Factors indicative of long-term survival after lung transplantation: A review of 836 10-year survivors

Data source 

We were provided Standard Transplant Analysis and Research (STAR) files from the UNOS registry. The data set comprises an open cohort of all patients undergoing LTx in the United States. No patient or center identifiers were included in this analysis and the study was granted Investigational Review Board exemption at our institution.

Study design 

The design was a case-control study embedded within the prospectively collected UNOS open cohort (case-cohort design). Cases were adult patients (aged >17 years) receiving initial LTx between 1987 and 1997 who survived at least 10 years after LTx (Figure 1). Controls were patients who received an allograft within the same era (1987–1997) and died between 1 and 5 years after LTx. Patients were excluded who survived less than 1 year and those who survived between 6 and 9 years. Also excluded were patients who survived 1 to 5 years but were lost to follow-up to limit potential selection bias in the analysis relating to factors predicting loss to follow-up rather than factors predicting death.

View Large Image Download to PowerPoint

Figure 1. Flowchart shows the study design.

Analysis 

The outcome of interest was 10-year survival vs 1- to 5-year survival. A multivariable logistic regression model predicting the odds of 10-year survival was constructed incorporating covariates of interest. After all covariates were tested in a univariate analysis, the multivariable logistic regression model was developed by incorporating well-represented covariates (<10% missing data) with the potential for confounding based on biologic plausibility. Lowess smoothing plots evaluated linear breakpoints for continuous variables, and spline terms were added to the model when appropriate. The likelihood ratio test for significance was used in a nested model approach to assess appropriateness of covariate inclusion.

Two models were constructed from the data set. The initial model, Model A, incorporated pre-operative, operative, and donor-specific factors. The second model, Model B, incorporated those factors in Model A plus additional post-operative factors All models were examined using the area under the receiver operating curve (ROC) along with the Akaike information criterion and Hosmer-Lemeshow (HL) goodness of fit test. Model accuracy was assessed using “leave one out” cross validation with 1,845 separate replications.9 A cross-validated HL goodness of fit chi-square statistic was calculated for this purpose. Collinearity was assessed by incorporating the final covariates into a multivariable linear regression model predicting a randomly generated number allowing calculation of variance inflation factors. To assess importance of missing data, multiple imputation was performed on the final model, and significant covariates were compared with the model constructed using case-wise deletion.

The final model (Model B) included the following covariates: age, age spline term with a cut-point at 35 years, recipient sex, recipient primary diagnosis, mean annual institutional volume, bilateral LTx (BLT) vs single LTx (SLT), HLA level, panel reactive antigen (PRA) level, development of BOS during follow-up, and number of hospitalizations for rejection and infection.

Univariate analysis for continuous data was conducted using the t-test or Wilcoxon rank sum test for non-parametric data. The chi-square test was used for comparison of categoric variables. For all analyses, a 2-tailed value of p < 0.05 was considered significant. Means are presented with standard deviations, medians with interquartile ranges (IQR), and all odds ratios (OR) are presented with 95% confidence intervals (CI). Statistical analyses were performed with the aid of STATA 9.2SE software (StataCorp LP, College Station, TX).

Study groups 

From 1987 until 2002, 5,273 patients received LTx. Exclusion of 323 children, 133 re-LTx patients, and those who did not meet study eligibility (ie, 1,235 patients surviving <1 year after LTx, 846 surviving 6 to 9 years, and 554 lost to follow-up) yielded a study population of 2,182 (Figure 1). Of these, 836 patients (17.3%) were 10-year survivors. The controls were 1,346 patients who survived 1 to 5 years. Although the number of 10-year survivors increased dramatically between 1987 and 1990, the number remained constant after 1990, comprising approximately 15% to 20% of all LTx recipients (Figure 2). The number of 5-year survivors increased steadily during the decade.

View Large Image Download to PowerPoint

Figure 2. Percentage of lung transplant (LTx) recipients surviving 10 years and 5 years from 1987 to 1997.

Baseline characteristics 

Cases and controls differed in several important baseline characteristics. Specifically, 10-year survivors were younger, more likely to have cystic fibrosis (CF) as their primary diagnosis, and were more likely to be white than controls (Table 1). Conversely, fewer patients in the 10-year survival group had idiopathic pulmonary fibrosis (IPF) or chronic obstructive pulmonary disease (COPD) as the primary diagnosis. Variables relating to acuity such as creatinine, oxygen requirement at baseline, and percentage of patients with pre-operative diabetes mellitus did not differ between cases and controls (Table 1).

