Site selection

The IICT Expert Panel agreed that a solid infrastructure and close working relationships between clinicians from different specialties were essential for sites where intratumoral immunotherapy studies are to be conducted. Those in favor of conducting trials only at academic hospitals pointed out the importance of integrated site expertise (eg, close working habits in the constraints of clinical trials between oncologists, interventional radiologists, radiologists, and surgical oncologists), having a pre-existing immune-monitoring infrastructure, the need for provider experience in delivering all types (including prior lines) of immunotherapy, and the necessary availability of specialists outside of medical oncology to be “modality-specific champions” of the delivery technique or the management of organ specific adverse events (AEs) (eg, dermatologists, interventional radiologists, surgeons, pulmonologists, etc). Some panel members further asserted that more complex or earlier phase trials should initially enroll only at comprehensive cancer centers. Those in favor of conducting trials at any center with the interest in and capability of performing intratumoral immunotherapy trials (ie, including community hospitals) argued that if site selection is too restrictive, then resulting data may not be generalizable, particularly if an intratumoral agent is targeted for eventual use in the community. Regardless of site, the group agreed that interventionalists (eg, interventional radiologists, pulmonologists, gastroenterologists) should be named co-investigators as they need to be invested in good injection technique and could complete a credentialing program and use a template structure for documenting each injection.

Exclusion criteria

With regard to exclusion criteria, the panel identified several factors that would necessarily preclude intratumoral immunotherapy trial participation, including difficult to access lesions, painful or large necrotic lesions, the presence of non-injectable lesions (depending on trial endpoints), proximity to large vessels or critical structures, extensive visceral disease, symptomatic CNS disease, extensive tumor burden, and lesions (injected and non-injected) that cannot be measured. Patient-specific exclusion criteria included coagulopathy, bleeding disorders, life-threatening autoimmune disease, poor ECOG PS, and multiple comorbidities. Further, special attention should be paid to the use of specific intratumoral immunotherapy agents in patients with compromised immune systems (eg, solid organ transplant patients, chronic immunosuppression).

Stratification factors

The panel agreed that stratification should account for the disease setting and experimental agent and should be modeled after previous studies to predict which factors will have the greatest impact on efficacy. In addition to patient-specific factors such as age and ECOG PS, the group identified the following factors that could be used for stratification: the presence of visceral (particularly liver) metastases, tumor markers (eg, LDH above normal40, etc), tumor immunogenicity (potentially defined by tumor mutational burden (TMB)), stage, size or number of lesions, disease burden (eg, prostate-specific antigen (PSA) elevated in prostate cancer), distant metastases, previous response to (ICB) therapy, HPV status for certain cancers, tumor type, or PD-L1 status.

Safe and effective dose to take into definitive trials

The optimal biological dose of intratumoral therapies may sometimes be informed by pharmacodynamic readouts of preclinical assays (as demonstrated in Maurer et al,41 and dose escalation should be done with a significant number of patients per dose level to ensure that the optimal therapeutic dose is reached42 43), which is particularly important when the maximum tolerated dose (MTD) is not expected to be the same as the optimal biological dose. The majority of the expert panel agreed that the safe and effective dose of an intratumoral agent to take into definitive trials should be based on data from an intratumoral monotherapy cohort followed by an intratumoral therapy plus systemic therapy combination cohort, pointing out the need to clearly demonstrate single-agent activity independent of synergism with a systemic agent, particularly for non-injected lesions. The remainder of the group elected for determination of this dose to be based on an intratumoral monotherapy cohort with simultaneous/parallel enrollment into an intratumoral therapy plus systemic therapy combination cohort, arguing that in most cases, overlapping intratumoral and systemic agent toxicities are expected to be limited. If the development goal is to combine the intratumoral agent with a systemic agent, however, then the intratumoral monotherapy portion of the study could be limited. Depending on strength of preclinical characterization, it may be advisable to robustly investigate multiple doses and/or schedules prior to moving to definitive trials in order to satisfy regulatory considerations.

