Introduction

Over the past two decades, the deployment of gene- and immune-targeted therapies has revolutionized the therapeutic landscape for both solid and hematologic malignancies. Historically, using one gene-targeted agent to impact one molecular alteration led to meaningful clinical benefit for patients, but this strategy has its limitations, as resistance usually develops. Even so, the exploitation of next-generation sequencing (NGS) to identify potential actionable biomarkers has informed novel treatment strategies for patients with cancers and led to a multitude of gene-targeted approvals by The Food and Drug Administration (FDA). NGS has also identified immunotherapeutic biomarkers, including tumor mutational burden (TMB), mismatch repair deficiency (dMMR), and microsatellite instability (MSI)1,2, which can be targeted with immune checkpoint inhibitors (ICIs).

Strategies using ICIs have generally been used as single agent and usually in later lines of therapy. Over the last several years, efforts to include immunotherapy in the treatment-naïve setting as well as in combination with gene-targeted therapies has offered higher response rates as well as higher complete remission rates3. However, despite the improvements in response rates and survival outcomes, many of these studies do not include a biomarker for inclusion4 and a majority of patients do not benefit.

Evidence that certain genomic alterations lead to increased expression of various immunoregulatory molecules is of interest for combining gene- and immune-targeted therapeutics5. Moreover, in cancers such as melanoma, FDA approval of atezolizumab (programmed cell death ligand 1 [PD-L1] inhibitor), cobimetinib (MEK inhibitor), and vemurafenib (BRAF V600E inhibitor) for advanced BRAF V600-mutated melanoma illustrates the value of combining these modalities6. This approach is an example of using a genomic biomarker for inclusion (BRAF V600E); however, there was no specific inclusion criteria for patients to receive atezolizumab. Indeed, rarely do trials include a biomarker for immunotherapy and it is even less common to perform trials to include a biomarker for both a targeted therapy and an immunotherapy when both drugs are combined4. This is especially important since not all patients derive clinical benefit and the potential risk for side effects from immunotherapies is both significant and able to occur at any time point after receiving therapy7.

We sought to identify patients who were presented at The University of California, San Diego (UCSD) Molecular Tumor Board (MTB) and were given immunotherapy combined with gene-targeted therapies based on their cancers’ immune and genomic markers. Herein, clinical outcomes including responsiveness, survival, and toxicities from dual-matched therapies were assessed.

Results

Patient characteristics

Among 429 patients who were assessable for therapeutic clinical outcome after MTB discussion8, 17 patients were eligible for the present analysis. Patient flow diagram is provided in Supplementary Fig. 1. The median age was 67 years (range 20–86), and 9 (53%) of them were female. Approximately 30% of the patients (29% [5/17]) received three or more lines of therapy before undergoing dual-matched therapy. Treatments were administered between October 2016 and August 2018. The most common cancer diagnosis was gastroesophageal cancer (24% [4/17]) and hematologic malignancies (24% [4/17]), followed by bladder/ureter cancer (18% [3/17]) and gynecologic cancer (18% [3/17]). Upon treatment recommendation by MTB, 5 of 17 patients (29%) received all recommended agents, and the other (71% [12/17]) received part of MTB recommended agents (Supplementary Table 1). All patients with solid tumors had advanced metastatic disease (i.e., not oligometastatic), and all patients with lymphoma showed diffuse organ involvement.

Each patient had a biomarker-based rationale for the utilization of both gene-targeted agents and ICIs

The biologic rationale for the selection of administered agents is outlined in Table 19,10 and depicted in Fig. 1, based on molecular profiling by NGS and IHC. Each patient had at least one predictive immune biomarker for ICIs (Fig. 1): Eight patients had only PD-L1 IHC as an immune biomarker; four had positive PD-L1 IHC along with other immune markers; and five had only markers other than PD-L1 IHC that were positive (e.g., high TMB or MSI-high or ARID1A alterations). As to genomic aberrations, the patterns of aberrations were mostly distinct from patient to patient. Only two out of 17 patients had the same genomic aberration patterns (ERBB2 amplification), although their co-positive immune markers were different. Nivolumab (anti-PD-1) was the most frequently used ICI (59% [10/17]). Figure 2 illustrates the targeted agents utilized for treatment and their combination patterns with ICIs. Detailed information regarding genomic aberrations, overexpression, TMB, MSI status, and PD-L1 IHC is provided in Supplementary Table 2.

