Abstract
Sorafenib is the standard treatment for patients with hepatocellular carcinoma (HCC) with advanced stage disease. Although its effectiveness has been demonstrated by randomized clinical trials and confirmed by field practice studies, reliable markers predicting therapeutic response have not yet been identified. Like other tyrosine kinase inhibitors, treatment with sorafenib is burdened by the development of adverse effects, the most frequent being cutaneous toxicity, diarrhoea, arterial hypertension and fatigue. In recent years, several studies have analysed the correlation between off-target effects and sorafenib efficacy in patients with HCC. In this review, an overview of the studies assessing the prognostic significance of sorafenib-related adverse events is provided.
Introduction
Sorafenib (Nexavar; Bayer HealthCare Pharma-ceuticals, Montville, NJ, USA), an orally active multitargeted tyrosine kinase inhibitor (TKI), currently sets the standard of treatment for patients with hepatocellular carcinoma (HCC) with preserved liver function [Child–Pugh (CP) A class] and advanced stage [Barcelona Clinic Liver Cancer (BCLC-C)] [Bruix and Sherman, 2005].
The clinically significant antitumoural activity demonstrated by two international, randomized, controlled phase III trials can, however, be hampered by the rise of adverse events (AEs) occurring in up to 80% of patients [Llovet et al. 2008; Cheng et al. 2009].
The most common sorafenib-related AEs are diarrhoea, fatigue, anorexia, hypertension (HTN) and dermatological toxicities, mainly hand–foot skin reaction (HFSR) and rash/desquamation, as reported by the randomized controlled trials and confirmed by several real-life studies [Iavarone et al. 2011; Lencioni et al. 2013; Cho et al. 2013; Di Marco et al. 2013].
The development of treatment-related toxicity is of primary importance because it worsens patient quality of life and is often the cause of dose reduction or treatment discontinuation.
Moreover, the need for additional drugs to manage AEs can also alter patient compliance with the targeted therapy. All this can, in turn, be a potential cause of treatment failure.
On the one hand, these observations have emphasized the need to improve the prevention and management of AEs in order to achieve the maximum therapeutic benefit and, on the other hand, they have stimulated studies to look at the correlation between AE development and treatment outcome [Vincenzi et al. 2010; Song et al. 2011; Otsuka et al. 2012; Bettinger et al. 2012; Shomura et al. 2014].
In fact, based on the concept that the occurrence of AEs may be related to the sorafenib mechanism of action and would be due to the inhibition of one or more drug targets, such as the vascular endothelial growth factor receptor family (VEGFR1, 2, 3), platelet-derived growth factor receptor family (PDGFR-β), stem-cell growth factor receptor (c-KIT), Fms-like tyrosine kinase 3 (FLT-3), the receptor encoded by the ret proto-oncogene (RET) and Raf serine/threonine kinase activity in normal organs, several studies have tried to assess whether the off-target effects of sorafenib predict antitumoural efficacy [Zhao et al. 2012; Kamba and McDonald, 2007; Song et al. 2011; Koschny et al. 2013; Wu et al. 2008].
Thus, the potential advantage of such a predictive significance would lie in the ability to highlight the efficacy of sorafenib at an early stage, thus not discouraging treatment maintenance in patients experiencing drug-related AEs.
In this review, a summary of the sorafenib toxicity profile and the recent data found in the literature regarding the potential role of AEs as surrogate markers for treatment efficacy is presented.
Safety profile
Randomized controlled studies
Two randomized controlled trials, the Sorafenib HCC Assessment Randomized Protocol (SHARP) and the Asia-Pacific trials demonstrated a similar safety profile. In particular, both registered a high rate of AEs which were generally well tolerated and rarely led to treatment discontinuation [Llovet et al. 2008; Cheng et al. 2009].
In the SHARP trial, the overall incidence of treatment-related AEs was 80% in the sorafenib group and 52% in the placebo group, with serious AEs in 52% and 54% of the treated and the placebo groups, respectively. However, grade 3 treatment-related AEs were more common in the sorafenib group and included diarrhoea (8%), HFSR (8%), HTN (2%) and abdominal pain (2%). The rate of patients who discontinued treatment due to toxicity was 38% in the sorafenib patients versus 37% in the placebo patients.
AE-related dose reductions occurred in 26% of the sorafenib group versus 7% of the placebo group, and were due to diarrhoea (8%), HFSR (5%) and skin toxicities (3%) [Llovet et al. 2008].
