Collateral damage: toxic effects of targeted antiangiogenic therapies in ovarian cancer

Abstract

First-line chemotherapy fails in more than 20% of patients with epithelial ovarian cancer and about 40–50% of women who respond to initial treatment relapse within 2 years. In the recurrent setting, second-line chemotherapeutic agents have a 15–20% response rate with no cures. Fortunately, clinical investigations that have assessed the efficacy of new, biologically targeted therapies have reinvigorated therapeutic options for patients living with ovarian and other malignancies. In view of the fact that ovarian cancer is one of the most angiogenic neoplasms, there is great hope that implementing targeted agents with antiangiogenic properties will improve outcomes. However, as experience grows with the antitumour activity of these drugs, new toxic effects are emerging. The effects of antiangiogenic agents on molecules and processes that also have physiologically important roles in healthy tissues are at the crux of these toxic effects, or “collateral damage”. This review discusses the leading toxic effects encountered and anticipated in clinical investigation and practice with antiangiogenic agents in patients with ovarian cancer, with particular focus on potential management strategies.

Introduction

Although tumour reductive surgery and cytotoxic chemotherapy have been the mainstay of treatment for ovarian cancer, the therapeutic benefit of this treatment might be reaching a ceiling. Minimum change in disease-specific mortality over the past three decades underscores this idea. Consequently, the pursuit of new treatment strategies needs to focus on targeting specific biological processes that drive the growth and progression of this malignant disease. Among these new strategies, therapies targeting tumour-supportive angiogenesis and associated growth factors, such as vascular endothelial growth factor (VEGF), are showing promise in clinical trials (webappendix pp 1–2). In view of a reported response rate of 16–21%^1–3 for single-agent bevacizumab as second-line therapy for ovarian cancer and its clinical profile in recently completed and ongoing phase 3 trials, bevacizumab is poised to be the first targeted agent approved for the treatment of ovarian cancer. The inherent hope with targeted therapies, such as bevacizumab, is to achieve an antitumour response, while largely sparing damage to healthy tissues. However, considering the roles that many new drug targets have in healthy physiological processes, it is becoming increasingly apparent that “collateral damage” might occur. This review will focus on some of the unintended effects of antiangiogenic therapies and will suggest strategies for their management. Although this paper largely encompasses toxicity data that have emerged from phase 1 and 2 trials of agents with antiangiogenic properties in ovarian cancer, it also considers toxic effects that have been recorded with these agents in other tumour types, because such effects are likely to be recapitulated in ovarian cancer. As clinical experience with targeted agents evolves, it is evident that patients with ovarian cancer might be more uniquely susceptible to specific adverse events, such as intestinal perforation. However, other toxic effects, such as those involving the cardiac and endocrine systems, are less likely to depend on primary disease site. Thus, we anticipate that many of the core strategies delineated in this review for coping with toxic effects associated with antiangiogenic therapies will be applicable to a wide variety of tumour types. As we strive to integrate these promising agents into clinical practice, our ability to maximise their therapeutic potential will depend on effective management of often unfamiliar adverse effects.

Hypertension

Hypertension is one of the most prevalent comorbid conditions identified in patients on cancer registries (38%), and has emerged as one of the most common side-effects of antiangiogenic therapy.⁴ VEGF antagonists are the most culpable agents, whereas other agents, such as vascular damaging agents, epidermal growth factor receptor (EGFR) inhibitors, matrix metalloproteinase inhibitors (MMPIs), and monoclonal antibodies directed against integrins are rarely associated with worsening blood pressure (table 1). The pathogenesis of angiogenesis inhibitor-induced hypertension is not thoroughly understood. VEGF antagonism can cause a decrease in nitric-oxide production by inhibition of nitric-oxide synthase. Suppression of nitric oxide leads to vasoconstriction and decreased sodium ion renal excretion, which culminates in elevated blood pressure.³⁵ Vascular rarefaction, a functional decrease in the number of arterioles and capillaries that results in increased peripheral vascular resistance, is another prevailing hypothetical mechanism. For many VEGF antagonists, the occurrence of hypertension is dose-dependent. For example, the overall incidence of hypertension in patients receiving low-dose (3, 5, or 7·5 mg/kg per dose) versus high-dose (10 or 15 mg/kg per dose) single-agent bevacizumab is 2·7–32% and 17·6–36%, respectively.³⁶ In view of the fact that patients with ovarian cancer are routinely treated with high-dose bevacizumab, it is not surprising that hypertension is a significant adverse event in this patient population. A review^{3–7,9–12,37–43} of 16 clinical studies involving nearly 600 patients with epithelial ovarian, primary peritoneal, or fallopian-tube cancer, treated with bevacizumab alone or in combination with various cytotoxic agents or the EGFR inhibitor, erlotinib, showed that grade 3 or 4 hypertension (blood pressure >150/100 mm Hg requiring more than one antihypertensive or hypertensive crisis) affects up to 21% of patients (table 1). Although the incidence of grade 3 or 4 hypertension with either single-agent bevacizumab or combination treatment with erlotinib is in the range of 8–10%, higher rates are noted when cytotoxic agents, shown to have distinct antiangiogenic activity, such as taxanes and topotecan, are added to bevacizumab or aflibercept, a VEGF ligand-binding fusion protein. A phase 3 study of weekly paclitaxel plus bevacizumab versus weekly paclitaxel alone for metastatic breast cancer support this idea. Nearly 15% of women in the combination group developed grade 3 or 4 hypertension compared with none in the paclitaxel-alone group.⁷ Similar rates can be projected for the upcoming phase 2 study of bevacizumab, carboplatin, and weekly paclitaxel for upfront treatment of ovarian cancer.

