Adipose tissue quantification and primary graft dysfunction after lung transplantation: The Lung Transplant Body Composition study

Abstract

Background:

Obesity is associated with increased risk of primary graft dysfunction (PGD) after lung transplantation. The contributions of specific adipose tissue depots is unknown.

Methods:

We performed a prospective cohort study of adult lung transplant recipients at four U.S. transplant centers. We measured cross-sectional areas of subcutaneous (SAT) and visceral adipose tissue (VAT) on chest and abdominal CT scans and indexed each measurement to height². We used logistic regression to examine associations of adipose indices and adipose classes with grade 3 PGD at 48 or 72 hours, and Cox proportional hazards models to examine survival. We used latent class analyses to identify patterns of adipose distribution. We examined associations of adipose indices with plasma biomarkers of obesity and PGD.

Results:

262 and 117 subjects had available chest CT scans and underwent protocol abdominal CT scans, respectively. In adjusted models, greater abdominal SAT index was associated with an increased risk of PGD (OR 2.0, 95% CI 1.03 to 4.1, p=.04) but not with survival time. VAT indices were not associated with PGD risk or survival time. Greater abdominal SAT index correlated with greater pre- and post-transplant leptin (r=.60, p<0.001, and r=0.42, p<0.001), pre-transplant IL-1RA (r=.25, p=0.04), and post-transplant ICAM-1 (r=.25, p=0.04). We identified three latent patterns of adiposity. The class defined by high thoracic and abdominal SAT had the greatest risk of PGD.

Conclusions:

Subcutaneous, but not visceral, adiposity is associated with increased risk of PGD after lung transplantation.

Introduction

Primary graft dysfunction (PGD), defined as acute lung injury within 72 hours of lung transplantation (1), affects 12 to 32% of all lung transplants, is a major cause of death within the first year after transplantation (2–5), and is a risk factor for bronchiolitis obliterans syndrome (6). There are no targeted therapies for PGD; treatment relies on supportive therapy focused on prevention of additional injury while awaiting lung recovery. A better understanding of risk factors for PGD will help identify patients at increased risk before transplant, as well as target pathways for prevention and treatment.

Obesity is a chronic pro-inflammatory state associated with changes in adipose tissue inflammation, and a risk factor for PGD after lung transplantation (7). Adipocyte hypertrophy, resulting from storage of excess fatty acids, produces interleukin-6 (IL-6), tumor necrosis factor alpha (TNF-α), and monocyte chemoattractant protein-1 (MCP-1) which along with increased free fatty acids (8–11), and increased leptin (12), recruits pro-inflammatory adipose tissue macrophages (ATMs) (8, 13). Pro-inflammatory ATMs produce interleukin-1 beta (IL1-β) which further stimulates adipocyte production of IL-6 and MCP-1 (14), a process that is decreased in the presence of interleukin-1 receptor antagonist (IL-1RA) (15). Visceral adipose tissue (VAT) contains a greater number of pro-inflammatory ATMs, although subcutaneous adipose tissue (SAT) in the obese also has an inflammatory phenotype with a similar number of ATMs (16) and increased T cell chemokines (17).

Subcutaneous and visceral adipose depots in both abdominal and thoracic compartments are associated with disease states and outcomes. Increased abdominal SAT is associated with increased cardio-metabolic risk factors (18, 19), incident heart failure (20), and acute kidney injury (21). Increased abdominal VAT is associated with decreased survival after liver transplantation (22, 23), while increased intrathoracic VAT has been associated with increased risks of atrial fibrillation (24), coronary artery disease (25, 26), systemic hypertension (25), and reduced right ventricular mass (27). The role of various adipose depots and obesity-related inflammation in PGD remains unknown.

Although obesity, as defined by the body mass index (BMI) is a poor surrogate for adipose depot distribution (28–34), it remains the primary metric used by centers to establish transplant candidacy. A better measure of obesity could identify novel targets for prevention and treatment and ultimately improve risk stratification prior to lung transplantation. Thus we examined associations between the size of various adipose tissue depots on computed tomography (CT) imaging and PGD risk after lung transplantation. We secondarily examined the association of both adipose indices and biomarkers of adiposity with PGD. Our a priori hypotheses specified that greater SAT and VAT in either compartment would be associated with an increased risk of PGD after lung transplantation.