Fig. 1: The rationale for the dual-matched therapy with ICI and targeted agent.
figure 1

This heatmap illustrates gene aberrations and the corresponding targeted agents utilized for treatment. Each column within the heatmap represents an individual patient. For instance, patient #1 was treated with vismodegib for PTCH1 mutation. Additionally, patient #1 exhibited positive immune markers, including TMB ≥ 10 (muts/Mb), an ARID1A alteration, and MSI-high, and thus, received immunotherapy. ICI immune checkpoint inhibitor, IHC immunohistochemistry, Mb megabase, MSI microsatellite instability, muts mutations, TMB tumor mutation burden.

Full size image
Fig. 2: Utilization of genomically targeted drugs and immune checkpoint inhibitors in the context of dual-matched therapy.
figure 2

The upper segment of the figure depicts the types of the immune checkpoint inhibitors implemented in the dual-matched therapy. Nivolumab and pembrolizumab are anti-PD-1 agents; atezolizumab is an anti-PD-L1 agent. The lower portion of the figure depicts individualized boxes representing the targeted agent(s) used in dual-matched therapy. The background colors of molecular targeted agent(s) denote the simultaneously administered immune checkpoint inhibitor with the same background color (blue, nivolumab; yellow, pembrolizumab; orange, atezolizumab). This tailored combination approach was customized for each patient’s unique and complex molecular profile. Patients could receive a single drug that targeted more than one biomarker (since small molecule inhibitors often have multiple targets) or more than one drug that targeted the same biomarker.

Full size image
Table 1 Characteristics of 17 patients who underwent dual matched targeted agent and immune checkpoint inhibitor simultaneously
Full size table

Response to dual-matched therapy

Clinical benefit (defined as stable disease (SD) ≥ 6 months as well as complete or partial response (CR or PR)) were seen in nine of 17 patients (53%): one, with complete response (6% [1/17]); three, had partial response (18% [3/17]), five, stable disease ≥6 months (29% [5/17]). Eight patients had progressive disease (47% [8/17]). Overall, clinical benefit was observed in 53% [9/17] of the patients (Fig. 3A).

Fig. 3: Clinical outcomes of the dual-matched therapy.
figure 3

A Clinical benefit rate (SD ≥ 6 months/CR/PR) of the 17 patients who underwent dual-matched therapy was 53% (9/17). B The swimmer’s plot provides information on the duration of survival and the timing of progression for each patient. The arrow indicates that the patients were progression-free at the date of their last follow-up. Two patients did not show apparent progression, but their therapy was suspended due to an adverse event (ID 8) or lost to follow-up (ID 17). C, D Kaplan–Meier curves for progression-free survival (PFS, Panel C) and overall survival (OS, Panel D) of 17 patients who received dual-matched therapy for treatment-refractory diverse cancers. Median follow-up period was 8.1 months (range 2.1–62.1). The pale red area indicates the 95% confidence interval (CI). Median PFS was 6.1 months (95% CI, 2.9–not estimable), and median OS was 9.7 months (95% CI, 6.7–not estimable). CNS central nervous system, CR complete response, PD progressive disease, PR partial response, SD stable disease.

Full size image

Survival analysis

Figure 3B displays patients’ time to progression and last follow-up. Median follow-up period was 8.0 months (range 2.2–62). Three patients (ID 6, 10, 17) had a durable PFS (more than 18.3 months [three times the median PFS]) (PFS of 33.0, 59.7 and 23.4+ months) and four patients (ID 1, 6, 10, and 17) had a prolonged survival (more than 19.3 months [two times the median OS]) (OS of 37.7+, 43.6, 62.1+ and 23.4+ months). Three patients with durable response in both PFS and OS (ID 6, 10, 17) survived more than 3 years after the initiation of dual-matched therapy. Kaplan–Meier curves showed that median PFS was 6.1 months (95% confidence interval [CI], 2.9–not estimable [NE]; Fig. 3C) and median OS was 9.7 months (95% CI, 6.7–NE; Fig. 3D).

