TIGER platform

What the registry holds, and what the platform does with it

An international multicentre registry for oesophageal cancer, and the analysis built on top of it. Every figure below is read live from the public statistics endpoint when this page loads.

Patient records: Loading…
Institutions: Loading…
Countries: Loading…
Recorded edits: Loading…

What the platform does

Each card describes a capability of the platform — the schema it implements, the analysis it performs, the formats it exchanges — and links to the documentation for it. These are statements about the software, not measurements of the registry.

Anatomical lymph-node map

Every metric painted onto the anatomical station diagram, station by station.

41 station slots, joined to the diagram by anatomical label and side — never by number.

eCRF station schemaRead more

Cohort comparison

Build two cohorts from eCRF fields and compare them side by side.

Cohorts are defined from the data dictionary, so a comparison is reproducible from its definition alone.

Analytics guideRead more

Survival analysis

Kaplan-Meier curves with strata and at-risk tables.

Curves carry the number still at risk at each milestone, so a tail drawn from a handful of patients reads as one.

Analytics guideRead more

AJCC 8th edition staging

Stage derived from T, N and M rather than typed by hand.

Implemented against the AJCC 8th edition oesophageal tables; the derivation is shown alongside the stage.

eCRF data dictionaryRead more

Data-entry provenance

A per-record audit trail: when it was entered, when it was revised, how often.

Reconstructed from the submission log, which is authoritative for edits from the first logged event onward.

Entry-provenance serviceRead more

Import and export

CSV and XLSX in through a validating wizard; CSV, XLSX, CDISC ODM and FHIR out.

Imports are staged, validated and signed before promotion, so a bad file never lands half-applied.

Import and export guidesRead more

Kaplan-Meier survival

The platform estimates survival over a cohort and splits it by any of the registry’s oncological variables — T stage, AJCC 8th stage group, resection margin, histology, neoadjuvant therapy, nodes harvested, and the one below. Pathologic N stage is the registry’s primary nodal variable and the stratifier the dashboard opens on.

The curves are drawn from a synthetic cohort committed to this repository, and the figure says so. No survival data is published on the public statistics endpoint: a Kaplan-Meier plot draws every death as a visible step, so a registry curve served to an anonymous visitor would publish the patients it was estimated from. What is shown here is the shape of the output and the apparatus around it, not a result.

Overall survival by pathologic N stage

Illustrative curves — shape only, not registry data

Registered members analyse their own data in real time and compare it with the global registry

Number of patients still at risk at each milestone, by stratum
Number at risk	0	12	24	36	48	60
pN0	200	171	141	118	108	100
pN1	150	111	76	61	46	42
pN2	100	65	42	21	18	15
pN3	50	27	16	6	5	4

Time in months. “—” means the milestone is beyond that stratum’s observed follow-up — it is not a zero.

The same estimates as the figure above, in figures. Illustrative curves — shape only, not registry data. Registered members analyse their own data in real time and compare it with the global registry.
pathologic N stage	Patients	Deaths	Median survival	12 months	36 months	60 months
pN0	200	67	not reached	90.7%	69.0%	63.0%
pN1	150	84	36.1 months [23.4, 46.4]	79.8%	50.7%	36.8%
pN2	100	77	21.8 months [15.9, 27.0]	67.3%	24.8%	17.7%
pN3	50	44	13.4 months [8.8, 22.3]	55.2%	12.8%	8.5%

Steps fall at observed deaths; vertical ticks mark censored follow-up — a patient still alive when the registry last heard from them. Between two deaths the estimate does not change, so each curve runs flat and stops at the end of the follow-up that was observed. The shaded area around each curve is its Greenwood 95 % interval; it widens as the number still at risk falls, which is why the row of counts under the axis is read with the tail rather than after it. Across the strata, log-rank p < 0.001 (χ² = 104.0 on 3 df). A median of “not reached” means the curve never fell to 50 %, which is not a missing value.

Over registry records the same figure is drawn from real follow-up rather than a synthetic cohort, and any stratum with fewer than five patients, or fewer than five deaths, is dropped from the estimate rather than drawn. Investigators reach it from the analytics dashboard, scoped to their own institution or to the whole registry.

Go deeper

Inside the registry

Geographic reach, and the data-entry provenance story: how much of this registry has been revised since it was first entered, and what the edit log can and cannot tell you.

View the registry

The lymph-node station map

The anatomical diagram, the 41 station slots recorded against it, and the three metrics the platform paints onto it — with the plain-English reading of each.

View the node map

Number at risk	0	12	24	36	48	60
pN0	200	171	141	118	108	100
pN1	150	111	76	61	46	42
pN2	100	65	42	21	18	15
pN3	50	27	16	6	5	4

Number at risk	0	12	24	36	48	60
pN0	200	171	141	118	108	100
pN1	150	111	76	61	46	42
pN2	100	65	42	21	18	15
pN3	50	27	16	6	5	4

Number at risk	0	12	24	36	48	60
pN0	200	171	141	118	108	100
pN1	150	111	76	61	46	42
pN2	100	65	42	21	18	15
pN3	50	27	16	6	5	4