Table 1.

Incidence of grade 3/4 hypertension with antiangiogenic agents

	Incidence, %
Single-agent bevacizumab1,2	9–10
Bevacizumab plus carboplatin plus paclitaxel5–7	4–13
Bevacizumab plus weekly paclitaxel8	14·8
Bevacizumab plus topotecan9	21
Bevacizumab plus cyclophosphamide10	11
Bevacizumab plus carboplatin plus gemcitabine11	9
Bevacizumab plus erlotinib12	8
Bevacizumab plus sorafenib13	26
Single-agent aflibercept14	9
Aflibercept plus cytotoxic chemotherapy15	13–32
Single-agent sunitinib16–18	0–10
Single-agent sorafenib16,19,20	<5
Sorafenib plus gemcitabine21	12
Single-agent pazopanib22	0–8
Single-agent cediranib23–25	33–48
Single-agent SU666826	0
Single-agent vandetanib27	≤5
Imatinib±docetaxel28	0
EGFR TKIs with or without cytotoxic chemotherapy28	0
Cetuximab with or without cytotoxic chemotherapy28	0
Fosbretabulin and DMXAA29–32	0
Volociximab33	<5
Vitaxin34	0

EGFR=epidermal growth factor receptor. TKI=tyrosine-kinase inhibitor.

DMXAA= 5,6-dimethylxanthenono-4-acetic acid.

The incidence and severity of hypertension vary across VEGF-receptor tyrosine-kinase inhibitors (RTKIs), and hypertension is not an inevitable side-effect of other antivascular agents. For example, in phase 1 studies of vascular damaging agents, fosbretabulin and 5,6-dimethylxanthenone-4-acetic acid (DMXAA), hypertension at less than grade 3 was noted in 4–35% of patients with advanced solid malignancies.^30,31,44 In a phase 2 study of fosbretabulin with paclitaxel and carboplatin, in patients with recurrent platinum-resistant ovarian cancer, grade 1–2 hypertension was noted in 22% of patients. There were no grade 3 or 4 events, although four patients were treated with nitrates transiently.⁴⁵ Grade 3 or 4 hypertension was not noted in five phase 1 or 2 studies of imatinib with or without docetaxel²⁶ or in ten phase 2 prospective clinical trials of EGFR-targeted therapy with or without cytotoxic agents in ovarian cancer.²⁷

Currently, patients with pre-existing hypertension, or those treated with antihypertensive drugs, are restricted from participating in several early clinical studies that are assessing various angiogenesis inhibitors. However, in view of the pervasiveness of hypertension in patients with cancer, these criteria might be too exclusionary.⁴⁶ Algorithms have been developed for the management of drug-induced hypertension to aid with continued administration of angiogenesis inhibitors without dose reduction or interruption, in the interest of improving both safety and efficacy.^47,48 Bevacizumab-induced hypertension can emerge at any time during therapy and an increase in blood pressure with VEGF RTKIs, such as sorafenib, typically occurs within 3 weeks of beginning therapy.^49,50 Due to the paucity of studies addressing blood-pressure surveillance, no clear recommendation can be made. However, it seems reasonable to follow blood pressure on a weekly basis for 6 weeks after initiating an angiogenesis inhibitor, with scheduled blood-pressure checks every 3 weeks thereafter. Even in this specialised patient population, blood-pressure goals should still be set using updated clinical practice guidelines, such as the Seventh Report of the Joint National Committee (JCN 7).⁵¹

Given the aforementioned potential mechanisms for angiogenesis inhibitor-induced hypertension, angiotensin-converting enzyme (ACE) inhibitors, angiotensin II receptor blockers, and calcium-channel blockers might be effective first-line antihypertensive agents in bevacizumab trials. Dihydropyridines, such as amlodipine and felodipine, are currently the preferred class of calcium-channel blockers, because non-dihydropyridines, non-dihydropyridines, such as verapamil and diltiazem, are known inhibitors of cytochrome P450 3A4 (CYP3A4).^36,46,48 Because VEGF RTKIs, such as sunitinib and sorafenib, are substrates for CYP3A4, inhibitors of this enzyme can substantially increase blood levels of these agents.¹⁶ Additionally, nifedipine may not be an optimal dihydropyridine for long-term blood pressure control, as it is know to induce VEGF. Although there is general consensus that angiogenesis inhibitors should be stopped if hypertension cannot be controlled using the standard oral dosing of two antihypertensive agents, there have been reports of substantial responses to long-acting oral nitrates in patients refractory to adequately dosed ACE inhibitors and calcium-channel blockers, who are continuing on anti-VEGF treatment.⁵² Additionally, the infrequently used blocker, nebivolol, which has a haemodynamic profile combining -adrenergic blocking activity with a vasodilating effect, mediated by the endothelial L-arginine nitric-oxide pathway, might prove effective.^49,53 Permanent discontinuation of angiogenesis inhibitors is recommended in patients who have hypertensive crisis.⁵⁰ The use of angiogenesis inhibitors should also be stopped in the less than 1% population of patients with bevacizumab-associated, suntinib-associated, or sorafenib-associated reversible posterior leucoencephalopathy (RPLS; webappendix p 3).⁵⁴ Clinical experience suggests that at least sunitinib therapy can be resumed once the syndrome resolves and blood pressure is adequately controlled.¹⁶ Lastly, although there is limited evidence that prophylactic antihypertensive therapy does not reduce dose interruptions of angiogenesis inhibitors, more research is needed in this area.⁵⁵