Methods

Study Participants

The Lung Transplant Body Composition (LTBC) study is an NHLBI-funded multi-center prospective cohort study designed to examine the relationship between adiposity and PGD after lung transplantation. LTBC is nested within the Lung Transplant Outcomes Group (LTOG) study, which examines risk factors for PGD (2). Subjects were eligible for enrollment in LTBC if they were at least 18 years of age and undergoing lung transplant evaluation between 2011 and 2014 at the University of Pennsylvania, Columbia University, or Duke University. Additional subjects from the University of California at San Francisco who were enrolled in LTOG and had available clinically indicated chest CT scans were added to the cohort after LTBC enrollment had completed.

LTBC and LTOG studies were approved by Institutional Review Boards at all participating centers, and all participants provided informed consent prior to enrollment. Details of LTOG data collection and variable definitions have been previously described (2, 7, 35).

Measurement of Abdomen CT Adipose Tissue Area

A subset of LTBC participants consented to undergo a single slice (10 mm) abdominal CT scan at the level of the L4/L5 vertebrae using Siemens or GE scanners. Techniques used to identify abdominal subcutaneous and visceral adipose tissues have been previously described and are provided in the Supplement (21, 36) (Figure 1A–B).

Figure 1:

Selected axial images from computed tomography scan of (A) abdominal visceral adipose tissue at L4/L5 vertebral level, (B) abdominal subcutaneous adipose tissue at L4/L5 vertebral level, (C) thoracic visceral adipose tissue at mid-T7 vertebral body, and (D) thoracic subcutaneous adipose tissue at mid-T8 vertebral body.

Measurement of Chest CT Adipose Tissue Area

Clinically-indicated non-research pre-transplant unenhanced full inspiration CT scans were used for thoracic adipose quantification. Using a standardized anatomic space approach (37, 38), our group previously identified that the maximum correlation between total volume of thoracic SAT and single slice area of thoracic SAT is found at the mid-T8 vertebral level (Pearson’s r= 0.97), and maximum correlation between total volume of thoracic VAT and single slice area of thoracic VAT is found at the mid-T7 vertebral level (Pearson’s r= 0.86) (37). Using this technique, we identified the single slice cross-sectional area of thoracic VAT and thoracic SAT (Figure 1C–D). There was high interrater reliability with an intra-class correlation coefficient of .98 for thoracic SAT and .97 for thoracic VAT (Supplement Figures 1A–B). Readers were blinded to PGD status.

Plasma biomarkers

Plasma samples were obtained prior to and 24 hours after transplantation and stored at −80°C as previously published (39). Circulating biomarkers of obesity (leptin), coagulation (plasminogen activator inhibitor 1, PAI-1), cell adhesion (intracellular adhesion molecule-1, ICAM-1), innate immunity (long pentraxin-3, PTX3), inflammatory modulation (IL-1RA), and macrophage recruitment and T cell differentiation (IL-6), were chosen based on known associations with PGD and/or obesity (7, 39–45). Biomarker levels were measured on multiplex assays using Meso Scale Discovery platforms and were log₂ transformed to normalize the distributions. Sample concentrations below the lower limit of detection were assigned values equal to halfway between zero and the lower limit of detection for the assay.

Primary Graft Dysfunction

All chest radiographs were collected at the time of transplantation and transmitted to the University of Pennsylvania where they were reviewed by two independent observers with adjudication as previously published (2, 35). The primary outcome was grade 3 PGD at 48 or 72 hours after transplantation defined as a PaO₂/FiO₂ ratio <200 with parenchymal opacities consistent with diffuse pulmonary edema (1, 35, 46, 47).

Analysis Approach

We operationalized adipose area in two fashions: (1) absolute cross-sectional area, and (2) area indexed to height-squared (m²) (48). We log₂ transformed the adipose measures to normalize distributions and allow estimation of associations per doubling of each independent variable of interest. We evaluated the convergent and divergent validity of abdominal and thoracic adipose indices with other measures of body composition using Pearson correlation coefficients.

