Citations

Abstract Background Artificial intelligence (AI) augmented laboratory tests can improve quality, efficiency, cost-effectiveness, and staff satisfaction. However, the clinical success of these tests can introduce unforeseen challenges for model retraining. This study describes the discovery of an “AI label feedback loop” in a clinically implemented AI-augmented kidney stone composition test. Methods An AI-augmented kidney stone composition test (V1) has been previously deployed for clinical kidney stone characterization. After several years of clinical use, a retrained model (V2) was developed using 6 times more data. Model performance of both V1 and V2 were evaluated across 3 datasets: a recent production validation (hold-out) set (mostly AI-influenced labels), the original V1 validation set (pre-AI, entirely human-labeled), and a subset of recent cases with exclusively human-generated or human-corrected labels. Results V2 demonstrated a 10% lower concordance rate than V1 when evaluated on the recent production hold-out set, despite a much larger training dataset. Performance between V1 and V2 was similar when applied to the pre-AI validation set. Notably, V2 outperformed V1 on the recent subset of cases with human-only or human-corrected labels, particularly for less-common stone types. These findings revealed an AI label feedback loop, confounding retraining and evaluation. Conclusions The integration of AI into clinical practice can potentially influence reported test results, complicating the development and evaluation of future models. To mitigate AI label feedback loops, ongoing human annotation and careful validation set construction are essential. These strategies can ensure reliable performance assessment and support the safe evolution of clinical AI systems.

Background/Objectives: To evaluate risk factors for postoperative neurological deficits following cervical endoscopic unilateral laminotomy for bilateral decompression (CE-ULBD) and to determine whether intraoperative neurophysiological monitoring (IONM) can predict neurological compromise. Methods: A multicenter retrospective review was performed on 42 CE-ULBD procedures conducted between 2016 and 2024; 33 cases met the inclusion criteria with available imaging and electromyography data. Demographic, operative, and neurophysiological variables were analyzed. Preoperative stenosis severity was graded using the Kang MRI system. Intraoperative IONM data, including electromyography firing and motor evoked potential (MEP) changes, were correlated with new postoperative weakness. Results: The cohort (69.1% male, mean age 70.2 ± 1.7 years, mean BMI 29.6 ± 1.1) included 56 decompressed levels. The most common operative levels were C3-4 (37%) and C4-5 (24%). Postoperative weakness occurred in four patients (12.1%), all of whom had severe (Grade 3) preoperative stenosis. Among these, 50% exhibited preoperative weakness. Neuromonitoring changes correlated significantly with postoperative weakness (Fisher’s Exact, p < 0.001); 100% of patients with new post-operative weakness had sustained MEP decrease at the time of closure. Conclusions: Patients with severe cervical stenosis and preoperative weakness are at heightened risk of postoperative neurological deficits following CE-ULBD. Elevated epidural pressure from continuous irrigation in a constricted canal may exacerbate cord compression, particularly in those with preexisting myelopathy. IONM changes strongly correlate with new deficits and may exacerbate cord compression, particularly in those with preexisting myelopathy, and may serve as an early warning system for impending neurological injury. Surgeons should exercise caution and maintain low irrigation pressures in patients with severe stenosis undergoing endoscopic cervical decompression.

ABSTRACT Despite advances in therapy, extramedullary disease (EMD) remains an aggressive form of multiple myeloma associated with poor outcomes. Patients with true EMD, in which plasmacytomas have become completely independent of bone, have a particularly poor prognosis. The pathogenesis of EMD is driven by complex mechanisms involving loss of adhesion molecules, heterogeneous genetic and epigenetic changes, and a solid tumor‐like architecture within the microenvironment. Although the introduction of advanced imaging techniques and immunotherapy has led to improved detection and more promising outcomes and regimens, respectively, more prospective studies dedicated to true EMD are needed.

