Genetically Predicted Neutrophil-to-Lymphocyte Ratio and Coronary Artery Disease: Evidence From Mendelian Randomization

nflammation contributes to atherosclerosis and coronary artery disease (CAD). To help identify therapeutic targets, it is important to ascertain whether biomark-ers associated with CAD risk are causal. In a recent meta-analysis of clinical trials, neutrophil-to-lymphocyte ratio (NLR) was associated with increased cardiovascular risk. 1 We investigate a potential causal nature of this relationship by performing Mendelian randomization (MR) analyses.

I nflammation contributes to atherosclerosis and coronary artery disease (CAD). To help identify therapeutic targets, it is important to ascertain whether biomarkers associated with CAD risk are causal. In a recent meta-analysis of clinical trials, neutrophil-to-lymphocyte ratio (NLR) was associated with increased cardiovascular risk. 1 We investigate a potential causal nature of this relationship by performing Mendelian randomization (MR) analyses.
All participants provided prior consent for each study included, and the studies were approved by the relevant review committees. Nonpublic estimates are available from the corresponding author upon reasonable request. To identify genetic variants associated with NLR, we obtained published genome-wide association study summary data for neutrophil count and lymphocyte count in 361 194 European ancestry participants from http:// www.nealelab.is/uk-biobank/, which were selected based on self-reported ancestry and genetic principal components. The propagation of error method was used to estimate the association of all available single-nucleotide polymorphisms (SNPs) with lymphocyte count subtracted from neutrophil count, resulting either in a positive value (which represents an increasing NLR) or a negative value (representing a decreasing NLR). Instruments were selected for MR by clumping all common SNPs (minor allele frequency, >0.01) at genome-wide significance (P<5×10 −8 ) to pairwise linkage disequilibrium threshold r 2 <0.01 (using European participants from the 1000Genomes project as reference). We estimated the association of each candidate instrument with NLR in individual participant data on 396 020 UK Biobank participants of White British descent with similar genetic ancestry (as reported by the UK Biobank resource), for whom neutrophil count, lymphocyte count, and data on all selected SNPs were available and had no extreme NLR value (defined as either NLR=0 or NLR=infinity). 2 To estimate the association of each SNP with NLR, we used log-linear regression with adjustment for age, sex, the first 10 principal components of genetic ancestry, and measurement batch. We chose log-linear regression because NLR can only be positive, with some values close to zero, rendering normal linear regression inappropriate. The Figure (A) shows the distribution of NLR, which is similar to the distributions from various clinical trials including CANTOS (Canakinumab Anti-Inflammatory Thrombosis Outcome Study). 1 Statistical power calculations for the minimum detectable odds ratio were performed for a power of 80% and type 1 error rate of 0.05. 3 We performed 2-sample MR analyses, where we considered the following outcomes: CAD, myocardial infarction, circulating CRP (C-reactive protein), IL-6 (interleukin-6), and fibrinogen levels (Figure). CRP, IL-6, and fibrinogen were considered to investigate the effect of NLR on inflammatory biomarkers. 4 Finally, to investigate the validity of the NLR instruments, we performed analyses with neutrophil count and lymphocyte count as outcomes. If data on a SNP were unavailable in one specific genome-wide association study, we searched Neutrophil-to-Lymphocyte Ratio and Heart Disease for an available proxy in high linkage disequilibrium (r 2 >0.9), and if unavailable, we omitted the SNPs for that specific analysis. All analyses were performed using the package TwoSampleMR in R v4.0.3. Results are presented per 1-unit increase in genetically predicted log-transformed NLR.
Power calculations showed that we had 80% power to detect a minimum odds ratio of 1.07 for CAD. For myocardial infarction, power calculations showed 80% power to detect an odds ratio of 1.08. After clumping and omitting multiallelic SNPs, a total of 182 uncorrelated SNPs were selected as potential instruments in MR analyses given their genome-wide significant association with lymphocyte count subtracted from neutrophil count. In primary analyses (Figure [B]), we observed strong evidence of an association between genetically predicted NLR and neutrophil count (2. In sensitivity analyses, we observed similar results using the MR Egger method. Excluding outliers, MR-Pleiotropy Residual Sum and Outlier showed similar results to the primary inverse-variance weighted results, with the exception of CRP (Figure [B]).
This MR study did not identify evidence to support that NLR is causally related to risk of CAD and myocardial infarction. Moreover, we did not find consistent evidence of a causal association of NLR on CRP, IL-6, or fibrinogen levels. This contrasts with previous reports that have suggested a potential causal role for NLR in CAD, given the finding that both neutrophils and lymphocytes are involved in atherogenesis. 1,5 The discrepancy may be attributable to the associations identified in epidemiological studies arising due to confounding and reverse causation. However, even if not causally related to CAD, NLR could still be used as a predictive measure for cardiovascular disease risk. To what extent NLR has added value in future cardiovascular risk prediction over and above CRP levels remains to be fully explored.
Most data are retrievable from the public domain. Summary data for the genetic instrument are available from the corresponding author. Neutrophil-to-Lymphocyte Ratio and Heart Disease