Telomere Length and Use of Immunosuppressive Medications in Idiopathic Pulmonary Fibrosis

Abstract

Rationale: Immunosuppression was associated with adverse events for patients with idiopathic pulmonary fibrosis (IPF) in the PANTHER-IPF (Evaluating the Effectiveness of Prednisone, Azathioprine and N-Acetylcysteine in Patients with IPF) clinical trial. The reason why some patients with IPF experience harm is unknown.

Objectives: To determine whether age-adjusted leukocyte telomere length (LTL) was associated with the harmful effect of immunosuppression in patients with IPF.

Methods: LTL was measured from available DNA samples from PANTHER-IPF (interim analysis, n = 79; final analysis, n = 118). Replication cohorts included ACE-IPF (Anticoagulant Effectiveness in Idiopathic Pulmonary Fibrosis) (n = 101) and an independent observational cohort (University of Texas Southwestern Medical Center-IPF, n = 170). LTL-stratified and medication-stratified survival analyses were performed using multivariable Cox regression models for composite endpoint-free survival.

Measurements and Main Results: Of the subjects enrolled in the PANTHER-IPF and ACE-IPF, 62% (49/79) and 56% (28/50) had an LTL less than the 10th percentile of normal, respectively. In PANTHER-IPF, exposure to prednisone/azathioprine/N-acetylcysteine was associated with a higher composite endpoint of death, lung transplantation, hospitalization, or FVC decline for those with an LTL less than the 10th percentile (hazard ratio, 2.84; 95% confidence interval, 1.02–7.87; P = 0.045). This finding was replicated in the placebo arm of ACE-IPF for those exposed to immunosuppression (hazard ratio, 7.18; 95% confidence interval, 1.52–33.84; P = 0.013). A propensity-matched University of Texas Southwestern Medical Center IPF cohort showed a similar association between immunosuppression and composite endpoints (death, lung transplantation, or FVC decline) for those with an LTL less than the 10th percentile (hazard ratio, 3.79; 95% confidence interval, 1.73–8.30; P = 0.00085). An interaction was found between immunosuppression and LTL for the combined PANTHER-IPF and ACE-IPF clinical trials (P_interaction = 0.048), and the University of Texas Southwestern Medical Center IPF cohort (P_interaction = 0.00049).

Conclusions: LTL is a biomarker that may identify patients with IPF at risk for poor outcomes when exposed to immunosuppression.

At a Glance Commentary

Scientific Knowledge on the Subject

The PANTHER-IPF (Evaluating the Effectiveness of Prednisone, Azathioprine and N-Acetylcysteine in Patients with IPF) clinical trial was a landmark study that demonstrated combination therapy with prednisone, azathioprine, and N-acetylcysteine was associated with harm in patients with idiopathic pulmonary fibrosis (IPF). The reason why some patients with IPF experience harm is unknown.

What This Study Adds to the Field

In this study, we demonstrate an interaction between short leukocyte telomere length and treatment with immunosuppressive medications across two large multisite clinical trials and one academic medical center cohort. These findings suggest that short leukocyte telomere length may explain why administration of immunosuppressive medication is associated with adverse outcomes in patients with IPF. Additional studies need to determine whether leukocyte telomere length is associated with a higher risk of adverse outcomes from immunosuppressive medications in patients without IPF.

Idiopathic pulmonary fibrosis (IPF) is a progressive fibrosing lung disease that often leads to respiratory failure and death within 3–5 years (1, 2). Kindred studies have implicated telomere dysfunction caused by pathogenic variants in genes encoding components of the telomere maintenance pathway (TERT, TERC, PARN, RTEL1, NAF1, DKC1, and TINF2) in the disease mechanism of familial pulmonary fibrosis (3–9). Rare variants in several of these genes have been found in adult patients with nonfamilial IPF (4, 10, 11). Shortened age-adjusted leukocyte telomere lengths (LTLs) are commonly found in patients with IPF (11–13) and are associated with reduced survival (14–16).

Telomeres are specialized structures at the ends of chromosomes and are markers of human aging. Telomeres shorten with each round of cell division and short telomeres trigger a DNA damage response leading to cellular senescence (17, 18). Age-dependent progressive shortening of telomeres signals as a “mitotic clock” (19–21), counterbalanced by the activity of telomerase, which catalyzes the addition of nucleotide repeats to the ends of chromosomes (22, 23). Increased oxidative stress and cigarette smoking have been linked to telomere attrition (24–26). Overall, telomere length reflects an integration of the starting telomere set point, rounds of cell division, endogenous telomerase activity, telomere stability, and various environmental effects.

We have previously found that individuals with pathogenic telomere-related mutations, short telomere lengths, and non-IPF diagnoses had similar survival characteristics as those with a diagnosis of IPF (27). Because of the small numbers and variable clinical practices, we could not determine whether mortality of the patients without IPF was associated with specific medications. Patients with end-stage pulmonary fibrosis with short telomere lengths who undergo lung transplantation and subsequent long-term treatment with immunosuppressive medications have worse survival and more post-transplant complications than patients with pulmonary fibrosis with preserved telomere lengths (28–30). Given that the administration of a combination of prednisone, azathioprine, and N-acetylcysteine was associated with increased mortality, hospitalizations, and treatment-related severe adverse events in patients with IPF enrolled in a large multisite clinical trial (31), we hypothesized that adverse outcomes of patients with IPF treated with immunosuppressive medications relate to LTL.

