Routman, David M.

Abstract Objective To describe the presenting symptoms, treatment regimens, and outcomes associated with oropharyngeal adenoid cystic carcinoma. Data Sources Ovid MEDLINE(R), Ovid Embase, Ovid Cochrane Central Register of Controlled Trials, and Scopus via Elsevier. Review Methods This review followed the Preferred Reporting Items for Systematic Reviews and Meta‐analyses. A medical librarian conducted the literature search using controlled vocabulary and keywords, with investigator input. Eligible articles were clinical trials and observational studies reporting on clinical features, treatment, or outcomes for oropharyngeal adenoid cystic carcinoma. Results Our search identified 233 unique citations, with 44 articles included covering 195 patients. The female‐to‐male ratio was 1.7:1, and the base of the tongue was the most common subsite (73%). Presenting symptoms were mainly pain/discomfort (n = 31), followed by a palpable mass (n = 19) and dysphagia (n = 12). At diagnosis, 22.3% (29 patients) had positive lymph nodes. Treatment was primarily surgical (n = 102), with definitive radiation therapy used in 58 cases. Disease progression was noted in 86 patients, 36% (n = 31) of whom developed distant metastases—primarily to the lungs (n = 26). Long‐term outcomes (n = 76) showed 3‐ and 5‐year overall survival rates of 71.5% to 100% and 47.6% to 79.0%, and disease‐free survival rates of 54.7% to 70.0% and 21.9% to 44.0%, respectively. Conclusion Among patients with available data, oropharyngeal adenoid cystic carcinoma appeared more common in women and has a predilection for the base of the tongue. However, this finding should be interpreted with caution, given the high proportion of missing data. These tumors appear to carry a higher risk for nodal metastases and may lead to poor outcomes.

Abstract Large Language Models (LLMs) demonstrate strong capabilities in healthcare applications. However, recruiting human evaluators to assess usability and effectiveness becomes challenging in the large-scale LLM-generated outcomes. Existing metrics are often outdated for evaluating LLM-based healthcare technologies. This paper proposes a Research through Evaluation (RtE) approach that can refine metrics and improve assessments of LLM-generated outputs. We deploy RtE in a retrospective comparative study of prostate cancer patient inquiries, comparing responses from human clinical care teams with those generated by a GPT-4-based in-basket bot. Through three rounds of co-evaluations with clinical professionals and an LLM-as-graders study with four LLMs, the RtE method shows that starting with small, iterative evaluations helps human graders adjust metrics to suit their daily practices better. Additionally, in the LLM-as-graders study, LLMs exhibit similar behavior, preferring the in-basket bot's responses over the clinical care team. Our methodology and findings demonstrate that the RtE method can enhance the transparency and robustness of evaluation metrics for LLM-generated content, with the potential to generalize to evaluations in other clinical practices.

1633 Background: CAR-T cell therapy has transformed outcomes of hematologic malignancies and is poised to expand into solid malignancies. However, broader use of CAR-T remains constrained by the need for timely cytokine release syndrome (CRS) detection. Current episodic monitoring may miss early physiologic changes. Wearable digital health technologies (DHTs), offering continuous monitoring and real-time alerts, could enable earlier CRS detection, further support outpatient delivery, and expand access to CAR-T and potentially other immunotherapies associated with CRS. Methods: This single-center observational study (NCT05018208) enrolled patients receiving FDA-approved CAR-T products in the outpatient setting at Mayo Clinic, Rochester MN. Management of post-CAR-T complications (e.g. CRS) followed the institution’s standard of care. Patients were monitored from day -5 to day +30 post infusion with intermittent remote patient monitoring (SoC) and an arm-worn wearable device for continuous monitoring (DHT). Patients were included in this analysis if DHT data were available for ≥ 5h of data/day or ≥ 2h of data/night on ≥2 days or nights for baseline and ≥1 day or night within 72h prior to CRS, or at least 2h within 4h of CRS onset, or if DHT was available for ≥7 days/nights post infusion with no CRS event. CRS detection time by DHT was defined as the onset of pyrexia or threshold-based deviation from an individual’s baseline pulse distributions. Results: 32 patients were included in this report (median age 64 years, gender 12 F/20 M, diagnoses 18 MM/14 NHL). 25 patients experienced a CRS event, with maximum CRS grades ranging from 1-3 (16 G1, 8 G2, 1 G3). 24/25 patients with CRS required tocilizumab/steroid. Median CRS onset time was 3.4 days (range 0.4-18.4). 12/32 patients had sufficient DHT data (7 CRS, 5 no CRS) for analysis. 5 patients with CRS had a DHT detection time in advance of the SoC by a median of 13h (range 2.8 - 69.2). The DHT prediction of CRS by baseline HR change was earlier than the first episodic fever by a median of 14.5h (range 13-77.5). The remaining 2/7 patients who did not have CRS detected by the DHT had missing DHT data for more than 45h before a treated CRS event. DHT detected CRS in 2/5 patients without CRS according to SoC. One patient had fever caught with intermittent monitoring. The other was treated for neutropenia fever with suspicion of CRS and/or infection. Conclusions: These results offer evidence that continuous DHT can identify CRS events earlier than intermittent monitoring (SoC). Despite limitations related to compliance with wearable device use and data missingness, and variability in clinical management of CRS, we highlight the usability of DHT and its potential for future studies with optimized strategies to address these limitations. Clinical trial information: NCT05018208 .

