TL;DR
This paper uses natural language processing (NLP) to systematically analyze 449 digital therapeutic (DTx) clinical trials registered on ClinicalTrials.gov. The study reveals key characteristics of DTx trials, including their geographical distribution, sponsor types, and common eligibility criteria, highlighting challenges in creating inclusive trials and offering insights for future research and development.
What Problem Does This Paper Address?
Digital therapeutics (DTx) – FDA-regulated software designed to prevent, manage, or treat diseases – are rapidly emerging as a novel class of healthcare interventions. While they hold immense potential for personalized and scalable care, understanding the landscape of clinical trials evaluating these technologies is crucial. Existing reviews often rely on manual data extraction, which is time-consuming and difficult to keep up-to-date with the fast-paced development of DTx. This paper addresses the need for a comprehensive, data-driven characterization of DTx clinical trials, leveraging advanced NLP techniques to analyze both structured and unstructured data from ClinicalTrials.gov. The goal is to provide a clearer picture of the current state of DTx clinical development, identify potential barriers to inclusivity, and inform future research directions.
Methods
The researchers employed a systematic review methodology combined with natural language processing (NLP) techniques. They identified 449 DTx clinical trials from ClinicalTrials.gov using a set of 27 search terms and specific filters to ensure they were FDA-regulated and not for basic science purposes.
Key NLP techniques used include:
- SciSpacy: This open-source library was used for text analysis pipelines and NLP models pretrained on biomedical datasets. It was instrumental in extracting and mapping medical conditions from free-text fields to standardized MeSH terms.
- BERTopic: This topic modeling technique was applied to the eligibility criteria (inclusion and exclusion) of the DTx trials. BERTopic clusters text embeddings to identify interpretable, semantically cohesive topics, allowing the researchers to understand the most common themes within these criteria.
The analysis also involved standard statistical methods for descriptive statistics, correlation analysis (Spearman's rank correlation), and group comparisons (Mann-Whitney U tests).
Datasets
The primary dataset used in this study was the ClinicalTrials.gov registry. The researchers extracted data for 449 DTx clinical trials that met their inclusion criteria. This dataset was supplemented with US Census Bureau population data for geographical analysis and the Neighborhood Atlas for area deprivation index (ADI) calculations.
Key Insights and Ideas
- Geographical Concentration: DTx clinical trials are heavily concentrated in a few populous US states, particularly California, New York, and Florida, suggesting potential disparities in access to trial participation.
- Sponsor Landscape: The majority of DTx trials are sponsored by academic medical centers ("Other") or industry, with a significant portion conducted by a single sponsor without collaborators.
- Eligibility Criteria as Barriers: A major finding is that DTx trials frequently exclude participants based on factors like age, comorbidities, pregnancy, language barriers, and crucially, "digital determinants of health." This includes requirements for smartphone ownership, data plans, and technological literacy, which could inadvertently widen health disparities.
- Focus on Specific Disease Areas: The most common medical specialties targeted by DTx trials are nervous system diseases and nutritional and metabolic diseases.
- Short Trial Durations: Interventional DTx trials tend to be short, with a median duration of about one year, raising questions about understanding long-term usage and efficacy.
- Interactive Dashboard: The authors have developed an interactive dashboard, available on GitHub, to allow researchers and clinicians to explore the DTx clinical trial data themselves, fostering further research and transparency.
Key Results
- 449 DTx trials were identified and included in the systematic review.
- Trials were predominantly interventional (88.2%) compared to observational (11.8%).
- The top states for DTx clinical trial locations were California (135 facilities), New York (58), and Florida (55).
- "Other" (academic medical centers) was the most common sponsor type (65%), followed by Industry (33%).
- The most frequent MeSH categories for DTx trials were nervous system diseases (19%) and nutritional and metabolic diseases (10%).
- Key inclusion criteria topics included clinical factors, ability to provide informed consent, age, smartphone access, and English fluency.
- Key exclusion criteria topics included medical history, pregnancy, allergies, and blood pressure.
- The study found a significant inverse correlation between the number of clinical trial locations and the Area Deprivation Index (ADI), indicating fewer trials in socioeconomically disadvantaged areas.
Limitations and Future Work
The authors acknowledge several limitations:
- Geographical Scope: The study focused on US-based trials registered on ClinicalTrials.gov, potentially missing DTx trials regulated or conducted outside the US.
- FDA Regulation: While efforts were made to focus on FDA-regulated devices, some DTx cleared via the 510(K) pathway might not have registered pre-approval trials, or post-marketing trials might have been missed.
- Data Completeness: The analysis relies on sponsor-provided data, which may not always be up-to-date, accurate, or standardized.
- Safety vs. Efficacy: The study could not differentiate between safety and efficacy studies based on the available data fields.
- MeSH Mapping: The conservative approach of mapping to broad MeSH terms might miss nuances, and allowing trials to map to multiple MeSH categories could provide a more comprehensive view.
Future work suggested includes:
- Investigating funding sources and costs of DTx trials.
- Further research into the geographical and demographic distributions of DTx users.
- Initiatives to assess the role of digital determinants of health in DTx effectiveness.
- Exploring other clinical vocabularies beyond MeSH for more nuanced metadata analysis.
- Expanding the analysis to include DTx trials regulated outside the US.
Summary updated