Table 1. Baseline Characteristics and Demographics Among 10-year Survivors and 1- to 5-year Survivors (1987–1997)

	Variables	10-year survivors N = 836	1–5 year survivors N = 1,346	p-valuea
	Recipient demographics
	Age, mean ± SD, years	44.0±11.5	47.8±12.5	<0.001
	Females, No. (%)	445/836(53.2)	680/1346(50.3)	0.28
	White race, No. (%)	777/836(92.9)	1214/1343(90.4)	0.04
	Mean institutional volume	15.1±11.7	13.3±10.7	<0.001
	Primary diagnosis, No. (%)
	Cystic fibrosis	170/836(20.3)	170/1346(12.6)	<0.001
	COPD	254/836(30.4)	623/1346(46.3)	<0.001
	Idiopathic pulmonary fibrosis	92/836(11.0)	205/1346(15.2)	0.005
	Primary pulmonary hypertension	66/836(7.9)	77/1346(5.7)	<0.001
	Other	254/836(30.4)	271/1346(20.1)	0.05
	Recipient O2 (L/min) support, comorbidities
	O2 (L/min) requirement, mean ± SD, liters/min	2.35±1.64	2.50±1.71	0.09
	Body mass index, mean ± SD, kg/m2	22.5±5.1	23.2±5.3	<0.001
	Diabetes mellitus, No. (%)	20/404(5.0)	39/664(5.9)	0.52
	Creatinine, (mg/dL) mean ± SD, mg/dl	0.89±0.50	0.96±0.94	0.70
	Transplant variables, No. (%)
	BLT (overall)	437/836(52.3)	428/1344(31.8)	<0.001
	BLT by age in years, No. (%)
	18–35	166/196(84.7)	190/232(81.9)	0.44
	35–50	177/310(57.1)	128/360(35.6)	<0.001
	>50	71/290(24.5)	87/589(12.6)	<0.001
	BLT by major diagnosis types, No. (%)
	Cystic fibrosis	168(99.4)	168(99.0)	0.60
	COPD	67(26.4)	97(15.6)	<0.001
	Idiopathic pulmonary fibrosis	27(29.4)	30(14.6)	0.003
	Primary pulmonary hypertension	41/66(62.1)	43/76(56.6)	0.50
	Other	133/254(52.4)	90/270(33.3)	<0.001
	HLA match, No. (%)
	0 antigen match	168/688(24.4)	341/1123(30.4)	0.005
	1 antigen match	258/688(27.5)	451/1123(40.2)	0.20
	2 antigen match	182/688(26.5)	235/1123(20.9)	0.01
	≥3 antigen match	80/688(11.6)	96/1123(8.6)	0.04
	Panel reactive antibody levels
	0%	637/787(80.9)	1003/1279(78.4)	0.38
	1% to 10%	124/787(15.8)	208/1279(16.3)	0.69
	11% to 25%	13/787(1.7)	36/1279(2.8)	0.09
	>25%	13/787(1.7)	32/1279(2.5)	0.19

BLT, bilateral lung transplantation; COPD, chronic obstructive pulmonary disease; HLA, human leukocyte antigen; SD, standard deviation.

a Based on results of either t-test or Wilcoxon rank sum test as appropriate for continuous variables or chi-square test for categoric variables. Values of p < 0.05 are significant.

Overall, 52.3% of the 10-year survivors received bilateral lung transplantation (BLT) compared with 31.8% in controls (p < 0.001; Table 1). When this was stratified by age, the association between BLT and 10-year survival only occurred for those patients aged older than 35. In addition, when stratified by primary diagnosis, only patients with COPD (26.4% vs 15.6%, p < 0.001), IPF (29.4% vs. 14.6%, p = 0.003), and other pathologies (52.4% vs. 33.3%, p < 0.001) had an increased use of BLT. Thus, BLT did not appear to be associated with 10-year survival in younger patients and those with CF.

HLA match levels appeared to be associated with 10-year survival, because cases were more likely to have at least 2 HLAs matched between donor and recipient than controls (Table 1). Although no significant association between PRA and 10-year survival was noted, only 1.7% of 10-year survivors had PRA levels greater than 10%.

