Intratumoral Immunotherapy: From Trial Design to Clinical Practice

Over the last decade, the advent of immune checkpoint targeted therapies (ICT) with mAbs directed at CTLA-4, PD-1, and PD-L1 have revolutionized the field of oncology. First, after 50 years of therapeutic efforts at trying to destroy cancer cells, the fact that immune-targeted therapies could put patients with relapsing/refractory metastatic cancer into durable remissions has been an incredible paradigm shift in terms of therapeutic strategies and how we conceive cancer as a disease (1).

Three dominant approved strategies are now competing on how to best use these new therapies. The first strategy is to combine anti-PD(L)1 antibodies with chemotherapy. This strategy keeps the usual systemic toxicities of chemotherapies, might increase the proportion of patients who could respond to the treatment and could provide benefits in progression-free survival. However, the median duration of responses obtained upon anti-PD(L)1 combinations with chemotherapies is close to the median duration of responses obtained with chemotherapies alone (e.g., which is significantly shorter than the median duration of responses obtained upon ICTs without chemotherapy. Indeed, in the randomized phase III keynote-048 trial, the median duration of response (mDoR) in patients with head and neck squamous cell carcinoma was approximately 7 months in the pembrolizumab + chemotherapy combination versus approximately 4 months in the chemotherapy alone group versus approximately 23 months in the pembrolizumab alone group (2). In non–small cell lung cancers, the mDoR obtained with pembrolizumab + chemotherapy was 7.4 months versus 5.4 months for chemotherapy alone whereas pembrolizumab alone has mDoR at 10.4 months in previously treated patients, and 23.3 months in previously untreated patients (3, 4). In triple-negative breast cancer randomized phase III trial IMpassion130, the mDoR was 7.4 months in the atezolizumab–nab-paclitaxel combination trial, versus 5.5 months with placebo–nab-paclitaxel whereas the mDoR remained “not reached” for atezolizumab or pembrolizumab without chemotherapy in the same disease population (5–7). The rationale supporting such combination comes from the demonstration that some chemotherapies can induce an immunogenic cell death (ICD; ref. 8). However, the in vivo rationale for chemotherapy-induced ICD is based on preclinical mouse models where chemotherapies were injected directly into the tumors of mice thereby preventing the systemic toxic effects of chemotherapies on white blood cells (9, 10). Indeed, the high dose-intensity chemotherapy regimens that are conventionally done in humans are cytotoxic for white blood cells and might therefore hamper the duration of immunotherapy responses. Chemotherapies can also have immune modulatory properties depending on their dose and regimen, which could diminish the immunosuppressive activities of tumor-infiltrating regulatory T cells or myeloid-derived suppressor cells (11). Therefore, new combinations of chemotherapies with ICT with lower chemotherapy doses and a maximum of one or two cycles might eventually show benefits both in terms of objective response rates but also in terms of DoR.

The second strategy currently pursued in the clinic is to combine anti-PD(L)1 with anti-CTLA4 antibodies. This strategy improves the response rates obtained with anti-PD(L)1 alone, generates long durability of responses, and might be even more efficient when increasing the dose of anti-CTLA-4 (12, 13). However, they expose patients to a significant level of grade 3–4 immune-related adverse events (irAE) which are also dependent on the systemic dose of anti-CTLA-4 (14–16). Indeed, in the recently published update of the randomized phase III Checkmate 067 trial, the level of Common Terminology Criteria for Adverse Events grade 3–4 irAEs was at 23% for nivolumab (anti-PD-1), 28% for ipilimumab (anti-CTLA-4), and 59% for the combination of the two drugs (14).

The third strategy is to combine anti-PD(L)1 antibodies together with systemic antiangiogenic compounds (17). This strategy has shown promising activity in the first line treatment of renal cell cancers with a combination of intravenous pembrolizumab (anti-PD-1) with oral axitinib [antiangiogenic tyrosine kinase inhibitor (TKI)], and in the first-line treatment of hepatocarcinomas with a combination of intravenous atezolizumab (anti-PD-L1) together with intravenous bevacizumab (anti-VEGF mAb; refs. 18, 19). Noteworthy, the duration of responses do not seem to be affected by the addition of antiangiogenic compounds so far in those early reports.

An alternate path forward is currently under strenuous investigation: intratumoral (IT) delivery of immunotherapies. Indeed, the number of clinical trials dedicated at testing such therapeutic strategy is currently booming and has been supported recently by the FDA approval in 2015 of the intratumoral immunotherapy Talimogene laherparepvec (T-VEC), an oncolytic virus (OV; Fig. 1A). Although the types of intratumoral immunotherapies was limited in the past to intratumoral Bacillus Calmette-Guerin (BCG) and intratumoral cytokines (e.g., IL2, IFNa; ref. 20), this clinical research effort concerns now multiple types of innovative immunotherapies from immunomodulatory nucleic acids, proteins, pathogens, and cell therapies (Fig. 1B; refs. 20–23).

