Incidence and relative risk of hepatic toxicity in patients treated with anti-angiogenic tyrosine kinase inhibitors for malignancy

Abstract

Aim

The aim of this study is to investigate the incidence and risk of hepatic toxicity in patients receiving tyrosine kinase inhibitors (TKIs) through a large up-to-date meta-analysis of available clinical trials.

Methods

PubMed was reviewed for phase III randomized trials with axitinib, pazopanib, sorafenib, sunitinib, regorafenib or vandetanib. The characteristics of each study and incidence of all and high grades of ALT, AST and total bilirubin increase were collected.

Results

A total of 3691 patients was available for meta-analysis, 1170 had metastatic renal cell carcinoma; 950 had advanced non-small cell lung carcinoma, 454 had hepatocarcinoma, 753 had metastatic colorectal cancer and 362 had metastatic soft-tissue sarcoma. The incidence of ALT, AST and bilirubin increase of any grade in patients treated with TKIs was 34.0% (95% CI 31.6, 36.3), 39.2% (95% CI 36.7, 41.6) and 21.8% (95% CI 19.9, 23.7), respectively. The incidence of the high grade increase was 5.2% (95% CI 4.2, 6.4), 5.0% (95% CI, 3.8, 6.2) and 1.7% (95% CI 1.1, 2.4), respectively. The relative risk of ALT, AST and total bilirubin increase was 1.85, 2.19 and 1.79 for any grade and 2.75, 2.39 and 1.65 for high grade, respectively.

Conclusions

Hepatotoxicity is a relative common event occurring in 23–40% of patients treated with TKIs. Despite this, only 5% of patients have had high grade of toxicity. A better knowledge of this phenomenon may prevent high grade toxicity and reduce treatment discontinuation due to this adverse event.

Introduction

Angiogenesis had been regarded as a possible target for treatment of solid tumours considering that it is over-expressed in up to 60% of human cancers 1. Preliminary data in 1971 hypothesized that tumour growth is dependent on angiogenesis and by their effectors such as the vascular endothelial growth factor (VEGF) and its receptor (VEGFR). Despite this evidence, no angiogenic inhibitors existed before 1980, and the first one was reported from Folkman's laboratory 2. The concept of using anti-VEGF strategies in cancer treatment was soon developed and current strategies were against VEGF by monoclonal antibodies or against VEGFR by tyrosine kinase inhibitors (TKIs).

Several TKIs have been tested in recent years in large randomized controlled trials for treatment of solid tumours and some of these such as sorafenib, sunitinib, axitinib, pazopanib, regorafenib and vandetanib, received approval for treatment of metastatic renal cell carcinoma (mRCC), hepatocellular carcinoma (HCC), neuroendocrine tumours (NET), gastrointestinal stromal tumours (GIST), metastatic colorectal cancer (mCRC) and tyroid cancer.

The use of these agents requires early and appropriate management of side effects, such as hepatic toxicity, in order to avoid unnecessary dose reductions and transitory or definitive treatment discontinuations. Hepatic toxicity arising during TKI therapy can be related to pre-existing diseases (such as viral or drug-induced cirrhosis or Gilbert's disease) or to a metastatic invasion that can affect the clearance, metabolism and absorption of TKIs so that many clinical trials exclude patients with significant liver dysfunction 3.

Toxicity related to the use of TKIs were recently reviewed and defined as ‘on target’ and ‘off target’ based on their dependence with the pharmacological effect. Hepatotoxicity was reported as an adverse event common to the VEGFR and to the epidermal growth factor receptor (EGFFR) inhibitors 4,5. Despite these data, the risk associated with the use of anti-VEGFR TKIs has not yet been defined. Therefore, we sought to investigate the incidence and risk of developing this toxicity in patients receiving axitinib, pazopanib, sorafenib, sunitinib, regorafenib or vandetanib through a large up-to-date meta-analysis of available clinical trials.

Methods

Definition of the outcome

Treatment with the anti-VEGFR TKIs was considered as the experimental arm and the other treatments as the standard one. Three variables were separately considered as an expression of hepatotoxicity, the increase of alanine aminotransferase (ALT), the increase of aspartate aminotransferase (AST) and the increase of total bilirubin. For each variable, we consider the increase of all grades and grade 3–4 as the main outcomes and the analysis was conducted in order to find a significant difference between the two arms. Adverse events were defined as per version three of the National Cancer Institute's Common Terminology Criteria for Adverse Events (CTCAE) criteria because of its use in the selected trials 6.