We used logistic regression models to examine associations between each independent variable of interest and PGD risk. We performed our primary analyses using generalized covariate balanced propensity scores to adjust for groups of covariates using a single propensity score (49–51) in order to avoid over-parameterized models. We performed multiple imputation with chained equation in ten datasets using the “MICE” package in R (51). We derived propensity scores in each imputed dataset and included the propensity scores as imputed covariates in our models. We used directed acyclic graphs (DAG) to identify the following confounders (Supplement Figure 2A–B), which we included in a “minimally adjusted” propensity score: intraoperative transfusion requirement, allograft ischemic time, intraoperative pulmonary artery systolic pressure, intraoperative use of extracorporeal support, and center. We also adjusted models by covariate groups defined as recipient characteristics (sex, pre-transplant diagnosis, lung allocation score (LAS), center), donor characteristics (total ischemic time, donor smoking), and operative characteristics (single vs. double lung transplantation, intraoperative use of cardiopulmonary bypass or extracorporeal membrane oxygenation, intraoperative transfusions greater than 1 liter of packed red blood cells). In order to demonstrate the durability of the effect estimates, (1) we performed sensitivity analyses with models adjusted for individual covariates, and (2) we have presented our results in a “model-building” format in which we sequentially adjusted for groups of covariates. We explored the use of non-linear terms for adipose indices using generalized additive models with the “GAM” function in R (52). In order to compare the relative contributions of the SAT index and BMI to PGD risk, we calculated R-squared values for models of PGD with: (1) minimally adjusted covariates, (2) minimally adjusted covariates plus BMI, and (3) minimally adjusted covariates plus the abdominal SAT index.

We evaluated the association between adipose depots associated with PGD and biomarkers of adiposity and PGD using Pearson correlation coefficients.

We performed latent class analysis to identify subgroups of subjects that were similar on observable characteristics (abdominal and thoracic SAT area, abdominal and thoracic VAT area, age, and sex) (53, 54). We compared the Bayesian Information Criteria to identify the optimal number of classes and the entropy to assess adequacy of fit. Each subject was assigned to the class to which they had the highest predicted probability of belonging.

We used multivariable adjusted Cox proportional hazards models to examine associations between adipose tissue area and survival time (time from transplantation to death) after transplantation. We confirmed the proportional hazards assumption by regressing Schoenfeld residuals over time.

All analyses were performed in STATA/IC version 15.1 (StataCorp, LP College Station, TX) (55) and R version 3.3.1 (R Foundation for Statistical Computing)

Results

Study Participants

Three-hundred fifty-six enrolled subjects underwent lung transplantation and had either an available chest CT or underwent single slice CT of the abdomen as part of the research protocol. Among these subjects, one was missing PGD assessment, 6 were missing chest imaging, and 21 were missing height, leaving 328 for inclusion in thoracic VAT analyses (Figure 2A). An additional 66 subjects had clinically-indicated chest CTs that cut off a portion of SAT, leaving 262 scans for inclusion in analyses of thoracic SAT. A total of 120 subjects consented to undergo abdominal imaging, of whom 117 had available measures for abdominal SAT and 116 had available measures for abdominal VAT (Figure 2B). Plasma biomarkers were available on a subset of 130 subjects. Due to availability of adequate plasma volume, some subjects did not have all biomarkers analyzed (Supplement Table 1).

Figure 2:

Study flow for (A) thoracic measures, and (B) abdominal measures.

Footnote:

Definition of abbreviations: PGD=primary graft dysfunction; VAT Index = visceral adipose tissue area/height²; SAT Index= subcutaneous adipose tissue area/height².

