The performance status gap in immunotherapy for frail patients with advanced non-small cell lung cancer.

PURPOSE: In advanced non-small cell lung cancer (NSCLC), immune checkpoint inhibitor (ICI) monotherapy is often preferred over intensive ICI treatment for frail patients and those with poor performance status (PS). Among those with poor PS, the additional effect of frailty on treatment selection and mortality is unknown. METHODS: Patients in the veterans affairs national precision oncology program from 1/2019-12/2021 who received first-line ICI for advanced NSCLC were followed until death or study end 6/2022. Association of an electronic frailty index with treatment selection was examined using logistic regression stratified by PS. We also examined overall survival (OS) on intensive treatment using Cox regression stratified by PS. Intensive treatment was defined as concurrent use of platinum-doublet chemotherapy and/or dual checkpoint blockade and non-intensive as ICI monotherapy. RESULTS: Of 1547 patients receiving any ICI, 66.2% were frail, 33.8% had poor PS (≥ 2), and 25.8% were both. Frail patients received less intensive treatment than non-frail patients in both PS subgroups (Good PS: odds ratio [OR] 0.67, 95% confidence interval [CI] 0.51 - 0.88; Poor PS: OR 0.69, 95% CI 0.44 - 1.10). Among 731 patients receiving intensive treatment, frailty was associated with lower OS for those with good PS (hazard ratio [HR] 1.53, 95% CI 1.2 - 1.96), but no association was observed with poor PS (HR 1.03, 95% CI 0.67 - 1.58). CONCLUSION: Frail patients with both good and poor PS received less intensive treatment. However, frailty has a limited effect on survival among those with poor PS. These findings suggest that PS, not frailty, drives survival on intensive treatment.

Risk of severe coronavirus disease 2019 despite vaccination in patients requiring treatment with immune-suppressive drugs: A nationwide cohort study of US Veterans.

BACKGROUND: Patients taking immune-suppressive drugs are at increased risk of severe coronavirus disease 2019 (COVID-19), not fully ameliorated by vaccination. We assessed the contributions of clinical and demographic factors to the risk of severe disease despite vaccination in patients taking immune-suppressive medications for solid organ transplantation (SOT), rheumatoid arthritis (RA), inflammatory bowel disease (IBD), or psoriasis. METHODS: Veterans Health Administration electronic health records were used to identify patients diagnosed with RA, IBD, psoriasis, or SOT who had been vaccinated against severe acute respiratory syndrome coronavirus 2, were subsequently infected, and had received immune-suppressive drugs within 3 months before infection. The association of severe (defined as hypoxemia, mechanical ventilation, dexamethasone use, or death) versus non-severe COVID-19 with the use of immune-suppressive and antiviral drugs and clinical covariates was assessed by multivariable logistic regression. RESULTS: Severe COVID-19 was more common in patients with SOT (230/1011, 22.7%) than RA (173/1355, 12.8%), IBD (51/742, 6.9%), or psoriasis (82/1125, 7.3%). Age was strongly associated with severe COVID-19, adjusted odds ratio (aOR) of 1.04 (CI 1.03-1.05) per year. Comorbidities indicating chronic brain, heart, lung, or kidney damage were also associated with severity, aOR 1.35-2.38. The use of glucocorticoids was associated with increased risk (aOR 1.66, CI 1.39-2.18). Treatment with antivirals was associated with reduced severity, for example, aOR 0.28 (CI 0.13-0.62) for nirmatrelvir/ritonavir. CONCLUSION: The risk of severe COVID-19 despite vaccination is substantial in patients taking immune-suppressive drugs, more so in patients with SOT than in patients with inflammatory diseases. Age and severe comorbidities contribute to risk, as in the general population. Oral antivirals were very beneficial but not widely used.

Antidepressant exposure and long-term dementia risk in a nationwide retrospective study on US veterans with midlife major depressive disorder.

