Published ahead of print on January 21, 2005, doi:10.1164/rccm.200409-1248OC
© 2005 American Thoracic Society doi: 10.1164/rccm.200409-1248OC
Graft Ischemic Time and Outcome of Lung TransplantationA Multicenter AnalysisDivision of Pulmonary Medicine and Thoracic Surgery, Beaujon Hospital, Clichy; Division of Thoracic Surgery, Marie-Lannelongue Hospital, Le Plessis-Robinson; Division of Thoracic Surgery, Haut L'Evêque Hospital, Bordeaux; Division of Pulmonary Medicine and Intensive Care Unit, Foch Hospital, Suresnes; Division of Pulmonary Medicine, Louis Pradel Hospital, Lyon; Division of Pulmonary Medicine and Thoracic Surgery, Laënnec Hospital, Nantes; Division of Pulmonary Medicine and Thoracic Surgery, Albert Michalon Hospital, Grenoble; and Division of Thoracic Surgery, Ste. Marguerite Hospital, Marseille, France Correspondence and requests for reprints should be addressed to Gabriel Thabut, M.D., Service de Pneumologie et Réanimation, Hôpital Beaujon, 100 avenue du Général Leclerc, 92110 Clichy, France. E-mail: gabriel.thabut{at}bjn.ap-hop-paris.fr
Rationale: The effect of graft ischemic time on early graft function and long-term survival of patients who underwent lung transplantation remains controversial. Consequently, graft ischemic time has not been incorporated in the decision-making process at the time of graft acceptance. Objectives: To investigate the relationship between graft ischemic time and (1) early graft function and (2) long-term survival after lung transplantation. Measurements and main results: The data from 752 patients who underwent single lung transplantation (n = 258), bilateral lung transplantation (n = 247), and heart-lung transplantation (n = 247) in seven French transplantation centers during a 12-year period were reviewed. Independent data quality control was done to ensure the quality of the collected variables. Mean graft ischemic time was 245.8 ± 96.4 minutes (range 50–660). After adjustment on 11 potential confounders, graft ischemic time was associated with the recipient PaO2/FIO2 ratio recorded within the first 6 hours and with long-term survival in patients undergoing single or double lung transplantation but not in patients undergoing heart-lung transplantation. The relationship between graft ischemic time and survival appears to be of cubic form with a cutoff value of 330 minutes. These results were unaffected by the preservation fluid employed. Conclusions: The results of this large cohort of patients suggest a close relationship between graft ischemic time and both early gas exchange and long-term survival after single and double lung transplantation. Such relationship was not found in patients undergoing heart-lung transplantation. The expected graft ischemic time should be incorporated in the decision-making process at the time of graft acceptance.
Key Words: lung transplantation • prognosis • acute lung injury • reperfusion injury Ischemia/reperfusion injury is a frequent complication after lung transplantation (LT) (1–3). In particular, the most severe form may lead to primary graft failure, which is responsible for a significant morbidity and mortality (4). On a theoretical basis, an influence of graft ischemic time on the occurrence of ischemia/reperfusion injury is expected. However, although data from experimental LT and other solid-organ transplantations support a deleterious role of ischemic time, its influence on early and late outcome in clinical LT remains debated. Some studies found that ischemic time has no impact on early graft function or survival (4–10), whereas others suggested that ischemic time per se or in combination with other donor characteristics has a negative influence on the outcome after LT (11–17). By consequence, the place of ischemic time in the decision-making process at the time of graft proposal is unclear. Answering the question of the role of ischemic time on graft outcome has potential clinical implications. Demonstrating that the role of ischemic time on outcome is minor could allow for improved geographic sharing of organs and increased donor lung use. On the other hand, if ischemic time adversely affects graft outcome, the tolerance of long ischemic time could lead to an increased risk of ischemia/reperfusion injury that should be weighed at the time of graft proposal. Using a large database, the aim of this study was to answer the following questions: is there an association between graft ischemic time and the outcome of LT in terms of early graft function and long-term survival? If yes, is there a threshold above which recipients face a sharp increase in mortality? Part of this study has been previously presented in abstract form (15).
Patients We used data collected as part of the study of the predictive factors of early mortality after LT designed by the French Lung Transplantation Group (15). All adult patients who underwent a first single LT (SLT), bilateral LT (BLT), or heart-LT (HLT) in seven centers in France between January 1987 and December 1998 were included in this cohort. Among the 785 patients included, graft ischemic time was reported in 752 patients who form the basis of the present study. This study has been made in accordance with good clinical practices and with the recommendations concerning human research that are contained in the Declaration of Helsinki.
Data Quality Control
Definition of Graft Ischemic Time
Studied Endpoints
Statistical Analysis
Best PaO2/FIO2 within the first 6 hours.