Among the entire cohort, the median (interquartile range, IQR) age was 61 (53–66) years, 59% were male, 75% white, 9% African-American, 17% had chronic obstructive pulmonary disease (COPD), 64% had interstitial lung disease (ILD), and 53% required intraoperative mechanical support. The median LAS at time of transplantation was 41.6 (36.1 to 51.0) and the median BMI was 25.1 kg/m² (IQR 22.0 to 28.2). Donors were 83% male, 25% had a history of cigarette smoking, 42% were white, and 16% were African-American (Supplement Table 2). Among subjects with abdominal CT scans, there was a median (IQR) of 182 days (83-397) between the abdominal CT scan and transplantation. Among subjects with thoracic CT scans, there was a median (IQR) of 150 days (79-294) between the chest CT and transplantation. Twenty percent (n=70) developed grade 3 PGD at 48 or 72 hours after transplantation. The one-year mortality rate was 7.5 deaths per 100 person-years (95% CI 4.9 to 10.9). A total of sixty-three (18%) subjects died over a median 2.7 years after transplantation.

Compared to those in the thoracic SAT cohort, subjects in the abdominal SAT cohort were more likely to be white, more likely to have a male donor, more likely to have a donor with a history of smoking, more likely to have a white donor, and less likely to require intraoperative mechanical circulatory support (Table 1). Baseline characteristics by availability of plasma biomarkers and tertile of adipose measures are available in the Supplement (Supplement Tables 3–4).

Table 1:

Overall subject characteristics for chest and abdomen cohorts

	Abdominal SAT cohort (n=117)	Thoracic SAT Cohort (n=262)
Recipient Variables
Age, years	62 (48-68)	62 (53-67)
Male gender	71 (61)	164 (63)
LAS at transplantation	39.5 (35.0 to 44.7)	40.3 (35.7–47.7)
Diagnosis
COPD	24 (21)	53 (20)
Interstitial lung disease	63 (54)	156 (60)
Sarcoidosis	5 (4)	10 (4)
Cystic fibrosis	15 (13)	25 (10)
Pulmonary arterial hypertension	3 (3)	7 (3)
Other	7 (6)	11 (4)
Race
White	101 (86)	214 (82)
African American	11 (9)	24 (9)
Other	5 (4)	15 (6)
Height, cm	172.7 (162.6 to 177.8)	172.0 (164.0-177.8)
Weight, kg	75.1 (60.3 to 83.9)	74.2 (61.5-83.3)
BMI, kg/m2	24.9 (21.5 to 27.9)	24.7 (21.5-27.6)
BMI Category
<18.5	11 (9)	26 (10)
18.5-25	49 (42)	115 (44)
25-30	39 (33)	83 (32)
30-35	16 (14)	36 (14)
>35	2 (2)	2 (1)
Abdominal VAT area (cm2)	85.5 (44.0 to 133.6)	88.1 (44.9-139.0)
Abdominal VAT Index (cm2/m2)	30.0 (16.0 to 45.3)	30.2 (16.5-47.2)
Abdominal SAT area (cm2)	220.6 (160.4 to 316.1)	215.2 (154.7-313.0)
Abdominal SAT Index (cm2/m2)	72.2 (53.7 to 109.9)	71.4 (52.2-104.5)
Thoracic VAT area (cm2)	11.4 (5.3-22.4)	13.5 (6.7-23.7)
Thoracic VAT Index (cm2/m2)	4.1 (1.9-7.7)	4.5 (2.3-8.1)
Thoracic SAT area (cm2)	119.5 (75.2-193.8)	122.3 (75.6-176.0)
Thoracic SAT Index (cm2/m2)	40.5 (25.6-66.8)	42.1 (25.4-61.2)
Donor Variables
Gender, male	73 (62)	128 (49)
Any smoking history	49 (42)	75 (29)
Race
White	78 (67)	122 (47)
African American	27 (23)	39 (19)
Other	12 (10)	91 (35)
Procedure Variables
Ischemic time, minutes	369 (301 to 420)	340 (271-402)
Transplant type - double	79 (68)	177 (68)
PRBC > 1 L	16 (14)	41 (16)
CPB or ECMO	43 (37)	124 (47)
PASP, mmHg	38 (31 to 47)	38 (30-47)
Center
Center A	47 (40)	57 (22)
Center B	19 (16)	61 (23)
Center C	51 (44)	62 (24)
Center D	NA	82 (31)

Continuous variables are presented as median (interquartile range). Categorical variables are presented as percentages. Percentages may not exactly equal 100% because of rounding.