In this study, we performed a post hoc analysis of subjects participating in the PANTHER-IPF (Evaluating the Effectiveness of Prednisone, Azathioprine and N-acetylcysteine in Patients with IPF) clinical trial (31) to determine whether LTL was associated with differential risk of composite endpoint-free survival for those treated with immunosuppressive medications as compared with placebo. Findings were replicated using subjects randomized to the placebo arm of the ACE-IPF (Anticoagulant Effectiveness in Idiopathic Pulmonary Fibrosis) clinical trial who were or were not taking immunosuppressive medications. In addition, we evaluated for an interaction between LTL and immunosuppressive medications on clinical outcomes of patients with IPF from an independent single-site academic cohort.

Methods

Study Populations

The study was approved by the institutional review board at the University of Texas Southwestern Medical Center (UTSW). Each subject provided written informed consent and a blood sample. Subjects from the PANTHER-IPF and ACE-IPF clinical trials (NCT00650091 and NCT00957242, respectively) included only those patients who consented to participate in both the parent study and the optional genetics substudy. Clinical data from PANTHER-IPF and ACE-IPF subjects were obtained from the NHLBI Biologic Specimen and Data Repository Information Coordinating Center.

Subjects from the interim analysis of PANTHER-IPF with LTL measurements who were randomized to prednisone/azathioprine/N-acetylcysteine or placebo (n = 79) were used to analyze the association between immunosuppression and outcomes in subjects stratified by LTL. Subjects from ACE-IPF who were randomized to placebo (n = 50) and treated with or without immunosuppression were used for this analysis. Subjects from ACE-IPF (n = 101) were used to analyze the interaction of LTL with warfarin. Subjects from the final analysis of PANTHER-IPF (n = 118) were used to analyze the interaction of LTL with N-acetylcysteine. An independent cohort included subjects with IPF participating in a longitudinal observational study at the UTSW (UTSW-IPF, n = 170). Subjects in the UTSW-IPF cohort who were either exposed or not exposed to immunosuppression (medical record documentation of ≥6 mo treatment with single-agent or combination of prednisone, azathioprine, or mycophenolate mofetil) were retrospectively identified. A composite outcome, including death, lung transplantation, and a relative FVC decline greater than or equal to 10% from baseline within 60 weeks of either enrollment into observational cohort study (for unexposed patients) or from the time of initiation of immunosuppression (for exposed patients) was evaluated.

Telomere Length Measurement

Genomic DNA was isolated by using QIAamp DNA Blood Maxi kit (Qiagen) from frozen blood samples collected from PANTHER-IPF and ACE-IPF subjects. Genomic DNA was isolated from blood leukocytes from the UTSW cohort using an Autopure LS instrument (Qiagen). The LTL was measured at UTSW for all samples using a quantitative polymerase chain assay and the RotoGene real-time PCR system (Qiagen) as previously described (12, 14, 32). The correlation between this method and the Southern blot method (terminal restriction fragment length analysis) was robust for 387 genomic DNA samples (Spearman rank correlation = 0.83; P < 2.2 × 10⁻¹⁶) (32). Briefly, the LTL was expressed as a logarithm-transformed ratio of telomere to single-copy gene [ln(T/S)] and this value was compared with normal control subjects (n = 201 unrelated multiethnic individuals from Dallas, TX, ranging in age from 19 to 89 yr) to calculate the age-adjusted measurement, or the leukocyte telomere length percentile for a given age. LTL was not measured for samples with a genomic DNA concentration less than 50 ng/μl (two from PANTHER-IPF interim, one from ACE-IPF, and seven from PANTHER-IPF final). The intraclass correlation for leukocyte telomere length measurements was 0.983 (95% confidence interval [CI], 0.976–0.987), 0.990 (95% CI, 0.986–0.994), and 0.980 (95% CI, 0.984–0.992) for the PANTHER-IPF, ACE-IPF, and the UTSW IPF cohort, respectively.

Statistics

Baseline differences between patients were compared using chi-square or Fisher exact test (categorical variables) and Student’s t test or one-way ANOVA (continuous variables). Adjusted P values were calculated using the Bonferroni correction when more than two pairwise comparisons were performed.

The primary goal of this study was to determine whether LTL was associated with differential risk for those treated with prednisone/azathioprine/N-acetylcysteine as compared with placebo in the PANTHER-IPF clinical trial for composite endpoint-free survival (defined as time from enrollment to death, lung transplantation, hospitalization, or decline in FVC ≥10% from baseline). The association between LTL and the combined outcome was assessed by constructing Kaplan-Meier survival curves for patients stratified by an LTL of less than the 10th or greater than or equal to the 10th percentile and comparing those randomized to prednisone/azathioprine/N-acetylcysteine with placebo using the log-rank test. We also assessed the association between LTL and the combined outcome using a multivariable Cox proportional hazards regression model that included age and baseline FVC % predicted as covariates. The number of potential covariates included in the models was limited by the number of events, therefore they were chosen based on a P less than 0.1 in univariable models for composite endpoint-free survival in the PANTHER-IPF cohort (baseline FVC % predicted P = 0.025, age P = 0.093, baseline Dl_CO % predicted P = 0.14, sex P = 0.58). An independent model included the GAP score, which accounts for all four of these variables (33). For the analysis of the combined PANTHER-IPF and ACE-IPF clinical trial cohort, a multivariable Cox model was used with stratification by clinical trial. The statistical interaction between LTL and the two therapies (prednisone/azathioprine/N-acetylcysteine or placebo) on composite endpoint-free survival was assessed by including an LTL-treatment interaction term in the multivariable Cox model (LTL, treatment, and LTL-treatment) and was considered significant if P_interaction was less than 0.05. The proportional hazards assumption was met for all Cox models. We assessed the statistical interaction between LTL and N-acetylcysteine in the PANTHER-IPF study final analysis along with LTL and warfarin in the ACE-IPF study.