Abstract INTRODUCTION Plasma phosphorylated tau at threonine 217 (p‐tau217) has emerged as one of the most promising blood‐based biomarkers for Alzheimer's disease (AD). While rare, heterophilic antibodies (HAb) are a persistent potential confounding factor in immunoassays. METHODS Potential HAb interference in the Lumipulse G pTau217 assay in samples exhibiting concentrations > 10 pg/mL were investigated. Samples were subjected to HAb blocking reagent (HBT), and in some cases, polyethylene glycol (PEG) precipitation and serial dilutions. RESULTS In 14 of the 15 suspected HAb cases, HBT treatment greatly reduced measured p‐tau217 concentrations. and in one of three specimens selected for further investigation, reduced Aβ42 concentrations PEG precipitation in the three selected samples also reduced p‐tau217 concentrations, while serial dilution yielded mixed effects. Control specimens were unaffected by sample treatment or dilutions. DISCUSSION HAb may result in falsely elevated p‐tau217 and p‐tau217/Aβ42 ratio. HAb interference should be considered in cases of unusually high p‐tau217 concentrations prior to clinical interpretation. Highlights Potential heterophile antibody (HAb) interference in the Lumipulse p‐tau217 immunoassay was investigated. Investigation was performed using HAb blocking reagent, and in some cases also using PEG precipitation . Patients with p‐tau217 concentrations > 10 pg/mL exhibited positive HAb interference in 14 of 15 cases. HAb interference can lead to falsely positive p‐tau217/Aβ42 ratios.

Abstract Background The polymorphic APOE gene is a susceptibility gene for Alzheimer's disease (AD). The ε4 allele is associated with increased risk for AD, particularly late‐onset disease, in a dose‐dependent manner. Additionally, homozygous ε4/ε4 individuals have increased risk of adverse events when undergoing amyloid lowering treatments. APOE genotyping is considered the gold standard for determination of ApoE4 zygosity. Recently, proteotype assays for ApoE4 have become available for high‐throughput immunoassay analyzers. These assays utilize the same specimen type as AD blood biomarkers and can potentially offer improved turnaround times compared to genotyping. In this study, we compare the performance of two automated ApoE4 proteotype research use only (RUO) immunoassays against APOE genotyping. Method EDTA‐plasma samples ( n = 65) from individuals undergoing ApoE PCR‐based genotyping for clinical purposes were evaluated. Samples were tested with the Beckman Coulter APOE ε4 RUO assay on the Access 2 and DXI 9000 instruments (Beckman Coulter, Inc), which provides algorithm calculated results for determining ApoE4 zygosity. Manufacturer generated data from 300 genotyped samples was used to calculate estimated cut‐points for ApoE4 zygosity for this study with results <0.55 classified as no E4, ≥0.55 to <4.55 as heterozygous E4, and ≥4.55 as homozygous E4. Samples were also tested with the Lumipulse G ApoE4 and Pan‐ApoE RUO assays on the Fujirebio Lumipulse G1200 analyzer. The ratio of ApoE4/Pan‐ApoE was utilized for determining zygosity with results <5% classified as no E4, ≥5% to <75% as heterozygous E4, and ≥75% as homozygous E4, as listed in the manufacturer IFU. Result APOE genotype distributions were (n, %): ε3ε4 (30, 46%), ε3ε3 (18, 28%), ε4ε4 (12, 18%), ε2ε3 (3, 5%), ε2ε4 (2, 3%). ApoE4 zygosity was categorized as no ε4 (21, 32%), heterozygous ε4 (32, 49%), or homozygous ε4 (12, 19%). The Beckman Coulter APOE ε4 RUO assay classified 65/65 (100%) patients correctly using manufacturer generated cut‐points on both instruments. The Lumipulse G ApoE4 and Pan‐ApoE RUO assays classified 61/65 (94%) patients correctly using the manufacturer listed cut‐points. Conclusion ApoE4 proteotype immunoassays may represent simplified and more accessible testing alternatives to genotyping for the assessment of ApoE4 zygosity in individuals being considered for amyloid lowering treatments.

