About

Nicholas J Seewald, PhD, is an Assistant Professor of Biostatistics and Epidemiology at the Hospital of the University of Pennsylvania and a Senior Scholar at the Center for Clinical Epidemiology and Biostatistics. His research expertise involves developing and applying statistical methodologies to address key questions in public health and medicine through thoughtful study design and data analysis, in collaboration with applied scientists. His work spans a wide range of applications, including physical activity, oncology, and substance use, and covers the entire investigative process from formulating research questions to study design and data analysis. His methodological focus is primarily on causal inference, particularly in the context of complex repeated-measures data. He aims to develop statistical tools that enable scientists to make impactful contributions in their fields by creating accessible and understandable methods to address important statistical issues.

Research topics

• Medicine
• Internal medicine
• Computer science
• Oncology
• Psychology

Selected publications

• Plasma cell‐free DNA markers predict occult metastases in patients with resectable pancreatic ductal adenocarcinoma

  Clinical and Translational Medicine · 2026-01-01

  articleOpen access

  Dear Editor, Detecting pancreatic ductal adenocarcinoma (PDAC) early can yield dramatic improvements in overall survival (OS). Curative intent resection is typically indicated when the disease is localised to the pancreas. However, standard of care imaging lacks sensitivity to detect smaller occult metastases, often resulting in patients undergoing an unnecessary and morbid surgery, followed by early recurrence.1, 2 While we have previously demonstrated detection of early-stage PDAC using exocrine pancreas methylation markers in cfDNA,3 here we show that methylation markers, when combined with circulating tumour KRAS mutation detection and imaging measurements, can predict the presence of occult metastatic disease before curative intent surgery. A convenience sample of patients was enrolled with written informed consent at the University of Pennsylvania Hospital (Philadelphia, PA), under IRB Protocol #822028, NCT02471170. Patients had previously untreated PDAC or were seen in the endoscopy clinic for routine screening (healthy controls) or non-cancer disease evaluation and monitoring (disease controls). Disease control patients’ diagnoses included pancreatic cyst, pancreatitis, intraductal papillary mucinous neoplasm, and other non-cancerous pancreatic conditions. Patients with PDAC were excluded for 1) insufficient imaging surveillance to identify occult metastases within 120 days of surgery or 2) receiving therapy for a second primary tumour ≤5 years of PDAC diagnosis. Clinical and demographic data were abstracted from the electronic medical record, including the presence of metastases within 120 days of surgery. Pathologic staging (pT and pN) was obtained for patients who completed surgery; otherwise, clinical staging was used. CA19-9 values for 69 of 75 naive resectable PDAC patients were abstracted from the medical record for a timepoint within 40 days of surgery. For 6 patients, an aliquot of previously frozen plasma was provided to the clinical laboratory at the University of Pennsylvania and analysed using the clinical protocol. See Supplemental Digital Content for elaboration of study methods. This study was performed in accordance with STARD 2015 guidelines. We analysed plasma from a cohort of 176 patients, including PDAC and non-PDAC controls (Figure S1 and Tables S1 and S2), to explore whether cfDNA methylation markers (Figure S2 and Table S7), independent of tumour genomic profiling, distinguished PDAC patients with and without occult metastases. Building on the previous pancreas tissue methylome analysis,3 we identified methylated or unmethylated loci in liver and lung tissue, the two most common sites of distant metastases for PDAC. We then adapted our methods to detect these loci in plasma cfDNA. For 75 patients with PDAC who had surgery without receiving neoadjuvant therapy (“naïve resectable”), the cfDNA concentration from exocrine pancreas, hepatocytes, and lung epithelium (expressed as genome equivalents or copies per ml) was significantly higher than disease (see Methods) or healthy controls (Figure 1A). While pancreas and lung concentrations were significantly higher in 24 patients with imaging-confirmed metastatic disease at diagnosis (“metastatic”) (Table S3) compared to naïve resectable patients, there was no significant difference in liver cfDNA (Figure 1A). Among 75 patients with naïve resectable PDAC, 25 had occult metastases, with six discovered intraoperatively and 19 by imaging within 120 days postoperatively. Overall survival (OS) was significantly shorter for patients with versus without occult metastases (Figure S3A). Liver was the most prevalent site of first detected occult metastases (Tables S4 and S5). Pancreas and lung cfDNA were significantly higher for patients with versus without occult metastases (p = 0.0007 and p = 0.0090, respectively). However, there was no significant difference in hepatocyte copies for these two groups (p = 0.6210, Figure 1B). Pancreas and lung copies had significant AUCs for predicting occult versus no occult metastases (p = 0.0009 and p = 0.0096, respectively), but hepatocyte copies did not (p = 0.6170, Figure 1C; associated cutoffs and statistics shown in Figure S4A). Kaplan-Meier analysis for time to any metastases (TTM) for all 75 naïve resectable patients showed that pancreas copies above versus ≤ median was associated with significantly shorter median TTM. Similar results were obtained for lung cfDNA, while no significant difference was found for hepatocyte cfDNA (Figure 1D; Kaplan-Meier analysis using AUC-derived cutoffs shown in Figure S4B). Patients with both pancreas and lung cfDNA above median had the shortest median TTM compared to those with either or neither marker above median (Figure 1E), with similar results for OS (Figure S3B). cfDNA methylation markers remained significantly associated with TTM and OS when patient characteristics were added in a multivariable analysis (Table S6). We next assessed two additional blood-based markers, pre-surgery circulating tumour DNA-based KRAS mutation detection (ctKRAS) and CA19-9, as well as primary tumour volume as measured from pre-surgery imaging. Among naïve resectable patients, the proportion of patients with detected ctKRAS was significantly higher for those with versus without occult metastases (p < 0.0001); however, there was no difference in CA19-9 (p = 0.4161, Figure 2A). Primary tumour volume for naïve resectable patients with occult metastases was higher than for those without (p = 0.0260, Figure 2A). ROC analysis was consistent with this for CA19-9 and tumour volume (Figure 2B), as was Kaplan-Meier analysis for TTM and OS for ctKRAS, CA19-9, and tumour volume (Figure 2C and Figure S5). Given that none of the continuous variables were significantly correlated with the cfDNA markers (Figure 2D), we assessed whether combining markers could improve prediction. A least absolute shrinkage and selection operator model selected pancreas and lung copies/mL, ctKRAS, and tumour volume for predicting occult metastases (Figure 2E,F). As an exploratory analysis, we analysed cfDNA markers for 27 patients who received neoadjuvant therapy; however, the markers were not predictive of occult metastases (Figure S6). Plasma cfDNA methylation markers may improve identification of patients with occult PDAC metastases, providing a potentially actionable biomarker for patient stratification, independent of tissue molecular analysis. These results are consistent with the recent finding that pre-operative ctDNA levels improved disease stratification for patients with early-stage non-small cell lung cancer4 and suggest that tumours with occult metastases associate with higher rates of cellular turnover in the primary or metastatic sites. More work is needed to examine whether the concept applies to tumours beyond PDAC, and to reproduce the findings in additional patient populations to facilitate clinical implementation. E.L.C. and Y.D. conceptualised the research plan. O.G.-R., D.B., S.C., M.Y., K.T., M.S., S.G., H.E.S., K.N.K., M.H.O., J.E.T., C.V., R.S. and H.S. generated data by experimentation or medical chart review. E.L.C., Y.D., M.Y. and R.S. managed the project. E.L.C., R.S. and N.J.S. led the primary data analysis. E.L.C. drafted the initial manuscript. S.G., R.S., E.L.C., J.E.T., O.G.-R. and H.E.S. produced figures and tables. S.C., S.G., H.E.S. and O.G.-R performed data organisation and conducted additional analysis. J.E.T. and E.L.C. completed the final review. This study was supported by the James and Marlene Scully Liquid Biopsy Innovation Fund and Penn Pancreatic Cancer Research Centre Netter Fund to Erica L. Carpenter. The Hale Family Centre for Pancreatic Cancer Research, Lustgarten Foundation Dedicated Laboratory Program, and National Cancer Institute of the National Institutes of Health Award CA210171 supported Brian Wolpin. Grants from EU (PANCAID, 101096309), the Soyka Pancreatic Cancer Fund and the Israel Innovation authority supported Yuval Dor. All subjects received written informed consent according to the Declaration of Helsinki, under University of Pennsylvania Institutional Review Board protocol 822028. The data produced and analyzed for this study are included in Table S1. Please note: The publisher is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.

