Across 28 European Phase I cardiology trials, the median CTIS/EU submission-to-first-authorization review was 66 days (SD 41). Across 39 country-specific Part II records, the median was 185 days (SD 201), and only 15/39 (38.5%) finished within 90 days. Multicountry submissions, larger site footprints and dose-escalation designs carried the clearest delay signals.
The trial-level end-to-end CTIS interval was substantially shorter and less variable than the country-specific Part II interval. The end-to-end distribution spanned 11–133 days, while Part II spanned 2–687 days.
Country-specific Part II was 2.8× longer at the median and almost five times more variable by SD, making national Part II execution the less predictable CTIS planning component.
End-to-end authorization accumulated steadily: 7/28 (25.0%) by 30 days, 13/28 (46.4%) by 60 days and 18/28 (64.3%) by 90 days. Country Part II progressed more slowly, reaching 15/39 (38.5%) by 90 days and 28/39 (71.8%) by one year.
A 90-day planning assumption covered nearly two-thirds of end-to-end authorizations but fewer than two-fifths of country Part II decisions. For Part II, a one-year contingency still left 11/39 (28.2%) beyond the threshold.
Median country-specific Part II duration ranged from 11 days in Sweden to 441 days in Greece. Among countries represented by at least four records, Belgium had a median of 83 days, the Netherlands 88 days, Spain 113 days, Germany 265 days and France 422 days.
| Country | N | Median days | SD days | Range days |
|---|---|---|---|---|
| Sweden | 1 | 11 | — | 11–11 |
| Hungary | 1 | 35 | — | 35–35 |
| Denmark | 1 | 36 | — | 36–36 |
| Belgium | 5 | 83 | 143 | 12–355 |
| Netherlands | 6 | 88 | 279 | 2–687 |
| Spain | 4 | 113 | 164 | 7–366 |
| Latvia | 1 | 128 | — | 128–128 |
| Austria | 1 | 137 | — | 137–137 |
| Czechia | 2 | 193 | 233 | 28–358 |
| Poland | 2 | 238 | 213 | 87–388 |
| Germany | 7 | 265 | 241 | 10–628 |
| Italy | 3 | 350 | 198 | 202–595 |
| France | 4 | 422 | 84 | 264–440 |
| Greece | 1 | 441 | — | 441–441 |
Country choice materially changed the observed Part II planning range. The most stable multi-record benchmark was France at a high 422-day median with 84-day SD; Belgium and the Netherlands were faster at the median but had wider dispersion because of long-tail records.
Four descriptive signals aligned with shorter-than-median approval: Q1 submission, a pure Phase I stage, small-molecule modality and non-industry sponsorship. The cohort median used to define “faster” was 65.5 days.
The most actionable upstream signal was submission timing: 4/5 Q1 submissions (80.0%) finished below the cohort median and none reached 90 days. Modality and stage also mattered, while the sponsor-type result should be treated as a marker of cohort complexity rather than a causal sponsor effect.
Substantial delay was defined as ≥60 days and ≥90 days. Multicountry scope, more than three total sites, dose escalation and combination treatment all increased both the median review time and the share crossing the 90-day threshold.
Geographic and protocol complexity compounded. Every multicountry trial and every combination trial reached 90 days, while dose escalation doubled the 90-day delay rate from 31.8% to 66.7%. These are planning signals, not proof of causation.
Country site count was the clearest continuous Part II correlate: Spearman ρ=0.37, p=0.021. Multi-site country submissions had a median of 264 days versus 102 days for single-site country submissions.
Part II duration rose with national site burden: multi-site country records were 162 median days longer. Q2 submissions were fastest at 36 days, while Q1 and Q3 were slowest at 310 and 321 days, respectively.
The full cohort scan found little end-to-end relationship with planned sample size, endpoint count or eligibility-criteria count. By contrast, the number of countries showed a moderate positive association with longer review.
Protocol scale was not equivalent to regulatory delay: planned sample size had ρ=−0.04, endpoint count ρ=0.06 and eligibility-criteria count ρ=−0.06. Geographic breadth was more informative at ρ=0.43.