Selection of the studies

We reviewed PubMed for citations from January 2006 to November 2012. The search criteria included articles published in the English language with the following anti-VEGFR TKIs, axitinib, pazopanib, sorafenib, sunitinib, regorafenib and vandetanib, used for treatment of solid cancers. The search was restricted to randomized controlled trials (RCTs) and phase III trials in which a TKI was compared with a non-TKI. If more than one publication was found for the same trial, the most recent was considered for the analysis. Study quality was assessed by using the Jadad 7-item scale that included the randomization, double-blinding and withdrawals. The final score was reported between 0 and 5 7.

Data extraction

Data extraction was conducted independently by two authors (RI and AP) and according to the Preferred Reporting Items for Systematic review and Meta-Analysis (PRISMA) statement 8, and any type of discrepancy was resolved by consensus. Data extracted for each trial were: first author's name, year of publication, trial phase, the number of valuable patients, number of arms, randomization rate, drugs used in the experimental and standard arm, dosage, type of cancer, stage of disease, median age, percentage of liver metastases, median follow-up, median treatment duration, median progression free survival (PFS) and reported data on ALT, AST and total bilirubin increase.

Statistical methods

To calculate the incidence of hepatic toxicity and 95% confidence interval (CI), the total number of patients treated with TKIs and control therapies with their relative number of patients developing toxicity were extracted from the safety profile of each selected trial. We have also calculated, for each study, the relative risk (RR) and the CIs of events in patients assigned to TKIs compared with control. To calculate the 95% CIs, the variance of a log-transformed study specific RR was derived using the delta method 9.

Statistical heterogeneity between trials included in the meta-analysis was assessed using Cochrane's Q statistic, and inconsistency was quantified with the I² statistic (100% × [Q − d.f.)/Q]) 10. The assumption of homogeneity was considered invalid for P values less than 0.1. Summary incidence and RRs were calculated using random or fixed effects models depending on the heterogeneity of included studies. When substantial heterogeneity was not observed, the pooled estimate was calculated based on the fixed effects model and reported using the inverse variance method.

When substantial heterogeneity was observed, the pooled estimate was calculated based on the random effects model and reported using the method of Der Simonian et al. 11, which considers both within and between study variations. Publication bias was evaluated using funnel plots (plots of study results against precision) and with the Begg et al. 12 and Egger et al. 13 tests. A two-tailed P value of less than 0.05 was considered statistically significant. All data were collected using Microsoft Office Excel 2007. Statistical analyses were performed using PASW statistic software (version 18) and MIX 2.0, a professional software for meta-analysis in Excel, Version (2.0.1.4) 14,15.

Results

A total of 166 citations resulted from the electronic search. After the selection procedure, 34 full text articles were considered for further assessment; nine were excluded because their first results were updated in following articles. Of remaining 25 articles, 24 are potentially eligible for meta-analysis and one was excluded because it compared two TKIs. At the end of the selection procedure 18 articles were excluded because of the lack of data related to hepatic toxicity. A total of six articles was included in the final analysis because of their adequate quality (Figure 1) 16–21. All of them were randomized phase III trials and five were double-blind. Two trials randomized patients with mRCC and the other patients were affected by HCC, non-small cell lung cancer (NSCLC), mCRC and metastatic soft-tissue sarcoma (mSTS), respectively. The characteristics of each study are presented in Table 1.

Figure 1.

Selection of randomized controlled trials (RCTs) included in the meta-analysis. TKI tyrosine kinase inhibitor

Table 1.