In the Asia-Pacific trial, the overall incidence of treatment-related AEs was 81.9% in the sorafenib group compared with 38.7% in the placebo group, and the most frequent grade 3/4 drug-related AEs in the sorafenib group were HFSR (10.7%), diarrhoea (6.0%) and fatigue (3.4%). The rate of patients who discontinued treatment for toxicity was 19.5% and 13.3% in the sorafenib and the placebo arms, respectively.
Dose reductions owing to AEs occurred in 30.9% of the sorafenib group and in 2.7% of the placebo group, mostly related to the development of HFSR (11.4%) and diarrhoea (7.4%) [Cheng et al. 2009].
A more detailed safety profile of the two studies is shown in Table 1.
Safety profiles of sorafenib in patients with HCC from the SHARP and Asia-Pacific trials.
AE, adverse event; GI, gastrointestinal tract; HCC, hepatocellular carcinoma; HFSR, hand–foot skin reaction; SHARP, Sorafenib HCC Assessment Randomized Protocol.
Field practice studies
Whereas patients enrolled in clinical trials do not represent the real-life scenario, observational postmarketing studies can provide data regarding drug safety and efficacy in larger cohorts of patients who present comorbidities and are not selected according to the very restrictive enrolment criteria of phase III studies.
The SOFIA (Sorafenib Italian Assessment) study was a multicentre, observational, noninterventional study conducted in Italy and aimed at evaluating safety and efficacy of sorafenib in clinical practice. A total of 296 patients with HCC (88% CP A) with advanced (BCLC-C, 75%) or intermediate (BCLC-B, 25%) stage disease were enrolled.
The median treatment duration was 4.2 months [95% confidence level (CI) 3.4–5] in the 260 CP A patients and 2 months (95% CI 0.2–3.8) in the 36 CP B patients [Iavarone et al. 2011].
Sorafenib was permanently discontinued in 44% of patients due to disease progression: in 40% due to AEs, mainly fatigue (6%), and in 16% for liver function deterioration. The overall incidence of AEs was 91%, 45% of which were grade 3/4 and included fatigue (25%), HFSR (9%), arterial HTN (7%), weight loss (6%), diarrhoea (6%) and bleeding (5%).
Treatment was down dosed in 54% of patients due to AEs and liver function deterioration in 83% and 17%, respectively. The most frequent AEs leading to dose reduction were fatigue (39%), HFSR (18%) and diarrhoea (14%).
A total of 77 (26%) patients received a half dose of sorafenib for more than 70% of the treatment period (median 6.8 months, 95% CI 4.2–9.4), whereas 136 patients maintained the full dose for a median of 3 months (95% CI 2.2–3.8) and 83 patients received a half dose for less than 70% of the entire treatment period of 3 months.
To summarize, from the safety point of view, the SOFIA study differed from the two randomized controlled trials in terms of rates of toxicities causing dose reduction or treatment interruption, with a lower safety profile. Only 44% of the patients were able to continue sorafenib treatment in the absence of tumour progression, whereas the majority of the patients had to discontinue therapy due to AEs or liver deterioration (Table 2).
Safety profiles of sorafenib in patients with HCC from the GIDEON and SOFIA real-world studies.
HCC, hepatocellular carcinoma; GIDEON, Global Investigation of Therapeutic Decisions in Hepatocellular Carcinoma and of its Treatment with Sorafenib; HFSR, hand–foot skin reaction; SOFIA, Sorafenib Italian Assessment.
Interestingly, a post hoc analysis in the SOFIA study pointed out an increase in survival rates for patients (n = 77) who received a half dose of sorafenib for more than 70% of the treatment period compared with those (n = 219) treated with a full dose for more than 70% of the treatment period or receiving a half dose for less than 70% (21.6 months versus 9.6 months, p = 0.0006). Moreover, the full dose of sorafenib was independently associated with mortality in a multivariate analysis [hazard ratio (HR) 1.6, p = 0.031].
Since the main reason for dose reduction was the development of AEs, some authors have suggested that the longer median survival of patients receiving a half dose of sorafenib for over 70% of the treatment period could provide indirect evidence for the positive correlation between AEs and efficacy [Zhao et al. 2012].
The GIDEON study (Global Investigation of Therapeutic Decisions in Hepatocellular Carcinoma and of its Treatment with Sorafenib) is a phase IV, international, multicentre, prospective, open-label, noninterventional study aimed at evaluating the use of sorafenib under real-life clinical practice conditions and gathering more comprehensive data on this therapy in patients with CP B class who were excluded from the randomized clinical trials [Lencioni et al. 2013].
A total of 3371 patients were enrolled, including 666 CP B and 74 CP C patients. In the 3202 patients evaluated for safety, the overall rate of AEs and drug-related AEs was 85% and 66%, respectively. The study showed that the incidence of drug-related AEs was generally consistent across the CP A and B groups, although serious AEs were more common in CP B patients (60% versus 36%).
