High blood eosinophil counts can predict an enhanced response to inhaled corticosteroids (ICS) in patients with chronic obstructive pulmonary disease (COPD), but additional insight in treatment effect heterogeneity is needed to optimize clinical disease management. We investigated if causal machine learning models can detect heterogeneity in the effect of two ICS-containing therapies on both time to first exacerbation and exacerbation rate, and if this approach can identify additional predictors of ICS response. Baseline characteristics from patients in the fluticasone furoate/umeclidinium/vilanterol (FF/U/V) and U/V arms of the InforMing the PAthway of COPD Treatment (IMPACT) trial (ClinicalTrials.gov identifier NCT02164513) were used to train a causal survival forest to estimate the effect of fluticasone furoate (an ICS) on time to first exacerbation for each patient. Similarly, a causal forest was trained for exacerbation rate. Results were averaged over 100 train and validation sets in a Monte-Carlo cross-validation approach. The analysis was repeated for a comparison of the FF/V and U/V arms. This analysis included 4048 FF/U/V, 4034 FF/V and 2025 U/V patients. Significant p-values for AUTOC (area under the targeting operator characteristic) curve indicated that all models ranked patients well according to individualized treatment and that significant heterogeneity was detected in the effect of ICS on both exacerbation outcomes. Respiratory symptoms, reversibility of airflow limitation and lung function, in addition to eosinophils, were identified as important predictors by all models. The contribution of additional predictors in determining the effect of ICS was most apparent for patients at the lower end of the eosinophil spectrum. This hypothesis-generating secondary analysis of IMPACT detected significant heterogeneity in the effect of ICS and indicated that, in addition to eosinophils, other clinical patient characteristics also contribute to the prediction of ICS response on both time to first exacerbation and exacerbation rate. The causal machine learning approach in this analysis can identify treatment effect predictors and patterns of response, to enhance insight in COPD disease phenotypes and heterogeneous treatment effects.

OBJECTIVE:Long-term azithromycin treatment effectively prevents acute exacerbations of chronic obstructive pulmonary disease (COPD). However, patients would benefit from better identification of responders and non-responders to minimise unnecessary exposure. We aimed to assess treatment effect heterogeneity and estimate individual treatment effects (ITEs) to distinguish patients most likely to benefit from prophylactic treatment. METHODS:We used data from 1025 patients of the MACRO trial to assess the ITE of azithromycin on annual exacerbation rate. A Causal Forest was used as a causal machine learning model. We independently validated our findings using data from 83 patients of the COLUMBUS trial. RESULTS:The tertile of patients with the best predicted ITE within MACRO and within the COLUMBUS independent validation cohort showed significant and substantially greater reductions in annual exacerbation rates (in MACRO -0.50, rate ratio 0.70, p=0.01, in COLUMBUS: -2.28, rate ratio 0.43, p<0.001) compared with the average treatment effect across the entire cohort (MACRO -0.35, rate ratio 0.83, p=0.01 and COLUMBUS -1.28, rate ratio 0.58, p=0.001). Conversely, no significant treatment effect was observed in the remaining two-thirds of patients. Primary determinants of ITE included respiratory symptoms, white blood cell count, haemoglobin, C-reactive protein and forced vital capacity. Smoking status did not emerge as a significant predictor. CONCLUSION:Based on five easily obtainable parameters to predict ITE, we identified treatment effect heterogeneity in COPD subjects treated with azithromycin maintenance therapy and found a small subgroup of responders driving the average reduction in exacerbations reported in previous trials.

Monoclonal antibodies recognizing nonprotein antigens remain largely underrepresented in our understanding of the molecular repertoire of innate and adaptive immunity. One such antibody is Mannitou, a murine IgM that recognizes paucimannosidic glycans. In this work, we report the production and purification of the recombinant antigen-binding fragment (Fab) of Mannitou IgM (Mannitou Fab) and employ a combination of biochemical and biophysical approaches to obtain its initial structural characterization. To this end, recombinant Mannitou Fab comprising the light chain (VL-CL) and heavy chain (VH-Cμ1) was produced in HEK293 FreeStyle cells and purified by cobalt-affinity chromatography followed by size-exclusion chromatography (SEC), which revealed two distinct oligomeric states consistent with a predominant monomeric form and a minor dimeric form. We employed SEC inline with multi-angle light scattering (SEC-MALS) and SEC coupled to small-angle X-ray scattering (SEC-SAXS) to establish that Mannitou Fab indeed adopts monomeric and dimeric forms in solution. Interestingly, Mannitou Fab is N-glycosylated at Asn164 of the heavy chain via HexNAc(5)Hex(6)Fuc(1–3) as revealed by mass spectrometry. We leveraged this information in conjunction with predicted structures of Mannitou Fab to facilitate the interpretation and modelling of SAXS data, leading to a plausible model for glycosylated Mannitou Fab. Analysis of the two chromatographically isolatable forms of Mannitou Fab using synchrotron-radiation circular dichroism revealed that the heat-denaturated Mannitou Fab monomer shares similar secondary-structural elements with the Mannitou Fab dimer, indicating that the latter may be misfolded. Collectively, the findings of this study will set the stage for future structural studies of Mannitou Fab and contribute to our understanding of possible side products due to misfolding during the production of recombinant Fabs, highlighting the importance of glycosylation in obtaining stable and monodisperse monomeric forms of recombinant Fabs.

