Citations

Abstract Error analysis is a critical step in evaluating and improving clinical concept extraction models, the most common clinical natural language processing (NLP) task. Unlike corpus annotation, which follows standardized protocols for creating gold-standard datasets, error analysis requires nuanced judgment grounded in both clinical expertise and NLP knowledge. This is especially important given the heterogeneity of clinical text, where variations in documentation style, note structure, and terminology can substantially influence model behavior. Despite its importance, there is currently no standardized, user-level framework to support systematic error analysis in clinical concept extraction task. In this study, we developed and validated MedError, a machine-assisted, human-in-the-loop framework designed to standardize and enhance error analysis for clinical concept extraction tasks. We collected and manually curated a corpus of 1,187 unique errors from a total of 4,237 notes across three different distinct hospitals. The error categories were defined using our previously validated error taxonomy and included 480 false negatives and 707 false positives across 25 error types and 48 clinical concept categories. We evaluated the performance of three proprietary and three open-source large language models (LLMs) in automatically classifying these errors into 26 and 15 predefined categories. We further developed a machine-assisted framework, MedError, which integrates best practices in error analysis, LLM-assisted classification and reasoning, and a user-friendly interface to enable more efficient, reproducible, and context-aware error analysis. The framework supports both single-site and federated multisite error analysis, facilitating the effective translation of clinical NLP systems into real-world settings.

Importance In the literature, there is a lack of data reporting tumor control rates after radiotherapy in actively growing vestibular schwannomas (VS). Data for this rarely studied population are needed. Objective To estimate tumor control rates in radiologically growing VS treated with first-line radiotherapy. Design, Setting, and Participants This international, multicenter cohort study used prospectively collected data from patients with growing unilateral VS treated first-line with radiotherapy between January 2000 and December 2023 from 8 tertiary referral skull base units. The data were analyzed in June 2025. Exposures Radiotherapy as an initial treatment for VS. Main Outcomes and Measures The primary outcome was treatment failure, ie VS growth postradiotherapy, which was predefined as an increase in maximum intracranial tumor diameter (ICTD) of 3 mm or greater within the first 2 years after radiotherapy or 2 mm or greater thereafter. Secondary outcomes were treatment failure based on different definitions of VS growth: (1) an increase in ICTD of 2 mm or greater, (2) an increase in ICTD of 3 mm or greater, and (3) conversion to surgery. Results A total of 1883 patients (975 female individuals [51.8%]; median age at diagnosis, 63 years [IQR, 53-71 years]) were included in the study. Using the primary definition of treatment failure (an increase in ICTD of ≥3 mm within the first 2 years postradiotherapy or ≥2 mm thereafter), the Kaplan-Meier estimate yielded a 10-year tumor control rate of 76.1% (95% CI, 72.7%-79.2%). For secondary outcome definitions, 10-year tumor control rates were 60.1% (95% CI, 57.5%-64.3%) for an ICTD increase of 2 mm or greater, 78.3% (95% CI, 75.0%-81.2%) for an increase of 3 mm or greater, and 92.6% (95% CI, 90.4%-94.3%) for conversion to surgery. Neither pretreatment tumor size nor tumor location (intracanalicular vs extracanalicular) were significantly associated with treatment failure. Conclusions and Relevance The results of this cohort study provide tumor control outcomes for radiologically growing VS treated with radiotherapy using several clinically relevant definitions of growth. By focusing exclusively on this rarely isolated subgroup, the findings offer targeted data to potentially inform treatment expectations and future research.

Background/Objectives: Results from a well-designed trial provide evidence to support approval of truly effective treatments or discontinuation of ineffective treatments. However, the information available at the time of trial design may be limited which may lead to underpowered trials. This work aims to evaluate the impact of design assumption misspecifications on the statistical power of randomized trials with survival outcomes. Methods: The impact of the design assumption misspecifications on statistical power of four different statistical methods was investigated in a simulation study. The methods include the log-rank test, MaxCombo test, the test of difference in survival probability, and test of difference in restricted mean survival time (RMST). The deviations considered include the survival rate in the control arm, the expected treatment effect in terms of magnitude and pattern, accrual rate, and drop-out rate. Results: Deviations in the control arm’s survival distribution have no impact on the power of the log-rank and MaxCombo tests but it affects the trial duration since trials designed with these tests require the total number of events to be met before the final analysis can be conducted. Misspecified treatment effect has similar effect on the statistical power of all four methods. When the proportional hazards assumption is misspecified, the RMST is more robust with a larger early treatment effect, while the survival probability and the MaxCombo tests are more robust with a larger late treatment effect and crossing hazards. Conclusions: Selecting the appropriate statistical tests to design a trial depends on the goal of the trial, the mechanism of action of the experimental treatment, the survival quantity of clinical interest, and the pattern of the expected treatment effect. The final design should be based on assumptions that are as accurate as possible, and the potential impacts of deviations from these assumptions on the trial’s statistical power should be carefully considered.