Importance Standard adjuvant therapy for human papillomavirus–associated oropharyngeal squamous cell carcinoma (HPV-OPSCC) is radiation therapy 60 Gy with or without cisplatin delivered in 30 fractions during 6 weeks. A phase 3 randomized clinical trial demonstrated decreased toxic effects and improved patient-reported outcomes with de-escalated adjuvant radiation therapy (DART) compared with standard therapy; however, outcomes for patients treated in the clinical practice setting following trial adoption remain unknown. Objective To compare characteristics and oncologic outcomes of patients treated with DART during clinical trials (on study) and in clinical practice settings (off study) and report progression-free survival (PFS) by risk category. Design, Setting, and Participants This retrospective cohort study included patients treated with DART in the MC1273 and MC1675 trials and in clinical practice after study completion at a multisite tertiary care center. Patients with surgically resectable HPV-OPSCC who underwent margin-negative resection between August 20, 2013, and March 21, 2023, and were treated with DART were included in the analysis. Those with prior head and neck cancer, pT0 or pT4 disease, unknown HPV status, metastatic disease at presentation, synchronous primary tumors, induction chemotherapy, or treatment at an outside institution were excluded. Data were analyzed from July 1, 2024, to July 18, 2025. Interventions DART consisting of 30 to 36 Gy in 1.5- to 1.8-Gy fractions twice daily with docetaxel 15 mg/m 2 on days 1 and 8. Main Outcomes and Measures Progression-free survival as measured via Kaplan-Meier analysis. Results A total of 282 patients were included in the analysis: 176 (62.4%) in the on-study group and 106 (37.6%) in the off-study group (252 [89.4%] male; mean [SD] age, 59.1 [9.3] years). Both groups were predominantly male (157 [89.2%] and 95 [89.6%], respectively) and consisted mostly of never smokers (118 [67.0%] and 62 [58.5%], respectively). Pathologic features were similar between groups except for nodal stage (pN1: 155 [88.1%] on study vs 102 [96.2%] off study; pN2: 21 [11.9%] on study vs 4 [3.8%] off study; P = .02). The median follow-up times were 5.2 years (95% CI, 5.2-5.8 years) in the on-study group and 2.2 years (95% CI, 1.9-2.4 years) in the off-study group. Two-year PFS was similar (93% [95% CI, 88%-97%] on study vs 93% [95% CI, 88%-99%] off study). For patients with pN1 and no extranodal extension (ENE), 2- and 5-year PFS were 97% (95% CI, 93%-100%) and 96% (95% CI, 92%-100%), respectively. Patients with pN1 and ENE had a 2- and 5-year PFS of 93% (95% CI, 89%-98%) and 87% (95% CI, 81%-94%), respectively. Conclusions and Relevance In this retrospective cohort study of patients receiving DART on and off study, oncologic outcomes were similar between the 2 groups. Fewer patients with pN2 received DART off study, given the interim results of MC1675. Patients with pN1 and no ENE treated with DART had excellent 2- and 5-year PFS, supporting DART as a compelling treatment option in well-selected patients. An additional study that is currently underway will evaluate biomarker risk-stratified treatment of patients with pN1 and ENE receiving DART.

