Rationale: The optimal timing for listing of cystic fibrosis patients for lung transplantation is controversial.
Objectives: We conducted a retrospective cohort study of 343 patients listed for lung transplantation at four academic medical centers to identify risk factors for death while awaiting transplantation.
Methods: Data on possible risk factors were abstracted from medical records.
Measurements: Time to death, patient demographic characteristics, and risk factors for death while awaiting transplantation were assessed. Univariate and multivariate survival analyses were performed using Cox regression.
Results: By univariate analyses, FEV1 ⩽ 30% predicted (HR, 3.8; 95% CI, 2.0–7.5), PaCO2 ⩾ 50 mm Hg (HR, 1.85; 95% CI, 1.1–3.0), and shorter height (HR, 1.8; 95% CI, 1.1–3.0) were associated with a higher risk of death. Referral from an accredited cystic fibrosis center was associated with a lower risk (HR, 0.53; 95% CI, 0.30–0.92). The final multivariate model included referral from an accredited cystic fibrosis center (HR, 0.5; 95% CI, 0.3–1.0) and listing year after 1996 (HR, 0.4; 95% CI, 0.2–0.7); both were associated with a lower risk of death. FEV1 ⩽ 30% predicted (HR, 6.8; 95% CI, 2.4–19.3), PaCO2 ⩾ 50 mm Hg (HR, 6.9; 95% CI, 1.5–32.1), and use of a nutritional intervention (HR, 2.3; 95% CI, 1.3–4.1) were associated with increased risk. Patients with FEV1 > 30% predicted had a higher risk of death only when their PaCO2 was ⩾ 50 mm Hg (HR, 7.0; 95% CI, 1.5–32), while the increased risk of death with FEV1 ⩽ 30% was not further influenced by the presence of hypercapnia.
Conclusions: We identified risk factors for waiting list mortality that could impact on transplant listing and allocation guidelines.
Cystic fibrosis (CF) is a multisystem autosomal recessive disease with a median survival of 33 yr. Lung transplantation is offered to patients with advanced CF lung disease as a potential life-extending and life-improving treatment (1). Because about 90% of patients with CF die from end-stage obstructive lung disease, CF is now the third leading indication for lung transplantation (2).
For each patient with CF there is a “transplant window” that opens when the risks of delay in referral for lung transplant outweigh those of the transplant procedure and closes when the patient is too ill, or too high an operative risk, to be an acceptable candidate for transplant. In general, patients must be sufficiently ill to require transplantation, but not so sick that they cannot tolerate the procedure. Commonly used guidelines for referring a patient for lung transplant include (1) FEV1 less than 30% predicted or rapidly progressive respiratory deterioration with an FEV1 greater than 30% predicted, (2) PaCO2 greater than 50 mm Hg, (3) PaO2 less than 55 mm Hg on room air, and/or (4) female sex and age less than 18 yr with FEV1 greater than 30% predicted, and deteriorating rapidly. These guidelines are largely based on data published in 1992 from the University of Toronto (Toronto, ON, Canada) (3–5). However, by using these criteria up to 30% of listed patients with CF die before transplantation (6). These high waiting list mortality data may reflect the inaccuracy of identifying the transplant window, or may highlight the limitations of waiting time–based donor allocation. We therefore conducted a 10-yr retrospective multicenter study to better identify predictors in a population “selected” by their physicians for lung transplantation, with the aim of obtaining a better understanding of the factors related to mortality while awaiting transplant. Recognition of these factors may lead to better lung allocation algorithms that maximize the survival benefit of transplantation.
Two groups have published models of predicted mortality based on retrospective analysis of the Cystic Fibrosis Foundation (Bethesda, MD) Patient Registry. The model of Liou and coworkers (7) predicts 5-yr survival on the basis of 10 characteristics. In a study employing this model, Liou and coworkers (8) demonstrated that only those adult patients with CF with a predicted 5-yr survival of less than 50% and who do not have Burkholderia cepacia colonization or CF-related arthropathy at the time of transplant derive a survival advantage from transplantation.
Mayer-Hamblett and coworkers (9) developed a multiple logistic regression model for predicting 2-yr mortality in patients 6 yr of age or more, based on 1996 U.S. Cystic Fibrosis Foundation Patient Registry data. This model included five variables in addition to mean FEV1. Unfortunately, this model had no better predictive accuracy for 2-yr mortality than FEV1 of less than 30% predicted alone. Both percent predicted FEV1 and the model presented had low positive predictive values. This study reaffirms the need to identify better predictors of short-term outcome in an attempt to improve mortality on the waiting list.
These data suggest that criteria by which to best identify patients with CF appropriate for lung transplant can still be improved. Because Liou and coworkers (8) and Mayer-Hamblett and coworkers (9) used Cystic Fibrosis Foundation Patient Registry data, there may be misclassification bias present (e.g., lack of validated diagnostic criteria for risk factors such as CF-related diabetes mellitus). Our study used careful review of charts from each center and standardization of criteria for categorization of risk factors. Clinical risk factors potentially associated with higher risk of death in patients with CF in previous studies are listed in Table 1. The purpose of this study was to identify other important risk factors for death specifically in patients with CF awaiting lung transplantation. Our current criteria for listing are quite poor at predicting who will survive to transplantation and who will die. Thus, a 10-yr retrospective, multicenter study was done to identify predictors in a population “selected” by their physicians. This may allow modification of who is listed as well as timing of listing to attempt to improve survival on the waiting list.