Trial design

While there are multiple options for the design of phase 1 trials, some of which were discussed by the expert panel, no consensus was reached among the group on optimal phase 1 intratumoral trial design. The group agreed that a typical 3+3 design is not well adapted for studying intratumoral immunotherapy as more patients per dose level, more data points, and a different assessment of pharmacodynamic impact and level of immune activation associated with the drug are needed. Trials of intratumoral therapy require more of a composite analysis including early efficacy, safety, pharmacodynamics, and markers of immune activation from blood and tissue analyses of injected and most importantly non-injected lesions. There was consensus that adequate safety and biology be explored at each dose level and that the optimal study design might vary. This is particularly important in intratumoral therapies where the MTD may not be the most biologically active dose.

Measuring response in phase 2 neoadjuvant and unresectable/metastatic trials

The majority of the panel thought that phase 2 trials in the neoadjuvant and unresectable/metastatic settings should include a measure of response as a primary endpoint preferably in both injected and non-injected tumors to ascertain systemic efficacy (see the discussions on Sample collection and Efficacy assessment, below). In this setting, local pathological responses need to be scored and their association to recurrence-free survival (RFS) established; currently, perioperative or neoadjuvant immunotherapy effect on pathological complete response (pCR) is incommensurate with RFS. This was exemplified in the primary analysis7 of neoadjuvant talimogene laherparepvec (T-VEC), where many patients who did not achieve pCR at surgery remained recurrence free at the time of analysis and, as the authors noted, the presence of apparent visible residual tumor at surgery did not associate with a poor response. Indeed, durable improvements in RFS and overall survival (OS) were also demonstrated at the 5-year follow-up of neoadjuvant T-VEC,44 despite a low rate pCR at the early analysis. Therefore, as some analyses of resected tumors are not predictive of OS, caution should be used in interpretation of these readouts. Also, event-free survival (EFS) is an important measure to capture those who progress prior to surgery, as well as events defined in the RFS. With regard to trials conducted in the unresectable/metastatic setting, response rates can be misleading and DCB or landmark PFS are preferred when high ORRs have not translated into phase 3 trial success, and where phase 3 trial successes have been realized with low response rates.45 46 With regard to measuring response on imaging, in addition to traditional endpoints based on RECIST v1.1, investigators may collect data that would allow exploratory analysis using novel immunotherapy criteria such as iRECIST and itRECIST. This means allowing therapy past RECIST v1.1 progression in clinically stable subjects, measuring new lesions when possible, and carefully documenting which lesions are injected at each treatment visit, especially those chosen for quantitative assessment (ie, “target” lesions in RECIST terms).

Randomization of phase 2

The panel agreed that randomization of phase 2 trials, possibly with two dose levels, is generally preferred as randomized data are more likely to support FDA approvals. If randomized phase 2 data do not lead directly to approval, then the potential for expansion into a phase 3 trial could still expedite drug development. Furthermore, single-arm data for any agent can be misleading, and comparison to SOC is preferred when the intratumoral drug is planned for use in combination with a systemic agent. Randomization should be used as an opportunity to introduce surrogate biomarkers of efficacy as well. The group also agreed that the decision to randomize phase 2 trials depends on the nature, intent, design, and goal of the trial as well as existing indications and unmet need for the specific cancer under study. For example, for patients whose cancer has progressed through multiple lines of therapy, particularly prior immunotherapy, an SOC comparison may not exist. Also, for biomarker based (tumor agnostic) patient selection, a standard or ethical comparison may not be established. Furthermore, there can be value in determining efficacy of an intratumoral drug regimen in a phase 2 study prior to randomization, particularly when a homogeneous population is needed to demonstrate effect and leverage a phase 3 trial. The group agreed that a Simon 2 stage design47 with clear futility rules is optimal for phase 2 trials of intratumoral agents.