Spontaneous protein crystallization is a rare event, yet protein crystals are frequently found in eosinophil-rich inflammation. In humans, Charcot-Leyden crystals (CLCs) are made from galectin-10 (Gal10) protein, an abundant protein in eosinophils. Although mice do not encode Gal10 in their genome, they do form pseudo-CLCs, made from the chitinase-like proteins Ym1 and/or Ym2, encoded by Chil3 and Chil4 and made by myeloid and epithelial cells respectively. Here, we investigated the biological effects of pseudo-CLCs since their function is currently unknown. We produced recombinant Ym1 crystals which were shown to have identical crystal packing and structure by X-ray crystallography as in vivo native crystals derived from murine lung. When administered to the airways of mice, crystalline but not soluble Ym1 stimulated innate and adaptive immunity and acted as a type 2 immune adjuvant for eosinophilic inflammation via triggering of dendritic cells (DCs). Murine Ym1 protein crystals found at sites of eosinophilic inflammation reinforce type 2 immunity and could serve as a surrogate model for studying the biology of human CLCs.

Hormone absence or inactivity is common in congenital disease, but hormone antagonism remains controversial. Here, we characterize two novel homozygous leptin variants that yielded antagonistic proteins in two unrelated children with intense hyperphagia, severe obesity, and high circulating levels of leptin. Both variants bind to the leptin receptor but trigger marginal, if any, signaling. In the presence of nonvariant leptin, the variants act as competitive antagonists. Thus, treatment with recombinant leptin was initiated at high doses, which were gradually lowered. Both patients eventually attained near-normal weight. Antidrug antibodies developed in the patients, although they had no apparent effect on efficacy. No severe adverse events were observed. (Funded by the German Research Foundation and others.).

Background and aimsPulmonary hypertension due to left heart disease (PH-LHD) is the most frequent form of PH. As differential diagnosis with pulmonary arterial hypertension (PAH) has therapeutic implications, it is important to accurately and noninvasively differentiate PH-LHD from PAH before referral to PH centres. The aim was to develop and validate a machine learning (ML) model to improve prediction of PH-LHD in a population of PAH and PH-LHD patients.MethodsNoninvasive PH-LHD predictors from 172 PAH and 172 PH-LHD patients from the PH centre database at the University Hospitals of Leuven (Leuven, Belgium) were used to develop an ML model. The Jacobs score was used as performance benchmark. The dataset was split into a training and test set (70:30) and the best model was selected after 10-fold cross-validation on the training dataset (n=240). The final model was externally validated using 165 patients (91 PAH, 74 PH-LHD) from Erasme Hospital (Brussels, Belgium).ResultsIn the internal test dataset (n=104), a random forest-based model correctly diagnosed 70% of PH-LHD patients (sensitivity: n=35/50), with 100% positive predicted value, 78% negative predicted value and 100% specificity. The model outperformed the Jacobs score, which identified 18% (n=9/50) of the patients with PH-LHD without false positives. In external validation, the model had 64% sensitivity at 100% specificity, while the Jacobs score had a sensitivity of 3% for no false positives.ConclusionsML significantly improves the sensitivity of PH-LHD prediction at 100% specificity. Such a model may substantially reduce the number of patients referred for invasive diagnostics without missing PAH diagnoses.

