Combining human and artificial intelligence

James Fleck, MD, PhD: Anticancerweb 08 (09), 2023

Double-blind randomized clinical trials (RCTs) are considered the most reliable method to assess the effectiveness of therapeutic interventions. They are associated with the highest level of evidence and are widely used for the validation of new drugs or treatment advances. However, they require large numbers of volunteer participants and close monitoring to ensure the results are reliable and meaningful. Because advances in patient care are based on small increments achieved in study-defined outcomes, RCTs are sometimes limited in statistical power due to unexpected limitations in the number of patients enrolled. Meta-analyze consists of a statistical resource used to bring together several RCTs with similar objectives in order to increase the n (number of suitable patients), hopefully seeking more consistent results.

As a research methodology, randomized clinical trials (RCTs) were developed in the middle of the last century. The first RCT was conceived in 1946 and chaired by Sir Geoffrey Marshall, assisted by two statisticians called Sir Austin Bradford Hill and Philip Hart. Currently, most RCTs require at least an average follow-up of 5 years to produce mature results in palliative care and up to 10 years for curative interventions. The conclusion is always probabilistic and driven by statisticians. Meta-analyses also have important limitations due to the high degree of heterogeneity often observed among included RCTs.

This old-fashion methodology no longer supports high-speed dynamics in disease's phenotypic expression. In addition, clinical trials only select patients in good clinical conditions, making most real patients ineligible, which makes recruitment difficult and limits the validity of the results. Due to their high cost and slowness, they can be considered as "bottlenecks" for clinical research. 

Artificial intelligence (AI) might represent an excellent alternative, addressing a new methodology based on big data analytics. AI can easy identify precise patterns for the most prevalent diseases and their behavior towards innovations. The approach replaces RCTs representative samples, composed of a few hundred ideal patients, with a set of available real data.  

The downside is the quality of medical records, which suffers from a lack of standardization and poor interoperability. These disadvantages can only be addressed through human intelligence, creating a better global standard for electronic personal health records. Global e-PHR is actively working in this direction. A new approach is proposed, supported by the well-known problem-oriented medical record. The global e-PHR also puts patients in a central position, acting as protagonists to better assess the quality of their medical data. The file would be responsive and available worldwide in real time. Based on patient informed consent, anonymized bigdata would identify disease patterns and outcomes of medical interventions.

Should we remain silent and not explore new methodological approaches to overcome the limitations of medical science? Reflecting on this question, I come across the recently unveiled artwork by Jaume Plensa, called Water's Soul, erected on the banks of the Hudson River in Newport, Jersey City. It is a call to empathetic self-reflection, guiding human behavior. Plensa translates his artistic sensitivity by stating: “One drop of water is quite alone, like a single person, but many drops together can create a tidal wave, and form immense rivers and oceans; When individuals come together to exchange ideas and create community, we can build something incredibly powerful.”

References:

1. MRC Streptomycin in Tuberculosis Trials Committee. Streptomycin treatment of pulmonary tuberculosis. BMJ 2:769-83, 1948

2. Global e-PHR (Personal Health Record): www.ephr.org

3. Gray Gallery: Jaume Plensa Water’s Soul, Newport, Jersey City, Oct 21, 2021