© 2002 American Thoracic Society
An Elevated Breathing Reserve Index at the Lactate Threshold Is a Predictor of Mortality in Patients with Cystic Fibrosis Awaiting Lung TransplantationPulmonary and Critical Care Unit, General Medical Services; and Lung Transplantation Program, Massachusetts General Hospital and Harvard Medical School, Boston, Massachusetts Correspondence and requests for reprints should be addressed to Leo C. Ginns, M.D., Massachusetts General Hospital, Lung Transplantation Program, Pulmonary and Critical Care Unit, Bigelow 808, Boston, MA 02114. E-mail: lginns{at}partners.org
The proportion of cystic fibrosis (CF) patients dying while on the lung transplant wait list remains high; identification of such patients remains difficult. The breathing reserve index (BRI = minute ventilation/maximal voluntary ventilation) at the lactate threshold (LT) is a predictor of a pulmonary mechanical limit to incremental exercise. We hypothesized that an elevated BRI at the LT in patients with CF awaiting lung transplantation would be a predictor of wait list mortality. Forty-five consecutive patients with CF completed cardiopulmonary exercise testing as part of their pretransplant assessment. We evaluated BRI at LT, baseline demographic characteristics, pulmonary function, and other exercise parameters via Cox proportional hazards modeling. Fifteen patients died while awaiting transplant. Twenty one were transplanted, and nine still awaited transplantation. Relative risks from the multivariate model included (95% confidence interval in parentheses) BRI at LT, 17.52 (2.45–123.97); resting PaCO2, 1.29 (1.10–1.49); resting PaO2, 0.97 (0.90–1.05); and forced expiratory volume at one second as a percent of predicted, 1.19 (1.05–1.34). BRI at LT not only provided the highest point estimate of risk for wait list mortality but also identified a physiologically significant threshold value (0.70 or more) for those at risk. This measurement may allow improved timing of listing for transplantation, including consideration for living donor transplantation.
Key Words: cystic fibrosis • lung transplantation • exercise • anaerobic threshold • mortality
Since the mid-1980s, over 1,100 lung transplants have been performed in the United States on patients with cystic fibrosis (CF) (1). In the appropriate candidate, bilateral lung transplantation can improve pulmonary function, exercise tolerance, quality of life, and survival (2). Nevertheless, the optimal timing of referral of patients with CF for transplantation remains difficult because both the natural history of CF is highly variable (2–4) and transplant waiting times continue to increase (1, 3). A referral that is too late confers increased mortality while awaiting transplantation; a transplant that is too early minimizes the benefits of the procedure, while subjecting the patient to the numerous potential post-transplantation complications. Current guidelines for cadaveric lung transplantation suggest that referral for patients with CF is appropriate when the forced expiratory volume at one second (FEV1) is 30% or less of predicted or more than 30% with rapid clinical deterioration. Other criteria include resting hypercapnia (PaCO2 > 50 mm Hg), resting hypoxemia (PaO2 < 55 mm Hg), and consideration of earlier listing for female patients (5). These guidelines are based largely on the study by Kerem and colleagues (6). In a Canadian CF population, they reported a 2-year mortality of more than 50% in patients whose FEV1 as a percent of predicted (FEV1%) fell to less than 30% for height and sex. Nevertheless, others have demonstrated a low discriminatory power of the FEV1 for predicting survival at less than 2 years (4, 7). For living donor lobar transplantation, no specific guidelines are available, other than a general consensus that patients not likely to remain a viable candidate for cadaveric transplantation should be considered. The ability to identify such patients with objective measurements should assist in the evaluation of appropriate candidates for living donor transplantation. The search for better predictors of mortality in the CF population continues. The cardiopulmonary exercise test would seem to be ideally suited for screening patients for transplant eligibility, as it measures the patient's pulmonary and cardiovascular reserve. Prior studies have demonstrated an association between oxygen uptake at peak exercise (V·O2 max) and overall mortality in CF (8, 9). We investigated the role of a novel marker of exercise tolerance, the breathing reserve index (BRI) at the lactate threshold (LT), in patients with CF awaiting lung transplantation. The BRI at LT is defined as the minute ventilation (V·E)/maximal voluntary ventilation (MVV) at the LT (10). When measured at the LT, an elevated BRI may be a more pure, reproducible, less effort-dependent index of deranged pulmonary mechanics (10), as it is measured during a submaximal realm of exercise before the sustained increase in blood lactate and ventilatory drive that is thought to be related to circulatory and skeletal muscle dysfunction (11). We hypothesized that those CF patients with an elevated BRI at LT would be at increased risk for death while on the transplant wait list.
