Bronchiolitis-obliterans syndrome-free survival following lung transplantation - an International Society for Heart and Lung Transplantation (ISHLT) Thoracic Transplant Registry analysis

Abstract

Background:

Lung transplant (LTx) recipients have low long-term survival and a high incidence of bronchiolitis-obliterans syndrome (BOS). However, few long-term, multicenter, and precise estimates of BOS-free survival (a composite outcome of death and BOS) incidence exist.

Methods:

This retrospective cohort study of primary LTx recipients (1994–2011) reported to the ISHLT Thoracic Transplant Registry assessed outcomes through 2012. For the composite primary outcome of BOS-free survival, we used Kaplan-Meier survival and Cox proportional hazards regression, censoring for loss-to-follow-up, end-of-study, and retransplantation. While standard Registry analyses censor at the last consecutive annual complete BOS status report, our analyses allowed for partially missing BOS data.

Results:

Due to BOS reporting standards, 99.1% of the cohort received LTx in North America. During 79,896 person-years of follow-up, single LTx (6,599/15,268; 43%) and bilateral LTx (8,699/15,268; 57%) recipients had a median BOS-free survival of 3.16 (95% CI, 2.99–3.30) years and 3.58 (95% CI, 3.53–3.72) years, respectively. Almost 90% of the single and bilateral LTx recipients developed the composite outcome within 10 years of transplantation. Standard Registry analyses “overestimated” median BOS-free survival by 0.42 years and “underestimated” the median survival after BOS by about a half-year for both single and bilateral LTx (p<0.05).

Conclusions:

A majority of LTx recipients die or develop BOS within 4 years, and very few remain alive and free from BOS at 10 years post-transplantation. Less inclusive Registry analytic methods tend to overestimate BOS-free survival. The Registry would benefit from improved international reporting of BOS and other chronic lung allograft dysfunction (CLAD) events.

INTRODUCTION

Lung transplantation (LTx) recipients have 1-, 5-, and 10-year unadjusted survival rates of 80%, 54%, and 32%, respectively.¹ Chronic lung allograft dysfunction (CLAD) causes most deaths after the first post-transplant year,^1–3 and most CLAD has an obstructive phenotype known as bronchiolitis obliterans syndrome (BOS).⁴ Obliterative bronchiolitis (OB), the histologic hallmark of BOS, consists of a fibrotic luminal obliteration of the respiratory and terminal bronchioles.⁵ The patchy nature of OB reduces the diagnostic sensitivity of transbronchial lung biopsies. Thus, clinicians use a thoroughly explored and unexplained drop in lung function to diagnose BOS, the surrogate of OB.^3,6–8 CLAD also includes a less common, restrictive phenotype known as restrictive allograft syndrome (RAS),^2,3,9–11 and some LTx recipients have a combined syndrome of RAS and BOS.³

Many studies and Registry analyses have reported the incidence or prevalence of BOS, and freedom from BOS after LTx.^1,12–15 However, incidence estimates of BOS and Kaplan-Meier estimates of freedom from BOS ignore an important and frequent competing event (i.e., death). Utilizing BOS-free survival after LTx allows the incorporation of both the death and BOS endpoints into a composite post-LTx outcome, and provides a much more robust estimate than either of its components.^16–22 Surprisingly, only a few single-center or small multi-center cohort studies^15,23,24 have reported long-term BOS-free survival rates, and these studies have not used uniform methodology or reporting.^25–27 As a result, estimates of BOS-free survival have varied widely, ranging from 34–75% at 5 years.^26,28,29 Importantly, since the remaining native lung in a single LTx recipient may influence the accuracy of the BOS diagnosis,^30,31 and single and bilateral LTx have different death and BOS incidences,^32–34 studies should be reporting these outcomes stratified by transplant procedure type. Yet, limited data exist regarding BOS-free survival rates in subgroups of LTx recipients, such as those receiving single versus bilateral LTx.^20,35,36

Given the knowledge gaps regarding BOS-free survival, we conducted this study to estimate its incidence in adult primary LTx recipients and stratified the analyses by transplant procedure type. We used more data-inclusive methods than those used in the Registry’s annual reports to assess the effects of the assumptions made by the Registry about the presence or absence of BOS and its occurrence date.^37–40 Since stakeholders may use data from the International Society for Heart and Lung Transplantation (ISHLT) Thoracic Transplant Registry as benchmarks or for clinical trial planning,^41,42 we compared effects of different approaches for handling partially missing data on BOS-free survival.

METHODS

Study design

Using data provided by the multinational ISHLT Thoracic Transplant Registry (www.ishlt.org/registries/) via a data sharing agreement, we conducted a retrospective cohort study of LTx reported during the era January 1, 1994 to December 31, 2011. We assessed follow-up data reported through December 31, 2012 (Supplement). The Washington University School of Medicine Human Research Protection Office/Institutional Review Board deemed this study exempt since it utilized de-identified data.