Factors Associated with Higher Risk of Death in General CF Population | Factors Associated with Higher Risk of Death in CF Population on Lung Transplant List | ||||
---|---|---|---|---|---|
Factor | Reference(s) | Factor | Reference(s) | ||
1. Pulmonary function and exercise capacity | 3, 7, 9, 14, 18–29 | 1. Yearly rate of decline in percent predicted FEV1 | 19 | ||
2. Microbiology (i.e., Burkholderia cepacia,Pseudomonas aeruginosa) | 30–35 | 2. Shorter 6-min walk distance | 14 | ||
3. Nutritional status (weight) | 36, 37 | 3. Presence of pulmonary hypertension | 14 | ||
4. Age | 18, 19, 23, 24, 28, 38 | 4. Presence of diabetes mellitus | 14, 26 | ||
5. Female sex | 13, 32, 33, 37 | 5. Shorter height | 26, 44 | ||
6. Pancreatic insufficiency | 35, 39 | ||||
7. CFRDM | 13, 29, 35 | ||||
8. Lower socioeconomic status | 40 | ||||
9. Pulmonary hypertension | 14, 41 | ||||
10. Number of acute exacerbations | 13 | ||||
11. Geographic location | 42 | ||||
12. Care in CF center | 32, 43 |
Some of the results of our study have been previously reported in the form of abstracts (10, 11).
The study population consisted of 343 adult and pediatric patients with CF listed for lung, heart–lung, or heart–lung–liver transplantation at the University of Pennsylvania Medical Center (Philadelphia, PA), Stanford University Medical Center (Stanford, CA), Children's Hospital of Philadelphia (Philadelphia, PA), Toronto General Hospital (Toronto, ON, Canada) and the Hospital for Sick Children in Toronto (Toronto, ON, Canada) from January 1990 to December 2002. Heart–lung transplant candidates were included in this study because this procedure was routinely performed at Stanford University Medical Center in the early 1990s. Institutional review board approval was obtained at each institution. Data were collected by chart review, using standardized procedures (see the online supplement for the detailed manual). A 10% sample of charts was reabstracted. Three patients were excluded because either the listing date or removal date was unknown.
We collected demographic, body measurement, pulmonary function, and hemodynamic data at the time of listing and data on microbiology, nutritional status, CF-related diabetes mellitus, osteoporosis, and use of preventive therapies before or at the time of listing. See Table E3 of the online supplement for a complete list of data items collected. A description of reference equations for percent predicted FEV1 is also presented in the online supplement.
Death while listed for transplant was the primary endpoint for all analyses. Patients who underwent lung transplantation were censored on the date of their operation, and those alive and waiting for transplantation were censored at the end of the study on December 31, 2002. Patients removed from the transplant waiting list for other reasons (voluntary removal, transfer to another center, “too well,” or “too sick”) were censored at the time of removal.
Statistical analysis was performed with Stata 7.0 (StataCorp, College Station, TX). All statistical tests were two-sided.
Comparisons of categoric variables between survivors and nonsurvivors were performed by χ2 or Fisher's exact test. Continuous variables were summarized as mean ± SD and 95% confidence interval (95% CI), or as median, range, and interquartile range (25–75%). Comparisons between survivors and nonsurvivors were performed by t test for normally distributed variables and by Wilcoxon rank-sum test for other variables.
Each potential risk factor was first analyzed separately by the Kaplan-Meier method to estimate survival, by univariate Cox regression to estimate the hazard ratios, and by log rank test to test survival differences between groups. Hazard ratios are presented for a 10% decrease in percent predicted FEV1 and FVC and for a 10–mm Hg increase in PaO2 and PaCO2.
We used Cox regression for multivariate analyses to determine hazard ratios, and the Wald χ2 test to calculate p values. The final model was generated by a backward model selection procedure from variables with p < 0.20 from the univariate Cox regression analyses, significant confounders, and biologically plausible effect modifiers. Variables were eliminated from highest to lowest p value, but remained in the model if the p value was less than 0.05. Variables believed to be highly clinically significant or found to be independent predictors of death in prior studies were included in these analyses irrespective of their p value. Minus log–minus log plots were used to confirm that the proportional hazards assumption was met.
To assess the impact of potential informative censoring, a sensitivity analysis was performed to determine the effect of two extreme cases (12). The impact of missing data was also assessed by comparing characteristics of patients included in the final multivariate model with those removed because of missing data. Analyses were also performed for each center separately to assess intercenter variability.
A total of 346 patients were identified; three patients were excluded because either their listing date or removal date was unknown. A total of 343 patients were analyzed (University of Pennsylvania Medical Center, n = 85; Stanford University Medical Center, n = 94; Children's Hospital of Philadelphia, n = 26; and Toronto General Hospital and the Hospital for Sick Children in Toronto, n = 138). Seventy-eight patients died (nonsurvivors group; n = 78). Two hundred and fifteen patients received a transplant, with patients receiving a bilateral lung transplant (n = 205), heart–lung transplant (n = 7), lung–liver transplant (n = 2), or heart–lung–liver transplant (n = 1). Thirty-four were still waiting at the end of the study and 16 were removed for other reasons (survivors group; n = 265).
