypertension, increased frequency of exacerbations or hospitalizations, sputum culture positive for Burkholderia cepacia, massive hemoptysis, and reduced 6-minute-walk test distance (3, 4, 7–11). Despite these data, estimating the time until death or LTx in patients with CF is exceedingly difficult, and care teams need more and better tools to prognosticate in this patient population.

In this issue of the Journal, Hebestreit and colleagues (pp. 987–995) present a multicenter, international, retrospective study of clinically indicated cardiopulmonary exercise testing (CPET) for individuals with CF (12). Ten centers (in Europe, Australia, and North America) contributed CPET data from over 500 individuals with CF, age ≥10 years, between 2000 and 2007. Data from a valid maximal CPET were available for 433 individuals, with follow-up of the cohort through 2014. The subjects selected were relatively healthy despite having a clinical indication for CPET (mean FEV₁, 73% predicted; 5-year survival rate, 93%). The investigators found that $\dot{\text{V}}$o_2peak, work_peak, $\dot{\text{V}}$e/$\dot{\text{V}}$o₂, and $\dot{\text{V}}$e/$\dot{\text{V}}$co₂ were all associated with the composite outcome after adjustment for other known predictors of death and/or LTx in multivariable models. Using Ward’s hierarchical clustering, the investigators identified four clusters, which included continuous and binary clinical and physiological parameters. This cluster analysis identified a group of individuals with low FEV₁, low body mass index z-scores, and worse CPET performance with dismal outcomes over the course of 10 years (63% death or LTx). Although FEV₁ was the most important variable for clustering, the CPET-derived parameters had a stronger influence on clustering than other traditional risk factors for death or LTx.

This study has several strengths. First, this is the largest study of CPET in CF, and it confirms prior single-center findings regarding the prognostic value of CPET-derived parameters in adults and adolescents with CF (13–15). Because of the large sample size in this study, analyses could be adjusted for important potential confounders of the association between CPET performance and death or LTx. The investigators identified strong relationships between CPET variables and death or LTx in the entire cohort that were independent of FEV₁, as well as among individuals with advanced lung disease (FEV₁ < 40%) and in short-term (2 yr) sensitivity analyses. Second, this study had long-term follow-up of clinical outcomes, with very few individuals lost to follow-up or missing primary endpoint data 5 years or more after CPET (n = 58, excluded from analyses). Third, the use of cluster analysis highlights the importance of focusing prognostication on the highest-risk group (individuals with low FEV₁, malnutrition, and poor CPET performance). One of the key benefits of CPET is that it represents a functional and dynamic assessment of the cardiopulmonary system. Such an evaluation provides important clinical variables that are unavailable during a static test of airflow, such as office spirometry.

One of the fundamental challenges of prognosticating in CF is that the event rate (death or LTx) in the overall population is low. Prognostication is most relevant for individuals with an imminent risk of death, to avoid missing the opportunity for LTx in the appropriate individuals with CF. Incorporation of CPET could augment the complex decision-making that occurs around the timing of evaluation and listing for LTx. Interestingly, because of longstanding evidence of CPET-derived parameters (e.g., $\dot{\text{V}}$o_2peak and $\dot{\text{V}}$e/$\dot{\text{V}}$co₂ slope) as predictors of death in patients with systolic heart failure, the selection of heart transplant candidates has incorporated CPET for more than a decade (16), and carefully collected prospective data support the prognostic value of CPET for these individuals (17–19). The identification of threshold values for CPET parameters to guide the timing of listing individuals with CF for LTx could be invaluable because of the documented prolonged survival with low lung function and the poor positive predictive value of FEV₁ <30% predicted (10).

Some key weaknesses of the study were acknowledged by the authors. One concern raised was the potential for confounding by CF center practices. They identified significant differences in outcomes at the CF center level. This in turn led the investigators to adjust for clinical site in their models. Although this analysis can take into account within-site correlation of participants, it cannot address potential differential indication bias. Individuals who underwent CPET at each site may have had different disease severities or clinical indications that could not be accounted for in the analysis. When indication bias occurs in an observational study, it remains a challenge to address analytically. The investigators would have needed a separate control population of individuals who had an equal probability (potentially via the propensity score) of undergoing CPET but did not receive the test. Differential outcome ascertainment is also a potential source of bias for this study, as the investigators attempted to minimize the risk of bias from loss to follow-up or informative censoring, but may have introduced ascertainment bias when the cohort was limited to individuals with a minimum of 5 years of follow-up at the testing CF center (e.g., healthier individuals may have moved away from the center). Thus, it remains challenging to generalize the results of this study to the greater CF population. Despite these limitations, the data presented provide strong observational evidence for the potential role of CPET in risk stratification for individuals with CF.

In conclusion, CPET adds prognostic information beyond the FEV₁ and could be a dynamic marker of disease progression in CF. The study by Hebestreit and colleagues is a call to action to perform a prospective study of CPET for individuals with CF—ideally, individuals with severe CF. CPET is another tool in the prognostication tool kit for CF and prospective research is imperative for individuals with advanced lung disease approaching LTx.