Clinical chemistry最新文献

Background: Untargeted metabolomics has shown promise in expanding screening and diagnostic capabilities for inborn errors of metabolism (IEMs). However, inter-batch variability remains a major barrier to its implementation in the clinical laboratory, despite attempts to address this through normalization techniques. We have developed a sustainable, matrix-matched reference material (RM) using the iterative batch averaging method (IBAT) to correct inter-batch variability in liquid chromatography-high-resolution mass spectrometry-based untargeted metabolomics for IEM screening.

Methods: The RM was created using pooled batches of remnant plasma specimens. The batch size, number of batch iterations per RM, and stability compared to a conventional pool of specimens were determined. The effectiveness of the RM for correcting inter-batch variability in routine screening was evaluated using plasma collected from a cohort of phenylketonuria (PKU) patients.

Results: The RM exhibited lower metabolite variability between iterations over time compared to metabolites from individual batches or individual specimens used for its creation. In addition, the mean variation across amino acid (n = 19) concentrations over 12 weeks was lower for the RM (CVtotal = 8.8%; range 4.7%-25.3%) compared to the specimen pool (CVtotal = 24.6%; range 9.0%-108.3%). When utilized in IEM screening, RM normalization minimized unwanted inter-batch variation and enabled the correct classification of 30 PKU patients analyzed 1 month apart from 146 non-PKU controls.

Conclusions: Our RM minimizes inter-batch variability in untargeted metabolomics and demonstrated its potential for routine IEM screening in a cohort of PKU patients. It provides a practical and sustainable solution for data normalization in untargeted metabolomics for clinical laboratories.

Background: Celiac disease (CeD) has an estimated prevalence of 1%-3%. The classical clinical presentation is malabsorption, but now patients may present with more subtle symptoms such as constipation, osteoporosis, or iron deficiency anemia. Children may also present with poor growth.CeD has a strong genetic component, and high-risk groups include first-degree relatives with CeD, patients with co-existing autoimmune diseases, and patients with chromosomal aberrations.

Content: Diagnostic tests for CeD include duodenal histology, serology, and genetic testing. Duodenal histology has traditionally been the gold standard of diagnosis. However, serological tests, especially IgA tissue transglutaminase antibodies (TTG-IgA), are widely used and diagnostic algorithms are based primarily on TTG-IgA as a starting point. Human leukocyte antigen typing may also be incorporated to determine genetic risk for CeD. Guidelines for children endorse biopsy avoidance provided high levels of TTG-IgA, with diagnostic accuracy being comparable to duodenal biopsy. Confirmation may be achieved by identifying IgA endomysial antibodies in a separate blood sample. Subjects with low positive TTG-IgA levels and subjects with IgA deficiency need a biopsy to establish a diagnosis of CeD. The clinical follow-up of CeD usually includes a repeat TTG-IgA examination. In adults, healing may be delayed or incomplete, and a rare consequence of refractory celiac disease is transformation to enteric T-cell lymphoma.

Summary: Laboratory testing, in particular TTG-IgA, plays a central role in the diagnosis and has an accuracy comparable to histology. Diagnostic algorithms utilizing laboratory testing are critical for the development of novel strategies to improve diagnosis.

Background: In clinical laboratories, the precision and sensitivity of autoverification technologies are crucial for ensuring reliable diagnostics. Conventional methods have limited sensitivity and applicability, making error detection challenging and reducing laboratory efficiency. This study introduces a machine learning (ML)-based autoverification technology to enhance tumor marker test error detection.

Methods: The effectiveness of various ML models was evaluated by analyzing a large data set of 397 751 for model training and internal validation and 215 339 for external validation. Sample misidentification was simulated by random shuffling error-free test results with a 1% error rate to achieve a real-world approximation. The ML models were developed with Bayesian optimization for tuning. Model validation was performed internally at the primary institution and externally at other institutions, comparing the ML models' performance with conventional delta check methods.

Results: Deep neural networks and extreme gradient boosting achieved an area under the receiver operating characteristic curve of 0.834 to 0.903, outperforming that of conventional methods (0.705 to 0.816). External validation by 3 independent laboratories showed that the balanced accuracy of the ML model ranged from 0.760 to 0.836, outperforming the balanced accuracy of 0.670 to 0.773 of the conventional models.

Conclusions: This study addresses limitations regarding the sensitivity of current delta check methods for detection of sample misidentification errors and provides versatile models that mitigate the operational challenges faced by smaller laboratories. Our findings offer a pathway toward more efficient and reliable clinical laboratory testing.

Background: Conventional population-based reference intervals (popRIs) are established on the ranking of single measurement results from at least 120 reference individuals. In this study, we aimed to explore a new model for popRIs, utilizing biological variation (BV) data to define the reference interval (RI) limits and compared BV-based popRI from different sample sizes with previously published conventional popRIs from the same population.

Methods: The model is based on defining the population set point (PSP) from single-measurement results of a group of reference individuals and using the total variation around the PSP, derived from the combination of BV and analytical variation, to define the RI limits. Using data from 143 reference individuals for 48 clinical chemistry and hematology measurands, BV-based popRIs were calculated for different sample sizes (n = 16, n = 30, and n = 120) and considered acceptable if they covered 90% of the population. In addition, simulation studies were performed to estimate the minimum number of required reference individuals.

Results: The median ratio of the BV-based to conventional RI ranges was 0.98. The BV-based popRIs calculated from the different samples were similar, and most met the coverage criterion. For 25 measurands ≤16 reference individuals and for 23 measurands >16 reference individuals were required to estimate the PSP.

Conclusions: The BV-based popRI model delivered robust RIs for most of the included measurands. This new model requires a smaller group of reference individuals than the conventional popRI model and can be implemented if reliable BV data are available.