INTRODUCTION: The use of antidepressants in major depressive disorder (MDD) has been reported to influence long-term risk of Alzheimer's disease (AD) and AD-related dementias (AD/ADRD), but studies are conflicting. METHODS: We used inverse probability weighted (IPW) Cox models with time-varying covariates in a retrospective cohort study among midlife veterans with MDD within the US Veterans Affairs healthcare system from January 1, 2000 to June 1, 2022. RESULTS: A total of 35,200 patients with MDD were identified. No associations were seen regarding the effect of being exposed to any antidepressant versus no exposure on AD/ADRD risk (events = 1,056, hazard ratio = 0.94, 95% confidence interval: 0.81 to 1.09) or the exposure to specific antidepressant classes versus no exposure. A risk reduction was observed for female patients in a stratified analysis; however, the number of cases was small. DISCUSSION: Our study suggests that antidepressant exposure has no effect on AD/ADRD risk. The association in female patients should be interpreted with caution and requires further attention. HIGHLIGHTS: We studied whether antidepressant use was associated with future dementia risk. We specifically focused on patients after their first-ever diagnosis of depression. We used IPW Cox models with time-varying covariates and a large observation window. Our study did not identify an effect of antidepressant use on dementia risk. A risk reduction was observed in female patients, but the number of cases was small.

Venous thromboembolism risk in cancer patients receiving first-line immune checkpoint inhibitor versus chemotherapy.

It remains unclear if immune checkpoint inhibitor (ICI) therapy is associated with higher rate of venous thromboembolism (VTE) compared with cytotoxic chemotherapy (chemo) in patients with comparable cancer type, staging, and comorbidities. Using the national Veterans Affairs healthcare system database from 2016 to 2021, we performed a propensity score (PS)-weighted retrospective cohort study to compare the incidence of VTE in patients with selected stage III/IV cancer receiving first-line ICI versus chemo. The PS model utilized overlap weights to balance age, sex, race, treatment year, VTE history, paralysis/immobilization, prolonged hospitalization, cancer type, staging, time between diagnosis and treatment, and National Cancer Institute comorbidity index. Weighted Cox regressions with robust standard error were used to assess the hazard ratio (HR) and 95% confidence interval (CI). We found that among comparable advanced cancers, first-line ICI (n = 1823) and first-line chemo (n = 6345) had similar rates of VTE (8.49% for ICI and 8.36% for chemo at 6 months). The weighted HR was 1.06 (95% CI 0.88-1.26) for ICI versus chemo. In a subgroup analysis restricted to lung cancers, first-line ICI/chemo (n = 828), ICI monotherapy (n = 428), and chemo monotherapy (n = 4371) had similar rates of VTE (9.60% for ICI/chemo, 10.04% for ICI, and 8.91% for chemo at 6 months). The weighted HR was 1.05 (95% CI 0.77-1.42) for ICI versus chemo, and 1.08 (95% CI 0.83-1.42) for ICI/chemo versus chemo. In conclusion, ICI as a systemic therapy has a similarly elevated risk as cytotoxic chemo for VTE occurrence in cancer patients. This finding can inform future prospective studies exploring thromboprophylaxis strategies.

Validation of algorithms to select patients with multiple myeloma and patients initiating myeloma treatment in the national Veterans Affairs Healthcare System.

BACKGROUND: We aimed to evaluate and compare the performance of multiple myeloma (MM) selection algorithms for use in Veterans Affairs (VA) research. METHODS: Using the VA Corporate Data Warehouse (CDW), the VA Cancer Registry (VACR), and VA pharmacy data, we randomly selected 500 patients from 01/01/1999 to 06/01/2021 who had (1) either one MM diagnostic code OR were listed in the VACR as having MM AND (2) at least one MM treatment code. A team reviewed oncology notes for each veteran to annotate details regarding MM diagnosis and initial treatment within VA. We evaluated inter-annotator agreement and compared the performance of four published algorithms (two developed and validated external to VA data and two used in VA data). RESULTS: A total of 859 patients were reviewed to obtain 500 patients who were annotated as having MM and initiating MM treatment in VA. Agreement was high among annotators for all variables: MM diagnosis (98.3% agreement, Kappa = 0.93); initial treatment in VA (91.8% agreement; Kappa = 0.77); and initial treatment classification (87.6% agreement; Kappa = 0.86). VA Algorithms were more specific and had higher PPVs than non-VA algorithms for both MM diagnosis and initial treatment in VA. We developed the "VA Recommended Algorithm," which had the highest PPV among all algorithms in identifying patients diagnosed with MM (PPV = 0.98, 95% CI = 0.95-0.99) and in identifying patients who initiated their MM treatment in VA (PPV = 0.93, 95% CI = 0.90-0.96). CONCLUSION: Our VA Recommended Algorithm optimizes sensitivity and PPV for cohort selection and treatment classification.

