Real-world data from a molecular tumor board demonstrates improved outcomes with a precision N-of-One strategy

Next-generation sequencing (NGS) can identify novel cancer targets. However, interpreting the molecular findings and accessing drugs/clinical trials is challenging. Furthermore, many tumors show resistance to monotherapies. To implement a precision strategy, we initiated a multidisciplinary (basic/translational/clinical investigators, bioinformaticians, geneticists, and physicians from multiple specialties) molecular tumor board (MTB), which included a project manager to facilitate obtaining clinical-grade biomarkers (blood/tissue NGS, specific immunohistochemistry/RNA expression including for immune-biomarkers, per physician discretion) and medication-acquisition specialists/clinical trial coordinators/navigators to assist with medication access. The MTB comprehensively reviewed patient characteristics to develop N-of-One treatments implemented by the treating physician’s direction under the auspices of a master protocol. Overall, 265/429 therapy-evaluable patients (62%) were matched to ≥1 recommended drug. Eighty-six patients (20%) matched to all drugs recommended by MTB, including combinatorial approaches, while 38% received physician’s choice regimen, generally with unmatched approach/low degree of matching. Our results show that patients who receive MTB-recommended regimens (versus physician choice) have significantly longer progression-free (PFS) and overall survival (OS), and are better matched to therapy. High (≥50%) versus low (<50%) Matching Score therapy (roughly reflecting therapy matched to ≥50% versus <50% of alterations) independently correlates with longer PFS (hazard ratio [HR], 0.63; 95% confidence interval [CI], 0.50–0.80; P < 0.001) and OS (HR, 0.67; 95% CI, 0.50–0.90; P = 0.007) and higher stable disease ≥6 months/partial/complete remission rate (52.1% versus 30.4% P < 0.001) (all multivariate). In conclusion, patients who receive MTB-based therapy are better matched to their genomic alterations, and the degree of matching is an independent predictor of improved oncologic outcomes including survival.

N ext-generation sequencing (NGS) has allowed the identification of novel potential targets for patients with cancer. Examples of successful matching of tumor molecular alterations to cognate drugs include: NTRK inhibitors larotrectinib and entrectinib in multiple solid tumors with NTRK fusions 1,2 , ROS1 inhibitors entrectinib and crizotinib in nonsmall cell lung cancer (NSCLC) with ROS1 alterations 3,4 , and the FGFR inhibitor erdafitinib in FGFR-altered urothelial carcinoma 5 . Unfortunately, the majority of patients still do not benefit from single-agent targeted therapies, and most patients who do respond eventually develop resistance [6][7][8] .
Biologic factors that may limit responsiveness to matched targeted monotherapies include genomic heterogeneity and complexity, as well as the fact that advanced tumors often have unique (N-of-One) molecular profiles 9 . Moreover, it may be challenging to differentiate driver from passenger molecular alterations in tumors with complicated genomic portfolios [10][11][12][13][14] . Several lines of evidence suggest that optimized therapy may require a customized/personalized combinatorial approach 15,16 .
In order to facilitate implementation of a precision medicine strategy in the clinic, we initiated a multidisciplinary molecular tumor board (MTB) [17][18][19] . The MTB represents a vehicle to integrate a comprehensive review of the patient characteristics, including clinical history, imaging, pathology, laboratory results, and molecular profiling, in the presence of specialists from diverse medical fields with the expertise of basic and translational scientists and computational biologists. The goal of the MTB was to develop an N-of-One treatment plan that could be initiated by the patient's physician under the auspices of a master protocol, with the assistance of clinical trial coordinators/navigators and medication acquisition specialists to facilitate drug availability. Of note, the MTB served as an advisory board, with the final decision made by the patient's physician. Hence, our MTB reflects a realworld experience with molecular profiling and patient treatment in the context of an academic medical center. Herein, we present the outcome of 715 patients with advanced cancer presented at our MTB, and demonstrate that adherence to MTB recommendations in order to match patients with targeted therapies was associated with higher degrees of matching and improved outcomes.