Data Source

National and multinational organ/data exchange organizations (“collectives”) and individual centers outside of collectives may voluntarily submit data to the Registry. Since its inception, almost 250 LTx centers have reported data to the Registry.¹ The Registry requires participating collectives/centers to report Tier 1 data elements, such as death, whereas they do not require reporting of the optional Tier 2 data elements such as BOS (Supplementary Table 1; https://www.ishlt.org/registries/heartLungRegistry.asp).

Participants and Study Eligibility Criteria

We analyzed Registry data for all adult LTx reported during the study era. Figure 1 shows the study eligibility criteria and the derivation of the study cohort. The Supplement provides details regarding methods to minimize bias and to determine “patient status” (‘living’, ‘dead’, ‘lost to follow-up’, or ‘retransplanted’) at the time of follow-up.^1,37–40

Figure 1: Derivation of study cohort from the ISHLT Thoracic Transplant Registry.

Flowchart of lung transplant recipients and application of eligibility criteria that resulted in the final study cohort that we stratified by transplant procedure type. BOS assessments “not reported” refers to sites or transplant collectives that do not report BOS information as part of their standard data reporting. *LTx= lung transplantation; Tx=transplantation; HLTx=heart-lung transplantation; BOS=bronchiolitis obliterans syndrome.

Outcomes

We used BOS-free survival as the primary outcome variable. A LTx recipient achieves the composite outcome if either BOS or death occur, whichever comes first (Supplement). Each transplant center determines if and when BOS has occurred. The Registry does not validate the BOS diagnosis. The Registry records BOS status (i.e., ‘yes’ [present] or ‘no’ [absent]) information during each post-transplant reporting period (i.e., transplant hospitalization and annually thereafter), but does not collect a specific date for BOS (Supplement). The Registry also does not formally collect data for the broader term “CLAD”.^8,39

The Registry receives death status information on all patients during each reporting period. However, unlike BOS, death reporting includes the date of death. Thus, the time of death has more precision than the time of BOS (Supplement).

Statistical Methods

This study analyzed the BOS status and event time data in ways that differ from the annual report.^1,37–40 For time to event analyses that use BOS as part (i.e., BOS-free survival) or all (freedom from BOS) of the primary outcome, the Registry had typically censored BOS data at the time of the last consecutive follow-up that reported “no BOS” (Scenario 1; Table 1).^1,37–40 For this study, we made clinically reasonable assumptions about missing data in the setting of consistent reporting (Scenario 2; Table 1). For Scenario 1 (typical Registry methodology) and Scenario 2 (this study’s methodology), we used the first report of “Yes BOS” as the endpoint for freedom from BOS analyses. Scenario 1 did not allow for missing responses, so it used consecutive reporting of BOS status to find the first report of “Yes BOS,” or it censored follow-up at the last consecutive report of “No BOS”. Scenario 1 used the midpoint of each reporting period as the date of BOS occurrence. In comparison, Scenario 2 allowed for a missing BOS response during one or more reporting periods, as long as the missing response(s) preceded a report with a “No BOS” or the first “Yes BOS” response. For the utilized “Yes BOS” events, Scenario 2 assumed that BOS occurred midway between the first “Yes BOS” and the previous “No BOS”.

Table 1:

Bronchiolitis obliterans syndrome status determination Scenarios and associated time to event calculations^*

Scenario 1

Assumptions used for the annual Registry report (Allows for missing values prior to “NO BOS” or “YES BOS”)

• BOS status determined based on:

○ Non-missing responses for BOS (YES/NO).

○ If the first annual follow-up has a missing BOS response, then the LTx recipient becomes excluded from outcome assessments.

○ Otherwise, at first missing response, BOS status is determined by the previous known responses.

• Follow-up time for time to BOS analyses:

○ Use time to the first “YES BOS” follow-up if preceded by non-missing BOS response(s) of “NO BOS.”

○ Otherwise, use time to the last consecutive “NO BOS.”

Scenario 2

Assumptions used for this study Scenario 1 + allows for missing values prior to “NO BOS” or “YES BOS”

• BOS status determined based on:

○ Non-missing responses for BOS (YES/NO),

○ A “No BOS” response preceded by missing BOS (missing BOS response[s] prior to a “NO BOS” treated as “NO BOS”), or

○ The “YES BOS” response preceded by any missing BOS response(s).

• Follow-up time for time to BOS analyses:

○ Use time to the first “YES BOS” follow-up if preceded by non-missing BOS response(s) of “NO BOS.”

○ Otherwise, use time to last “NO BOS.”