Dosing of the dual-matched therapy and adverse events

We assessed the actual dose of each administered agent as a percentage of the FDA-approved dose. Dose percentage of ICI (N = 17) administered was 100% in all patients, and that of gene-targeted agents was 50% (median) (range, 25–100%) (Supplementary Table 3; and Fig. 4); physicians routinely started with a lower dose of the gene-targeted agent to avoid toxicity for the combination. Serious adverse events (SAEs), at least possibly related to the dual-matched therapy, occurred in four patients (24% [4/17]): one had Grade 4 (encephalitis and sepsis), while the other three had Grade 3 (acute kidney injury [N = 1], hypothyroidism [N = 1], pancytopenia [N = 1]). All four were hospitalized, but none of the cases were fatal. The causes of hospitalization included acute kidney injury, which necessitated a 75% dose reduction of cabozantinib (ID 2); hypothyroidism combined with poor compliance in taking levothyroxine, resulting in the termination of pembrolizumab (ID 8); sepsis and encephalitis, leading to the discontinuation of the therapy (ID 12); and pancytopenia, which led to the suspension of palbociclib with recovery of blood counts (ID 14). During their course of therapy, six patients (35% [6/17]) had to reduce their treatment doses.

Fig. 4: Dose percentage of each administered agent.
figure 4

This heatmap describes the initial dose percentage of each drug given. Each cell contains dose percentage with its corresponding color. Each row indicates each patient, and each column indicates each administered agent. Dose percentage of immune checkpoint inhibitors (N = 17) was 100% in all patients, and that of targeted agents (N = 22) was 50% (median) and ranged from 25% to 100%. Patients whose ID are with an asterisk had serious adverse events (Grade ≥3) during their treatment course.

Full size image

Trial search identified only a small portion of trials that employed biomarkers for both gene-targeted therapy and immunotherapy in patient selection

Our initial search (conducted March 24, 2023; ClinicalTrials.gov) yielded a total of 1702 registered clinical trials. Based on screening and one additional record identified through expert input, 314 trials appeared to be eligible for further evaluation (see Methods) and were assessed in detail (Fig. 5A). Among these 314 trials included for detailed assessment, only 1.3% (4/314) of trials employed a biomarker for patient inclusion for both the ICI and the gene-targeted agent (Fig. 5B). Targeted genes were HER2 in two trials, KRAS G12C in one trial, and BRAF V600E in one trial. All four trials assessed immune markers: three assessed the PD-L1 positivity for patient inclusion (one trial requires PD-L1 TPS ≥ 1%, and two trials with any PD-L1 positivity) while one required MSI-high/dMMR. Meanwhile, 75% (235/314) of the trials did not assess any biomarkers for patient inclusion. The details of four included trials are shown in Supplementary Table 4.

Fig. 5: Trial search results.
figure 5

A PRISMA flow chart of the trial search (See Methods). Search was conducted on Mar 24, 2023, for ClinicalTrials.gov, using the search terms “(PD-1 OR PD-L1 OR CTLA4 OR LAG3) AND (targeted OR combination)” with a start date of Jan 1, 2018. After the screening, 1389 records did not meet the inclusion criteria and were therefore excluded from the further assessment. B Proportion of trials that utilized biomarkers for both the immune checkpoint inhibitor and the gene-targeting agent. Of the 314-trials assessed, 235 trials (75% [235/314]) did not assess biomarkers for patient inclusion. Forty-six trials checked gene alterations or immunohistochemistry for the targeted gene. The details of the four trials that assessed biomarkers for both the immune- and the gene-targeted therapy are presented in Supplementary Table 4. ICI immune checkpoint inhibitor.

Full size image

Discussion

The present study demonstrated the clinical outcomes and safety of combined genomic and immune-matched therapy—dual-matched therapy—in a pan-cancer setting. Both targeted agents and ICIs were selected with a specific biomarker-based rationale in each patient. Despite multiple lines of previous therapy (29% [5/17] had ≥3 lines of therapy), the disease control rate (includes SD ≥ 6 months or objective response) was 53%. The 17 patients analyzed in the present study represents only a fraction (2.4%) of the total patients (N = 715) presented during the study period at MTB meetings8; however, our findings indicated that a certain proportion of patients may derive benefits from dual-matched therapy.