SESSION TITLE: Late-Breaking Insights in Pulmonary Medicine Posters SESSION TYPE: Original Investigation Posters PRESENTED ON: 10/10/2023 12:00 pm - 12:45 pm PURPOSE: Long-term azithromycin treatment is effective to prevent acute exacerbations of COPD (AECOPD). However, selecting patients who are likely to benefit from the therapy is crucial to avoid unnecessary exposure. We aimed to evaluate treatment effect heterogeneity and estimate individual treatment effects (ITE) of azithromycin maintenance therapy on annual exacerbation rate. METHODS: We used data from 1025 subjects of the MACRO trial (NCT00325897) to develop an estimation model and assess the ITE of azithromycin on annual exacerbation rate. The Causal Forest model was used to estimate ITE. We independently validated the model using data from 83 subjects of the COLUMBUS trial (NCT00985244). RESULTS: Respiratory symptoms, white blood cell count, haemoglobin, C-reactive protein, and forced vital capacity emerged as the primary determinants of the ITE in MACRO. When ranking subjects based on their estimated ITE in MACRO, the tertile of subjects with the strongest ITEs had a significantly lower rate ratio 0.45 (p=0.003) compared to the average treatment effect across entire cohort (rate ratio 0.73, p=0.001). The model was confirmed in COLUMBUS with a rate ratio of 0.43 (p=0.003) in the tertile with highest ITE. No significant treatment effects were observed in the remaining two tertiles. CONCLUSIONS: We identified heterogeneity in treatment effects of long-term azithromycin therapy for preventing AECOPD and found that the average treatment effect reported in previous trials is driven by a small subgroup of responders. Furthermore, we successfully predicted ITE based on a limited number of parameters readily available in routine clinical practice and demonstrated the generalizability of the model. CLINICAL IMPLICATIONS: Azithromycin to prevent acute exacerbations of COPD should be restricted to individuals with highest predicted treatment benefits. Machine learning models help to identify these subjects. DISCLOSURES: No disclosure on file for Richard Albert No relevant relationships by John Connett Consultant relationship with ArtiQ Please note: 01/07/2022 till now Added 06/01/2023 by Maarten De Vos No relevant relationships by Remco Djamin No relevant relationships by Iwein Gyselinck No relevant relationships by Helene Huts No relevant relationships by Wim Janssens No relevant relationships by Sarah Lindberg No relevant relationships by Menno van der Eerden No relevant relationships by Kenneth Verstraete

Background Parameters from maximal expiratory flow-volume curves (MEFVC) have been linked to CT-based parameters of COPD. However, the association between MEFVC shape and phenotypes like emphysema, small airways disease (SAD) and bronchial wall thickening (BWT) has not been investigated. Research question We analyzed if the shape of MEFVC can be linked to CT-determined emphysema, SAD and BWT in a large cohort of COPDGene participants. Study design and methods In the COPDGene cohort, we used principal component analysis (PCA) to extract patterns from MEFVC shape and performed multiple linear regression to assess the association of these patterns with CT parameters over the COPD spectrum, in mild and moderate-severe COPD. Results Over the entire spectrum, in mild and moderate-severe COPD, principal components of MEFVC were important predictors for the continuous CT parameters. Their contribution to the prediction of emphysema diminished when classical pulmonary function test parameters were added. For SAD, the components remained very strong predictors. The adjusted R 2 was higher in moderate-severe COPD, while in mild COPD, the adjusted R 2 for all CT outcomes was low; 0.28 for emphysema, 0.21 for SAD and 0.19 for BWT. Interpretation The shape of the maximal expiratory flow-volume curve as analyzed with PCA is not an appropriate screening tool for early disease phenotypes identified by CT scan. However, it contributes to assessing emphysema and SAD in moderate-severe COPD.

Performance of AI in fracture detection on radiography and its effect on the performance of physicians: a systematic review This systematic review has a twofold objective regarding the evaluation of the use of artificial intelligence (AI) for fracture detection on radiography. The first is to examine the performance of the current AI algorithms. The second concerns an evaluation of the effect of AI support on the performance of physicians in fracture detection. A systematic literature search was performed in 4 databases: PubMed, Embase, Web of Science and CENTRAL. Fourteen studies met the inclusion and exclusion criteria. The studies were divided into 2 categories: a first group in which a comparison was made between the performance of AI and the performance of physicians and a second group comparing the performance of physicians with and physicians without AI aid. Seven studies reported a comparable or superior fracture detection performance for AI compared to physicians, including radiologists. One study established a comparable performance on the internal test. On the external test, a lower AI performance was found compared to physicians. The second group of 6 studies reported a positive effect on the fracture detection performance of physicians when aided by AI. The current AI algorithms have a fracture detection performance comparable with physicians. At present, AI can be used as an aid in fracture detection. The potential impact of AI as an aid is greater with regard to less experienced doctors. The biggest hurdle of the current AI algorithms is the lack of large quantities of high-quality training data. Prospective studies, as well as further development and training of detection algorithms are needed in the future, in addition to larger datasets.