Long-term survival.
Determination of a cutoff value of graft ischemic time to discriminate long-term survivors.
Role of the Funding Source
The current study included 752 patients whose main characteristics are summarized in Table 1. The types of performed procedures according to each transplantation period and the characteristics of the centers regarding the type and the number of transplantations are given in Table E1 and Figure E1 of the online supplement. Mean graft ischemic time for the whole cohort was 245.8 ± 96.4 minutes (median 240; range 50–660). It was 236.5 ± 83.8 minutes (median 240; range 50–510) for SLT, 317.4 ± 90.9 minutes (median 315; range 100–660) for BLT, and 184 ± 60.9 minutes (median 180; range: 60–360) for HLT (analysis of variance for comparisons between the three groups: p = 0.0001).
PaO2/FIO2 Ratio Within the First 6 hours after Transplantation Median of PaO2/FIO2 ratio within the first 6 hours was 261 (25th–75th percentile: 150–374) for SLT, 339 (183–428) for BLT, and 292 (152–390) for HLT (analysis of variance for comparisons between the three groups: p = 0.02). After adjustment on the 11 variables described in METHODS, the parameter estimate for a 1-hour increase in graft ischemic time was –13.3 (95% confidence interval [CI]: –26.0 to –0.6; p = 0.04) for SLT, –20.8 (–35.3 to –6.3; p = 0.005) for BLT, and –7.1 (–17.3 to +2.9; p = 0.16) for HLT. In other words, after adjustment for all potential confounding factors, a 1-hour increase in graft ischemic time was associated with an absolute decrease in PaO2/FIO2 ratio of 13.3, 20.8, and 7.1 for SLT, BLT, and HLT, respectively. We did not find significant differences in early gas exchange when comparing extracellular and intracellular preservation solutions.
Long-term Survival
Association Between Graft Ischemic Time and Survival for Each Transplantation Procedure
BLT. Using Cox regression analysis, unadjusted hazard ratio for long-term survival was 1.13 (95% CI: 1.08–1.18; p = 0.0001) for each 1-hour increment of graft ischemic time. After adjustment for all potential confounding variables listed in METHODS, the adjusted hazard ratio was 1.19 (95% CI: 1.04–1.38; p = 0.01). The use of fractional polynomials suggested that a cubic modeling of graft ischemic time results in a better fit of the model than a linear modeling (p = 0.01) (Figure 1). The use of cubic smoothing splines has led to similar conclusions (data not shown). Here again, these results seem robust because a polynomial modeling was found statistically better than a linear modeling in 81% of the 1,000 bootstrap samples. Table 2 illustrates the implication of this modeling in terms of the relative risk of death according to graft ischemic time. Using threshold determination, repeated on the 1,000 bootstrap samples, a cutoff of 330 minutes (95% CI: 180–360) was found to best discriminate between long-term survivors and nonsurvivors. Patients with graft ischemic time higher than this threshold (n = 110) had a relative risk of death of 1.55 (95% CI: 1.15–2.08; p = 0.004). Figure 2 displayed the survival of patients according to this cutoff. No formal violation of the proportional hazard assumption was found. As was the case in SLT recipients, the effect of graft ischemic time on the relative risk of death seems to peak in the first year after LT and to wear off quickly thereafter (Figure E2). When the post-transplant period was portioned into different segments, the relative risks of death associated with a 6-hour graft ischemic time compared with a 2-hour graft ischemic time were 1.98 (1.69–2.32) in the 90 postoperative days, 1.45 (0.96–2.18) between 90 days and 1 year, and 1.26 (0.74–2.14) after 1 year.
HLT.
Influence of the transplant period and of the preservation solution on the results.
