Data coverage and limitations

Voxsanity is built on public registries. Those registries are incomplete, and so is this site. This page says where the gaps are, so you are not left assuming a completeness that does not exist.

If you are making a decision about your own care, treat this site as a starting point for a conversation with your doctor, never as the full picture.

The scope: Australia, and trials you could still join

This site covers clinical trials that are open and have at least one site in Australia. A trial that is closed, or that runs only overseas, does not get a page here.

That is a deliberate narrowing, and it is worth being plain about what it costs. A trial with no Australian site can still matter to you. Your specialist may know of one, and some trials accept participants from overseas. Not finding a trial here does not mean none exists.

How a trial is counted

Two conditions have to hold before a trial appears on this site, and both are read from the registry record rather than decided by us.

  • The registry lists it as open. That means one of three statuses: recruiting, not yet recruiting, or enrolling by invitation. Those are the registry's own words for a study that has not closed its doors. Anything else, including completed, terminated, withdrawn, suspended and active but not recruiting, is not counted as open.
  • The registry lists at least one site in Australia. This is a count of trials, not of sites. One trial with six Australian hospitals counts once.

Both facts come from the trial record and change when the sponsor updates it, which can be long after the reality changed. A trial listed as recruiting may have stopped recruiting weeks ago. Always check with the trial site before acting on a status shown here.

As at this build, 2,067 trial pages are published, of which 261 are for trials the registry lists as something other than recruiting.

The one figure, and where it comes from

One number is used everywhere this site counts Australian trials, and this is its definition. A trial counts once if the registry lists it as open — recruiting, not yet recruiting, or enrolling by invitation — and lists at least one site in Australia. It is counted by registry identifier, so a trial running at six Australian hospitals counts once and a trial tagged to three conditions counts once. The same list is what this site builds a page for, so the figure on the home page, on the trial finder, on trials in Australia, on the conditions index and on the rare disease hub is also the number of trial pages published, and a reader can walk to every one of them.

Two figures on this site are deliberately not that number, and both say so where they appear. The dated pages under today's update state what was measured on the day they cover, which is a historical record and not a claim about today. Per-condition counts on a condition page count only that condition's trials. Nothing else on this site states a total of Australian trials.

How a trial is assigned to a condition

This is the part most likely to produce a wrong result, so here is how it actually works.

Every condition we cover is mapped to one or more MeSH codes. MeSH is the National Library of Medicine's medical vocabulary, and ClinicalTrials.gov publishes the MeSH codes it has assigned to each study alongside the study record. A trial is assigned to a condition when the study's MeSH codes include one of that condition's codes, or a code filed underneath it. That second part matters: a non-small-cell lung cancer study carries the code for non-small-cell lung cancer, not the code for lung cancer, and it belongs on the lung cancer page.

A code is an identifier rather than a spelling, so this does not depend on how a sponsor happened to word their condition field. That was the previous method and it produced wrong results we could measure: trials for a nerve condition called CIDP were landing on the multiple sclerosis page, because the word "demyelinating" appears in both.

Where ClinicalTrials.gov has assigned no usable MeSH code, and only there, we fall back to reading the sponsor's own condition field. About one study in six carries no MeSH code at all, and some carry codes too general to place, so without this fallback those trials would reach no condition page. The fallback can only add a condition; it never removes or overrides one the codes established.

Both steps are deliberately generous. The codes bring in recognised subtypes, and the fallback accepts common abbreviations and accepted spectrum terms, because hiding a trial somebody could join is a worse failure than showing one that is only loosely related. So a condition page can list a trial that turns out not to be relevant to you. Read the trial record before acting on it.

The ANZCTR gap, which is the biggest one

The Australian New Zealand Clinical Trials Registry, ANZCTR, is the Australian trial registry. This site does not include it.

The ANZCTR terms of use prohibit automated collection of material from its website without its approval, and we have not sought and obtained that approval. So an Australian trial registered only with ANZCTR, and not also with ClinicalTrials.gov, does not appear here at all. Many Australian trials register with both. Some do not, and we cannot tell you how many.

If you are looking for a trial in Australia, search these as well as this site:

Do not treat our list as the complete set of Australian trials. It is not, and the missing part is Australian.

Conditions: 149, not all of them

Voxsanity covers 149 conditions. There are thousands of recognised medical conditions. If yours is not among the 149, that is not a signal that no research exists. It means we have not built a page for it.

We intend to expand the list. We are not going to promise a date, because we would only be guessing.

How freshness is measured, and what the date means

Two different dates appear on this site, and they answer different questions.

  • The date the registry processed the record. This is the registry's own stamp. Where ClinicalTrials.gov publishes one it is shown on the trial page under "Sources and dates", and it is the date to trust: it tells you when the record was last updated at the source. About one record in five carries no such date at the registry, and those pages leave the line out rather than putting a date of ours under the registry's name.
  • The date we last read it. A nightly job refreshes trial records. That is a fact about the job, not about every record: a record is only as fresh as the last time the sponsor changed it.