Study Population Between November 1990 and March 1997, 51 patients with the primary diagnosis of CF completed transplant evaluation at the Massachusetts General Hospital. The cardiopulmonary exercise test is a routine part of the transplantation evaluation at this institution. Four patients were lost to follow-up. Two patients had no pretransplant exercise data. Therefore, 45 patients comprise this study. These were divided into two groups: those who survived to transplantation or were alive and awaiting transplantation at the time of analysis (n = 30) and those who died awaiting transplantation (n = 15). The minimum duration of listing for analysis of those alive and awaiting transplant was 18 months, set to equal the minimum expected wait time for transplantation (2).
Procedures
Data Analysis V·O2 max was defined as the highest V·O2 measured during the symptom-limited exercise test. Predicted values for maximal oxygen uptake (V·O2 max) were those of Hansen and colleagues (15). The LT was defined by the "join point" of the log–log plot of lactate concentration versus V·O2 (16). In five cases of indeterminate LT, the "V-slope" gas exchange method of determining anaerobic threshold was employed (17–19). The BRI was calculated as V·E/MVV at maximum exercise (BRImax) and at the LT.
Survival analysis by Cox proportional hazards modeling was done using time to event from the initial cardiopulmonary exercise test date. Patients still awaiting transplantation were censored at the time of analysis. Initial modeling evaluated the following univariate predictors: FEV1, PaCO2, PaO2, P(A-a)O2, body mass index (BMI), sex, V·O2 max, BRImax, LT, V·D/V·T, and ventilatory equivalent for CO2 (V·E/V·CO2). BRI at LT was initially modeled as a linear variable, but threshold effects were found to be equally predictive. A threshold value of 0.70 was subsequently chosen because it represents the value associated with a pulmonary mechanical limit (12, 20–22). Univariate predictors of moderate statistical significance (p values
Baseline patient characteristics are shown in Table 1. Demographic data were remarkable for approximately equal age, sex, and BMI distribution in the groups. The presence of Burkholderia cepacia was also assessed. Only one patient in each of the groups developed sputum cultures positive for these bacteria. Baseline spirometric and gas exchange data are also shown on Table 1. These parameters form the basis of the nonexercise variables used in the univariate proportional hazards regression model. Both groups began with an FEV1 of less than 30% of predicted as expected based on inclusion criteria; the difference in FEV1 between the groups was not statistically significant. Resting PaO2, PaCO2, and P(A-a)O2 were assessed as markers of gas exchange. The mean time of death in the group that died was 325 days, whereas the mean wait time was 826 days for those subsequently transplanted and 823 days for those still on the wait list.
Baseline exercise data from the two patient groups are shown in Table 2. Besides the primary predictor of interest, the BRI at LT, other exercise predictors included in the univariate proportional hazards model include V·O2 max, V·O2 max/kg/min, LT, LT/kg/min, BRImax, V·E/V·CO2 at maximal exercise and at the LT, and V·D/V·T.
Univariate time-sensitive analysis of covariates of interest is shown in Table 3 . Values are expressed as relative risk ratios, assuming a null value of one; p values and 95% confidence intervals for the relative risk ratios are also shown. In the univariate analysis, the only significant predictors of mortality in this population of pretransplant CF patients were resting PaCO2, resting PaO2, and the BRI at LT. In the univariate analysis, our predictor of interest, the BRI at LT, was highly predictive of pretransplant mortality with a relative risk of 9.82 (95% confidence interval, 2.69–35.64) at a threshold value of 0.70. Several other variables approached statistical significance. These included the FEV1, FEV1%, P(A-a)O2, V·O2 max/kg/min, and V·D/V·T. The FEV1 and P(A-a)O2 were dropped because of high collinearity with the common historic predictors of the FEV1% and the PaO2. The other covariates were included in the initial multivariate model. V·O2 max, LT, and V·E/V·CO2 were not significant (p > 0.25) in the univariate analysis and were therefore excluded in the multivariate model. Although age, sex, and BMI were also not significant, they were included in the initial multivariate model because of their historical significance and the potential for confounding of the model effects without the inclusion of these covariates.