Quality Control
The main results of this study are the following: in patients undergoing SLT or BLT, (1) graft ischemic time is associated with early gas exchange and long-term mortality; (2) the relationship between graft ischemic time and long-term survival is not of linear but of cubic shape; (3) there is a steep increase in the relative risk of death when graft ischemic time is more than 6 hours; and (4) the harmful effect of graft ischemic time on the relative risk of death peaks in the first year after LT and vanishes as time elapses. In patients who underwent HLT, we failed to detect a relationship between graft ischemic time and both early postoperative gas exchange and long-term survival. Although there is much theoretical rationale supporting a harmful effect of graft ischemic time on the outcome of LT, conflicting results have been observed in the clinical setting. Many clinical case reports exist in the literature of excellent outcomes despite the markedly prolonged pulmonary graft ischemic times; in some cases, beyond 11 hours. Several studies aiming to analyze the outcome of patients with long ischemic time found that graft ischemic time did not adversely affect early graft function or survival (4–10). Moreover, until the most recent report, the data from the International Society of Heart and Lung Transplantation (ISHLT) Registry did not indicate that ischemic time per se had an influence on the 1-year or 5-year survival. Conversely, other groups, including ours, have found a negative influence of graft ischemic time on early graft function (11–15). In 1999, the ISHLT Registry (5,052 patients) identified ischemic time as a risk factor of mortality in combination with donor age (17) and in the 2003 report (12,532 patients), ischemic time per se was found as a risk factor of 1-year mortality (16, 17). We feel that several factors may account for the discrepancies observed among these studies. First, most studies were of limited sample size, resulting in limited statistical power. Second, in most cases, patients were put into two to three groups according to arbitrary cutoffs of graft ischemic time, further diminishing the power of the study. Third, most studies did not take into account potential confounding factors by means of multivariable analysis. Fourth, data coming from the ISHLT Registry must be interpreted within the limitations of a voluntary and uncontrolled database. To overcome these limitations, we have designed a large cohort of all consecutive patients who underwent a first LT in seven centers in France. A data quality control ensured the reliability of the collected data. Approximately one third of the transplantation procedures in this study were SLT, one third BLT, and one third HLT. Because each of these procedures markedly differed regarding graft ischemic time and early oxygenation, they were analyzed separately. We made a special effort to better assess the true relationship between graft ischemic time and long-term survival. We have found that, in patients who underwent SLT or BLT, the relationship between graft ischemic time and the relative risk of death was of cubic form. In other words, and as indicated in Figure 1, although patients who undergo LT experience a slight increase in mortality risk when graft ischemic increases from 1 to 4 hours, there is a rapid increase in the relative risk of death when graft ischemic time is longer than 6 hours. We also found that the harmful effect of graft ischemic time on the relative risk of death peaks in the first years after LT. This result is in line with the results from the ISHLT Registry in which graft ischemic time has been found to be associated with 1-year but not with 5-year survival. It could suggest that graft ischemic time may affect early- but not late-graft function. We failed to detect a significant relationship between graft ischemic time and the outcome of HLT. Several reasons might explain these findings: (1) such a relation does not exist for patients undergoing HLT; (2) our analysis lacks the power to detect a relationship; or (3) the rather short graft ischemic time observed in HLT patients in our cohort does not allow us to detect a relationship. In particular, only three HLT patients had graft ischemic time longer than 330 minutes, the threshold found to best discriminate between survivors and nonsurvivors in the SLT and BLT cohorts. Several groups, including ours, have found that the type of preservation fluid could influence the occurrence of ischemia/reperfusion injury. Namely, these studies suggest that the use of extracellular-type preservation fluid protects the graft against ischemia/reperfusion injury (14, 21). It could be argued that when using extracellular-type preservation solutions, the influence of graft ischemic time on outcome would be less pronounced. In the present study, no association was found between the use of extracellular preservation fluid and early gas exchange. Moreover, the type of preservation fluid did not affect the relationship between graft ischemic time and long-term survival. Despite improvement in the quality of preservation, our opinion is that the potential deleterious role of graft ischemic time cannot be neglected. Several limitations of our study could be underlined. Obviously, its retrospective design could bias the results. We did our best to account for confounding factors by the use of multivariate analysis. Moreover, we compared the characteristics of donors and recipients according to the duration of graft ischemic time (longer than or shorter than 6 hours) and found that they did not differ (data not shown). However, we cannot exclude that other factors not studied in our analysis or unidentified may have biased our results. That our survival rates are somewhat lower than those reported in the International Registry (16) could also be viewed as another limitation of this study. In their simplest forms, all usual regression models assume that for a certain scale of the dependent variable (the log of the instantaneous risk of death for the Cox model), each predictor variable is linearly related to it. Although this approach is easy to implement and to understand, it may result in oversimplification of the true phenomenon under observation and even worse, in misleading conclusions. Many techniques are available to better assess the true relationship between a continuous predictor and an outcome variable, including fractional polynomials, smoothing splines or other smoothing functions. The use of these techniques in the present study suggests that the shape of the relationship between graft ischemic time and long-term survival is not of linear but of cubic form. These results also suggest that SLT and BLT patients face a rapid increase in the risk of death when graft ischemic time is longer