Among the exposures of WBC counts, the genetically predicted counts of most WBCs may not be associated with the risk of HCC except the sum basophil neutrophil counts, based on both datasets. Lower total basophil neutrophil counts may be an independent risk factor for HCC. Univariable MR analysis based on one dataset suggested that eosinophil counts had a significant association with HCC, but the association became nonsignificant after adjustments for other traits.

Univariable MR analysis of exposures on HCC risks in the discovery stage

To characterize the relationship between WBC counts and the risk of HCC, we constructed a genetic instrument for WBC counts using 73–173 independent SNPs associated with the above 8 traits at a genome-wide level of significance (P < 5 × 10−8), which accounted for 1.24–33.86% of the variability in the WBC counts. The mean F-statistic ranged from 68.63 to 115.78, suggesting that the risk of weak instrument bias was low. Univariable MR analysis identified a lower sum of basophil neutrophil counts as a risk factor for HCC. Briefly, each 1-SD increase in the sum basophil neutrophil count could help reduce the risk of HCC (OR = 0.400, P = 0.0001, CI 0.233–0.684). We found no strong evidence to support causal associations between HCC risk and the other WBC subtypes including eosinophil counts (OR = 0.560, P = 0.059, CI 0.307–1.022), basophil counts (OR = 1.039, P = 0.939, CI 0.384–2.811), neutrophil counts (OR = 1.144, P = 0.832, CI 0.328–3.989), lymphocyte counts (OR = 0.636, P = 0.198, CI 0.320–1.266), monocyte counts (OR = 1.213, P = 0.472, CI 0.717–2.051), sum eosinophil basophil counts (OR = 0.691, P = 0.236, CI 0.375–1.273), and sum neutrophil eosinophil counts (OR = 0.806, P = 0.544, CI 0.403–1.614) (Table 2).

Table 2 The effect estimates, test of heterogeneity and test of pleiotropy of white cell counts on HCC (discovery)Multivariable MR analysis of exposures on HCC risks in the discovery stage

Then, we explored the causal relationship between WBC counts and HCC by conducting a multivariable MR analysis. We observed that sum basophil neutrophil counts had an independent causal effect on the occurrence of HCC: eosinophil counts (OR = 0.502, P = 0.159, CI 0.192–1.309), basophil counts (OR = 1.088, P = 0.873, CI 0.386–3.065), neutrophil counts (OR = 0.605, P = 0.459, CI 0.160–2.288), lymphocyte counts (OR = 0.666, P = 0.266, CI 0.326–1.363), monocyte counts (OR = 1.201, P = 0.511, CI 0.696–2.072), sum eosinophil basophil counts (OR = 1.145, P = 0.786, CI 0.431–3.044), and sum neutrophil eosinophil counts (OR = 1.245, P = 0.746, CI 0.330–4.705). The OR of HCC decreased per 1-SD increase in the sum basophil neutrophil counts (OR = 0.437, P = 0.003, CI 0.252–0.757) (Fig. 2).

Fig. 2

The forest plot of the multivariable Mendelian randomization results based on the discovery dataset. Odds ratio (OR) is per 1-SD increase; 95% LCI is the lower limit of the 95% confidence interval; 95% UCI is the upper limit of the 95% confidence interval

Univariable MR analysis of exposures on HCC risks in validation stage

In the validation stage, we successfully replicated the MR results of WBC counts. We validated these risk factors for HCC by MR analysis. Univariable MR results demonstrated that sum basophil neutrophil counts had a significant causal effect on HCC. The OR of HCC decreased per 1-SD increase in the sum basophil neutrophil counts (OR = 0.573, P = 0.018, CI 0.361–0.909). Eosinophil counts had a weak protective effect on HCC: eosinophil counts (OR = 0.568, P = 0.033, CI 0.338–0.954). Other WBC subtypes have no causal relationship with HCC risk: basophil counts (OR = 0.509, P = 0.123, CI 0.215–1.202), neutrophil counts (OR = 0.817, P = 0.509, CI 0.449–1.488), lymphocyte counts (OR = 0.944, P = 0.851, CI 0.521–1.711), monocyte counts (OR = 0.671, P = 0.085, CI 0.426–1.056), sum eosinophil basophil counts (OR = 0.596, P = 0.054, CI 0.352–1.009), and sum neutrophil eosinophil counts (OR = 0.693, P = 0.230, CI 0.381–1.262) (Table 3).

Table 3 The effect estimates, test of heterogeneity and test of pleiotropy of white cell counts on HCC (validation)Multivariable MR analysis of exposures on HCC risks in validation stage

Multivariable MR analysis in validation stage also revealed that the sum of the basophil neutrophil counts was an independent HCC risk factor. It is worth noting that after adjustments for other traits, the association between eosinophil counts and HCC became nonsignificant: eosinophil counts (OR = 0.596, P = 0.214, CI 0.264–1.348), basophil counts (OR = 0.547, P = 0.180, CI 0.226–1.320), neutrophil counts (OR = 1.806, P = 0.308, CI 0.580–5.623), lymphocyte counts (OR = 0.982, P = 0.954, CI 0.535–1.805), monocyte counts (OR = 0.723, P = 0.175, CI 0.453–1.155), sum eosinophil basophil counts (OR = 0.996, P = 0.992, CI 0.434–2.285), sum neutrophil eosinophil counts (OR = 0.462, P = 0.183, CI 0.148–1.440), and sum basophil neutrophil counts (OR = 0.574, P = 0.021, CI 0.358–0.920) (Fig. 3).

Fig. 3

The forest plot of the multivariable Mendelian randomization results based on the validation dataset. Odds ratio (OR) is per 1-SD increase; 95% LCI is the lower limit of the 95% confidence interval; 95% UCI is the upper limit of the 95% confidence interval

Sensitivity analysis

We observed that the confidence interval of the exposures was relatively wide, which was considered to be caused by low sample size. It cannot be ruled out that there would be weak connections between the other WBC count traits and HCC. Another possibility of the null findings observed in our MR analyses could be explained by the low proportion of variances in some of the exposures (F statistics  < 100). On the other hand, we deemed that sum basophil neutrophil counts had a causal relationship with HCC.

There was heterogeneity of neutrophil counts (P < 0.05) in the discovery stage. All the results of these risk factors were the MR-PRESSO corrected results if outliers were detected. No significant horizontal pleiotropic effects were detected in the MR-Egger test (for the intercept of MR-Egger, all P values were more than 0.05). The statistical power of these exposures was 100%.