We could not obtain accurate hospitalization data for the UTSW-IPF cohort. Therefore, the composite endpoint for this cohort was time from enrollment to death, transplant, or FVC decline greater than or equal to 10% from baseline. Because there were significant differences between the UTSW patients treated with immunosuppressive medications compared with those that were not treated, propensity score matching was performed to account for these baseline differences. Multivariable logistic regression was used to estimate probability of receiving immunosuppression. The covariates in the model included age, sex, smoking status, ethnicity (non-Hispanic white vs. other), baseline FVC % predicted, and baseline Dl_CO % predicted. Each patient in the UTSW cohort who received immunosuppressive medications was matched 1:1 to a patient who did not receive immunosuppressive medications using the nearest neighbor algorithm (34).

Results

LTL was measured from genomic DNA for approximately half of subjects enrolled in the PANTHER-IPF clinical trial (NCT00650091; 51% [79/155] of those included in the interim analysis) and 70% (101/145) of subjects enrolled in the ACE-IPF clinical trial (NCT00957242) (Figure 1). Demographic and baseline characteristics of PANTHER-IPF and ACE-IPF subjects for whom DNA was available or not are shown in Tables E1 and E2 in the online supplement, respectively. There was a slight difference in the ages of subjects across the different groups in PANTHER-IPF, but this difference was not observed when comparing those for whom DNA was or was not available (67.7 [7.7] DNA available vs. 68.0 [7.8] DNA not available; P = 0.76). No differences were seen in ACE-IPF.

Figure 1.

Various cohorts used to investigate an interaction between age-adjusted leukocyte telomere length (LTL) and medications with regard to clinical outcomes. Subjects participating in the PANTHER-IPF (Evaluating the Effectiveness of Prednisone, Azathioprine and N-Acetylcysteine in Patients with IPF) (NCT00650091) and ACE-IPF (Anticoagulant Effectiveness in Idiopathic Pulmonary Fibrosis) (NCT00957242) clinical trials with available genomic DNA were included in this study. Subjects from the interim analysis of PANTHER-IPF were used to determine whether LTL was associated with differential risk of composite endpoints for those treated with prednisone/azathioprine/N-acetylcysteine (“triple” therapy) as compared with placebo. Subjects from ACE-IPF who were randomized to placebo and treated with or without immunosuppression were also used for this analysis. A propensity-matched cohort collected from an academic medical center was used as a third cohort to analyze the association between immunosuppression and composite endpoints in subjects with IPF stratified by LTL. Subjects from ACE-IPF were used to analyze if LTL was associated with outcomes for those treated with warfarin as compared with placebo. Subjects from the final analysis of PANTHER-IPF were used to analyze if LTL was associated with outcomes for those treated with N-acetylcysteine as compared with placebo. NAC = N-acetylcysteine; UTSW = University of Texas Southwestern Medical Center.

PANTHER-IPF

We found that 62% (49/79) of subjects in the PANTHER-IPF cohort had an LTL less than the 10th percentile of normal (Table 1). There was no difference in LTLs between those randomized to prednisone/azathioprine/N-acetylcysteine (1.07; SD, 0.21) as compared with those randomized to placebo (1.11; SD, 0.22; P = 0.36). After stratifying by treatment and LTL, we found that those with an LTL less than the 10th percentile randomized to prednisone/azathioprine/N-acetylcysteine had a similar baseline FVC % predicted (66.8; SD, 16.2) as those randomized to placebo (70.3; SD, 9.5; P_adjust = 0.27). Likewise, those with an LTL greater than or equal to the 10th percentile randomized to placebo had similar baseline FVC % predicted (71.9; SD, 19.5) than those randomized to prednisone/azathioprine/N-acetylcysteine (80.2; SD, 11.0; P_adjust = 0.73). At baseline, we found no differences in the numbers of subjects with a cytopenia (anemia, leukopenia, thrombocytopenia) or red blood cell macrocytosis for those with an LTL less than or greater than or equal to the 10th percentile. However, we found a higher peak red blood cell mean corpuscular volume in the LTL less than the 10th percentile group randomized to prednisone/azathioprine/N-acetylcysteine compared with the other groups (Table 1).

Table 1.