○ If missing BOS values occur immediately prior to the first “Yes BOS”, then assume BOS occurred midway between the first “Yes BOS” and the last “No BOS.”

^*Both BOS status determination Scenarios attempt to identify BOS based on the first reporting of “YES BOS.” Both Scenarios assume the LTx recipient has not been reported as dead to the Registry. Otherwise, death prior to the last follow-up used in each Scenario described above uses date of death for censoring in freedom from BOS analyses and for time of the event in BOS-free survival analyses.

Statistical Analysis

To describe the study cohort and its subgroups, we used descriptive statistics for continuous variables and counts and proportions for categorical variables. To compare baseline data among groups, we used Wilcoxon rank sum test for continuous data and the Chi-Square test for categorical data. To estimate the incidence of events after transplant, we used time-to-event Kaplan-Meier analyses, and we censored for retransplant and last follow-up reported (end of study follow-up or loss to follow-up).⁴⁰ We used the log-rank test to compare Kaplan-Meier curves between or among groups. Based on an a priori statistical analysis plan, we stratified the outcome data for LTx procedure type (i.e., single or bilateral).³⁷ We compared estimates of BOS-free survival between the Scenarios with the Kruskal-Wallis test. We used Cox proportional hazards regression to adjust BOS-free survival analyses for factors associated with the outcome. We used SAS Enterprise Guide 5.1 (SAS Institute, Inc., Cary, NC). SPSS Statistics version 23 (IBM Corp., Armonk, N.Y., USA) for statistical analyses and GraphPad Prism version 7.00 (GraphPad Software, La Jolla California USA) for plotting. We defined statistical significance a priori for a P-value of <0.05. We did not adjust the critical p-value for multiple baseline comparisons between groups.

RESULTS

LTx Recipients and Study Eligibility

The ISHLT Registry query contained data from 38,395 adults who had a primary LTx reported in the 18-year study (Figure 1). Nearly half of the recipients had LTx at a site where the site or its collective did not report BOS to the Registry. When evaluating the three geographic regions typically described by the Registry, North America had the highest BOS reporting rate (84.5%), while Europe (0.13%) and Others (8.46%) had much lower rates. The study eligibility criteria produced a final cohort of 15,268 LTx recipients eligible for analysis (Figure 1). Of these, 15,135 (99.1%) of the LTx occurred at centers in in North America. Greater than 98% of eligible LTx recipients had the required BOS and survival status data available for analysis.

Of the 15,268 LTx recipient cohort, 6,599 (43.2%) received a single LTx and 8,669 (56.8%) received a bilateral LTx. Compared to single LTx recipients, bilateral LTx recipients had a 1) younger age, 2) more frequent use of life support going into transplantation, 3) high-risk behavior donor use, 4) female donor, 5) CMV mismatch status (recipient seronegative; donor seropositive), and 6) longer total ischemic time (Table 2). During the study era, the predominant LTx procedure switched from single LTx to bilateral LTx.

Table 2:

