ATS 2021 will feature presentations of original research from accepted abstracts. Mini Symposia and Thematic Poster Sessions are abstract based sessions.
Please use the form below to browse scientific abstracts and case reports accepted for ATS 2021. Abstracts presented at the ATS 2021 will be published in the Online Abstract Issue of the American Journal of Respiratory and Critical Care Medicine, Volume 203, May 3, 2021.
Search Tips:
- Use the keyword search to search by keyword or author's name.
- Filter your search results by selecting the checkboxes that apply.
- Click on "Clear" to clear the form and start a new search. .
Search results will display below the form.
Real-Time Clinical Assessment and Temporal Predictions of CompEx Events
Session Title
TP14 - TP014 DIAGNOSTIC AND SCREENING INSIGHTS IN PULMONARY, CRITICAL CARE, AND SLEEP
Abstract
A1584 - Real-Time Clinical Assessment and Temporal Predictions of CompEx Events
Author Block: B. Toro1, J. Morrill2, K. Qirko3, A. Jauhiainen4, I. Psallidas5, S. Necander6, H. Forsman7, C. A. Da Silva8, S. Swaminathan9; 1E-Thera Inc, New York, NY, United States, 2Oxford University, Oxford, United Kingdom, 3Booking.com, Amsterdam, Netherlands, 4Early Biostats and Statistical Innovation, AstraZeneca, Gothenburg, Sweden, 5AstraZeneca, Cambridge, United Kingdom, 6Clinical Development, Research and Early Development, Respiratory and Immunology (R&I), AztraZeneca, Gothenburg, Sweden, 7Clinical Development, AztraZeneca, Gothenburg, Sweden, 8AstraZeneca, Gothenburg, Sweden, 9Vironix Health, AUSTIN, TX, United States.
Rationale
CompEx is a novel endpoint, developed to capture clinically relevant deteriorations of asthma that, when combined with severe exacerbations, creates a composite outcome. Previous retrospective investigations have shown that CompEx strongly mirrors results seen with severe exacerbation-validated outcomes, suggesting clinical trials of shorter duration and fewer patients could potentially emerge from the use of CompEx as an endpoint. Thus far, real-time evaluation of CompEx in clinical trials as well as temporal predictions of these events are yet to be conducted. We evaluated real-time CompEx data from a trial of asthmatics and determined 1) the quality of CompEx as a marker of material, asthma-related deterioration of health 2) the concordance of CompEx with other predictive algorithms for exacerbations, and 3) the scope for predicting CompEx with events in advance of their onset.
Methods
A machine learning backed triage application was retrofit with the requisite morning and evening diary assessments needed for evaluating CompEx each day. The study population included 27 English-speaking patients with asthma from the phase 2B GRANIT trial [NCT03622112; mean age 53 years, 23 (83%) female]. A python-based automated cloud program was developed to take application data in real time and calculate CompEx events. Simultaneously, patient symptom, demographic, and vital sign data were collected and computed to execute machine learning-based exacerbation and triage algorithms trained on physician opinion. Data output from these models was used to train a final temporal classifier to predict CompEx events early.
Results
The triage application platform showed an ability to automate and calculate CompEx events within seconds of patient completion of diary events. Though low patient enrollment should be noted, the study revealed 8 total CompEx events among 27 participants under 3 months of observation while only 2 severe exacerbations were documented. Analysis of the data showed that exacerbations predicted by independent flare-up detection algorithms that were trained on physician opinion showed a strong concordance with CompEx events.
Conclusions
Comparison of patient data input with triage application-captured symptom, demographic, and baseline health assessments with CompEx showed evidence that CompEx events are marked by asthma-related deteriorations and severe exacerbations. Finally, the temporal CompEx predictor showed significantly elevated probabilities preceding each of the 8 CompEx events. A threshold can be set which can be tuned to provide warnings of such events while maintaining a clinically relevant specificity. This approach could be further validated for robustness with a larger CompEx event dataset.
Author Block: B. Toro1, J. Morrill2, K. Qirko3, A. Jauhiainen4, I. Psallidas5, S. Necander6, H. Forsman7, C. A. Da Silva8, S. Swaminathan9; 1E-Thera Inc, New York, NY, United States, 2Oxford University, Oxford, United Kingdom, 3Booking.com, Amsterdam, Netherlands, 4Early Biostats and Statistical Innovation, AstraZeneca, Gothenburg, Sweden, 5AstraZeneca, Cambridge, United Kingdom, 6Clinical Development, Research and Early Development, Respiratory and Immunology (R&I), AztraZeneca, Gothenburg, Sweden, 7Clinical Development, AztraZeneca, Gothenburg, Sweden, 8AstraZeneca, Gothenburg, Sweden, 9Vironix Health, AUSTIN, TX, United States.
Rationale
CompEx is a novel endpoint, developed to capture clinically relevant deteriorations of asthma that, when combined with severe exacerbations, creates a composite outcome. Previous retrospective investigations have shown that CompEx strongly mirrors results seen with severe exacerbation-validated outcomes, suggesting clinical trials of shorter duration and fewer patients could potentially emerge from the use of CompEx as an endpoint. Thus far, real-time evaluation of CompEx in clinical trials as well as temporal predictions of these events are yet to be conducted. We evaluated real-time CompEx data from a trial of asthmatics and determined 1) the quality of CompEx as a marker of material, asthma-related deterioration of health 2) the concordance of CompEx with other predictive algorithms for exacerbations, and 3) the scope for predicting CompEx with events in advance of their onset.
Methods
A machine learning backed triage application was retrofit with the requisite morning and evening diary assessments needed for evaluating CompEx each day. The study population included 27 English-speaking patients with asthma from the phase 2B GRANIT trial [NCT03622112; mean age 53 years, 23 (83%) female]. A python-based automated cloud program was developed to take application data in real time and calculate CompEx events. Simultaneously, patient symptom, demographic, and vital sign data were collected and computed to execute machine learning-based exacerbation and triage algorithms trained on physician opinion. Data output from these models was used to train a final temporal classifier to predict CompEx events early.
Results
The triage application platform showed an ability to automate and calculate CompEx events within seconds of patient completion of diary events. Though low patient enrollment should be noted, the study revealed 8 total CompEx events among 27 participants under 3 months of observation while only 2 severe exacerbations were documented. Analysis of the data showed that exacerbations predicted by independent flare-up detection algorithms that were trained on physician opinion showed a strong concordance with CompEx events.
Conclusions
Comparison of patient data input with triage application-captured symptom, demographic, and baseline health assessments with CompEx showed evidence that CompEx events are marked by asthma-related deteriorations and severe exacerbations. Finally, the temporal CompEx predictor showed significantly elevated probabilities preceding each of the 8 CompEx events. A threshold can be set which can be tuned to provide warnings of such events while maintaining a clinically relevant specificity. This approach could be further validated for robustness with a larger CompEx event dataset.