Demographic and clinical profiles of the survivors and nonsurvivors are shown in Table 2. One notable difference between the groups was percent-predicted FEV1. There was a significant difference in the mean percent-predicted FEV1 and proportion of patients with an FEV1 of less than 30% predicted between survivors and nonsurvivors (p < 0.01; Table 2). Nonsurvivors had a greater 6-min walk distance, were more likely to have multidrug-resistant Pseudomonas (defined as resistant to three or more different antibiotics), and were more likely to use inhaled recombinant human DNase. Height was similar between the groups (p = 0.3). Survival rates for the entire cohort were 80% at 1 yr, 64% at 2 yr, and 50% at 3 yr. Mean time to death among the nonsurvivors was 347 ± 314 d, which was similar to the average time to transplantation among the survivors (368 ± 351 d). Median time to death in the nonsurvivors group was 267 d (interquartile range, 92–489 d), whereas the median time to transplant in the survivors group was 251 d (interquartile range, 100–510 d). These results are further detailed in Table 2. The mean waiting time for those still waiting at the end of the study was 551 ± 483 d.
Nonsurvivors | Survivors | ||||||
---|---|---|---|---|---|---|---|
General Characteristics | Value | No. | Value | No. | p Value | ||
Age, yr: mean ± SD | 27 ± 9 | 76 | 26.9 ± 8 | 262 | 0.90 | ||
Sex, female: no. (%) | 40 (51) | 78 | 122 (46) | 265 | 0.44 | ||
Height, cm: mean ± SD | 163 ± 13 | 78 | 164 ± 14 | 259 | 0.30 | ||
Race, white: no. (%) | 76 (100) | 76 | 253 (77) | 264 | 0.47 | ||
BMI, mean ± SD | 19.4 ± 3 | 78 | 19.3 ± 3 | 258 | 0.93 | ||
Pulmonary function | |||||||
FEV1% predicted, mean ± SD | 24 ± 7 | 75 | 27 ± 8 | 257 | < 0.01 | ||
FEV1 ⩽ 30% predicted, no. (%) | 65 (87) | 75 | 171 (67) | 257 | < 0.01 | ||
FVC% predicted, mean ± SD | 42 ± 11 | 75 | 44 ± 14 | 222 | 0.13 | ||
Gas exchange, mean ± SD | |||||||
PaO2, mm Hg | 64 ± 12 | 70 | 64 ± 10 | 195 | 0.46 | ||
PaCO2, mm Hg | 47 ± 9 | 71 | 46 ± 9 | 224 | 0.22 | ||
SIXMW, ft | 1,021 ± 451 | 63 | 900 ± 520 | 214 | 0.06 | ||
Microbiology, no. (%) | |||||||
Burkholderia cepacia | 14 (19) | 75 | 57 (22) | 257 | 0.63 | ||
Stenotrophomonas maltophilia | 1 (1) | 76 | 24 (9) | 257 | 0.02 | ||
Achromobacter xylosoxidans | 2 (3) | 76 | 8 (3) | 257 | 0.99 | ||
MDR Pseudomonas | 19 (25) | 76 | 50 (20) | 257 | 0.33 | ||
MRSA | 1 (1) | 76 | 16 (6) | 257 | 0.13 | ||
Aspergillus spp. | 7 (9) | 76 | 49 (19) | 256 | 0.05 | ||
CMV (IgG) | 21 (31) | 68 | 102 (39) | 261 | 0.26 | ||
Hemodynamics, no. (%) | |||||||
Pulmonary hypertension | 6 (21) | 29 | 29 (29) | 100 | 0.48 | ||
RV dysfunction | 7 (21) | 33 | 76 (51) | 150 | < 0.01 | ||
Presence of, no. (%) | |||||||
Osteoporosis | 5 (23) | 21 | 25 (23) | 108 | 0.99 | ||
Chronic sinusitis | 47 (65) | 72 | 154 (63) | 245 | 0.78 | ||
Nutritional intervention | 25 (33) | 75 | 55 (27) | 201 | 0.37 | ||
Liver disease | 5 (7) | 75 | 28 (11) | 259 | 0.38 | ||
CF-related diabetes mellitus | 16 (21) | 76 | 65 (25) | 260 | 0.54 | ||
Use of | |||||||
Inhaled TOBI, no. (%) | 43 (57) | 75 | 161 (65) | 247 | 0.22 | ||
Inhaled rhDNase, no. (%) | 45 (60) | 75 | 114 (46) | 247 | 0.05 | ||
Waiting time, d | |||||||
Mean | 347 | 78 | 369 | 265 | 0.87 | ||
Median | 251 (100–510)* | 78 | 267 (92–489)* | 265 |
Table 3 depicts the results of the univariate analyses. An FEV1 of less than 30% predicted was associated with a higher risk of death (see also Figure 1), as was a 10% decrease in FEV1 and FVC. Hypercapnia was also associated with a higher risk of death; a 10–mm Hg rise in PaCO2 was associated with a hazard ratio (HR) of 1.3 and a 95% CI of 1.1 to 1.7, whereas PaCO2 of 50 mm Hg or greater (compared with PaCO2 < 50 mm Hg) had an HR of 1.8 and a 95% CI of 1.1 to 3.0. Referral from an accredited CF center was associated with a lower risk of death (HR, 0.5; 95% CI, 0.3–0.9). Listing year during or after 1996 was associated with a lower risk of death (HR, 0.4; 95% CI, 0.3–0.7). Also, the use of a nutritional intervention was potentially associated with a higher risk of death (HR, 1.4; 95% CI, 0.9–2.2). The presence of Stenotrophomonas maltophilia (HR, 0.2; 95% CI, 0.02–1.2) and Aspergillus (HR, 0.6; 95% CI, 0.3–1.3) may be associated with a lower risk of death. The presence of B. cepacia colonization was not associated with a higher risk of death (HR, 1.1; 95% CI, 0.6–1.9). The univariate results did not appreciably change when only the 230 patients with complete data on all variables included in the final multivariate model were used (see Tables E2 and E3).
