A Genetic Approach to the Association Between PCSK9 and Sepsis

Key Points Question In patients admitted to the hospital with serious infection, is there a significant association between PCSK9 genetic variation and risk of sepsis? Findings In this cohort study of 10 922 patients, the risk of sepsis was not significantly associated with PCSK9 functional variants, PCSK9 genetic risk score, or genetic estimation of PCSK9 expression levels. Meaning These findings suggest that PCSK9 genetic variations are not significantly associated with the risk of sepsis in patients hospitalized with infection.

This supplementary material has been provided by the authors to give readers additional information about their work.

eMethods. Cohort Identification and Power Calculations Cohort identification
Sepsis was defined as concurrent infection and organ dysfunction occurring within one day of hospital admission (days -1, 0, and +1) using an algorithm to detect sepsis using EHR data 1 with minor modifications. The modifications were as follows: in the definition of sepsis we included (1) septic shock defined by presence of ICD9 codes 995.92 and 785.52, or ICD10 codes R65.20 and R65.21 because these codes are very specific; 1 (2) vasopressor initiation identified by use of levophed (noprepinephrine bitartrate), or use of dobutamine or dopamine and a billing code for administration of a vasopressor (ICD9-CM procedure code 00.17 or ICD10-PCS procedure codes 3E033XZ, 3E043XZ, 3E053XZ and 3E063XZ) because dobutamine or dopamine alone had low specificity for identifying patients in whom it was used as a pressor; (3) we did not use serum lactate levels as a criterion because they were seldom available; (4) we excluded individuals who had scheduled cardiothoracic surgery because the algorithms did not reliably identify the reason for artificial ventilation or ICU admission and some of these patients had an infection and received an antibiotic. The methods used have been described in detail previously. 2

Associations between LDL-C and (1)4 PCSK9 functional variants (2) PCSK9 GRS and (3) predicted PCSK9 expression
For LDL-C analyses (other than the gene expression analysis), we did not restrict our analysis of the relationships between the genetic instruments and LDL-C to the sepsis cohort (n=10922) because some patients may only have had LDL-C measured after hospital admission (or during sickness) which would confound the association test. Instead, we used all available data in BioVU for individuals with LDL-C and PCSK9 genotypes. We have used all BioVU individuals who had both PCSK9 genotypes and an LDL-C measurement. We used the median LDL-C for those with multiple measurements.
The number of individuals in each analysis varies due to data availability in BioVU. Specifically, the 4 functional PCSK9 variants were extracted from genome-wide platforms and the ExomeChip (N=22,995). The 6 SNPs for PCSK9 GRS were not available on the ExomeChip and could not be imputed from it. Therefore, the association between the PCSK9 GRS and measured LDL-C was evaluated within those on genome-wide platforms (n=15,387). For the association between estimated PCSK9 expression and LDL-C, we used the sepsis cohort because gene expression had been calculated in this cohorts. Within the sepsis cohort, 3630 individuals had both predicted PCSK9 expression and measured LDL-C.

Power calculations
There was adequate power to detect small differences between the study groups; the detectable odds ratio for sepsis in the loss-of-function (LOF) carriers relative to non-carriers was 1.15. We estimated this detectable difference as a post-hoc calculation using PC software 3  Patients who were gain-of-function carriers were excluded from this power calculation.

Figure. statistical power in current study
We adopted the PCSK9 GRS based on LDL-C levels from previous high impact publications. 4,5 Specifically, the four functional PCSK9 variants were used in Walley's paper in Science Translation Medicine. 4 Although there was no quantification of PCSK9's effect for removal of LPS in vivo, the same genetic instruments demonstrated the relationship between PCSK9 variants and mortality in a cohort of ~500 individuals with septic shock. Furthermore, the GRS used in our manuscript was also significantly associated with both myocardial infarction and type II diabetes mellitus. 5