Erratum: Ocular toxicities associated with targeted anticancer agents: an analysis of clinical data with management suggestions

[This corrects the article DOI: 10.18632/oncotarget.17634.].


INTRODUCTION
Targeted chemotherapy agents are becoming increasingly important in the clinical management of cancer. Of 42 novel oncologic drugs approved since 2008, 30 were either antibodies or kinase inhibitors targeting specific receptors or unique intracellular signal transduction pathways [1]. This number of approved oncologic drugs is significantly higher than the 15 agents approved between 2000 and 2008. Whereas the toxicity profiles of traditional oncologic agents are known and relatively well-described, the toxicity profiles of targeted therapy agents are not as well-known and include adverse sight-threatening events [2][3][4]. Ocular toxicities are among the most common adverse events associated with targeted agents [2][3][4][5][6][7]. This high frequency of ocular adverse events can be partially attributed to the delicate homeostatic environment of growth factors, cell receptors, and vascular formation in the eye, a unique microenvironment that is disrupted by many targeted agents [4,[8][9][10][11][12][13]. Currently, there is a paucity of data documenting particular ocular toxicities of targeted agents [2,4,6]. The purpose of this study is to provide a better understanding of the toxicities that have been observed in various targeted therapies and to provide recommendations for screening, surveillance, and management of these events.

Drug selection and FDA label review
Of the 138 agents that were screened for inclusion in the study, 34 were palliative or non-anticancer agents, 12 were duplicates that were approved for separate indications, and 46 were cytotoxic, non-targeted, or conjugated agents and were therefore excluded; the remaining 46 agents were reviewed and screened for their association with ocular toxicities, as described in the Methods section.
FDA labels were reviewed for mention of ocular toxicities, from which 20 agents were associated with some form of ocular toxicity. Three agents (bortezomib, pertuzumab, and dabrafenib) were associated with minor ocular adverse events according to the FDA label, but limited evidence of ocular toxicity was evident upon an independent survey of the literature. Multiple rigorous case reports have associated bortezomib with eyelid chalazia [14][15][16] and dabrafenib with uveitis and cystoid macula edema [17], however, due to lack of quantified trial data, these agents were excluded from Table 2. Similarly, two agents (idelalisib and ibrutinib) were found to have adverse events from our literature review, but the FDA labels included no mention of ocular toxicity; these agents were excluded from Table  1 but were included in subsequent analysis. A total of 16 agents, including 12 small-molecule drugs and 4 monoclonal antibodies, were analyzed in this study for ocular toxicity profiles based on evidence from FDA labels and clinical trials. Therefore, of the original 46 targeted medications, 18 of 30 small-molecule drugs (60%) and 12 of 16 monoclonal antibodies (75%) were not associated with ocular toxicity and were thus excluded. A summary of ocular toxicities from the FDA label review is included in Table 1.

Ocular adverse events
We conducted a review of 217 independent studies, 32 of which met inclusion criteria. The results and ocular adverse events described in these studies are included in Table 2.
Ocular events were scored according to severity and CTCAE grade, as detailed in Table 3. Severe ocular adverse events that were not given a grade were also reported. The most common severe adverse event included severe conjunctivitis, associated with 9 (19.6%) of the total targeted agents, and blurred vision, connected with 10 agents (21.7%). Imatinib had the highest incidence of grade 3 or higher events among all targeted agents, with 3% of patients experiencing grade 3 or higher periorbital edema. Imatinib and crizotinib had the highest incidence of ocular toxicity overall, with 70% and 62-64% of patients experiencing some form of ocular toxicity per FDA statistics and an independent review of available data (Tables 1 and 2). Acute vision-threatening events (including retinal vascular occlusion, retinal pigment epithelial detachment, corneal membrane ulceration and perforation, and blindness) were rare, almost always occurring <1% of the time in their respective drug classes. Only 5 drugs (10.9%) were associated with these vision-threatening events, namely erlotinib, gefitinib, trametinib, vemurafenib, and ipilimumab.
The most severe ocular adverse events occurred with EGFR, MEK and CTLA-4 inhibitors and targeted antibodies. These events included corneal perforation and retinal vascular occlusion. Retinal vascular occlusion had an incidence of 0.8%. Corneal perforation was described primarily in case reports and case studies; however, 3 patients in a large phase 1 trial of gefitinib with a cohort of 221 patients experience grades 1-3 corneal erosion and corneal defects. Use of ipilimumab resulted in the most severe adverse event, causing blindness in 2 out of 393 patients [18].
Small-molecule drugs appear to have a higher incidence of ocular toxicity than do monoclonal antibodies. The percentage of small-molecule drugs associated with ocular toxicity (37.5%) was higher than that associated with monoclonal antibodies (28.6%). Even within the same class of action, monoclonal antibodies resulted in fewer adverse events than did their smallmolecule counterparts.