Main characteristics of the included studies

First author (year)	Trials design							Type of patients				Study results			Jadad score
	Phase	Number° of patients	Number° of arms	Randomization rate	Drug	Dosage	Control arm	Type of cancer	Phase of disease	Median age, range (years)	Hepatic metastases (%)	Median follow-up (months)	Median treat duration (weeks)	Median PFS (months)
Motzer et al. (2009) [17]	III	750	2	1:1	SU	50 mg day−1, 4/6 weeks	IFN	mRCC	Advanced	62 (27–87)	25.2	NR	44 vs. 12	11 vs. 5	3
Sternberg et al. (2010) [16]	III	435	2	2:1	PZ	800 mg day−1	Placebo	mRCC	Advanced	58 (28–85)	24.6	NR	29.6 vs. 15.2	9.2 vs. 4.2	5
Kudo et al. (2011) [18]	III	458	2	1:1	SO	400 mg twice daily	Placebo	HCC	Advanced	69 (29–86)	100	NR	17 vs. 20	5.4 vs. 3.7	5
Scagliotti et al. (2012) [19]	III	960	2	1:1	SU	37.5 mg day−1	Erlotinib	NSCLC	Advanced	61 (31–85)	19.1	21.3 vs. 22.0	18.6 vs. 19	3.6 vs. 2.0	5
van der Graaf et al. (2012) [20]	III	369	2	2:1	PZ	800 mg day−1	Placebo	mSTS	Advanced	55 (44–64)	28.2	14.9 vs. 14.6	16.4 vs. 8.1	4.6 vs. 1.6	5
Grothey et al. (2013) [21]	III	760	2	2:1	RE	160 mg day−1	Placebo	mCRC	Advanced	61 (54–67)	–	–	2.8 vs. 1.8	1.9 vs. 1.7	5

HCC, hepatocellular carcinoma; IFN, interferon; mCRC, metastatic colorectal cancer; NSCLC, non-small cell lung cancer; PFS, progression free survival; PZ, pazopanib; RCC, renal cell carcinoma; RE, regorafenib; SO, sorafenib; STS, soft tissue sarcoma; SU, sunitinib.

A total of 3689 patients was available for meta-analysis: 1170 had mRCC; 950 had advanced NSCLC; 753 had mCRC; 454 had HCC and 362 had mSTS.

Incidence and relative risk of ALT increase

For the incidence of any grade of ALT increase, a total of 2938 patients was included in the final analysis. An increase in ALT was reported in 546 out of 1606 patients treated with TKIs compared with 324 out of 1332 patients in the control arm, with an incidence of 34.0% (95% CI 31.6, 36.3) in TKI treated patients and 24.3% (95% CI 22.0, 26.7) in the control arm. The RR (random effect) to develop any grade of ALT increase was 1.85 (95% CI 1.21, 2.82; P = 0.004) in patients treated with TKIs compared with controls (Q = 35.6%, P < 0.001, I² = 88.8%) (Figure 2A). When TKIs were compared with placebo the RR was 2.71 (95% CI 2.05, 3.58; P < 0.001) and when compared with the active control the RR was 1.16 (95% CI 1.02, 1.32; P = 0.02) without significant heterogeneity in both cases. When stratified by type of drug, sorafenib and pazopanib, but not the sunitinib, increased the RR for all grade ALT elevations (Table 2).

Figure 2.

Relative risk of ALT aminotransferase increase of all grades (A) and high grade (B)

Table 2.

Incidence and relative risk of liver toxicity by tyrosine kinase inhibitors (only drugs with at least two parameters were reported)

Hepatic parameters	Grade of toxicity	Sorafenib			Sunitinib			Pazopanib
		Number of patients°	Incidence (95% CI)	RR (P value)	Number of patients	Incidence (95% CI)	RR (P value)	Number of patients	Incidence (95% CI)	RR (P value)
ALT	All	456	21.0% (15.5, 26.5)	4.33 (P < 0.001)	1685	38.1% (34.8, 41.4)	1.14 (P = 0.28)	797	33.1% (29.0, 37.2)	2.42 (P < 0.001)
	3–4	456	8.3% (4.5, 12.1)	4.71 (P = 0.004)	1685	2% (1.9, 2.1)	1.51 (P = 0.28)	435	12.1% (8.1, 16.0)	8.75 (P = 0.002)
AST	All	456	24.9% (19.1, 30.7)	5.14 (P < 0.001)	1685	47.3% (47.2, 47.3)	1.41 (P < 0.001)	797	32.3% (32.2, 32.4)	2.91 (P < 0.001)
	3–4	456	13.1% (8.5, 17.7)	4.25 (P < 0.001)	1685	1.8% (1.7, 1.8)	1.33 (P = 0.74)	435	7.9% (4.6, 11.2)	11.5 (P = 0.016)
Total Bilirubin	All	–	–	–	1685	28.2% (28.1, 28.2)	3.60 (P = 0.21)	797	23.6% (23.5, 23.7)	3.31 (P < 0.001)
	3–4	–	–	–	1685	1.1% (1.0, 1.1)	1.39 (P = 0.56)	435	3.1% (0.9, 5.3)	1.5 (P = 0.54)

ALT, alanine aminotransferase; AST, aspartate aminotransferase; CI, confidence interval; RR, relative risk.