The aim of this article is to provide clinical investigators, academic or industrial clinical trial sponsors, and regulatory agencies with a comprehensive presentation of the critical points which need to be addressed when considering developing intratumoral immunotherapies.

The principle of human intratumoral immune therapies (HIT-IT) is to deliver immunostimulatory products directly into a tumor lesion (primary or metastatic) to prime and/or boost an antitumor immune response. The aim of HIT-IT is to generate a systemic immune response against the tumor upon local immune stimulation thanks to the circulation in the blood and lymph of immune cells and antibodies. Therefore, the objective of such therapeutic strategy is to use the tumor as its own vaccine, and generate a polyclonal adaptive immune response mediated by T and/or B cells against preexisting tumor-specific antigens or tumor-associated antigens.

The advantage of the HIT-IT strategy over regular cancer vaccines is that it uses off-the-shelf immune stimulatory products and requires neither pretreatment molecular target identification nor human leukocyte antigen patient selection. Therefore, HIT-IT is more of a universal therapeutic strategy rather than a personalized cancer vaccine approach (20). The other theoretical advantage of such approach is that it lets the immune system reacting against the most immunogenic epitopes of a given tumor lesion. Indeed, cancer cells accumulate mutations over time and the mutational profile varies across metastasis (24–26). The immune reactivity against those diverse cancer cells epitopes also varies across tumor lesions (27–29) Therefore, injections of multiple tumor sites could better address the issue of the heterogeneity of cancer cells by simultaneously targeting multiple cancer cells epitopes while providing durability of tumor responses thanks to the memory features of the adaptive immunity.

At this stage of scientific and clinical knowledge, we cannot exclude that some immune therapies might only have an immunostimulatory effect on innate immune cells and a therapeutic effect limited to the injected tumor lesions. However, the safety and activity of such limited type of treatment could still be valuable for specific clinical situations (e.g., preoperative sarcomas, multiples skin metastasis, oligometastatic elderly patients,…).

Many intratumoral immunotherapies have now been tested at the preclinical level, including nucleic acids (e.g., encoding mRNAs or as DNA/RNA sensor agonists; refs. 30–32), proteins (e.g., cytokines or mAbs; ref. 33), cells (e.g., dendritic cells or CAR-T cells; ref. 34), virus (e.g., OV or viral vectors for gene therapy; refs. 35, 36), bacteria (e.g., BCG; ref. 37), … (see also reviews in refs. 21, 23). These approaches have been developed preclinically in immunocompetent murine models, with either transplantation of syngeneic murine cancer cell lines or sometimes with carcinogen-induced tumors or oncogene-driven spontaneous tumors. Most of them have reported to generate potent antitumor immune responses with tumor growth delays, or shrinkage and even eradication of the injected tumor lesions.

Interestingly, some tumor models are sensitive to a simple monotherapy, for example, intratumoral TLR9 agonist (38), or intratumoral agonistic anti-CD40 mAbs (39, 40) or peritumoral antagonistic anti-CTLA-4 mAbs (41, 42). Some tumor models are resistant to monotherapies but become sensitive to combinations of intratumoral immunostimulatory products (e.g., tumor response to only triple combinations of agonistic anti-CD137 + antagonistic anti-PD-1 + antagonistic anti-CTLA-4; ref. 43). Depending on the type of immunotherapy injected in situ, the target cells could be different, but not necessarily located in the tumor microenvironment (TME). Indeed, local injections could allow for the immunotherapy to diffuse in the stroma and around the tumor where, for instance, tertiary lymphoid structures or T cells could be located. Intratumoral injections are also the best way of optimizing the therapeutic index of an immunotherapy not only in the tumor but also in the tumor-draining lymph nodes (44). In those murine models, such local treatments have been shown to generate also systemic protective antitumor immunity. Of note, the eradication of the injected tumor lesion usually comes together with tumor-specific protection against the same tumor rechallenge.

Some local therapies have the ability to generate abscopal effects against distant noninjected tumor sites. Depending on the tumor model and type of treatment, these abscopal effects can range from a tumor growth delay to a complete eradication of the noninjected tumor sites (38, 45). These responses against distant noninjected tumor lesions are tumor specific, and intratumoral immunotherapy do not generate abscopal responses against an irrelevant distant tumor type (45). Most interestingly is the ability of some of these intratumoral immunotherapies to overcome the resistance to ICT therapies. Indeed, many mouse syngeneic tumor models do not respond to systemic anti-CTLA4 or anti-PD(L)1 therapies, either alone or even in combinations. But some intratumoral immune stimulation can overcome and synergize with such ICT therapies. This has been shown notably upon intratumoral immunotherapies with pattern recognition receptor agonists (PRRa) such Toll-like receptor (TLR) agonists (45–47) or stimulator of IFN genes (48, 49). This has also been shown with multiple intratumoral OV strategies (50–52). However, it is not clear yet whether this effect rely on the oncolytic properties of the virus or to its PRRa effects. Indeed, we and others have shown that inactivated OV lose their oncolytic properties in vivo but keep their ability to synergize with ICT antibodies through the stimulation of type I IFN pathways (51, 53). Overall, these models could be useful to better understand the mechanisms of resistance to anti-PD(L)1/anti-CTLA4 and allow for an effective translation into the clinic.

