Sorafenib use in hepatocellular carcinoma: More questions than answers


  • Ghassan K. Abou-Alfa M.D.

    Corresponding author
    1. Department of Internal Medicine, Memorial Sloan-Kettering Cancer Center, New York, NY
    2. Department of Internal Medicine, Weill Cornell Medical College, New York, NY
    • Address reprint requests to: Ghassan K Abou-Alfa, M.D., Memorial Sloan-Kettering Cancer Center, 300 East 66th St., New York, NY 10065. E-mail:; fax: 646-888-4255.

    Search for more papers by this author

  • Potential conflict of interest: Dr. Abou-Alfa received grants from Bayer.


overall survival


phosphorylated ERK


partial response


Response Evaluation Criteria in Solid Tumors


vascular endothelial growth factor.

Almost 2 years ago, an article with the title “Field-Practice Study of Sorafenib Therapy for Hepatocellular Carcinoma: A Prospective Multicenter Study in Italy” by Iavarone et al. was published in this journal.[1] To date, the conclusions of this article continue to stimulate discussion.

In their attempt to better identify the role of sorafenib in the setting of broad nonclinical trial use, Iavarone et al. conducted an observational study right after the phase III randomized study of sorafenib versus placebo (SHARP trial) reported an improved median overall survival (OS) outcome of 10.7 versus 7.9 months in favor of sorafenib.[2] The key outcomes of this observational study conducted by Iavarone et al. were a similar median survival of 10.5 months and a toxicity profile analagous to the SHARP trial.

The key issues under question are the correlation of improved outcome in association with a lower dose of sorafenib; the prediction of outcome based on side effects; and the use of the modified Response Evaluation Criteria in Solid Tumors (mRECIST).

The intent of this editorial is to help elucidate those points and draw some practical conclusions that apply to the use of sorafenib in the clinical setting.

Sorafenib Dosing and Outcome

The observation of Iavarone et al. regarding sorafenib dosing and outcome is intriguing—but it is also limited by its retrospective nature, as the authors fairly acknowledge. The duration of treatment between patients who underwent a dose reduction and patients who received full dose sorafenib resulted in prolonged treatment exposure (6.8 versus 3 months), even though not necessarily an increased cumulative dose. OS was 21.6 months (95% confidence interval [CI] 13.6-29.6) for patients who received a half-dose of sorafenib for 70% of the treatment period, compared to 9.6 months (95% CI 6.9-12.3) for the patients who maintained full dosing or had a dose reduced for <70% of the treatment period.

This finding may imply that targeted drugs may not follow the therapeutic paradigm of cytotoxic anticancer drugs, whose clinical activity correlates with dosing and achievement of clinically effective blood levels.

In the first interim analysis of the Global Investigation of therapeutic DEcisions in hepatocellular carcinoma and Of its treatment with sorafeNib (GIDEON) study, it was observed that dosing of sorafenib differed by specialty. Eighty-three percent of hepatologists and gastroenterologists started patients on 800 mg/day compared to 63% of medical oncologists,[5] yet no outcome data have been reported. Thus, so far no data support the use of reduced dose of sorafenib and the retrospective analysis of Iavarone et al. should be interpreted within that context. Indeed, in the study of Iavarone et al. the starting dose was 800 mg daily, which was only reduced as required by collateral effects or intolerance.

One may consider to resolve the paradox of improved results in the face of reduced doses by hypothesizing that collateral effect may be a pharmacodynamic indicator of an individual's susceptibility to the drug. Thus, the Raf-inhibitory effects of a reduced dose in a patient experiencing collateral effects could actually be higher than those of full dose in a patient who does not have any collateral effect. To add to the complexity of the issue, Raf inhibitors may have a paradoxical effect if the dose is lower than that able to efficiently inhibit both B-Ras and C-Ras, with activation of c-Raf.[3] A significant increase in cell proliferation and high expression of phosphorylated ERK (pERK) 1/2, phosphorylated mammalian target of rapamycin (mTOR), hypoxia inducible factor 1 alpha (HIF1α), vascular endothelial growth factor (VEGF), and VEGF receptor-2 was shown with lower doses of sorafenib in a rodent model of polycystic liver disease mice who show a severe liver phenotype, a high proliferation rate of the cystic epithelium.[4] These observations are limited, similar to the phase I pharmacodynamics data published on sorafenib by the assumption of one target of sorafenib,[5] being Raf in this instance, while obviously antiangiogenesis and other targets may be central to those observations. Of note, the maximum mean AUC0-12 values were obtained at 600 mg bid, although the difference between 400 and 600 mg bid was only marginal. Add to this, a complete inhibition of phorbol myristate acetate (PMA)-stimulated ERK phosphorylation was achieved at the 400 mg bid dose level. While similar results were obtained at 600 mg bid, the fact that a partial response in hepatocellular carcinoma (HCC) occurred at the 400 mg bid dosing sealed the deal. This question, however, will remain unanswered as the data on dose escalation in the setting of renal cell carcinoma continues to emerge.[6, 7]

In this light, Child-Pugh B patients pose another dilemma, when it comes to dosing of sorafenib. The first analysis from the GIDEON study did not provide any insight on Child-Pugh status pertaining to the starting dose of sorafenib8; however, several other studies have shown that the outcome of patients with Child-Pugh B cirrhosis fare worse than their Child-Pugh A counterparts.[1, 9] Iavarone et al. reported hepatic function to have deteriorated in 15% of Child-Pugh A patients and in 8% of Child-Pugh B patients after 12 weeks of sorafenib therapy and in 16% of Child-Pugh A patients and 40% of Child-Pugh B patients after 24 weeks of therapy. It is difficult to understand the significance of these time patterns. In CALGB 60301, a phase I study evaluating sorafenib in patients with differing degrees of hepatic or renal dysfunction, hyperbilirubinemia was identified as a dose-limiting toxicity for sorafenib and guidelines for dosing sorafenib in the setting of hepatic dysfunction were provided.[10] While this study addressed safety, it did not obviate the paradoxical proliferative effect sorafenib may have at those lower doses that may be contribute to the worse outcome these patients may have.