Proteinuria

Proteinuria in response to bevacizumab and aflibercept treatment can occur as a result of interference with VEGF-dependent glomerular endothelial integrity.⁵⁶ Bevacizumab-related proteinuria can also, in part, stem from thrombotic microangiopathy. Glomerular capillary endothelial injury by this process can manifest as protein-uria and could represent one permutation of vascular thromboembolic disease particular to bevacizumab.⁵⁷ In clinical trials of bevacizumab in ovarian cancer, grades 1 or 2 and 3 or 4 proteinuria occurred in 3–40% and 1–4% of patients, respectively.^{3–7,9–12,37–43} The incidence of low-grade proteinuria is similar with bevacizumab alone and in combination with cytotoxic agents and erlotinib.^{3–7,9–12,37–43} A review¹⁵ of toxicity data from aflibercept trials showed grade 2 proteinuria in 17% and grade 3 or 4 proteinuria in 4–13% of patients, with aflibercept monotherapy associated with the lowest incidence (4%) of grade 3 or 4 proteinuria. Proteinuria with angiogenesis inhibitors other than bevacizumab and aflibercept is rare. The proteinuria is predominantly asymptomatic and detectable only by laboratory analysis. Thus, monitoring by use of regular dipstick urinalyses should be considered in patients being treated with bevacizumab or aflibercept. Those with a dipstick reading of 2 or more should undergo a 24-h urine total-protein collection. Bevacizumab and aflibercept should be interrupted in patients excreting at least 2 g of protein in a 24-h period. However, treatment could be reinstituted after recovery in those gaining benefit from therapy, if the patient does not progress to nephrotic syndrome and recovery occurs within 3 weeks.⁵⁰ No standardised recommendations exist for the pharmacological management of patients who develop drug-induced proteinuria. However, on the basis of extensive clinical experience with diabetic-induced nephropathy and kidney disease, anti-angiotensin agents could be considered as first-line agents, provided specific criteria are met, including serum creatinine levels less than 176·8 mmol/L, normal potassium levels, and absence of renal artery stenosis.³⁶

Cardiac toxicity

Retrospective studies show that cardiovascular disease is frequently present at the time of cancer diagnosis and can have a substantial effect on long-term survival after completion of therapy.⁵⁸ The high prevalence of underlying cardiac disease is an important consideration at initiation of, and during, antiangiogenic therapy given the spectrum of reported acquired cardiac abnormalities. Left-ventricular dysfunction from a targeted agent was first reported for trastuzumab, with an incidence ranging from 4–7% when trastuzumab was used as a single agent to 27% when it was administered concurrently with anthracycline-based chemotherapy.⁵⁹ By contrast, cardiac toxicity associated with other antiangiogenic agents, including bevacizumab, is not well-characterised, because clinical trials have not included predefined cardiac endpoints. Congestive heart failure (CHF; defined as grade 2–4 left-ventricular dysfunction) was identified as a rare, but serious toxic effect of bevacizumab in a phase 3 metastatic breast cancer trial of weekly paclitaxel with or without bevacizumab. Grade 3 or 4 CHF transpired in three of 365 patients (0·8%)treated with bevacizumab versus one of 346 (0·3%) in women who received weekly paclitaxel alone.⁸ Cardiac toxicity has been more widely recognised with sunitinib and sorafenib. With sunitinib, decrements in left-ventricular ejection fraction (LVEF) of 10% or more have been noted in up to 28% of treated patients, but only 3–8% actually developed class III–IV New York Heart Association (NYHA) heart failure.⁵⁸ Previous anthracycline exposure and coronary artery disease seem to be major independent risk factors.⁶⁰ However, certain features of sunitinib-induced cardiac dysfunction suggest that it is linked to potentially reversible myocardial mitochondrial dysfunction and energy depletion, distinguishing it from anthracycline-mediated cardiac toxicity due to irreversible free-radical cardiomyocyte damage.^61–63 The mean time to onset of CHF due to sunitinib is highly variable, ranging from 22 days to 27 weeks after introduction of the drug, with varying degrees of recovery.^60,61,64 Reductions in LVEF of 10% or more are less common with sorafenib treatment, occurring in about 5% of patients. However, a precise understanding of which patients will progress to heart failure is absent.⁵⁸ Electrophysiological abnormalities, such as QTc prolongation, have been noted with some agents, such as fosbretabulin and DMXAA. Although QTc prolongation associated with these agents is typically moderate and transient, prolongation of cardiac repolarisation has the potential to cause serious ventricular arrhythmias, in particular torsades de pointes.^29,65

In view of the fact that oncologic disease progression can mimic cardiac toxicity by causing symptoms, such as a decline in performance status, dyspnoea, or chest discomfort related to pleural or pericardial effusions, it is important to maintain a high index of suspicion so that symptoms are not misattributed to malignancy alone. For this reason, baseline and periodic assessment of LVEF might be prudent in patients treated with anti angiogenic agents. Serial monitoring of B-type natriuretic peptide or troponin, which might serve as biomarkers of myocardial injury, is under investigation.⁶⁰ Discontinuation of antiangiogenic agents is recommended if CHF and arrhythmias precipitated by QTc prolongation develop. Circumspect resumption of therapy at a reduced dose is a consideration in patients without clinical evidence of CHF, but with a decrease of LVEF by an absolute value of 20% and to below the lower limit of normal (50%) provided careful blood pressure and cardiac monitoring are done. Thyroid function should also be assessed, because hypothyroidism is a rare cause of cardiomyopathy and an increasingly recognised toxic effect associated with sunitinib.⁶⁶ Avoidance of electrolyte disarray, in particular potassium and magnesium depletion, is key in keeping the risk of torsades with QTc prolongation to a minimum. ACE inhibitors and -blockers are the core pharmacotherapy for left-ventricular dysfunction.⁶⁷ Drugs, such as carvedilol and simvastatin, which have been shown to have cardio-protective effects, might also prove useful. Additionally, in view of the heightened risk for cardiac ischaemia due to microembolic arterial events with angiogenesis inhibitors, concomitant treatment with low-dose aspirin (81 mg daily) might also be beneficial.⁶⁴ Lastly, new strategies for structural reengineering of toxic drugs, such as imatinib, might help to improve their therapeutic index.⁶⁸