Most of the abscopal responses reported so far in humans have been obtained with physical methods such as irradiation, cryotherapy, or radiofrequency (54, 55). Such abscopal responses are rare, limited mostly to case reports, but dedicated prospective trials are now being developed and reported (49 active clinical studies registered in clinicaltrials.gov as of June 2020; refs. 56–58). All these studies refer to a putative role of the immune system reacting against the tumor upon such physical destruction or alteration of the tumor (59, 60). Therefore, direct immune stimulation against the tumor with immunomodulatory compounds injected in situ is a logical step forward. The first FDA/European Medicines Agency (EMA)-approved intratumoral immunotherapy is an herpes-derived OV genetically modified to express human GM-CSF called T-VEC. It has shown objective abscopal responses and overall survival benefits in stage IIIb-IVM1a melanoma (61, 62). Other OV such as Pexa-Vec, a vaccinia-derived OV also encoding for human GM-CSF, have shown their ability to generate abscopal responses in hepatocarcinomas (63). Most interestingly is the potential of intratumoral OV to synergize with systemic immune checkpoint blockade as shown in preclinical data (50, 51). For instance, some early clinical results in metastatic melanoma have shown superior activity of systemic anti-CTLA4 when combined to intratumoral OV with objective response rates at around 50% (where systemic anti-CTLA4 alone is usually below 20%; ref. 64). The same trend has been observed with anti-PD1, and is now tested in a randomized phase III clinical trial (NCT02263508; ref. 65). Besides OVs, intratumoral oncolytic peptides have also been reported to have the ability to generate de novo T-cell responses and clinical activity in patients with melanoma and sarcoma (66, 67). Intratumoral pathogens or pathogen extracts (bacteria and mycobacteria) have been tested in the past to treat several types of cancers (22, 68, 69). More recently, synthetic molecules mimicking the properties of pathogens' extracts (so-called pathogen-associated molecular patterns) with the aim of stimulating PRRs have been developed in patients with refractory/relapsing advanced cancers and have shown monotherapy activity (70–72). Many of these intratumoral immunotherapies are currently developed with systemic ICTs (see refs. 23, 73 for reviews of ongoing trials). This type of intratumoral + intravenous combinations requires particular attention and therefore specific wording. Indeed, a tumor response in a noninjected lesion could in that situation be the effect of the systemic therapy and not from an abscopal effect (Fig. 2). Such types of responses in distant noninjected lesions should therefore be called “anenestic” (from the Greek “noninjected”) and not abscopal as recommended here (74).

The most frequent question about intratumoral immunotherapy relates to its feasibility. However, if oncological units are able to put an 18G (1.02 mm) needle into a tumor lesion in order to establish a cancer diagnosis, they should also be able to put a much thinner 22G (0.64 mm) needle into a lesion to inject some immunotherapies in it. Therefore, the actual question is not about the feasibility of injecting an immunotherapy product into a lesion but rather about how often we need to inject such product in order to obtain a clinical benefit for the patient.

The second most frequent question about HIT-IT relates to its acceptability by patients. In our experience, it is extremely rare that a patient refuses a treatment targeted to where his disease is located, as far as we can guarantee that the maximum will be done to perform a quick and seamless procedure without any pain.The third criticism against HIT-IT relates to its practical implementation. As discussed further in those lines, the expertise for putting a needle into a tumor lesion is already in place in the vast majority of hospitals making oncology diagnosis. Therefore, the issue is more about the logistics (including access to sophisticated imaging units) and how to get interventional radiologists involved into such HIT-IT clinical trials and later on clinical practice procedures.

Last but not least, a usual criticism from pharmaceutical companies used to be the inability to register an intratumoral drug product. The recent approval of T-VEC both by FDA ad EMA has now modified this false perception.

In fact, the actual challenges for HIT-IT are related to defining which pharmacokinetic/pharmacodynamic markers could help to determine the appropriate dose and regimen of the intratumoral immunotherapeutic product. Other challenges include the fact that conventional trial designs (e.g., 3+3) aiming at establishing the MTD and dose-limiting toxicities (DLT), together with the historical RECIST1.1 radiological assessment criteria, might no longer be valid for intratumoral immunotherapies (Table 1).

The first objective of a conventional first-in-human clinical trial in oncology is to establish the safety of a new therapeutic product and to determine the MTD. This is usually done by following a stepwise dose escalation within a 3+3 clinical trial design (3 patients per dose level) where investigators aim to determine for each dose level if patients are developing DLTs (75). This approach has now been widely validated for systemic therapies (either oral or intravenous) where the expected volume of distribution of the therapeutic product is the whole body (± the central nervous system).