Summary and management recommendations
A summary of ocular events and management recommendations are presented in Table 4. While there is a paucity of data on this subject, our recommendations are based on available data from clinical trials, case reports and series, clinical experience, expert opinion, and existing national guidelines. One of the central underlying themes in the management recommendations is the concept of establishing a pretreatment baseline to better gauge subsequent ocular adverse events. Screening suggestions were modeled after preferred practice patterns on the management of ocular toxicities of hydroxychloroquine, issued by the American Academy of Ophthalmology, [19] and of ethambutol, issued by the Hong Kong Ophthalmologic Society [20]. Guidelines were referenced according to frequency of screening in drugs that showed a similar incidence and severity of ocular toxicities. From this, specific screening parameters were formulated based on severity of adverse Dabrafenib (Tafinlar) [34] Uveitis/Iritis 1% n=586. Can cause cystoid macula edema.

DISCUSSION
Ocular toxicities are becoming increasingly relevant with the increased use of targeted agents. Sparse data exist with regard to specific recommendations for the screening and management of these toxicities. The goal of this study was to review data from FDA labels and from independent studies to determine which agents need ophthalmologic management and to provide recommendations for the screening and management of patients receiving these agents. Management recommendations have already been explored by various groups. The FDA labels include detailed recommendations for at least two agents, namely trametinib and ipilimumab. These recommendations are incorporated into this study. Van der Noll et al. presented a review of ocular toxicity with management recommendations and algorithms for serous retinal detachment and retinal veinous occlusion in 2013 [21]. Our study provides a comprehensive and agent-specific set of recommendations for the management and screening of known ocular toxicities.
Our data analysis showed that for the majority of agents, the most common types of ocular toxicity were low-grade in severity, primarily grade 1 or 2 according to the CTCAE [22]. Table 5 provides a detailed list of adverse events and the agents that cause them. The most common complications among all agents (both small-molecules agents and monoclonal antibodies) were conjunctivitis and "visual disturbances." However, the progression of even the most common toxicities was not consistent, and some toxicities were severe, as evidenced by at least two case reports in which rapid progression to blindness occurred with use of ipilimumab and crizotinib [18,23]. As such, the establishment of a visual baseline is important to assess the severity and progression of signs and symptoms that may arise throughout the course of management. We believe that the incidence of ocular toxicities in certain agents and the potential for severity is enough to merit ophthalmic referral when prescribing these agents. The most important suggestion based on these data is the establishment of an ophthalmic baseline for targeted anticancer drugs known to cause ocular toxicities.
Unlike their cytotoxic counterparts, targeted inhibitors tend to be highly specific for their molecular targets (more than 99% in most cases). As such, toxicities tend to be much more focal than systemic. Examples of this phenomenon can be seen in these data, in particular among the most severe adverse events, such as corneal ulceration, blindness, or retinal artery or vein occlusion ( Figure 1), which were associated with a minority of agents. EGFR inhibitors and MEK inhibitors as a class were associated with a high proportion of severe ocular toxicities. It is known that EGFR is intimately involved in angiogenesis and wound healing [4,24]; as such, it is not surprising that the most severe ocular toxicities associated with this class of agent included complications with delayed wound healing, such as corneal ulceration. Corneal microcysts ( Figure 2) have also been seen in some patients undergoing EGFR inhibitor therapy in Phase I trials. MEK inhibitors inhibit a key step in the MAPK signal transduction pathway, and an increasing body of evidence suggests that this pathway is involved in the maintenance and repair of the retina [4,12,13]. Trametinib, a MEK 1/2 inhibitor, is known to increase the risk of severe retinal issues (Figure 3), such as retinal detachment and retinal vascular occlusion.
Given the high degree of specificity that targeted agents exhibit, clinicians are better able to direct their screening of toxicity toward these specific regions of the eye that certain agents are known to affect. As such, mechanism-based screening guidelines may be appropriate for screening toxicity in targeted agents. An example taken from our management suggestions Therapy FDA Label Notes