PANTHER-IPF Baseline Characteristics of Patients Stratified by Treatment Arm and LTL

	Prednisone/Azathioprine/N-Acetylcysteine*		Placebo*		P Value†
	LTL < 10th %ile (n = 25)	LTL ≥ 10th %ile (n = 14)	LTL < 10th %ile (n = 24)	LTL ≥ 10th %ile (n = 16)
Age, mean (SD)	69.8 (7.1)	69.9 (6.7)	64.7 (8.2)	66.9 (8.3)	0.08
Male sex, n (%)	20 (80)	9 (64)	18 (75)	11 (69)	0.71
Non-Hispanic white, n (%)	25 (100)	13 (93)	23 (96)	14 (88)	0.34
Ever smoker, n (%)	16 (64)	10 (71)	19 (79)	11 (69)	0.70
Baseline PFT, mean (SD)
FVC % predicted	66.8 (16.2)	80.2 (11.0)	70.3 (9.5)	71.9 (19.5)	0.058
DlCO % predicted	41.6 (11.2)	47.9 (10.3)	45.5 (10.1)	45.7 (14.9)	0.41
Telomere length
ln(T/S), mean (SD)	0.96 (0.13)	1.27 (0.16)	0.99 (0.136)	1.30 (0.13)	<0.0001
Hematologic parameters
Hb, g/dl, mean (SD) nadir value	13.5 (1.2)	14.2 (1.4)	13.3 (1.3)	13.8 (1.2)	0.20
White blood cell count, 103/μl, mean (SD) nadir value	6.8 (1.6)	6.4 (1.3)	6.5 (1.1)	6.5 (1.2)	0.25
Platelet count, 103/μl, mean (SD) nadir value	197 (56)	186 (35)	194 (47)	210 (75)	0.59
Mean corpuscular volume, fl, mean (SD) peak value	103 (7)	98 (3)	97 (5)	95 (6)	0.00052
Safety endpoints, %
Death
From any cause	4 (16)	0	1 (5)	0	0.20
From respiratory cause	4 (16)	0	1 (5)	0	0.20
Hospitalization
From any cause	13 (52)	1 (7)	2 (8)	2 (13)	0.00079
From respiratory cause	6 (24)	0	0	0	0.0036
Acute exacerbation	4 (16)	0	0	0	0.046
FVC ≥10% decline	3 (12)	3 (21)	3 (13)	1 (6)	0.68
Serious adverse event‡	13 (52)	1 (7)	3 (13)	2 (13)	0.0012
Composite endpoint§	15 (60)	4 (29)	6 (25)	3 (19)	0.023

Definition of abbreviations: %ile = percentile; ln(T/S) = log-transformed telomere length relative to a single-copy gene; LTL = age-adjusted leukocyte telomere length; PANTHER-IPF = Evaluating the Effectiveness of Prednisone, Azathioprine and N-acetylcysteine in Patients with IPF; PFT = pulmonary function test.

^*Only subjects included in the interim analysis of PANTHER-IPF who were randomized to prednisone/azathioprine/N-acetylcysteine or placebo were included.

^†P values were calculated by ANOVA (continuous variables) and chi-square or Fisher exact test (categorical variables) to represent differences across four groups.

^‡Serious adverse events include those that 1) are fatal, 2) are life threatening, 3) require nonelective hospitalization >24 hours, 4) result in persistent or significant disability or incapacity, or 5) are other important medical events.

^§The composite endpoint includes death, lung transplantation, hospitalization, and FVC ≥10% decline.

Kaplan-Meier survival curves of subjects stratified by telomere length and treatment demonstrate worse composite endpoint-free survival for those randomized to the prednisone/azathioprine/N-acetylcysteine arm compared with the placebo arm with an LTL less than the 10th percentile (P = 0.0039) (Figure 2A). In contrast, we find that there is no difference in composite endpoint-free survival for patients randomized to the prednisone/azathioprine/N-acetylcysteine or placebo arms who have an LTL greater than or equal to the 10th percentile (P = 0.49) (Figure 2B).

Figure 2.

Composite endpoint-free survival of patients with idiopathic pulmonary fibrosis (IPF) from the PANTHER-IPF (Evaluating the Effectiveness of Prednisone, Azathioprine and N-Acetylcysteine in Patients with IPF) (NCT00650091) and ACE-IPF (Anticoagulant Effectiveness in Idiopathic Pulmonary Fibrosis) (NCT00957242) clinical trials. (A–D) Kaplan-Meier composite endpoint-free survival estimates of patients with IPF stratified by an LTL less than the 10th percentile or greater than or equal to the 10th percentile and exposure to immunosuppressive medications from the PANTHER-IPF (A and B) and ACE-IPF (C and D) clinical trials. The composite endpoints for these clinical trials include death, lung transplantation, FVC ≥10% decline, and hospitalization. AZA = azathioprine; IS = immunosuppressive; LTL = age-adjusted leukocyte telomere length; NAC = N-acetylcysteine; Pred = prednisone.

The association between treatment with immunosuppressive medication and composite endpoint events (death, transplant, hospitalization, and FVC decline) remained significant in the LTL less than the 10th percentile group after adjusting for age and baseline FVC % predicted (hazard ratio [HR], 2.84; 95% CI, 1.02–7.87; P = 0.045) (Table 2). For PANTHER-IPF patients with an LTL less than the 10th percentile, sensitivity analysis demonstrated that the harm associated with immunosuppression was mainly driven by hospitalizations because the association with the combined endpoint of death, transplant, or FVC decline was not significant (see Table E3).