Cohort pre-transplant characteristics, stratified by procedure type

	Single LTx (N=6599)	Bilateral LTx (N=8669)	p value
RECIPIENT
AGE AT TRANSPLANT			<0.0001
Median (5th-95th percentile), years	59.0 (43.0 – 69.0)	50.0 (22.0 – 66.0)
GENDER			0.8026
Female	3001 (45.5%)	3960 (45.7%)
ABO BLOOD GROUP			0.1784
A	2707 (41.0%)	3574 (41.2%)
AB	252 (3.81%)	356 (4.11%)
B	689 (10.4%)	980 (11.3%)
O	2950 (44.7%)	3759 (43.4%)
Missing	1 (.%)	0 (.%)
PRIMARY DIAGNOSTIC INDICATION FOR LTX			<0.0001
COPD, non-A1ATD	3245 (49.2%)	2225 (25.7%)
IIP	2238 (33.9%)	1739 (20.1%)
Cystic fibrosis	7 (0.11%)	2183 (25.2%)
COPD, A1ATD	356 (5.39%)	470 (5.42%)
ILD, non-IIP	208 (3.15%)	255 (2.94%)
Primary pulmonary hypertension	45 (0.68%)	351 (4.05%)
Sarcoidosis	126 (1.90%)	322 (3.71%)
All other diagnoses	374 (5.67%)	1123 (13.0%)
Missing	0 (.%)	1 (.%)
ERA OF TRANSPLANT			<0.0001
1/1/1994 through 12/31/2002	2938 (44.5%)	2236 (25.8%)
1/1/2003- 12/31/2011	3661 (55.5%)	6433 (74.2%)
ON LIFE IMMEDIATELY PRIOR TO LTx			<0.0001
No	6393 (97.6%)	7869 (91.6%)
Yes	158 (2.39%)	724 (8.35%)
Missing	48 (.%)	76 (.%)
MOST RECENT PRA PERCENT - CLASS 1			0.2039
Median (5th-95th), %	0.0 (0.0 – 18.0)	0.0 (0.0 – 25.0)
MOST RECENT PRA PERCENT - CLASS 2			0.8872
Median (5th-95th), %	0.0 (0.0 – 23.0)	0.0 (0.0 – 18.0)
DONOR
AGE			0.6085
Median (5th-95th), years	30.0 (15.0 – 57.0)	30.0 (15.0 – 57.0)
GENDER			<0.0001
Female	2411 (36.5%)	3596 (41.5%)
Male	4188 (63.5%)	5073 (58.5%)
ABO BLOOD GROUP			0.0067
A	2404 (36.4%)	3227 (37.2%)
AB	129 (1.96%)	203 (2.34%)
B	649 (9.83%)	954 (11.0%)
O	3417 (51.8%)	4285 (49.4%)
HISTORY OF HIGH RISK BEHAVIOR			0.0325
No	2917 (44.2%)	5364 (61.9%)
Yes	199 (3.01%)	442 (5.09%)
Missing	3483 (52.8%)	2863 (33.0%)
ISCHEMIC TIME FOR LAST LUNG IMPLANTED			<0.0001
Median (5th-95th), hours	3.9 (2.0 – 6.3)	5.4 (3.1 – 8.4)
DONOR / RECIPIENT CMV STATUS			<0.0001
Negative / Negative	883 (13.4%)	1544 (17.8%)
Negative / Positive	1557 (23.6%)	1609 (18.6%)
Positive / Negative (“mismatch”)	1177 (17.8%)	2001 (23.1%)
Positive / Positive	2391 (36.2%)	2727 (31.5%)
Missing	591 (8.96%)	788 (9.09%)

After excluding missing cases, P-values calculated using Wilcoxon rank sum test for continuous variables and Chi-square test for categorical variables. IIP limited to those with a diagnosis of “UIP/IPF” in the ISHLT Registry. Abbreviations: A1ATD, alpha-1-antitrypsin deficiency; CMV, cytomegalovirus; COPD, chronic obstructive pulmonary disease; IIP, idiopathic interstitial pneumonia; ILD, interstitial lung disease; IPF, idiopathic pulmonary fibrosis; PRA, panel reactive antibody; UIP, usual interstitial pneumonitis.

Follow-up after LTx

The study cohort had 79,896 person-years of follow-up (single LTx 33,934 person-years; bilateral LTx, 45,962 person-years). The cohort had only 279/15,268 (1.83%) cases lost to follow-up prior to the study end date, and the lost events typically occurred after years of reported follow-up (median: 4.76; interquartile range, 4.02 years).

Post-lung transplantation outcomes

BOS-free survival

A majority of LTx recipients developed the composite outcome (death or BOS, whichever occurred first) within 4 years after transplantation (Figure 2A-C) and almost 90% developed the composite outcome within 10 years of transplantation (Figure 2A, Table 3A). The median time to BOS-free survival for bilateral LTx was 4.00 years when using Scenario 1, the approach typically used by the Registry, compared to 3.58 years when using Scenario 2, which accounted for partially missing BOS data. In comparison, the median time to BOS-free survival for single LTx was 3.19 years (Scenario 1) compared to 3.16 years (Scenario 2) (p<0.001 for bilateral vs single LTx in each Scenario; Figure 2A, Table 3A).

Figure 2: BOS-free survival after lung transplantation, overall and stratified by transplant procedure type.

Kaplan-Meier survival methods estimated the proportion in whom bronchiolitis obliterans syndrome (BOS) and death had not occurred at follow-up after lung transplantation (LTx), stratified by the two BOS status determination Scenarios [Scenario 1 (orange) versus Scenario 2 (green); see Table 1 for Scenario definitions] that we used to determine the presence of absence of BOS. Numbers below X-axis time points represent those at risk.

Cohorts consisted of primary LTx reported from 1/1/1994 to 12/31/2011, with outcomes assessed through 12/31/2012. Analyses censored for retransplant, end of study follow-up, and loss to follow-up (i.e., used the mid-point of the reporting period last reported as alive).

Figure 2A shows unadjusted BOS-free survival in the entire eligible study cohort (Single LTx and Bilateral LTx combined), while Figures 2B and 2C show estimates stratified by transplant procedure type (Single LTx, Figure 2B; Bilateral LTx, Figure 2C). Single LTx and bilateral LTx recipients had a median time to BOS or death of 3.2 for Scenario 1 to 3.3 years for Scenario 2, and 3.6 years for Scenario 1 to 4.0 years for Scenario 2 respectively. For each Scenario, bilateral LTx had greater BOS-free survival than single LTx (log rank p < 0.001).