For intratumoral immunotherapies, there is an additional constraint which is the volume of the injected tumor lesions. Indeed, the aim of HIT-IT is to maximize the local bioavailability of an immunotherapy. But the volume of tumor lesions is highly variable within a given patient and across patients. Also, this volume increases exponentially with the size of its diameter (Fig. 3).

Therefore, when considering a dose-escalation trial for HIT-IT, there are three options to manage such dose escalation. First, is to work at every dose level with a fixed volume of drug to be injected and increase the concentration of the product at each dose level (Fig. 4, left column).

The alternate strategy is to decide to work with a fixed concentration of the therapeutic product and increase the volume of injection at each dose level (Fig. 4, middle). For these two first strategies, we do not take into consideration that the tumor volume to be injected might be variable across patients and that the volume of injection will not be adapted to the size of the tumor lesion. However, with these two strategies, patients from the same dose level will receive the same local concentration of the therapeutic product.

A third option would be to adapt the volume of injection to the volume of the tumor lesion to be injected (Fig. 4, right). This strategy would allow to fill the volume of the injected lesions and to expose all the tumor cells to the therapeutic product. However, the local concentration (and potential bioactivity) of the therapeutic product will be variable across patients, even across patients from the same dose level.

It is important to note that for these three therapeutic designs, the same amount of therapeutic product can be delivered at each dose level. However, the local bioavailability of the product might be different across patients and dose levels whether we reason in terms of volume of injection versus concentration to the injected product.

Depending on the therapeutic product, one strategy could be preferred out of the other two ones. Indeed, if the therapeutic product has a wide range of equivalent bioactive concentrations (e.g., OVs), then the aim would be fill the entire volume of the injected lesions and adjust the volume of injection to the size of the injected lesions. Alternatively, if a therapeutic product has a clear concentration/efficacy relationship, then one of the two first strategies could be preferred.

The issue of the variability of tumor volumes across patients could be, in those settings, controlled by limiting the tumor lesions eligible for intratumoral injections. A maximal volume of tumor lesions eligible for IT injections could then be set at 3–4 cm in diameter (i.e., tumor volume <30 mL). Also, when the patients presents with small tumor lesions (e.g., skin metastases <1 cm in diameter), it can be recommended per protocol to split the volume of injection to several tumor lesions with the aim of filling the volume of a maximum number of tumor lesions. Also, protocol writers need to take into consideration the fact that deep-seated tumor lesions need at least 1–1.5 cm in diameter to secure the injections under radiological guidance (74).

The historical development of cytotoxics in oncology and their dose-dependent toxicity profile have driven the conventional 3+3 designs of phase I trials in oncology. However, when working with immunotherapies, there might not be a correlation between dose and toxicity [e.g., anti-PD-(L)1 antibodies: no correlation between dose, toxicity and efficacy has been found within their early phase trials; refs. 76, 77]. Moreover, at high dose, some immunostimulatory products might actually trigger physiologic/homeostatic negative feedback loops with pharmacodynamic readouts showing a bell shape curve effects rather than a linear correlation between dose and activity (e.g., for STING and CD40 agonists; refs. 78, 79). Also, at high dose, some immunotherapies could lead to the internalization of the therapeutic target and therefore hamper the bioactivity of the therapeutic product.

Therefore, during an HIT-IT dose escalation, the pharmacodynamic readouts are critical to determine the optimal biological dose (OBD) rather than the MTD. Because most of the on-target activity is expected to happen in the tumor, tumor (rather than blood) pharmacodynamic readouts might be critical for OBD determination. Therefore, pretreatment and on-treatment tumor biopsies of injected lesions (± noninjected lesions for distant effects assessments) are mandatory to determine what is the OBD of a given immunostimulatory product.

Also, to generate scientifically relevant datasets and have evidence-based decisions for recommended phase II doses (RP2D), having only 3 patients (i.e., 3 data points) per dose level is insufficient (especially when considering the level of noncontributing tumor biopsies). Therefore, it is recommended to pick a rolling-six trial design (80), or to allow to backfill previous dose levels which have been cleared for safety in the classical 3+3 designs. An alternate way to proceed would be to have an approach in the spirit of the modified toxicity probability interval trial design which would not only take into consideration safety but also pharmacodynamic readouts (81).

Another inheritance of cytotoxics in oncology is the notion of DLTs during dose-escalation trials to determine the RP2D. Most current immunotherapies have not shown DLTs (e.g., ICT antibodies,…) in their first in human studies (82–84). Also, when patients develop severe adverse events (SAE) when treated with cytotoxics, the toxicity is rather homogenous across patients (e.g., decrease in white blood cell counts) and is clearly related to the dose of the treatments (e.g., dose adjustments for subsequent cycles of chemotherapies). For many immunotherapies, the SAEs that patients are developing are mostly related to the patients' genomic or immunologic background rather than the dose of the immunotherapy (e.g., wide diversity of autoimmune SAEs seen in patients treated with antiimmune checkpoint mAbs across dose levels; refs. 85, 86) Therefore, investigators shall be cautious when concluding about the relationship between a SAE and the actual dose of the injected product. Again, an increased number of patients per dose level should help to determine such dose/toxicity relationship. Also, investigators should take into consideration differentially the toxicities that are related to the investigational product versus the ones that are caused by the needle of the intratumoral injection procedure (e.g., pneumothorax or bleeding immediately following the intratumoral injection).