Sample collection

The expert panel agreed that mandatory serial tissue collection of injected lesions is recommended in early-phase intratumoral immunotherapy clinical trials for target and pharmacodynamics assessments unless it is unsafe (eg, a difficult trajectory or proximity to large vessels) or not feasible (eg, when tissue collection would compromise the ability to inject or assess a lesion). Tumor biopsies of an uninjected lesion can also be performed for documentation of systemic antitumor effects (abscopal (tumor response in uninjected lesions without concomitant systemic therapy) or anenestic (tumor response in uninjected lesions in the context of concomitant systemic therapy such as intravenous anti-PD-1 such that the effect might be related to the systemic therapy rather than a consequence of the intratumoral therapy)), although priority should be given to analysis of the injected lesion. Furthermore, any tissue collection, but particularly collection of non-injected lesions, should involve thorough patient education to facilitate informed consent. While serial tissue collection is generally not feasible for larger phase 3 trials, it is compatible with phase 1 and 2 studies. For example, serial tissue collection can be used during dose escalation to inform further dosing (eg, to monitor for a bell shape curve effect), in a randomized phase 2 substudy, or during a phase 1 study at a dose selected for further evaluation to determine which biomarkers are most relevant to carry through to larger studies. Regardless of when it is introduced, the group agreed that serial tissue sampling of non-injected lesions (when present) would help to establish a systemic and potentially durable response that could reinforce the value of the treatment for a registration trial. Additional tissue collection factors to consider include the use of fresh tissue biopsy (supernatant) analysis, which is still experimental but may provide valuable information, and evaluation of TDLN tissue. Peripheral blood should be also collected to measure systemic immune responses with markers including plasma (eg, for cytokines and soluble factor titration), serum (eg, for antibody titration), peripheral blood mononuclear cells (PBMCs), circulating tumor DNA (ctDNA), and coding DNA (cdDNA).

Efficacy assessment

The duration and magnitude of response overall and of injected versus non-injected lesions separately should be assessed, taking into account how representative the target lesions are for the total burden of potentially injectable lesions. While there are several exploratory endpoint response criteria that can be useful for monitoring and recording responses of lesions to intratumoral immunotherapy during trial investigation, such as itRECIST,48 irRECIST,49 and iRECIST,50 some of the panelists recognized that these tools can be very cumbersome for investigators. Using standard RECIST v1.151 allowing for treatment and response collection beyond progression could be used as the primary method for response assessment recording injected and non-injected lesions and measuring new lesions when possible. However, the protocol should detail how the data beyond progression should be reported in the case report form, notably when uninjected lesions become injected and especially if these lesions were initially part of the RECIST target lesions.

The authors also acknowledge the benefit of highly active injected local immunotherapies that may have limited impact on OS or systemic efficacy but still have value in some clinical situations. For example, local intratumoral treatment of cutaneous metastases (which are often difficult to manage) in patients with triple-negative breast cancer could significantly improve the quality of life of patients. Thus, repeated local administration should be permitted when there is benefit to factors such as quality of life. This is a limitation with current traditional response criteria and may require revisiting appropriate endpoints and assessments to capture all types of benefit.

Biomarkers

While the expert panel was unable to identify a core set of universal biomarkers of response to intratumoral immunotherapy, there was agreement that biomarkers should match the mechanism of action of the drug. Several important tissue-based biomarkers (which may depend on the agent under study) were identified, including markers of T-cell clonality, TME immunophenotype, tumor-infiltrating lymphocytes (TILs), regulatory T-cells (Tregs), dendritic and T-cell activation markers, markers of exhaustion/suppression, IFNγ gene expression signature, and spatial profiling (RNA, protein). Important blood-based biomarkers of response were also identified, including PBMCs, clonality of T-cell response, ctDNA, cdDNA, immunophenotyping, circulating cytokines/chemokines, gene expression signatures, neoantigen expression, and viral titers and antidrug antibodies for oncolytic viruses. Molecular imaging might aid the identification of local versus abscopal effect from locally delivered immunotherapies.52 However, further research is needed to qualify optimal radiotracers to be used for immune monitoring of these patients.

