Lieve Van Hoovels

Background: In the American College of Rheumatology/European Alliance of Rheumatology Associations 2010 classification criteria for rheumatoid arthritis (RA), rheumatoid factor (RF) and anti-citrullinated protein/peptide antibodies (ACPA) are assigned the same weight, although ACPA have proven more specific than RF. We aimed to refine the serological weight factors for RA classification. Materials and methods: Diagnostic samples from 398 patients with RA and 448 diseased controls (DC) were evaluated with RF and ACPA assays from five different manufacturers. A grid search was conducted across all possible integer score pairs for RF and ACPA to identify the weighting combination that optimised the specificity of RA classification. This combination was applied in a refined scoring system (RSS) for three confirmatory cohorts comprising, respectively, 67, 157 and 132 patients with RA and 441, 153 and 193 DC. Findings: In the investigative cohort, the optimal combination of weighting scores was characterised by higher scores for ACPA, higher antibody levels and double RF/ACPA positivity. Applying the optimal combination of weighting scores decreased misclassification compared with the 2010 RA classification criteria, minimally affecting sensitivity. For seropositive patients, RSS significantly lowered misclassification from 38.3–55.8% to 15.0–26.9%. The improved specificity of RSS (compared with the 2010 criteria) was validated in the confirmatory cohorts. Considering only seropositive patients, the RSS significantly decreased RA misclassification (8.6–36.8% vs 20.0–78.9%), minimally affecting the sensitivity (86.7–97.3% vs 86.7–96.0%). Interpretation: Serological weight factors for RA classification can be improved by taking into account the antibody type, the antibody levels and single or combined positivity. The application of an RSS reduced the number of RA misclassifications.

Background: Urinalysis is a standard clinical test that includes the microscopic examination of urinary sediment to identify formed elements. Manual evaluation by laboratory technicians is time-intensive and subject to human error. Automated analysis using digital microscopy images presents a potential alternative. This study evaluates the integration of a deep learning approach to automatically classify urinary sediment images in the clinical laboratory, including independent prospective validation of its performance. Methods: An annotated data set comprising 13 classes of urinary sediment elements was created from a database of Sysmex UD-10 digital microscope images. An EfficientNet-based model was trained and tested across three experimental scenarios to evaluate the effects of data collection strategies on performance. Uncertainty calibration was examined. The model’s robustness and interpretability were examined using gradient-weighted class activation mapping (Grad-CAM) to visualize influential image regions and t-distributed stochastic neighbor embedding (t-SNE) to analyze learned feature embeddings. Lastly, a graphical user interface was developed for a prospective evaluation in the laboratory. Results: The model achieved approximately 97% overall accuracy on the test set. Experiments revealed sensitivity to data set variability, suggesting that performance may improve by integrating additional training examples. Confidence scores aligned with accuracy, and interpretability analyses showed that the model focused on relevant image regions and learned embeddings demonstrated clear class separation. In the prospective evaluation, top 1 and top 3 accuracies decreased to approximately 78% and 92%, respectively. Conclusions: Our results indicate that a lightweight deep learning model can achieve high performance in classifying urine particles. Analysis of discrepancies between retrospective and prospective evaluations provides important insights toward reliable clinical application.

Article Applying the likelihood ratio concept in external quality assessment for ANCA was published on November 27, 2024 in the journal Clinical Chemistry and Laboratory Medicine (CCLM) (volume 0, issue 0).

Classification criteria have been developed for rheumatoid arthritis (RA) and other rheumatic diseases in order to gather a homogeneous patient population for clinical studies and facilitate the timely implementation of therapeutic measures. Although classification criteria are not intended to be used for diagnosis, they are frequently used to support the diagnostic process in clinical practice, including clinical decision-making. The 2010 American College of Rheumatology (ACR)/European Alliance of Associations for Rheumatology (EULAR) classification criteria for RA are capable of identifying the majority of symptomatic patients with RA already in the earliest stages of the disease who are not yet showing radiographic changes. These patients will also profit from the early implementation of therapy with disease-modifying antirheumatic drugs (DMARDs). However, the risk of misclassification is higher as compared with the former 1987 ACR criteria, which were considerably less sensitive to the recognition of patients with early RA. Of note, the presence of rheumatoid factors (RFs) and anticitrullinated protein antibodies (ACPAs) has been attributed equal weight in the 2010 ACR/EULAR criteria and may contribute up to 50% of the score needed for being classified as RA. However, while ACPAs have been proven to be the most specific serological markers of RA, the specificity of RF is moderate, especially at lower titres. This may lead to the misclassification of RF-positive patients and, consequently, the unjustified implementation of DMARD therapy. Therefore, issues arise on how comprehensive the criteria should be and whether they should be updated and adapted to findings from the past two decades that might increase both their specificity and sensitivity.

Background: Circulating calprotectin (cCLP) is a promising diagnostic and prognostic biomarker in neutrophil-related inflammation, with most evidence in rheumatoid arthritis (RA). In the current study we aimed to investigate the impact of pre-analytical and analytical confounders on the inter-study heterogeneity of recent publications (01/2000–02/2023). Methods: A systematic literature search was performed in the electronic databases Medline, Embase and Cochrane databases (March 23st, 2023). After removal of duplicates, article eligibility was assessed using following inclusion criteria: (I) English (II) full-texts, describing (III) unique study results obtained from (IV) clinically defined RA patients and healthy control (HC) cohort, based on (V) clearly specified sample types (serum/plasma) and (VI) analytical method, and (VII) providing quantitative cCLP data. MEDCALC version 22.021 (MedCalc Software Ltd., Belgium) was applied for meta-analysis, including the Egger regression test to evaluate the publication bias. Results: Our literature search resulted in a total of 646 titles: 44 titles were retrieved for full-text screening and 18 studies including 21 cCLP datasets were eligible for inclusion in the meta-analysis. Totally, cCLP was measured in 2,627 RA and 897 HC and cCLP levels were significantly higher in RA patients [estimated standard mean difference (SMD) =1.032; 95% confidence interval (CI): 0.767–1.297]. Significant (P2 statistic =88.31%; 95% CI: 83.52–91.71%). cCLP analyses was tested using 14 different cCLP assays. Only in 9 out of the 18 included studies, adherence to pre-analytical conditions was specified. Subgroup analysis of the datasets (n=12) of these 9 studies, still showed significant heterogeneity (Cochran’s Q =96.107; I² statistic =88.55%; 95% CI: 81.91–92.76%). Excluding the Bühlmann (n=4) datasets, the datasets compliant to pre-analytical recommendations (n=9), using eight different cCLP assays on samples of 856 RA patients and 425 HC, showed absence of heterogeneity (Q =4.84; I² statistic =0.00; 95% CI: 0.00–42.28), whilst in the other subgroup (n=8) significant inter-study heterogeneity persisted (Q =66.03; I² statistic =89.40%; 95% CI: 81.48–93.93%). cCLP concentrations remained significantly higher in RA patients compared to HC (estimated SMD =0.698; 95% CI: 0.551–0.844). Conclusions: The inter-study variability of cCLP in RA is, besides analytical issues, mainly caused by pre-analytical confounders. Adherence to pre-analytical recommendations, significantly reduced inter-study heterogeneity and provided a reliable SMD, accurately revealing higher cCLP levels in RA versus HC.

Article Discriminating signal from noise: the biological variation of circulating calprotectin in serum and plasma was published on December 12, 2023 in the journal Clinical Chemistry and Laboratory Medicine (CCLM) (volume 0, issue 0).