Missing data for abdominal cohort: 5 are missing donor smoking, 10 are missing pulmonary artery systolic pressure, 11 are missing transfusion, 4 missing abdominal VAT measurement, 6 are missing thoracic SAT measurement, 5 are missing thoracic VAT measurement.

Missing data for chest cohort: 2 are missing sex, 9 are missing race, 62 are missing donor sex, 10 are missing donor smoking, 14 are missing total ischemic time, 32 are missing pulmonary artery systolic pressure, 17 are missing transfusion, 4 are missing use of CPB or ECMO, 161 are missing abdominal SAT measures, , 162 are missing abdominal VAT measurements.

Definition of abbreviations: LAS = lung allocation score at transplantation; COPD = chronic obstructive pulmonary disease; VAT = visceral adipose tissue area; VAT index = visceral adipose tissue area divided by height²; SAT= subcutaneous adipose tissue area; SAT index= subcutaneous adipose tissue area divided by height²; CPB = intraoperative use of cardiopulmonary bypass; ECMO = intraoperative use of extracorporeal membrane oxygenation; PRBC = packed red blood cells; PASP = pulmonary artery systolic pressure.

Convergent and Divergent Validity of CT Measures of Adiposity

Abdominal and thoracic VAT and SAT indices were highly correlated with each other as well as BMI, waist circumference, hip circumference, and waist to hip ratio (Table 2). BMI was most closely correlated with SAT area in both the abdomen (r=.74) and chest (r=.68). Waist circumference was most closely correlated with VAT area in the abdomen (r=.74) and SAT area in the chest (r=.65). Hip circumference was most closely correlated with SAT area in the abdomen (r=.77) and chest (r=.65). Waist to hip ratio was most closely correlated with VAT area in the abdomen (r=.57) and chest (r=.47).

Table 2:

Convergent and divergent validity of adipose area and indexed adipose area with other measures of body composition. Values represent Pearson correlation coefficients.

	Abdomen				Chest
	VAT Area	VAT Index	SAT Area	SAT Index	VAT Area	VAT Index	SAT Area	SAT Index
Abdomen VAT Area	1
Abdomen VAT Index	0.99*	1
Abdomen SAT Area	0.62*	0.64*	1
Abdomen SAT Index	0.56*	0.60*	0.98*	1
Chest VAT Area	0.62*	0.59*	0.43*	0.36*	1
Chest VAT Index	0.62*	0.60*	0.46*	0.40*	0.99*	1
Chest SAT Area	0.66*	0.68*	0.87*	0.87*	0.52*	0.53*	1
Chest SAT Index	0.60*	0.64*	0.85*	0.88*	0.46*	0.49*	0.99*	1
Body mass index	0.58*	0.59*	0.74*	0.71*	0.47*	0.47*	0.67*	0.65*
Waist Circumference	0.74*	0.71*	0.69*	0.62*	0.61*	0.59*	0.65*	0.57*
Hip Circumference	0.58*	0.56*	0.77*	0.72*	0.43*	0.42*	0.69*	0.62*
Waist to Hip Ratio	0.57*	0.55*	0.25*	0.19	0.47*	0.46*	0.25*	0.20*
Six-minute walk distance	0.02	−0.01	−0.06	−0.10	0.12	0.10	−0.18*	−0.21

Definition of abbreviations: VAT = visceral adipose tissue area; VAT Index = visceral adipose tissue area divided by height²; SAT= subcutaneous adipose tissue area; SAT Index = subcutaneous adipose tissue area divided by height²

Waist circumference was available for 181 subjects, hip circumference was available for 182 subjects, waist to hip ratio was available for 181 subjects, six-minute walk distance was available on 239 subjects.

^*p ≤ 0.05

Subcutaneous adipose tissue

Among subjects with abdominal SAT measured, 23 (20%) developed grade 3 PGD at 48 or 72 hours after transplantation. Each doubling in the abdominal SAT index was associated with 1.9-fold increased odds of PGD in both minimally (OR 95% CI 1.02 to 3.4, p=.04, Table 3, Figure 3A) and fully adjusted models (OR 95% CI 1.1 to 3.4, p=.03). The association was similar in analyses adjusted for propensity scores for recipient characteristics, donor characteristics, and operative characteristics (Table 3) and in additional sensitivity analyses (Supplement Table 5–6). Results were similar when evaluating the association between abdominal SAT area (rather than index) and PGD risk (Supplement Table 7, Supplement Figure 3A).