Most of the treatment regimens in oncology have been inherited from past clinical developments of cytotoxic therapies: every 2 or 3 weeks for chemotherapies or daily intake for oral medications (chemo or TKIs). The pharmacokinetic profiles can vary greatly from one immunostimulatory products to another (e.g., half-life of 5 days for anti-PD-L1 avelumab vs. 3 weeks for anti-PD-L1 atezolizumab vs. <2 hours for intravenous GM-CSF; refs. 87–89). Also, those pharmacokinetic profiles might be very different depending on the route of administration (e.g., half-life of GM-CSF becomes 20 hours when administered subcutaneously; ref. 87). The pharmacodynamic effects can also vary greatly across families of immunostimulatory products, from minutes (e.g., cytokine peaks upon treatment with CAR-T cells or pattern recognition receptor agonists) to weeks (e.g., T-cell infiltrates in tumor biopsies upon treatment with ICT therapies). Also the expected pharmacodynamic readouts could be different in injected tumor lesions (e.g., recruitment and activation of antigen-presenting cells, development of tertiary lymphoid structures) versus distant noninjected tumor lesions (e.g., increase in cytotoxic T cells, decrease in cancer cell Ki67, necrosis; ref. 74). Therefore, the treatment regimen for HIT-IT trials should rely on strong preclinical rationale and be validated by the pharmacokinetic/pharmacodynamic profiles seen in patients during the phase I trial. This means that several regimens could be explored within the same trial to determine what would be the best treatment regimen.

Also, protocol writers should take into other aspects when choosing a drug regimen with HIT-IT. First the fact that many immunostimulatory products might face tachyphylaxia, which means that the ability to stimulate the immune system could decrease over time, either because of the level of drug exposure (e.g., downregulation of OX40 on T cells beyond 40% of target occupancy; ref. 90) or related to epigenetic modifications in immune cells upon repeated drug exposures (e.g., chromatin changes upon TLR agonist therapy; ref. 91). Second, depending on the therapeutic product, and especially for biologics, the risk of generating antidrug antibodies might impair the bioavailability of the therapeutic product after several days or weeks (92).

Although systemic blood pharmacokinetic assessments shall not drive the development of intratumoral immunotherapies they can sometimes be monitored notably to assess the level of systemic exposure and the potential contribution to the activity in noninjected tumor lesions. Also, those systemic blood pharmacokinetic assessments could provide an order of magnitude on the kinetics of the leakage of the intratumoral immunotherapy toward the systemic blood compartment and extrapolate on the duration of exposure of the immunotherapy in the TME.

The pharmacodynamic assessments are more critical to determine the ability of an intratumoral immunotherapy to activate the immune system. As above mentioned, this assessment should be done on tumor biopsies, using either transcriptomic, epigenetic of proteomic readouts. But HIT-IT can also generate systemic pharmacodynamic readouts (e.g., transcriptomic changes in whole blood, circulating cytokines levels, phenotype of white blood cells,…) which could be really useful to demonstrate the ability of intratumoral immunotherapies to generate systemic immune responses (93).

To perform intratumoral immunotherapies, the patients should present with injectable tumor lesions (i.e., having adequate tumor size) that are safe to be injected (i.e., excluding some risky tumor locations). Injectability criteria should be clearly defined in HIT-IT protocols: accessibility of lesions, injectable lesion sizes, superficial versus deep-seated lesions, specific precautions for lesions close to large blood vessels, close to hollow organs, beneath the liver capsule or the pleura, concomitant medications and risk of hemorrhagic complications (e.g., anticoagulants),… Some expert recommendations have already been published elsewhere on those topics (74) and have been summarized in Table 2.

A critical challenge for HIT-IT clinical trials is their ability to provide reliable assessments of treatment efficacy and clinical benefit to patients. Indeed, the conventional RECIST criteria are not adapted to the atypical patterns of response seen with immunotherapies such as pseudoprogressions, mixed responses, or transient draining lymph node swelling (Fig. 5; ref. 94).

Also iRECIST criteria cannot capture the efficacy seen in injected tumor lesions as, per RECIST spirit and like for irradiation, these treated lesions should not be chosen as target lesions (95). However, a HIT-IT strategy which would be able to put only injected lesions in durable complete remission (CR) could be of great therapeutic value for some oligometastatic or symptomatic patients and therefore challenge or be complementary to other local therapies (e.g., irradiation of cryoablation). As previously recommended, HIT-IT trials shall eventually be able to report separately the objective response rates seen on injected versus noninjected tumor lesions (74). This achievement requires that, per protocol, the diameters of injected and noninjected lesions are separately collected prospectively in the Case Report Form (CRF) of the trial.