Table 2.

Association between Immunosuppression and Composite Endpoint-Free Survival in Patients with IPF Stratified by LTL using Multivariable Cox Regression Models Adjusted for Age and Baseline FVC % Predicted

	PANTHER-IPF* (Triple Therapy and Placebo)			ACE-IPF† (Pred/AZA or No Pred/AZA)			Combined PANTHER and ACE Clinical Trial Cohorts‡			UTSW IPF§
	n (Events)	HR (95% CI)	P Value	n (Events)	HR (95% CI)	P Value	n (Events)	HR (95% CI)	P Value	n (Events)	HR (95% CI)	P Value
Telomere length, <10th %ile	49 (21)	2.84 (1.02–7.87)	0.045	28 (9)	7.18 (1.52–33.84)	0.013	77 (30)	3.91 (1.61–7.98)	0.0015	67 (33)	3.79 (1.73–8.30)	0.00085
Telomere length, ≥10th %ile	30 (7)	4.10 (0.63–26.7)	0.14	22 (6)	2.00 (0.38–10.26)	0.41	52 (13)	1.81 (0.64–5.91)	0.29	103 (33)	0.35 (0.17–0.75)	0.0066

Definition of abbreviations: %ile = percentile; ACE-IPF = Anticoagulant Effectiveness in Idiopathic Pulmonary Fibrosis; AZA = azathioprine; CI = confidence interval; HR = hazard ratio; LTL = age-adjusted leukocyte telomere length; PANTHER-IPF = Evaluating the Effectiveness of Prednisone, Azathioprine and N-Acetylcysteine in Patients with IPF; Pred = prednisone; UTSW = University of Texas Southwestern Medical Center.

^*The PANTHER-IPF cohort includes only subjects from the interim analysis with LTL measurements. These subjects had been randomized to prednisone/azathioprine/N-acetylcysteine (triple therapy) or placebo. The combined endpoint for the PANTHER-IPF cohort includes death, lung transplantation, hospitalization, and FVC ≥10% decline.

^†The ACE-IPF cohort includes only subjects with LTL measurements. These subjects had been randomized to placebo who were exposed or unexposed to prednisone or azathioprine. The combined endpoint for the ACE-IPF cohort includes death, lung transplantation, hospitalization, and FVC ≥10% decline.

^‡Analysis of the combined clinical trial cohorts was performed using a Cox proportional hazards model with stratification by cohort.

^§The UTSW IPF cohort was propensity matched for age, sex, ethnicity, smoking history, baseline FVC % predicted, and baseline Dl_CO % predicted. The combined endpoint for the UTSW cohort includes death, lung transplantation, and FVC ≥10% decline.

ACE-IPF

Because warfarin was associated with harm in the ACE-IPF study (35), we included only patients randomized to the placebo arm of this study to assess for the effect of LTL and immunosuppression on outcomes (Table 3). Review of the records of these subjects identified 36% (18/50) who were taking prednisone or azathioprine. We find a similar number of subjects with an LTL less than the 10th percentile in ACE-IPF (56%; 28/50) as was found in PANTHER-IPF (62%). There was no difference in demographics or baseline lung function across the different groups.

Table 3.

ACE-IPF Cohort Baseline Characteristics of Patients Stratified by Treatment Arm and LTL

	ACE-IPF*				P Value†
	Pred/AZA		No Pred/AZA
	LTL < 10th %ile (n = 9)	LTL ≥ 10th %ile (n = 9)	LTL < 10th %ile (n = 19)	LTL ≥ 10th %ile (n = 13)
Age	65.6 (5.4)	67.2 (9.3)	66.3 (7.4)	68.5 (7.4)	0.79
Male sex	7 (78)	7 (78)	17 (89)	9 (69)	0.54
Non-Hispanic white	8 (89)	6 (67)	18 (95)	13 (100)	0.079
Ever smoker	8 (89)	7 (78)	15 (79)	11 (85)	1.0
Pred	8 (89)	7 (78)	0	0
AZA	3 (33)	6 (67)	0	0
Baseline PFTs, mean (SD)
FVC % predicted	56.1 (16.7)	51.8 (16.7)	57.0 (16.6)	63.9 (17.3)	0.41
DlCO % predicted	29.5 (20.0)	28.1 (11.6)	36.7 (10.2)	39.7 (15.3)	0.17
Telomere length, mean (SD)
ln(T/S)	1.04 (0.15)	1.40 (0.21)	1.02 (0.12)	1.32 (0.15)	<0.0001
Safety endpoints, %
Death
From any cause	3 (33)	0	0	0	0.0023
From respiratory cause	3 (33)	0	0	0	0.0023
Hospitalization
From any cause	4 (44)	3 (33)	0	0	0.00019
From respiratory cause	1 (11)	0	0	0	0.36
Acute exacerbation	1 (11)	0	0	0	0.36
FVC ≥10% decline	2 (22)	1 (11)	2 (10)	3 (23)	0.73
Serious adverse event‡	4 (44)	3 (33)	0	2 (11)	0.021
Composite§	6 (67)	3 (33)	3 (16)	3 (23)	0.047

Definition of abbreviations: %ile = percentile; ACE-IPF = Anticoagulant Effectiveness in Idiopathic Pulmonary Fibrosis; AZA = azathioprine; ln(T/S) = log-transformed telomere length relative to a single-copy gene; LTL = age-adjusted leukocyte telomere length; PFTs = pulmonary function tests; Pred = prednisone.