  PublisherOA PDFDOI

• Impact of State Telemedicine Policies on Substance Use Disorder Treatment During the COVID-19 Pandemic

  Journal of General Internal Medicine · 2026-04-17

  articleOpen access

  IMPORTANCE: The COVID-19 pandemic disrupted traditional substance use disorder (SUD) treatment modalities, prompting innovative telemedicine solutions. OBJECTIVE: To evaluate the association between state-level telemedicine policies and SUD treatment during the COVID-19 pandemic. DESIGN: An augmented synthetic control analysis comparing changes in SUD treatment before and after the implementation of telemedicine policies in states that adopted these policies to changes in SUD treatment in comparison states without these policies from 2018 to 2022. SETTING: This study utilized a comprehensive policy database merged with de-identified patient claims data from the OptumLabs® Data Warehouse. PARTICIPANTS: Individuals aged 18 in only fully insured commercial, private plans subject to state insurance policies between January 2018 and December 2022. EXPOSURE: Combined implementation of three state telemedicine policies during the early months of the COVID-19 pandemic (March-July 2020), including telemedicine coverage parity, telemedicine payment parity, and in-person relationship waivers. MAIN OUTCOMES: SUD treatment initiation and continuation, measured at the patient-month level and aggregated to the state-month level for analyses, with subgroup analyses for opioid use disorder (OUD) and alcohol use disorder (AUD) treatment. RESULTS: Implementation of the three policies was associated with average effects of less than 0.006 percentage points in the proportion of adults initiating SUD, AUD, or OUD treatment between March/April 2020 and December 2022, with confidence intervals not exceeding a 0.02 percentage point increase or decrease (p > 0.05). Among adults receiving treatment prior to the COVID-19 pandemic, these policies were associated with an estimated average change of less than 1.5 percentage points in the proportion of adults receiving SUD, AUD, or OUD treatment, with confidence intervals not exceeding a 4.5 percentage point increase or decrease (p > 0.05). CONCLUSIONS AND RELEVANCE: Our study did not identify impacts of state telemedicine coverage parity, payment parity, and in-person relationship waiver policies on SUD treatment during the COVID-19 pandemic.

  PublisherDOI

• Waveform data capture substantial variation in tidal volume and other respiratory parameters

  Annals of the American Thoracic Society · 2026-02-11

  articleOpen access

  RATIONALE: Patients receiving invasive mechanical ventilation (IMV) require accurate assessments of ventilator parameters. Documentation of these parameters in standard practice may fail to capture meaningful variation due to intermittent missingness. OBJECTIVES: To assess variation in continuously measured ventilator parameters and agreement with measurements documented as part of routine care in the electronic health record (EHR). METHODS: We performed a retrospective cohort study of patients receiving IMV in a medical intensive care unit from November 2024 through March 2025. We compared the observed tidal volume, minute ventilation, peak inspiratory pressure, and positive end-expiratory pressure, measured continuously from device waveforms with intermittent EHR documentation. We calculated descriptive statistics and measures of agreement between these sources. RESULTS: For 59 encounters, the median age was 65 years (IQR, 59-72 years), 33 (56%) patients were male, and 17 (29%) were Black. Thirty-four (58%) patients died or were discharged to hospice. Among 358 patient-days of data, continuous measurements captured significantly more variation than EHR-documented measurements across all parameters. The largest errors were in observed tidal volume (mean absolute error, 69 mL [95% CI, 62-77 mL]; correlation coefficient 0.540). Agreement in tidal volume was worse among patients receiving mandatory modes of ventilation (correlation coefficient 0.454). CONCLUSIONS: Intermittent measurement of ventilator parameters fails to capture large variability observed in continuous, waveform-derived measurements. Poor agreement in parameters like tidal volume, even in mandatory modes of ventilation, highlights the potential for ventilator waveform data to improve care and advance research for patients with acute respiratory distress syndrome and others receiving IMV.

  PublisherDOI

• Associations between patterns of neuroendocrine liver metastatic burden and outcomes after liver-directed therapy, systemic chemotherapy and peptide receptor radionuclide therapy.

  Neuroendocrinology · 2026-04-16

  article

  Introduction Patients with liver-dominant metastatic neuroendocrine tumor (NET) have multiple treatment options. Post hoc analysis of NETTER-1 suggested that tumor size but not tumor burden predicted progression-free survival (PFS), contrary to prior publications. Analysis of imaging datasets from two completed multicenter prospective clinical trials for capecitabine-temozolomide (CapTem) and liver-directed therapy (LDT), and an institutional cohort of patients treated with peptide receptor radionuclide therapy (PRRT) was performed to investigate whether subgroups of liver metastatic disease based on lesion size, lesion number and tumor burden may guide treatment selection. Methods Image review from a similar number of patients with liver metastases from each cohort were categorized by number, maximum diameter, and tumor burden as a fraction of liver volume (n=219). Morphologic categories were then correlated with response and PFS by RECIST criteria. Descriptive and graphical analyses were followed by multivariable modeling to test treatment-by-stratum interaction. Results Imaging features were not associated with statistically significant differences in response or PFS for the three therapies or the entire analyzed population. The ORR for LDT, PRRT and CapTem were 65%, 38% and 25% (p<0.001), respectively, with an odds ratio favoring LDT of 5.45 vs. CapTem and 3.0 vs. PRRT. The respective median PFS were LDT 18.9 months [95%CI 16.3-24], PRRT 21.6 [14.3-26.7], and 16.6 [11.5-29] for CapTem (p=0.99). Conclusion LDT had the highest response rate of these distinct cohorts. PFS was not different between modalities. Imaging features did not predict treatment outcome within a particular modality nor to favor one over another when triaging patients.