The final multivariate analysis results are shown in Table 4. This model contains three significant predictors of mortality in CF: the baseline resting PaCO2, the FEV1%, and the BRI at LT. One nonsignificant predictor whose inclusion enhanced the overall model, the baseline resting PaO2, is also included in the model. Although age, sex, BMI, V·D/V·T, and V·O2 max/kg/min were included in the initial multivariate model, they did not add predictive value and were dropped from the final model. As shown, the relative risks for the model include (95% confidence intervals in parentheses): the BRI at LT, 17.52 (2.45–123.97); PaCO2, 1.29 (1.10–1.49); PaO2, 0.97 (0.90–1.05); and FEV1%, 1.19 (1.05–1.34). It should be noted that modeling excluding the BRI at LT yielded a nonsignificant result for the relative risk of the FEV1% in both the saturated (RR of 1.11, p = 0.08) and the final (RR of 1.08, p = 0.14) models.
Our data show that an elevated BRI at LT is associated with a markedly increased relative risk of death in patients with CF awaiting lung transplantation. In both our univariate and multivariate Cox regression model, the BRI at LT was the factor displaying the highest relative risk of death. Despite its ability to predict a pulmonary mechanical limit to exercise in chronic obstructive pulmonary disease (10), this is the first demonstration, to our knowledge, of the discriminatory power of the BRI at LT in predicting mortality. The physiologic abnormalities during exercise in patients with CF, including those pretransplant, have been well described (12, 23). There is a pulmonary mechanical limit to exercise before transplantation, implying that airway obstruction prohibits patients from generating a minute ventilation (V·E) sufficient to meet the increased metabolic demands of exercise. In general, pretransplant patients with CF also develop an early lactate acidemia (12) and thus an early LT. The presence of a pulmonary mechanical limit at such an early time during the exercise cycle implies ventilatory impairment with the mildest degree of exertion. The physiologic basis for the use of the BRI at LT as a prognostic marker in patients with CF awaiting lung transplantation rests on the uniqueness of the BRI to measure simultaneously alterations in V·E with limitations in pulmonary mechanics (as measured by the FEV1). By definition (BRI = V·E/MVV; MVV = FEV1 x 40) any increase in V·E or decrease in FEV1 at a given time point during cardiopulmonary exercise test will result in a simultaneous increase in the BRI. The rationale for the simultaneous measurement of FEV1 and V·E in the end-stage CF population follows. It is clear that progression of disease in CF is accompanied by decrements in the FEV1; the leading cause of mortality for these patients remains respiratory failure (24). Population-based cohort studies of individuals with CF have consistently demonstrated a survival benefit associated with increasing FEV1 (6, 7, 9, 25–27). The FEV1 as a percent predicted remains a prime determinant in considering CF patients for transplantation (3, 5). Although early studies also supported the predictive value of the FEV1 in CF patients awaiting lung transplantation (28, 29), they were limited by the inclusion of very ill subjects with very short survival times. This was emphasized by Ciriaco and colleagues (28). In analyzing 67 CF patients listed for transplantation between 1990–1993, they noted a mean FEV1 of 16.8% predicted in those who died awaiting transplant and concluded that patients were being referred too late. All of the more recent studies support the notion that once the FEV1 drops below 30% predicted (the primary criteria for transplant listing [5]), its discriminatory power to predict mortality becomes poor (4, 30–34). The need for other markers once patients meet this criterion for transplant listing has been noted (33). In fact, an overreliance on the FEV1 may enhance the possibility of adverse outcomes in these patients (30, 32, 34). Not unexpectedly, the FEV1% did not predict mortality among the subjects in our study. Although our final multivariate model does include the FEV1% as "significant," inclusion of the FEV1% without BRI at LT, as would be done in any study not evaluating this unique marker, produced a nonsignificant relationship (p = 0.14) between FEV1% and mortality. Models including the BRI at LT without the FEV1, however, do maintain a significant association (p = 0.02) (data not shown). Moreover, if the predictive value of the BRI on CF mortality were solely due to the effects of FEV1, it would be expected that the BRImax would have been as predictive as the BRI at LT. Multiple studies have also demonstrated that V·E in CF is elevated, both at rest (35, 36) and during exercise (37–39). Moreover, these increases in V·E tend to parallel disease severity (38, 39). Physiologic dead space increases with worsening airflow obstruction in CF (38, 39). This has led to the hypothesis that the increased V·E noted in the CF population is a reflection of the increased dead space caused by limitations in airflow (36, 38). However, other mechanisms by which the V·E in such patients can be elevated exist. First, it has been noted that alterations in carbon dioxide chemosensitivity occur during exercise in CF resulting in an abnormally high V·E for a given alteration in V·CO2 (37). Second, although cor pulmonale has long been implicated in the prognosis of CF patients (40), it has been recently demonstrated that even subclinical pulmonary hypertension is also associated with increased mortality (41, 42). Pulmonary hypertension, in turn, has been consistently