Variable | No. (343 total) | Hazard Ratio (95% CI) | p Value |
---|---|---|---|
Pulmonary function | |||
FEV1% predicted (10% ↓*) | 332 | 2.1 (1.5–3.0) | < 0.01 |
FVC% predicted (10% ↓) | 297 | 1.3 (1.1–1.6) | 0.01 |
FEV1 ⩽ 30% predicted | 332 | 3.8 (2.0–7.5) | < 0.01 |
Ventilation/oxygenation | |||
PaCO2, mm Hg (10 mm Hg ↑*) | 295 | 1.3 (1.1–1.7) | 0.02 |
PaCO2 ⩾ 50 mm Hg | 295 | 1.8 (1.1–3.0) | 0.02 |
PaO2, mm Hg (10 mm Hg ↓) | 265 | 1.2 (0.9–1.4) | 0.19 |
Stature | |||
Shortest height quartile† | 343 | 1.4 (0.9–2.4) | 0.18 |
Age less than 18 yr | 338 | 0.9 (0.5–1.7) | 0.76 |
BMI | 336 | 1.0 (0.9–1.1) | 0.91 |
Microbiology | |||
Stenotrophomonas maltophilia | 333 | 0.2 (0.02–1.2) | 0.07 |
Burkholderia cepacia | 332 | 1.1 (0.6–1.9) | 0.83 |
Achromobacter xylosoxidans | 333 | 0.9 (0.2–3.6) | 0.86 |
MDR Pseudomonas | 333 | 1.1 (0.7–1.9) | 0.64 |
MRSA | 333 | 0.3 (0.04–12.3) | 0.26 |
Aspergillus spp. | 332 | 0.6 (0.3–1.3) | 0.19 |
CMV | 329 | 0.8 (0.5–1.4) | 0.50 |
Hemodynamics | |||
Pulmonary hypertension | 129 | 2.7 (0.8–9.4) | 0.11 |
RV dysfunction | 183 | 0.4 (0.2–1.0) | 0.06 |
Others | |||
Accredited CF center | 334 | 0.5 (0.3–0.9) | 0.03 |
Six-minute walk, ft | 277 | 1.0 (0.99–1.0) | 0.19 |
Presence of osteoporosis | 129 | 1.1 (0.40–3.0) | 0.88 |
Nutritional intervention | 276 | 1.4 (0.90–2.2) | 0.19 |
Use of TOBI | 322 | 0.8 (0.50–1.30) | 0.40 |
Use of rhDNase | 322 | 1.0 (0.63–1.6) | 0.98 |
Presence of CFRDM | 336 | 1.1 (0.60–1.9) | 0.50 |
Presence of liver disease | 334 | 0.8 (0.3–2.0) | 0.67 |
Listing year (continuous variable) | 321 | 0.9 (0.8–0.9) | < 0.01 |
Listing year ⩾ 1996 | 321 | 0.4 (0.3–0.7) | < 0.01 |
The results of the multivariate analysis are shown in Table 4. Of the original cohort of 343 patients, 230 had complete data for all variables and were included in the final model. An FEV1 of less than 30% predicted is an independent predictor of death (HR, 6.8; 95% CI, 2.4–19.3). However, there was a significant interaction between PaCO2 and FEV1. Patients with FEV1 > 30% predicted had a significantly higher risk of death when their PaCO2 was 50 mm Hg or greater versus those patients with PaCO2 of less than 50 mm Hg (HR, 7.0; 95% CI, 1.5–32.0). Conversely, there was no significant difference in the risk of death between those patients with PaCO2 of 50 mm Hg or greater and those with PaCO2 of less than 50 mm Hg when their percent predicted FEV1 was less than 30 (HR, 1.4; 95% CI, 0.4–5.0).
Variable | Hazard Ratio for Death (95% CI) | p Value |
---|---|---|
FEV1 ⩽ 30% predicted | 6.8 (2.4–19.3) | < 0.01 |
PaCO2 ⩾ 50 mm Hg | 6.9 (1.5–32.1) | 0.01 |
FEV1 ⩽ 30% predicted × PaCO2 > 50 mm Hg* | 0.2 (0.04–1.0) | 0.05 |
In patients with FEV1 > 30% predicted, PaCO2 ⩾ 50 mm Hg (n = 5) versus PaCO2 < 50 mm Hg (n = 62) | 7.0 (1.5–32) | 0.01 |
In patients with FEV1 ⩽ 30% predicted, PaCO2 ⩾ 50 mm Hg (n = 52) versus PaCO2 < 50 mm Hg (n = 111) | 1.4 (0.4–5.0) | 0.7 |
Listing year ⩾ 1996 | 0.4 (0.2–0.7) | < 0.01 |
Nutritional intervention | 2.3 (1.3–4.1) | < 0.01 |
Accredited CF center | 0.5 (0.3–1.0) | 0.06 |
The presence of S. maltophilia colonization in the respiratory tract, shorter 6-min walk distance, and the presence of hypoxemia were all removed from the multivariate model as their p value increased to more than 0.05 during the backward selection procedure. Because B. cepacia has been associated with a higher risk of death in prior studies (9, 13), the presence of B. cepacia was forced into the multivariate model. CF-related diabetes mellitus was a significant risk factor for death in the one study (14) and thus was also forced into the model. Although the presence of pulmonary hypertension was found to be an independent predictor in the same study, we had too many missing data to include it in our final model. The presence of either B. cepacia and/or CF-related diabetes mellitus still did not significantly alter the other covariates when forced into the final multivariate model, and therefore these parameters were ultimately excluded from the model. Also, age, both as a continuous variable and as a dichotomous variable (using 18 yr as the cutoff) was forced into the model (age < 18 yr: HR, 1.8; 95% CI, 0.8–3.9; p = 0.13). Although there was a trend toward a higher risk of death for patients whose age was less than 18 yr, when removed from the model the hazard ratios of the other variables did not appreciably change and thus it was excluded from the model.