Table 3A:

Kaplan-Meier BOS-free survival after primary lung transplantation

BOS-free survival
	1 year*	3 year*	5 year*	10 year*	Mediana
Bilateral Lung Transplant (n=8,699)% (95% CI)
Scenario 1b	90.2	61.3	41.3	14.3	4.00 years
	(89.6–90.9)	(60.3–62.4)	(40.2–42.5)	(13.2–15.5)	(3.98–4.03)
Scenario 2b	85.3	58.3	40.0	15.3	3.58 years
	(84.6–86.1)	(57.3–59.4)	(38.9–41.1)	(14.2–16.4)	(3.53–3.72)
Single Lung Transplant (n=6,599)% (95% CI)
Scenario 1b	88.0	54.8	32.8	7.1	3.19 years
	(87.2–88.8)	(53.5–56.0)	(31.6–34.1)	(6.19–7.94)	(3.09–3.40)
Scenario 2b	83.1	51.2	31.7	7.8	3.16 years
	(82.1–84.0)	(49.9–52.4)	(30.5–32.9)	(7.01–8.67)	(2.99–3.30)
Single or Bilateral Lung Transplant (n=15,268)% (95% CI)
Scenario 1b	(88.8–89.8)	(57.7–59.3)	(36.7–38.4)	(10.0–11.5)	(3.75–3.91)
	89.3	58.5	37.5	10.8	3.84 years
Scenario 2b	84.4	55.2	36.3	11.6	3.48 years
	(83.8–84.9)	(54.4–56.0)	(35.5–37.1)	(10.9–12.3)	(3.45–3.50)

BOS, bronchiolitis obliterans syndrome; CI, confidence interval.

Composite outcome event: BOS or death, whichever comes first.

Censoring: retransplant, end of study or loss to follow-up.

^*Values denote cumulative percent (95% CI) without the composite outcome.

^aMedian time to composite outcome (95%CI).

^bBOS status determination Scenario definitions shown in Table 1.

Single LTx recipients had significantly worse unadjusted BOS-free survival than bilateral LTx recipients, regardless of the Scenario used to determine BOS status (for both Scenarios, HR 1.29, 95% CI 1.24–1.34; log-rank test p<0.001; Table 3A, Figure 2B and 2C, Supplementary Figure 1), and the divergence in BOS-free survival between the single and bilateral LTx groups widened during follow-up.

To compare BOS-free survival between lung transplant procedure types, we adjusted for baseline (time of transplant) variables that have a known impact on BOS and survival, focusing on Tier 1 elements that had complete data in the Registry. After adjusting for recipient age at transplant, gender, era and primary diagnostic indication for transplantation, only transplant procedure type had a significant association with BOS-free survival. Both Scenarios showed similar results favoring bilateral over single LTx (for single LTx, Scenario 1 HR_adj 1.32, 95% CI 1.26–1.38, p <0.001; Scenario 2 HR_adj 1.31, 95% CI 1.25–1.37, p <0.001).

Death and BOS

Regarding composite endpoint components, 46.8% (8,128/15,268) of the recipients died during the observational period, with nearly 40% of the deaths occurring within 5 years after LTx (Figure 3, Table 3B). Median time to death was 5.75 years for single LTx recipients compared to 8.48 years for bilateral LTx recipients (p<0.0001, Figure 3, Table 3B). Among those who lived for 10 years after LTx, almost all developed BOS (Figure 4A-C). The median time to BOS after LTx ranged from 5.00 for Scenario 1 to 4.67 years for Scenario 2 (p<0.0001 for both single and bilateral LTx; Figure 4A, Table 3C).

Figure 3: Survival after lung transplantation, stratified by transplant procedure type.

Kaplan-Meier survival for single and bilateral lung transplant (LTx). Cohorts consisted of primary LTx reported from 1/1/1994 to 12/31/2011, with outcomes assessed through 12/31/2012. Analyses censored for retransplant, end of study follow-up, and loss to follow-up (i.e., used the mid-point of the reporting period last reported as alive), but not censored for BOS.

Bilateral LTx (blue) had better unadjusted survival than single LTx (red), and the discrepancy became larger during follow-up (log-rank test p < 0.001; unadjusted hazard ratio for death in single LTx versus bilateral LTx 1.56 [95% CI 1.49–1.63]). Thin lines surrounding each curve represent the upper and lower 95% CI limits.