Vascular thromboembolism

Although the background incidence of venous thromboembolic events (VTE) is up to 16% in patients with ovarian carcinoma, reports of arterial thromboembolic events (ATE) are rare.⁶⁹ However, there is growing evidence that an increased risk of ATE is one of the more serious toxic effects associated with bevacizumab therapy. In a recent analysis of pooled data of 1745 patients with metastatic colorectal cancer, non-small-cell lung cancer (NSCLC), or breast cancer from five randomised trials, the addition of bevacizumab to chemotherapy was associated with an increased risk of ATE (overall incidence 3·8% with bevacizumab vs 1·7% in the chemotherapy-alone group), but not VTE. Most ATE episodes were myocardial or cerebrovascular events, but peripheral vascular and mesenteric clots have been reported. On multivariate analysis, only exposure to bevacizumab, age 65 years or greater, and a previous history of ATE were identified as significant risk factors.⁷⁰ Up to now, a total of 30 vascular thromboembolic events with two related deaths (one attributed to cerebrovascular ischaemia and another to a right-ventricular thrombus) have been reported in nearly 600 patients with ovarian cancer treated with bevacizumab (table 2).

Table 2.

Vascular thromboembolic events in patients with ovarian cancer treated with bevacizumab

	Number of events/numbers of patients
Myocardial ischaemiaor infarction1,10	4/114
Cerebrovascular ischaemia or infarction1,8,10,11	4/149
Ventricular thrombus10	1/70
Pulmonary embolus2,5,7	3/154
Deep-venous thrombosis1,2,6,10	5/194
Vascular-access-related venous clot6,10	2/88
Non-specified vascular thromboembolic events8,5,10,11,37,71	11/252

Accumulating data suggest that the incidence of VTEs might be less with aflibercept compared with bevacizumab. Only two VTEs were reported in toxicity data from five phase 1 or 2 trials of aflibercept.¹⁵ A hypothetical explanation for this difference is that anti-VEGF antibodies, such as bevacizumab, are more prone to form multimeric complexes, lodge in the vasculature, and serve as nidus for clot formation.⁷² The incidence of thromboembolic complications is also low for sorafenib and sunitinib, although a few cases of ATE, especially cardiac ischaemia and myocardial infarction have been reported. In the two sorafenib registration trials, the incidence of treatment-emergent cardiac ischaemia or infarction was higher in patients receiving sorafenib (2·9%) compared with those receiving placebo (0·4%).⁵⁸ Among the other VEGF RTKIs being assessed in ovarian cancer, isolated reports of vandetanib-related pulmonary emboli, deep-venous thrombosis (DVT), and intestinal ischaemia are the most noteworthy.²⁷ The relative risk of VTEs can also be increased over baseline with anti-integrin antibodies.^73,74 ATEs in the form of cardiac and intestinal ischaemia have been noted with fosbretabulin and the findings of several phase 1 studies will probably aid in further defining the risk of ATE with this agent.^30,31,65

Considering the already increased risk of thrombosis in patients with cancer and the high morbidity and mortality associated with this complication, there is growing momentum for instituting prophylaxis in patients treated with targeted therapies. Prophylaxis options include low-molecular-weight heparin (LMWH; eg, enoxaparin 40 mg daily), warfarin, or aspirin (81–200 mg daily). Information pertaining to the relative efficacy of each of these options in clot prevention has largely been supplied by studies assessing thalidomide with or without chemotherapy, because of the high incidence of vascular thromboembolism in patients with multiple myeloma treated with these regimens.⁷⁵ Thromboprophylaxis should be carefully balanced with the paradoxical increased risk of bleeding in patients treated with angiogenesis inhibitors. Fortunately, patients treated with aspirin (≤325 mg daily) for arterial thromboprophylaxis, while receiving bevacizumab, did not have more haemorrhagic events.⁷⁶ In a subgroup of patients treated with bevacizumab and concomitant full-dose warfarin for thrombotic events, bleeding complications did not seem to be significantly increased.^75,77 However, a large randomised trial showed that full-dose warfarin is associated with a higher bleeding rate than LMWH (1·5 mg/kg daily) in patients with venous thromboembolism and cancer, suggesting that LMWH might be a safer and equally as effective option.⁷⁸

In the absence of consensus guidelines for managing patients with thromboembolic disease on therapy, treatment plans from clinical trials assessing bevacizumab in ovarian cancer provide a useful framework. Patients who have grade 3 or 4 ATEs, including cerebrovascular ischaemia, transient ischaemic attack, cardiac ischaemia or infarction, and peripheral or visceral ischaemia, should be taken off study. Treatment should also be permanently discontinued in cases of new or worsening grade 2 ATEs since beginning therapy. In the event of VTEs, treatment can be temporarily held for grade 3 or asymptomatic grade 4 incidents. If the planned duration of full-dose anti coagulation is 3 weeks or under, treatment should be held until the period of full-dose anticoagulation is over. If the planned duration of full-dose anticoagulation exceeds 3 weeks, treatment can be resumed during the period of full-dose anticoagulation, provided that all of the following criteria are met: biochemical testing consistently within therapeutic range on a stable dose of anticoagulant; absence of pathological conditions that carry high risk of bleeding (eg, tumours involving major blood vessels); no history of haemorrhagic events while on study; and evidence of benefit from treatment (no evidence of disease progression).

