A trial using a language model to raise physical activity in type 2 diabetes was withdrawn without enrolling anyone — the research that did not happen, kept in the public registry (NCT06596330)
A randomized trial that would have built automatically generated coaching prompts from a language model into a smartphone app, aiming to raise physical activity and improve glycemic control in people with type 2 diabetes. Enrollment is 0 and the status is recorded as withdrawn.
Trial overview (primary data)
- StatusWithdrawn
- ConditionsType2diabetes
- InterventionsBEHAVIORAL: Validation of language model prompts in increasing short-term physical activity, BEHAVIORAL: Assessment of long-term changes to physical activity and glycemic control
- SponsorStanford University
- Period2025-07-01 〜 2029-07-01
Key points
- A randomized trial that would have used automatically generated coaching prompts from a language model to raise physical activity and improve glycemic control.
- The status is withdrawn and enrollment is 0 — it ended without a single participant.
- The reason for withdrawal is not part of this registry record.
- Of the 750 medical-AI trials this site holds as of 2026-08-28, only 13 — 4 withdrawn, 6 terminated, 3 suspended, or 1.7 percent — are recorded as stopped.
- The record states that even modest improvements such as 500 additional steps a day are associated with decreased diabetes and mortality.
- Of the 803 medical AI trials held as of 2026-09-01, suspended, terminated and withdrawn records total just 17.
1What it means for research that never happened to be on the record
A clinical trial registry does not exist only to count studies that produced results. Studies that were planned, registered, and then never carried out remain in the same ledger. This trial is recorded with zero enrollment and a status of withdrawn — it ended without a single participant. In research, attempts that do not work out tend to go unpublished and disappear.
A public registry is a mechanism that partly halts that disappearance, and this record is exactly what that mechanism produces.
2Stopped trials account for only 1.7 percent
Of the 750 medical-AI clinical trials this site holds as of 2026-08-28, 4 are withdrawn, 6 terminated and 3 suspended — 13 in all, or 1.7 percent of the total. Meanwhile 247 are recruiting and 215 have not begun recruiting. Read at face value the distribution suggests medical-AI trials almost never founder, but in practice ideas that collapse at the planning stage are never registered at all. What reaches the ledger is only what got as far as registration.
3What the record does not say
The important point is that this record does not say why it was withdrawn. Funding, staffing, a technical judgment — there is nothing left to decide between them. To fill that in would be to manufacture a reason that may never have existed.
What can be said with certainty about this trial is only this: a plan to raise physical activity and improve glycemic control using prompts generated by a language model was registered, and it was withdrawn without a single participant. The substance of the plan does survive in the record, and for anyone designing a trial along the same lines it remains a usable starting point.
4The states of trials on the ledger
A trial registry is not only for counting studies that produced results. Studies planned, registered and never carried out remain on the same ledger.
That is the breakdown across the 803 medical AI trials this site holds as of 2026-09-01. Records amounting to discontinuation total 17. Read plainly the distribution suggests medical AI trials almost never founder, while in reality concepts that lapse at the planning stage are never registered at all. Only what reached registration appears.
What can be said with certainty here is that a plan was registered and withdrawn without a single participant.
Why it matters
The value of a public registry is not only that successful research can be counted. Because research that never happened stays in the same ledger, it guards against judging the momentum of a field from its successes alone. In deciding whether to adopt medical AI, what was attempted and did not continue carries as much information as the list of achievements.
FAQ
Why was this trial withdrawn?
Is there value in covering a withdrawn trial?
Do medical-AI trials rarely stop?
Sources (primary)
Source: ClinicalTrials.gov (U.S. NIH/NLM, public domain). This site does not provide medical advice. Verify the latest and exact details with the official source. This site is not endorsed or certified by the NIH/NLM.
- ClinicalTrials.gov (study record, original)
- NCT ID: NCT06596330