Risk factors of SARS-CoV-2 infection and complications from COVID-19 in lung cancer patients.

BACKGROUND: Identifying lung cancer patients at an increased risk of getting SARS-CoV-2-related complications will facilitate tailored therapy to maximize the benefit of anti-cancer therapy, while decreasing the likelihood of COVID-19 complications. This analysis aimed to identify the characteristics of lung cancer patients that predict for increased risk of death or serious SARS-CoV-2 infection. PATIENTS AND METHODS: This was a retrospective cohort study of patients with lung cancer diagnosed October 1, 2015, and December 1, 2020, and a diagnosis of COVID-19 between February 2, 2020, and December 1, 2020, within the Veterans Health Administration. Serious SARS-CoV-2 infection was defined as hospitalization, ICU admission, or mechanical ventilation or intubation within 2 weeks of COVID-19 diagnosis. For categorical variables, differences were assessed using Χ2 tests, while Kruskal-Wallis rank-sum test was used for continuous variables. Multivariable logistic regression models were fit relative to onset of serious SARS-CoV-2 infection and death from SARS-CoV-2 infection. RESULTS: COVID-19 infection was diagnosed in 352 lung cancer patients. Of these, 61 patients (17.3%) died within four weeks of diagnosis with COVID-19, and 42 others (11.9%) experienced a severe infection. Patients who had fatal or severe infection were older and had lower hemoglobin levels than those with mild or moderate infection. Factors associated with death from SARS-CoV-2 infection included increasing age, immune checkpoint inhibitor therapy and low hemoglobin level. CONCLUSIONS:  The mortality of lung cancer patients from COVID-19 disease in the present cohort was less than previously reported in the literature. The identification of risk factors associated with severe or fatal outcomes informs management of patients with lung cancer who develop COVID-19 disease.

Factors associated with the speed and scope of diffusion of COVID-19 therapeutics in a nationwide healthcare setting: a mixed-methods investigation.

BACKGROUND: The global COVID-19 pandemic is an opportunity to evaluate factors associated with high levels of adoption of different therapeutics in a real-world setting. The aim of this nationwide, retrospective cohort study was to evaluate the diffusion and adoption of novel therapeutics with an emerging evidence basis and to identify factors that influenced physicians' treatment decisions. METHODS: Cohort creation: A cohort of Veteran patients with a microbiologically confirmed diagnosis of SARS-CoV2 were identified, and cases were classified by disease severity (outpatient, inpatient with mild and severe disease, intensive care unit ICU]). After classification of disease severity, the proportion of cases (outpatients) and admissions (inpatients) in each category receiving each type of medication were plotted as a function of time. Identification of milestones and guidance changes: Key medications used for the management of COVID-19 milestones in the release of primary research results in various forms (e.g. via press release, preprint or publication in a traditional medical journal), policy events and dates of key guidelines were identified and plotted as a timeline. After a timeline was created, time points were compared to changes in medication use, and factors potentially impacting the magnitude (i.e. proportion of patients who received the treatment) and the speed (i.e. the slope of the change in use) of practice changes were evaluated. RESULTS: Dexamethasone and remdesivir, the first two medications with clinical trial data to support their use, underwent the most rapid, complete and sustained diffusion and adoption; the majority of practice changes occurred after press releases and preprints were available and prior to guideline changes, although some additional uptake occurred following guideline updates. Medications that were not "first in class", that were identified later in the pandemic, and that had higher perceived risk had slower and less complete uptake regardless of the strength and quality of the evidence supporting the intervention. CONCLUSIONS: Our findings suggest that traditional and social media platforms and preprint releases were major catalysts of practice change, particularly prior to the identification of effective treatments. The "first available treatment in class" impact appeared to be the single most important factor determining the speed and scope of diffusion.