Table 3B:

Kaplan-Meier survival after primary lung transplantation

Survival
	1 year*	3 year*	5 year*	10 year*	Mediana
Bilateral Lung Transplant (n=8,699)% (95% CI)
All Scenariosb	93.5	78.8	66.2	43.3	8.48 years
	(93.0–94.1)	(77.9–79.7)	(65.1–67.3)	(41.9–44.8)	(8.09–8.73)
Single Lung Transplant (n=6,599)% (95% CI)
All Scenariosb	91.3	72.8	55.8	24.7	5.75 years
	(90.6–92.0)	(71.8–74.0)	(54.5–57.0)	(23.4–26.1)	(5.57–5.92)
Single or Bilateral Lung Transplant (n=15,268)% (95% CI)
All Scenariosb	92.5	76.2	61.5	34.1	6.83 years
	(92.1–93.0)	(75.5–76.9)	(60.7–62.3)	(33.0–35.1)	(6.69–7.02)

CI, confidence interval.

Outcome event: death.

Censoring: retransplant, end of study, or loss to follow-up.

^*Values denote cumulative percent (95% CI) alive.

^aMedian time to death (95%CI).

All BOS Scenarios use the same survival data.

Figure 4: Freedom from BOS after lung transplantation, overall and stratified by transplant procedure type.

Kaplan-Meier survival methods estimated the proportion in whom bronchiolitis obliterans syndrome (BOS) had not occurred at follow-up after lung transplantation (LTx), stratified by the two BOS status determination Scenarios [Scenario 1 (orange) versus Scenario 2 (green); see Table 1 for Scenario definitions] used for determining the BOS status (present or absent). Numbers below X-axis time points represent those at risk.

Cohorts consisted of primary LTx reported from 1/1/1994 to 12/31/2011, with outcomes assessed through 12/31/2012. Analyses censored for death, retransplant, end of study follow-up, and loss to follow-up (i.e., used the mid-point of the reporting period last reported as alive).

Figure 4A shows estimates of overall freedom from BOS in the entire eligible study cohort (Single LTx and Bilateral LTx combined), while Figures 4B and 4C show estimates stratified by transplant procedure type (Single LTx; Figure 4B; Bilateral LTx, Figure 4C).

For each Scenario, bilateral LTx had greater freedom from BOS than single LTx (log rank p < 0.001). For each Scenario, single LTx recipients had a shorter median time to BOS than bilateral LTx recipients, with the median time to BOS ranging from 0.14 years (Scenario 1) to 0.78 years (Scenario 2).

Table 3C:

Kaplan-Meier Freedom from BOS after primary lung transplantation

Freedom from BOS
	1 year*	3 year*	5 year*	10 year*	Mediana
Bilateral Lung Transplant (n=8,699)% (95% CI)
Scenario 1b	95.6 (95.1–96.1)	71.3 (70.3–72.4)	52.1 (50.9–53.4)	22.3 (20.7–23.9)	5.06 years (5.02–5.23)
	(95.1–96.1)	(70.3–72.4)	(50.9–53.4)	(20.7–23.9)	(5.02–5.23)
Scenario 2b	90.7 (90.1–91.4)	68.4 (67.4–69.5)	51.1 (49.8–52.3)	24.1 (22.6–25.7)	5.29 years (4.92–5.42)
	(90.1–91.4)	(67.4–69.5)	(49.8–52.3)	(22.6–25.7)	(4.92–5.42)
Single Lung Transplant (n=6,599)% (95% CI)
Scenario 1b	95.3	68.4	47.9	17.5	4.92 years
	(94.7–95.8)	(67.1–69.7)	(46.4–49.4)	(15.8–19.3)	(4.80–4.99)
Scenario 2b	90.4	64.8	47.1	19.7	4.51 years
	(89.6–91.1)	(63.6–66.1)	(45.7–48.6)	(18.0–21.3)	(4.46–4.56)
Single or Bilateral Lung Transplant (n=15,268)% (95% CI)
Scenario 1b	95.5	70.1	50.3	20.2	5.00 years
	(95.1–95.8)	(69.3–70.9)	(49.3–51.3)	(19.0–21.4)	(4.98–5.04)
Scenario 2b	90.6	66.9	49.4	22.2	4.67 years
	(90.1–91.1)	(66.1–67.7)	(48.5–50.3)	(21.1–23.3)	(4.56–5.10)

BOS, bronchiolitis obliterans syndrome; CI, confidence interval.

Outcome event: BOS.

Censoring: death, retransplant, end of study or loss to follow-up.

^*Values denote cumulative percent (95% CI) without BOS.

^aMedian time to BOS (95%CI).

^bBOS status determination Scenario definitions shown in Table 1.