Discontinuation of treatment due to AEs was more common in CP B patients (40% versus 28.9%) [Marrero et al. 2013]. Final data are awaited.
The prevalence of AEs in the real-life studies is summarized in Table 2.
Prognostic significance of AEs
Many other clinical practice studies have been published in recent years. Some of these assessed the correlation between the development of one or more AEs and clinical benefit in terms of longer time to progression (TTP), disease control rate (DCR) and overall survival (OS), however with conflicting results [Vincenzi et al. 2010; Song et al. 2011; Otsuka et al. 2012; Bettinger et al. 2012; Koschny et al. 2012; Cho et al. 2013; Estfan et al. 2013; Shomura et al. 2014; Yada et al. 2014; Reig et al. 2014; Akutsu et al. 2015].
Skin toxicities
Dermatologic AEs mainly include HFSR which is characterized by erythema, dysesthesia or paresthesia on the palms and soles, and a rash. HFSR can progress to swelling of the skin, desquamation, ulceration and blistering, causing great impairment of the patient’s quality of life [Miller et al. 2014].
HFSR is a frequent AE of sorafenib and sunitinib, both of which act as inhibitors of VEGFR, PDGFR, c-KIT and FLT-3. This may indicate that inhibition of one or more of these receptors/pathways plays a role in HFSR development [Robert et al. 2005].
Since sorafenib targeting of VEGFR, PDGFR and c-KIT produces tumour vessel regression be means of the inhibition of endothelial cells, the capillary endothelium might be the first target in HFSR development. This effect, which is more frequently localized in areas of friction or trauma where the affected vessels and fibroblasts are unable to be repaired due to the continuous stimulus, leads to inflammation.
Prophylactic measures are crucial to treatment strategy. In a recent large, prospective, randomized controlled trial, the twice-daily application of prophylactic urea-based creams significantly reduced the incidence and severity of sorafenib-related HFSR in patients with HCC [Ren et al. 2015]. However, most of the studies here reported do not describe the preventive strategies that were adopted to avoid HFSR occurrence.
Several studies have found a positive correlation between the development of skin AEs and TTP or OS. Otsuka and colleagues retrospectively evaluated 94 patients with HCC treated with sorafenib. They found that patients developing skin toxicities (62%) showed significantly longer survival than patients without cutaneous AEs (HR 0.449; 95% CI 0.256–0.786; p = 0.005). Skin toxicities were also a significant prognostic factor in a multivariate analysis (HR 0.522; 95% CI 0.274–0.997; p = 0.049) [Otsuka et al. 2012].
In a retrospective study of 99 patients with advanced (BCLC-C) HCC, Cho and colleagues found that presence of HFSR was predictive of a prolonged TTP (OR 0.40; 95% CI 0.19–0.82; p = 0.007) and better OS (OR 0.40; 95% CI 0.24–0.67; p = 0.001) [Cho et al. 2013].
In a retrospective study, Vincenzi and colleagues demonstrated that early skin toxicities of all grades predicted an improved DCR and TTP in 65 patients with advanced HCC treated with sorafenib [Vincenzi et al. 2010].
Similar results were reported by Shomura and colleagues in a retrospective analysis of 37 patients with HCC showing that a skin toxicity of over G2 was a significant predictor of longer OS, and by Yada and colleagues in a retrospective study of 46 patients with HCC reporting a significantly longer TTP of patients with HFSR compared with those without [Shomura et al. 2014; Yada et al. 2014].
Recently, Reig and colleagues prospectively assessed 147 patients with HCC with the aim of validating prior evidence regarding the prognostic significance of AEs which had emerged from the above-mentioned retrospective studies [Reig et al. 2014].
A time-dependent covariate analysis identified dermatologic AEs (HFSR, rash, oedema erythema, folliculitis) requiring dose adjustment within the first 60 days (HR 0.58; 95% CI 0.36–0.92; p = 0.022) to be an independent predictor of better outcome associated with better survival (p = 0.0270). The predictive significance of dermatologic AEs for survival was confirmed by this landmark analysis (p = 0.027) which was used to exclude time-dependent bias. Interestingly, other early AEs did not have an impact on the outcomes.
Hypertension
Arterial HTN is typically considered as a class-specific toxicity of antiangiogenic treatments. Although widely reported during such therapies, a recent meta-analysis involving 13,555 sorafenib-treated patients found a significantly higher incidence of all-grade and high-grade HTN in patients with renal cell carcinoma (RCC) compared with those with non-RCC malignancies (24.9% versus 15.7% and 8.6% versus 1.8%, respectively) [Li et al. 2014]. It has been speculated that patients with RCC are more susceptible to developing HTN due to concomitant renal dysfunction or a previous nephrectomy.