^*The ACE-IPF cohort includes only subjects randomized to placebo who were exposed and unexposed to Pred or AZA.

^†P values were calculated by ANOVA (continuous variables) and chi-square or Fisher exact test (categorical variables) to represent differences across four groups.

^‡Serious adverse events include those that 1) are fatal, 2) are life threatening, 3) require nonelective hospitalization >24 hours, 4) result in persistent or significant disability or incapacity, or 5) are other important medical events.

^§The composite endpoint includes death, lung transplantation, hospitalization, and FVC ≥10% decline.

Kaplan-Meier survival curves of subjects from ACE-IPF stratified by telomere length and immunosuppression exposure demonstrated worse composite endpoint-free survival for those taking prednisone or azathioprine with an LTL less than the 10th percentile (P = 0.0049) (Figure 3A). In contrast, we find no difference in composite endpoint-free survival for those either taking or not taking prednisone or azathioprine with an LTL greater than or equal to the 10th percentile (P = 0.54) (Figure 3B).

Figure 3.

Composite endpoint-free survival of propensity-matched patients with idiopathic pulmonary fibrosis from a longitudinal academic cohort. (A and B) Kaplan-Meier composite endpoint-free survival of patients with idiopathic pulmonary fibrosis stratified by an LTL less than the 10th percentile (A) or greater than or equal to the 10th percentile (B) and exposure to immunosuppressive medications. Patients with idiopathic pulmonary fibrosis are propensity-matched for age, sex, ethnicity, smoking history, baseline FVC % predicted, and baseline Dl_CO % predicted. The composite endpoints include death, lung transplantation, and FVC ≥10% decline. AZA = azathioprine; IS = immunosuppressive; LTL = age-adjusted leukocyte telomere length; Pred = prednisone; UTSW = University of Texas Southwestern Medical Center.

The association between immunosuppressive medication and composite endpoint events (death, transplant, hospitalization, and FVC decline) remained significant in the group with shorter LTL after adjusting for age and baseline FVC % predicted in ACE-IPF (HR, 7.18; 95% CI, 1.52–33.84; P = 0.013) and the combined PANTHER and ACE clinical trial cohort (HR, 3.91; 95% CI, 1.68–9.10; P = 0.0015) (Table 2). Similar results were obtained when adjusting for the GAP score (33), which accounts for four clinical predictors of mortality (data not shown).

Independent IPF Cohort

We used a longitudinal academic IPF cohort that has been enrolling since 2003 (before the release of the PANTHER-IPF clinical trial results in 2012 [31]) to independently assess the effect of exposure to immunosuppressive medications. We found significant differences in age (P = 0.0019) and baseline lung function (FVC % predicted, P = 0.0032; Dl_CO % predicted, P = 0.00018) for patients with IPF prescribed one or more immunosuppressive medications (prednisone n = 43, azathioprine n = 28, and mycophenolate mofetil n = 14) for more than 6 months (Table 4). After propensity score matching for age, sex, ethnicity, smoking history, baseline FVC % predicted, and baseline Dl_CO % predicted, differences between drug-exposed and no drug-exposed groups were no longer evident. In this propensity-matched cohort, exposure to immunosuppression was associated with a worse composite endpoint-free survival (defined as death, transplant, or FVC ≥10% decline) within 60 weeks in the LTL less than the 10th percentile group (P = 0.0014) (Figure 3A) but was associated with an improved outcome in the LTL greater than or equal to the 10th percentile group (P = 0.0057) (Figure 3B). The associations remained significant after adjusting for age and baseline FVC % predicted (HR, 3.79; 95% CI, 1.73–8.30; P = 0.00085 for the LTL <10th percentile group) (HR, 0.35; 95% CI, 0.17–0.75; P = 0.0066 for the LTL ≥10th percentile group).

Table 4.

Unmatched and Propensity-matched UTSW IPF Cohort Baseline Characteristics of Patients Stratified by Treatment

	Unmatched				Propensity Matched*
	Drug† (n = 88)	No Drug (n = 140)	SMD	P Value	Drug† (n = 85)	No Drug (n = 85)	SMD	P Value
Age	63.7 (9.2)	67.7 (9.5)	0.428	0.0019	63.9 (9.0)	66.1 (9.1)	0.236	0.13
Male sex	68 (77)	102 (73)	0.102	0.56	65 (76)	65 (76)	<0.001	1.0
Non-Hispanic white	74 (84)	122 (87)	0.087	0.65	71 (84)	73 (86)	0.065	0.83
Ever smoker	60 (68)	83 (59)	0.185	0.23	57 (67)	53 (62)	0.098	0.63
Baseline PFTs, mean (SD)
FVC % predicted	59.4 (16.9)	67.5 (22.9)	0.403	0.0032	59.4 (16.9)	62.9 (19.0)	0.202	0.19
DlCO % predicted	35.9 (14.9)	45.1 (20.4)	0.514	0.00018	35.9 (14.9)	38.1 (13.0)	0.161	0.30
Telomere length, mean (SD)
ln(T/S)	1.27 (0.33)	1.24 (0.31)	0.079	0.57	1.27 (0.34)	1.23 (0.31)	0.119	0.44
<10th percentile	35 (40)	52 (38)	0.037	0.89	33 (39)	34 (40)	0.024	1.0
Endpoint, n (%)
Death	17 (19)	17 (12)	0.197	0.20	16 (19)	14 (16)	0.061	0.84
FVC ≥10% decline	17 (19)	19 (14)	0.149	0.35	17 (20)	14 (16)	0.091	0.69
Composite‡	36 (41)	42 (30)	0.228	0.12	34 (40)	32 (38)	0.048	0.87