Racial and Ethnic Disparity for Cancer Mortality in General and Single-Payer Healthcare Systems in the United States.

BACKGROUND: It remains unclear what factors significantly drive racial disparity in cancer survival in the United States (US). We compared adjusted mortality outcomes in cancer patients from different racial and ethnic groups on a population level in the US and a single-payer healthcare system. PATIENTS AND METHODS: We selected adult patients with incident solid and hematologic malignancies from the Surveillance, Epidemiology, and End Results (SEER) 2011-2020 and Veteran Affairs national healthcare system (VA) 2011-2021. We classified the self-reported NIH race and ethnicity into non-Hispanic White (NHW), non-Hispanic Black (NHB), non-Hispanic Asian Pacific Islander (API), and Hispanic. Cox regression models for hazard ratio of racial and ethnic groups were built after adjusting confounders in each cohort. RESULTS: The study included 3,104,657 patients from SEER and 287,619 patients from VA. There were notable differences in baseline characteristics in the two cohorts. In SEER, adjusted HR for mortality was 1.12 (95% CI, 1.12-1.13), 1.03 (95% CI, 1.03-1.04), and 0.91 (95% CI, 0.90-0.92), for NHB, Hispanic, and API patients, respectively, vs. NHW. In VA, adjusted HR was 0.94 (95% CI, 0.92-0.95), 0.84 (95% CI, 0.82-0.87), and 0.96 (95% CI, 0.93-1.00) for NHB, Hispanic, and API, respectively, vs. NHW. Additional subgroup analyses by cancer types, age, and sex did not significantly change these associations. CONCLUSIONS: Racial disparity continues to persist on a population level in the US especially for NHB vs. NHW patients, where the adjusted mortality was 12% higher in the general population but 6% lower in the single-payer VA system.

Severe COVID-19 in Vaccinated Adults With Hematologic Cancers in the Veterans Health Administration.

Importance: With SARS-CoV-2 transforming into an endemic disease and with antiviral treatments available, it is important to establish which patients remain at risk of severe COVID-19 despite vaccination. Objective: To quantify the associations of clinical and demographic variables with odds of severe COVID-19 among patients with hematologic cancers. Design, Setting, and Participants: This case-control study included all patients with hematologic malignant neoplasms in the national Veterans Health Administration (VHA) who had documented SARS-CoV-2 infection after vaccination. Groups of patients with severe (cases) vs nonsevere (controls) COVID-19 were compared. Data were collected between January 1, 2020, and April 5, 2023, with data on infection collected between January 1, 2021, and September 30, 2022. All patients with diagnostic codes for hematologic malignant neoplasms who had documented vaccination followed by documented SARS-CoV-2 infection and for whom disease severity could be assessed were included. Data were analyzed from July 28 to December 30, 2023. Exposures: Clinical (comorbidities, predominant viral variant, treatment for malignant neoplasm, booster vaccination, and antiviral treatment) and demographic (age and sex) variables shown in prior studies to be associated with higher or lower rates of severe COVID-19. Comorbidities included Alzheimer disease or dementia, chronic kidney disease, chronic obstructive pulmonary disease, diabetes, heart failure, and peripheral vascular disease. Main Outcome and Measures: The main outcome was severe COVID-19 compared with nonsevere SARS-CoV-2 infection. Severe COVID-19 was defined as death within 28 days, mechanical ventilation, or hospitalization with use of dexamethasone or evidence of hypoxemia or use of supplemental oxygen. Multivariable logistic regression was used to estimate the associations of demographic and clinical variables with the odds of severe COVID-19, expressed as adjusted odds ratios (aORs) with 95% CIs. Results: Among 6122 patients (5844 [95.5%] male, mean [SD] age, 70.89 [11.57] years), 1301 (21.3%) had severe COVID-19. Age (aOR per 1-year increase, 1.05; 95% CI, 1.04-1.06), treatment with antineoplastic or immune-suppressive drugs (eg, in combination with glucocorticoids: aOR, 2.32; 95% CI, 1.93-2.80), and comorbidities (aOR per comorbidity, 1.35; 95% CI, 1.29-1.43) were associated with higher odds of severe disease, whereas booster vaccination was associated with lower odds (aOR, 0.73; 95% CI, 0.62-0.86). After oral antiviral drugs became widely used in March 2022, 20 of 538 patients (3.7%) with SARS-CoV-2 infection during this period had progression to severe COVID-19. Conclusions and Relevance: In this case-control study of patients with hematologic cancers, odds of severe COVID-19 remained high through mid-2022 despite vaccination, especially in patients requiring treatment.

