Australian Clinical-Trials Intelligence
22,006 trials · with a NSW governance lens

Australian clinical-trials intelligence

Every clinical trial in Australia, in one view

A national intelligence platform over the whole field: every Australian clinical trial with at least one state-based site, triangulated across four sources. Start here for the national picture, then explore and cross-filter all 22,006 trials - or drop into the NSW governance lens below.

Explore every Australian trial Register · live cross-filter · per-trial provenance
0
Australian trials
with ≥1 state-based site
0
Interventional
drug, device & procedure studies
0
Observational
cohort & registry studies
0
NSW trials
counted per state · the national leader by volume

Trial volume by year

Registrations 2016-2026 · interventional vs observational

This timeline plots 17,660 trials by registration year; 4,346 of the 22,006 total were registered before 2016 or carry no registration year and are not shown here.

Where trials run, by state

Reach per state · NSW highlighted

22,006 distinct trials. A trial is counted in every state it runs in, and 8,428 run in 2+ states, so these bars overlap and intentionally sum to more than 22,006 - this is each state's reach, not a split of the total.

National phase pipeline

All 22,006 Australian trials · normalised registry phase

“Not applicable” is a real registry phase, not missing data: non-drug interventions (behavioural, device, surgical, dietary, procedural) and observational studies have no clinical-trial phase - and Australia runs a large volume of these. Drug-trial phases 1-4 are shown above.

Source coverage

Trials carrying each source · a trial can appear in several

Explore - every Australian trial

Slice the whole national field

One shared filter over all 22,006 trials. Pick states, phases, registries, sponsors or a therapeutic area - or just search a title - and every chart, count and table row below recomposes live. A trial is counted in each state it runs in, so the state bars double-count multi-state trials. NSW is just one of eight quick filters.

Loading 22,006 trials…

Fetched once, on demand - this large national register is kept out of first paint.

Sites - every research facility

Every site that runs a trial

All 2,233 physical research facilities in Australia's clinical-trial register (556 in NSW), entity-resolved to single physical sites. Filter by state, sector, governance or therapeutic area - or search a name - then open any site for its full dossier: trial volume, phase & therapeutic-area mix, the sponsors running trials there, and the trials it conducts (each opens its own dossier). Sector is NSW-specific; interstate sites are shown unclassified. These counts reconcile with the NSW lens (556 NSW facilities, 199 public NSW Health).

Loading 2,233 sites…

Fetched once, on demand - kept out of first paint.

Research intelligence - the analytical heart

Who runs Australian trials, where, with whom

Beyond counts: the institutions and sponsors that actually move the national field, and the collaboration ties between research sites. Entity-resolved across all 22,006 trials - explore the sponsor and institution leaderboards, drill into a dossier, trace the connected ecosystem in the site collaboration network, or deep-dive a therapeutic area.

Loading research intelligence…

Four entity-level bundles (~230 KB) - sponsors, institutions, the collaboration network and therapeutic areas - fetched once, on demand.

Where the trials actually run

The footprint, mapped

Three lenses on the same geography. Switch between the national distribution, NSW's Local Health Districts, and every mappable research facility - sized by trial volume and coloured by sector.

Australian trials by state (counted per state)

Where every trial is recorded

Four systems, one register

Every Australian trial in the register is triangulated across up to four systems - the two public registries (ANZCTR, ClinicalTrials.gov) and NSW Health's governance records (CTMS, REGIS). This shows how many distinct trials each system holds and exactly where they overlap - the de-duplicated truth behind the register. CTMS & REGIS are NSW-only by nature, so their sets are small and NSW-bound.

Computing source overlaps…

A lens within the platform

The NSW governance lens

National data answers what is running. This lens asks a sharper question of one jurisdiction: of the trials with a NSW site, which ones does NSW Health actually govern - and which sit outside its oversight line entirely? The story, the provenance and the NSW-scoped cross-filter all live below.

The NSW governance blind spot

The blind spot in clinical-trial governance

Within its governance line, NSW Health sees the trials run across its own public facilities. Beyond that line sits a second research economy - commercial, academic and unclassified sites - that the same oversight never touches.