Recently, a consensus work has generated guidelines for an intratumoral RECIST (itRECIST; ref. 96). These guidelines should be implemented in future HIT-IT trials to harmonize the way HIT-IT clinical results are reported. Briefly, those guidelines follow the historical RECIST guidelines by recommending to first defining measurable and nonmeasurable tumor lesions. Then, within measurable lesions, the guidelines recommend to define target and nontarget lesions. Finally, the itRECIST guidelines require to define injected and noninjected tumor lesions for both target and nontarget tumor lesions. Two important rules have been formulated in those guidelines. The first rule is that noninjected lesions (either target or nontarget) could become injected per protocol during the treatment course if they are enlarging or if no other lesion is available. The second rule is that a nontarget lesion (either injected or noninjected) cannot become a target lesion and vice versa. The itRECIST guidelines still offer the opportunity to provide an overall RECIST 1.1 response rate by pooling the information collected (see ref. 96 for more details).

Also, it is critical for the pharmacodynamic and imaging assessments to correctly collect the information regarding which tumor lesion has been biopsied and/or injected versus not biopsied and/or not injected. Moreover, it is important to realize that the efficacy (or tumor progression) of an HIT-IT strategy can be captured prior the protocol tumor assessments. Indeed, during every intratumoral injection, the operator can capture data (visual, ultrasound, or CT scan) about the status of the injected lesion (notably response or progression) earlier than the protocol assessments. This data could also be collected in the clinical trial CRF (e.g., ultrasound measurement of lesion diameter at every timepoint of injection). However, it has to be clearly defined within the protocol which assessment (predefined CT scans timepoints or imaging assessments along the cycles of injection) will be used to potentially calculate the volume/dose of injection per lesion and assess the efficacy of the treatment. Indeed, there are always discrepancies between CT scans and ultrasound size assessments and this discrepancy should be admitted rather than generating queries in the CRF of the clinical trial. Last but not least, PET scan assessments have been shown to provide valuable information for the assessment of immunotherapy efficacy (97, 98). In the context of intratumoral immunotherapies, PET-scans could help to decipher atypical tumor responses especially between the metabolic activities of injected versus noninjected tumor lesions.

Some techniques of intratumoral injections and medical devices for intratumoral administrations have been already reported (63, 99). We currently do not know whether a specific method of intratumoral injection is better than another to efficiently stimulate the antitumor immune response or to better deliver in situ immune stimulatory products.

Within HIT-IT protocols, rules for injections (needle size, volume, flow, pressure, single, or multiple injections in the same tumor) must be clearly stated to avoid too much variability between operators but also reported in the final publication of the results to allow comparisons of techniques and strategies.

Also, rules for the management of injected lesions shall be defined within the HIT-IT protocols. Should investigators keep injecting the same tumor lesions until complete disappearance (i.e., inject the same site with no tumor visible on imaging) or should they aim to inject as many lesions as possible across treatment cycles? Indeed, injecting multiple tumor lesions at different timepoints could have a prime-boosting effect, that is, amplifying preexisting antitumor immunity and priming a response against nonshared neoepitopes arising from new cancer cell clones (Fig. 6). This strategy could better address the clonal heterogeneity of metastatic cancers. But it needs to be specified within the protocol and collected in the patient CRF.

In addition, rules for lesion prioritization could be useful to test specific hypothesis (activity within lymph node metastasis vs. liver metastasis, for example). Indeed, if not specified, investigators will always aim at treating the most convenient (usually superficial) tumor lesions, and will strive to inject that same lesion at every cycle until a radiological CR is obtained. However, we currently do not know whether a specific tumor location (e.g., skin metastasis vs. liver metastasis) would best ignite the antitumor immune response upon intratumoral immunotherapy (Fig. 7). Also, knowing the evolution status of the injected lesion (progressive or stable) before the injection could be of value for better understanding the outcome.

Now that an intratumoral immunotherapy has become standard of care for metastatic melanoma (100), many oncology hospitals might face operational issues in order to efficiently treat their patients.

Most (if not all) oncology hospitals have interventional radiologists, radiologists, and endoscopists that are able to stick a 18G needle into a tumor lesion to perform a biopsy and establish a cancer diagnosis. Therefore, these same experts could be involved into therapeutic rather than diagnosis gestures by injecting intratumoral immunotherapies with even thinner needles (e.g., 22G). Therefore, to develop intratumoral immunotherapies, oncologists need to involve their colleagues from radiology, interventional radiology, and endoscopists in their clinical practice and HIT-IT clinical trials. However, such collaboration will not be possible in every hospital because of absence of such expertise on site, or minimal available time from these experts, or because of a lack of resources to fund those intratumoral activities (e.g., nurses, assistants, study coordinators,…). Also, the cost of such intratumoral procedures and the inability to treat patients in an outpatient setting are additional factors which could limit the implementation of intratumoral immunotherapies in the routine practice.