Grade 3–4 of the ALT increase was evaluable in 2576 patients with 71 events out of 1367 patients treated with TKIs compared with 17 out of 1209 patients in the control arm. The incidence of high grade of ALT increase was 5.2% (95% CI 4.2, 6.4) for TKIs compared with 1.4% (95% CI 0.7, 2.1) in the control arm. The RR to develop grade 3–4 of ALT increase was 2.75 (95% CI 1.56, 4.82; P < 0.001) in patients treated with TKIs compared with controls (Figure 2B). No significant heterogeneity was observed in the RR analysis for grade 3–4 (Q = 6.01, P = 0.11, I² = 50.1%). When TKIs were compared with placebo the RR was 5.89 (95% CI 2.52, 13.77; P < 0.001) and when compared with the active control the RR was 1.51 (95% CI 0.71, 3.20; P = 0.28) without significant heterogeneity in both cases.

As reported for all grades of ALT elevation, a significant increase in the RR of grade 3–4 toxicity was present for sorafenib and pazopanib but not for sunitinib (Table 2). No differences for any grade and high grades of ALT increase were found based on the type of disease (Table 3).

Table 3.

Incidence and relative risk of liver toxicity in patient affected by metastatic renal cell carcinoma (mRCC) or not

Toxicity	mRCC				Non-mRCC
	Number of patients	RR	95% CI	P value	Number of patients	RR	95% CI	P value
Any grade
Increased ALT	1170	1.71	0.93, 3.15	0.085	1768	2.24	0.74, 6.76	0.153
Increased AST	1170	1.99	1.06, 3.76	0.033	1768	2.97	0.98, 8.95	0.053
Increased bilirubin	1170	5.69	1.93, 16.80	0.002	2065	2.56	1.00, 6.53	0.049
High grade
Increased ALT	1170	2.58	1.18, 5.64	0.017	1406	2.94	1.30, 6.61	0.009
Increased AST	1170	3.01	0.31, 29.5	0.343	1406	2.36	0.67, 8.27	0.180
Increased bilirubin	1170	2.00	0.61, 6.51	0.250	1703	1.46	0.56, 3.79	0.440

Incidence and relative risk of AST increase

For incidence of all grades of AST increase, a total of 2938 patients was included in the final analysis. An increase in AST was reported in 629 out of 1606 patients treated with TKIs compared with 320 out of 1332 patients in the control arm, with an incidence of 39.2% (95% CI 36.7, 41.6) and 24.0% (95% CI 21.7, 26.4), respectively. The RR (random effect) to develop any grade of AST increase was 2.19 (95% CI 1.49, 3.23; P < 0.001) in patients treated with TKIs compared with controls. When TKIs were compared with placebo the RR was 3.61 (95% CI 2.31, 5.64; P < 0.001) and when compared with the active control the RR was 1.41 (95% CI 1.26, 1.59; P < 0.001) without significant heterogeneity in both cases. A higher RR to have any grade of AST elevation was present independent of the type of drug used (Table 2).

Grade 3–4 of the AST increase was evaluable in 2576 patients with 68 events out of 1367 patients in the treatment group compared with 21 out of 1209 patients in the control arm. The incidence of high grade of AST increase was 5.0% (95% CI 3.8, 6.2) for TKI compared with 1.7% (95% CI 1.0, 2.5) in the control arm. The RR to develop a grade 3–4 AST increase (random effect) was 2.39 (95% CI 0.94, 6.09; P = 0.067) in patients treated with TKIs compared with controls. Significant heterogeneity was observed with a fixed model in the analysis for all grades (Q = 30.3%, P < 0.001, I² = 86%) and grade 3–4 (Q = 8.31, P = 0. 040, I² = 63.9%) of AST increase.

A different RR for grade 3–4 AST increase was found when a TKI was compared with placebo (RR = 4.88, 95% CI 2.32, 10.27; P < 0.001) or with the active control (RR = 1.13, 95% CI 0.54, 2.37; P = 0.74), no heterogeneity was found in the first case (Q = 0.82, P = 0.36, I² = 0%). A higher RR to have grade 3–4 of AST elevation was present in patients treated with sorafenib or pazopanib but not in those treated with sunitinib (Table 2).

No differences for any grade and high grades of AST increase were found based on the type of disease (Table 3).

