Observed Compliance
Machines can now watch a dose being swallowed, yet the largest tuberculosis trials yet run, covering more than 25,000 patients in four countries, found smart pillboxes and coded labels did not improve treatment outcomes. The episode examines face-reading software, ingestible sensors and smart pillboxes, the ethics of autonomy, bias and data, and the strongest case against all of it. It closes with a measured verdict on the series' thesis.
Who's who
- Daniel LabovitzLead author of a 2017 trial of AiCure software in stroke patients
- Sara BrowneLead author of a 2019 San Diego trial of sensor-tagged TB pills
- Degu Jerene+1 more episodeLead author of the 2025 four-country smart pillbox trials
- Katherine Fielding+1 more episodeSenior author of the China, four-country and Ethiopian trials
- Adithya CattamanchiLead author of a 2021 Uganda trial of a pill-packet call-in system
- Joy BuolamwiniComputer scientist whose Gender Shades study measured face-software bias
- Timnit GebruComputer scientist and co-author of Gender Shades
- Craig KlugmanCo-author of a 2018 paper on the ethics of smart pills
Where it happens
- San Diego, California+2 more episodesSite of the 2019 wirelessly observed therapy trial
- Philippines+1 more episodeOne of four countries in the 2025 smart pillbox trials
- South AfricaOne of four countries in the 2025 smart pillbox trials
- Tanzania+1 more episodeOne of four countries in the 2025 smart pillbox trials
- Ukraine+1 more episodeOne of four countries in the 2025 smart pillbox trials
Timeline
- July 30, 2012Proteus Digital Health announces FDA clearance of its ingestible sensor
- November 13, 2017FDA approves Abilify MyCite, the first drug with a digital ingestion tracking system
- June 2020Proteus Digital Health files for Chapter Eleven bankruptcy
- March 2023CDC recommends video observation be considered equivalent to in-person observation
- March 11, 2025Four-country trials published: smart pillboxes and labels did not reduce poor outcomes
- September 2025Ethiopian trial in nearly 3,900 people also finds no reduction in poor outcomes
Still debated
Why better recording of doses has not yielded better outcomes is debated: the tools may not change behavior, standard care may already have been good, or trials may have measured the wrong outcomes. Whether watching means surveillance, observation or support is the crux, and readers may weigh the evidence differently.