Abstract Purpose: Circulating tumor fraction estimate (ctFE) is a machine learning–derived composite metric of circulating tumor DNA (ctDNA) burden. We hypothesized that pre-treatment and early on-treatment ctFE could robustly risk-stratify patients with locally advanced (LA) and oligometastatic non–small cell lung cancer (NSCLC) treated with radiotherapy (RT). Experimental Design: In a prospective phase II clinical trial (NCT03916419), 26 patients with unresectable stage IIB to IIIC NSCLC received magnetic resonance (MR)-guided hypofractionated chemoradiotherapy (chemoRT) followed by immunotherapy. Plasma ctDNA was profiled at baseline and mid-treatment (days 10–14) to derive ctFE and maximum variant allele frequency (Max VAF). A burden-based ctFE threshold derived from baseline samples was applied unchanged to mid-treatment samples and validated in two external cohorts: LA NSCLC treated with chemoRT (LA-RW; n = 94) and oligometastatic NSCLC treated with RT (OM-RW; n = 309). Results: Pre- and mid-treatment ctFE burden strongly stratified overall survival (OS) and progression-free survival (PFS), outperforming Max VAF and ctFE detectability. The baseline ctFE was prognostic for OS (HR 5.93, P = 0.005) and PFS (HR 11.08, P < 0.001) and remained significant at mid-treatment (OS: HR 7.08; PFS: HR 12.06; both P < 0.001). Early ctFE dynamics defined three molecular response groups with OS separation (median OS 60.8 vs. 13.0 vs. 2.9 months; P < 0.001). ctFE remained associated with survival in both validation cohorts. Conclusions: Early ctFE derived from a clinically available, tumor-naïve ctDNA assay enables noninvasive risk stratification in LA and oligometastatic NSCLC treated with RT, supporting its use as a practical biomarker for precision treatment adaptation.

Management of locoregionally advanced human papillomavirus-positive oropharyngeal squamous cell carcinoma (HPV(+) OPSCC) can include induction chemotherapy followed by definitive chemoradiation. The standard induction regimen of cisplatin, docetaxel, and 5-fluorouracil (TPF) is associated with high toxicity. Given the chemosensitivity of HPV(+) OPSCC, platinum doublets are frequently used as induction therapy with potentially less toxicity. This retrospective study aimed to compare outcomes between treatment-naive HPV(+) OPSCC patients receiving induction TPF and those receiving an induction platinum doublet. Data collected included tumor characteristics, response after chemoradiation, hospitalization rates, and overall survival (OS). Fifty-five patients (18 TPF and 37 platinum doublet) were included. There was no significant difference in response after completion of definitive chemoradiation (TPF: CR 83.3%, PR 5.6%, progression or metastasis 11.1% vs. platinum doublet: CR 75.7%, PR 16.2%, progression or metastasis 8.1%; p = 0.5241). There were also no differences in hospitalizations for adverse events (38.9% in TPF vs. 40.5% in platinum doublet, p = 0.907) or recurrence (11.1% in TPF vs. 2.7% in platinum doublet, p = 0.198). The 5-year OS was 84.6% in the TPF group and 81.5% in the platinum doublet group (p = 0.581). Induction platinum doublet regimens offer comparable OS, response, and hospitalization rates to TPF in locally advanced HPV(+) OPSCC. Induction with a platinum doublet may be a viable de-escalation strategy for patients who are not candidates for TPF.

Abstract Background: HPV16 (human papillomavirus) is a leading cause of oropharyngeal cancer (OPC) in the US. Antibodies to HPV16 early antigens are detectable in serum samples up to two decades before diagnosis and are potential biomarkers for early detection. In this study, we have developed and compared three custom HPV16 serology platforms for detection of OPC: MSD (Meso Scale Discovery) electrochemiluminescent detection, RAPID ELISA, and a novel semi-quantitative lateral flow assay (LFA) using a custom android application. Methods: For the MSD platform, recombinant HPV 16 antigens (E1, E2, E6, E7) were synthesized in mammalian cells and printed on multiplexed arrays. In RAPID ELISA, GST-tagged antigens (E1, E2, E6, E7) were expressed using mammalian in vitro transcription and translation and bound IgG detected. For the LFA, His-tagged recombinant proteins (HPV16 CE2 184-365aa and HPV16 E7) were synthesized, printed, and antibodies detected with protein G conjugated nanoparticles. Archived serum samples drawn from newly-diagnosed OPC patients (n = 60), with matching saliva samples (n=54, MSD assay only) were obtained from Mayo Clinic. Unmatched controls (n = 46) were obtained from AT Still. For the lateral flow assay, colorimetric image analysis was done using a titration curve of known concentration of monoclonal antibody. We optimized the biomarker panel (E2, E6, and E7 for MSD and ELISA, and for LFA E2, and E7) by using multiparametric analysis and plotting the receiver operating curves to the test performance, and assays were compared using the Cohen’s kappa statistic. Results: At 98% specificity, the MSD HPV16 serologic assay achieved 88% sensitivity, similar to RAPID ELISA, and the LFA assay at 85% sensitivity. The signal intensities for saliva samples were weaker at 24% sensitivity. The Cohen’s kappa value ranged from 0.74 to 0.78 for the comparison between all three assays. There was only one sample among all the cases that was negative by all three serology tests. Conclusion: All three assays have strong concordance for the detection of HPV16 early antigen serology. All the tests can facilitate large scale HPV serology research. Citation Format: Akshansh Kaushik, David Routman, Tej Patel, Joshua Eger, Padhmavathy Yuvaraj, Kathleen R. Bartemes, Marisa D. Griesel, Danielle E. Hunter, Katie Van Abel, Anu Mathew, Mingyue Wang, Leonid Dzantiev, Martin Stengelin, Jacob N. Wohlstadter, Ann Spolarich, Kristina R. Dahlstrom, Jennifer Blain Christen, Erich Sturgis, Karen Anderson. HPV serology for rapid detection of HPV positive oropharyngeal cancer [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2026; Part 1 (Regular Abstracts); 2026 Apr 17-22; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2026;86(7 Suppl):Abstract nr 5099.