Impaired angiogenesis leading to a decrease in the density of microvessels, the endothelial dysfunction associated with decreased nitric oxide production and activation of the endothelin 1 system, a potent vasoconstrictor, are the proposed mechanisms; however, the exact cause of sorafenib-related HTN remains to be defined [Wu et al. 2008; Bair and Choueiri, 2013; Hamnvik et al. 2015].
The development of sorafenib-related HTN has been extensively correlated with treatment efficacy, both in patients with HCC and those with other cancers. Estfan and colleagues studied 41 patients with advanced HCC who received sorafenib, and found significantly longer OS in patients who experienced any grade of HTN during treatment than in patients who did not develop HTN (median OS 18.2 versus 4.5 months, p = 0.0165).
The authors, however, recognized that their analysis was limited by blood pressure measurement at irregular intervals, due to the retrospective nature of the study, and by the relative small sample number [Estfan et al. 2013]. Similarly, Akutsu and colleagues found that, in 38 patients with advanced HCC, the appearance of HTN within 2 weeks of the beginning of sorafenib treatment correlated with a better TTP (153 days in the HTN group versus 50.5 days in the nonhypertension group, p = 0.017) and OS (1329 days in the hypertension group versus 302 days in the nonhypertension group, p = 0.0039) [Akutsu et al. 2015].
Unlike skin toxicity, in the previously mentioned study by Otsuka and colleagues, treatment-related HTN did not show any correlation with clinical outcomes [Otsuka et al. 2012].
In a more recent retrospective analysis of 1120 patients with RCC (32.2%), HCC (11.6%), gastrointestinal stromal tumours (GISTs, 12.5%) and other sarcomas (15.3%) treated with various anti-VEGF drugs, mainly sunitinib (52%), sorafenib (25.9%) and pazopanib (18%), the development of HTN predicted a survival advantage (25.6 versus 11.2 months, HR 0.76; 95% CI 0.65–0.89) [Hamnvik et al. 2015].
Diarrhoea
Diarrhoea is one of the main drawbacks for patients with HCC on sorafenib and was the main cause of dose reduction in the SHARP trial. It occurred in all grades of diarrhoea in 39% of the patients and in grade 3 in 8% of the population treated [Llovet et al. 2008].
Real world data from the SOFIA study showed similar results [Llovet et al. 2008; Iavarone et al. 2011].
The underlying pathophysiology of sorafenib-related diarrhoea is still unknown. However, different mechanisms involving VEGF inhibition have been hypothesized. As VEGF plays a role in maintaining parts of the normal adult vasculature, VEGFR inhibition by sorafenib could cause diarrhoea by significantly reducing the capillary network in the intestinal villi.
Other studies have supported the hypothesis that sorafenib might cause diarrhoea by inducing pancreatic exocrine dysfunction since VEGFR inhibitors can reduce the density of the capillaries in pancreatic islets and decrease zymogen granules [Fan and Iseki, 1998; Kamba et al. 2006; Mir et al. 2012; Hescot et al. 2013]
Few studies have reported a significant correlation between diarrhoea development and antitumour efficacy. Bettinger and colleagues retrospectively assessed such a correlation in 112 patients with advanced HCC. They found that diarrhoea was an independent positive prognostic factor (HR 0.41; p = 0.001) at multivariate Cox regression models, and patients with diarrhoea had a significantly longer median OS than patients without diarrhoea (14.1 months versus 7.1 months, p = 0.011). Otherwise, the HFSR did not show any prognostic significance [Bettinger et al. 2012].
Koschny and colleagues prospectively collected data regarding 46 patients with advanced HCC who were treated with sorafenib. They found that patients developing grade 2/3 diarrhoea at any time during sorafenib treatment (41%, n = 19) had increased OS compared with patients with grade 0 or 1 diarrhoea (p = 0.009). Treatment-emergent HFSR, however, did not show any correlation [Koschny et al. 2013].
In a recent observational study, Di Costanzo and colleagues tested the ability of pretreatment and in-treatment clinical variables for predicting survival. The study included a training (226 patients) and a validation (54 patients) cohort [Di Costanzo et al. 2015].