Haemorrhage

A tendency for increased bleeding with VEGF antagonism has been attributed to decreased renewal capacity of endothelial cells.⁷⁹ Serious bleeding complications have been reported in patients with renal-cell carcinoma (RCC) and NSCLC treated with anti-VEGF therapy. For example, life-threatening episodes of haemoptysis occurred in 2–9% of patients enrolled in NSCLC trials. Centrally located tumours located near great vessels, cavitated, and presenting with squamous histology are the most at risk.^54,80 Although bleeding events are often reported in studies with anti-VEGF therapy in ovarian cancer (eg, <10% to 90% with bevacizumab), most are mild and most commonly present as grade 1 or 2 mucosal surface bleeding, such as epistaxsis, gingival bleeding, and haematuria. Rare bevacizumab-related grade 3 gastrointestinal haemorrhage has been reported.^5,10,38,71 In a phase 2 trial assessing aflibercept monotherapy in 162 patients with recurrent, platinum-resistant, ovarian cancer, bleeding was not a significant adverse event.¹⁴ However, in a phase 1 study of aflibercept plus oxaliplatin, folinic acid, and fluorouracil (FOLFOX4) in 32 patients with solid tumours, 9% of patients had grade 3 or worse haemorrhagic events, including one fatal haemorrhagic stroke.⁸¹ Fatal intracerebral haemorrhage has been especially concerning in patients with brain metastases treated with sorafenib and sunitinib. Five of 67 patients (7%) with metastatic RCC treated with sorafenib or sunitinib died of intracerebral haemorrhage during therapy, four with known brain metastases and two with a history of hypertension.⁸² Despite the infrequency of CNS metastases in some tumour types, including ovarian cancer (0·5–2·2% of cases), the threat of intracerebral haemorrhage seems to exist even in the absence of lesions. For example, two grade 4 cerebral haemorrhages in patients without evidence of brain metastases occurred in phase 1 or 2 studies of cediranib.^23,83 Overall, the incidence of haemorrhagic events in patients other than those with NSCLC receiving VEGF RTKIs vandetanib, SU6668, and pazopanib seems to be similar to the incidence in those receiving placebo. This also seems to be the case with imatinib, EGFR-targeted agents, fosbretabulin, and DMXAA.

Management of bleeding events associated with angiogenesis inhibitors largely centres around treatment interruption and administration of standard supportive care. Most clinicians recommend that treatment with angiogenesis inhibitors be suspended at least 30 days before and after surgery. Caution should also be exercised in patients with congenital bleeding diathesis, acquired coagulopathy, and those receiving full-dose anticoagulation for the treatment of thromboembolism.⁵¹ Initially, few data existed and there was much concern about the safety of using angiogenesis inhibitors in patients with CNS metastases. However, experience over the past 5 years suggests it is at least safe to administer bevacizumab to patients with CNS metastases who have been primarily treated with surgery or radiotherapy.³⁶ Data leading to recent accelerated US Food and Drug Administration (FDA) approval of bevacizumab for recurrent malignant gliomas further support the safety of bevacizumab in patients with large CNS disease burden.

Gastrointestinal toxicity

Arguably one of the most worrisome complications of antiangiogenic agents can be intestinal perforation. A connection between intestinal perforation and bevacizumab was first made in trials involving patients with colorectal cancer. Intestinal perforation has been noted in 1–4% of patients receiving bevacizumab in combination with fluorouracil-based chemotherapy and is thought to be more prevalent in those with acute diverticulitis, intra-abdominal abscess, gastrointestinal obstruction, tumour at perforation site, abdominal carcinomatosis, and previous abdominal or pelvic radiotherapy.^84,85 Use of aspirin and non-steroidal antiflammatory drugs, colonoscopy or sigmoidoscopy done within 1 month of starting bevacizumab therapy, and partial or complete response to therapy are also potential risk factors.^46,86 Patients with recurrent ovarian cancer have exceptional risk factors, including previous gastrointestinal surgery, carcinomatosis compromising overall bowel function, intermittent or chronic bowel obstruction, and poor nutrition. Perhaps the highest rates of perforation (0–15%) have occurred in patients with ovarian cancer who were receiving bevacizumab. In an industry-sponsored trial, five patients (11%) had intestinal perforation within 51–178 days of starting bevacizumab, one of which was fatal. All had radiographic evidence of bowel involvement at study entry and a trend towards increased frequency of intestinal perforation was noted for bowel-wall thickening and bowel obstruction. Exposure to three rather than two previous chemotherapeutic regimens was most strongly associated with intestinal perforation.¹ Other studies of single-agent bevacizumab for recurrent ovarian cancer have not shown high rates of intestinal perforation. For example, there were no perforations in a Gynecologic Oncology Group phase 2 study (GOG 170-D)³ and only one enterocutaneous fistula in a separate retrospective study describing bevacizumab use in 32 patients who had failed a median of five previous chemotherapy regimens.^2,3 Similarly, aflibercept monotherapy carries an increased risk of intestinal perforation in women with recurrent ovarian cancer.^14,87 The hazard of intestinal perforation in women receiving bevacizumab with chemotherapy seems to be comparable to monotherapy.^5,10 Sfakianos and colleagues⁸⁸ retrospectively compared the incidence of perforation in patients with recurrent ovarian cancer receiving chemotherapy with or without bevacizumab. The relative risk for developing a perforation or fistula in patients receiving bevacizumab was 1·09 (95% CI 0·40–2·96), suggesting that bevacizumab might not increase the incidence of perforation when added to salvage chemotherapy. However, combining antiangiogenic agents might pose a greater risk. For example, four perforations, two of which were fatal, occurred in 26 patients treated with combination bevacizumab and erlotinib.^12,39 Furthermore, of 39 patients with refractory solid tumours treated with low-dose bevacizumab and sorafenib, two of 13 patients with ovarian cancer developed either an enterocutaneous or an enterovaginal fistula 2 weeks into therapy (table 3).¹³ There have been no reported cases of bowel perforation in patients with ovarian cancer treated with either VEGF RTKIs or EGFR-targeted agents as monotherapy. Likewise, there are no reports of perforation associated with cases of intestinal ischaemia from vascular damaging agents, such as fosbretabulin.⁴⁴