  PublisherDOI

• QIM26-279: Increasing Equitable Adherence to Annual Lung Cancer Screenings and Diagnostic Surveillance: A Pragmatic Factorial Trial

  Journal of the National Comprehensive Cancer Network · 2026-03-27

  article
  PublisherDOI

• Supplementary Table S3 from Corticosteroid-Dependent Association between Prognostic Peripheral Blood Cell-Free DNA Levels and Neutrophil-Mediated NETosis in Patients with Glioblastoma

  2025-04-01

  supplementary-materialsOpen access

  &lt;p&gt;Array consistent autosomal CpGs from which deconvolution signature was selected.&lt;/p&gt;

  PublisherDOI

• Supplementary Table S8 from Corticosteroid-Dependent Association between Prognostic Peripheral Blood Cell-Free DNA Levels and Neutrophil-Mediated NETosis in Patients with Glioblastoma

  2025-04-01

  supplementary-materialsOpen access

  &lt;p&gt;List of all methylGSA results for (Gene Ontology,GO, gene sets) with differential methylation for % neutrophil ccfDNA dichotomized at median.&lt;/p&gt;

  PublisherDOI

• LODESTAR: A single-arm phase II study of rucaparib in solid tumors with pathogenic germline or somatic variants in homologous recombination repair genes.

  Journal of Clinical Oncology · 2025-05-28

  article

  3151 Background: To explore PARP inhibitor (PARPi) utility across solid tumors and identify biomarkers that predict sensitivity. Methods: This single-arm phase II study assessed rucaparib monotherapy in patients with solid tumors and pathogenic variants (PVs) in BRCA1, BRCA2, PALB2, RAD51C, RAD51D (Cohort A) or BARD1, BRIP1, FANCA, NBN, RAD51B (Cohort B). The primary endpoint was ORR in Cohort A. Secondary endpoints included DCR, PFS, OS and safety. A scar-based HRD signature (HRDsig) and platinum sensitivity status were explored post-hoc. Results: Fifty-one patients in Cohort A and 12 in Cohort B were evaluable for efficacy. ORR of cohort A was 18% (95% CI 10-30%). A significantly higher ORR was observed with HRDsig+ tumors compared to HRDsig- tumors (32%, 95% CI 15-54, vs. 0%, 95% CI 0-14%, p &lt; 0.01). In the entire study population: DCR of 65% (95% CI 53-76%), mPFS of 5.5 mo (95% CI 3.68-7.82), and mOS of 12.1 mo (95% CI 10.6 – inf). PFS and OS were significantly longer for platinum sensitive tumors (mPFS: 7.8 mo vs. 3.5 mo, p = 0.02; mOS: NR vs 5.45mo, p = 0.01). Tumor histology was not independently predictive of outcome. Tumors with PVs in Cohort A genes were more likely to be HRDsig+ than tumors with PVs in Cohort B genes. Analysis of a large commercial database showed that in non-canonical tumors with BRCA PVs, 30.2% were HRDsig+. Conclusions: Rucaparib has activity in HRDsig+ solid tumors with PVs in HRR genes, regardless of histology. Platinum sensitivity correlated with improved outcomes. Clinical trial information: NCT04171700 .

  PublisherDOI

• Supplementary Table S10 from Corticosteroid-Dependent Association between Prognostic Peripheral Blood Cell-Free DNA Levels and Neutrophil-Mediated NETosis in Patients with Glioblastoma

  2025-04-01

  supplementary-materialsOpen access

  &lt;p&gt;List of all methylGSA results for (Gene Ontology,GO, gene sets) with differential methylation by pre-operative steroid exposure (exposed vs unexposed).&lt;/p&gt;

  PublisherDOI

• Supplementary Table S2 from Corticosteroid-Dependent Association between Prognostic Peripheral Blood Cell-Free DNA Levels and Neutrophil-Mediated NETosis in Patients with Glioblastoma

  2025-04-01

  supplementary-materialsOpen access

  &lt;p&gt;Published reference methylomes used in deconvolution&lt;/p&gt;

  PublisherDOI

Frequent coauthors

• Anna Maria Storniolo

  13 shared

• Vered Stearns

  13 shared

• Kelley M. Kidwell

  University of Michigan–Ann Arbor

  13 shared

• David A. Flockhart

  12 shared

• N. Lynn Henry

  University of Michigan–Ann Arbor

  12 shared

• Daniel F. Hayes

  University of Michigan–Ann Arbor

  12 shared

• Kunal C. Kadakia

  Levine Cancer Institute

  11 shared

• Claire Snyder

  Johns Hopkins Medicine

  8 shared

Education

• Ph.D., Statistics

  University of Michigan

  2021

• Master of Arts, Statistics

  University of Michigan

  2018

• Master of Science, Biostatistics

  University of Michigan

  2015

• Bachelor of Science, Mathematics

  University of Notre Dame

  2013