Follow-up 

An examination of follow-up variables revealed further differences between cases and controls (Table 2). The 10-year survivors had fewer hospitalizations for rejection but more hospitalizations for infection compared with controls. Overall oxygen requirements during the course of follow-up were 0.8 liters/min lower in 10-year survivors than in the 1- to 5-year survivors (p < 0.001). BOS developed in 10-year survivors with increased frequency compared with 1- to 5-year survivors, likely relating to their longer survival (60.1% vs 45.3%, p < 0.001). Despite this increased prevalence, however, 10-year survivors did not die of BOS as frequently as controls (14.6% vs 29.8%, p < 0.001). The leading cause of death in the 10-year survivor group was non-specific respiratory failure, and a greater percentage of 10-year survivors died of malignancy than 1- to 5-year survivors.

Table 2. Univariate Comparison of Follow-up Variables Between 10-year Survivors and 1- and 5-year Survivors

	Variable	Survival		p-valuea
		10-years	1–5 years (1987–1997)
		N = 836	N = 1,346
	BOS development	503(60.1)	702(45.3)	<0.001
	Hospitalizations, mean ± SD, No.
	For infection	1.65±1.73	1.01±1.07	<0.001
	For rejection	0.53±0.88	0.66±0.85	<0.001
	O2 (L/min) requirement at F/U, mean ± SD	0.19±0.54	0.99±1.81	<0.001
	Patients who died, No. (%)	268(32.1)	1346(100)	NAb
	Causes of death, No. (%)
	BOS	39(14.6)	401(29.8)	<0.001
	Acute rejection	0(0)	18(1.3)	0.02
	Infection	39(14.6)	291(21.6)	0.003
	Respiratory failure non-specific	43(26.0)	122(9.1)	0.008
	Malignancy	32(11.9)	91(6.8)	0.003
	Renal failure	8(3.0)	24(1.8)	0.2
	Multisystem organ failure	14(5.2)	55(4.1)	0.5
	Cardiovascular disease	12(4.5)	45(3.3)	0.5
	Other known	57(21.3)	231(17.1)	0.21
	Unknown	24(9.0)	68(5.1)	0.001

BOS, bronchiolitis obliterans syndrome; F/U, follow-up; SD, standard deviation.

a Value of p based on results of t-test for continuous variables or chi-square test for categoric variables. Significance set at p < 0.05. b This value of p not applicable due to method of control selection.

Multivariable logistic regression 

Multivariable logistic regression models were constructed comparing 10-year survivors with controls (Table 3). The initial model was calibrated with a set of covariates followed by extension to post-operative factors (Models A and B). On univariate analysis, as well as in both the base model (Model A) and extended model (Model B), age 18 to 35 years was associated with an increase in the odds of 10-year survival. In addition, the spline term at age 35 was significant; indicating that 35 years is indeed a significant breakpoint in prediction of 10-year survival. In all models, the odds of 10-year survival were decreased in patients aged older than 35. Furthermore, all models revealed BLT to be strongly associated with an increase in the odds of 10-year survival. Mean center volume (OR, 1.02; 95% CI, 1.01–1.03; p < 0.001 for each one case/year decrease) and higher levels of HLA matching (OR, 1.61; 95% CI, 1.07–2.42; p = 0.02) were both strongly associated with an increase in the odds of 10-year survival. On a graphic examination of mean hospital volume, no level was found to serve as a threshold for optimizing the odds of 10-year survival.

Table 3. Univariate and Multivariable Logistic Regression Predicting Odds of 10-year Survival vs 1- to 5-year Survival