Study population

The panel prioritized the development of intratumoral immunotherapy for premalignant lesions with a higher risk of progression (eg, non-muscle invasive bladder cancer, high-risk cirrhosis, congenital giant nevi), that are not amenable to surgery (eg, familial polyposis syndromes with numerous lesions such as neurofibromatosis), or for which surgery would incur a high level of morbidity (eg, genital, oral cavity/oropharyngeal, and facial lesions). They also prioritized peritumoral and/or regional application of drug (eg, draining LNs) in this setting.

Endpoints for non-randomized trials

Most of the panel agreed that a change in pretreatment versus post-treatment characteristics of the injected premalignant tissue was the best primary endpoint, while a smaller set of respondents prioritized a change in pretreatment versus post-treatment systemic biomarkers or other endpoints. The local effect measured could include reduced dysplasia, immune infiltrate, etc, and serve as proof of efficacy, particularly as there has been a lack of systemic response for intratumoral immunotherapy in this setting. Disease-dependent time to event endpoints identified included time to progression/relapse, development of frank malignancy/invasive disease, or development of new lesions. Regardless of the chosen endpoint, the panel agreed that response should be measured in injected and non-injected lesions whenever possible.

Endpoints for randomized trials

Most of the expert panel agreed that EFS was the best primary endpoint with the remaining panel members prioritizing other endpoints, such as time to progression on first subsequent therapy, disease-specific progression-free survival/death due to disease, progression due to malignancy, local invasion/development of invasive lesions, emergence of histologically similar lesions, or a need for subsequent treatment or imaging. The group agreed that the focus should be on endpoints that are acceptable to regulatory agencies and will lead to approvals, which would most likely be EFS potentially represented by a composite of events in order to expedite study completion. The group agreed that endpoints are necessarily tumor-specific, but overall should have a theme of time to avoiding something morbid (eg, surgery, radiation, malignant transformation, loss of heterozygosity, or progression or recurrence of disease on imaging). The group agreed that quality of life may be a relevant endpoint for some diseases, and that sporadic versus genetic lesions will have different endpoints. As the effect of local immune changes on systemic efficacy is still unknown, local immune responses should ideally be developed with a biological correlate.

Sample collection timing

Although the panel agreed that the timing of tissue and/or blood collection depends on the agent and disease under study, they agreed that the ideal time for tissue analysis is prior to the onset of necrosis but after TILs can be demonstrated (a time frame of roughly 48 hours to 1-week postinjection for most agents, and likely within 3–4 weeks of injection for all agents because of the time needed to mount de novo adaptive immunity). Initial tissue sampling should occur when the tumor is still intact (eg, at time of diagnostic biopsy, screening, or enrollment, or prior to the first injection). Blood samples, on the other hand, may be obtained sooner in order to demonstrate an immediate effect. Although serial tissue or blood sampling is ideal, these data may not be logistically feasible (eg, when treatment of the premalignant lesion involves complete resection). When feasible, serial tissue and/or blood samples should be collected at regular intervals for up to 1-year status post (first) injection, taking practical considerations into account (eg, biopsies at the time of procedures such as colonoscopies, injections, etc). The intervals between sample collections will depend on the agent and disease under study.

Randomized study arms

The panel agreed that the control arm for randomized studies of intratumoral immunotherapy used to treat clinically non-metastatic disease should be SOC treatment, which may entail adjuvant only, neoadjuvant only, or neoadjuvant followed by adjuvant systemic therapy. The administration of neoadjuvant intratumoral immunotherapy in the experimental arm should occur at least 1–3 weeks prior to local treatment (eg, surgery, radiation) to optimally boost/prime adaptive immunity. Continuation of the intratumoral agent into the adjuvant setting depends on the SOC for the disease under study (see box 3).

Intratumoral placebo control

The panel was divided about whether or not an intratumoral placebo control was required or even preferred in the control arm of a randomized study of intratumoral immunotherapy. Another question raised was the correct vehicle for an injectable drug, for example, saline versus a blank virus. When considering whether or not to use an injectable placebo control, the group agreed that it is also important to consider context (eg, accessibility of lesions, reliability of other biomarkers, etc). Additionally, it is recommended that patient advocacy groups be involved in the discussion of trials and the ethics of including placebo controls from a patient perspective.