In the present analysis, we focused on patients who received biomarker-based targeted therapy and immunotherapy––the dual-matched therapy––upon MTB recommendations. The MTB took into account the fact that advanced cancers are individually distinct and complex when it comes to genomic and immune markers11, and facilitates the best treatment options with in-depth assessment of actionable biomarkers and other patient-specific information12. In particular, our prior studies show that targeting tumors with a higher degree of drug matching to alterations (reflected by a higher Matching Score) was associated with better clinical outcomes13,14. Overall, eight of 17 patients showed clinical benefit (53%); one patient attained a complete response; three, a partial response; and five had stable disease for 6 months or more. This is despite the fact that half of them (4/8) having had ≥3 lines of prior therapy. Targeted genes included CCND3, MYD88, NF1, KIT, ERBB2, NRAS, and PTCH1. Survival outcomes were comparable to those reported in observational studies evaluating second-line or later immunotherapy15,16. Among the eight patients with clinical benefit, four patients had durable survival exceeding 1.5 years.

Genomic alterations in metastatic cancers are diverse, distinct, and unique—akin to malignant snowflakes11. The present study is no exception; all 17 patients had unique genomic aberration patterns, which were matched in an N-of-1 approach (Fig. 1). Specific rationales to select treatment drugs are of the essence, and every drug given had a biomarker-based rationale (Fig. 2). In addition to genomic alterations, immunotherapy biomarkers were fully considered. In this analysis, PD-L1 IHC was positive in 71% (12/17) of patients17,18; TMB ≥ 10 (muts/MB), in 35% (6/17)19,20; ARID1A mutation, in 35% (6/17)21; and MSI-high in 12% (2/17)2,22. Of note, six patients had multiple positive immune markers; this highlights the importance of incorporating comprehensive immune analysis. Appreciation of the unique footprint of each cancer provides clues to effectively combat the disease itself. Molecular profiling can shed light on the precise evaluation of these footprints, and our data showcases an example of exploiting these clues across cancer types.

FDA-approved immune markers currently used to predict the response to anti-PD-1/PD-L1 therapy include PD-L1 IHC, TMB, and MSI status. Although a high PD-L1 IHC score generally predicts a positive response, as shown in this analysis, some patients with low PD-L1 expression still respond to immunotherapy17. In contrast, a high PD-L1 expression score does not guarantee a response to anti-PD-1 therapy, perhaps because of intratumoral heterogeneity. In the present study, 8 of the 17 patients were positive for PD-L1 IHC alone, without other immune markers; of these, 4—including one with a high PD-L1 IHC score (>25% TC)—did not attain clinical benefit. PD-L1 is used as one of the biomarkers for immunotherapy, but it has limitations17. Regarding MSI, only about 2% of all tumors are MSI-high23. High TMB is predictive of therapeutic efficacy, yet there is ongoing debate regarding the threshold for its clinical utility1,24,25. In the present study, the clinical decision to use immunotherapy was mainly based on such available markers. Indeed, these biomarkers have led to selection of patients who may potentially benefit from immunotherapy, as certain patients experienced a durable response. However, it is known that each tumor may have a unique portfolio of immune markers26. To further personalize cancer immunotherapy, it is essential to explore beyond existing biomarkers, including transcriptomic analysis using RNA sequencing27,28,29 and analysis of soluble immune markers30 to tailor immunotherapy treatments. With the expanding use of agents targeting CTLA-4, LAG-3, and TIGIT, a key question is whether biomarker-based selection will improve outcomes. As predictive markers remain limited for these newer targets, further data are needed to guide precision immunotherapy.

One patient with MSI-high and high TMB (36 mutations/Mb) did not respond to PD-1 blockade (ID 15). This patient harbored a truncating JAK1 mutation, potentially impairing interferon-γ signaling and contributing to resistance31 despite strong indicators of immunogenicity. This case highlights the importance of identifying genomic alterations that may confer resistance to ICIs, even in tumors with otherwise favorable biomarkers. Systemic aggregation and evaluation of such discordant cases may help refine patient selection.