In the final model, year of listing (1996 and beyond compared with before 1996) was associated with a lower risk of death on the transplant waiting list (HR, 0.4; 95% CI, 0.2–0.7). Also, the use of a nutritional intervention was associated with a higher risk of death (HR, 2.3; 95% CI, 1.3–4.1). The final model is shown in Table 4.
We assessed whether transplant candidates from accredited CF centers differ from candidates not from accredited CF centers in markers of more severe disease (Table 5). Patients referred from an accredited CF center had a higher percent-predicted FEV1 (27 ± 8) than did those that were not referred from an accredited CF center (25 ± 8).
Variable | Referred from Accredited CF Center (n = 277) | Not Referred from Accredited CF Center (n = 57) | p Value |
---|---|---|---|
FEV1% predicted, mean ± SD | 27 ± 8 | 25 ± 8 | 0.02 |
PaCO2 (mm Hg), mean ± SD | 46 ± 9 | 48 ± 11 | 0.3 |
Listing year ⩾ 1996, % | 30 | 37 | 0.3 |
Nutritional intervention, % | 31 | 22 | 0.3 |
The sensitivity analysis tested the effect on the hazard ratios of two extreme situations concerning the patients removed from the transplant list for reasons other than death or transplant, such as transfer to another center and voluntary removal. In one case we assumed all patients in the “other” category experienced the event (death) after the longest survival time of any individual in the data set, that is, the censoring time is replaced by the longest survival time (1,783 d). In the second case, we assumed the “other” patients died on the day of censoring. In both situations, the hazard ratios for patients with FEV1 of less than 30% predicted with PaCO2 of more than 50 mm Hg decreased compared with the original analysis but were still significantly greater than 1.0. For the other variables, there was no substantial change in the hazard ratios. These data suggest that informative censoring did not influence our model (results shown in Tables E2A and E2B).
Because of the possibility of confounding by center, we performed the univariate analyses on the data from each center with sufficient numbers of patients (all centers other than Children's Hospital of Philadelphia). There were no substantial differences between centers in the HR for death using FEV1 of 30% predicted or less and PaCO2 of 50 mm Hg or greater. At one site (Stanford), referral from an accredited CF center showed a trend toward an increased risk of death (HR, 1.4; 95% CI, 0.2–11). However, only eight patients were not referred from a CF center, so the estimate is imprecise (see Table E4).
Also, the effect of missing data was assessed. The main difference between those patients excluded from the final model because of missing data and those included with complete data was a shorter waiting time (see Table E5).
Accurate prediction of the appropriate time for listing a patient with CF with end-stage lung disease is of paramount importance to minimize mortality on the transplant waiting list and to maximize the survival benefit after transplantation. An FEV1 of less than 30% predicted has been the guidepost for referral for lung transplantation, based on the landmark study published by Kerem and coworkers (3). Despite this practice, mortality while on the waiting list remains high. Subsequent studies by Liou and coworkers (7) and Mayer-Hamblett and coworkers (9) may allow better prediction of mortality among patients with CF, using multiple variables. Although the model by Liou and coworkers seems useful it deals with transplantation as an on-demand treatment, and does not consider the issues of a waiting list. The model developed by Mayer-Hamblett and coworkers (9) is more targeted at the usual time of waiting for transplantation once listed, but unfortunately is no better at predicting 2-yr mortality compared with FEV1 of less than 30% predicted alone. The goal of this study was to further identify who on the waiting list should be prioritized for transplantation.
By inclusion of data not collected by the Cystic Fibrosis Foundation Patient Registry, this multicenter study identifies some new clinical risk factors while confirming other clinical risk factors for waiting list mortality. Consistent with the findings of Kerem and coworkers (3), we observed that FEV1 of less than 30% predicted was a highly significant risk factor for death among patients with CF awaiting lung transplantation, even though the multivariate results for FEV1 of less than 30% predicted are misleading without the interaction terms. Also, to our knowledge prior studies such as that of Kerem and coworkers did not look extensively for interactions (the only potential interactions tested were between sex and percent predicted FEV1 and between sex and age), so it is not known whether additional interactions were present. We observed that hypercapnia may be an independent risk factor for death among patients with CF with better lung function. Hypercapnia was one of the main factors identified by Kerem and coworkers and that criterion was included in the International Society for Heart and Lung Transplantation guidelines (15, 16). In patients with severely impaired lung function as indicated by an FEV1 of less than 30% predicted, and thus advanced obstructive lung disease, hypercapnia seems not as important a marker for higher risk of death, and may be in fact caused by poor airflow obstruction. These data suggest a hypothesis that hypercapnia in the face of better airflow may have a fundamentally different physiologic basis.
