How often clinical trial completion dates slip
The primary completion date is when a trial expects to collect its last primary-endpoint data, and for many sponsors it is the date that matters most. Across every version of every interventional trial on ClinicalTrials.gov, nearly half have pushed it back at least once.
Headline numbers
- 46.9% of the 461,435 interventional trials have moved their primary completion date later at least once.
- 12.3% have pushed it back three or more times.
- Among trials that slipped, the median net delay from the first registered date to the latest is 334 days.
- For 23.3% of trials, the latest primary completion date is a year or more after the first one registered.
- Slippage is more common in trials that ended badly: 63.2% of trials now marked terminated had slipped, against 52.6% of completed ones.
By sponsor type
Industry sponsors revise the date more often but by less. NIH-sponsored trials slip less often than industry trials, but when they do, the delays are much longer.
| Lead sponsor class | Trials | Ever pushed | Pushed 3+ times | Median net delay (days) | Net delay ≥ 1 year |
|---|---|---|---|---|---|
| All interventional trials | 461,435 | 46.9% | 12.3% | 334 | 23.3% |
| Industry | 113,523 | 52.5% | 15.0% | 183 | 18.6% |
| Academic & other | 320,449 | 45.4% | 11.3% | 365 | 24.6% |
| NIH | 7,881 | 46.3% | 19.1% | 614 | 41.4% |
| U.S. federal (non-NIH) | 4,310 | 54.7% | 19.3% | 366 | 32.8% |
| Networks, individuals, other | 15,272 | 37.0% | 7.5% | 366 | 21.3% |
Interventional trials, classified by study type and lead sponsor class in each trial's latest version. "Median net delay" is the latest primary completion date minus the first one registered, among trials pushed at least once. "Net delay ≥ 1 year" is among trials with a primary completion date on record.
How it was measured
A trial counts as pushed when any version's primary completion date is later than the previous version's. Estimated dates that later become actual dates count too, which is what you want when the question is how often plans come in late. The core of the calculation is a single window function over the version table:
WITH v AS (
SELECT nct_id, nct_version,
primary_completion_date AS pcd,
LAG(primary_completion_date) OVER (
PARTITION BY nct_id ORDER BY nct_version
) AS prev_pcd
FROM trial_versions
)
SELECT nct_id,
COUNT(*) FILTER (WHERE pcd > prev_pcd) AS n_pushes,
SUM(pcd - prev_pcd) FILTER (WHERE pcd > prev_pcd) AS days_pushed
FROM v
GROUP BY nct_id;You can't reproduce this from a single current-state snapshot of the registry. The latest record shows only where the date ended up, not the earlier dates it replaced or when each change was posted.
Why it matters
- Investors and analysts track readout timing; a slipped primary completion date usually means a later readout.
- Meta-researchers study registration quality and whether registered timelines are credible.
- Operations and feasibility teams benchmark realistic timelines by phase, sponsor type and therapeutic area.
- Modelers use a trial's slippage so far as an early signal of its eventual outcome.
More background: how ClinicalTrials.gov record versions work.