associated with an increased exercise V·E for V·CO2 (43–45), in part because of its independent effect on physiologic dead space (44–46). Finally, it is possible that ventilatory drive in CF may be increased secondary to alterations in skeletal muscle oxygen metabolism (47–49), which have been noted even during low-level exercise. The resulting decreased intracellular pH during exercise is linearly related to increased V·E in normals and can drive ventilation independent of arterial pH even before the LT (50). Thus, the predictive ability of the BRT at LT rests on the simultaneous incorporation of a classic measure of baseline lung function, the FEV1, with that of a broader measure of exercise capacity, the V·E, which can represent alterations in pulmonary mechanics, V/Q matching, pulmonary hypertension, and skeletal muscle oxidative metabolism. It is likely that the finding of a relatively elevated BRI could be found at any submaximal exercise period in the patients at risk; however, the LT is a convenient, reproducible reference point of cardiopulmonary function (16, 51, 52). The LT has also been advocated as a minimum training threshold to enhance pulmonary clearance in patients with CF (53). Thus, one additional speculation for the increased mortality in patients with an elevated BRI at LT is that these patients have a decreased capability to clear secretions adequately caused by failure of pulmonary mechanics. We have chosen a threshold level for BRI at LT of 0.70 to match reference definitions for a pulmonary mechanical limit (12, 20–22). The BRI at LT was dichotomized using this definition in the univariate and multivariate analyses performed. In both the univariate and multivariate analyses, the BRI at LT yielded the highest point estimates and confidence intervals for the relative risk of death. The difference in survival of those patients at or exceeding this threshold at the time of transplant evaluation compared with those below this threshold is shown in Figure 1 . However, it should be noted that there was enough of a difference between those patients with a high BRI at LT and those with a lower one that similar results could have been obtained using alternate values for the threshold. Univariate results using varying threshold levels are shown in Table 5.
One criticism regarding the generation of the numerical data involved in the BRI at LT might be the inclusion of patients in whom no discrete LT was measured. In these patients, the V-slope anaerobic threshold was used as the LT point. This was true of five patients, two in the group that died awaiting transplantation and three in the group that survived. Use of this determination seems to be a conservative estimate of the LT in CF. In a study of 32 patients with CF, Nikolaizik and colleagues (53) noted that the V-slope method demonstrated lower values than the LT. Assuming that the V-slope method does underestimate the LT, this would have resulted in a lower BRI at LT for our patients in whom the LT could not be determined. Thus, the presence of an elevated BRI at the lower V-slope anaerobic threshold points appears to be a valid marker for a premature pulmonary mechanical limit. There are other potential limitations to this study. Data collection was done retrospectively, although techniques were used to minimize errors. Overall, missing data were small, and analyses with and without adjustments for missing data did not alter results. Numerically, this study was relatively small (45 patients analyzed) and may not have been sufficiently powered to allow for significance among some of the covariates. Sample size may have also led to false positives because of the number of covariates included in the initial analysis. However, the strength of the association of our primary outcome of interest, as well as the historical precedence of our other significant predictors in our multivariate analysis, lends credence to our conclusions. In summary, an elevated BRI at LT represents reduced pulmonary mechanical reserve at low-level exercise, likely caused by a combination of increased ventilatory demand and decreased capacity. We have found that an elevated BRI at LT is highly predictive of mortality in patients with CF awaiting lung transplantation. The presence of an overt threshold level that distinguishes between those at risk for wait list mortality and those likely to survive to transplantation makes this marker unique. Use of this measurement may facilitate better prioritization of transplant listing, addressing the concern about the poor discriminatory ability of the FEV1 alone. It may also allow patients at risk for premature deaths to contemplate other options, such as living donor transplantation.
This work was supported by the American Heart Association (96-50406; D. S.) and the National Institutes of Health (K24: HL04022-02) and also in part by the Nirenberg Center for Advanced Lung Disease and the Levi Montiero Lung Transplant Fund. Received in original form May 21, 2001; accepted in final form March 26, 2002
This article has been cited by other articles:
|
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
TOP
ABSTRACT
INTRODUCTION
0.25) and those of historical significance (age, BMI, and sex) were included in the initial multivariate model. Goodness of fit testing using likelihood ratio testing allowed compilation of the final multivariate predictive model. Computations were made with the SAS statistical program (version 6.12; SAS Institute, Cary, NC). Results are expressed as relative risks and their 95% confidence intervals. A p value of less than 0.05 was considered significant. 