Results
Patient characteristics. Among 715 patients with diverse malignancies, the median age was 61 years (range: 3-92 years), and 58.7% (N = 420) were women. Patients had advanced/metastatic disease, and the majority of patients had ≥2 prior therapies. The most common diagnosis was breast cancer (18% [129/715]), followed by colorectal cancer ( Table 1). The physician, per their clinical judgment of necessity, ordered molecular tests. Physicians often presented patients upon receipt of molecular results, regardless of whether or not progressive disease was apparent. In general, the physician did not change therapy unless there was progressive disease on the current therapy or the current therapy was not tolerable.
Overall, 429 patients were evaluable for therapy after MTB discussion. Patients were not evaluable mostly because they received no further therapy after the MTB discussion or because their therapy was not changed within six months after the MTB (see Supplementary Fig. 1).
Utilization of variety of molecular profiling laboratories. During the MTB discussion, all profiling reports were included for the discussion if the testing was performed at a CLIA-certified (clinical-grade) laboratory. Tissue NGS was performed on 646 patients at seven different laboratories. Blood-derived cell-free DNA (cfDNA) was evaluated in 309 patients at two laboratories. Additionally, mRNA expression analysis, immunohistochemistry (IHC) and immunotherapy-associated markers ( Table 3).

) (Supplementary
Among patients tested for targets associated with immunotherapeutic implications (N = 362), TMB was high in 5.7% (16/280) (definition of TMB-high varied from each laboratory, e.g., Foundation One defined it as ≥20 mutations/megabase while Caris defined it as ≥17 mutations/megabase). MSI was high in 3.2%  Matching patients to drugs was feasible after MTB discussion. Overall, 265 of 429 therapy evaluable patients (62%) were matched to at least one drug recommended by the MTB, including 86 of 429 patients (20%) matched to all recommended drugs including a combination therapy approach. The other patients (N = 164, 38%) received a physician's choice regimen (generally unmatched or low match) (Supplementary Methods and Supplementary Fig. 1).

Discussion
Along with advances in molecular profiling technologies, targeted therapy and checkpoint blockade have revolutionized outcomes for some patients. However, treatment response can be short lived and substantial numbers of patients have primary or acquired resistance [1][2][3][4][5]20 . Although a number of basket and umbrella trials have been developed 21,22 , challenges include low patient enrollment, which may be related to strict eligibility criteria. With the goal of improving patient management, multiple institutions are now implementing MTBs [17][18][19][23][24][25][26] . MTBs can facilitate clinical trial enrollment and following the MTBs recommendation may improve outcomes [23][24][25][26] .
The MTB experience at the UCSD Moores Cancer Center for Personalized Therapy demonstrated that evaluation of different clinical-grade testing modalities (including tissue NGS, cfDNA, mRNA and IHC [including immune and chemotherapy biomarkers]) was feasible and facilitated the MTB discussion (Supplementary Table 1). Types and frequencies of genomic alterations identified were similar to previous reports (Fig. 1 Fig. 1).
Physicians were permitted to choose which therapy they considered best for their patients regardless of MTB discussion, which was considered advisory. Similar to the previous report by Hoefflin et al. 23 , patients who received an MTB-recommended regimen had significantly improved PFS and OS when compared to patients who received a physician's choice regimen (Fig. 2). Patients whose physicians adhered to MTB recommendations were more likely to receive matched targeted therapies that covered a larger fraction of their tumor's molecular alterations, yielding a high matching score, which may explain the improved clinical outcome ( Table 6). Consistent with this notion, significant improvement in PFS and OS were observed along with the step-wise increase in matching score (Fig. 3). Notably, patients who received therapy with high (≥50%) matching score  Sixty of the 80 patients receiving immunotherapy were matched based on profiling.
CI confidence interval,  (Table 2 and Fig. 3c). Overall, our MTB experience suggests that greater degrees of matching of tumors to drugs, including with customized N-of-one recommended combinations, was independently associated with better outcomes. These results are consistent with previous preliminary work from our group 27 as well as from our prospective trial-Investigation of Profile-Related Evidence Determining Individualized Cancer Therapy (I-PREDICT) [NCT02534675]) trial, where 73 patients with treatment-refractory solid tumors were treated with a combination-based strategy using their unique tumor genomic signatures 15 . Patients whose genomic aberrations were targeted with high matching scores demonstrated significantly better clinical outcomes including response rate, PFS, OS when compared to the lower matching group. Similar results were seen in the WINTHER trial (NCT01856296) where patients were navigated to therapy on the basis of DNA as well as RNA profiling 16 . Further investigations with larger sample size are required to determine if certain combination approaches are more efficacious than others. Of note, in the current study, there were 10 patients who had single gene alterations that were matched with targeted therapies, and 8/10 such patients (80%) attained clinical benefit including N = 5 partial responses (PFS: 8, 11 + , 19, 20+ and 36 months) and N = 3 durable stable disease ≥6 months (PFS: 11 + , 15 and 22+ months) (with "+" indicating ongoing responses). Hence, single gene targeting may be sufficient to achieve clinical response among patients harboring a limited number of genomic alterations.
There are several obstacles to the implementation of MTBs: (i) MTBs require multidisciplinary expertise; smaller practices may benefit from collaborations by remote video conference/virtual tumor boards 28 ; (ii) lack of access to drugs or clinical trials that limits patients being treated with the recommended regimen; clinical trials need to be available locally and expanded use of existing anticancer therapies is needed as leveraged by the Drug Rediscovery protocol (DRUP) developed in the Netherlands 29  There were several limitations to our current study. First, it was not a randomized controlled trial but rather reflects a real-world experience. Second, the number of cancer types included in the study was based on the treating physicians who requested MTB discussion, thereby predisposing to selection bias. Third, molecular analysis was obtained at various time points in relationship to the clinical history. Lastly, differences in molecular profiling platforms may have affected the detection of targetable markers.
In summary, our MTB successfully facilitated the interpretation of multiple testing modalities including tissue NGS, cfDNA, mRNA and IHC. Patients whose physicians followed the MTB discussion recommendations received therapy that was better matched to their alterations and achieved significantly better clinical outcomes, including longer PFS and OS when compared to patients who received physician's choice regimen. Moreover, patients who received treatment that matched a larger fraction of identified molecular anomalies (i.e., had higher matching scores) did better, and the degree of matching was an independent predictor of improvement in all outcome parameters, including OS, in multivariate analysis. These data are consistent with previous observations indicating that patients with advanced/metastatic disease often have complex and distinct molecular alterations 11,12,14,30 that require Nof-One matched treatments 15 , rather than standard monotherapies or unmatched therapy based on non-biomarker based population trials. Further clinical investigation is warranted in order to validate these findings, as well as to determine if there are matching score thresholds that determine the utility of precision therapies.