Those in favor of using an intratumoral placebo argued that stress and inflammation related to the injection itself may independently alter the TME and confound the effect of the experimental agent. Furthermore, drug approval is often contingent on use of a placebo. Those opposed to using an intratumoral placebo argued that if intratumoral immunotherapy is considered to be a procedure plus drug combination—with the injection being an integral component of the therapy—then the control arm should not include an injectable placebo. Patients may also be unwilling to enroll in studies where they receive an injection without guaranteed drug delivery.

Using a placebo intratumoral vehicle in the control group for a randomized phase 2 study but then forgoing the intratumoral placebo in the larger phase 3 study was one strategy discussed. One instance identified as an acceptable “open label” scenario was the addition of an intratumoral agent to a systemic SOC treatment compared with SOC systemic treatment alone. The group also agreed that it may be more important to have an injectable placebo if the endpoint is more short-term (eg, EFS or degree of tumor regression), however, if the endpoint is more long-term (eg, OS, PFS), then having an injectable placebo may not be as important. Besides the potential negative impact on patient enrollment, if an intratumoral placebo is being considered for incorporation into the control arm of a randomized study, then the potential for harm to the patient with the injection should be carefully weighed against the benefits of study blinding, and these risks should be discussed with patients and regulatory agencies. Alternatively, agencies might consider intratumoral placebo as a sham surgery bringing futile risks and treatment burden to patients, and prefer evaluating the intratumoral injection procedure and the injected therapy as a whole.

Endpoints for non-randomized trials

The panel was divided over the most important primary endpoint for non-randomized studies of neoadjuvant intratumoral immunotherapy. Many members agreed that pathological response of the injected lesion(s) is a reliable primary endpoint. Some members strongly argued that EFS should be included (and correlated to pCR and partial pathological response (pPR) rates in injected and non-injected lesions) as a primary endpoint in non-randomized studies given that there is no established association between pathological response and survival endpoints for intratumoral therapy, and therefore, pathological response is not an accurate endpoint. Overall, the group still favored this endpoint particularly in instances where a good ORR had been demonstrated in earlier phase trials. As pathological response can be nuanced, the group agreed that it should ideally be evaluated centrally with established criteria,56–59 presuming that tissue transfer for central evaluation would not cause a prohibitive delay in tissue assessment. The remainder of the group prioritized clinical response on imaging (potentially including positron emission tomography(PET) scan results),60 61 ctDNA,62 EFS, and disease-free survival (DFS) or time to relapse. RFS for the intention-to-treat (ITT) population was another proposed endpoint, although this endpoint may be difficult to interpret in the absence of a control arm. R0 resection was another proposed endpoint, although it can be heavily dependent on the surgeon and pathologist.

Endpoints for randomized trials

Many expert panel members agreed that EFS was the most important primary endpoint for randomized trials of neoadjuvant intratumoral immunotherapy, while many others prioritized OS,63 DFS, RFS (for the ITT population), distant metastasis-free interval (DMFI), response in the injected lesion(s), response in the non-injected lesion(s), response measured in all lesions, or progression of residual tissue, pCR (with a survival secondary endpoint), ctDNA (as a tertiary endpoint), or treatment/procedure-related AEs (TRAEs). The group agreed that survival endpoints such as RFS and EFS are more likely to lead to regulatory approvals.

Sample collection timing

Although the panel agreed that the timing of tissue and/or blood collection depends on the agent under study, they agreed that the ideal time for first tissue collection is when the tumor is still intact (eg, at time of diagnostic biopsy, screening, or enrollment), 2 weeks prior to the first injection, at the time of the first injection, status post neoadjuvant therapy (including prior to and 3 weeks postsurgery), and at the time of progression or recurrence. Blood samples should be obtained just prior to biopsies. Because antitumor efficacy of immunotherapies is time dependent, the timing of surgical resection of the injected lesion is also critical. Because it takes about 3 weeks to mount de novo immunity, significant time might be needed between the intratumoral immunotherapy, the local response, and the systemic antitumor effect. Therefore, adaptive trial designs and associated tissue sampling timing could be envisioned where the local response is monitored and the time of surgery is adapted to the kinetics of tumor responses (eg, delay surgery until kinetics, as measured by ctDNA or imaged tumor diameter, is declining).