0
Trials within governance across 0 public NSW Health facilities
0
Trials outside governance across 0 facilities NSW Health doesn't oversee
0 current interventional trials across NSW · 0 facilities in total
How we found them Triangulated across two trial registries - the overlap is the cross-validated core.
Within governance Outside governance - the blind spot
Where the evidence lives

Where the evidence lives

Three systems, one truth-unevenly shared

Every NSW trial leaves a trace in at least one of three systems: the two public registries - ANZCTR and ClinicalTrials.gov - and NSW Health's own governance records (CTMS & REGIS). Where they overlap, a trial is cross-validated. Where they don't, something is invisible to someone. Full NSW universe - 9,626 trials, all statuses and study types.

0
Governance-only - the public blind spot

Public-sector LHD trials recorded in NSW governance systems that were never registered in either public registry. Invisible to anyone reading the registries.

0
Registry trials outside governance

Registered with ANZCTR and/or ClinicalTrials.gov, with a NSW site, yet absent from NSW Health's governance records - the registry-side blind spot.

0
Cross-validated by governance

Recorded in NSW governance and at least one public registry - the reconciled core where both views agree.

Tip: click a registry-only region (ANZCTR-only or ClinicalTrials.gov-only) to filter the register below. The register holds the 3,189 current-interventional trials - a subset of this 9,626-trial universe - so the filtered count is smaller than the region's figure here.

NSW lens · slice the cohort

The anatomy of the activity

This cross-filter is scoped to the NSW governance cohort - the 3,189 current interventional trials with a NSW site, the set where governance status is known. (For the all-Australia cross-filter over every 22,006 trials, use Explore above.) Click any bar - or a sector in the map legend - and every figure, chart, marker and row recomposes live.

3,189 trials of 3,189
2,438 within governance
751 outside governance
295 facilities touched
1,175 distinct sponsors

Therapeutic area

Click a bar to filter · Cancer dominates the cohort

Phase

Normalised registry phase

Recruitment status

Live enrolment state

Sponsor class

Who runs the trial

Registry source

The live triangulation · gold = cross-validated

Full cohort - not affected by the filters above

These two views are measured per site and per state, not per trial, so a single trial can land in several districts and states at once. They cannot be cross-filtered without double-counting - they stay fixed as full-cohort context.

Local Health District

NSW Health internal footprint · current trials per LHD

NSW in national context

Current interventional trials by state

NSW lens · down to the row

The NSW trial register

Every trial in the filtered NSW cohort, one row each. Brush any chart above or the map legend and this table follows. Sort any column, page through the set, or export exactly what you're looking at as a CSV. (The full national register of all 22,006 trials lives in Explore.)

Showing 1 to 25 of 3,189
Page 1

How this platform is built

Methods & data provenance

A transparent account of every data source, definition, linkage rule and limitation behind this platform. It is built for scrutiny: each figure on this site traces back to the rules set out here.

This platform is part of RAMS, the Research Activity Monitoring System, developed exclusively for NSW Health. Every figure is derived from a validated, reproducible pipeline over public trial registries and NSW Health governance records. Counts reflect the current data build and are revised as the sources refresh.

1 · Data sources

The platform triangulates five systems: two public clinical-trial registries (global and Australian) and three NSW Health governance or operational records. The registries establish which trials exist; the NSW systems establish which were ethically approved and site-authorised to run in NSW Health facilities.

SourceCustodianCoverageVariables (held / used)RefreshRole here
ClinicalTrials.govUS NIH / NLM~1996 to 2027100+ protocol fields / ~20 coreDaily API (v2)Global registry
ANZCTRNHMRC (AU / NZ)2007 to 202565 trial fields across 19 linked tables / core setPeriodic bulk exportAU / NZ registry
CTMSNSW Health OHMRtrials span 1997 to 202617 fields / key setPeriodic exportNSW trial-management operations
REGIS · SSANSW Health (REGIS)May 2018 to May 202626 fields / key setPeriodic exportNSW site authorisation
REGIS · ETHNSW Health (REGIS)May 2019 to May 202619 fields / key setPeriodic exportNSW research ethics

REGIS (the NSW research-governance system) records submissions only from mid-2018 onward, whereas the public registries reach back much further (ANZCTR to 2007, ClinicalTrials.gov to the late 1990s). This is why many early-year trials appear in the registries with no NSW governance record.

2 · The universe: what is counted

The headline 22,006 is the de-duplicated set of clinical trials with at least one Australian site, pooled across the four registry and governance systems. It is:

  • Cumulative and all-time, not a single year. Registration dates span the late 1990s to 2027; the time-series charts are deliberately windowed to 2016 to 2026 (17,660 trials) for readability.
  • All study statuses (completed, recruiting, terminated, withdrawn and so on), not active trials only. 9,534 are currently active, as defined below.
  • Both study types: 19,398 interventional and 2,352 observational, plus a small residual of expanded-access and untyped records.