While simultaneous combination therapy for patients with dual-positive biomarkers may appear rational, it remains unclear whether it consistently yields superior outcomes compared to sequential approaches. Given the relatively higher rates of ≥G3 toxicity observed with the dual-matched therapies (24%) and the possibility of reduced dosing of targeted agents (Fig. 4), sequential treatment strategies might allow for full-dose administration, possibly with fewer adverse events and without compromising effectiveness. Although historically it was combination therapies that led to cures in childhood lymphomas and leukemias32, further studies will need to address the benefits of combined versus sequential therapy.

Our clinical trial search revealed that only a small portion (~1.3%) of the 314 clinical trials found in ClinicalTrials.gov used biomarkers both for targeted agents and for immunotherapy in patient enrollment. Since the introduction of ICIs, the idea of augmenting the effectiveness of gene-targeted therapies with ICIs has naturally arisen, usually based on preclinical data. Various trials have been conducted, yet there has been little implementation of combined gene-targeted therapies based on clear biomarker-based evidence along with ICIs selected based on immunomic markers, as our systematic search showed. The lessons learned from our prior MTBs, and suggested herein, are that biomarker-based combination treatment may correlate with better outcomes than standard monotherapies or unmatched therapies typically based on non-biomarker-driven trials8.

Further studies with a basket design or platform design may be helpful in combining targeted and immunotherapy markers. Basket studies targeting mutations such as BRAF V600E, RET, NTRK, and immune markers such as high TMB or MSI-high have been reported. However, most studies are designed to impact one target at a time25,33. A platform designed trial could allow for patients to receive a targeted therapy up front based on a biomarker and if no response or progression and the tumor has an immune biomarker adding an immunotherapy to the regimen. Other such studies could try to compare patients with both immune and targeted therapy markers who receive a customized combination versus investigators choice of targeted therapy, immunotherapy, or broad-spectrum chemotherapy (e.g., carboplatin/paclitaxel, gemcitabine/cisplatin). A study such as the latter could help elucidate if there is a survival advantage to individualizing therapy for patients based on their genomic profiles. However, because drug availability for individual patients may vary and be limited, and because the biomarker field is moving forward quickly, a randomized trial may be difficult to operationalize.

One of the real-world limitations for MTB-directed therapy includes drug availability and its acquisition process. Limitations of the current study include lack of a control group, small sample size, retrospective nature of the analysis, and tumor heterogeneity. Given the period during which patients were enrolled in this study, potential biomarkers for ICI response such as interferon-γ-related signatures34 or melanoma-based nomograms35, as well as certain immune evasion-related gene alterations (e.g., B2M alteration), could not be incorporated. Despite the limitations, the strength of our study lies in the fact that molecular profiling was exploited to choose drugs. While our study was based on tumor-only panel data, we acknowledge that matched tumor-normal sequencing may enhance the interpretation of genomic data36, and that the integration of RNA sequencing may play an increasingly important role in future precision oncology assays12,36. Also, while our current framework excluded chemotherapy-based regimens and concomitant radiotherapy due to the lack of predictive biomarkers, future studies will need to evaluate their roles when used alongside targeted therapies. In particular, the emergence of antibody–drug conjugates emphasizes the need to re-evaluate the role of the chemotherapy backbone within the evolving framework of precision medicine37.

In summary, our data demonstrate the potential effectiveness of dual-matched therapy, using a biomarker-based approach for combining gene-targeted agents and ICIs. Even so, there has been limited use of biomarkers for patient inclusion in recently conducted clinical trials, per our clinical trials search. Further prospective trials should be initiated in order to further elucidate how biomarker-based treatment selection can enable optimization of individualized combination cancer therapy.