Incidence and relative risk of total bilirubin increase

For incidence of all grades of bilirubin increase a total of 3235 patients was included in the final analysis. The increase of the total bilirubin of all grades was reported in 409 out of 1877 patients treated with TKIs compared with 152 out of 1358 patients in the control arm, with an incidence of 21.8% (95% CI 19.9, 23.7) in TKIs and 11.2% (95% CI 9.5, 12.9) in the control group. The RR (random effect) to develop any grade of bilirubin increase was 1.79 (95% CI 1.50, 2.12; P < 0.001) in patients treated with TKIs compared with controls (Figure 3C). Significant heterogeneity was observed with a fixed model in the analysis for all grades (Q = 40.1%, P < 0.001, I² = 73%).

Figure 3.

Relative risk of total bilirubin increase of all grades (A) and high grade (B)

No different RR for all grade bilirubin increase was found when a TKI was compared with placebo (RR = 3.65; 95% CI 2.40, 5.48; P < 0.001) or with the active control (RR = 3.61; 95% CI 0.50, 26.25; P = 0.20), and no heterogeneity was found in the first case (Q = 0.97, P = 0.61, I² = 0%). A higher RR to have any grade of bilirubin elevation was present in patients treated with pazopanib or regorafenib (RR = 5.69, 95% CI 2.07, 15.7; P < 0.001), but not in those treated with sunitinib (Table 2).

Grade 3–4 of the total bilirubin increase was evaluable in 28 out of 1638 patients in the treatment group compared with 10 out of 1235 patients in the control arm, with an incidence of 1.7% (95% CI 1.1, 2.4) and 0.8% (95% CI 0.3, 1.4), respectively. The RR to develop grade 3–4 of total bilirubin increase was 1.65 (95% CI 0.79, 3.47; P = 0.19) (Figure 3D) and no differences were found when a TKI was compared with placebo or with an active control or comparing patients treated with sunitinib, pazopanib or regorafenib.

A higher RR for any grade of bilirubin increase was found in mRCC compared with non-mRCC patients but no differences were found with high grade of disease (Table 3).

Study quality

Randomized treatment allocation sequences were generated in all trials. The control arm was placebo in four studies (in mCRC, mRCC, HCC and mSTS), interferon was used in one study with mRCC and another one compared the combination of sunitinib plus erlotinib with the erlotinib in NSCLC. The hepatic toxicity, as defined by the CTCAE version 3 criteria, was reported in all the included trials. One trial only reported all grades of toxicity 20 and another one only reported the increase of bilirubin 21. Follow-up time was reported only in three trials. Jadad scores for each trial are listed in Table 1. The mean score was 4.7 and all were high quality trials (Jadad score 3 to 5).

Publication bias

No significant publication biases were detected for all grades of hepatic toxicities: P values from Begg's and Egger's test were 0.32 and 0.14 for ALT increase, 0.32 and 0.06 for AST increase and 0.80 and 0.052 for total bilirubin increase, respectively. Similarly, no significant publication biases were detected for high grades of ALT, AST and bilirubin increase: P values from Begg's and Egger's test were 0.17 and 0.44, 1.0 and 0.80, 0.09 and 0.09, respectively.

Discussion

In this study, we report the incidence and the relative risk of hepatic toxicity in patients treated with anti-VEGFR TKIs. To our knowledge this is the first study that meta-analyzed this aspect considering all the approved TKIs for treatment of solid tumours. Single trials have reported that hepatic toxicity is a relatively common effect with a variable incidence ranging from 23% to 40% of treated patients, and it represents a serious event in about 5% of cases. Moreover, we report that the incidence of ALT and AST increase was quite similar, and the RR for high grade toxicity was 2.75 and 2.39 respectively. We also found that this toxicity did not depend on the type of disease, mainly if patients were divided by those affected by mRCC or not. In our analysis, the incidence of all grade events was greater for sunitinib and pazopanib compared with sorafenib. Even if the RR of high grade of hepatic toxicity is greater in patients treated with sorafenib and pazopanib compared with sunitinib, these data must be carefully interpreted considering the higher RR of hepatic toxicity found in trials comparing a TKI with placebo instead of an active control. Therefore, any conclusions about the different RR of high grade events between sunitinib (compared with interferon) and pazopanib (compared with placebo), should take into account this evidence.

The mechanism underlying TKI-related hepatic toxicity is still not well clarified. At present, no polymorphisms in molecules involved in TKI pharmacokinetic and pharmacodynamic pathways have been associated with transaminase elevation, while pazopanib-induced hyper-bilirubinaemia seems to be frequently associated with the Gilbert's uridine-diphosphoglucuronate glucuronosyltransferase 1A1 (UGT1A1) polymorphism 22.