Prediction of Outcome Based on Side Effects

In follow-up to the dosing and outcome discussions herein, a correlation between worsening selective adverse events for sorafenib and improved outcome has been suggested. Hypertension as a pharmacodynamic marker for improved outcome with tyrosine kinase inhibitors has been studied extensively.[11-14] Maitland et al.[15] have shown that escalating the sorafenib dose from 400 twice daily, to 600 twice daily, and 400 three times daily did not result in any statistically significant increased in diastolic blood pressure. A correlation has not been ruled out; however, a more complex pharmacodynamic-based interaction over dose-to-hypertension is a potential explanation.[16]

A link between disruption of VEGF-related angiogenesis and hypertension was not observed in 20 patients who received sorafenib at a standard dose of 400 mg twice daily and who had serial evaluations of their blood pressure and heart rate in addition to measurement of VEGF, catecholamines, endothelin I, urotensin II, renin, and aldosterone.[16] Similarly, there was no correlation with any other clinical or humoral factors. Possible other mechanisms could explain these observations of improved outcome in the setting of increased hypertension; however, to understand these associations an evaluation of vascular rarefaction (a decrease in vascular density in peripheral skeletal muscle), endothelial dysfunction, and altered nitrous oxide metabolism, among many other putative mechanisms need further study.[16, 17]

Predicting the outcome of sorafenib based on rash and hand-foot skin reactions has also been suggested, with the observation that improvement in time to progression from 4 months in patients without skin toxicity to 8.1 months in those with skin toxicity.[18] A plausible mechanism for this observation, however, is lacking, but possible explanations may relate to hyperkeratosis and vascular effects.[19]

Other adverse events have also been implicated as independent prognostic factors for improved overall survival using sorafenib, including diarrhea. No mechanism of action has been reported yet. Clearly, these symptoms plus possibly others may be a pharmacodynamic reading of a sorafenib effect.


mRECIST was first reported in 2010 as a way to overcome the limitations of RECIST by defining only the contrast-enhanced portion of the lesion as a target lesion as opposed to the whole lesion.[20, 21] To date, we lack an independent prospective validation of mRECIST in HCC. mRECIST may have merit, based on its correlation with OS compared to RECIST, as noted by Edeline at al.[22] Recently, a second-line phase II study evaluating brivanib in patients with advanced HCC had a post-hoc exploratory analysis of efficacy using mRECIST.[23] The use of mRECIST doubled the partial response (PR) from 4.2% based on mWHO criteria to 10.4% based on mRECIST, and an increase in the stable disease fraction from 41.3% to 60.9%, with a parallel decrease in the fraction of patients with progression of disease from 41.3% to 15.2%. The main implications with this use of mRECIST in this example is the retrospective application and the incorrect translation of the mRECIST outcome into a time to tumor progression (TTP) figure, as patients may well have been off therapy based on any revised TTP figures. In fact, neither study reported the duration of therapy. The test of these challenges against the routine use of mRECIST pertains to the outcome of the randomized phase III study of brivanib versus placebo in a second-line setting, which demonstrated a median treatment duration for brivanib of 3.1 months (95% confidence interval [CI], 2.6 to 4.0 months), and a median TTP of 4.1 months.[6] As Llovet et al.[24] acknowledged: “Further research is needed to determine whether mRECIST for HCC is a valid marker of antitumor activity in patients with HCC.”

Previous efforts evaluating tumor necrosis as a correlate of response have been reported as early as the first phase II study evaluating sorafenib.[25] However, we await a pending validation study regarding a tumor necrosis/volume correlative study embedded within CALGB 80802, a randomized phase III study of sorafenib plus doxorubicin versus sorafenib trial. ( NCT01015833).

In conclusion, congratulations to Iavarone et al. for describing their experience with sorafenib in Italian patients, and bringing forth critical questions about sorafenib's use in HCC. The starting dose of sorafenib should remain at the original 400 mg twice daily. From the safety and tolerability standpoint, lower starting doses along with close observation may well be considered in patients with more advanced liver function; however, this may paradoxically negate the hoped-for benefit of sorafebnib. Sorafenib side effects may well help with a pharmacodynamic reading of sorafenib effects; however, the validity of such a clinical marker is yet to be defined. Lastly, while one may acknowledge the limitations of the current radiologic assessment tools we have at hand, mRECIST still requires further validation and should be reported in conjunction with conventional readings of oncologic effect.

  • Ghassan K. Abou-Alfa, M.D.1,2

  • 1Department of Internal Medicine, Memorial Sloan-Kettering Cancer Center, New York, NY

  • 2Department of Internal Medicine, Weill Cornell Medical College, New York, NY