Table 3.

Incidence of intestinal perforation in patients with ovarian cancer treated with bevacizumab or aflibercept

	Treatment regimen	Number of events/number of patients	Previous chemotherapy regimens, n	Comments
Monk3	Single-agent bevacizumab for platinum-refractory disease	1/32	Median 5 (range 2–10)	One enterocutaneous fistual in a patient who had seven previous debulking surgeries
Burger2	Single-agent bevacizumab for recurrent disease	0/62	1 (for 21 patients) and 2 (for 41 patients)	··
Cannistra1	Single-agent bevacizumab for platinum-resistant disease	5/44	2 (for 23 patients) and 3 (for 21 patients)	All had bowel involvement at study entry, had received three previous chemotherapy regimens, and had stable disease
Campos5	Carboplatin/paclitaxel plus bevacizumab as first-line therapy	2/62	NA	45 of 62 patients had optimally cytoreduced tumours
Micha6	Carboplatin/paclitaxel plus bevacizumab as first-line therapy	0/20	NA	17 of 20 patients had optimally cytoreduced tumours
Penson7	Carboplatin/paclitaxel plus bevacizumab as first-line therapy	0/30	NA	··
Herzog40	Oxaliplatin/docetaxel plus bevacizumab as first-line therapy	1/59	NA	One colonic fistula
Wright37	Bevacizumab plus cytotoxic agents for recurrent disease	4/62	Median 5 (range 1–15); 8·5 in patients with gastrointestinal perforation	Bowel obstruction in six patients, only one of whom developed a gastrointestinal perforation
Sfakianos88	Bevacizumab plus cytotoxic agents for recurrent disease	5/68	Mean 5	Three of five patients with a history of obstruction had a gastrointestinal perforation
Richardson11	Gemcitabine/carboplatin plus bevacizumab for recurrent disease	3/35	Median 2 (range 1–11)	Two gastrointestinal perforations with one at an anastomotic site created >1 year previously and one associated with a perforated diverticular abscess in patients with complete response who had received one to two previous regimens, one enterovaginal fistula
Cohn41	Bevacizumab plus weekly taxane for platinum-refractory disease	0/10	Media 4 (range 3–8)	Eight patients with carcinomatosis when bevacizuamb initiated
Garcia10	Bevacizumab plus cyclophosphamide for recurrent disease	4/70	Median 2 (range 1–3)	Three gastrointestinal perforations and one fistula, bevacizumab resumed in one patient without further gastrointestinal complications
Chura38	Bevacizumab plus cyclophosphamide for recurrent disease	0/15	Median 8 (range 5–15)	All had intra-abdominal disease, but no bowel obstruction at initiation
Nimeiri12	Bevacizumab plus erlotinib for recurrent disease	2/13	Median 2 (range 1–3)	Two fatal gastrointestinal perforations in patients who had two previous regimens and pre-existing evidence of small-bowel involvement
Friberg39	Bevacizumab plus erlotinib for recurrent disease	2/13	Median 2 (range 1–3)	Both had two previous regimens and known small-bowel obstruction
Azad13	Bevacizumab plus sorafenib for recurrent disease	2/13	NA	One enterocutaneous fistula and one enterovaginal fistula in patients with rapid tumour regression
Tew14	Aflibercept for recurrent disease	2/162	Range 2–3	··
Colombo87	Aflibercept for recurrent disease and ascites	1/10	Range 2–9	··

NA=not applicable.

Although the mechanisms underlying bowel perforation with bevacizumab or aflibercept are not fully understood, a possible explanation could be bowel-wall weakening in tumour-containing segments.⁶² Risk reduction might be achievable with careful patient selection, such as limiting bevacizumab treatment to patients without clinical symptoms of bowel obstruction, evidence of rectosigmoid involvement on pelvic examination, and bowel involvement on CT.⁸⁹

Maintaining a high index of suspicion in patients presenting with fever, leucocytosis, or abdominal pain is also warranted. Some VEGF RTKIs and vascular-damaging agents are known to induce varying degrees of abdominal discomfort or pain, potentially signifying other serious underlying medical conditions. For example, acute hepatic toxicity and pancreatitis have been reported with imatinib and sorafenib, respectively.^90,91 In the case of vascular-damaging agents, treatment-related tumour pain is a commonly reported side-effect.^31,32 Laboratory and imaging studies are useful diagnostic procedures. Radiographic signs of perforation on plain abdominal films and contrast-enhanced CT include free peritoneal air, extraluminal contrast, fistula, and abscess formation (webappendix p 4). Pneumatosis intestinalis, subserosal and submucosal gas-filled cysts in the gastrointestinal tract, might also be a variant of bevacizumab-related perforation.⁴²

