AssetMemo
AssetsMechanismsCompaniesCalendarChartsWatchlistAsk AIMethodology
  • Cardiovascular & Metabolic469
  • Dermatology230
  • Gastroenterology & Hepatology177
  • Hematology45
  • Immunology176
  • Infectious Disease145
  • Nephrology130
  • Neurology478
  • Oncology1,186
  • Ophthalmology75
  • Psychiatry141
  • Rare & Genetic Diseases141
  • Respiratory162
◧AssetMemo
⌕

Evidence-graded drug-asset intelligence
◧AssetMemo
AssetsMechanismsCompaniesReadout CalendarCompare ChartsWatchlist & AlertsAsk AI
Therapeutic areas
  • Cardiovascular & Metabolic469
  • Dermatology230
  • Gastroenterology & Hepatology177
  • Hematology45
  • Immunology176
  • Infectious Disease145
  • Nephrology130
  • Neurology478
  • Oncology1,186
  • Ophthalmology75
  • Psychiatry141
  • Rare & Genetic Diseases141
  • Respiratory162
ReferenceMethodology & RubricUs vs. AI ToolsUK IBD Access GuideAbout the founder

Educational data tool. Evidence graded; securities never are.

BrowseAll assetsAll mechanismsAll companiesReadout calendarReadouts by yearResults by diseaseDelayed readoutsOpen dataCompare share pricesWhat's new
ReferenceMethodology & sourcesUs vs. the AI toolsUK IBD Access GuideAbout the founderAsk the AIReport a problem / suggest a feature

Educational research only, not investment advice. Evidence quality is graded; securities never are. · Terms · Privacy · Logos by Logo.dev

Ask

Opening the console…

How the trial evidence is graded

Prediction services score a drug's odds of approval. This site grades the quality of the evidence itself, and publishes both the rubric and the per-trial reasoning.

Rubric by Eric Lloyd.

How this site derives its two most-scrutinized numbers: each indication's total attributable healthcare cost and evidence grade. The grade rubric is below, with sources and the exact criteria used; the market-size derivations have their own page.

Evidence gradingData sourcesMarket sizingPer-indication derivations ↗Glossary ↗

How evidence grades and signals are calculated

Every graded trial receives an evidence quality grade - strong, moderate, or weak - displayed across the platform as investor-facing signal tiers: Reliable signal, Partial signal, or Weak signal. Grading is based strictly on how the trial was designed to prove its result, not on how favorable that result looks. A single-arm study reporting a 90% response rate is exactly as uncontrolled as one reporting 10%. The response rate itself never moves a trial between grades.

Reliable signalgrade: strong

Confirmatory meaning: The completed evidence has the design features that historically survive confirmatory trials.

Design criteria: Randomized, blinded (double- or quadruple-blind), and placebo- or active-controlled, adequately powered (a power calculation for the primary endpoint is reported, or the trial enrolled at least its planned number of participants, and the trial's own documents do not say it was underpowered or stopped early), with a hard primary endpoint, and that primary comparison was statistically significant in the favorable direction (or prespecified equivalence/non-inferiority margin test succeeded). "Hard" means a clinical event, symptom or function outcome (clinical remission, endoscopic healing, exacerbations, survival) or an objective, disease-defining physiological measure that regulators accept as an approvable primary endpoint for that disease: FEV1 in asthma and COPD, blood pressure in hypertension, HbA1c in type 2 diabetes, LDL cholesterol or triglycerides for lipid-lowering therapies, % weight change in obesity, HIV-1 RNA suppression, albuminuria in kidney disease. Good design alone is never enough: a well-run trial whose primary result was null or mixed is capped at Partial signal (moderate).

Partial signalgrade: moderate

Confirmatory meaning: Positive or informative evidence, but with design limits that temper confidence.

Design criteria: Real but limited evidence, for one of several specific reasons: (a) a genuine open-label extension or roll-over of an earlier randomized trial's own participants; (b) real randomization and blinding intact, but the trial is not adequately powered: no power calculation and no planned enrolment is found in the sources checked, it enrolled fewer participants than planned with no power calculation, or its own documents say it was underpowered or stopped early; (c) a surrogate primary endpoint - a proxy biomarker below the disease definition (CRP, fecal calprotectin, target engagement, pharmacokinetics/pharmacodynamics), as opposed to the disease-defining measures listed under Reliable signal - or a primary endpoint that measures only safety or tolerability; (d) an otherwise well-designed randomized trial whose primary comparison was unclear, non-significant, mixed, or only partially met (good design does not upgrade a null or mixed result); or (e) a randomized, controlled trial that is open-label, when its primary endpoint is objective: a measure that knowing the treatment cannot change, such as overall survival, a laboratory value, or imaging read by a blinded central reviewer.

Weak signalgrade: weak

Confirmatory meaning: Results come from study designs that frequently fail to replicate under controlled conditions.

Design criteria: No randomization, no blinding, and no comparator arm (a standalone open-label single-arm study that is not a confirmed extension of an earlier randomized trial), or a trial severely underpowered for its own comparison, terminated early, or without a usable result. Uncontrolled single-arm trials default to Weak signal regardless of sample size or favorable reported rates. This stays true when the FDA accepted a single-arm study as the basis for an approval. The grade describes the design, and the drug page shows the note "single-arm, accepted for approval" beside it.

A drug's headline - the Evidence Outlook at the top of its profile - is stage-aware: the drug's development stage picks what kind of verdict applies, and the per-trial grades above only pick the tier within it:

  • Approved drugs: The regulator's approval verdict is the headline. The historical registrational trials are preserved below with their design-strength track record (e.g. track record: Reliable signal).
  • Phase 3 (Registrational): Shows how well completed mid-stage evidence supported entering a registrational trial (Reliable, Partial, or Weak signal), alongside expected Phase 3 readout timing.
  • Completed mid-stage readouts: Displays a signal tier - Reliable signal, Partial signal, or Weak signal - mapped directly from the single most decision-relevant graded trial (the highest-graded, most recent among ties). A missed primary comparison is stated plainly as a negative readout rather than tiered.
  • Ungraded evidence: When completed mid- or late-stage studies exist on the registry but have not yet been graded, the platform states this honestly rather than coercing an unknown readout to "Weak".
  • Early-stage programs: Phase 1 and first-in-human studies establish safety, tolerability, and dosing - they are not designed to demonstrate efficacy, so no signal tier is assigned.

Most grades on this site were first drafted by an AI model reading ClinicalTrials.gov's own posted design and results data directly, against the exact criteria above. Where the registry has no posted results, the result comes from a cited publication or company announcement, and the trial says so. Figures are never invented and never drawn from outcome favorability alone. Each draft is either reviewed and edited by hand before publishing, or, where a second independent model reaches the same grade and result on a re-read with no reviewable judgment call remaining, published without a manual step. Every draft keeps its full reasoning and the source data snapshot it was written from, for audit.

Where the data comes from

Trial records come from ClinicalTrials.gov's public API: every registered study for every listed drug, re-checked daily, including registered arms, endpoints and posted results. Company financials come from SEC EDGAR company facts for US filers, and from primary filings for foreign ones. Approval status is curated per indication from regulator and company sources. Market-size inputs are taken from the peer-reviewed epidemiology and cost literature cited below, never from market-research press releases. Where a number cannot be traced to a source, the site shows nothing rather than an estimate: every "Not sourced" on this page is deliberate.

Editorial standards

How often the data updates. Company press releases, SEC 8-K and 6-K filings and wire news are checked every hour. ClinicalTrials.gov records, posted trial results, closing share prices, insider filings and earnings dates are checked every day. US approvals and EDGAR financials are checked every week, and newly registered drugs every two weeks. Each change the feeds apply is dated, and the what's new page and the readout delay tracker list them.

Facts, not advice. The site grades the design and results of trials and shows sourced figures. It never says whether to buy or sell a security, never ranks stocks to buy and never gives a price target. Valuation figures such as cash and market value are shown as data, with no recommendation attached.

Sources. Every figure links to its source, and a figure with no source is left blank. AI-drafted grades and key results go through a review queue before they are published. If you find an error, use the feedback form. Each report is read, and a confirmed error is fixed in the data.

How share-price comparisons and moves are measured

Prices. Charts use end-of-day closing prices for US listings and US-listed ADRs, delayed by 24 hours. Prices are adjusted for stock splits. Each correction is stored with the filing or exchange notice that states the split, and applies only where the stored prices show the jump. Compare charts show each line as the percent change from its first close in the range.

Sector fund. Each company is compared with one of three sector funds: XBI (S&P Biotech), IBB (Biotechnology) or XLV (Health Care). The site uses the fund whose daily returns had the highest correlation with the company's daily returns over the past 250 trading sessions. A company with fewer than 120 shared sessions uses XBI, and the page says so. Large pharmaceutical companies usually match XLV, large biotechnology companies IBB, and smaller biotechnology companies XBI.

Own move. A company's own move on a day is its daily return minus its fund's daily return. Fund prices on this site start on 1 July 2024, so own moves start then.

Unusual move. A day is unusual when the own move is more than twice the standard deviation of the company's own moves over the previous 60 sessions. The tested day is not part of that 60-session window, and a day is not tested until 40 earlier sessions exist.

Events on the same day. For each unusual day, the site lists dated items from the previous session's date to the day itself, because a release after the close moves the next session. Items are grouped by how closely they are linked to the company: its own news, filings, earnings dates and financings; news from other publishers that names its drugs; clinical and regulatory news about other companies' drugs in the same indications (chosen by words in the headline, such as phase, topline, approval or FDA); and a day when the sector fund itself made an unusual move.

Price moves around an event. Figures are cumulative from the close of the session before the event session. Announcement dates have no time of day, so the event session and the next session are both shown. A weekend or holiday event starts on the next trading session.

A drug's share of a company's unusual movement. A drug's events are news items about the drug from a company that holds rights to it, and its dated approvals. Earnings, buyback and director-dealing releases count as company news. Each event covers its session and the next session. The share is the part of the company's squared own moves that fell on those sessions, so large moves count for much more than small ones. A session that belongs to several drugs is split equally between them. The share is shown when a drug has at least 8 event days. Below that, its events are listed with the moves on each.

Limits. Prices are daily, and a move is matched to a trading day. Stored news starts at different dates for different sources, and a day with no listed event can still have news the site does not store. Companies listed only outside the US have no price series here.

Causes. An event on the same day as a price move does not show that the event caused the move. These figures describe past prices. They are not a forecast or a recommendation to buy or sell any security.

Market sizing

Each indication's total attributable cost is its US patient population times the extra healthcare cost per patient that the disease causes, with drug spend per patient shown beside it. The current figures, the caveats to read before citing them, and the derivation and source for each indication are on their own page.

Market sizing: figures, caveats and derivations ↗

← back to asset universe