Abstract Background: Liquid biopsy approaches for cancer detection often rely on tissue-specific methylation or fragmentation features to infer tumor origin. However, a pan-aerodigestive cancer detection strategy is clinically valuable given shared risk exposures, field cancerization, and frequent diagnostic ambiguity among lung, esophageal, and head & neck cancers. We hypothesized that advanced-stage disease exhibits convergent chromatin remodeling, producing a dominant malignancy-associated fragmentomic signal that supersedes subtler tissue-of-origin patterns. Methods: Whole-genome EM-seq (∼15×) was performed on plasma cfDNA from 120 patients with stage III/IV aerodigestive cancers (lung, esophageal, head & neck; squamous and adenocarcinoma) and 60 age-matched controls. Four fragmentomic modalities were assessed: (1) genome-wide fragmentation in 5-Mb bins (DELFI-like), (2) NMF spectral components, (3) 4-mer end motifs, and (4) promoter-centric nucleosome scores (Griffin). All preprocessing occurred strictly within each training fold of a repeated stratified 10-fold CV (100 repeats) to prevent data leakage. No test-sample information was used in training. Hyperparameters were tuned only within training partitions. Random Forest classifiers were trained under this fully nested scheme. Results: The multimodal classifier robustly detected advanced aerodigestive malignancies with a mean AUROC 0.86 and AUPRC 0.94. However, modality dissection revealed a clear performance hierarchy: nucleosome positioning (Griffin) alone achieved AUROC 0.85 and AUPRC 0.93, nearly matching the full integrated model, whereas end motifs performed poorly (AUROC 0.61). Biological interpretation demonstrated prominent nucleosome depletion centered at promoters of proliferation-associated genes, indicating a convergent epigenetic state of promoter opening across late-stage aerodigestive malignancies. Conclusions: Advanced aerodigestive cancers exhibit a dominant, convergent chromatin remodeling signature detectable in plasma cfDNA, driven primarily by altered nucleosome positioning. In late-stage disease, this shared malignancy-associated fragmentomic signal outweighs tissue-of-origin-dependent patterns. These findings highlight the utility of nucleosome-based fragmentomics not merely for detection, but for resolving diagnostic ambiguity in patients with indeterminate masses and for monitoring total tumor burden in high-risk aerodigestive malignancies. Citation Format: Axel Misael Hidalgo, Ayesha Hashmi, Jeff Szymanski, Pradeep Chauhan, Lilli Greiner, Faridi Qaium, Daniel J. Ma, David M. Routman, Katie M. Van Abel, Aadel A. Chaudhuri.. Convergent chromatin remodeling enables pan-aerodigestive detection of advanced malignancies in plasma cfDNA [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2026; Part 1 (Regular Abstracts); 2026 Apr 17-22; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2026;86(7 Suppl):Abstract nr 3862.