Referral from an accredited CF center to one of the transplant centers included in this study was associated with a lower risk of death in both the univariate and multivariate analyses (although not strictly significant in the latter). These observations may reflect earlier recognition or aggressive management of advanced disease in patients referred from accredited CF centers. There are 117 accredited specialized CF care centers in the United States and 38 in Canada that offer comprehensive care to patients with CF and their families. To our knowledge, no studies have shown that outcomes of patients cared for at these centers are better than those of patients cared for by physicians without specialized training in the care of patients with CF and without access to the team approach offered at the accredited CF centers. The previous models for predicting survival also may have excluded subjects who were not cared for in CF care centers, as data from these patients may not have been included in the U.S. Cystic Fibrosis Foundation Registry that served as the source data for these models (9, 13). We also assessed whether patients not referred from CF centers may have markers of more severe disease (Table 5). Patients who are referred from nonaccredited CF centers do have lower percent predicted FEV1 values. This may reflect later recognition of more severe disease by physicians not as familiar with CF.
Our observation that nutritional intervention, defined as use of an appetite stimulant (i.e., megestrol acetate [Megace]), placement of a gastrojejunostomy tube, or use of parenteral nutrition, is associated with a higher risk of death while awaiting lung transplantation is, at first glance, somewhat surprising. Interestingly, use of these interventions does not correlate with body mass index (which was not found to be associated with a higher risk of death) in this study as well as in another (9). One hypothetical explanation of these data is that requirement for additional nutritional support is a marker of a more ill patient. Such a situation may occur if a patient has increased nutritional demand due to, for example, a higher level of ongoing inflammation from lung disease. However, we observed no differences in markers of more severe disease in patients who received a nutritional intervention and those who did not with respect to markers of more severe disease (see Table E6).
Year of listing was also an important factor for survival on the waiting list, with listing during or after 1996 reducing the risk of death. There are a number of potential explanations for this observation. It may be a function of improved survival over the last 10 yr in general for patients with CF. Alternatively, it may merely reflect physician practice to list patients for transplant earlier (when they are presumably less ill) with recognition of increased waiting time for suitable donor organs. The study by Liou and coworkers (8) showed improved posttransplant survival in more recent years as well.
It was surprising that none of the microbiological data were predictive of survival (i.e., presence of B. cepacia or methicillin-resistant Staphylococcus aureus). This may have to do with listing practices at the U.S. centers (in more recent years no one with B. cepacia on sputum culture is listed for transplant) whereas in Toronto, where B. cepacia patients are eligible for transplant listing, the overall waiting list mortality is low (15%).
In the newly implemented United Network for Organ Sharing (UNOS; Richmond, VA) allocation model (http://www.unos.org/SharedContentDocuments/Calculation_Guide.pdf), candidates for lung transplantation are assigned priority for transplant based on a lung allocation score (17). The factors used to predict the risk of death on the lung transplant waiting list included FVC, pulmonary artery systolic pressure, O2 required at rest, age, body mass index, insulin-dependent diabetes, functional status (New York Heart Association class), 6-min walk distance, and ventilator use. The actual Cox proportional hazard model used to determine these risk factors has not been published. The risk factors used in their model are not specific to CF but are used more broadly for all disease populations. Also, the UNOS allocation model is based on a retrospective evaluation of variables collected by UNOS for the purpose of listing patients and has not been prospectively validated. Surprisingly, FEV1, traditionally the benchmark for prediction of death among patients with CF being considered for transplant, is not included in this allocation model. Hypoxemia, body mass index, presence of CF- related diabetes mellitus, and 6-min walk distance were all evaluated in this study and not found to be predictors of increased risk of death. Pulmonary artery systolic pressure in our study had too many missing values to lead to a meaningful conclusion, while functional status and ventilator use were not assessed in this study. Thus, we cannot compare our data with the UNOS model.
The main limitation of our study is missing data. Table E2 shows a comparison of patients with missing data and not included in the final multivariate model and those with complete data and included in the final model. Because of missing data we could not include some potential risk factors for death, such as the presence of pulmonary artery hypertension and reduced right ventricular ejection fraction. This study evaluated a selected population: those listed for transplantation. The rationale was to identify the factors that led to death before transplantation as a way to identify patients with CF who might require a higher priority for lung allocation. Differences in listing criteria for lung transplant between the centers were carefully considered. For instance, Toronto has a large population of patients with B. cepacia and lists and transplants patients with this organism, whereas most U.S. transplant centers, including the ones in this study, do not list patients with B. cepacia. Also, Toronto has an allocation system that is partially based on medical urgency and with generally shorter waiting times (median, < 6 mo). It is, therefore, possible that patients listed at Toronto have different risk factors for waiting list mortality compared with the other centers in the study. However, no differences were found among the listed patients at the four sites. We therefore repeated a Cox proportional hazards model including only patients from Toronto, and found that B. cepacia was not a significant risk factor for death in this population.
The results of our study do not agree with the results of a single-center study by Vizza and coworkers (14), in which a Cox proportional hazards model identified short 6-min walk distance, higher pulmonary artery systolic pressure, and diabetes mellitus as significant risk factors for death while waiting for lung transplantation. These differences may have occurred for a few reasons. First, we had a significant number of missing echocardiograms and thus pulmonary hypertension could not be included in our model. Second, our definition of diabetes mellitus included patients who were taking insulin or oral hypoglycemic medications at any point in time before listing and not taking concomitant corticosteroids, so our classification of patients with CF-related diabetes mellitus is different. Third, in our study short 6-min walk distance showed no significant association with a higher risk of death (HR, 0.98; 95% CI, 0.96–1.0) in either our univariate or multivariate analyses. It is possible this may be related to differences in rehabilitation practice at the centers included in our study.