than 330 minutes. The pitfall of these methods is the risk of overfitting (i.e., to model not only the true phenomenon under observation, but also some noise). We used bootstrap resampling techniques to deal with these limitations. Because a polynomial model fits better than a linear one in most bootstrap samples and the better fit was for polynomial modeling after removing the most influential patients, we feel confident in the present results. The proportional hazard assumption is a major assumption underlying the Cox model. It implies that the ratio between the risk of death of two patients is constant regardless of time. For instance, based on this model, a SLT patient with a 4-hour graft ischemic time is expected to face a 13% increase in its risk of death compared with a patient with a 2-hour graft ischemic time, whatever the time elapsed from LT may be. More subtle assessment of the relative risk of death suggests that the harmful effect of graft ischemic time peaks in the early postoperative period and vanishes over time. The mechanism underlying the influence of graft ischemic time on outcome of LT is not fully understood. As shown by several experimental and clinical studies, including ours (12, 14, 22, 23), graft ischemic time is associated with the occurrence of early graft dysfunction, which is believed to be mainly an expression of lung ischemia/reperfusion injury. Because the occurrence of early graft dysfunction increases mortality (4, 14, 15, 24, 25), we hypothesize that the impact of graft ischemic time on survival is mediated by an increased incidence of early graft dysfunction. This hypothesis is consistent with the results of the present study suggesting that the impact of graft ischemic time on the relative risk of death peaks in the first year after LT. To this date, an association has been found between 1-year survival after LT and several variables which are either donor-related (such as age), surgery-related (such as retransplantation), or recipient-related (such as underlying disease, CMV mismatch or ventilator-dependency). According to our results graft ischemic time should be added to the list of risk factors of mortality. However, a question arises: is the graft ischemic time predictable in a given patient and do our results have a potential clinical utility? In other words, could the expected graft ischemic time be integrated in the decision-making process at the time of graft proposal? In France, when a lung graft is proposed by the national organ sharing organization, we are able to some extent to estimate the graft ischemic time, taking into account several logistic parameters (geographic origin of the donor, expected hour of the beginning of harvesting, and time needed by the recipient to reach the transplant center) and operative considerations (type of transplantation, expected difficulty in the dissection). Based on the collected information, a graft ischemic longer than 6 hours may be expected in some patients. In that case, our opinion is that, for a given patient, the risk of dying while on a waiting list should be balanced against the added mortality risk represented by a graft ischemic time exceeding 6 hours, particularly when other risk factors of mortality are also present at the time of graft proposal. For instance, the added risk could be judged too high in an emphysematous patient in a stable condition, but not in a fibrotic patient with a deteriorating condition. From a clinical point of view, another issue that should be stressed is that graft ischemic time longer than 6 hours is not associated in all cases with the development of an ischemia/reperfusion injury in the recipient (this fact explains why clinical success has been reported after a long ischemic time). Conversely, severe forms of primary graft failure have been observed after ischemic time of less than 2 hours. In summary, we found a significant relationship between graft ischemic time and the outcome after SLT and BLT. Moreover, the risk of death is sharply increased when graft ischemic time exceeds 6 hours. These findings suggest that the expected graft ischemic time should be included in the decision-making process at the time of graft acceptance.
Supported by Etablissement français des Greffes (EFG). This article has an online supplement, which is accessible from this issue's table of contents at www.atsjournals.org Conflict of Interest Statement: G.T. does not have a financial relationship with a commercial entity that has an interest in the subject of this manuscript; H.M. does not have a financial relationship with a commercial entity that has an interest in the subject of this manuscript; J.C. does not have a financial relationship with a commercial entity that has an interest in the subject of this manuscript; P.D. does not have a financial relationship with a commercial entity that has an interest in the subject of this manuscript; C.D. does not have a financial relationship with a commercial entity that has an interest in the subject of this manuscript; J.-F.V. does not have a financial relationship with a commercial entity that has an interest in the subject of this manuscript; M.S. does not have a financial relationship with a commercial entity that has an interest in the subject of this manuscript; P.L. does not have a financial relationship with a commercial entity that has an interest in the subject of this manuscript; G. L. does not have a financial relationship with a commercial entity that has an interest in the subject of this manuscript; M.B. does not have a financial relationship with a commercial entity that has an interest in the subject of this manuscript; J.-F.M. does not have a financial relationship with a commercial entity that has an interest in the subject of this manuscript; A.H. does not have a financial relationship with a commercial entity that has an interest in the subject of this manuscript; P.D. does not have a financial relationship with a commercial entity that has an interest in the subject of this manuscript; C.P. does not have a financial relationship with a commercial entity that has an interest in the subject of this manuscript; D.B. does not have a financial relationship with a commercial entity that has an interest in the subject of this manuscript; M.R.-G. does not have a financial relationship with a commercial entity that has an interest in the subject of this manuscript. Received in original form September 21, 2004; accepted in final form January 5, 2005
This article has been cited by other articles:
|
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
TOP
ABSTRACT
METHODS
2 of the log-rank test. Analyses were performed using STATA 8.1 for Windows (Stata Corp., College Station, TX). Further details of the statistical analysis can be found in the online supplement. 
