Discussion

OVs represent a new class of cancer therapeutics with considerable flexibility to induce tumor cell death in a manner that promotes both innate and tumor-specific adaptive immune responses. Indeed, preclinical studies have identified a large number of viral species with native permissiveness for tumor cell replication.1 17 In addition, viruses can be easily modified to improve antitumor activity by promoting preferential entry and replication in cancer cells and enhancing antitumor immunity through altering antiviral immune responses or expressing eukaryotic genes that increase direct cell killing or promote host immunity. The clinical implementation of OV therapy, despite preclinical support for their antitumor properties and tolerable safety profile, has been slow. Challenges include an incomplete understanding of the underlying mechanisms of tumor regression with specific OV agents, lack of biomarkers to better match effective viral species with permissive tumor types and patient features and limited standardization of immune correlates in clinical trials. In order to better understand the current clinical landscape of OV clinical development, we reviewed 97 published OV clinical trials and confirmed a wide range of approaches are under study spanning many different viruses, cancer populations, routes of delivery and combination treatment strategies. Most of the studies were early phase trials likely reflecting the novelty of this approach, but it is also possible that later stage, negative studies may not have been published creating a missed opportunity to better understand the complex responses of OV treatment in patients with cancer. Furthermore, we found that while most studies did evaluate antiviral humoral responses and made limited attempts to identify viral shedding, few studies focused on defining cellular immune responses against virus or tumor cells.

The most common viruses used in OV cancer clinical trials were adenovirus, HSV-1, reovirus and poxviruses. This likely reflects the better understanding of DNA viruses and the ability of large DNA viruses to allow deletion of non-essential viral genes to alter cell selective replication and reduce pathogenicity while being able to accept foreign transgenes for expression. We also found that approximately two-thirds (65%) of the OV studies conducted used some type of viral modification. While transgene expression was reported in 40 clinical trials encompassing 51 specific recombinant genes, the majority were GM-CSF (see figure 2), which is a cytokine thought to promote recruitment and maturation of dendritic cells, and ultimately to help generate adaptive immune responses by promoting cross presentation of tumor antigens.1 18 Other commonly used transgenes included LacZ, included for viral detection purposes, and enzymes designed to activate pro-drugs that aid in tumor cell killing. GM-CSF was used in T-VEC, and the common inclusion of GM-CSF likely represents its integration in the T-VEC product, but is was surprising that other cytokines and genes encoding immune stimulatory proteins were not commonly used in viruses entering the clinic. Although not as amenable to genetic manipulation, reovirus was one RNA virus that has been widely studied in the clinic. The selection of the optimal virus and potential transgenes should be based on further biological analyses of tumor cells, host factors and mechanisms that promote Th1 and CD8+ effector T cell immune responses. Studies have identified intracellular sensors, such as the cGAS-STING complex and Toll-like receptors that are critical for tumor cell induction of host innate immunity.19 Interestingly, these same sensors are used for detection of DNA and RNA viruses, and the status of these intracellular sensors in human cancers are not well defined.20 Further translational and biomarker studies are needed to better define the status of these antiviral machinery elements across human cancers and preclinical studies should continue to evaluate which viruses and genetic modifications are best utilized in tumor cells based on their genomic profile at the time of treatment. While most studies used exogenous OVs, we also identified two trials that used non-traditional approaches to OV treatment. In one clinical trial focused on HIV-positive relapsed/refractory lymphoma, investigators used the proteasome inhibitor, bortezomib, to activate latent gammaherpesviruses (GHVs) in tumor cells.21 The investigators hypothesized that bortezomib would induce lytic activation of the GHVs promoting lymphoma cell killing and might also inhibit HIV infection by restoring the APOBEC3G cytidine deaminase, a natural antiretroviral. The trial did confirm activation of Kaposi sarcoma herpesvirus and Epstein-Barr virus and reported responses in 17 of 22 patients, but treatment also allowed for chemotherapy and rituximab. Strategies to activate latent viruses may be an interesting approach for other cancers as well, and other drugs may be useful for viral reactivation. Panobinostat, a histone deacetylase inhibitor, was reported to reactivate latent HIV in aviremic adult patients, which was used as a way to more completely eradicate HIV with antiretroviral therapy.22 We also found a study that used OVs to infect tumor cells ex vivo as a strategy to optimize autologous cell vaccination.23 These studies suggest that other approaches with OVs are possible and may merit further development.