The histologic alterations of other TKI related hepatotoxicity have already been reported in the medical literature. These describe sinusoidal congestion, necrosis of hepatocytes, inflammation, and hepatocyte drop out around the hepatic venule consistent with drug toxicity following the use of imatinib, a drug which targets the c-kit receptor, BCR-Abl, PDGFR and other tyrosine kinases 23. The histologic alterations due to pazopanib-related hepatic toxicity have recently been described in two cases that showed mild active cholestatic hepatitis with inflammation that predominantly involved portal tracts. This was composed of lymphocytes, histiocytes, neutrophils and numerous eosinophils. Bile duct epithelial injury was also present and associated with a lymphocytic infiltrate within the duct epithelium, bile ductular proliferation and cholangiolitis. Mild centrilobular intrahepatic cholestasis was described, which was consistent with general bile duct damage 24.

Classification of hepatic toxicity is a current challenge. The common terminology criteria for adverse events (CTCAE) identify four grades of toxicity based on the level of the increase compared with the upper limit of normal of liver transaminase (AST or ALT), serum bilirubin (direct and indirect), PT/INR, aPTT and albumin 5. If only the changes in transaminases were considered, the CTCAE is useful to categorize the level of toxicity and to express the hepatic histological injury, but it does not directly account for the liver function. This can be better reported by a multiparametric index such as Child-Pugh which includes, other than the transaminases, the protein synthesis activity expressed by the coagulation function and the value of the albumin 3. Additionally, the role of these classifications to affect TKI dose-modification is still undefined.

In fact, current guidelines consider other intervals to decide the drug discontinuation or its interruption. These recommend the monitoring with a serum liver test every 4 weeks for the first 4 months and periodically in the following months. If an isolated increase of serum aminotransferase less than eight-fold the upper limit of normal (ULN) is reported without changes in the value of total bilirubin, treatment may be continued but weekly monitoring of liver tests is recommended. If the aminotransferase level is greater than three times the ULN and there is a concomitant increase of the total bilirubin greater than two times the ULN, the dosage of fractioned bilirubin is required. Subsequently, if the direct bilirubin is greater than 35% of the total value, treatment has to be discontinued. Interruption of treatment is mandatory if aminotransferase increases to greater than eight times the ULN until recovery of toxicity or a grade less than the grade 1. Treatment re-introduction may be considered at lower doses and with weekly monitoring of liver tests if a new increase of the aminotransferase greater than three times the ULN is recorded 25.

Despite these guidelines, several authors suggest the monitoring of the liver tests every 2 weeks for the first 8 weeks because the measurement at 4 weeks may lead to the precocious diagnosis of hepatic toxicity 24. The use of steroids or N-acetylcysteine and glutathione to manage this toxicity is suggested by some authors 24,25, but no experimental evidence is available.

It is important to note some limitationsof our study. First, this was a meta-analysis based on studies and not on the patients' data, then confounding variables such as patient's comorbidities, previous chemotherapeutic exposure, concomitant treatments that affect liver function, could not be incorporated into the analysis. Second all the included studies were conducted in patients with adequate organ function at study entry, so data about the incidence and severity of hepatic toxicity might be not directly related to the overall population treated with TKIs in clinical practice. Third we were unable to consider patients with the concomitant increase of two or more liver tests with the intent to describe more dangerous conditions.

In conclusion, the use of anti-VEGFR TKIs is associated with a significant increase in the incidence and risk of developing hepatic toxicity. The different toxic profile of these molecules should be taken into account in the choice of treatment when possible, and according to the patient's comorbidity and drug availability. Finally, we concur with the need to examine the existence of a relationship between serum drug concentrations and hepatic toxicity as recently suggested 24, as well as the evaluation of potential predictive factors in order to avoid toxicity and premature drug discontinuation.

Competing Interests

All authors have completed the Unified Competing Interest form at http://www.icmje.org/coi_disclosure.pdf (available on request from the corresponding author) and declare no support from any organization for the submitted work. RI and GP had specified relationships with AVEO, GSK, Bayer and Pfizer in the previous 3 years and no other relationships or activities that could appear to have influenced the submitted work.

Supporting Information

Additional Supporting Information may be found in the online version of this article at the publisher's web-site: