Outcome of heart–lung and bilateral sequential lung transplantation for cystic fibrosis: a UK national study

Cystic fibrosis (CF) is the most common, life-threatening, inherited disease in Caucasians. It affects 7,500 patients at any time in the UK 1 and ∼30,000 patients in the USA, with 10 million Americans being symptomless carriers of the defective CF gene 2. Despite the multisystem nature of the disease, pulmonary complications are the cause of death in >95% of CF patients 3. Currently, transplantation is the only effective treatment for CF patients with respiratory failure 4–6 and constitutes a third of the bilateral sequential lung transplantation (BSLT) procedures in children and adults reported to the International Society for Heart and Lung Transplantation 7, 8. CF patients are the largest single group of patients referred for transplantation with end-stage suppurative lung disease 9. Currently, as many as 40–50% of CF patients die on the waiting list due to a paucity in the supply of donor lungs 6.

Survival benefits and pulmonary function improvement for CF patients have been reported separately after heart–lung transplantation (HLT) 10–12 and BSLT 13, 14. Initial results for pulmonary transplantation from the UK national audit have been reported in a preliminary form 15. In this study, the current authors have investigated whether the type of operation (BSLT or HLT) affects the outcome following transplantation for CF, using the UK national audit database.

METHODS

The UK Cardiothoracic Transplant Audit is a national prospective cohort study that has been collecting data since April 1995 on all intrathoracic transplantation procedures in the UK. The audit collects data when patients are registered on the national waiting list for heart and/or lung transplantation, at the time of transplantation and at selected (90 days and yearly thereafter) follow-up periods.

In the current study, adult CF patients (aged ≥16 yrs) undergoing first-time, isolated lung transplantation or HLT from cadaveric donors between July 1995 and June 2002 were analysed retrospectively. Seven en bloc double-lung transplants carried out during this period were excluded.

Survival to 4 yrs was the primary end-point of the study. Surviving patients were censored at their last reported follow-up or at 4 yrs (whichever was earlier). Secondary outcomes were time to the first treated rejection episode, number of treated infection episodes, change in forced expiratory volume in one second (FEV₁) and use of the domino (HLT) or donor heart (BSLT).

Donor and pre-transplant recipient characteristics were compared using the Chi-squared/Fisher's exact tests (categorical data) or the Kruskal-Wallis test (continuous variables). Fisher's exact test was chosen in preference to the Chi-squared test when >20% of the expected frequencies were less than five.

All analyses were adjusted for possible confounding variables. Differences in recipient covariates were summarised using propensity scores 16, 17. Derived using logistic regression, the propensity score is the probability of receiving HLT (rather than a BSLT), which is conditional on the recipient's risk profile. Propensity-score centiles were used to adjust the treatment effect (HLT versus BSLT) for differences in recipient covariates not accounted for by other variables included in the outcome analysis. Propensity-score quintiles have been shown to reduce 80% of the bias in the variables used to derive the score 16. Limiting the number of bias-reducing covariates included in the outcome analyses in this way is particularly relevant when the study is small and events are few. Recipient variables included in the propensity score were as follows: age (grouped as 16–22 yrs, 23–30 yrs, and >30 yrs); sex; body mass index (BMI; grouped as <17.5, 17.6–19.3, and >19.4 kg·m⁻²); difference in the recipient's weight at transplant and the ideal weight for age and sex (three categories; a difference of >15 kg, 10–15-kg difference, and <10-kg difference) 18; percentage predicted FEV₁ and forced vital capacity (FVC; <30% compared with >30% predicted for FEV₁ and <40% compared with >40% predicted for FVC); history of domiciliary oxygen support; history of previous thoracotomy; diabetes; creatinine clearance (<75 mL·min⁻¹, 75–100 mL·min⁻¹, >100 mL·min⁻¹); bilirubin; infection; and admission to hospital prior to transplantation. History of prior resuscitation and ventilation prior to transplantation were not included in the propensity-score calculation as all the events occurred in the HLT group. The ability of the model to discriminate between BSLT and HLT was evaluated using the area under the receiver-operating characteristic (ROC) curve, and the model fit was assessed using the Hosmer–Lemeshow Chi-squared test.

Cox regression 19 was used to compare survival between the two groups, with and without adjustment for centre volume (six categories; three centre categories with >40 transplants each, two centre categories with 10–15 procedures each and a sixth category of two centres who had performed less than five transplants each). The analyses were adjusted for the following: propensity-score quartiles; history of prior resuscitation; home oxygen support; ventilation prior to transplantation; diabetic status; FEV₁ and ideal weight groups (as described previously) in the recipient; donor variables; year of transplant (in two eras; July 1995–June 1999 and July 1999–June 2002); and time from listing to transplant (in five groups; ≤30 days, 31–90 days, 91–180 days, 181–365 days, and >365 days). The following donor variables were included in the model: donor age (grouped as <22 yrs, 23–34 yrs, and >35 yrs); sex; cytomegalovirus (CMV) mismatch (CMV-positive donor, CMV-negative recipient); donor past history (including history of smoking, alcohol and/or drug abuse); use of inotropic support in the donor and ischaemia time (grouped as <3 h, 3–4 h, and >4 h). Ischaemia time for BSLT was taken as the time from donor cross-clamp to reperfusion of the first lung. Missing ischaemia times were imputed, using partial data, where available. For example, if the time from organ arrival to the reperfusion was known, but the cross-clamp time was missing, the median travel time from the donor to recipient hospital was added to the recorded time.

Time to the first rejection episode was also compared using Cox regression, after adjustment for propensity-score quartiles and CMV mismatch. The proportional-hazards assumption was checked for all survival models fitted.

Post-transplant infection rates between the two groups were compared using negative binomial regression after adjusting for propensity-score quartiles, CMV mismatch and length of follow-up. Model fit was assessed using the Chi-squared test.

Waiting time to transplant and waiting-list mortality were compared in the cohort of adult CF patients (≥16 yrs) registered on the waiting list for either HLT or BSLT between July 1995 and June 2002 using the Kaplan–Meier method.

RESULTS

A total of 181 (88 BSLT and 93 HLT) first-time adult (aged ≥16 yrs) CF patients undergoing transplantation between July 1995 and June 2002 in seven centres across the UK were studied (fig. 1⇓).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 1—

Denomination of the study population in cystic fibrosis (CF) patients undergoing transplant from the total UK CF transplant population. BSLT: bilateral sequential lung transplantation; HLT: heart–lung transplantation.

In the UK, HLT was the dominant procedure until 2000, but, since then, BSLT has become more common (p = 0.03, Chi-squared test for trend).

Donor and recipient characteristics were similar across the two groups of CF patients (table 1⇓). The only differences were history of resuscitation for a sudden near-death episode, use of ambulatory oxygen support at home, creatinine clearance and ischaemia time. The numbers of patients who had suffered near-death episode(s) (p = 0.004) and required long-term domiciliary oxygen support (p = 0.001) were higher in the HLT group. Calculated creatinine clearance (calculated using the Cockroft–Gault formula 20) was also higher in the HLT group (p = 0.02), whereas ischaemia time was longer in the BSLT group (p<0.001). There were also more diabetics (p = 0.07) and more patients were receiving hospital treatment at the time of transplantation (p = 0.08) in the HLT group. Significant differences in centre preference towards the type of transplantation were noted (p<0.001). BSLT was carried out at all seven centres, but only three centres performed HLT; one of the largest centres performed only BSLT. In the three centres that performed both procedures, HLT was more common (fig. 2⇓). The model to derive propensity scores from the recipient covariates was well calibrated (lack of fit, p = 0.64), and showed good discrimination (area under the ROC curve = 0.77).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 2—

Number of bilateral sequential lung (□) and heart–lung () transplantations performed in the centres.

Survival

At analysis, 72 CF recipients had died. The actuarial 4-yr survival for the whole cohort was 56% (95% confidence interval (CI) 47–63). The median follow-up of survivors was 48 and 35 months in the HLT and BSLT groups, respectively. Unadjusted Kaplan–Meier survival revealed a tendency towards better survival in the BSLT group (p = 0.09; log rank test; fig. 3⇓). The 4-yr unadjusted survival for HLT and BSLT groups was 51% (95% CI 40–61) and 60% (95% CI 46–71), respectively. After adjustment for propensity scores, time on the waiting list, specific recipient factors, ischaemia time and donor factors, there was no significant difference in survival between the two groups (hazard ratio (BSLT versus HLT) = 0.64; 95% CI 0.35–1.18; p = 0.15). When the analysis was also adjusted for centre, the risk of death in the HLT group was reduced further (hazard ratio (BSLT versus HLT) = 0.77; 95% CI 0.29–2.06; p = 0.60). The proportional-hazards assumption was tenable for both models (test for nonproportionality, p = 0.42 and p = 0.44, respectively).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 3—

Unadjusted Kaplan–Meier survival estimates for bilateral sequential lung (-----) and heart–lung (––––) transplantation in cystic fibrosis recipients (p = 0.09). ═: 30 days.

DISCUSSION

Although Hardy et al. 22 performed the first lung transplantation in 1963, it was not until the initial experience with combined HLT in the early 1980s that replacement of lungs was successfully accomplished in humans 23–25. Techniques of isolated single- and double-lung (DLT) transplantation and BSLT were subsequently developed. CF patients suffer from respiratory failure associated with bilateral lung sepsis and, hence, require replacement of both lungs, leading to the necessity of HLT, BSLT or en bloc DLT. Isolated single-lung transplantation and contralateral pneumonectomy for CF have been described, but have only been used in rare circumstances 26. DLT has been advocated in the past, but concerns about healing led to its discontinuation 14. DLT was excluded from the current study because of the small number of transplants performed in the study period. BSLT avoids the risk of cardiac allograft vasculopathy, which may be as high as 12% after combined HLT 27. The risk of transplanting a denervated heart and its attendant concerns are also avoided 9. Re-operation rates due to bleeding may be as high as 28% in HLT 28 and initial results of HLT for CF patients were disappointing in some centres 29, while those achieved by others were more encouraging, with 5-yr survival rates up to 52% 11, 30. Early reports documented a high incidence of airway complications following BSLT, which was not seen in HLT, but the incidence of airway complications in BSLT has decreased subsequently 14. One of the attractive features of HLT is the potential domino donor heart, which is fully prepared for transplantation and from a living donor. The potential advantage of a domino donor heart over a cadaveric heart has been highlighted in a recent study 31.

In the current study, HLT and BSLT recipients were similar in their pre-transplant characteristics, although HLT recipients received home oxygen support more frequently and had an increased incidence of prior resuscitation. HLT recipients had better creatinine clearance than BSLT recipients. Also, there were more diabetics in the HLT group and proportionally more recipients in this group were receiving hospital treatment at the time of transplant, although this was not statistically significant. Significant centre preferences for one procedure over the other were observed, and this was evaluated by analysing survival with and without an adjustment for centre. A significant reduction in the number of HLT procedures since 2000 was noted in this study, and it is believed that this may be due to a general decrease in transplant activity in the UK, along with changing trends in organ availability and demand.

It was found that survival to 4 yrs, unadjusted for differences in comorbidity between the two groups, tended to favour BSLT. However, when survival was adjusted for comorbidity and donor factors, no significant survival advantage was seen after either BSLT or HLT. Time to the first rejection episode and 4-yr infection-episode rates were compared and were found to be similar in both groups. Information on post-operative FEV₁ has only been collected since January 2001, and was similar across the groups, although the data were too few to allow a definitive conclusion to be drawn.

Analysis of the use of hearts from BSLT donors and domino hearts from HLT recipients revealed a higher proportion of hearts being used from domino donors, ameliorating the impact of heart usage in HLT.

Waiting times in the two groups were different in that CF patients who were registered for BSLT waited longer for their transplant and also lived longer on the waiting list. A small number of patients who were registered for HLT went on to receive a BSLT and vice versa, suggesting that HLT and BSLT have been used in a comparable way in the UK. Patients on the waiting list for HLT died significantly earlier than patients waiting for BSLT, despite being on the waiting list for a shorter duration, supporting the view that they were sicker at the time of listing and, hence, at transplantation.

The present study has limitations that would be expected of an observational multicentre study. The type of operation was not randomly assigned. The current authors have attempted to adjust for bias due to differences in risk between the two groups through the use of propensity scores and the inclusion of key donor and recipient factors in the regression models. Regression modelling can never entirely account for group imbalances, and the adjusted estimates may still be subject to some residual confounding. However, such confounding is unlikely to change the conclusion. The data used in the current study was collected as part of a multicentre audit of thoracic organ transplantation, and was not designed specifically to address the choice of operation for CF patients. As a result, the present authors were not able to provide details of cardiac allograft vasculopathy in the domino hearts, although a recent study from one of the centres involved in the audit reported results on this complication 31. Post-operative management was specific to each centre, was not recorded in detail and, therefore, could not be accounted for in the present study. Nevertheless, the data reflects the outcome of a large unselected and consecutive series of CF patients treated in the UK. Data in the audit are properly validated by means of regular computer validation programs, performed bimonthly, and case-note validation performed biannually, for accuracy and consistency. The possibility that a small difference in the outcome of the two procedures will become apparent when a much larger number of cases have been accrued cannot be excluded.