By system the universe holds ANZCTR 11,989 · ClinicalTrials.gov 9,980 · CTMS 1,441 · REGIS 1,713. These sets overlap, so they sum to more than 22,006. A further 841 REGIS-governed studies matched no public registry and sit deliberately outside the 22,006 (surfaced on the Sources tab).

3 · Definitions

Active / current registry-derived

A trial is "current" when its status is recruiting, not yet recruiting, active not recruiting, or enrolling by invitation. Each registry's native status vocabulary is normalised to a single controlled list via a published crosswalk. One judgement to note: NSW operational wind-down statuses (for example "pending closeout") are counted as current, so "current" leans inclusive.

Interventional vs observational registry-native

Taken directly from each registry's study type field. The headline "clinical trials" framing centres on interventional studies; observational studies are retained and labelled.

Phase registry-native

From the registry phase field, normalised to Phase 1 to 4 (with span buckets such as Phase 2-3), "Not Applicable", or "Unknown". Early-phase and phase-0 records are rolled into Phase 1.

Sponsor class: industry vs non-industry our classification

The "industry" flag is anchored on ClinicalTrials.gov's own funder type (agencyClass = INDUSTRY) plus ANZCTR's "Commercial sector / Industry" sponsor type. The binary split itself is ours, not an external standard: every non-commercial sponsor type (government, academic, hospital, network, individual) collapses into a single broad "non-industry" residual. Treat it as a useful first cut, not a definitive funder taxonomy.

Therapeutic area hybrid

For ANZCTR-registered trials the area is the registry's own Health Condition category (an authoritative coding). For ClinicalTrials.gov-only trials, a keyword classifier maps free-text conditions onto the same category names. Trials with no usable condition text are labelled "Uncategorised", and area shares are computed over the categorised subset, not the full universe.

Governance-only defined rule

A trial is "confirmed" only if it appears in CTMS, ANZCTR or ClinicalTrials.gov. NSW site-authorisation records (REGIS-SSA) that match none of these are reported separately as governance-only (841 studies) and are not part of the 22,006.

4 · Record linkage & entity resolution

Across registries. A trial registered in more than one system is merged into a single canonical record. Linkage is deterministic first: a shared primary registry ID (NCT or ACTRN), including cross-references one registry records about another (a ClinicalTrials.gov record naming its ANZCTR number, or the reverse). A conservative fuzzy fallback (near-identical title, plus sponsor agreement, plus matching year) catches dual registrations that share no ID.

Why ANZCTR and ClinicalTrials.gov barely overlap. They are separate registries, and most Australian trials register in only one. Even when a trial is dual-registered, the second registry's ID is frequently left blank: only 87 of about 12,200 ClinicalTrials.gov Australian records declare an ANZCTR number, and only 46 of about 15,600 ANZCTR records declare an NCT. Beyond these declared IDs, the probabilistic stage below recovers same-trial pairs from title, sponsor and year, so the small overlap is a real property of the data, not a matching failure. Read it as a lower bound: only dual registrations with titles too divergent to clear the threshold, and no shared ID, remain unlinked.

Across facilities. Site names are resolved to physical facilities by normalising punctuation and apostrophes, expanding abbreviations (RPA, RNSH and the like), reconciling compound and proper-name variants, and splitting by state. This collapses, for example, more than 100 spellings of Royal Prince Alfred Hospital into a single facility.

4a · The RAMS entity-resolution cascade

Deciding when two records describe the same real-world trial is run as a named, reproducible method: a two-stage cascade that links on exact identifiers first, then recovers same-trial pairs that carry no shared ID. It is precision-first by design - a false merge fuses two distinct trials and corrupts the headline count, so the cascade would rather leave a pair unlinked, and escalate it for human review, than risk a wrong merge.

Source records ANZCTR · ct.gov CTMS · REGIS STAGE 1 Deterministic identity shared-ID union-find (NCT, ACTRN) → exact identity clusters unresolved pairs STAGE 2 Probabilistic matching block → score (τ, σ) → decision D merge review reject Canonical trials
Figure 1. The two-stage cascade. Stage 1 forms exact identity clusters by union-find over registry IDs; Stage 2 scores only the cross-source pairs Stage 1 left unresolved, routing each to merge, human review, or reject.