As mentioned above, the possibility of informative censoring exists, where the reason for removal from the list is not independent of the outcome (in this case, death). The results of our sensitivity analysis are reassuring in that informative censoring in the group of patients removed from the list for “other” reasons does not occur.
Another possible limitation is the inclusion of patients with heart–lung, lung–liver, or heart–lung–liver transplants. To assess the possible impact on our results from including such patients, the multivariate modeling was performed only with patients who had lung transplantation only (excluding heart–lung and heart–lung–liver transplants). There was no significant difference in the hazard ratios for any of the variables in the final model.
Under the previous time-based allocation system employed in the United States, patients waited up to 2 to 3 yr from listing to transplantation. The system has changed and now allocation of lungs is based on severity of disease (priority-based system) rather than waiting time. Thus, improved prediction models are critical in determining anticipated mortality among patients with CF being considered for lung transplantation. On the basis of these data, the presence of hypercapnia in the setting of better lung function and the requirement for nutritional interventions are additional factors that should be considered in developing future models to be used in allocating organs to the CF population.
1. | Charman SC, Sharples LD, McNeil KD, Wallwork J. Assessment of survival benefit after lung transplantation by patient diagnosis. J Heart Lung Transplant 2002;21:226–232. |
2. | Zuckerman JB, Kotloff RM. Lung transplantation for cystic fibrosis [review]. Clin Chest Med 1998;19:535–554, vii. |
3. | Kerem E, Reisman J, Corey M, Canny GJ, Levison H. Prediction of mortality in patients with cystic fibrosis. N Engl J Med 1992;326:1187–1191. |
4. | Arcasoy SM, Kotloff RM. Lung transplantation. N Engl J Med 1999;340:1081–1091. |
5. | Geller DE, Kaplowitz H, Light MJ, Colin AA; Scientific Advisory Group, Investigators, and Coordinators of the Epidemiologic Study of Cystic Fibrosis. Allergic bronchopulmonary aspergillosis in cystic fibrosis: reported prevalence, regional distribution, and patient characteristics. Chest 1999;116:639–646. |
6. | Hosenpud JD, Bennett LE, Keck BM, Edwards EB, Novick RJ. Effect of diagnosis on survival benefit of lung transplantation for end-stage lung disease. Lancet 1998;351:24–27. |
7. | Liou TG, Adler FR, Cahill BC, FitzSimmons SC, Huang D, Hibbs JR, Marshall BC. Survival effect of lung transplantation among patients with cystic fibrosis. JAMA 2001;286:2683–2689. |
8. | Liou TG, Adler FR, Huang D. Use of lung transplantation survival models to refine patient selection in cystic fibrosis. Am J Respir Crit Care Med 2005;171:1053–1059. |
9. | Mayer-Hamblett N, Rosenfeld M, Emerson J, Goss CH, Aitken ML. Developing cystic fibrosis lung transplant referral criteria using predictors of 2-year mortality. Am J Respir Crit Care Med 2002;166:1550–1555. |
10. | Belkin RA, Singer LG, Henig NR. Height, Aspergillus and MRSA: potential predictors of death in CF patients on the lung transplant waiting list [abstract]. Pediatr Pulmonol 2001;32:317. |
11. | Belkin RAB, Singer LG, Rubenstein RC, Xie SX, Kotloff RM, Lipson DA, Henig NR. Elevated PaCO2 as a risk factor for death in CF patients awaiting lung transplantation [abstract]. Pediatr Pulmonol 2004;38(Suppl. 27):317. |
12. | Collett D. Informative censoring. In: Collett D, editor. Modelling survival data in medical research. London: Chapman & Hall; 2003. p. 319. |
13. | Liou TG, Adler FR, Fitzsimmons SC, Cahill BC, Hibbs JR, Marshall BC. Predictive 5-year survivorship model of cystic fibrosis. Am J Epidemiol 2001;153:345–352. |
14. | Vizza CD, Yusen RD, Lynch JP, Fedele F, Alexander Patterson G, Trulock EP. Outcome of patients with cystic fibrosis awaiting lung transplantation. Am J Respir Crit Care Med 2000;162:819–825. |
15. | Yankaskas JR, Mallory GB Jr. Lung transplantation in cystic fibrosis: consensus conference statement. Chest 1998;113:217–226. |
16. | Maurer JR, Frost AE, Estenne M, Higenbottam T, Glanville AR; International Society for Heart and Lung Transplantation, American Thoracic Society, American Society of Transplant Physicians, European Respiratory Society. International guidelines for the selection of lung transplant candidates. J Heart Lung Transplant 1998;17:703–709. |
17. | United Network for Organ Sharing. Organ distribution. Richmond, VA: United Network for Organ Sharing; 2004. Policies 3.5–3.9. Available from: http://www.unos.org/policiesandbylaws/policies.asp?resources=true (accessed January 6, 2006). |
18. | Kotloff RM, Zuckerman JB. Lung transplantation for cystic fibrosis: special considerations. Chest 1996;109:787–798. |