Abstract Background Motor disturbances are common in neurodegenerative disorders. Digital health technologies can be a tool for remote, accessible, and low‐burden motor assessments. The ALLFTD Mobile App includes a finger‐tapping test (mApp‐FTT) to assess dexterity. We evaluated the reliability and construct validity of smartphone‐based measures of finger tapping in an in‐person sample. Methods Participants ( n = 33; mean age=68.2 [SD=8.6]) completed the mobile finger‐tapping test (mApp‐FTT) on iOS and Android devices (two counterbalanced 20‐second trials per hand) and the mechanical finger‐tapping test (FTT). The sample included cognitively unimpaired participants (22.2%) and those diagnosed with frontotemporal dementia‐spectrum disorders (41.7%), Alzheimer's dementia (22.2%), and progressive supranuclear palsy or corticobasal syndrome (13.9%). A motor domain was calculated based on neurological exam and added to the Clinical Dementia Rating® plus National Alzheimer's Coordinating Center–Frontotemporal Lobar Degeneration (CDR®+NACC‐FTLD). Pearson's correlations assessed associations between mApp‐FTT performance and FTT performance, clinical severity (CDR®+NACC‐FTLD Box Score), motor rating scales (Progressive Supranuclear Palsy Rating Scale [PSPRS] and Unified Parkinson's Disease Rating Scale [UPDRS]), and cognitive function. Due to small sample size, PSPRS subdomains were dichotomized, and t‐tests were performed to evaluate group differences based on symptom presence. Results Total mApp‐FTT taps on iOS and Android were strongly correlated ( r = 0.86, 95% CI [0.76, 0.92], p < .001). mApp‐FTT was associated with manual FTT on iOS ( r = 0.70, 95% CI [0.52, 0.82], p < .001) and Android ( r = 0.62, 95% CI [0.41, 0.77], p < .001). Among iOS users, total PSPRS score was negatively correlated with mApp‐FTT ( r = ‐0.60, 95% CI [‐0.75, ‐0.39], p < .001). Worse mApp‐FTT performance was observed in participants with higher limb rigidity (mean difference=15.46, 95% CI [3.40, 27.52], p < .001), limb dystonia (mean difference=19.22, 95% CI [5.67, 32.77], p < .001), and finger tapping (mean difference=12.74, 95% CI [3.42, 22.06], p < .001). Total UPDRS score correlated with lower mApp‐FTT ( r = ‐0.48, 95% CI [‐0.66, ‐0.28], p < .001), as did disease severity (CDR®+NACC‐FTLD Box Score; r=‐0.46, 95% CI [‐0.66, ‐0.22], p < .001). Conclusion Smartphone‐based finger‐tapping assessments effectively detect motor and cognitive impairment, with stronger associations for motor function. Findings in iOS users highlight the need to assess operating system‐related biases and validate in Android. Remote smartphone assessments offer a scalable tool for tracking disease progression, warranting larger longitudinal studies.

Abstract Background Chronic kidney disease (CKD) is associated with increased blood‐based biomarker concentrations, potentially confounding their utility in Alzheimer's disease (AD) evaluation. While plasma phosphorylated tau217 ( p ‐tau217) is elevated in the presence of amyloid related pathology, amyloid beta 1‐42 (Aβ42) is decreased. We investigated the association of plasma p ‐tau217, non‐phosphorylated tau217 (np‐tau217), Aβ42, Aβ40, as well as the Aβ42/Aβ40 and p ‐tau217/Aβ42 ratios, versus decreased estimated glomerular filtration rate (eGFR in mL/min/1.73 m 2 ) in individuals without amyloid pathology. Method This study included participants with CKD stages 1‐2 (eGFR 60‐98, n = 50), and stages 3‐4 (eGFR 25‐59, n = 33), who were Pittsburgh 11 C‐Compound B amyloid‐PET negative (Centiloid of < 25). Plasma p ‐tau217, np‐tau217, Aβ42, and Aβ40 were measured by immunoprecipitation mass spectrometry (C 2 N Diagnostics). %p‐tau217 [( p ‐tau217/np‐tau217) x100], Aβ42/Aβ40 ratio, and p ‐tau217/Aβ42 ratio were calculated. Log‐Log (power function) best‐fit regression analysis was used to quantitatively predict effects of changes in eGFR (2021 CKD‐EPI equation) versus each single analyte or ratio (Figure 1). Result Selected best‐fit analyte equations versus eGFR are shown in Table 1. The predicted percent increase of p ‐tau217 as eGFR decreases from 60 to 45 was 52% (95% CI: 30–78%). Significantly smaller increases of 16% were predicted for Aβ42 (95% CI: 12–21%), and 19% for Aβ40 (95% CI: 16–23%) over the same eGFR range. The best‐fit predicted increase for the p ‐tau217/Aβ42 ratio and %p‐tau217 as eGFR decreases from 60 to 45 was 31% (95% CI: 12–53%), and 19% (95% CI: 2–38%), respectively. While not recommended for assessment of amyloid pathology, the Aβ42/Aβ40 ratio did not exhibit a predicted increase (‐2.3% decrease, 95% CI: ‐5.2–0.6%) versus decreasing eGFR from 60 to 45. Conclusion The use of either %p‐tau217 or the p ‐tau217/Aβ42 ratio (to a lesser degree) reduced the confounding effects of decreasing eGFR as compared to p ‐tau217 alone. Aβ42 and Aβ40 exhibited similar reduced predicted increases versus decreasing eGFR as compared to p ‐tau217. Relative to the predicted effects of decreasing eGFR on p ‐tau217 measurements, the p ‐tau217/Aβ42 ratio relative increase was reduced by 40%, while the relative increase was reduced by 63% for %p‐tau217 measurements.