Comparison of methodologies for using data from LTx recipients who have missing BOS status

BOS-free survival

Different Scenarios for using data from LTx recipients who had missing BOS status during any eligible follow-ups produced significantly different outcomes for each transplant procedure type (Supplementary Table 2A-C). For example, Scenario 1, the approach typically used by the Registry, yielded significantly higher BOS-free survival by about 5% at 1 year and by about 3% at 3 years for each of the single and bilateral LTx groups, in comparison to Scenario 2, which accounted for partially missing BOS data (Supplementary Figure 1, Supplementary Table 2A). There was no overlap in the 95% confidence intervals for 1- and 3-year BOS-free survival between Scenario 1 and 2 (Supplementary Table 2A). Scenario 1 also had a significantly higher median BOS-free survival time when compared to Scenario 2 (e.g., for bilateral LTx, 0.42 years longer; Supplementary Table 2A). These Scenario differences in BOS-free survival occurred mainly in the first 4 years post-transplantation (Figure 2B-C, Supplementary Table 2A) and became less apparent at later time points.

Survival

By definition, the Scenarios used to determine BOS had no effect on survival. Single LTx recipients had significantly worse survival (2.7 year lower median survival; HR_adj for death 1.45, 95% CI 1.37–1.53; p<0.001; Figure 3, Table 3B; Supplementary Table 2B) compared to the bilateral LTx recipients.

Freedom from BOS

Regardless of the Scenario used, single LTx recipients had significantly worse unadjusted freedom from BOS compared to bilateral LTx recipients (Scenario 1, HR for BOS 1.14, 95% CI 1.09–1.19; p<0.001; Scenario 2, HR for BOS 1.13, 95% CI 1.08–1.18; Figure 4B and C, Table 3C) compared to the bilateral LTx recipients and had a shorter median time to BOS (0.14 years; Supplementary Figure 2, Supplementary Table 2C). This discrepancy became even more apparent with Scenario 2 analyses (0.78 years) (Table 3C, Figure 4B and C). Both Scenarios showed similar results favoring bilateral over single LTx when adjusted for age at transplant, gender, diagnosis and era (for single LTx, Scenario 1 HR_adj 1.29, 95% CI 1.16–1.30, p <0.001; Scenario 2 HR_adj 1.21, 95% CI 1.15–1.28, p <0.001).

Post-BOS survival

After the diagnosis of BOS, LTx recipients had a low Kaplan-Meier survival. Bilateral LTx recipients had a longer median survival time than single LTx recipients, with Scenario 1 having at least half-a-year lower survival than Scenario 2 for both transplant procedure types (Bilateral LTx, Scenario 1, 3.8 years and Scenario 2, 4.3 years; Single LTx Scenario 1, 2.6 years, and Scenario 2, 3.1 years; Supplementary Table 3). Given the incompleteness of BOS stage reporting, we did not analyze outcomes based on BOS stage at first BOS diagnosis, trajectory of BOS stages, or effect of treatments on the development of BOS.

DISCUSSION

To address long-term outcomes in lung transplant recipients, we utilized the data from the ISHLT Thoracic Transplant Registry, which uses a type of real-world evidence data, which when compared to clinical trial data, can more closely describe how a technology will perform in a broader, more representative population over a longer timeframe.⁴³ This analysis of this Registry data demonstrated the disappointingly high incidence of both death and BOS, and consequently low BOS-free survival for recipients of primary LTx. The majority of single and bilateral LTx recipients had BOS or death reported within 4 years after LTx, and almost 90% of recipients experienced at least one of these outcomes within 10 years after LTx.

We believe that standard Registry LTx recipient analyses (Scenario 1) overestimate BOS-free survival produced by not utilizing data after a recipient’s first missing report of BOS status.^1,38–40 In comparison to our analytic approach (Scenario 2) that incorporated some common sense assumptions, we detected that Scenario 1 significantly “overestimated” BOS-free survival by about 5% at 1 year and about 3% at 3 years for both of the single and bilateral LTx groups. Also, compared to our approach, Scenario 1 significantly “overestimated” the median BOS-free survival time by 0.42 years for bilateral LTx and by 0.03 years for single LTx and “underestimated” the median survival after BOS by about a half-year for both procedure types.

The majority of previous multicenter cohort studies and Registry analyses have focused on freedom from BOS and/or overall survival rather than BOS-free survival.^1,14,44,45 This study demonstrated the important discrepancy between freedom from BOS and BOS-free survival. Since freedom from BOS Kaplan-Meier analyses censor for death and BOS-free survival analyses include death as part of the composite endpoint,^21,46 the freedom from BOS analyses have much better outcomes than the BOS-free survival analyses. For example, even at 3 years after LTx, freedom from BOS was 70%, while BOS-free survival was only 55%, and this disparity increased during follow-up. This demonstrates the importance of incorporating survival when presenting outcome data.

