How Predictive Drug Adherence Can Forecast Patient Dropout in Asthma Clinical Trials

Poor medication adherence has a significant impact on the viability of asthma clinical trials. Inhaled corticosteroids, for instance, have an adherence rate of ∼50%, according to a systematic review of 51 observational studies.1

A problem for asthma researchers is that it can be difficult to identify if a patient is truly refractory to the investigational drug or if they have simply not been taking their treatment. This complication was highlighted in a separate study that found only 27% of patients labeled as having severe uncontrolled asthma were truly refractory when adherence and technique were accounted for.2

Low rates of adherence are linked to greater risk of patient dropout, which results in lost data and increased recruitment costs. Today, however, innovations in computer vision, machine learning, and predictive analytics can improve medication adherence and forecast patient attrition. Predictive adherence technology from AiCure can identify at-risk patients during the execution of a study while there is still time to drive supportive intervention by clinical sites. This forward-looking, proactive capability can guide strategies to improve adherence and patient retention in asthma clinical research.

Poor adherence in asthma clinical trials is characterized by underuse of inhaled corticosteroids, often accompanied by over-reliance on short-acting β2-agonists for symptom relief.3 Due to the episodic nature of asthma, many patients feel that their daily life is not substantially impacted; consequently, many harbor doubts about the accuracy of their diagnosis or are in denial about the impact of the disease and, in turn, the need for long-term treatment. This hesitancy, revealed in a thematic analysis of 21 relevant articles, can negatively impact medication adherence in clinical trials and lead to early discontinuation.

AiCure technology allows for visual confirmation of drug adherence using computer vision on patients’ cellphones. Its machine learning capabilities provide insights into medication adherence patterns that could be used in asthma clinical trials to reduce the cost and delay caused by early discontinuation. This technology leverages predictive analytics to better identify patient dropout risks. Sponsors can use a placebo lead-in period of 10 to 14 days to understand patient dosing behavior and select patients for study inclusion who are likely to remain adherent to their medication over the course of the trial. Ultimately, the technology aims to improve clinical trial participation by making sure that asthma study participants have what they need to take their medication and stay engaged in the trial as it progresses.

Collaboration with Boehringer Ingelheim Predicts Patient Attrition

A recent collaboration with Boehringer Ingelheim (BI) offers insight into how medication adherence can be used to predict patient dropout in clinical trials.4 In two separate BI studies, predictive analytics models and dosing data from the AiCure platform accurately and reliably identified study participants who were at risk of dropping out. Armed with this knowledge, BI and the clinical sites had more nuanced information to develop effective strategies for reducing medication non-adherence and preventing dropouts. The performance of the models used in the two BI collaborations shows the power of this technology for medication adherence monitoring and predictive analysis of patient behavior.

Watch this video to learn more about the AiCure collaboration with Boehringer Ingelheim.

Most drivers of poor medication adherence in asthma were found to be deliberate in nature and may therefore be modifiable through strategies involving patients, physicians, and the healthcare system.2 Patient-centered interventions with proven effectiveness include patient engagement and education initiatives, such as routine counseling and training for patients to better understand their condition, training on inhaler technique, and the introduction of programs rewarding patients for optimal adherence and outcomes.2

AiCure provides an additional level of insight that can help sites proactively manage medication adherence in asthma clinical trials. The Site Services team helps clinical sites interpret medication adherence data and develop patient engagement strategies that can improve outcomes. For instance, AiCure data can indicate that an asthma trial participant is struggling to use the study inhaler. When non-adherence reaches the threshold level pre-specified in consultation with the sponsor, study teams receive a red alert and study coordinators are instructed to follow up with the patient and record the intervention in the site dashboard.

This extra level of service supports site staff to maximize the efficiency of clinical trial operations, making it a crucial resource for sponsors of asthma research.

To discuss medication adherence in asthma clinical trials and the potential to reduce patient dropout using predictive adherence technology, please contact us.

1 Brennan V, Mulvey C, Costello RW. The clinical impact of adherence to therapy in airways disease. Breathe (Sheff). 2021 Jun;17(2):210039. doi: 10.1183/20734735.0039-2021. PMID: 34295431; PMCID: PMC8291957.

2 Sulaiman I, Greene G, MacHale E, et al.. A randomized clinical trial of feedback on inhaler adherence and technique in patients with severe uncontrolled asthma. Eur Respir J 2018; 51: 1701126. doi: 10.1183/13993003.01126-2017

3 Amin S, Soliman M, McIvor A, Cave A, Cabrera C. Understanding Patient Perspectives on Medication Adherence in Asthma: A Targeted Review of Qualitative Studies. Patient Prefer Adherence. 2020 Mar 10;14:541-551. doi: 10.2147/PPA.S234651. PMID: 32210541; PMCID: PMC7071882.

4 Michael Sand, Zheng Zhu, Lei Guan, Shaolei Feng, Kristen Daniels, Dooti Roy. AI-based Adherence Prediction for Patients: Leveraging a Mobile Application to Improve Clinical Trials. Poster presented at the Annual American Society of Clinical Psychopharmacology (ASCP) Meeting, Scottsdale, Arizona, May 31–June 03, 2022.