How the figures are worked out

Every date is the registry's own: when a study was registered and first posted, its start, its primary completion and its completion, and when results were posted. The registry marks most dates as actual or estimated. The timeline draws what has happened as a solid bar and what is still planned as a hatched one.

Primary completion is when the last participant was examined or treated to collect data for the study's main question, its primary outcome. Completion is the last participant's last visit for every outcome, secondary ones and side effects included: the end of the study. A study that stopped early gives the dates its data collection stopped. The comparisons below are made on primary completion, the date a study's main result depends on.

About 177,000 older records give dates without saying which they are. Those dates are listed and drawn in grey, but nothing that depends on a date having happened is said about them.

Dates are read to the month. A date given only as a year is not drawn on the bar.

Records are refreshed from ClinicalTrials.gov every night. Since Sep 25, 2026 each change has also been recorded, which is where a line like “primary completion moved from June 2027 to December 2027” comes from. Changes before that date are in the record history on ClinicalTrials.gov, which each timeline links to.

A trial is compared with other studies that share three things with it:

• Its condition. The first of its MeSH condition terms that has a condition page here. The terms are the registry's own mapping of what the sponsor entered, the same ones used in the “In context” section and the breadcrumb.
• Its phase. Phase 1 (early phase 1 included), phase 1/2, phase 2, phase 2/3, phase 3, phase 4, interventional studies with no phase, or observational studies (prospective, or not stated). Combined phases are kept apart: a phase 2/3 study is planned as both and runs like neither.
• Its kind of lead sponsor, industry or not, where the group is large enough.

The sponsor matters more than anything else we measured. Phase 2 studies run by industry reached primary completion in a median of 21 months; the rest took 47. Compared with every sponsor at once, an ordinary industry trial would look quick and an ordinary academic one slow.

So there are two comparisons, the closer one first. Studies with the same kind of sponsor are used when 20 or more of them have reached primary completion. Studies of every sponsor follow, and are used on their own when the narrower group is too small, provided 10 have reached it. With fewer, no comparison is shown.

Some condition pages group many conditions rather than name one: a whole branch of MeSH, such as Neoplasms or Infections, or a trait many conditions share, such as Rare Diseases or Chronic Disease. A study whose condition is one of these is compared with studies of many different conditions, whose durations vary widely, and its timeline says so.

The figure is the time from a study's actual start to its actual primary completion, in whole months. A group's figure is the month by which half its studies had reached primary completion, with the months by which a quarter and three quarters had.

The half is the median, not a chosen half. Every study in the group counts. On a study's timeline the dot is that median, and the band runs from the quarter mark to the three-quarter mark: the middle half of the studies, with the quickest and the slowest quarters outside it. No study is dropped as an outlier; the only ones left out are those listed under what is left out.

Studies still running are counted. Each counts until its most recent registry update, the last time it is known to have been still under way. This is a standard method for durations that have not all ended yet (Kaplan–Meier). Counting only the studies that have finished gives a figure that is too short, because the recent studies that have finished are the quick ones: for phase 3 breast cancer studies, finished studies alone gave 45 months, against 58 counted this way.

A group's figure is shown only once half its studies have reached primary completion. Before that, the median is not known yet, and nothing is said.

The figures are recalculated every night with the records they come from.

A study does not count towards any comparison when:

• It stopped early (terminated, withdrawn or suspended). Its primary completion date marks when it stopped, not when it finished.
• Its start is not marked actual: an estimate, or a date the registry does not type.
• A date is given only as a year, so months cannot be counted.
• Its dates are out of order, such as primary completion before the start.
• The span is over 30 years, which is treated as an error in the record.
• It is an expanded access record, which has no study dates to compare.
• It is a retrospective or cross-sectional observational study. Its dates bound the records it reads, not the running of a study. Counted in, they shortened the observational figure: a median of 17 and 14 months, against 34 for prospective studies.

And a study's own page shows no comparison when:

• it stopped early, because setting when it stopped against how long studies take to finish would read as a verdict on it;
• its start or primary completion is not marked actual or estimated;
• it has no condition with a page here, or its group is below the thresholds above.

Only dates and the differences between them. A study is described as above, below or close to a group's figure, close meaning within 10%. It is never called late, slow, overdue or delayed: a registry date can move for many reasons, and the record does not say which.

Where context cuts the other way it is said in the same place. A gap before results were posted is stated next to any delay of results the sponsor requested.

• Registry records are entered by people. Dates go un-updated, conditions are listed loosely, and some records are simply wrong. The rules above catch the errors that can be detected from the record; they cannot catch a plausible date that is not true.
• The condition can be the wrong one. A study is grouped by the first of its MeSH terms with a page here, and that term is sometimes broad (“Neoplasms”) or off the point of the study. Some records list dozens of conditions, sometimes alphabetically. We are reviewing the terms by hand to improve this.
• Date changes are known only since Sep 25, 2026. “Moved once” means once since then.

If a figure looks wrong for a study you know, tell us at contact@clinicaltrials.gg. Where the records come from, and what is done to them, is on Where the data comes from.