Methods
Molecular tumor board. The molecular tumor board (MTB) (face-to-face) meetings were held for 1-1.5 h,~3 times/month at the UCSD Moores Cancer Center. Cases to be discussed were submitted by the treating physicians. A handout prepared by the MTB project manager included a meeting agenda, de-identified patient information (age and sex, physician's name, diagnosis and date of diagnosis, last treatment, biopsy site and date, molecular test used, molecular profile results, and comments), and a copy of the key parts of the molecular diagnostic report. The manager also facilitated screening laboratories for certification and assisted physicians with ordering tests and helped with obtaining consent when requested.
A senior and a mid-level medical oncologist experienced in clinical trials, genomics, and immunotherapy moderated the meeting. The MTB attendees included Table 3 Association between patient and treatment characteristics and clinical benefit rate (SD ≥ 6 months/PR/CR) (N = 405*). Patients. In the current study, electronic medical records were reviewed for patients' characteristics and outcome, for individuals presented at the MTB between December 2012 and September 2018. This study followed the guidelines of the IRB-approved UCSD-Profile Related Evidence Determining Individualized Cancer Therapy (PREDICT) study (NCT02478931, https://clinicaltrials.gov/ct2/ show/NCT02478931) and any investigational studies for which the patients gave consent. (Details of included/excluded patients are shown in Supplementary Fig. 1.) NGS of tissue and cell-free circulating tumor DNA (cfDNA). Tissue and blood NGS was conducted in one of several CLIA certified laboratories (see Supplementary  Table 1) (182 to 596 genes in tissue panels and 54 to 74 genes for blood-derived cfDNA), depending on the laboratory and time frame. mRNA and protein expression analysis (including for immune markers) were also evaluated in selected patients (Supplementary Table 1 Matching score. All NGS pathogenic variants (but no variants of unknown significance [VUS]) were included in the matching score calculation as previously described 15 . However, protein or RNA were only included in the calculation when they were targeted. Briefly, the matching score evaluated the number of pathogenic alterations targeted by drugs given divided by total number of pathogenic alterations: the higher the score, the better the match (0%, unmatched; 100%, completely matched). For example, if a tumor harbored eight genomic alterations and the patient received two agents that targeted four of these alterations, the score would be 50% (4 of 8). Investigators blinded to patient outcomes determined the scores. Since there can be heterogeneity between blood and tissue samples or between two tissue biopsies, if a patient had more than one NGS or other biomarker report, the alterations in each report were counted; however, if the two reports were from the same laboratory and of the same type (but from a different blood or tissue sample), the sample closest to the time of MTB was used.
Reporting summary. Further information on research design is available in the Nature Research Reporting Summary linked to this article.

Data availability
All data generated or analyzed during this study are included in this published article (and its supplementary information files).