Abstract Background Clinically available plasma p ‐tau217 assays may be utilized in the determination of eligibility for amyloid‐lowering therapies. Our institution recently introduced a plasma p ‐tau217 assay in clinical practice using a two‐cutpoint approach that optimizes the accuracy while minimizing false positive risk and categorizing ≤20% of results as intermediate. In this study, we prospectively evaluate the performance of a plasma p ‐tau217 assay for detection of amyloid pathology using these predefined cutpoints. Method Patients recruited were being considered for lecanemab therapy at a tertiary care subspecialty clinic. Plasma p ‐tau217 was measured using the clinically available Fujirebio Lumipulse p ‐tau217 assay at Mayo Clinic, and results (pg/mL) were classified as either negative (≤0.185), intermediate (0.186–0.324) or positive (≥0.325). Amyloid pathology was determined by amyloid‐PET and/or CSF p ‐tau181/Aβ42. Result Ninety‐three patients were recruited between November 2023 and December 2024. Of these, 49 underwent amyloid‐PET, 4 had CSF p ‐tau181/Aβ42 measurements, and 40 had both PET and CSF results (with 3/40 discordant cases), for use in the assessment of amyloid pathology. The three discordant cases were considered amyloid positive for purposes of plasma p ‐tau217 comparison. Overall amyloid positivity was 90%. Plasma p ‐tau217 results were positive in 83% (77/93), negative in 6% (6/93), and intermediate in 11% (10/93) of patients. An elevated p ‐tau217 predicted the presence of amyloid positivity with a diagnostic accuracy of 96% (95% CI 90% − 99%). In the intermediate range, 80% of patients were positive for amyloid pathology. In patients with a definitive p ‐tau217 result (positive or negative), the positive and negative predictive values were 97% and 83%. There were three patients with discordant results: one PET+/p‐tau217‐, one PET‐/p‐tau217+, and one CSF‐/p‐tau217+. Clinical presentation of these discordant cases was examined, and relevant findings will be discussed. Conclusion The Lumipulse plasma p ‐tau217 assay exhibited acceptable diagnostic accuracy for the determination of amyloid pathology when compared to either amyloid‐PET or CSF p ‐tau181/Aβ42. This prospective study confirms the utility a two‐cutpoint approach for interpreting plasma p ‐tau217 for the assessment of amyloid pathology eligibility for amyloid‐lowering therapies in a heterogenous clinical cohort.