	Covariates of interest	Univariate analysis OR (95% CI)	p-value	Multivariable analysisa OR (95% CI)	p-value	Multivariable analysisb OR (95% CI)	p-value
	Pre-op/operative variables
	Age 18–35 years	1.07(1.03–1.09)	<0.001	1.08(1.04–1.12)	<0.001	1.07(1.03–1.11)	0.01
	Age >35 years	0.95(0.94–0.96)	<0.001	0.98(0.96–0.99)	0.01	0.97(0.95–1.00)	0.03
	Age spline coef. at 35	0.89(0.86–0.92)	<0.001	0.91(0.87–0.95)	<0.001	0.92(0.88–0.96)	0.01
	Female sex			0.93(0.74–1.15)	0.50	0.92(0.74–1.16)	0.26
	Primary diagnosis
	COPD	Reference		Reference		Reference
	Cystic fibrosis	2.45(1.89–3.17)	<0.001	1.47(0.92–2.34)	0.11	1.54(0.95–2.51)	0.08
	Idiopathic pulmonary fibrosis	1.10(0.83–1.46)	0.51	1.02(0.70–1.48)	0.93	1.07(0.72–1.58)	0.74
	PPH	2.10(1.47–3.01)	<0.001	1.43(0.89–2.31)	0.14	1.74(1.06–2.85)	0.03
	Other	2.30(1.84–2.88)	<0.001	1.62(1.20–2.20)	0.002	1.74(1.26–2.39)	0.001
	Mean annual center volume	1.02(1.01–1.03)	<0.001	1.01(1.00–1.02)	0.009	1.02(1.01–1.03)	<0.001
	BLT	2.34(1.96–2.80)	<0.001	1.84(1.36–2.48)	<0.001	1.71(1.25–2.34)	0.001
	BMI	0.97(0.96–0.99)	0.002	0.98(0.96–1.00)	0.09	0.98(0.95–1.00)	0.07
	HLA match
	0	Reference		Reference		Reference
	1	1.16(0.91–1.48)	0.22	1.18(0.90–1.54)	0.23	1.12(0.85–1.49)	0.4
	2	1.57(1.20–2.05)	0.001	1.61(1.19–2.17)	0.002	1.58(1.16–2.17)	0.004
	≥3	1.69(1.19–2.40)	0.003	1.73(1.17–2.56)	0.006	1.61(1.07–2.42)	0.02
	Ischemic time
	<2.5 hours	0.62(0.44–0.87)	0.007	0.67(0.44–1.02)	0.06	0.59(0.38–0.94)	0.02
	≥2.5 hours	1.20(1.13–1.27)	<0.001	1.04(0.96–1.13	0.28	1.03(0.95–1.13)	0.4
	Spline at 2.5 hours of ischemic time	1.94(1.33–2.82)	<0.001	1.56(0.99–2.46)	0.06	1.73(1.08–2.78)	0.02
	PRA levels
	0%	Reference		Reference		Reference
	1%–10%	0.94(0.74–1.20)	0.61	1.00(0.74–1.34)	0.98	0.96(0.70–1.31)	0.79
	11%–25%	0.57(0.30–1.08)	0.09	0.56(0.27–1.15)	0.11	0.61(0.29–1.28)	0.19
	>25%	0.64(0.33–1.23)	0.18	0.46(0.21–1.00)	0.05	0.46(0.21–1.03)	0.06
	Follow-up variables
	BOS	1.78(1.48–2.10)	<0.001	NA		1.78(1.40–2.26)	<0.001
	Hospitalizations
	For infection	1.29(1.21–1.37)	<0.001	NA		1.40(1.27–1.54)	<0.001
	For rejection	0.80(0.72–0.89)	<0.001	NA		0.55(0.48–0.65)	<0.001
	Model fit
	AUROC	NA		0.67	0.73
	HL chi-square	NA		10.58	0.23	13.3	0.1
	McFaddens pseudo-R2	NA		0.05	0.1

AUROC, area under the receiver operating characteristic curve; BLT, bilateral lung transplantation; BMI, body mass index; BOS, bronchiolitis obliterans syndrome; CI, confidence interval; COPD, chronic obstructive pulmonary disease; HL, Hosmer-Lemeshow; HLA, human leukocyte antigen; NA, not applicable; OR, odds ratio; PPH, primary pulmonary hypertension; PRA, panel reactive antibody.

a Multivariable logistic regression model incorporating the following pre-operative and operative covariates: age 18–35, age spline at 35, sex, primary diagnosis, mean annual institutional volume, bilateral vs single LTx, BMI, HLA match level, ischemic time <2.5 hours, spline term at ischemic time of 2.5 hours, and PRA. b Multivariable logistic regression model incorporating all covariates in model a, plus the following follow-up covariates: number of hospitalizations for infections, number of hospitalizations for rejection, and development of any grade of BOS.