Definition of abbreviations: IPF = idiopathic pulmonary fibrosis; ln(T/S) = log-transformed telomere length relative to a single-copy gene; PFTs = pulmonary function tests; SMD = standardized mean difference; UTSW = University of Texas Southwestern Medical Center.

^*Cohort propensity matched for age, sex, ethnicity, smoking history, baseline FVC % predicted, and baseline Dl_CO % predicted.

^†Individuals in the drug group were prescribed prednisone (n = 43), azathioprine (n = 28), and/or mycophenolate (n = 14) for ≥6 months’ duration.

^‡The composite endpoint includes death, lung transplantation, and FVC ≥10% decline.

Interaction between LTL and Exposure to Immunosuppressive Medications

The interaction effect between LTL as a continuous variable and exposure to immunosuppressive medications was significant for the composite endpoint in the combined clinical trial cohort (P_interaction = 0.048) (Table 5). The interaction effect between LTL, either as a continuous variable (P_interaction = 0.00049) or a dichotomous variable (P_interaction < 0.0001), and exposure to immunosuppressive medications was significant for the composite endpoint in the propensity-matched UTSW cohort. In contrast, we find no evidence for an interaction between LTL and exposure to either N-acetylcysteine or warfarin on clinical outcomes (see Table E4).

Table 5.

Interaction Effects between LTL and Immunosuppressive Medication Exposure on Combined Endpoints

	PANTHER-IPF* (n = 79) (Events = 28)	ACE-IPF* (n = 50) (Events = 15)	Combined PANTHER and ACE Clinical Trials* (n = 129) (Events = 43)	UTSW Cohort† (n = 170) (Events = 66)
LTL continuous, ln(T/S)	0.12	0.11	0.048	0.00049
LTL dichotomous, <10th %ile	0.38	0.22	0.17	<0.0001

Definition of abbreviations: %ile = percentile; ACE-IPF = Anticoagulant Effectiveness in Idiopathic Pulmonary Fibrosis; ln(T/S) = log-transformed telomere length relative to a single-copy gene; LTL = age-adjusted leukocyte telomere length; PANTHER-IPF = Evaluating the Effectiveness of Prednisone, Azathioprine and N-Acetylcysteine in Patients with IPF; UTSW = University of Texas Southwestern Medical Center.

^*The PANTHER-IPF cohort includes only subjects from the interim analysis randomized to prednisone/azathioprine/N-acetylcysteine or placebo. The ACE-IPF cohort includes only subjects randomized to placebo who were exposed and unexposed to prednisone or azathioprine. The composite endpoint for these clinical trials includes death, lung transplantation, hospitalization, and FVC ≥10% decline.

^†The UTSW IPF cohort was propensity matched for age, sex, ethnicity, smoking history, baseline FVC % predicted, and baseline Dl_CO % predicted. The composite endpoint for the UTSW cohort includes death, lung transplantation, and FVC ≥10% decline.

Discussion

In this study we demonstrate an interaction in patients with IPF between LTL and exposure to immunosuppressive medications with regard to the composite endpoint of death, transplant, FVC decline, or hospitalization across two large multisite clinical trials (PANTHER-IPF and ACE-IPF). LTL has previously been shown to be a prognostic biomarker for patients with IPF (14–16) in predicting survival. This study provides a biologic explanation for the interim results of the PANTHER-IPF clinical trial and presents the novel finding that LTL is a pharmacogenomic biomarker that identifies patients with IPF at risk for adverse clinical outcomes when exposed to immunosuppressive medications.

We find that the numbers of patients with IPF from these two multisite clinical trials include a disproportionately large number of subjects (56–62%) with LTLs below the 10th percentile. In comparison, approximately 80% of subjects with pathogenic variants in telomere-maintenance genes (TERT, TERC, RTEL1, and PARN) have an LTL less than the 10th percentile (5). We have previously reported that up to 35% of patients with IPF collected from three independent academic sites have LTLs less than the 10th percentile (14). An LTL less than the 10th percentile has been found in 17–36% of patients with chronic hypersensitivity pneumonia (36). It is possible that the large number of patients with IPF with shortened LTLs in PANTHER-IPF and ACE-IPF was caused by increased enrollment of those with a family history of disease in the genetic substudies. Alternatively, because inclusion into both these studies required adjudication by the IPF Clinical Research Network (IPF-Net) to ensure a confident diagnosis of IPF (37), it is possible that these stringent inclusion criteria for the IPF-Net clinical trials preferentially included patients with shorter telomere lengths.