The first approved intratumoral immunotherapy is a pathogen (herpes derived OV) that has been genetically modified (GMO) to encode for human GM-CSF (100). GMOs and pathogens might generate regulatory constraints in terms of treatment preparation, product handling, staff exposure and patient management. Therefore, for that specific product but also for upcoming immunotherapy clinical trials using pathogens and/or GMOs as therapeutic products, oncologists need to anticipate the fact that their pharmacists, head of nurses, and hygienists need to be involved ahead of time.

Nurses also need to be trained for the specifics of HIT-IT to allow for a practical implementation of the therapeutic strategy, collaborating on the treatment procedures, monitoring, and taking care adequately of the patients (e.g., pain/stress assessment and treatment prior, during and after the intratumoral procedure) and anticipating the potential adverse events (hypotension, fever,…).

Last but not least, the evaluation of the efficacy of an HIT-IT strategy can only be accurate and relevant if the radiologists that are performing the radiological assessments have been precisely informed of which tumor lesions have been injected or not injected. Therefore, the dialog with diagnostic radiologists and their involvement in the therapeutic decision making is essential.

Obviously, all these organizational and logistical aspects should be taken into consideration by sponsors during their site selection of HIT-IT clinical trials.

For the practical implementation of HIT-IT clinical trials, a dedicated study coordinator is required to be fully aware of the specifics of the HIT-IT protocols and variability of endpoints across them. This person will also be critical to make sure that the data regarding injected/biopsied tumor lesion is correctly generated and reported in the patient CRF (precise location, size, number of biopsies, timings, etc.). The study coordinator will act as a bandmaster to organize regular HIT-IT board meetings (Fig. 8).

The HIT-IT board meetings needs to gather on a regular basis (e.g., once a week) the interventional radiologist, the radiologist, the nurse and the oncologist that are involved in the intratumoral immunotherapy clinical trials. These meetings will be the place to discuss upcoming trials and assess their feasibility, to agree on radiological targets, define and prioritize the injected versus noninjected tumor lesions, both during the screening period for eligibility assessment but also during the study for safety and efficacy assessments.

In case of academic HIT-IT trials where ancillary studies need to be run on fresh samples, having a scientist, engineer, or technician involved in the HIT-IT board meetings is also mandatory to make sure that the lab team is ready to process those fresh samples when the patients will be injected and collected.

Last but not least, the involvement of the hospital pharmacy is critical and there are specific considerations that are important to anticipate from a sponsor point of view. First, in terms of drug preparation, if the protocol allows (or requires) that the operator can inject several tumor lesions at a given treatment cycle, it will be important to determine whether the pharmacy shall prepare “X” syringes of “Y” mL (one per lesion) or a single syringe of “X*Y” mL (one syringe for all the lesions to be injected). Sometimes, the drug prescription will end up being an order to the pharmacy to provide bulk product and syringes and the operator will eventually be the one to both prepare the syringe with the right volume and proceed to the injection and deliver the amount of drug in accordance with the drug prescription. Second, depending on the diameter of the needles that are used for injection and their length, the dead volume needs to be anticipated by both the pharmacy and the operator to make sure that the right volume/dose of product has been correctly prepared/administered. Third, some injected product may flow back at the injection point making difficult to assess if the total dose has been correctly administered into one or several tumor lesions. Altogether, sponsors should realize that the exact monitoring (and subsequent deviations) of drug administration cannot be as stringent as for intravenous or oral deliveries.

Intratumoral immunotherapy is a novel treatment strategy that is currently thriving both in terms of number of clinical trials and by the diversity of immune products that are tested into patients with relapsing/refractory cancer. Its promise is to generate more efficacy and less toxicity of immunotherapies by leveraging a better therapeutic index upon in situ immune stimulation against cancer cells and use the tumor as its own vaccine. As it is the case for any new drug in oncology, most if not all the current developments are performed in the setting of patients with metastatic relapsing/refractory cancer. One important prospect is that neoadjuvant HIT-IT strategies will soon challenge neoadjuvant or adjuvant systemic immunotherapies as illustrated by the lower relapse rate of melanomas after neoadjuvant treatment with intratumoral OV or TLR agonists (101, 102). Indeed, as we are now treating many patients bearing local cancers with systemic ICT antibodies we are now facing the issue of exposing patients to severe acute or chronic irAEs although the conventional treatments could have been sufficient to cure them. The number of patients needed to treat with an ICT antibody to avoid a cancer relapse versus the number of patients who will develop a grade 3/4 irAE but could have been cured by conventional therapies is not currently taken into consideration. The dramatic increase in overall cancer survival postimmunotherapies will strengthen the patient advocacy in the near future. Now that we are advanced in the drug development of systemic immunotherapies for patients with metastatic and localized cancer, these toxicity/efficacy concerns shall influence more and more the drug development strategies in oncology to the benefit of HIT-IT strategies (Fig. 9), notably in the context of the routine use of neoadjuvant and adjuvant conventional cytotoxic treatments. Radiotherapy is of special interest for combinations with HIT-IT strategies for its ability to stimulate the immune system locally while preventing a systemic toxicity on white blood cells (103). Indeed, recent data support such irradiation plus HIT-IT combinations to prime systemic antitumor immune responses (93, 103, 104). However, the ideal dose/regimen for building on the immune stimulatory properties of radiation remains unclear and might depend on the cancer histology (105, 106).