Management of bowel complications after anti-VEGF therapy presents a dilemma, as a result of the increased likelihood of surgical and postoperative complications, such as bleeding, thrombosis, and compromised wound healing. Therefore, operative intervention versus conservative management should be carefully considered. Limited data suggest that initial management could consist of intravenous antibiotics, bowel rest with nasogastric tube placement, and percutaneous intraperitoneal catheter placement by an interventional radiologist. In a retrospective review of 20 patients managed non-operatively, only one patient ultimately required colonic resection and colostomy for extensive fistula and abscess formation in the setting of perforated diverticulitis. The 60-day post-perforation mortality rate was 25% (six of 24 patients). However, four of these patients were discharged with resolution of their acute symptoms before dying of tumour progression.⁹² Thus, a non-operative treatment approach might be preferable in select cases of bowel perforation, because it might spare patients with limited life expectancy from surgical morbidities.⁹² An experienced surgeon should decide on the benefit of surgery.

Increased risk of wound-healing complications in patients actively receiving angiogenesis inhibitors undergoing surgery is an important consideration when deliberating operative intervention. Scappaticci and colleagues⁹³ showed an increased risk of wound-healing complications in patients undergoing surgery within a 28–60 day window of receiving bevacizumab (6·7–10% with chemotherapy and bevacizumab vs 0% with chemotherapy alone).⁹³ On the basis of these data and the half-life of bevacizumab (21 days), many researchers recommend that the gap between bevacizumab discontinuation and major surgery be at least 30 days to keep the risk of surgical wound or bowel anastomotic healing complications to a minimum.⁵⁰ Patients who underwent bowel resection with primary reanastomosis as part of a tumour-reductive procedure are candidates for antiangiogenic therapy provided 30 days have elapsed from the time of surgery and they have recovered normal bowel function. In the event of surgical intervention for bowel perforation, a stoma might be favourable to primary anastomosis, in view of the negative effect of angiogenesis inhibitors on wound healing and increased risk of anastomotic breakdown.⁹⁴ Neither bevacizumab nor aflibercept seem to negatively affect wound healing after minor surgical procedures, such as vascular access device implantation.⁹⁵ Continuation of angiogenesis inhibitors is largely contraindicated in patients who develop treatment-related bowel perforations.⁸⁵

Gastrointestinal toxicity in the form of diarrhoea and mucositis is a common dose-limiting side-effect of several VEGF RTKIs. This finding is presumably related to their oral administration and interference with signalling pathways involved in sustaining mucosal health and repair. On average, 50% of patients on sorafenib or sunitinib report diarrhoea of any grade. 2 to 6% of patients have grade 3 or 4 diarrhoea (increase of ≥7 stools per day over baseline, hospitalisation, IV fluids >24 h, interference with activities of daily living, or life-threatening consequences, such as haemodynamic collapse). Although the incidence of grade 3 or 4 diarrhoea is comparable to single-agent imatinib and erlotinib, higher rates have been noted with cediranib (11%), pazopanib (13%), and BIBF-1120 (16%).⁹⁶ Oral hydration and antidiarrhoeal agents, such as loperamide, are the cornerstones of treatment for grade 1 or 2 diarrhoea and treatment should be interrupted until grade 3 or 4 diarrhoea resolves to grade 1 or below. Additionally, patients treated with sorafenib and sunitinib often complain of oesophagitis or gastritis, characterised by odynophagia and dyspepsia unaccompanied by mucosal changes at endoscopy. Empiric acid suppression with a proton-pump inhibitor seems to best counteract this side-effect.⁴⁹

Dermatological toxicity

The term hand–foot skin reaction (HFSR) or acral erythema refers to a group of signs and symptoms that bilaterally affect the hands or feet of patients receiving sorafenib and sunitinib. This problem is a relatively pathognomonic dermatological toxicity associated with sorafenib and sunitinib, because bevacizumab, aflibercept, and most VEGF RTKIs are not known to trigger this side-effect. HFSR has been reported with single-agent sorafenib treatment in 25–50% of patients.^19,97 Nearly one in five patients treated with sunitinib also develop HFSR. Common symptoms of HFSR include dysaesthesia and paraesthesia (which can occur before other symptoms are apparent), erythema, oedema, hyperkeratosis, dry or cracked skin, callous-like blisters (which usually do not contain fluid), and desquamation.⁴⁹ HFSR seems distinct from the palmar-plantar erythrodysaesthaesia often seen in patients receiving cytotoxic chemotherapy, such as liposomal doxorubicin and capecitabine. Although pressure areas are usually the most problematic, non-pressure-bearing areas, such as the lateral sides of fingers and periungual zones, can also be affected in patients exposed to sorafenib and sunitinib.⁹⁸ Symptoms typically appear during the first 6 weeks of treatment with these agents, and, although they frequently decrease in intensity during the course of therapy, prompt intervention is advised, because early symptoms often resolve quickly with minimum effort, allowing continuation of therapy without dose reduction. Pharmacologic interventions such as clobetasol cream, systemic corticosteroids, vitamin E, topical 99% dimethyl-sulfoxide, or pyridoxine in doses up to 400 mg, administered twice daily have also proved successful. Rapid symptom improvement can be achieved with temporary cessation of therapy, allowing reinstitution of the drug within 3–14 days.^46,49,99 Although some patients will remain asymptomatic for the remaining duration of therapy, others will repeatedly relapse despite dose reduction. With the third episode of grade 3 symptoms, permanent discontinuation of therapy is recommended.¹⁶ More recent studies have been identifying predictors of sorafenib-associated HFSR. Increasing cumulative sorafenib dose and combination therapy with other anti-VEGF agents, such as bevacizumab, seem to intensify the risk of sorafenib-associated HFSR. Occurrence of sorafenib-associated HFSR also strongly correlates with the development of, or worsened pre-existing, hypertension. Thus, a high index of suspicion for sorafenib-associated HFSR should be maintained in these instances.⁹⁹

Search strategy and selection criteria.