The data from this study suggest a pharmacogenomic interaction related to an individual’s genetic makeup (blood LTL) on his or her response to immunosuppressive medications. The TOLLIP SNP rs3750920 has been recently shown to demonstrate a gene-drug interaction with N-acetylcysteine but not immunosuppression in patients with IPF (38). The data presented in this study suggest that the pharmacogenomics interaction for LTL exists for drugs with an immunosuppressive class effect. We found a similar adverse effect of the combination of “triple” therapy (prednisone, azathioprine, and N-acetylcysteine) in PANTHER-IPF as was seen with single drug or combination drug therapy of prednisone and azathioprine in ACE-IPF or prednisone, azathioprine, and/or mycophenolate mofetil in the UTSW IPF cohort. Because too few patients were exposed to each individual immunosuppressive medication, we were unable to determine whether the telomere length–drug interaction was different for the various medications.

The analysis of data from both IPF-Net clinical trials suggests that hospitalizations and serious adverse events were associated with use of immunosuppressive medications in patients with LTL less than the 10th percentile. Most of the hospitalizations were not related to respiratory causes, but the exact nature of each were not documented. The small number of patients completing the 60-week follow-up point in the IPF-Net clinical trials limited our ability to analyze the effect of immunosuppression and LTL on the rate of lung function (FVC) decline. The same endpoints reported in the interim analysis of the PANTHER-IPF trial (death, hospitalization, and serious adverse events) were used in this study. Each of these variables has been shown to be clinically relevant for IPF (39–41). We used a composite endpoint to improve statistical power given the small numbers of patients and events.

There is ample evidence for a baseline immunodeficiency phenotype in patients with dyskeratosis congenita and extremely short telomere lengths. They often exhibit lymphopenia, reduced numbers of B and natural killer cells, impaired lymphocyte proliferation, and dysregulated T-cell signaling (42–45). Similarly, telomere shortening has been associated with T-cell replicative senescence in aging adults (46). Shortened telomere length is associated with higher rates of upper respiratory tract infections (47) and increased mortality in older individuals when infections occur (48). It is very likely that an immune dysfunction phenotype is unmasked when patients with IPF with short telomere length are exposed to cytotoxic or immunomodulatory medications.

There are several limitations of this study. This was a post hoc analysis of two different multisite clinical trials that were not designed to assess for genomic interactions with treatment-related clinical outcomes. Thus, these findings should be viewed as hypothesis generating and not indicative of a causal relationship. Only a portion of patients in each clinical trial provided a DNA sample for evaluation, which significantly limited the sample size and power to evaluate individual clinical endpoints. However, despite the small numbers of events and PANTHER-IPF subjects with available DNA, a significant association between telomere length and a reduced composite-endpoint-free survival was found in patients exposed to immunosuppressive medications. Unaccounted differences in subjects who agreed to participate in the genetics substudies, such as differences in enrollment by site, may bias these results. The telomere length assay we used is less precise and more variable than others (49), but was used because genomic DNA samples were available and fresh blood samples were not. We cannot exclude the possibility that the difference in the number of subjects with LTL less than the 10th percentile across the cohorts may be caused by technical variables. Because the ACE-IPF trial cohort was not designed to assess the efficacy of immunosuppression, use of these medications was not factored into subject randomization and the doses and total exposures were not available. The UTSW IPF cohort had significant baseline differences between those exposed and not exposed to immunosuppression and is subject to confounding by indication. Even though we performed propensity score matching using predictors that would influence the decision to start therapy or not, there may be unaccounted factors that could bias the composite outcomes. The timing of dose changes and the particular immunosuppressive medications for the UTSW cohort were not uniform.

Although our results are interesting, they are unlikely to change the management of patients with IPF in the current era of antifibrotic medications where at least one (pirfenidone) has been shown to be effective in patients regardless of telomere length (11). However, the use of immunosuppression remains part of the medical armamentarium for many progressive non-IPF forms of interstitial lung disease (ILD), including hypersensitivity pneumonitis and connective tissue disease–associated ILD. If the pharmacogenomics interaction between telomere length and immunosuppression medication is independent of ILD diagnosis, then it is possible that short telomere length patients without IPF may be harmed by these therapies. Additional study is needed especially for chronic hypersensitivity pneumonitis– and rheumatoid arthritis–associated ILD, which often behave like IPF with progressive fibrosis despite treatment with immunosuppressive medications (50). However, as this study has shown, accounting for differences in demographic characteristics, different dosages, or use of different immunosuppressive medications can plague any retrospective cohort study. Prospective studies for non-IPF diseases in which immunosuppression is currently used as standard practice is needed to determine whether LTL is a modifier of response to immunosuppressive therapy.

In summary, we find that the risk of the composite endpoint with treatment of immunosuppressive medications in patients with IPF is informed by age-adjusted telomere length across multiple independent cohorts. The division of patients with IPF into two groups by telomere length that demonstrate drastically different outcomes implores the practice of routine genetic sampling in future clinical trials. These findings recapitulate prior observations that shortened LTLs are associated with adverse clinical outcomes in IPF and implicate immunosuppressive medications as a contributing cause. Finally, this study highlights the genetic vulnerability of short telomere patients to immunosuppression. Individuals with less common manifestations of telomere syndromes (51), such as bone marrow failure and liver disease, may be similarly harmed by immunosuppressive therapies. These findings provide a basis for exploring treatment strategies tailored by telomere length for patients with non-IPF diagnoses.