Associations between single nucleotide polymorphisms in the FAS pathway and acute kidney injury

Study population

FACTT was a 2?×?2 factorial randomized trial, conducted by the National Heart, Lung,
and Blood Institute (NHLBI) Acute Respiratory Distress Syndrome (ARDS) Clinical Trials
Network. FACTT included 1,000 patients with ALI for 48 h, randomized to receive conservative
vs. liberal fluid management in one arm, and pulmonary artery catheter (PAC) vs. central
venous catheter (CVC) in the other arm; eligibility and exclusion criteria have been
previously described 20], 21]. To be eligible for enrollment in FACTT, patients were required to be intubated and
on positive-pressure ventilation, have a ratio of the partial pressure of arterial
oxygen (PaO2) to the fraction of inspired oxygen (FiO2) 300, and have bilateral infiltrates
on chest radiography consistent with pulmonary edema without evidence of left atrial
hypertension for 48 h. An additional enrollment requirement was the intent by the
primary physician to insert a CVC. Written informed consent was obtained from participants
or legally authorized surrogates in the original FACTT study and available to use
for secondary analysis. Further details about the consent process can be found in
the FACTT study 21]. There was no difference in the incidence of AKI or in receipt of dialysis between
the FACTT treatment arms, although these were secondary outcomes of the original studies
21], 22].

Genotyping

Among the 1,000 FACTT patients, 310 Caucasians and 91 African-Americans provided consent
for genetic testing. The University of Washington Institutional Review Board approved
this ancillary study. We previously showed that the baseline characteristics of genotyped
vs. non-genotyped patients in FACTT are very similar 15]. A priori, we performed our analyses separately by race, to avoid confounding by
race. We genotyped 367 SNPs in 45 genes (Additional file 1: Table S1), chosen from an annotated pathway diagram of genes involved in the Fas/FasL
pathway 8], 23]. We used re-sequencing information available through the NHLBI Program in Genomic
Applications 24], the National Institute of Environmental Health Sciences Environmental Genome Project
25], and the International HapMap Project (Release #19) 26] to calculate linkage disequilibrium (LD) bins (excluding SNPs with minor allele frequency
(MAF) 0.05 and r²
??0.8) for each of the 45 genes chosen, using LD SELECT on the genome variation server
27]–30]. We then selected the final 367 TagSNPs from within the identified LD bins by prioritizing
those previously reported to be associated with a disease or quantitative trait and
those more likely to have functional significance (non-synonymous??synonymous??untranslated
region). The selected TagSNPs covered over 95 % of the common LD bins within the candidate
genes. The SNPs and their associated genes are listed in the Additional file 1: Table S2. We used a commercially available Illumina GoldenGate BeadXpress (San Diego,
USA) system to genotype the selected SNPs 28], 29].

Quality control

Resulting genotype information underwent quality control at the SNP and subject level.
We excluded six patients for whom 80 % of the SNPs were successfully genotyped (mean
individual call rate 98.5 %) and we excluded 23 SNPs for which 80 % of patients were
successfully genotyped (mean SNP call rate 93.7 %). After stratifying the genotyped
patients by self-identified race (Caucasian or African-American), we excluded an additional
30 and 51 SNPs in Caucasians and African-Americans, respectively, for which the MAF
was 0.05. We excluded an additional 30 SNPs in Caucasians and 13 in African-Americans,
respectively, for which p values among controls for Fisher’s exact Hardy Weinberg equilibrium were 1?×?10
^?331]. After these quality control measures, 284 (77.4 %) and 280 (76.3 %) SNPs remained
from 307 Caucasian subjects and 88 African-American subjects, respectively. A flow
diagram of patient groups is provided in Fig. 1.

Fig. 1. Flow chart of patient allocation during the study. SNP single nucleotide polymorphism, FasL Fas ligand, AKIN Acute Kidney Injury Network

Outcome

Our definition of AKI was adapted from the AKI Network (AKIN) definition 32], 33]. While the AKIN definition utilizes a sequential, incremental increase in serum creatinine
(SCr) within 48 h, we used peak compared to trough SCr within the first 2 days following
study enrollment. In doing so, we aimed to increase the sensitivity for ALI-associated
AKI, which may have occurred before enrollment in FACTT, as patients could have enrolled
up to 48 h after the onset of ALI. Stage 1 and higher AKI was defined as an increase
in SCr??=0.3 mg/dl or??=50 %. Hourly urine output was not available in the FACTT
database.

Primary analysis

In our primary analysis, we assessed the risk of AKI by genotypes using logistic regression,
assuming an additive genetic effects model, i.e., each additional copy of the minor
allele (0, 1 or 2) is associated with an incremental change in risk of AKI. We performed
the analysis separately for each race (Caucasian and African-American). A priori,
we decided to adjust for gender, age, and FACTT treatment arm. We did not adjust for
sepsis because we were concerned it may sit within the causal pathway between genotype
and AKI. We corrected for multiple comparisons by using a false discovery rate (FDR)
threshold 0.1, which estimates that less than 10 % of the associations with an FDR
value at or below this level are false positives (29). This is intermediate between
using a raw p value (least conservative) and Bonferroni correction (most conservative), and accounts
for the fact that many of the genotypes are in partial LD and, thus, many of the hypothesis
tests are correlated. We used Fisher’s exact test to evaluate genotype frequencies
for Hardy-Weinberg equilibrium. Stata 11.2 and Golden Helix SNP Variation Suite
7 were used for the statistical analyses. Assuming an MAF of 20 % on average, we estimated
a power of 80 % to detect an odds ratio (OR) of 1.65 or greater in the Caucasian patients
in our study.

Sensitivity analysis

In sensitivity analysis, we compared the genotypes in those without AKI to those with
severe AKI (modified AKIN stage 2 and 3). In this analysis, patients subcategorized
as having severe AKI had an increase in SCr 100 %, or had a rise in SCr to 4.0 mg/dl
and an increase of at least 0.5 mg/dl. Patients with stage-1 AKI were excluded from
the sensitivity analysis, thus excluding patients who may have been misclassified
as having AKI with a minimal increase in SCr 34]. Similar to the primary analysis, we assessed the risk of AKI by genotypes using
logistic regression with an additive model, and stratified by race. Clinical outcomes
including receipt of dialysis, mortality, ventilator-free days, and ICU-free days
are outlined by AKI stage in Additional file 1: Tables S3a and b.