Panitimumab (Vectibix) [45] Eye/eyelid irritation (1%), conjunctivitis (4%), ocular hyperemia (3%), increased lacrimation (2%), in 463 patients Pertuzumab (Perjeta) [46] Mentions increased lacrimation (No evidence of this in literature) Rituximab (Rituxan, Mabthera) [47] Mentions uveitis and optic neuritis is the targeted surveillance of the retina in patients undergoing MEK inhibitor therapy. For drugs with known retinal toxicities such as trametinib, we suggest self-screening by the patient with an Amsler grid for visual field monitoring. Identification of any of these ocular symptoms merits urgent referral for additional ophthalmologic assessment. For agents known to be associated with more severe ocular toxicities (e.g., gefitinib, trametinib, vemurafenib, and ipilimumab), we suggest routine ophthalmic surveillance and baseline assessment by an ophthalmologist. A review of symptoms such as eye pain, redness, and changes in vision should be obtained by the medical oncologist at each follow-up visit. a Studies did not report individual ocular adverse events. b Studies did not report total ocular adverse events. Many of the ocular toxicities displayed by targeted agents are closely related to the drug's mechanism of action, a sentiment commonly echoed in relevant scientific literature [2][3][4][5][6]. However, even among drugs that target the same molecule with high specificity, variations in toxicity exist. Of note, the use of monoclonal antibodies did not typically lead to as many ocular adverse events as small-molecule agents did, even among those that shared the same mechanism of action (e.g., gefitinib and cetuximab). Ocular toxicities among monoclonal antibodies seemed to be lower in both incidence and frequency. A possible reason for this is that monoclonal antibodies are less able to permeate physiological barriers [25,26], such as the blood-brain or blood-ocular barriers, and are therefore less able to interfere directly with the delicate ocular microenvironment. The exception to this, however, is ipilimumab. Although it is believed that the antibody itself is unable to cross the blood-brain barrier, evidence suggests that activated T cells may be able to penetrate the brain [27][28][29]. This provides one possible explanation for why ipilimumab, among all of the monoclonal antibodies, is associated with the most severe ocular adverse events.
We attempted to conduct a comprehensive metaanalysis of targeted agents in the current literature; however, several limitations need to be acknowledged. The only agents that were included were those that were FDA-approved, so many experimental drugs were not included in this study. Most of these experimental drugs did not have sufficient study data and as such, an analysis was not feasible. A quantitative analysis was not possible due to insufficient data reported in the literature, with many agents and labels having few studies that report ocular toxicities.
Common to all retrospective analyses, variation in study quality was a limitation. Studies displayed variability and inconsistency in the reporting of ocular toxicities and ocular adverse events and in the methods of determining whether an ocular event could be attributed to the agent in question. However, the focus of the analysis was centered around phase 3 and above clinical trials validated by the FDA for quality; in this way, variability was attenuated as much as possible. Although these clinical recommendations were based on our best assessment, we recognize that many of these patients will not be receiving targeted therapy for a prolonged period and that accordingly, many of our screening recommendations may not apply in this therapeutic setting.
The CTCAE grading scale may be limited in its applications to ocular toxicity. For instance, the definition of a grade 4 adverse event, according to the CTCAE, is  Screening: Recommend pre-treatment ophthalmic exam including slit lamp exam and dilated fundoscopic exam. Patient screening for risk factors attributed to the development of severe events (e.g., hypertension, CAD, baseline visual deficits) should be considered before administration of MEK inhibitors. Baseline OCT, fundus photography, and Amsler grid may be performed to identify macular problems. Eye exams may be conducted every 3-6 months to monitor for severe OAEs. Management: The drug should be discontinued with signs of any serious ocular events and an ophthalmologist should be consulted. FDA label recommends holding drug for up to 3 weeks with signs of grade 2-3 retinal pigment epithelial detachment. Do not modify dabrafenib if used in combination. It is commonly believed that subretinal fluid associated with MEK inhibitors will resolve even with continuation of the drug.