Abstract Background: Personalizing chemoradiotherapy (chemoRT) for locally advanced non-small cell lung cancer (LA-NSCLC) is limited by the lack of early biomarkers that inform response during treatment. Current circulating tumor DNA (ctDNA) molecular residual disease assays predict post-treatment outcomes but lack mid-treatment utility. ctDNA burden estimation has traditionally relied on variant-centric metrics such as maximum variant allele fraction (Max VAF), which rely on mutation-specific signals and may be confounded by histology-dependent shedding variability, allelic imbalance, and clonal hematopoiesis. Here we present circulating tumor fraction estimate (ctFE), a machine-learning composite score that integrates VAF distributions, copy-number alterations, and germline B-allele frequency deviations to approximate global tumor burden using a widely available, tumor-naïve clinical platform. Methods: A burden-based ctFE threshold was derived using pre-treatment plasma from a prospective phase II clinical trial of MR-guided hypofractionated chemoRT (LA-WU, n=26), locked, and applied unchanged to early mid-treatment samples (day 10-14). To test scalability and generalizability, its prognostic performance was validated in two real-world cohorts: 94 LA-NSCLC patients receiving definitive chemoRT (LA-RW) and 309 oligometastatic NSCLC patients receiving consolidative RT (OM-RW). ctFE and Max VAF were compared as continuous predictors across cohorts. Results: ctFE consistently outperformed Max VAF. In LA-WU, pre-treatment ctFE was associated with overall survival (OS HR=1.15, p=0.04) and progression-free survival (PFS HR=1.84, p=0.009), whereas Max VAF was not. Mid-treatment ctFE remained prognostic for PFS (HR=1.14, p=0.026). Early ctFE dynamics defined three molecular response groups with marked OS separation: consistently low, responder, and nonresponder groups (median OS 60.8 vs 13.0 vs 2.9 months, respectively; p<0.001). In multivariable models including both biomarkers, higher ctFE remained independently associated with worse survival in LA-RW (OS HR=1.88, p=0.010) and OM-RW (OS HR=1.37, p=0.040; PFS HR=1.45, p=0.008), whereas Max VAF did not. The locked ctFE threshold stratified OS across all cohorts (LA-WU HR=5.93; LA-RW HR=9.08; OM-RW HR=2.26; all p<0.001). Conclusion: ctFE provides clinically meaningful pre- and mid-treatment risk stratification and consistently outperforms Max VAF across locally advanced and oligometastatic NSCLC cohorts. We show that ctFE is a biologically informative, clinically generalizable and scalable ctDNA burden metric measurable using a tumor-naïve, off-the-shelf assay, supporting its practical utility in biomarker-adapted radiotherapy strategies. Citation Format: Ayesha Hashmi, Jessica Linford, Pradeep S. Chauhan, Kaushal Parikh, Rotem Ben-Shachar, John Guittar, Malvika Pillai, Jyoti Patel, Matteo Bergsagel, Nicholas P. Semenkovich, Kenneth R. Olivier, Sean S. Park, Dawn Owen, David M. Routman, Katie N. Lee, Alex D. Sherry, Aaron S. Mansfield, Daniel Morgensztern, Ramaswamy Govindan, Clifford G. Robinson, Carmen Bergom, Saiama N. Waqar, Bruna Pellini Ferreira, Gregory R. Vlacich, Aadel A. Chaudhuri. Early ctDNA quantification by ctFE outperforms Max VAF for survival stratification across locally advanced and oligometastatic NSCLC treated with radiotherapy [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2026; Part 1 (Regular Abstracts); 2026 Apr 17-22; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2026;86(7 Suppl):Abstract nr 1183.

PURPOSE Cancer recurrence is a critical outcome for patients and physicians. Retrospective cancer recurrence data can evaluate recurrence-directed treatment and generate novel interventions targeting recurrent cancers. However, large cancer databases do not provide recurrence-related information, stymying study at scale, and consequently require significant manual record review. Automated evaluation of records may allow for the rapid generation of easily analyzed data sets, accelerating the evaluation of recurrence altogether. METHODS Patients treated with radiation therapy at one tertiary referral center from 2010 to 2018 with a verified cancer status (cancer recurrence v no cancer recurrence) were identified. Patients with recurrent disease were initially identified through manual record review, and the associated pathology report was collected. Google Automated Machine Learning with Natural Language Processing (AutoNLP) and Google Gemini 1.5 Pro were used to generate a model for binary classification, with comparison to the gold-standard manually developed data set. RESULTS A total of 7,054 patients were identified. 3,431 (48.6%) were female, with a median age of 64 years. Head and neck (1,482, 21%), breast (1,480, 21%), upper GI (1,307, 18.5%), and lung/thorax (1,107, 15.7%) were the most common disease sites. Recurrence was verified for 1,546 patients (21.9%) using pathology reports, of which 1,249 positive cases were paired with 651 negative pathology reports for model development. Google Gemini 1.5 Pro consistently outperformed AutoNLP across all measurements of accuracy, generating a greater absolute difference in precision, recall, negative predictive value, and specificity, and a higher likelihood of correct classification at the individual level, rendering Gemini superior in recurrence status extraction. CONCLUSION AutoNLP and Google Gemini 1.5 Pro are promising tools for identifying recurrence from pathology reports, with the latter demonstrating superior overall performance, making it particularly suitable for clinical translation.