Table 3:

Associations between subcutaneous adipose tissue index and grade 3 primary graft dysfunction at 48 or 72 hours.

	All subjects with available SAT Index
Model	Abdomen (N=117)			Chest (N=262)
	OR per doubling of SAT index	95% CI	P	OR per doubling of SAT index	95% CI	P
PGD models
Minimally Adjusted ¥	1.9	1.02 to 3.4	.04	1.1	0.8 to 1.5	.53
Adjusted for recipient variables ≠	1.9	1.1 to 3.4	.03	1.1	0.8 to 1.5	.63
+ Donor $	1.9	1.03 to 3.6	.04	1.1	0.8 to 1.5	.64
+ Operative§	2.0	1.1 to 3.6	.04	1.1	0.8 to 1.5	.53

Definition of abbreviations: SAT index = subcutaneous adipose tissue area divided by height², CI = confidence interval, OR = odds ratio.

^¥Minimally Adjusted Characteristics are transfusion requirement, allograft ischemic time, pulmonary artery systolic pressure, intraoperative use of cardiopulmonary bypass or extracorporeal membrane oxygenation, center

^≠Recipient Characteristics are diagnosis, gender, pulmonary artery systolic pressure, lung allocation score, center

^$Donor Characteristics are allograft ischemic time, donor smoking

^§Operative Characteristics are use of extracorporeal support, transplant type, intraoperative transfusions >1L packed red blood cells

Figure 3:

Continuous association between (A) Abdominal Subcutaneous Adipose Index and grade 3 PGD at 48 or 72 hours, and (B) Thoracic Subcutaneous Adipose Index and grade 3 PGD at 48 or 72 hours. P values for association are (A) p=0.045, and (B): p=0.54. Values differ from main analyses due to differences in modeling technique. Dark dotted black line represents the effect estimates. Surrounding thin lines represent 95% confidence bands. Each vertical line in the rug plot along the x axis represents a single study subject.

Among subjects with available thoracic SAT measured, 45 (17%) developed grade 3 PGD at 48 or 72 hours after transplantation. We did not detect significant associations between either thoracic SAT index (Table 3, Figure 3B) or thoracic SAT area (Supplement Table 7, Supplement Figure 3B) and PGD risk in any models. In analyses of abdominal or thoracic SAT index among subjects with all 4 available scans, we did not detect a significant association between SAT index or SAT area and PGD or death after transplantation (Supplement Tables 6–7).

Addition of the abdominal SAT index to our models improved the fit of the minimally adjusted model by 55%, while the addition of BMI improved the fit by 36% (Supplement Table 8).

Visceral adipose tissue and PGD risk

Twenty-three subjects (20%) with measured abdominal VAT and 63 subjects (19%) with measured thoracic VAT developed PGD. We did not detect associations between thoracic VAT index, thoracic VAT area, abdominal VAT index, or abdominal VAT area and PGD risk in any models (Supplement Tables 9–11).

Latent class analysis

Latent class analyses identified 3 distinct groups (Supplement Table 12). Class 1 (n=22), referred to going forward as the high SAT group, was characterized by high thoracic and abdominal SAT; class 2 (N=46) had intermediate SAT, and class 3 (N=30) had low abdominal and thoracic SAT (Supplement Table 13). Although lacking precision in the estimates in this small under-powered post-hoc subgroup analysis, those in low and intermediate SAT groups had lower adjusted risk of PGD compared to those with high SAT (Supplement Table 14).

Post-transplant Survival

We did not observe an association between adipose measures and survival time (Supplement Tables 6–7, 9–10).