Keratitis Keratitis
If mild, initiate artificial tears, refer for ophthalmology assessment within 1-2 weeks and continue drug. If severe, immediate referral to ophthalmology is merited with considerations to withhold the medication until assessment.

Keratoconjunctivitis Sicca Erlotinib
If mild, refer for ophthalmology assessment within 1-2 weeks and continue drug. If severe, immediate referral to ophthalmology is merited with considerations to withhold the medication until assessment.

Macular Edema Vemurafenib
If severe, refer for immediate assessment by ophthalmology and consider withholding medication until assessment.

Ocular Hemorrhage Gefitinib
If severe, refer for immediate assessment by ophthalmology and consider withholding medication until assessment.

Ocular Ischemia Gefitinib
If severe, refer for immediate assessment by ophthalmology and consider withholding medication until assessment.
Periorbital Edema Imatinib, Nilotinib, If mild, refer for ophthalmology assessment within 1-2 weeks and continue drug. If severe, immediate referral to ophthalmology is merited with considerations to withhold the medication until assessment.    an event with "life-threatening consequences," which is atypical of nearly all ocular adverse events. These limitations will be addressed in the upcoming version of the CTCAE.

Retinal Artery Occlusion
Studies referenced by FDA labels provide one perspective of toxicity incidence in the general population; however, all labels acknowledge that because clinical trials are conducted under widely varying conditions, adverse reaction rates observed in the clinical trials of a drug cannot be directly compared with rates in clinical trials of another drug and may not reflect the rates observed in practice. Preliminary data are already beginning to show rates of ocular toxicity that are higher than those reported in the labels. One such example of this is with the incidence of toxicity seen in MEK inhibitors. Preliminary data show that the incidence of retinal pigment epithelial detachment with or without central serous retinopathy (Figure 4) occurred in 30 of 94 patients (31.9%) across six phase 1 clinical trials. This is a substantial increase from the reported figure (0.8%).
As evidenced by our data, ocular adverse events are an increasingly common concern in treatment with targeted anticancer agents. We hope that this study will set the stage for further specific recommendations for the screening and management of ocular toxicities in this entire class of medications. With the increased incidence of ocular toxicities, assessment by and involvement of an ophthalmologist in the treatment of patients receiving agents known to cause ocular events is merited. As the landscape of oncologic management and adverse events changes with advancing therapy, a more multidisciplinary approach to the treatment of patients with cancer is a reasonable recommendation.

Drug retrieval
The CenterWatch database of U.S. Food and Drug Administration (FDA)-approved oncologic agents was reviewed by two independent authors (C. F. and J. L.) for potential candidate agents. 138 agents that were FDA-approved between January 1, 1998, and March 14, 2015, were reviewed and screened for inclusion. We excluded all non-chemotherapeutic and duplicate agents; cytotoxic, conjugate, and immunomodulating agents; and agents with no (or insufficient) evidence of ocular adverse events ( Figure 5). A review of the FDA labels for the remaining agents was performed to exclude agents for which no ocular adverse events had been reported. In addition, a simultaneous review of the literature was performed to identify any independent studies that observed these adverse events as well as any ocular adverse events observed after FDA approval of the agent. We searched Medline and Google Scholar for evidence of ocular adverse events in the remaining targeted agents. Search limits included all studies from drug inception to the present day. FDA labels for 46 targeted oncologic agents were retrieved and screened for ocular toxicity.

