How trial stature is scored
Ruleset version 1
Every trial we hold carries three independent measures — methodological rigour, scale, and investigator standing. They are computed and shown separately and are never combined into a single “quality” number: a small, exquisitely designed single-centre trial and a sprawling open-label registry study are not comparable on one dimension, and collapsing them would hide exactly the tradeoff a reader needs to see. These are descriptions of a trial, not recommendations about it.
Axis 1 — Methodological rigour
Registry fields cannot support a full risk-of-bias appraisal — that needs the published paper — but the design fields alone are strongly discriminating. Read this as a registry-data proxy for design rigour, not a Cochrane-grade assessment. Where a field is inferred (control arm, endpoint type) we infer conservatively.
| Component | Weight | Scoring |
|---|---|---|
| Randomised | 20 | Yes 20 · No 0 |
| Blinding | 20 | Quadruple 20 · Triple 16 · Double 13 · Single 7 · Open 0 |
| Control arm | 15 | Active comparator 15 · Placebo 13 · Historical 5 · None 0 |
| Primary-outcome specificity | 10 | Objective + timeframe 10 · Named but vague 5 · Absent 0 |
| Hard vs surrogate endpoint | 10 | Mortality/MACE 10 · Clinical event 7 · Surrogate 3 |
| Multi-centre | 8 | ≥10 sites 8 · 2–9 sites 5 · single 0 |
| Data monitoring committee | 7 | Present 7 · not stated 0 |
| Prospective registration | 5 | Before start 5 · within 30 days 3 · later 0 |
| SAP / protocol posted | 5 | Both 5 · one 3 · neither 0 |
Bands: DEFINITIVE ≥ 80 · STRONG 60–79 · MODERATE 35–59 · EXPLORATORY < 35. Rigour is always shown relative to phase: a phase 1 study is not deficient for lacking a control arm — that is the correct design for its purpose — so the score is reported against its phase cohort (“strong for a phase 2 study”) and never as an absolute ranking across phases.
Axis 2 — Scale
The straightforward axis. Enrolment, sites and sponsor track record are log-scaled so the difference between 20 and 200 participants counts for more than the difference between 20,000 and 20,200.
| Component | Weight | Scoring |
|---|---|---|
| Enrolment | 40 | Log-scaled: 20 → 10 pts, 200 → 22, 2,000 → 33, 20,000 → 40 |
| Site count | 25 | Log-scaled similarly |
| Country count | 15 | 1 → 0 · 2–5 → 7 · 6–15 → 12 · 16+ → 15 |
| Planned duration | 10 | Longer follow-up scores higher |
| Sponsor scale | 10 | By the sponsor's own prior trial count in our corpus |
Bands by enrolment, with site and country count as modifiers: MEGA ≥ 10,000 · LARGE 1,000–9,999 · MEDIUM 200–999 · SMALL 30–199 · PILOT < 30. We use actual enrolment once a trial has completed and its target while it is recruiting, and we always label which — a trial that targeted 2,000 and enrolled 340 is an important fact, so enrolment attainment is shown alongside.
Axis 3 — Investigator standing
The most valuable axis and the one requiring the most care. It is weighted toward what we can verify from our own corpus — the strongest signal is the one we compute ourselves, this person has led 47 trials, 41 of which completed and posted results — rather than citation counts, which measure publishing, favour long English-language careers, and so are weighted low on purpose.
| Component | Weight | Scoring |
|---|---|---|
| Prior trials led (our corpus) | 30 | Directly verifiable, no external dependency |
| Completed and posted results | 25 | Follow-through matters more than volume |
| Cumulative enrolment led | 15 | Ability to actually run large studies |
| Years active leading trials | 10 | |
| Depth in this therapeutic area | 10 | Specialisation beats generic prominence |
| Bibliometrics (h-index, citations) | 10 | External, optional, weighted low on purpose — we hold none, so this scores 0 |
Bands: FIELD_LEADING ≥ 80 · SENIOR 55–79 · ESTABLISHED 25–54 · UNKNOWN < 25 or no data. UNKNOWN predominantly means we have no data, which is often true for excellent early-career investigators and for records outside ClinicalTrials.gov — it is not a judgement.
Names are matched conservatively. An ORCID identifier is an exact match; a distinctive full name links across trials; but a common name given only as a surname and an initial (“Wang L”, “Kim S”) produces no link at all rather than a guess, because merging two clinicians into one profile is a reputational error, not a data-quality one. We never rank clinicians, publish a leaderboard, or build a browsable directory of people; investigator context appears only on a trial, in service of understanding that trial, and it never influences personal match ranking.
Honest caveats
- These scores are descriptive of registry data only. An absent field lowers a score, and absence often reflects registration practice rather than study quality.
- Today we hold ClinicalTrials.gov data only. Non-English registries record less structured design detail; when we add them we will normalise within registry rather than systematically penalise them.
- A high-stature trial is not necessarily safer, more effective, or a better choice. Enrolling in a large definitive trial can mean a higher chance of receiving placebo; a small early-phase study may be the only route to a novel therapy.
- Prominent investigators are more likely to have industry relationships, because industry funds the largest trials and recruits the most experienced leaders. That is neither disqualifying nor endorsing — it is context.
- An investigator may request review of a mis-attributed trial or a bad name merge; we handle it promptly and log it.