Title and Abstract

Determining the Causal Relationship Between COVID-19 and Acute Kidney Injury Using Mendelian Randomization

Abstract

Acute kidney injury (AKI) has been observed in over 20% of COVID-19 hospitalizations, occurring at twice the rate among patients admitted to intensive care units. Approximately 10% of these cases require dialysis, and AKI is linked to high mortality rates.1 While these clinical associations have been documented, the causal relationship between COVID-19 and AKI remains unclear due to potential confounding factors and mediators. To address this, we conducted a Mendelian randomization (MR) analysis to investigate the causality of relevant quantitative traits.

MR is a statistical method used to infer causal relationships between risk factors (exposures) and outcomes using genetic variants as instrumental variables (IVs). These genetic variants are assumed to be randomly assigned at conception, mimicking a randomized controlled trial and mitigating confounding and reverse causation. This analysis utilized comparisons of genome-wide association studies (GWAS) summary statistics from multiple COVID-19 studies across different hospitalization levels, integrated with GWAS summary statistics for AKI. By applying robust MR methods, such as inverse-variance weighted (IVW) analysis, MR-Egger regression, and weighted median estimation, we aimed to address potential pleiotropy and ensure reliable causal inference, shedding light on the genetic interplay between COVID-19 severity and AKI risk.

Citations:

  1. https://link.springer.com/article/10.1007/s10157-021-02092-x

Introduction

Background

In our analysis, the exposures are COVID-19 at three different levels of severity: Severe Respiratory COVID, Hospitalized COVID, and general population COVID. A potential casual relationship between the exposure and AKI is highly plausible given that incidence of AKI in COVID-19 cases has been highly documented in hospital settings. MR is a helpful tool in confirming a casual relationship as there may be intermediate mediators or confounders also associated with AKI.

Objectives

We aim to discover the true nature of the relationship between COVID-19 and AKI through a rigorous evaluation of MR results. MR is a method that, under specific assumptions, can provide such an estimate. We predict that COVID-19 onset causes AKI, based on previous observational data in a hospital setting.

Methods

Study Design and Data Sources

Data Source Setting Participants Selection of Variants Diagnostic Criteria Ethics
COVID-19 host genetics initiative Conducted meta-analysis using European studies, largely collected before the widespread distribution of the COVID-19 vaccination. Collected data for critical illness, hospitalization, and SARS-CoV-2 infection Very Severe Respiratory- Cases: 13769, Controls: 1072442; Hospitalized- Cases: 32519, Controls: 2062805; Infection- Cases: 122616, Controls: 2475240 “Meta-analysis was done with fixed effects inverse variance weighting. Results are available in genome builds 38 and 37. An AF filter of 0.001 and an INFO filter of 0.6 was applied to each study before meta.” (1)
FastGWA 2021 (Jiang L et al.) Used Genome-wide genotyping array to report the trait Acute Renal Failure (PheCode 585.1). Recruited from U.K (European). Cases: 1199, Controls: 455,149 Data obtained from UK Biobank. NR imputed quality control revealed 11842647 SNPs passing quality control.
SAIGE 2018 (Zhou W et al.) Used Genome-wide genotyping array to report the trait Acute Renal Failure (PheCode 585.1). Recruited from U.K (European) Cases: 4521, Controls: 397602 “Uses the saddlepoint approximation (SPA) to calibrate unbalanced case-control ratios in score tests based on logistic mixed models” (2) Data obtained from UK Biobank. Quality checked by Affymetrix, 28000000 SNPs passed QC (imputed)
Severe Respiratory Confirmed COVID vs Population

Hospitalized COVID vs Population

COVID-19 vs Population

FastGWA AKI Dataset (Jiang L et al.)

SAIGE AKI Dataset (Zhou W et al.)

Citations:

  1. https://www.covid19hg.org/results/r7/
  2. (Zhou et al.)

Assumptions

Three core assumptions for IVs:2

  1. Relevance: the genetic variants used as IVs are associated with the exposure

  2. Independence: there are no confounders between the IVs and outcome

  3. No horizontal pleiotropy: the only pathway from the IVs to outcome is through the exposure

Statistical Methods

We performed two sample Mendelian Randomization and ran the following tests: MR-Egger, MR-Egger Regression, MR median, MRweighted median, MR penalized weighted median, MR IVW, MR mode, and MR weighted mode.

Assessment of Assumptions

  1. Relevance: We only uses SNPs that are associated with a P values greater than 5E-8 in order to filter out IVs with a strong association with the exposure.

  2. Independence:

  3. No horizontal pleiotropy:

Results

Descriptive Data

Bibliography

“COVID-19 Host Genetics Initiative.” Covid19hg.org, covid19hg.org, 2022, www.covid19hg.org/results/r7/. Accessed 16 Mar. 2025.

Haycock, Philip C, et al. “Best (but Oft-Forgotten) Practices: The Design, Analysis, and Interpretation of Mendelian Randomization Studies.” The American Journal of Clinical Nutrition, vol. 103, no. 4, 9 Mar. 2016, pp. 965–978, https://doi.org/10.3945/ajcn.115.118216.

Kanai, Masahiro, et al. “A Second Update on Mapping the Human Genetic Architecture of COVID-19.” Nature, vol. 621, no. 7977, 1 Sept. 2023, pp. E7–E26, www.nature.com/articles/s41586-023-06355-3, https://doi.org/10.1038/s41586-023-06355-3. Accessed 7 Sept. 2023.

Zhou, Wei, et al. “Efficiently Controlling for Case-Control Imbalance and Sample Relatedness in Large-Scale Genetic Association Studies.” Nature Genetics, vol. 50, no. 9, 13 Aug. 2018, pp. 1335–1341, https://doi.org/10.1038/s41588-018-0184-y.

  1. https://link.springer.com/article/10.1007/s10157-021-02092-x↩︎

  2. (Haycock et al.)↩︎