We hypothesized that most OV clinical trials would focus on melanoma and other skin cancers given the easy accessibility of tumors for local injection. While melanoma was the most common tumor targeted in clinical trials (n=30), and represented the largest number of patients (n=1000), this was skewed by the phase III OPTiM clinical trial of T-VEC which enrolled 436 patients. There was only one clinical trial reporting OV treatment in cutaneous squamous cell carcinoma, although other studies in non-melanoma skin cancer are currently in progress.24 We found that the next most common cancer being targeted with OVs was GI tumors, including colorectal, pancreas, gastric, esophageal and anal tumors (see table 1 and figure 3). We also found many clinical studies were open to any solid tumor and the heterogeneity of patient populations in OV clinical trials may provide challenges to better defining early signals of response. A minority of studies focused on pediatric tumors and patients with a variety of hematological malignancies. Further preclinical studies would be useful to understand potential difference in viral permissiveness and replication across tumor types by individual OV agents.

Another controversial issue in OV clinical development is the selection of the best route of administration. While OVs are ideally suited for direct IT injection, this may limit the number and location of tumors that can be directly treated. While intravenous delivery offers the potential to infect metastatic lesions in multiple locations, administration of OVs into the circulation is limited by dilution in peripheral blood and potential clearance by pre-existing antibodies and other serum proteins.11 There is much interest in identifying novel delivery mechanisms that avoid premature viral clearance while promoting drug biodistribution to tumor sites, including viral envelope modifications, nanodelivery vehicles, integration into stem cells and other cellular carriers.25 Our data suggest that while IT was the most common mode of OV delivery representing 48 clinical trials and 1482 patients (45.8%), there are a large number of studies that have used intravenous administration (34 clinical trials and 1147 patients; 35.5%). Other modalities included hepatic artery infusion, intraperitoneal delivery, intravesical delivery, intradermal injection, direct injection into the tumor bed and convection-enhanced delivery. In some studies, both IT and intravenous administration was used. Overall, there were no major differences in safety or bioshedding that could be distinguished between intravenous and IT (online supplemental table 2) but this retrospective review was not powered to identify differences. Further studies to understand how delivery impacts viral distribution to the tumor site, viral clearance, antitumor immunity and safety are needed. Clinical trials may also consider alternative delivery routes to help define how this may impact therapeutic outcomes.

Most of the clinical trials were early phase studies making analysis of clinical endpoints difficult. Nonetheless, we attempted to explore therapeutic responses and found that the overall objective response across the studies was 9% with an additional 12% of patients achieving stable disease as the best response. While these numbers are low, it is important to remember that most of the studies were early phase and not designed to detect clinical responses. It was notable that only 53% of the reported studies used standardized RECIST criteria to report objective responses making analysis of clinical responses challenging (figure 8A). Of the responses reported the largest numbers appear to be in studies using HSV-1, adenovirus, reovirus and vaccinia virus but this may reflect the larger number of trials being conducted with these OVs (figure 8C). Indeed, HSV-1 includes the only phase III clinical trial which enrolled 436 patients.4 We also examined whether the route of administration impacted antitumor activity. Not surprisingly, objective responses were higher with IT delivery likely related to more rapid viral clearance and dilution when administered by the intravenous route (figure 8D). It was interesting, however, that intravenous administration was associated with objective responses in 52 patients suggesting that further investigation and optimization of intravenous delivery of OVs is warranted.