Biomarkers

Subjects with available biomarkers were more likely to be transplanted from a male donor or a white donor, and less likely to require intra-operative cardiopulmonary support but were otherwise similar to those without available biomarkers (Supplement Table 3). Both abdominal and thoracic SAT indices were positively associated with pre-transplant leptin (abdominal SAT index r=.61 p<0.0001, thoracic SAT index r=.47, p<0.0001, Figure 4A–B), post-transplant leptin (abdominal SAT index r=.44, p<0.44, and thoracic SAT index r=.41, p<0.0001, Supplement Figure 4), and pre-transplant IL1-RA (abdominal SAT index r=.25, p=0.04, thoracic SAT index r=.28, p-=0.005, Figure 4C, Supplement Figure 5). Abdominal SAT index was significantly associated with post-transplant ICAM-1 levels (r=.25, p=0.04, Supplement Figure 6). There was no significant association between changes in leptin levels from pre- to post-transplant (Supplement Figure 4). There was no significant association between abdominal or thoracic SAT index and pre-transplant levels, post-transplant levels, or change in levels from pre- to post-transplant for IL-6 (Supplement Figure 7), PTX3 (Supplement Figure 8), or PAI-1 (Supplement Figure 9).

Figure 4:

Scatter plots of (A) abdominal subcutaneous adipose index and pre-transplant leptin levels, (B) thoracic subcutaneous adipose index and pre-transplant leptin levels, and (C) abdominal subcutaneous adipose index and pre-transplant IL1-RA levels. Pearson correlation coefficients are (A) r= 0.61, p<0.0001, (B) r=0.47, p<0.0001, and (C) r=0.25, p=0.04.

Discussion

We found that a greater quantity of abdominal SAT was associated with an increased risk of PGD after lung transplantation. Neither thoracic nor abdominal VAT measures seem to be associated with PGD risk despite these CT measures having strong construct validity as measures of adiposity. SAT adipose measures were also associated with higher plasma levels of leptin, IL-1RA, and ICAM-1. Based on post-hoc latent class analyses, those with both high thoracic and abdominal subcutaneous adiposity seemed to be at increased risk of PGD compared to those with intermediate or low subcutaneous adiposity. Overall, our findings support the hypothesis that subcutaneous, but not visceral, adipose tissue may play a role in the development of PGD after lung transplantation.

While VAT may be considered a more inflammatory tissue, SAT has an inflammatory phenotype with a similar number of adipose tissue macrophages (16) and increased T cell chemokines in obese compared to lean adults (17). Increased abdominal SAT is associated with increased cardio-metabolic risk factors (18, 19), incident heart failure in the elderly (20), and acute kidney injury after trauma (21). In addition to its inflammatory effects, obese adipose tissue is characterized by impaired clearance of reactive oxygen species and increased circulating free fatty acids (56, 57), both of which are associated with PGD risk (58). Furthermore, total SAT area is significantly greater than VAT area which may additionally account for the differential PGD effects between the two depots.

Adipose tissue is a major source of IL1-RA (15) consistent with our finding of increased IL1-RA with greater SAT. Exogenous IL-1RA attenuates ischemia-reperfusion injury in a rat model of myocardial ischemia (59) and decreases acute inflammation in a bleomycin mouse model of idiopathic pulmonary fibrosis (60). Greater IL-1RA does not seem to have had protective effects in patients with greater SAT, which may be related to corticosteroid administration (61,62), or maximal IL-1RA production at baseline. Alternately, greater IL-1RA may be a plasma marker of higher IL1-β production as has been suggested in critical illness (63). Further investigation of the role of adipose-mediated changes in IL1-RA production and IL1-β kinetics may identify a modifiable target for PGD prevention.

ICAM-1 is a pulmonary vascular endothelial marker associated with PGD risk (39). Obesity alters endothelial cell function via increased endoplasmic reticulum stress and decreased expression of endothelial junctional adherence proteins, resulting in increased circulating levels of vascular endothelial markers including ICAM-1, and increased vascular injury in response to lipopolysaccharide in mouse models (69, 70). The direct association between SAT and post-transplant ICAM-1 suggests that SAT may increase PGD risk through altered pulmonary endothelial cell function.

