Insights methodology

Methodology version: v1.0 · Introduced: July 2026 · Next scheduled review: July 2027

Snapshot calculation methodology (v1.0)

Each night, Voxsanity computes a snapshot for every tracked condition and appends it to an append-only history table (condition_monthly_snapshots). A snapshot records, as at the run date (snapshot_date, in UTC):

• Active trials — trials tagged to the condition whose status is not withdrawn or terminated.
• Recruiting trials — trials with status recruiting.
• Australian-site trials — recruiting trials with at least one site located in Australia.
• Phase 1 / 2 / 3 counts — recruiting trials by phase (see definitions below).
• Completed trials, NIH funding total, Australian site penetration rate, and the top five sponsors by recruiting-trial count.

Three fields named here previously are no longer part of what a snapshot reports, and we say so rather than quietly dropping them from the list:

• Average FDA-to-PBS lag — removed from the snapshot on 25 July 2026 and no longer written. It averaged a drug-to-condition join we found to be unreliable. The real FDA-to-PBS figure is computed separately from the PBS Schedule (see below).
• Patent count — the field is written but is always empty, because patent data is deliberately not collected at all (see Data sources below). No patent figure appears anywhere on this site.
• New approvals this month — still written to the table but read by nothing. It rested on the same unreliable drug-to-condition join. Condition What changed pages now report newly subsidised on the PBS instead, which is a different measure drawn from the PBS Schedule.

Snapshots are append-only and never overwritten. Uniqueness is enforced on (condition_id, snapshot_date), so a condition has at most one snapshot per day. Snapshots run nightly (previously monthly; daily snapshots were introduced in July 2026).

Phase count definitions

Phase values come directly from the source registry (ClinicalTrials.gov). We map the raw registry enum to display labels without reinterpretation:

• Phase 1 — PHASE1 (and EARLY_PHASE1, shown as Early Phase 1). First-in-human safety and dosing.
• Phase 2 — PHASE2. Preliminary efficacy and side effects.
• Phase 3 — PHASE3. Large-scale efficacy; the stage closest to a regulatory submission. Our "most active research pipelines" ranking uses the Phase 3 count.
• Phase 4 — PHASE4. Post-approval studies.
• Not Applicable — NA. Studies for which a phase does not apply (e.g. many device or observational studies).

Unless stated otherwise, phase counts on the Insights Hub count recruiting trials only. The Phase 3 count (condition_monthly_snapshots.phase3_count) counts trials for the condition that are both in Phase 3 and currently recruiting, and is the figure used for the "most active research pipelines" ranking. It is recomputed nightly from the full trials table.

Australian site penetration formula

Australian site penetration measures how much of a condition's recruiting activity is accessible locally:

AU site penetration = recruiting trials with ≥1 Australian site ÷ total recruiting trials

The rate is null (not zero) when a condition has no recruiting trials, so "no data" is never displayed as 0%. A condition with high global Phase 3 activity but zero Australian sites appears in the research gap section.

Data sources used

• ClinicalTrials.gov — trial registrations, status, phase, sites, and eligibility criteria (public domain).
• openFDA — FDA drug approvals and labelling (CC0).
• NIH Reporter — US research funding totals and trends (free for commercial use).
• PatentsView / USPTO — patent filing counts as a research-activity signal (public domain). No patent figure is published anywhere on Voxsanity today, and the source is not fetched at all: patent data is deliberately not pursued unless a real patient-facing reason for it comes up. That is a product decision rather than a technical limitation, so nothing on this site currently rests on it.
• OpenAlex — academic paper volume (CC0). No paper-volume figure is published anywhere on Voxsanity today, and the source is not fetched at all. Paper counts were requested nightly up to 29 July 2026 but were never stored, so the field behind them was empty for every condition; the request was removed rather than left running, and it would only return alongside a real stored figure and a page that shows it. As with the patent case above, nothing on this site currently rests on this source.
• PBS — Australian subsidy listing status (© Commonwealth of Australia, CC BY 3.0 AU: attribution required, and the data must not be modified). The PBS attribution appears on every page that shows PBS data.

Australian TGA/ARTG registration data is not one of the sources above. TGA-hosted material is retransmission-restricted and our permission request is unanswered, so no TGA-sourced value is published anywhere on Voxsanity. The full licence position for every source is on the data sources page.

WHO ICTRP is deliberately excluded. All data is publicly available government registry data, republished in plain English. Voxsanity does not create or infer clinical conclusions.

Research equity metric

Research equity expresses clinical trial activity relative to how many people are affected by a condition:

Research equity = total recruiting trials ÷ estimated Australian patient population

A higher score means more research activity relative to the number of people affected. Conditions with few trials per patient are "research deserts". Source: ClinicalTrials.gov recruiting counts. Australian prevalence data is not yet loaded, so this metric has not launched. Updated monthly. Methodology version v1.0 (set July 2026); next review July 2027. This metric publishes once Australian prevalence data is loaded — until then the research equity page shows a placeholder.

FDA-to-TGA lag

Days between a drug's FDA approval date and its TGA approval date, computed only where both dates are available in the openFDA data. openFDA (US FDA) does not currently supply Australian TGA dates, so this metric is schema-ready but returns no values yet. Methodology version v1.0.

FDA-to-PBS lag

Days between a medicine's US FDA approval date and its first listing on the Pharmaceutical Benefits Scheme. It is matched against the current PBS Schedule by active-ingredient set, not by drug name and not from the fda_to_pbs_lag_days column, which is empty. Only an exact ingredient-set match contributes a value, because only then does that Schedule item's first-listed date belong to that medicine; a medicine whose ingredients are all subsidised but not together as that product contributes nothing rather than a misleading date.

The figure is computed but published only when the matched sample is robust — at least 40 medicines carrying a positive lag, and at least 60% of matched medicines carrying one. That threshold is not met today, so the statistics page withholds the figure rather than headlining a thin one, and no average FDA-to-PBS gap is stated anywhere on the site. Methodology version v1.0.

Top sponsors

Pharmaceutical companies and research institutions with the most recruiting trials per condition, ranked by recruiting-trial count and updated nightly. Where a single sponsor holds a majority of a condition's recruiting trials (above 50%), we flag it as high sponsor concentration.

What we do not compute

We do not predict drug-approval likelihood for individual trials. The phase success rates shown on trial pages (Phase 1 ~10%, Phase 2 ~30%, Phase 3 ~65%) are industry-wide historical averages from published research — not predictions for any specific trial.

When metrics were introduced

• Monthly condition snapshots — v1.0, first half of 2026.
• Daily snapshots — July 2026 (replaced the monthly cadence; snapshot_date replaced snapshot_month).
• Insight callouts & Insights Hub — July 2026.
• FDA-to-PBS access lag — computed from the PBS Schedule since July 2026; published only once the matched sample passes the robustness threshold described above, which it does not yet.

Annual review

This methodology is reviewed at least annually. The next scheduled review is July 2027. Material changes to any calculation are recorded in DATA_METHODOLOGY_CHANGELOG.md and the methodology version is incremented — we never change a calculation silently.

This page documents how Voxsanity computes the figures shown across the site. It does not constitute medical advice. Figures are indicators derived from public data, not predictions of clinical outcomes or regulatory decisions.