19. | Milla CE, Warwick WJ. Risk of death in cystic fibrosis patients with severely compromised lung function. Chest 1998;113:1230–1234. |
20. | Doershuk CF, Stern RC. Timing of referral for lung transplantation for cystic fibrosis: overemphasis on FEV1 may adversely affect overall survival. Chest 1999;115:782–787. |
21. | American Thoracic Society. ATS statement: guidelines for the six-minute walk test. Am J Respir Crit Care Med 2002;166:111–117. |
22. | Kadikar A, Maurer J, Kesten S. The six-minute walk test: a guide to assessment for lung transplantation. J Heart Lung Transplant 1997;16:313–319. |
23. | Nixon PA, Orenstein DM, Kelsey SF, Doershuk CF. The prognostic value of exercise testing in patients with cystic fibrosis. N Engl J Med 1992;327:1785–1788. |
24. | Sharma R, Florea VG, Bolger AP, Doehner W, Florea ND, Coats AJ, Hodson ME, Anker SD, Henein MY. Wasting as an independent predictor of mortality in patients with cystic fibrosis. Thorax 2001;56:746–750. |
25. | McAllister TA, Mocan H, Murphy AV, Beattie TJ. Antibiotic susceptibility of staphylococci from CAPD peritonitis in children. J Antimicrob Chemother 1987;19:95–100. |
26. | Egan TM, Bennett LE, Garrity ER, Grover FL, Ring WS, Robbins RC, Trulock E, Wood DE. Predictors of death on the UNOS lung transplant waiting list: results of a multivariate analysis. J Heart Lung Transplant 2001;20:242. |
27. | Hayllar KM, Williams SG, Wise AE, Pouria S, Lombard M, Hodson ME, Westaby D. A prognostic model for the prediction of survival in cystic fibrosis. Thorax 1997;52:313–317. |
28. | Augarten A, Akons H, Aviram M, Bentur L, Blau H, Picard E, Rivlin J, Miller MS, Katznelson D, Szeinberg A, et al. Prediction of mortality and timing of referral for lung transplantation in cystic fibrosis patients. Pediatr Transplant 2001;5:339–342. |
29. | Rosenecker J, Hofler R, Steinkamp G, Eichler I, Smaczny C, Ballmann M, Posselt HG, Bargon J, von der Hardt H. Diabetes mellitus in patients with cystic fibrosis: the impact of diabetes mellitus on pulmonary function and clinical outcome. Eur J Med Res 2001;6:345–350. |
30. | Ledson MJ, Gallagher MJ, Jackson M, Hart CA, Walshaw MJ. Outcome of Burkholderia cepacia colonisation in an adult cystic fibrosis centre. Thorax 2002;57:142–145. |
31. | Frangolias DD, Mahenthiralingam E, Rae S, Raboud JM, Davidson AG, Wittmann R, Wilcox PG. Burkholderia cepacia in cystic fibrosis: variable disease course. Am J Respir Crit Care Med 1999;160:1572–1577. |
32. | Davis PB. Clinical pathophysiology and manifestations of lung disease. In: Yankaskas JR, Knowles MR, editors. Cystic fibrosis in adults. Philadelphia, PA: Lippincott-Raven; 1999. pp. 45–67. |
33. | Demko CA, Byard PJ, Davis PB. Gender differences in cystic fibrosis: Pseudomonas aeruginosa infection. J Clin Epidemiol 1995;48:1041–1049. |
34. | Goss CH, Otto K, Aitken ML, Rubenfeld GD. Detecting Stenotrophomonas maltophilia does not reduce survival of patients with cystic fibrosis. Am J Respir Crit Care Med 2002;166:356–361. |
35. | Schaedel C, de Monestrol I, Hjelte L, Johannesson M, Kornfalt R, Lindblad A, Strandvik B, Wahlgren L, Holmberg L. Predictors of deterioration of lung function in cystic fibrosis. Pediatr Pulmonol 2002;33:483–491. |
36. | Huang NN, Schidlow DV, Szatrowski TH, Palmer J, Laraya-Cuasay LR, Yeung W, Hardy K, Quitell L, Fiel S. Clinical features, survival rate, and prognostic factors in young adults with cystic fibrosis. Am J Med 1987;82:871–879. |
37. | Rosenfeld M, Davis R, FitzSimmons S, Pepe M, Ramsey B. Gender gap in cystic fibrosis mortality. Am J Epidemiol 1997;145:794–803. |
38. | Lewis P. CF epidemiology: population perspective and impact [Symposium Session Summaries]. Annual North American Cystic Fibrosis Conference, 2000. |
39. | Gaskin K, Gurwitz D, Durie P, Corey M, Levison H, Forstner G. Improved respiratory prognosis in patients with cystic fibrosis with normal fat absorption. J Pediatr 1982;100:857–862. |
40. | Schechter MS, Shelton BJ, Margolis PA, Fitzsimmons SC. The association of socioeconomic status with outcomes in cystic fibrosis patients in the United States. Am J Respir Crit Care Med 2001;163:1331–1337. |
41. | Fraser KL, Tullis DE, Sasson Z, Hyland RH, Thornley KS, Hanly PJ. Pulmonary hypertension and cardiac function in adult cystic fibrosis: role of hypoxemia. Chest 1999;115:1321–1328. |
42. | Corey M, McLaughlin FJ, Williams M, Levison H. A comparison of survival, growth, and pulmonary function in patients with cystic fibrosis in Boston and Toronto. J Clin Epidemiol 1988;41:583–591. |
43. | Walters S, Britton J, Hodson ME. Hospital care for adults with cystic fibrosis: an overview and comparison between special cystic fibrosis clinics and general clinics using a patient questionnaire. Thorax 1994;49:300–306. |
44. | Beker LT, Russek-Cohen E, Fink RJ. Stature as a prognostic factor in cystic fibrosis survival. J Am Diet Assoc 2001;101:438–442. |