ABSTRACT Brain atrophy is a common feature of neurodegenerative disorders, including Alzheimer's disease, Parkinson's disease, multiple sclerosis, and more recently SARS‐CoV‐2 infection for which the neuroinflammatory mechanism is not fully understood. Nevertheless, neuroinflammation and CD8 T cell accumulation is frequently observed in postmortem brain tissue of patients with neurodegeneration. We therefore developed a murine model of brain atrophy using the Theiler's murine encephalomyelitis virus (TMEV) infection model of multiple sclerosis. We employ single major histocompatibility complex (MHC) class I molecule conditional knockout mice generated by our program to analyze the contribution of immune cell infiltration and onset of atrophy. TMEV infected C57BL/6 mice that singularly express H‐2K b or H‐2D b were evaluated. TMEV infection of these mice resulted in ventricular atrophy at 14 d.p.i. However, H‐2D b expressing mice presented with significantly greater and continuous ventricular atrophy compared to H‐2K b expressing mice and MHC class I knockout control mice. Flow cytometric analysis revealed H‐2D b expressing mice had greater brain infiltrating CD8 T cell responses at 7 and 28 dpi which correlated with the extent of ventricular atrophy. Meanwhile, H‐2K b and MHC class I knockout mice also had distinct positive correlations with ventricular atrophy and global neuroinflammation in general. These findings support the role of immune cell infiltration into the CNS as a putative mechanism of ventricular atrophy following TMEV infection, with the progressive enlargement of ventricular atrophy observed in H‐2D b mice being associated with the presence of long‐lived memory CD8 T cells residing in the brain.

Abstract Background FGFR alterations are known to be driver alterations in several tumor types. We aimed to assess the efficacy of pemigatinib, an oral FGFR1-3 inhibitor, in patients with metastatic or unresectable colorectal cancer whose tumors harbored FGF/FGFR alterations. Patients and Methods The ACCRU-GI-1701 is a single-arm phase II trial which enrolled patients with previously treated FGF/FGFR-altered metastatic colorectal cancer to receive oral pemigatinib daily in 21-day cycles. The primary endpoint is objective response. Secondary endpoints include clinical benefit, progression-free survival, overall survival, quality of life, and adverse events (AEs). This trial was registered with ClinicalTrials.gov (NCT04096417). Results Of the 14 patients included in the interim analysis, the objective response rate as well as clinical benefit rate were 0%. Given these results, the trial closed to enrollment after stage one due to futility. A total of 42.9% of patients had at least one grade 3 or higher AE, the most common being anemia and fatigue. Conclusion Pemigatinib monotherapy did not lead to objective responses in patients with chemorefractory metastatic colorectal cancer harboring FGF/FGFR alterations, although it was overall relatively well tolerated with no new safety signals. Notably, 93% (n = 13) of patients had only FGF/FGFR mutations and amplifications; one patient had an FGFR3-WHSC1 fusion at a low cfDNA percentage (0.02%).

Objective To provide a comprehensive analysis of initial suicide attempts, covering incidence, risk factors, outcomes, and healthcare use in the month before and the month after the attempts. Design Comprehensive analysis of the Swedish population that included three designs: a retrospective cohort study to investigate incidence and healthcare use, a nested case-control study to investigate risk factors, and a matched cohort study to examine subsequent suicide attempts and mortality. Setting Comprehensive Swedish national registers that include patient diagnoses from hospitals and specialist outpatient care, and cause of death information updated to the end of 2019. Participants 3.7 million people born in Sweden in 1963-98 and followed from age 10 to 57 years. Main outcome measure First lifetime suicide attempt identified in patient and death registers using ICD (international classification of diseases) codes for intentional self-harm, any self-harm with lethal methods or requiring hospital admission, or any self-harm resulting in death. Results The lifetime risk of an initial suicide attempt in the study population was 4.6%, with greater risk in females and highest risk between ages 18 and 24. One in 10 families in Sweden had at least one family member who attempted suicide. Overdose and poisoning were the most common methods. Previous psychiatric disorders, general medical diseases, and adverse life events were associated with increased risk of initial suicide attempt, while higher socioeconomic status was associated with decreased risk. People with an initial suicide attempt were at substantially increased risk of subsequent attempts (hazard ratio 23.4), death by suicide (16.4), and all cause mortality (7.3). At least 60% of those who made an initial suicide attempt had a healthcare contact in the month before the attempt. Conclusions This study provides comprehensive data on the incidence, risk factors, outcomes, and healthcare use of initial suicide attempts in the Swedish population, highlighting the need for systematic prevention efforts for people who have attempted suicide for the first time.