Line of therapy to prioritize for development

The expert panel was divided about which line of therapy to prioritize for development of intratumoral immunotherapy drugs. Those who prioritized development following progression on prior immunotherapy received to treat metastatic disease argued that the development of novel agents typically starts in later lines of treatment to establish safety and efficacy, with findings from these studies then applied to earlier lines of treatment. In some cases, intratumoral immunotherapy may be used to “prime” tumors that are insensitive to ICB.16 Acquired resistance to anti-PD-1 has been associated with disrupted type I IFN signaling secondary to acquired mutations in JAK1, JAK2, beta-2-microglobulin,65 and HLA class I loss particularly in intermediate TMB tumors.66 These changes, however, may result in tumors that are more sensitive to intratumoral immunotherapy agents. For example, oncolytic viruses may replicate more efficiently in tumor cells lacking intact IFN or JAK-STAT signaling.67

In addition to tumor cell status, line of treatment may also influence therapeutic outcomes. Indeed, some panelists argued that intratumoral immunotherapy is always more effective earlier in the disease course (ie, when the host immune system is most intact)23 66 68–70 and could be used to augment front-line SOC systemic therapy. Some panel members prioritized development following progression on prior immunotherapy received in the adjuvant setting, although this was agreed to be an area where further research is needed. The panel also agreed that some cancers, such as hormone receptor-positive breast cancer, are relatively insensitive to immunotherapy in any line of treatment, and this should also be considered. Conversely, advanced cancers that have acquired genomic instability have typically lost their type I IFN sensitivity.23 68–70 Therefore, some intratumoral immunotherapies which are type I IFN inducers (eg, TLR or STING agonists, oncolytic viruses) might not be adapted to such subsets of diseases.

Randomized study arms

The expert panel discussed their preferred study arms for randomized trials of intratumoral immunotherapy for unresectable/metastatic cancer in different lines of treatment. The group agreed that the control arm should receive SOC therapy regardless of treatment line. In the event that a SOC option does not exist, for example, for some patients whose cancer has progressed on ≥1 prior line of immunotherapy, physician’s choice of treatment is a viable alternative. In the case of patient enrollment according to a tumor agnostic biomarker, then the clinical trial design and size should be discussed with authorities first if the trial intent is registrational.

Most of the panel agreed that addition of intratumoral immunotherapy to SOC treatment should serve as the experimental arm for studies of patients who are immunotherapy-naïve. With the caveat that trial design may be dependent on phase 2 observations, the group agreed that intratumoral immunotherapy with or without SOC treatment should be the experimental arm for patients whose cancer has progressed on ≥ 1 prior line(s) of non-immunotherapy treatment. With the caveat that trial design is often context-dependent, many agreed that the intratumoral agent with or without the failed systemic immunotherapy should serve as the experimental arm for patients whose disease has progressed on a prior line(s) of immunotherapy.

Intratumoral placebo control

The panel agreed that an intratumoral placebo agent is more important for trials of unresectable or metastatic disease in the front-line setting, and regulatory agencies may require a placebo injection in the control arm regardless of treatment-line setting. A broader discussion about the necessity of intratumoral placebos with the FDA may be warranted, particularly for lesions that are not (safely) accessible (eg, esophageal lesions). Caveats of placebo control arms include the potential negative influence on patient enrollment and are still relevant in the metastatic setting, although less impactful. Also, in the context of a very active intratumoral therapy (eg, ORR≥50% in injected lesions), it might not be ethically acceptable to inject a placebo. Patient advocacy groups should also be involved in the discussion of trials and the ethics of including placebo controls.