Transfer Learning for Mortality Prediction in Non-Small Cell Lung Cancer with Low-Resolution Histopathology Slide Snapshots.

High-resolution whole slide image scans of histopathology slides have been widely used in recent years for prediction in cancer. However, in some cases, clinical informatics practitioners may only have access to low-resolution snapshots of histopathology slides, not high-resolution scans. We evaluated strategies for training neural network prognostic models in non-small cell lung cancer (NSCLC) based on low-resolution snapshots, using data from the Veterans Affairs Precision Oncology Data Repository. We compared strategies without transfer learning, with transfer learning from general domain images, and with transfer learning from publicly available high-resolution histopathology scans. We found transfer learning from high-resolution scans achieved significantly better performance than other strategies. Our contribution provides a foundation for future development of prognostic models in NSCLC that incorporate data from low-resolution pathology slide snapshots alongside known clinical predictors.

Natural Language Processing Algorithm to Extract Multiple Myeloma Stage From Oncology Notes in the Veterans Affairs Healthcare System.

PURPOSE: Stage in multiple myeloma (MM) is an essential measure of disease risk, but its measurement in large databases is often lacking. We aimed to develop and validate a natural language processing (NLP) algorithm to extract oncologists' documentation of stage in the national Veterans Affairs (VA) Healthcare System. METHODS: Using nationwide electronic health record (EHR) and cancer registry data from the VA Corporate Data Warehouse, we developed and validated a rule-based NLP algorithm to extract oncologist-determined MM stage. To that end, a clinician annotated MM stage within over 5,000 short snippets of clinical notes, and annotated MM stage at MM treatment initiation for 200 patients. These were allocated into snippet- and patient-level development and validation sets. We developed MM stage extraction and roll-up algorithms within the development sets. After the algorithms were finalized, we validated them using standard measures in held-out validation sets. RESULTS: We developed algorithms for three different MM staging systems that have been in widespread use (Revised International Staging System [R-ISS], International Staging System [ISS], and Durie-Salmon [DS]) and for stage reported without a clearly defined system. Precision and recall were uniformly high for MM stage at the snippet level, ranging from 0.92 to 0.99 for the different MM staging systems. Performance in identifying for MM stage at treatment initiation at the patient level was also excellent, with precision of 0.92, 0.96, 0.90, and 0.86 and recall of 0.99, 0.98, 0.94, and 0.92 for R-ISS, ISS, DS, and unclear stage, respectively. CONCLUSION: Our MM stage extraction algorithm uses rule-based NLP and data aggregation to accurately measure MM stage documented in oncology notes and pathology reports in VA's national EHR system. It may be adapted to other systems where MM stage is recorded in clinical notes.

Building Research Infrastructure to Develop Greater Learning Efficiencies (BRIDGE).

In this manuscript, we outline our developed version of a Learning Health System (LHS) in oncology implemented at the Department of Veterans Affairs (VA). Transferring healthcare into an LHS framework has been one of the spearpoints of VA's Central Office and given the general lack of evidence generated through randomized control clinical trials to guide medical decisions in oncology, this domain is one of the most suitable for this change. We describe our technical solution, which includes a large real-world data repository, a data science and algorithm development framework, and the mechanism by which results are brought back to the clinic and to the patient. Additionally, we propose the need for a bridging framework that requires collaboration between informatics specialists and medical professionals to integrate knowledge generation into the clinical workflow at the point of care.

