Granulocyte colony-stimulating factor (G-CSF) has been shown to have regenerative and immunomodulatory properties and is therefore an attractive therapeutic approach for patients with advanced liver disease [1]. Randomized clinical trials have shown that G-CSF has been successful in reducing mortality and increasing survival of patients with advanced liver disease such as acute-on-chronic liver failure and cirrhosis. However, the view was challenged by the results of a GRAFT study that failed to confirm the positive effects previously observed [2]. Interestingly, Jindal et al. [3] hypothesized inadequate patient selection was the reason for the lack of efficacy in this study. In addition, there is some doubt from animal studies that G-CSF alone increases mortality and promotes inflammation in acute-on-chronic liver failure (ACLF) models by sensitizing the liver to endotoxin [4,5]. In line with this, Shen et al. [6] concluded that G-CSF exacerbated liver injury (increasing alanine transaminase [ALT] and aspartate aminotransferase [AST] levels) in autoimmune hepatitis models. From a broad perspective, Shen hypothesized that G-CSF may not be suitable for all types of liver disease.

This study used publicly available de-identified data from previous studies approved by an ethical standards committee. Therefore, no further ethical approval or informed consent was required in this study.

In the Mendelian randomization (MR) paradigm, random allocation of genetic variants associated with an exposure (agonist/stimulator of a drug target such as G-CSF therapy) can be used to assess causal relationships with a phenotype (risk of cirrhosis, liver failure or hepatocellular carcinoma). Thus, mendelian randomization can provide quasirandomized evidence on the effect of G-CSF therapy in advanced liver disease.

Cirrhosis is the end-stage of viral hepatitis, alcohol-associated hepatitis, and autoimmune hepatitis, resulting in liver failure or hepatocellular carcinoma (HCC), which remains an important cause of morbidity and mortality in people with chronic liver disease worldwide [7]. Multi bio-bank summary GWAS data for decompensated liver disease by manual retrieval was obtained making it feasible to verify the effect of G-CSF signaling on them using MR. From this perspective, we use all available decompensated liver disease GWAS data to explore whether G-CSF signaling, proxied by neutrophils, using variants from the CSF3R gene region, can be expected to reduce risks of decompensated liver disease.

We extracted summary GWAS data from multi-biobank containing cirrhosis, liver failure and HCC patients. Furthermore, we included four liver enzymes GWAS data from all publicly available studies (including aspartate aminotransferase levels, alanine transaminase levels, gamma-glutamyl transferase levels, and alkaline phosphatase levels). We selected neutrophil cell count for the main analysis because CSF3R stimulator/agonist results in the mobilization of neutrophils from the BM into the blood, which physiologically is expected to have a direct effect on neutrophil cell count.

Single nucleotide polymorphisms (SNPs) were identified within the CSF3R gene region (GRCh37/hg19 Chr1: 36931644-36948879) and were robustly associated with neutrophil cell count at genome-wide significance (P<5×10−8) in a GWAS meta-analysis from the Blood Cell Consortium. They were further clumped to a LD threshold of r² <0.35 with a physical distance threshold of 10,000 kb and were selected as proxies for G-CSF drug targets.

After clumping, we included 4 SNPs as instruments. These SNPs had a mean F statistic of 171, the variants are therefore sufficiently strongly associated with neutrophils to give meaningful MR estimates. False discovery rate Inverse Quantile Transformation (FIQT) Winner’s Curse correction was used to adjust for Winner’s Curse bias. To generate mendelian randomization estimates, we estimated the Wald ratio for each genetic variant by dividing the variant-outcome association by the variant-neutrophils associ-ation. Standard errors for mendelian randomization were estimated as the standard error of the variant-outcome association divided by the variant-phenotype association. Pooled IVW estimates for each SNP were then meta-analyzed with a multiplicative random or fixed effects model (Fig. 1).

After combining GWAS summary data from different studies, we found no evidence of an association between genetically proxied CSF3R stimulator/agonist (per 10⁹/L increase in neutrophil count) and two outcomes considered: cirrhosis (odd ratio 1.34, 95% CI 0.57 to 3.15), liver failure (OR 2.54, 95% CI 0.73 to 8.83) (Fig. 2) while was negatively associated with hepatocellular carcinoma (OR 0.12, 95% CI 0.03 to 0.54) (Supplementary Fig. 1). Furthermore, it is important to note that the available evidence suggests a weak positive correlation between G-CSF therapy and ALT levels (MD 0.01, 95% CI 0.00 to 0.02). However, no significant associations have been observed between G-CSF therapy and otherliver enzymes, including AST levels (MD 0.04, 95% CI –0.02 to 0.11), gamma-glutamyl transferase levels (GGT, MD 0.02, 95% CI –0.03 to 0.06), and alkaline phosphatase levels (ALP, MD 0.02, 95% CI –0.04 to 0.08) (Fig. 3).

In this study, we did not detect clear associations between genetically predicted neutrophil-weighted G-CSF signaling and risks of decompensated liver disease, using all available GWAS data and gene-region significant variants in the CSF3R gene. Similarly, no significant associations were observed between G-CSF signaling and elevated levels of liver enzymes, including ALT, AST, ALP, and GGT.

To some extent, the question of whether G-CSF therapy has a positive effect on cirrhosis, liver failure, or liver enzymes can be addressed via mendelian randomization. In this study, neutrophil cell count was used as a biomarker to weight the effects of genetic variants, thus reflecting the effect of G-CSF therapy. However, it is important to note that neutrophil count does not necessarily represent the mechanism by which G-CSF therapy exerts its effects. This suggests that neutrophil function may not be the key factor underlying the potential beneficial effects of G-CSF therapy in patients with decompensated liver disease, such as cirrhosis and liver failure.That said, it is worth emphasizing that the four liver enzymes examined in this study (ALT, AST, ALP, and GGT) are insufficient to fully capture the severity of liver disease, particularly in advanced conditions such as ACLF and end-stage liver disease [8]. More robust indicators of liver function deterioration, such as albumin, prothrombin time, and bilirubin, are better suited for assessing disease progression [9]. Unfortunately, we were unable to identify eligible genetic instruments for these markers to conduct further analysis. The absence of these critical se-verity indicators represents a limitation of our study and somewhat restricts the evidence regarding the impact of G-CSF on liver function.

While our analysis suggested a potential protective effect of neutrophil-weighted signaling on HCC, this finding should be interpreted with caution. Emerging evidence indicates that G-CSF may also play a role in promoting HCC progression, potentially through its regulatory effects on tumor-associated macrophages rather than neutrophils [10]. Therefore, the relationship between G-CSF and HCC appears complex and context-dependent, warranting further investigation to clarify its dual roles in tumor biology.