Methods

Patient selection

From December 2012 to September 2018, 715 therapy-evaluable patients with diverse treatment-refractory cancers were presented at Molecular Tumor Boards at UC San Diego Moores Cancer Center8. Among these, patients who received ICIs combined with gene-targeted agents based on the following rationale were included in this analysis: (i) harboring TMB ≥ 10 Muts/Mb, MSI-high, positive PD-L1 immunohistochemistry (IHC), or an ARID1A gene alteration21; and (ii) harboring gene alteration(s) or protein overexpression(s) corresponding to the targeted agents used for (e.g. HER2 positive (ERBB2 mutations/amplifications or immunohistochemistry 3+) being targeted by anti-HER2 therapy). The expression level of PD-L1 with IHC was evaluated by the SP142 assay (Ventana Medical Systems, Inc., Tucson, Arizona, USA) or the 22C3 assay (Dako North America, Inc., Carpinteria, California, USA). PD-L1 status was decided based on the proportion of tumor area occupied by PD-L1-expressing tumor-infiltrating immune cell (% IC) or the percentage of PD-L1 expressing tumor cells (% TC) when evaluated with the Ventana PD-L1 SP142 assay, and was deemed positive if either percentage was ≥1%. For the 22C3 assay, tumor proportion score (TPS) or combination proportion score (CPS) was used, which we deemed positive if either score was ≥1%. All NGSs and IHCs were performed at Clinical Laboratory Improvement Amendments (CLIA)-certified laboratories. All investigations were conducted in accordance with the guidelines of the Institutional Review Board of University of California, San Diego for data collection (Study of Personalized Cancer Therapy to Determine Response and Toxicity, UCSD_PREDICT, NCT02478931) and for any investigational interventions for which patients or their legal guardians provided written informed consent. The study protocol was approved by the Institutional Review Board of University of California San Diego (UCSD Project 130794), and informed consent was obtained from all participants involved in the study. All procedures were carried out in accordance with the principles of the Declaration of Helsinki.

Definition of matched therapy

In this study, “matched” targeted therapy using a tyrosine kinase inhibitor (TKI) or other small molecule inhibitor and/or a monoclonal antibody (mAb) was defined as administration based on target gene alterations by NGS or overexpression detected by NGS or IHC. Matched immunotherapy with ICI was defined as administration of ICI for positive PD-L1 by IHC, TMB ≥ 10 (muts/MB), MSI-high, or ARID1A aberrations21. “Dual-matched” combination of targeted therapy and immunotherapy was defined here as simultaneous administration of both agents with a biomarker for both the targeted therapy and the ICI.

Dosing information and assessment of safety

We defined dose percentage as the proportion of actual dose given compared to the FDA-approved dose. If a single drug has several approved doses amongst cancer types, we used the highest approved dose as a denominator. Adverse event severity was assessed using the Common Terminology Criteria for Adverse Events (CTCAE) v4.038.

Search for clinical trials that investigate “dual-matched” approach targeting both genomic and immune markers

We conducted a database search to identify ongoing or completed clinical trials investigating a combination therapy of immunotherapy and targeted therapy for patients with neoplasms. We searched US National Institutes of Health Ongoing Trials Register (ClinicalTrials.gov) during the period from 1 Jan 2018 to 24 Mar 2023, employing the search terms “ (PD-1 OR PD-L1 OR CTLA4 OR LAG3) AND (targeted OR combination).” The yielded records of registered clinical trials were screened by the authors, using Rayyan39. Clinical trials investigating the simultaneous combination of targeted agents and ICIs were included. The exclusion criteria were: (i) trials incorporating concomitant radiotherapy or systemic cytotoxic chemotherapy in their treatment regimens, even if immunotherapy and targeted therapy were administered simultaneously; (ii) trials administering an investigational immunomodulatory drug, even if they assessed a combination with currently available ICIs and targeted drug(s); and (iii) trials without assessment of both genomic/protein and immune analysis for patient enrollment.

Outcome endpoints and statistical analysis

All patients were evaluated for the treatment response in accordance with RECIST 1.1 criteria. Progression-free survival (PFS) was defined as the interval between the date of treatment initiation and the earliest occurrence of disease progression or death from any cause. Overall survival (OS) was calculated as the duration between the date of treatment commencement and the last recorded follow-up or death. Data cutoff date was June 30, 2023. Patients still progression-free (for PFS) or alive (for OS) at time of last contact or data cut-off date, whichever was earlier, were censored at that time point. Durable response was defined as a patient with a PFS that exceeded three times the median PFS of all patients; durable or prolonged OS was defined as survival that exceeded two times the median OS of all patients in this cohort (per the definition of Pons-Tostivint et al.)40. A two-sided p value < 0.05 was deemed statistical significance. All statistical analyses were performed using R 4.2.3 (R Foundation for Statistical Computing, Vienna, Austria) and Python 3.10.5.