We identified a correlation between greater SAT and increased plasma leptin before and after transplantation. Leptin is produced by adipose tissue, may increase ATM recruitment (12), alters neutrophil chemotaxis (64,65), is increased in the BAL of patients with acute lung injury (66), and may increase macrophage-mediated cytokine production (67). Leptin resistance is protective from hyperoxia-induced lung injury and decreases both inflammation and fibrosis in a bleomycin-induced mouse model of acute lung injury (66, 68) suggesting that SAT-derived leptin may not only play a role in the pathogenesis of PGD but also link PGD to chronic lung allograft dysfunction.

Greater abdominal SAT may lead to misclassification of PGD through mechanical effects on the lungs including atelectasis and decreased chest wall compliance. However severe PGD is characterized by both radiologic changes and severe hypoxemia. Severe hypoxemia is unlikely to be caused by mechanical effects alone. Extensive work has validated the current diagnostic criteria for PGD (35, 46) suggesting that misclassification is unlikely to account for this association. Furthermore, prior latent class analysis evaluating sub-phenotypes of PGD identified classes that did not significantly differ by BMI as would be expected if chest wall compliance were leading to misclassification (47).

There are several limitations to the current study. First, our patient population is predominantly normal or overweight with few obese patients, which may limit generalizability of our findings particularly with greater degrees of adiposity present in the general population. However, despite our limited BMI range, we identified an increased risk associated with increased abdominal SAT suggesting that subcutaneous adipose area identifies additional PGD risk even among patients who appear low risk by BMI. Additionally, the population selected represents the overall transplant population and our approach, including the delay between imaging and transplantation, is pragmatic in the setting of unpredictable transplant timing thus making it particularly relevant to this question. Whether there are changes in adiposity prior to lung transplantation is unknown. Second, we have a relatively small cohort with few events, limiting the ability to evaluate effect modification by classic risk factors including cardiopulmonary bypass, diagnosis, and donor smoking. Notably, our main effect estimates remain significant even when including a large number of covariates in the model suggesting a durable effect despite small sample size. Third, we were only able to evaluate circulating biomarkers as a measure of systemic inflammation. This may fail to identify local effects of adipose-mediated inflammation. Fourth, measurement of single slice cross-sectional area of adipose may not entirely reflect total adiposity. Prior work has demonstrated a significant correlation between single slice cross-sectional area and total volume of adipose (37, 38); additionally, our measures of adipose were highly correlated with traditional measures of body composition suggesting that we are measuring the tissue of interest. Fifth, the divergent findings in abdominal and thoracic subcutaneous adipose could also reflect either (1) different populations rather than differential effects of the two distinct anatomic adipose depots or (2) differences in adipose measurement related to the use of clinically-indicated rather than protocolized research thoracic CT scans. Reassuringly, on clinically observable characteristics, the two groups appear similar. Sixth, both abdominal and thoracic imaging were only available on a subgroup of 98 subjects thus limiting interpretation of our latent class models. Seventh, while our analyses reflect the results of our a priori hypotheses, there remains a small risk of a false positive finding given the multiple comparisons made. Finally, our study was not designed to evaluate the role of abdominal SAT measurement in pre-operative risk stratification, while our data suggest that abdominal SAT may better model PGD risk compared to BMI, additional work is required before this could be integrated into routine clinical assessment.

The reason that SAT, but not VAT, appears to contribute to PGD risk remains unclear but may be related to (1) the larger volume of total SAT, (2) differences in systemic effects of SAT including increased reactive oxygen species and free fatty acids, (3) anatomic differences in drainage or changes in compliance, or (4) we have failed to detect an association between VAT and PGD where one truly exists.

In summary, we have shown that greater abdominal subcutaneous adipose tissue area was associated with increased risk of PGD after lung transplantation independent of known PGD risk factors in a multi-center prospective cohort study. We have also highlighted the potential utility of measuring body composition through advanced imaging techniques, rather than non-specific measurements such as BMI, in order to better understand mechanisms linking body composition and outcomes in lung transplantation. Further research is required to establish the mechanisms by which adipose tissue affects the pulmonary parenchyma including the potential role of changes in free fatty acid levels, reactive oxygen species, local changes in inflammation, and pulmonary endothelial cell injury. Additionally, future studies should focus on the role of abdominal SAT in risk stratification prior to lung transplantation.