Phenotyping Hepatic Immune-Related Adverse Events in the Setting of Immune Checkpoint Inhibitor Therapy.

PURPOSE: We present and validate a rule-based algorithm for the detection of moderate to severe liver-related immune-related adverse events (irAEs) in a real-world patient cohort. The algorithm can be applied to studies of irAEs in large data sets. METHODS: We developed a set of criteria to define hepatic irAEs. The criteria include: the temporality of elevated laboratory measurements in the first 2-14 weeks of immune checkpoint inhibitor (ICI) treatment, steroid intervention within 2 weeks of the onset of elevated laboratory measurements, and intervention with a duration of at least 2 weeks. These criteria are based on the kinetics of patients who experienced moderate to severe hepatotoxicity (Common Terminology Criteria for Adverse Events grades 2-4). We applied these criteria to a retrospective cohort of 682 patients diagnosed with hepatocellular carcinoma and treated with ICI. All patients were required to have baseline laboratory measurements before and after the initiation of ICI. RESULTS: A set of 63 equally sampled patients were reviewed by two blinded, clinical adjudicators. Disagreements were reviewed and consensus was taken to be the ground truth. Of these, 25 patients with irAEs were identified, 16 were determined to be hepatic irAEs, 36 patients were nonadverse events, and two patients were of indeterminant status. Reviewers agreed in 44 of 63 patients, including 19 patients with irAEs (0.70 concordance, Fleiss' kappa: 0.43). By comparison, the algorithm achieved a sensitivity and specificity of identifying hepatic irAEs of 0.63 and 0.81, respectively, with a test efficiency (percent correctly classified) of 0.78 and outcome-weighted F1 score of 0.74. CONCLUSION: The algorithm achieves greater concordance with the ground truth than either individual clinical adjudicator for the detection of irAEs.

Brain tumors in United States military veterans.

BACKGROUND: Comprehensive analysis of brain tumor incidence and survival in the Veteran population has been lacking. METHODS: Veteran data were obtained from the Veterans Health Administration (VHA) Medical Centers via VHA Corporate Data Warehouse. Brain tumor statistics on the overall US population were generated from the Central Brain Tumor Registry of the US data. Cases were individuals (≥18 years) with a primary brain tumor, diagnosed between 2004 and 2018. The average annual age-adjusted incidence rates (AAIR) and 95% confidence intervals were estimated per 100 000 population and Kaplan-Meier survival curves evaluated overall survival outcomes among Veterans. RESULTS: The Veteran population was primarily white (78%), male (93%), and between 60 and 64 years old (18%). Individuals with a primary brain tumor in the general US population were mainly female (59%) and between 18 and 49 years old (28%). The overall AAIR of primary brain tumors from 2004 to 2018 within the Veterans Affairs cancer registry was 11.6. Nonmalignant tumors were more common than malignant tumors (AAIR:7.19 vs 4.42). The most diagnosed tumors in Veterans were nonmalignant pituitary tumors (AAIR:2.96), nonmalignant meningioma (AAIR:2.62), and glioblastoma (AAIR:1.96). In the Veteran population, survival outcomes became worse with age and were lowest among individuals diagnosed with glioblastoma. CONCLUSIONS: Differences between Veteran and US populations can be broadly attributed to demographic composition differences of these groups. Prior to this, there have been no reports on national-level incidence rates and survival outcomes for Veterans. These data provide vital information that can drive efforts to understand disease burden and improve outcomes for individuals with primary brain tumors.

Matching Patients to Accelerate Clinical Trials (MPACT): Enabling Technology for Oncology Clinical Trial Workflow.