Endpoints for non-randomized trials

The panel discussed that in order to obtain (accelerated) FDA approval for an agent based on a single-arm, non-randomized study of unresectable or metastatic cancer, the efficacy of that agent must be assessed by a surrogate endpoint deemed reasonably likely to predict clinical benefit71 and a confirmatory randomized trial must be well underway at the time of accelerated approval71 (with the potential exception of true orphan diseases). However, it was recognized that conducting a randomized clinical trial in the same setting as the initial single arm trial, for example, where patients have already progressed on SOC, may be challenging, as there may be no SOC with which to compare. Furthermore, in later lines of therapy, the FDA may look for an OS endpoint, which can be challenging to obtain. And while imaging such as with Stein Fojo modeling72 that isolates and quantifies the behavior of the treatment-resistant fraction of the tumor can give much more “power per patient” in smaller, phase 2 studies, this approach does not scale up well.

Most panel members agreed that clinical response on imaging or, to a lesser degree, pathological response of injected lesion(s) and non-injected lesions, were the most important primary endpoints for non-randomized trials of intratumoral immunotherapy for unresectable/metastatic disease. The group pointed out that while it is necessary to establish local efficacy in this setting (particularly, eg, if the injected lesion is causing significant morbidity), it may not be sufficient. Response in non-injected lesions, as well as of overall disease, is particularly important to measure for metastatic disease, and duration and depth of response should be measured in addition to ORR, which can be misleading when reported in isolation. OS and PFS (as measured by an immune response criteria allowing for progression before response) were also discussed, although the group acknowledged that itRECIST and iRECIST are exploratory and regulatory agencies still require proof of efficacy via RECIST v1.1 (with the modification that progression before response should be allowed). A surrogate endpoint in phase 2 trials to determine efficacy in a different population than the phase 1 (safety/pharmacodynamics) population (eg, phase 1 population consists of pretreated patients, phase 2 population consists of treatment-naïve patients) was discussed, but ultimately if no efficacy is demonstrated in phase 1, it is difficult to support continuation of a drug into a phase 2 trial.

Endpoints for randomized trials

The panel agreed that OS, EFS (eg, PFS, disease metastases-free survival (DMFS), or potentially as a composite of progression, initiation of next line therapy, and death), and/or response in all lesions were the best primary endpoints for randomized trials of intratumoral immunotherapy in the unresectable/metastatic disease setting. The group discussed that a time-to-event endpoint is needed for approval, and that response in injected versus non-injected lesions should be measured separately with annotation for location (eg, liver). Response should be measurable by RECIST v1.1 and DOR should be measured as well. If PFS is to be used as an endpoint, then progression outside of the treatment field is especially important to measure. Of note, progression per RECIST v1.1 can be misleading with immunotherapy agents. Indeed, initial inflammation reactions to the intratumoral therapy can increase draining LNs above the 15 mm size limit or convert distant microscopic lesions to measurable, and therefore, count as a new lesions and classify the patient as a progressor although those can be due to transient inflammation reactions. Thus, initial progression before response should be allowed and confirmation of progression required when assessing both ORR and PFS.

Sample collection timing

The panel agreed that while the timing of tissue collection depends on the agent being studied, ideal times to collect samples included at screening (for tissue, to avoid confounding TRAEs and to avoid interfering with the first injection) and at the time of or just prior to each injection (for blood). The group discussed functional (eg, CD8-based or granzyme-based) imaging, for example, to coordinate sample collection with peak CD8 expansion73; even in just a subset of patients these data would be informative as the optimal time for biopsy varies widely based on disease and therapy. Clinical markers (eg, at time of best response and time of progression) could also be used to guide tissue acquisition. Several post-treatment time points for tissue collection were discussed, including monthly intervals up to 1 year, 1–3 weeks status post first injection (with a risk of necrosis and/or patient refusal beyond this point), and later time points if feasible to capture durable inflammatory immune responses. Importantly, the group agreed that non-injected lesions should be biopsied as well, particularly when correlated with functional imaging, as this tissue can provide great insight into the response and mechanism of action of the drug. Patients would need a particularly thorough explanation about biopsy of non-injected lesions as part of a shared informed consent and decision-making process.