The majority of studies reported were designed to evaluate safety of OV treatment, and our data confirm previous statements that OV treatment is tolerable across many different viral agents, combination strategies and delivery routes. As shown in table 2, the most common adverse events were low-grade constitutional symptoms, including fever, fatigue, chills/rigors, nausea/vomiting, non-specific pain and flu-like symptoms. Other common adverse events were low-grade injection site pain. We evaluated high-grade treatment emergent events and found most events related to disease progression or other drugs used in combination treatment regimens (online supplemental tables 2 and 3). Although there was one case of a grade 4 infusion reaction associated with intravenous delivery of a reovirus, no major differences were noted between intravenous and IT delivery or with specific viruses. Thus, OV therapy does appear to have an overall highly tolerable safety profile with largely non-overlapping toxicity with other cancer therapeutics. While most of the local and constitutional symptoms were low grade, occasional high-grade events were reported and attention to premedicating with acetaminophen and other analgesics might be considered to prevent or ameliorate these reactions.

Given the safety profile and mechanisms of action associated with OV, they have been suggested as good agents for combination studies. In fact, preclinical data and emerging clinical data support the combination strategy, perhaps most notable for OV and immune checkpoint blockade.17 An initial high rate of response of 50% in patients with advanced melanoma was reported for T-VEC in combination with ipilimumab in a small phase I study.5 This was later confirmed in a larger phase 2 study in which 198 patients with melanoma were randomized to receive T-VEC and ipilimumab or ipilimumab alone.26 In this study, the response rate was doubled from 18% for ipilimumab alone to 38% for patients treated with the combination. In another small phase I study, T-VEC and pembrolizumab was associated with a 62% objective response rate in patients with advanced melanoma, with a third achieving a complete response.6 A larger randomized clinical trial evaluating T-VEC and pembrolizumab versus pembrolizumab alone has completed enrollment and is awaiting follow-up (NCT02965716). Thus, we hypothesized that most of the combination studies would be OV and immunotherapy. However, in review of the 97 clinical studies, only 36 included a combination agent and only five were with immunotherapy (figure 4). Surprisingly, most of the combination studies were evaluating OV and cytotoxic chemotherapy. Prodrugs and radiation therapy were also being investigated in combination with OV; and there is clinical data supporting these concepts.27–29 The use of chemotherapy may be a reflection of the tumors being studied and indeed the most common regimens were not specified with the selection of the chemotherapy left to investigator choice or carboplatin/paclitaxel regimens, commonly used in studies of patients with non-small cell lung cancer. Chemotherapy administration did appear to be associated with expected adverse events but it did not appear to be significantly worse with associated OV treatment. Preclinical studies of how therapeutic responses are improved with specific OVs and other agents should be a high priority with more rapid translation of promising combinations into the clinic.

An important correlate of OV administration is to ensure delivery of virus to tumor cells and assess the degree and kinetics of viral shedding in body fluids and compartments. While we found that viral bioshedding was assessed in 71 studies, most limited this to evaluation of blood or serum in 56 studies, urine in 36 studies and tumor tissue in only 26 studies. Only a few studies included more extensive tissue evaluation, including saliva, fecal samples, skin, injection sites and sputum. Further, while detection of virus is important, detection of replication competent virus is more meaningful, and this requires plaque assays or other methods that can detect viral replication. In the studies included in this report, most investigators used PCR (n=58) to detect viral genome while 12 studies used both PCR and confirmatory plaque assay with one study relying solely on plaque assay for viral detection. Of note, there were no reports in any study of household or close contact transmission of OVs. These data suggest that biodistribution is not being reported routinely or in a standardized manner across most OV studies. An assessment of tumor biopsies for the presence of replicating virus should be considered as part of trial designs, especially in early phase clinical studies. Alternatively, dedicated bioshedding studies could be contemplated to provide a more comprehensive assessment of viral biodistribution and kinetics, and only one trial of T-VEC was dedicated to such an assessment.30