Individual trial records are refreshed on a rolling basis rather than all at once, so a record may be up to several weeks old. Each trial page and trial card shows the date its own record was last synced. Where a record has lagged, its status, sites and enrolment can differ from the current registry state in either direction, so check the registry link on the trial page before acting on them.

The data sources page names the source behind each part of the site and the terms it sets.

Trial phases, and what a phase count means

Phase comes straight from the registry. We map the registry's value to a label and do not reinterpret it.

  • Phase 1. First testing in people, usually a small group, looking at safety and dose.
  • Phase 2. Whether the treatment appears to work, and what the side effects are.
  • Phase 3. A larger comparison against the current standard of care. This is the stage closest to a regulator being asked to approve something.
  • Phase 4. Studies of a treatment that is already approved and in use.
  • Not applicable. The registry's own value for studies where a phase does not apply, which covers many device, behavioural and observational studies.

Where a condition page shows how many trials are at each phase, that is a count of the open Australian trials described above, taken from the registry's phase field.

Historical phase success rates, and why they are not shown beside a trial

Across the pharmaceutical industry, most treatments that enter clinical testing never reach approval, and the chance of getting through improves at each later stage. The rounded historical figures usually cited are:

  • Phase 1: around 10% of drugs entering Phase 1 eventually reach regulatory approval.
  • Phase 2: around 30% of drugs entering Phase 2 eventually reach regulatory approval.
  • Phase 3: around 65% of drugs entering Phase 3 eventually reach regulatory approval.

Where this comes from. Published estimates of the same quantity disagree widely. For the chance that a drug entering Phase 1 is eventually approved, studies report figures from about 7% to about 22%; for a drug entering Phase 3, from about 49% to about 70%. The rounded figures above are the ones conventionally quoted and sit inside those ranges. They are not taken from any one study, and we do not compute them ourselves. The largest published study of the question is Wong CH, Siah KW, Lo AW, Estimation of clinical trial success rates and related parameters, Biostatistics 2019;20(2):273-286 (read it, doi:10.1093/biostatistics/kxx069), which analysed 406,038 clinical trial records covering 21,143 compounds over 1 January 2000 to 31 October 2015 and found that 13.8% of all drug development programs eventually lead to approval, while oncology reaches 3.4% in the same sample.

These are averages across the whole industry over past decades, not a prediction about any single trial. They vary by disease area, by treatment type and by decade. They say nothing about one trial's design, sponsor or candidate, and nothing at all about whether a person who joins a trial benefits from it.

Three phase values have no rate: Phase 4 studies concern a treatment that is already approved, Early Phase 1 comes before the main pathway, and Not Applicable covers studies that are not part of the approval pathway at all.

We do not show these percentages beside an individual trial, on a trial page or on a trial card. A percentage next to one named trial invites a reader to treat it as that trial's odds, or as their own, and no caveat underneath reliably undoes that. If you see one attached to a named trial, that is a defect, and we would like you to tell us.

Medicine shortages, and what they cannot tell you

We publish the shortage reports the Therapeutic Goods Administration is currently publishing, reproduced with the TGA's written permission and free to everyone. Two limits are worth knowing.

  • A report appears on one of our medicine pages only where its active ingredient matches a medicine we already track, and we do not track every medicine supplied in Australia: as at this build that is 739 of 1,175 reports. A medicine with no shortage notice here is therefore not evidence that it is available, because this understates shortages and cannot overstate them.
  • These records are national and describe registered products. They cannot tell you what your own pharmacy has in stock, and a local supply problem that was never reported to the TGA does not appear at all. Your pharmacist is the person who knows.

To check one medicine directly, use the TGA's own Medicine Shortage Reports Database.

Rare disease reference data is thin

  • Orphanet reference data is matched for 61 of the 149 conditions, and even where it is matched we hold no usable prevalence figure. Where Orphanet publishes a prevalence it is mostly European, and for several of these conditions the matched Orphanet code covers only the rare form of the condition rather than all of it, so an Australian figure derived from it would understate the real population, in some cases by a wide margin. That is why we publish none.
  • Orphan drug designation is not covered. Do not read the absence of an orphan drug indicator as meaning anything.

Things this site does not hold at all

  • No PBS data. We publish no subsidy status, no price and no restriction text for any medicine. Use pbs.gov.au, or ask your pharmacist.
  • No TGA approval dates and no ARTG numbers. The Australian Register of Therapeutic Goods is where to check whether a product is registered here.
  • No Australian research funding figures. Nothing on this site reflects what Australia spends on research.
  • No AI-written text. Nothing a reader sees here is written by an AI model.

Two things to hold in mind about all of the above

Registry data is only as current as the people who file it. A trial's status changes when the sponsor updates the registry, which can be long after the reality changed.

Absence of data is not evidence of absence. If a condition page shows no trials, the honest reading is that the registries we use do not show us any, not that there are none.

If you find something on this site that is wrong, tell us. Every page has a way to report an issue, and there is also the contact page. We read every report and reply as soon as we can. What we have already corrected is on the corrections page.

Related: our data sources and our editorial policy.