LBA1 Background: Standard adjuvant chemotherapy of stage III colon cancer consists of a fluoropyrimidine plus oxaliplatin. In patients (pts) with stage III colon cancer and deficient mismatch repair (dMMR), the benefit of an immune checkpoint inhibitor combined with adjuvant chemotherapy is unknown. The phase III ATOMIC trial (NCT02912559) was conducted to determine whether atezolizumab (atezo), an anti-PD-L1 antibody, can improve pt outcomes when added to adjuvant 5-fluorouracil, leucovorin plus oxaliplatin (mFOLFOX6) in pts with stage III dMMR tumors. Methods: We conducted an NCI-sponsored, multicenter and randomized phase III trial in pts with surgically resected stage III dMMR colon adenocarcinoma (any T, N 1,2 M 0 ). Pts, age > 12 years (yr), were accrued at NCTN sites and the German AIO. Tumor dMMR was determined by local immunohistochemistry and centrally verified. Pts were randomized 1:1 to mFOLFOX6 plus atezo (840 mg IV q2 weeks) for 12 cycles (6 months)[mo] followed by atezo monotherapy for 13 cycles (12 mo total) versus mFOLFOX6 alone for 12 cycles. Randomization stratification factors were N-stage (N 1 /N 1c vs N 2 ), T-stage (T 1 -T 3 vs T 4 ) and site (proximal vs distal). The primary endpoint was disease-free survival (DFS); secondary endpoints were overall survival and adverse event (AE) profile (CTCAE, PRO-CTCAE). Primary efficacy analysis was done in the intent-to-treat population; DFS was compared by arm (stratified log-rank test). Hazard ratio (HR) and 95% confidence interval (CI) were calculated using a stratified Cox model; 3-yr DFS was determined by Kaplan-Meier method. Among 700 pts., 165 DFS events with two interim analyses (50% , 75% of events) yielded 90% power to detect HR 0.6 (3-yr DFS 75% vs. 84.2%) assuming exponential survival and 1-sided alpha (0.025). Results: From 9/2017 to 1/2023, 712 pts were randomized (1 pediatric) to either atezo plus mFOLFOX6 (n= 355; atezo arm) or mFOLFOX6 (n= 357). Median pt age was 64 yr. 55.1% were female. Among tumors, 83.8% were proximal, 46.1% were clinical low risk (T 1-3 N 1 ) and 53.9% high risk (T 4 and/or N 2 ). At the second interim analysis, median pt follow-up was 37.2 mo (interquartile range, 24.2 to 55.5) and 124 DFS events were observed. Three-year DFS was 86.4 % (95% CI, 81.8 to 89.9) in the atezo arm and 76.6 % (95% CI, 71.3 to 81.0) in the mFOLFOX6 arm (HR, 0.50; 95% CI, 0.35 to 0.72). Stratified log-rank p-value was <0.0001, crossing the pre-specified efficacy boundary of 0.009. Efficacy for the atezo arm was consistent across subgroups, including pts >70 yr and low- and high-risk groups. Treatment-related > grade 3 AEs occurred in 71.7 % of pts in the atezo arm vs 62.1 % in the mFOLFOX6 arm. Conclusion: The addition of atezolizumab to mFOLFOX6 significantly improved DFS and should be considered the new adjuvant standard of care for patients with dMMR stage III colon cancer. Support: U10CA180821, U10CA180882, U24CA196171; Genentech, a member of the Roche group; https://acknowledgments.alliancefound.org. Clinical trial information: NCT02912559 .

e23317 Background: mCRC poses significant challenges, particularly in patients (pts) who progress after multiple lines of therapy. Limited effective treatment options are available in the advanced chemotherapy-refractory setting. The FRESCO-2 trial demonstrated that Fruq, a selective VEGFR inhibitor, provides meaningful overall survival (OS) benefit in pts with refractory mCRC. However, there is limited real-world evidence regarding Fruq's efficacy in patients treated after approved third-line therapies. Methods: This retrospective, multi-site cohort study evaluated outcomes in 33 mCRC pts who received Fruq following treatment with Rego and/or TAS-102 with or without bevacizumab. Patients were stratified into two groups: Fruq as a second late-line therapy, after either Rego or TAS-102, (27 patients) and as a third late-line therapy, after both Rego and TAS-102, (6 patients). Overall survival (OS) was calculated from Fruq initiation to death, while time to treatment discontinuation (TTD) was defined as the duration from Fruq initiation to cessation for any reason, including death. Survival analysis was performed using the Kaplan-Meier method to estimate median survival, and the comparison between groups were done using the log-rank test. Results: Among the 33 pts, the median age was 58 years (range: 35–84), with 57.6% being male. Hepatic metastases were present in 78.8%, and 60.6% had left-sided primary tumors. RAS mutations were observed in 63.6%, and 97% were BRAF wild-type. Median OS was significantly longer in pts who received Fruq as a second late-line therapy (7.2 months) compared to those receiving it as a third late-line therapy (4.0 months) (p = 0.039). Median TTD was similar between groups (2.4 vs. 2.8 months, p = 0.916). Conclusions: This real-world analysis highlights the clinical benefit of Fruquintinib in mCRC patients following treatment with Regorafenib and/or TAS-102. The data suggest flexibility in treatment choices, with Fruquintinib showing improved OS when used earlier in the treatment sequence. These findings underscore the importance of exploring optimal sequencing strategies and evaluating Fruquintinib as a potential option before other approved agents in the refractory setting. Clinical outcomes with fruquintinib as second vs. third late-line therapy. Fruq as Second Late-Line Fruq as Third Late-Line p-value Median OS 7.2 months 4 months 0.039 Median TTD 2.4 months 2.8 months 0.916 OS: Overall Survival; TTD: Time to Treatment discontinuation.