Why Rare Cancer Patients Need a Different Information Architecture

Individually, each rare cancer affects few people. Collectively, rare cancers account for a substantial share of all cancer diagnoses. This produces a situation where a large number of patients each face a disease about which comparatively little is known.

The usual response is that rare cancers need more research. True, and insufficient. The evidence structure is different in kind, and systems designed around common disease fail in specific ways when applied to it.

How the evidence differs

For a common cancer, the evidence pyramid works as intended: randomised trials at the top, observational data below, case reports at the base, and the recommendation draws on the upper tiers.

For a genuinely rare tumour, the upper tiers may be empty and may remain so. Not because of neglect, but because a trial powered for a conventional endpoint may require more patients than are diagnosed globally in a decade.

An evidence framework that treats the absence of high-tier evidence as a deficiency to be corrected gives, in these diseases, a permanently unsatisfying answer. What is needed is a framework for reasoning well from what does exist.

Four consequences

Diagnosis carries more weight and is less reliable. Rare entities are misclassified more often, simply because fewer pathologists have seen many of them. Specialist review is not a courtesy in rare disease; it frequently changes the diagnosis, and the diagnosis determines everything downstream.

Expertise is geographically scattered. The people who have seen this entity are distributed across institutions and countries. For common cancers, local expertise is usually adequate. For rare cancers, the relevant expert is often elsewhere, and finding them is part of the problem.

The literature is fragmented and moves quickly relative to its size. A handful of new papers may represent a substantial proportion of everything published on an entity. Monitoring is more tractable than in common disease, but only if someone is actually doing it.

Trial discovery is disproportionately hard. Registry search assumes you know what to search for. In rare disease the relevant trial may be a basket trial recruiting by molecular alteration rather than by histology, which does not surface under a disease-name search.

Where molecular characterisation carries more weight

When disease-specific evidence is thin, molecular findings carry proportionally more of the reasoning load. Tissue-agnostic alterations — NTRK fusions, mismatch repair deficiency, high tumour mutational burden — matter more here precisely because the disease-specific literature offers less.

This makes comprehensive, fusion-capable profiling more rather than less important in rare tumours, which inverts the intuition that expensive testing should be reserved for cases where the yield is predictable.

What a better architecture looks like

It routes difficult diagnoses to people who have seen the entity. It searches trials by molecular alteration as well as by disease name, and internationally rather than locally. It synthesises small-series evidence explicitly, stating what tier the recommendation rests on. It treats "the evidence here is thin" as information to be communicated rather than a gap to be papered over. And it connects patients to the small number of others with the same disease, since registries and patient networks carry real information in these conditions.

The underlying point

Rare cancer is not a scaled-down version of common cancer. The information problem has a different shape, and building for it means designing differently rather than doing less of the same thing.