Information for this review was compiled from the PubMed and Medline databases using the search terms: “ovarian cancer”, “angiogenesis”, “antiangiogenesis”, “antivascular”, “molecular-targeted therapy”, “toxicity”, “adverse event”, “hypertension”, “cardiac,” “thrombosis”, “perforation”, “dermatological”, and “thyroid.” Papers published up to May, 2009, were included and no language restrictions were applied. Information regarding trials for antiangiogenic and targeted agents was obtained from: http://www.clinicaltrials.gov and http://www.nci.nih.gov/clinicaltrials. Relevant published abstracts from scientific oncological meetings were also considered.

Other sunitinib and sorafenib skin rashes, including erythema and maculopapular or seborrhic dermatitis, tend to decrease over time and rarely require dose reduction. Case reports of leucocytoclastic vasculitis, presenting as palpable, painful purpura and petichiae, also exist.^100,101 Subungual splinter haemorrhages can be another subtle finding in patients taking sorafenib and can represent thrombotic or embolic phenomena. Hair depigmentation can also follow treatment with sunitinib or other targeted agents that inhibit Ret or c-Kit. In the case of sunitinib, hair depigmentation generally appears after 5–6 weeks of treatment, but reverses as early as 2–3 weeks after treatment discontinuation.⁵³

Dermatological toxicity in the form of mild to moderately severe aseptic acneiform skin rash or folliculitis can appear in 50–100% of patients receiving EGFR-targeted agents (figure). The rash is characterised by erythematous follicular papules and pustules appearing in areas rich in sebaceous glands, such as the face, upper chest, and back. The abdomen, palms, and soles can also be involved. The rash usually appears several days after the start of therapy and is more intense during weeks 2–3 of anti-EGFR treatment. Management options include topical antiseptics or agents with anti-inflammatory properties. Colloidal oatmel and oral doxycycline treatment have also been used succussfully. Topical steroids should be avoided.

Figure.

Diffuse erythematous dermatologic rash on the posterior lower extremities in a patient with ovarian cancer treated with anti-epidermal growth factor receptor therapy

Other skin toxicities noted with anti-EGFR therapy include nail fragility, hair changes, such as hypertrichosis, and xerosis. Nail toxicity appears in 10–15% of patients and is characterised by a paronychial inflammation with painful fissures in the nail folds. Hair changes are hallmarked by trichomegaly, with rigid and longer eyelashes and more brittle, slow-growing scalp hair.⁵⁴ Interestingly, studies have shown that patients on cetuximab and erlotinib with any grade of rash have better median overall survival (months)¹⁰² than patients without a skin rash, and that patients with grade 3 rash benefit from the longest survival.¹⁰³ Studies have also reported a link between response to sorafenib and cutaneous adverse reactions.¹⁰⁴ This links the possibility that these rashes could serve as a surrogate marker of effective target inhibition and antitumour activity.

Endocrine toxicities

Thyroid dysfunction is an increasingly recognisable side-effect of sunitinib treatment. In the first published series of patients with gastrointestinal stromal tumours, 15 of 42 (36%) patients developed hypothyroidism after an average of 50 weeks of sunitinib therapy. Rini and colleagues¹⁰⁵ documented abnormal thyroid function testing in 85% of patients with RCC on sunitinib. Thus, regular surveillance of thyroid function with thyroid-stimulating hormone measurements at baseline and every 2–3 months in patients on sunitinib is recommended and thyroid hormone replacement should be pursued in those with hypothyroidism.^66,105,106

Hypoglycaemia is another important endocrine-related side-effect of sunitinib. In an analysis of 19 diabetic patients with metastatic RCC, who were treated with sunitinib, all had a decrease in blood glucose levels after 1 month of therapy and two patients were able to stop their antihyper-glycaemic treatment during the active treatment phase of each cycle. Thus, glycaemic control should be carefully monitored in diabetic patients treated with sunitinib.¹⁰⁷

Conclusion

Expanding the scope of treatment for ovarian and other cancers to include targeted therapies poses new challenges, including the threat of untoward, often unordinary, side-effects. In view of the fact that antiangiogenic agents dominate the current portfolio of targeted biological agents, many of the best characterised toxic effects at this point represent perturbations in vascular functioning. In this case, and others, some adverse effects of targeted agents can be so precisely linked to the biological activity of the target that emergent toxic effects might prove to be a sensitive gauge of therapeutic response. Thus, thoughtful consideration of the unique toxic effects related to these agents can aid with the discovery of much needed clinico-biological markers for predicting and monitoring response. Furthermore, the success of many targeted agents under assessment might lie in disease stabilisation, rather than disease eradication. Given that therapeutic strategies aimed at controlling cancer are likely to require frequent and prolonged drug administration, the matter of toxicity is bound to become more complex.¹⁰⁸ Currently, there is paucity of management strategies based on strong scientific evidence. Existing approaches are predominantly derived from level 4 and 5 evidence, because toxicity management has not been an endpoint in most studies done up to now. Thus, there is a fundamental need for investigations that will generate more evidence-based practice guidelines. Ultimately, vigilance and appropriate management by oncologists and other collaborating specialists is essential for obviating the high morbidity and potential mortality of toxic effects stemming from targeted and antiangiogenic agents.