Altogether, HIT-IT strategies with or without conventional cytotoxics (notably irradiation) could be ideal to prime safely a systemic antitumor immunity upon intratumoral injection of localized cancers while preventing their relapse after surgery. This endeavor of developing neoadjuvant HIT-IT strategies is currently under way (107). However, this clinical development will also raise specific questions such as determining the timing of IT injections prior to the surgery, the impact of resecting the draining lymph nodes, etc. Dedicated trials will be needed to answer those important questions.

S. Champiat reports personal fees from Amgen, AstraZeneca, Bristol-Myers Squibb, Janssen, MSD, Novartis, and Roche; other from Abbvie, Adaptimmune, Aduro Biotech, Agios Pharmaceuticals, Amgen, Argen-X Bvba, Arno Therapeutics, Astex Pharmaceuticals, AstraZeneca Ab, Aveo, Basilea Pharmaceutica International Ltd., Bayer Healthcare Ag, Bbb Technologies Bv, Beigene, Blueprint Medicines, Boehringer Ingelheim, Boston Pharmaceuticals, Bristol-Myers Squibb, Ca, Celgene Corporation, Chugai Pharmaceutical Co., Clovis Oncology, Cullinan-Apollo, Daiichi Sankyo, Debiopharm, Eisai, Eisai Limited, Eli Lilly, Exelixis, Forma Therapeutics, Gamamabs, Genentech, GlaxoSmithKline, H3 Biomedicine, Hoffmann La Roche Ag, Imcheck Therapeutics, Innate Pharma, Institut De Recherche Pierre Fabre, Iris Servier, Janssen Cilag, Janssen Research Foundation, Kyowa Kirin Pharm. Dev, Lilly France, Loxo Oncology, Lytix Biopharma As, Medimmune, Menarini Ricerche, Merck Sharp & Dohme Chibret, Merrimack Pharmaceuticals, Merus, Millennium Pharmaceuticals, Molecular Partners Ag, Nanobiotix, Nektar Therapeutics, Novartis Pharma, Octimet Oncology Nv, Oncoethix, Oncopeptides, Orion Pharma, Ose Pharma, Pfizer, Pharma Mar, Pierre Fabre, Medicament, Roche, Sanofi Aventis, Sotio A.S, Syros Pharmaceuticals, Taiho Pharma, Tesaro, and Xencor [as part of the Drug Development Department (DITEP), principal/subinvestigator of clinical trials]; grants from AstraZeneca, Bristol-Myers Squibb, Boehringer Ingelheim, Janssen Cilag, Merck, Novartis, Pfizer, Roche, and Sanofi [as part of the Drug Development Department (DITEP)]; and nonfinancial support from AstraZeneca, Bayer, Bristol-Myers Squibb, Boehringer Ingelheim, Medimmune, Merck, NH TherAGuiX, Pfizer, and Roche [as part of the Drug Development Department (DITEP); drug supplied] during the conduct of the study. L. Tselikas reports grants from Bristol-Myers Squibb Foundation and Terumo, and personal fees from AMGEN during the conduct of the study, as well as personal fees from Boston Scientific, Guerbet, and Sirtex, and personal fees and nonfinancial support from GE Healthcare outside the submitted work. C. Robert reports personal fees from Bristol-Myers Squibb (advisory board), MSD (advisory board), Novartis (advisory board), Roche (advisory board), Sanofi (advisory board), Pierre Fabre (advisory board), CureVac (advisory board), and Biothera (advisory board) during the conduct of the study, as well as personal fees from Lutris (advisory board) outside the submitted work. T. de Baere reports grants from Terumo (grant for preclinical research) and personal fees from H3 Biomedicine (advisory board), AstraZeneca (advisory board), and Johnson & Johnson (advisory board) during the conduct of the study. A. Marabelle reports grants, personal fees, nonfinancial support, and other from Bristol-Myers Squibb (drug supply for academic trial); grants from Fondation MSD Avenir; personal fees, nonfinancial support, and other from MSD (drug supply for academic trial); personal fees and nonfinancial support from Roche; personal fees from Pfizer, Lytix, and EISAI; other from Idera (drug supply for academic trial); grants, personal fees, and nonfinancial support from AstraZeneca; and grants, personal fees, and other from Transgene (drug supply for academic trial) outside the submitted work. No disclosures were reported by the other authors.

The intratumoral immunotherapy activities from Gustave Roussy are performed within the Centre d'Investigation Clinique BIOTHERIS and receives funding support from DGOS & INSERM (CIC BT 1428).