Since the outcome of OV treatment likely depends on the balance between antiviral and antitumor immunity, we assessed whether the OV clinical trials were reporting data on these responses. While 58 studies did report antiviral antibody titers, only 27 included specific assessment of neutralizing antibody titers. Analysis of cell-mediated viral responses was less common and was reported in only 10 clinical trials, although all were able to detect an increase in viral-specific T cells after treatment. Less than half of the clinical trials (n=48) attempted to evaluate antitumor immune responses with most focusing on more general descriptive analyses, such as immunohistochemical staining of tumor biopsies for immune cells, serum cytokine levels, and analysis of numbers and activation status of peripheral blood T cells (see table 3). Only seven studies directly assessed tumor-specific T cells, largely by interferon-gamma ELISPOT assay. Clearly, more efforts should be made to include biomarkers of both humoral and cellular virus-specific and tumor-specific immune responses in clinical OV studies.

This report has several important limitations, including the retrospective nature of the review, the focus on including only peer-reviewed published papers identified on PubMed, the heterogeneity of the clinical trials and the long period of time (20 years) encompassed by the review. We recognize that many OV clinical trials have not been reported and were, thus, not included in this review. For example, many studies of an oncolytic coxsackievirus have been conducted with interesting clinical and biomarker results but these have only been reported in abstract form, to date and thus were not included in our analysis.31 Further, negative studies, such as a recent Phase III clinical trial of an oncolytic vaccinia virus, Pexa-Vec in combination with sorafenib, in hepatocellular carcinoma was reported through press releases but data was not published.32 Although disappointing, publication of negative studies would be helpful to generate discussion and provide important information on how to improve OVs, patient selection and study designs for OV therapy.

Despite the limitations of the current report, our data support the need for a more organized effort to standardize clinical trial design for OV clinical trials. Investigators should consider using standardized endpoint response criteria when reporting clinical responses. Since RECIST appears to provide a similar pattern of response to that seen in studies not using RECIST, this may be an appropriate tool for monitoring OV-treated patients. We would also support efforts to work with regulatory agencies to identify better methods for assessing injected lesions independently from patient-level responses as this may be informative in small, early phase clinical studies. Collecting data on viral bioshedding is important and attempts to monitor virus through measurement of replicative virus rather than PCR across multiple anatomic sites should be the standard in phase I studies, and this might be omitted from later phase development unless a specific transmission concern is seen. While most trials did evaluate antiviral antibody titers, additional efforts need to be made to evaluate antiviral T cell responses. Whenever possible, assessment of both antiviral and antitumor cellular immunity in early phase trials may provide important insights into the underlying mechanisms of antitumor immunity with OVs and guide late stage development. Finally, while OVs are well suited for combination regimens, few published studies have explored OV with other immunotherapy agents and this would be a high priority for further clinical investigation.

In conclusion, we summarized the clinical experience with OVs over the past 20 years. While the data are not exhaustive, it provides a snapshot into the types of OVs being used in the clinic, tumors being targeted, combinations in development and demonstrates the status of correlative studies being done. OVs represent a novel approach and may require additional research to optimize the viral vectors, promote better patient selection and improve viral delivery and biomarker analysis. The unique mechanism of action may also require a change in current methodology used for clinical endpoint assessments, and recently investigators have proposed new criteria for RECIST monitoring of OVs and other IT agents (itRECIST).12 Our data suggests that currently most OV studies use large DNA viruses with modifications and transgene expression has largely employed GM-CSF with most given by IT delivery although there is an increasing number of IV studies. While most studies have used monotherapy OVs, published combination studies have largely been with chemotherapy. There is a need for more preclinical studies to better define the underlying biological mechanisms that OVs use to mediate antitumor activity and clinical studies need to consider more standardized approach to defining viral distribution and integrating appropriate biomarker studies that provide information on both the antiviral and antitumor immune responses. Investigators in the field should also be encouraged to publish their data, which will speed clinical development and help optimize the full potential of OVs for the treatment of patients with cancer.