When post-operative covariates were included (Model B), each increase of 1 hospitalization for rejection decreased the odds of 10-year survival by 45% (OR, 0.65; 95% CI, 0.48–0.65; p < 0.001). The final model incorporated 15 covariates with a C-index of 0.73 and pseudo-R2 value of 0.1 (Table 3). Furthermore, the non-significant HL chi-square statistic of 13.3 (p = 0.1) indicated that the final model appropriately fit the data.

Interaction between infection and rejection 

In Model B, each hospitalization for infection increased the odds of 10-year survival by 40% (OR 1.40; 95% CI, 1.27–1.54; p < 0.001). Because of the counterintuitive nature of this finding, an interaction term between hospitalizations for infection and rejection was added to investigate the relationship between these 2 variables. Incorporation of this term revealed a positive and significant interaction between hospitalization for infection and rejection (OR 1.20; 95% CI, 1.04–1.37; p = 0.01). In this new model, rejection persisted as a significant covariate, with a 59% lower odds of 10-year survival with each additional hospitalization for rejection (OR, 0.41; 95% CI; 0.28–0.63; p < 0.001). However, hospitalizations for infection no longer predicted the outcome (OR, 1.11; 95% CI; 0.93–1.33; p = 0.3). This indicates that the observed association between infection and improved survival may be mediated by the associated decrease in rejection among those with increased immunosuppression.

Multivariable model 

Several points can be made from an examination of significant covariates present in the multivariable model. Specifically, age 18 to 35 years was associated with an increase in the odds of 10-year survival, but age older than 35 did not show an effect. The statistically significant spline term at 35 years indicated that age 35 was an important cut-point in explaining the outcome of interest.

Hospital volume emerged as a predictor of 10-year survival, with each 1 case/year increase translating into a 2% increase in the odds of 10-year survival. Several studies have evaluated the importance of institutional volume in transplantation outcomes, but most have focused on short-term mortality rates.10, 11 This study emphasizes the protective role that volume may have in long-term survival and stresses the importance of specialized providers and support systems present in large volume institutions.

Bilateral vs single LT 

An important finding from this study is that a higher percentage of 10-year survivors received BLT than controls. In the primary comparison between 10-year survivors and same era controls, BLT emerged as a significant predictor of 10-year survival in all logistic regression models. This relationship persisted after controlling for CF, a potential confounder given the nearly uniform BLT rate in this group. Among patients who survived 1 year or more, BLT was associated with doubling of the odds of 10-year survival. Many centers have adopted the use of BLT in LTx.1, 12, 13, 14 Potential therapeutic benefits of BLT include a reduction in alveolar damage during reperfusion, improved pulmonary compliance and mechanics, and the avoidance of native lung pathology.12, 15

Hospitalizations for rejection and infection 

It was not surprising that 10-year survivors had fewer hospitalizations for treatment of acute rejection during follow-up. Several studies examining LTx have found associations between acute rejection and BOS.2, 16 More interesting was the association between infection and long-term survival. Studies investigating the risk of infection with BOS and death have had mixed results. Some have linked the development of post-LTx non-cytomegalovirus infections to BOS and death,17, 18 whereas others have not.19, 20 Although infection may be an important mediator of early death, its effect on long-term survival is less clear. Patients surviving 10 years or longer did have an increased number of infectious hospitalizations with decreased hospitalizations for rejection. We found this association quite interesting, and further examination using the incorporation of an interaction term revealed the connection to be explained by a concomitant effect on hospitalizations for rejection. Hence, these data support the importance of avoiding acute rejection during follow-up, and may suggest a role for more aggressive immunosuppression regimens post-operatively.

Limitations 

Our study is limited by its use of the case-cohort design, which is retrospective in nature. We do not have control of all confounders. UNOS data are limited in the variables collected; and hence, there may be important variables not included in our analysis. Large data sets, like the one used in this study, rely on accurate coding, and we cannot confirm that errors in coding do not exist. However, we rely on the assumption that these errors are unlikely to bias results. Finally, we acknowledge that in this analysis we have eliminated information for patients that survived between 5 and 10 years after LTx. However, this was an a priori design strategy to examine differences in long-term and intermediate-term survival.

Conclusions 

We have presented an initial examination of differences between extended long-term and intermediate-term survival. Seventeen percent of our cohort survived 10 years after LTx. Avoidance of acute rejection, use of BLT, and increased hospital volume may lead to improved long-term survival in LTx.