Abstract Background Recent technological advances have enhanced the precision of blood‐based multiplex proteomics and their application in Alzheimer's disease. However, these approaches remain underutilized in frontotemporal lobar degeneration (FTLD). Familial FTLD cohorts offer a unique opportunity to study presymptomatic periods of disease that cannot be studied in sporadic cases. We leveraged NUcleic acid Linked Immuno‐Sandwich Assay (NULISAseq) to identify blood‐detectable proteins that change prior to symptom onset in familial FTLD. Method 120 FTLD mutation carriers (40 C9orf72 , 40 GRN , 40 MAPT ) and 40 age and sex‐matched noncarrier controls from the ALLFTD Consortium completed blood draws and clinical assessment. Targeted plasma proteomics was performed via NUcleic acid Linked Immuno‐Sandwich Assay (NULISAseq; 132 proteins). Disease age estimates measuring predicted symptom onset were derived from a validated model incorporating FTLD gene‐specific clinical, neuroimaging, and biomarker data. Mass linear regression models identified disease age ‘thresholds’ where individual proteins significantly diverged between each mutation carrier group and controls. NULISAseq targets were validated against orthogonal assays, specifically Simoa (plasma) and SomaScan (cerebrospinal fluid [CSF]). Result Compared to controls, 14 plasma proteins significantly diverged ( p <.05) in C9orf72 , 24 in GRN , and 13 in MAPT . Beyond NEFL, 10 additional proteins significantly diverged in more than one mutation group, including markers of synaptic function (NPTX2, NPTXR) and vascular integrity (PGF) that deviated prior to estimated symptom onset (disease age range: ‐30 to 0). Proteins with gene‐specific signals included GFAP (increased in GRN at disease age ‐17), ENO2 (decreased in C9orf72 at disease age ‐7), and FLT1 (increased in MAPT at disease age ‐6). Plasma NULISA‐based NEFL and GFAP showed strong concordance when measured via plasma Simoa ( r s>0.88) and were among 38 targets with significant concordance when measured via CSF SomaScan ( r range: 0.16 to 0.75, p s<.05). Conclusion Alterations in plasma protein abundance are measurable in familial FTLD prior to predicted symptom onset. The validity and brain‐based relevance of blood‐detectable NULISA targets is bolstered by the observed convergence with plasma Simoa and CSF SomaScan. Replication of top hits in larger familial and sporadic FTLD sample sizes is warranted to determine generalizability and suitability for clinical trials.

Abstract Background Alzheimer's disease (AD) blood biomarkers (BBMs) are increasingly used in clinical trials and clinical practice. However, widespread adoption of BBMs requires further understanding of factors that might influence the interpretation of these biomarkers which would allow for the development of standardized sample collection protocols. For example, studies of the effect of fasting on AD BBM concentrations have yielded inconsistent findings. In this study, we evaluated plasma concentrations of amyloid‐beta(Aβ) 42, Aβ40, phosphorylated Tau ( p ‐tau) 181, p ‐tau217, and neurofilament light chain (NfL) in cognitively normal individuals under fasting and non‐fasting conditions. Method EDTA‐plasma was collected from healthy donors ( n = 14; age >55yrs old) two weeks apart, at the same time of day. The first sample collection was under fasting conditions in which donors did not eat or drink (except water) for at least 12 hours prior to collection. The second sample collection was under non‐fasting conditions in which donors were instructed to eat a meal in the morning, approximately half an hour prior to collection. Plasma Aβ42, Aβ40, p ‐tau181, p ‐tau217, and NfL concentrations were measured on the Fujirebio Lumipulse G1200 analyzer. Wilcoxon signed‐rank test was used to compare values between the fasting and non‐fasting results. Spearman's rank correlation was used to assess biomarker correlation between fasting status. Result Moderate to strong correlation was observed between the fasting and non‐fasting measurements: p ‐tau217 (r s : 0.86), p ‐tau181 (r s : 0.89), NfL (r s : 0.94), p ‐tau217/Aβ42 (r s : 0.80), Aβ42/40 (r s : 0.66), Aβ42 (r s : 0.77), and Aβ40 (r s : 0.92). Average percent difference between the 2 conditions was: p ‐tau217 27%, p ‐tau181 6%, NfL 9%, p ‐tau217/ Aβ42 25%, Aβ42/40 4%, Aβ42 2%, and Aβ40 ‐6%. Statistically significant differences ( p ≤ 0.05) between fasting and non‐fasting measurements were observed for p ‐tau217, NfL, p ‐tau217/Aβ42, Aβ42/Aβ40, and Aβ42; but not for Aβ40 or p ‐tau181. Conclusion This data suggests that for some AD BBMs, there may be statistically significant differences in concentrations of samples collected in fasting versus non‐fasting conditions. However, further studies are needed to determine if the observed differences may affect the clinical interpretation of these BBMs in individuals with amyloid related pathology.