Identification of ocular adverse events and study selection
Eligible studies for review included phase 1 or higher clinical trials that investigated targeted therapies as monotherapy, as well as a number of meta-analyses and pertinent reviews of the literature. Search terms included the drug name with Boolean operators AND phase NOT combination, and NOT plus. Studies were screened on the basis of relevance, patient demographics, study design, route of drug administration, and procedural integrity (i.e., randomized, double-blinded controlled trials). All studies that were included were monotherapy trials of the targeted agent. Studies that did not report ocular adverse events were excluded and were not retrieved. In addition, we excluded studies that had used small-molecule drugs in conjunction with Figure 5: Diagram for the inclusion of anticancer agents that were analyzed. All FDA-approved cancer-related agents were screened between January 1, 1998, and March 14, 2015. All non-chemotherapeutic agents, duplicate agents, and cytotoxic agents were excluded. FDA labels were retrieved for the remaining agents, and all agents that displayed evidence of ocular adverse events were included in the study. A total of 16 agents (4 monoclonal antibodies and 12 small-molecule targeted inhibitors) were initially included in the study. Four agents (bortezomib, pertuzumab, dabrafenib, and idelalisib) were associated with minor ocular adverse events according to the FDA label, but no evidence of ocular toxicity was evident upon an independent survey of the literature; these agents were therefore excluded. cytotoxic therapy, those that were not monotherapy, and those that failed to include adverse events below grade 3. Clinical studies were retrieved from Medline, Google Scholar, the Cochrane database, and the NIH Clinical Trials Database. Studies searched on the NIH database were limited to those with results and were reviewed for inclusion criteria and possible retrieval based on study details provided by the sponsor institution.
All ocular adverse event frequencies and severities were identified on the basis of data from both FDA labels and independent clinical studies. When available, FDA label-referenced studies were retrieved and used to create an overview of ocular toxicities across all FDA labelreferenced studies. If discrete FDA data were unavailable, data from independent studies were screened for inclusion criteria and reported.

Study retrieval and data pooling
Two authors independently reviewed the abstracts and figures of all eligible studies. All relevant mention of ocular adverse events were noted and retrieved. Of these studies, the authors reviewed the data and methods for sufficient rigor and independently assessed for risk of bias. All studies that failed to report nominal data, failed to partition the data into discrete adverse events, or failed to report total adverse events were excluded.
Studies were reviewed according to selection criteria. Of these, only studies that reported ocular adverse events were included. Meta-analyses and review articles were reviewed and retrieved as indicated. Data on ocular adverse events, including frequency of independent events, were gathered and reported from FDA labels and selected studies. If the cohort demographics and study designs were similar, the results were pooled. From these data of ocular adverse events, severity of toxicities were separated on the basis of Common Terminology Criteria for Adverse Events v4.03 (CTCAE) grade [22].

Drug comparison
Drugs were separated into two classes: smallmolecule targeted inhibitors and monoclonal antibodies. Data from the two groups of targeted therapies were analyzed to determine significant differences between them. Specific parameters examined included overall frequency of ocular adverse events, frequency of events graded at least CTCAE grade 3, the most frequent ocular adverse events in each class, and the percentage of drugs that were associated with ocular adverse events in each class. Significant findings were then re-examined between with agents that were matched by mechanism of action. Percentages of drugs per class causing ocular events were calculated as total number of agents with known ocular adverse events divided by total number of agents in the group.

Formulation of clinical recommendations
Known ocular adverse events were screened for most pertinent events. Ocular adverse events were included on the basis of the following characteristics: Incidence and frequency of the event, acuteness and aggressiveness of onset, severity of the event, and irreversibility of the event. Next, existing guidelines were retrieved for drugs that have well-documented ocular adverse events (e.g., hydroxychloroquine, ethambutol). A review of the literature was performed to identify relevant meta-analyses, randomized controlled trials, reviews, casecontrol or cohort studies, and case reports or case-series. Interventions and recommendations for rare or unique cases were identified in case studies. Recommendations were formulated with the input of an expert consultant with extensive expertise in dealing with ocular toxicities (DSG).