The occurrence of diarrhoea, skin toxicity and arterial hypertension within 1 month of treatment was evaluated. Three groups of patients were taken into account: patients without AEs (group 0), patients with one event (group 1) and patients with two to three events (group 2). In the training cohort, this classification correlated with 3-month progression of disease at imaging which was observed in 41.9%, 25.9% and 12.7% of the patients in groups 0, 1 and 2, respectively (p = 0.014). This trend was also confirmed in the validation cohort in which a progressive increase in median TTP and OS was observed from group 0 to group 2 (p = 0.000).
An overview of the studies assessing the prognostic significance of sorafenib-related AEs is reported in Table 3.
Overview of the studies assessing the prognostic significance of sorafenib-related AEs.
AEs, adverse events; DCR, disease control rate; HFSR, hand–foot skin reaction; OS, overall survival; TTP, time to progression.
Within the first 60 days of treatment.
Within 2 weeks of initiation of therapy.
Discussion
Pending the identification of biological markers capable of predicting tumour response, the search for clinical predictors is nowadays becoming a hot topic in oncology research.
The availability of reliable predictive markers would help in identifying patients who are likely to benefit from antitumoural treatment and to avoid unnecessary toxicity in potentially resistant patients. In this setting, understanding whether the development of AEs could act as a surrogate marker of sorafenib efficacy in patients with HCC would be clinically relevant.
Among the studies evaluating the prognostic significance of AEs in patients with HCC, skin toxicity seems to have a more convincing predictive role as it is also confirmed in a prospective analysis [Reig et al. 2014].
Such a hypothesis has been assessed in several other types of cancer. In particular, a correlation between the development of AEs and treatment efficacy has been reported in patients with breast cancer treated with endocrine therapy as well as in patients with non-small cell lung cancer treated with chemotherapy with or without VEGF inhibitors [Fontein et al. 2012, 2013; Goodwin et al. 2010].
Furthermore, a correlation between AEs and treatment efficacy has also been pointed out in patients with renal cancer treated with sorafenib and patients with colorectal cancer treated with cetuximab or bevacizumab [Cunningham et al. 2004; Di Fiore et al. 2011; Cai et al. 2013; Ravaud and Sire, 2009].
However, analysis of the studies on patients with HCC reviewed above requires some preliminary remarks. Almost all the studies were retrospective and enrolled a small number of patients. Since a longer treatment time increases the probability of developing AEs, unintended biases due to the retrospective study design could have affected the results.
There is a wide heterogeneity in the prevention, clinical assessment and management of sorafenib-related AEs, and this may have influenced the incidence and prevalence rates, especially in studies which investigated the prognostic role of AEs according to their severity or time of appearance [Edmonds et al. 2012].
Moreover, to date, variability in the sorafenib pharmacokinetic has been poorly investigated. The data available seem to indicate a large interpatient pharmacokinetic variability and would suggest a relationship between increased cumulated sorafenib exposure, and the incidence and severity of AEs [Boudou-Rouquette et al. 2012b].
A preliminary study has also pointed out the influence of pharmacogenetic variants on sorafenib-induced toxicity. In particular, a higher risk for grade 2 and higher diarrhoea in patients with variants of the uridine diphosphate glucuronyl transferase 1A9, a sorafenib metabolizing enzyme, has been reported [Boudou-Rouquette et al. 2012a].
An additional source of interindividual pharmacokinetic variability seems to be the albumin plasma level which influences sorafenib clearance and can vary among patients with HCC and cirrhosis [Tod et al. 2011].
Furthermore, genetic polymorphisms in VEGF and VEGFR have been correlated to both tumour neoangiogenesis and differences in the incidence of HFSR [Lee et al. 2013; Scartozzi et al. 2014].
Taken together, these findings suggest that the differences in the reported incidence, severity and prognostic significance of sorafenib-related AEs among studies may be related to multiple factors.
In order to definitively ascertain the prognostic role of sorafenib-related AEs, we suggest three possible major lines of development: standardization in prevention and detection measures as well as in the management of AEs, with the aim of accurately defining the incidence of these toxicities and the degrees of severity; correlation of toxicities with sorafenib plasma concentration determining the association between AEs and drug exposure; definition of pathogenic mechanisms underlying the development of AEs assessing the influence of genetic polymorphisms on the sorafenib pharmacokinetic and on prevalence and severity of AEs [Jain et al. 2010; Lee et al. 2013; Scartozzi et al. 2014].
Finally, multicentre and prospective studies enrolling an adequate number of patients are needed to further test and validate this prognostic correlation.
Footnotes
Funding
This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.
Conflict of interest statement
Prof. Luigi Bolondi: Bayer AG (speaker fee, advisory board), Bristol-Myers Squibb (research grant, advisory board), Bracco (research grant), Roche (speaker fee). The other authors declare that they have no conflicting interests.