Clinical trial enrollment is impeded by the significant time burden placed on research coordinators screening eligible patients. With 50,000 new cancer cases every year, the Veterans Health Administration (VHA) has made increased access for Veterans to high-quality clinical trials a priority. To aid in this effort, we worked with research coordinators to build the MPACT (Matching Patients to Accelerate Clinical Trials) platform with a goal of improving efficiency in the screening process. MPACT supports both a trial prescreening workflow and a screening workflow, employing Natural Language Processing and Data Science methods to produce reliable phenotypes of trial eligibility criteria. MPACT also has a functionality to track a patient's eligibility status over time. Qualitative feedback has been promising with users reporting a reduction in time spent on identifying eligible patients.

Baseline correlates of frailty and its association with survival in United States veterans with acute myeloid leukemia.

Frailty is an important construct to measure in acute myeloid leukemia (AML). We used the Veterans Affairs Frailty Index (VA-FI) - calculated using readily available data within the VA's electronic health records - to measure frailty in U.S. veterans with AML. Of the 1166 newly diagnosed and treated veterans with AML between 2012 and 2022, 722 (62%) veterans with AML were classified as frail (VA-FI > 0.2). At a median follow-up of 252.5 days, moderate-severely frail veterans had significantly worse survival than mildly frail, and non-frail veterans (median survival 179 vs. 306 vs. 417 days, p < .001). Increasing VA-FI severity was associated with higher mortality. A model with VA-FI in addition to the European LeukemiaNet (ELN) risk classification and other covariates statistically outperformed a model containing the ELN risk and other covariates alone (p < .001). These findings support the VA-FI as a tool to expand frailty measurement in research and clinical practice for informing prognosis in veterans with AML.

Trends in Fracture Rates Over Two Decades Among Veterans With Ankylosing Spondylitis.

OBJECTIVE: There is an increased risk of fracture in individuals with ankylosing spondylitis (AS) compared to the general population, possibly due to systemic inflammatory effects. The use of tumor necrosis factor inhibitors (TNFi) may reduce fracture risk by inhibiting inflammation. We assessed fracture rates in AS versus non-AS comparators and whether these rates have changed since the introduction of TNFi. METHODS: We used the national Veterans Affairs database to identify adults ≥18 years old with ≥1 International Classification of Diseases, Ninth Revision (ICD-9)/ICD-10 code for AS and at least 1 disease-modifying antirheumatic drug prescription. As comparators, we selected a random sample of adults without AS diagnosis codes. We calculated fracture incidence rates for AS and comparators, with direct standardization to the cohort structure in 2017. To compare fracture rates from 2000 to 2002 (pre-TNFi) versus 2004-2020 (TNFi era), we performed an interrupted time series analysis. RESULTS: We included 3,794 individuals with AS (mean age 53 years, 92% male) and 1,152,805 comparators (mean age 60 years, 89% male). For AS, the incidence rate of fractures increased from 7.9/1,000 person-years in 2000 to 21.6/1,000 person-years in 2020. The rate also increased among comparators, although the ratio of fracture rates (AS/comparators) remained relatively stable. In the interrupted time series, the fracture rate for AS patients in the TNFi era was nonsignificantly increased compared to the pre-TNFi era. CONCLUSION: Fracture rates have increased over time for both AS and non-AS comparators. The fracture rate in individuals with AS did not decrease after TNFi introduction in 2003.

Machine learning-based natural language processing to extract PD-L1 expression levels from clinical notes.

Introduction: PD-L1 expression is used to determine oncology patients' response to and eligibility for immunologic treatments; however, PD-L1 expression status often only exists in unstructured clinical notes, limiting ability to use it in population-level studies. Methods: We developed and evaluated a machine learning based natural language processing (NLP) tool to extract PD-L1 expression values from the nationwide Veterans Affairs electronic health record system. Results: The model demonstrated strong evaluation performance across multiple levels of label granularity. Mean precision of the overall PD-L1 positive label was 0.859 (sd, 0.039), recall 0.994 (sd, 0.013), and F1 0.921 (0.024). When a numeric PD-L1 value was identified, the mean absolute error of the value was 0.537 on a scale of 0 to 100. Conclusion: We presented an accurate NLP method for deriving PD-L1 status from clinical notes. By reducing the time and manual effort needed to review medical records, our work will enable future population-level studies in cancer immunotherapy.