Limited data exists regarding BOS-free survival in single and in bilateral LTx recipients. ^20,35,36 In addition, the few studies that have assessed BOS-free survival have not generated consistent estimates, with the estimates varying from 34–75% at 5 years post-LTx.^26,28,29 We found that only 31.7% of single LTx and 40.0% of bilateral LTx recipients (36.3% for all LTx) remained alive and free from BOS at 5 years post-LTx. While a component of BOS may be reversible upon use of azithromycin, the first randomized placebo-controlled trial of azithromycin to prevent BOS was published in 2011.³⁶ As our a priori study cohort was defined as those receiving a primary LTx in the era of 1/1/1994 through 12/31/2011, and recipient outcome assessment through 12/31/2012, it is extremely unlikely that a significant proportion of patients in our study were treated with chronic azithromycin to prevent BOS, since that practice only began gaining acceptance around the end of this study.^47,48 Even if centers did start giving azithromycin when lung function began to drop, it likely would have affected only a small proportion of the cohort since most patients had died or developed BOS by 2011. However, going forward, the use of azithromycin needs to be considered when interpreting long-term outcomes associated with BOS and/or CLAD.⁶

Previous Registry data assessments^32,33 and other studies^{18–20,35,36} have found a strong association between transplant procedure type and survival after LTx. This study provides new and robust data about BOS-free survival, adjusts for important baseline covariates and reinforces the discrepancy in outcomes between bilateral over single lung transplantation. Moreover, as the Registry data uses a type of real-world evidence data, local and patient-specific clinical practice (and not a rigid study protocol) determine diagnostic and treatment decision-making in this real-world setting.⁴⁹ Thus, secondary real-world evidence data obtained from Registries support retrospective analyses which may serve as inputs for prospective study designs.⁵⁰

This study of LTx that occurred over 18 years had 79,896 person-years of follow-up, and more than 98% of recipients had follow-up that allowed for assessment of the composite endpoint. Despite the large sample size of more than 15,000 recipients that included a majority of the world’s lung transplants during the study era, and the apparent precision of our estimates, we recognize limitations of Registry data analyses. Accuracy and precision tradeoffs include: 1) the lack of an exact date of BOS and assumptions about the timing of BOS, 2) lack of validation of the BOS diagnosis, 3) a significant amount of missing BOS stage information at the time of first BOS diagnosis, and 4) BOS trajectory information.²⁰ Regarding the limitations of the multivariable analyses, we restricted it to variables from the Tier 1 mandatory reporting elements that we thought might influence BOS-free survival; we did not include factors^51,52 that had a large amount of missing data. We did not have the ability to significantly assess for predictor variable data accuracy, though we attempted to assess some broad categories (such as primary diagnostic indication) to decrease the risk of misclassification.³⁷ We also did not have the option of including any variables that the Registry does not collect.

Regarding the primary composite outcome, we used the narrower outcome of BOS instead of CLAD because the Registry does not specifically collect CLAD data. However, some “yes BOS” status reporting may actually refer to non-BOS CLAD events.^8,39 Since CLAD includes BOS and restrictive allograft syndrome (RAS), the latter often having worse outcomes and treatment response compared to BOS,^10,11,4 the BOS-free survival estimates likely underestimated CLAD-free survival. However, given the very low BOS-free survival shown in this study, any worse estimates of CLAD-free survival do not change the bleak picture that begs for improvement. Additionally, this study did not assess for causes of death, as we did not aim to do so. The Registry’s annual reports already provide cause of death information.¹ Moreover, in our study, we would argue that the cause of death did not matter as we mainly wanted capture that a death occurred as an absolute competing event to BOS. As an aside, cause of death determination, even when assessed by blinded central adjudication, has significant validity concerns.⁵³

This Registry study utilized data that came almost exclusively from the U.S. and Canada. The lack of non-North American data in the Registry has resulted mainly from a combination of lack of BOS data collection by collectives and a lack of center reporting of BOS data to the Registry. Though the lack of non-North American BOS data in the Registry decreases the generalizability of this study, it does not affect its internal validity. Supporting the external validity of this study, the cohort likely included more than half of all LTx performed internationally. For example, in 2016, 55% (2,818/5,147) of the world’s LTx occurred in the U.S. and 6% (302/5,147) occurred in Canada.⁵⁴ Thus, this study generalizes to about two-thirds of the world’s LTx recipients.

To summarize, despite its limitations, this large study: 1) provides robust estimates of BOS-free survival, 2) describes the precision of these estimates, 3) shows the important differences between BOS-free survival and freedom from BOS, 4) points out the important effects of handling missing data, and 5) describes a clinically relevant association between transplant procedure type and BOS-free survival. The findings from this study support the vital need for new therapies to improve post-transplant survival and to reduce the incidence of BOS and other types of CLAD. Accordingly, we strongly advocate for expanded CLAD data collection by the Registry, the data collectives, and individual transplant centers. We propose that clinical trials of therapeutics aimed at mitigating CLAD or any of its subtypes should use event-free survival as a primary endpoint.