Lenvatinib is an inhibitor of multiple tyrosine kinases including VEGFRs and FGFRs. It was approved by the U.S. Food and Drug Administration (FDA) for the treatment of solid tumors including thyroid cancer (2015), renal cell carcinoma (2016) and more recently in first line for advanced HCC (2018) [4] and in combination with pembrolizumab for endometrial carcinoma (2019). After over 6 years of clinical experience, lenvatinib is described as a well-tolerated drug that causes AEs characteristic of angiogenesis inhibitors, including hypertension (68%), diarrhea (59%), palmar-plantar erythrodysesthesia syndrome (32%), and proteinuria (31%).

Here, we report what is to our knowledge the first series demonstrating the occurrence of early and very frequent erythrocytosis in HCC patients under lenvatinib, with a potential risk of thromboembolic complications. The relationship between erythrocytosis and lenvatinib and its regression at treatment discontinuation, documented by Naranjo adverse drug reaction probability scale, confirmed the direct implication of the drug in these AEs.

Several case reports of erythrocytosis occurring during antiangiogenic treatments in cancer (cediranib, vandetanib, axitinib, pazopanib, sunitinib, sorafenib or bevacizumab) have been published [14,15,16,17,18,19,20,21,22]. However, the high frequency of this AE (87%) is striking in the present series, despite use of the classical dosage. Interestingly, in comparison with sorafenib, sunitinib or regorafenib, lenvatinib inhibition of VEGFR on an HCC cell line (HepG2), assessed by the IC50 value, was reported to be 4–128 times more efficient [23]. This high VEGFR inhibition could explain the higher frequency of Hb level increase under lenvatinib. It cannot be excluded, however, that erythrocytosis could be a class effect of all anti-VEGF therapies (Table 3), but its magnitude and its frequency might depend on the IC50 value of the different molecules.

Interestingly, this AE of lenvatinib has not been reported in other cancers. For example, in radioactive iodine (RAI)-refractory thyroid cancer which was the first indication of lenvatinib, phase II and III clinical trials did not report any Hb increase [24,25,26]. Previous RAI treatment, known to induce bone marrow impairment [27,28,29], could be an explanation, although our results clearly demonstrate the role of HCC tumoral cells in yielding an EPO increase.

Indeed, a liver specificity should be considered. In adults, EPO is produced not only by renal peritubular cells, but also by the liver [30]. In fact, HCC could be associated with a paraneoplastic syndrome characterized by secondary erythrocytosis and high plasma EPO levels produced by the cancer cells [31,32,33]. Here, immunohistochemistry confirmed that HCC tumoral cells were able to produce higher levels of EPO, under lenvatinib, than non-tumoral hepatocytes in the cirrhotic liver (Fig. 2). Moreover, there is now increasing evidence showing the impact of hypoxia on hepatic EPO production. An increase of EPO secretion by hepatocytes and stellate cells has been reported in rats in hypoxia conditions [34,35,36,37]. It has also been shown that VEGF inhibition results in a large increase in EPO secretion from hepatic cells, leading to enhanced erythropoiesis and elevated circulating red blood cell counts [38]. Therefore, lenvatinib, one of the most potent VEGFR inhibitors, could potentiate EPO production by HCC cells subject to hypoxic-like conditions and explain the Hb/Ht increase. The doubling of EPO level under lenvatinib in one patient, leading to a high level of Hb, supports this hypothesis. EPO level monitoring might represent a simple surrogate marker for stringent blockade of VEGFR in HCC patients treated by lenvatinib.

Several reports have suggested that EPO and EPO stimulating agents could promote tumor cell proliferation through its specific receptor (EPOR) and hypoxia in head and neck tumors [39], breast cancer [40] and HCC [41]. This was shown to be associated with poorer overall survival rates in HCC patients [42]. However, other studies contradicted these findings [43, 44] and even suggested that the level of EPOR in the cirrhotic tissue could be correlated with tumor cell differentiation and a favorable outcome [45]. In the present study, erythrocytosis was not correlated with the tumor response nor with the occurrence of other AEs.

Erythrocytosis is associated with a risk of thrombosis in relation with an increase of blood viscosity and blood flow decrease [46,47,48]. Antiangiogenic drug therapy is also associated with an increased risk of thrombosis [49]. In thyroid cancer patients treated with lenvatinib, arterial thromboembolic events occurred in 5.4% of the cases (2.7% grade ≥ 3) and venous thromboembolic events in 5.4% (3.8% grade ≥ 3) [26]. Furthermore, HCC occurs in 80–90% of the cases in cirrhotic patients, who, due to complex coagulation disorders, present with a pro-thrombotic condition. For all these reasons, a close monitoring of hematologic parameters is recommended before and during lenvatinib treatment. In case of erythrocytosis, we propose to initiate thromboprophylaxis (i.e., low-dose ASA). Control of cardiovascular risk factors such as hypertension, hyperlipidemia, diabetes and smoking cessation should be emphasized if applicable.

In conclusion, in this cohort of HCC patients, a frequent and specific erythrocytosis was evidenced, possibly secondary to EPO secretion by tumor cells related to the antiangiogenic activity of lenvatinib. These results suggest that a close and early monitoring of hematologic parameters should be performed for such patients. Thromboprophylaxis by ASA should be prescribed in case of erythrocytosis and phlebotomy in case of symptomatic effects of erythrosis.