1Normalisation and significant tokens

K(s) = { w ∈ T(s) : |w| ≥ 4 }

T lowercases, maps punctuation to spaces and collapses whitespace; K(s) keeps tokens of at least four characters.

2Stage 1 - deterministic identity

i ∼ j ⇔ ∃ k ∈ {NCT, ACTRN} : k(i) = k(j) ≠ ∅

Canonical entities are the connected components of ∼ (union-find). A secondary identifier v (EudraCT, UTN) is admitted only when unambiguous:  admit(v) ⇔ |NCT(v)| ≤ 1 ∧ |ACTRN(v)| ≤ 1.

3Stage 2 - candidate blocking

C = { (m, c) : |ym − yc| ≤ 1 ∧ ∃ w ∈ K(m) ∩ K(c) : w ∉ Σ }

Only the cross-source pairs Stage 1 left unresolved enter C; y is the start year and Σ the stop-token set (corpus frequency > 200). Blocking cuts the comparison space from O(n2) to near-linear.

4Similarity scores

τ(m,c) = TSR(titlem, titlec) σ(m,c) = TSeR(sponsorm, sponsorc)
TSR(a, b) = 100 · ( 1 − d(a★, b★)|a★| + |b★| )

Both scores lie in [0, 100]. a★ is the string with its tokens sorted and rejoined; d is the indel (insertion / deletion) edit distance. The token-sort ratio (TSR) is word-order insensitive; the token-set ratio (TSeR) also ignores duplicated tokens, and σ = 100 when either sponsor is blank.

5Decision function

D(m,c) =mergeτ ≥ 92 ∧ σ ≥ 80review80 ≤ τ < 92rejectotherwise

Merge pairs are unioned into the entity; review pairs are escalated to a human-adjudicated queue and never auto-linked. On the current corpus the cascade confirms 151 cross-registered ANZCTR / ct.gov trials and surfaces 153 borderline pairs for review.

merge human review reject reject 60 80 92 100 0 80 100 title similarity τ (token-sort ratio) sponsor σ
Figure 2. Decision regions in title-similarity (τ) by sponsor-similarity (σ) space. Only the green corner auto-merges; the amber band is held for human adjudication; everything else is rejected. The narrow merge corner is the precision-first stance made visible. (τ-axis shown from 60.)
SymbolParameterValue
τ*title auto-merge threshold92
τrreview-band lower bound80
σ*sponsor agreement gate80
Δystart-year tolerance±1
|Σ|stop-token frequency cutoff200
min |w|significant-token length4

5 · Reading the numbers

  • Per-state counts double-count by design. A trial running in three states is counted in all three, so the state bars sum to far more than 22,006 (about 46,000). They measure trial-state presence, not distinct trials.
  • Headline subsets are narrower than the universe. NSW-governance views focus on current interventional NSW trials (3,189); the 22,006 is the full national, all-status pool.
  • Counts are lower bounds. About 1,569 NSW trial-site records lacked a usable facility name and could not be attributed; 202 facilities remain unclassified by sector.

6 · Limitations

  • Dual-registration overlap is a floor. Cross-registry overlap reflects declared IDs plus high-confidence fuzzy matches; an unknown residual of undisclosed dual registrations is not linked.
  • Sponsor class is a binary heuristic, not a funder taxonomy; "non-industry" is a broad residual, and sponsors are not yet entity-resolved, so the same sponsor under spelling variants can be over-counted.
  • NSW governance coverage is structural. REGIS covers only NSW Health public-site trials and begins mid-2018, so private-site and pre-2018 trials carry no NSW governance record by definition.
  • Therapeutic area for registry-silent trials is keyword-inferred and single-area per trial; condition-less records are "Uncategorised".
  • Status reconciliation favours the freshest source when registries disagree, and CTMS wind-down statuses count as current.

7 · Future directions

  • Linkage to grants, patents and publications for an end-to-end view of research translation.
  • Entity-resolved sponsors and investigators, replacing raw-string counts.
  • Automated source refresh with versioned data builds and published reconciliation reports.
  • Extension beyond NSW Health to other organisations and jurisdictions.

About

Developed by Dr Yagiz Aksoy, MD PhD

linkedin.com/in/yagizalpaksoy

Part of RAMS (Research Activity Monitoring System) · developed exclusively for NSW Health