How often do clinical trial completion dates actually slip?

What share of date changes push the date later?

93.5%. When a sponsor moves a primary completion date that is still a forecast, by 30 days or more, the move is a slip 93.5% of the time. Pull-ins are rare: of 102,588 recorded moves, 95,889 pushed the date later and 6,699 brought it earlier.

The practical reading: a forecast completion date on ClinicalTrials.gov is not a neutral estimate that might move either way. Empirically it is a floor. When it changes, it almost always changes in one direction, which is worth knowing every time you build a timeline on a competitor’s posted date.

How far do dates move when they slip?

The median slip is 335 days, almost exactly eleven months. A tenth of slips exceed 731 days, two full years. The median pull-in, for the 6.5% of moves that come earlier, is 154 days.

These are not small administrative corrections. The 30-day threshold already excludes date tidying, so everything counted here is a real schedule change, and the typical one costs the better part of a year.

Does phase 3 slip less than earlier phases?

Yes, and by enough to matter. Phase 3 dates are the most reliable in the registry: they still slip more often than not, but less often and less far than any other phase.

The likely explanation is unglamorous: phase 3 protocols are costed, powered and scheduled under more scrutiny than anything earlier, and their sponsors have the most riding on the date. Whatever the cause, the practical rule holds: an early-phase forecast date deserves roughly a year of scepticism, a phase 3 date somewhat less.

Do trials that slip once slip again?

Often. Of the 62,766 trials that moved a forecast date at least once, 24,730, or 39%, moved it again, and 9,526 moved it three or more times. A revised date is not a settled date. The first slip is the single best predictor that you are watching a schedule under pressure, which is exactly why a one-off check of the registry tells you so much less than watching it move.

Methodology

Source.ComplyRx’s change archive, built on 451,414 of the 455,360 interventional studies on ClinicalTrials.gov, combining reconstructed change history (from public registry archive data, at month precision) with live change detection. Window: 1 October 2022 to 1 August 2026.

What is counted. Changes to the primary completion date where both the old and new value were still forecasts (neither marked actual), and where the move was 30 days or larger in either direction. The symmetric threshold matters: the archive records small pull-ins more completely than small slips, so an unthresholded count would bias the direction split. Applying 30 days both ways removes that bias and excludes only 4 events out of 102,592.

What is not counted. Date changes where either side was marked actual (those are the subject of a separate article), and moves under 30 days.

Known limitation.The per-phase table requires joining each event to the trial’s current registry record for its phase, and that index under-represents terminated, withdrawn and unknown-status trials. The phase-level figures therefore describe trials that remained observable; the headline figures use no such join and carry no such bias. Archive-derived events are recorded at month precision, so day-level deltas older than July 2026 are approximate to within the month.