The disclosure calendar in oncology is unusually orderly. A large share of what a CI team will learn in a year becomes public at ASCO, ESMO, AACR, ASH and a handful of disease-specific meetings, on dates published a year in advance.
That predictability is the opportunity. It is also why congress coverage separates competent CI functions from busy ones: everyone gets the same data at the same moment, so the advantage is entirely in what you did beforehand and how fast you can interpret.
The cycle
| Stage | What becomes available | CI work |
|---|---|---|
| Abstract titles Weeks ahead |
Titles and session assignments, often without results | Triage. Which sessions matter, who presents, what is conspicuously absent |
| Abstract release Days ahead, under embargo |
Top-line results against an older data cut | First real read. Draft the interpretation now, not later |
| Late-breakers Often very late |
The submissions that could not make the deadline | Re-plan. This is where practice-changing data tends to land |
| Presentation On the day |
Mature data cut, full safety, subgroups, discussant's view | Compare against the abstract. The delta is the story |
| The room Same day |
Discussant framing, Q&A, corridor reaction | The part no database will ever carry |
| Publication Sometimes simultaneous |
Full methods and supplementary appendix | Verify. The appendix often contains what the slides omitted |
The abstract is not the data
This is the single most common error in congress coverage, and it produces confidently wrong analysis every year.
An abstract is written against a data cut months old, because submission deadlines are months before the meeting. By the time it is presented, the sponsor is usually showing a more mature cut: longer median follow-up, more events, sometimes a different point estimate. A hazard ratio can move meaningfully between abstract and podium, and occasionally the direction of the narrative moves with it.
Abstracts are also space-constrained in a way that systematically hides the interesting parts. Safety detail, discontinuation rates, dose reductions, subgroup consistency — all of that is usually on the slides and absent from the abstract. A team that publishes its competitive read off the abstract and then does not revisit it after the presentation is circulating a stale conclusion internally.
Practical rule. Treat the abstract read and the presentation read as two separate deliverables, and say which one a given document is. The most valuable single line in post-congress coverage is often "this changed from the abstract, and here is how."
What the room gives you that the slides do not
Three things, none of which are recoverable afterwards from any database:
The discussant. Major sessions pair a presentation with an independent expert's critique. That framing frequently determines how the data is received by the field — a discussant flagging a concern about a control arm can blunt an otherwise positive readout. Getting the discussant's slides and argument is high-value and easy to miss.
The Q&A. What the audience of practising oncologists chose to ask about tells you where the scepticism sits. Repeated questions about a particular toxicity or subgroup are a signal about adoption that the primary endpoint will not give you.
Corridor reaction. Whether investigators actually believe the result is clinically meaningful, and whether they would change practice. This is the input that most often contradicts the headline.
Preparing properly
- Build the target list from titles, before the abstracts drop. Session, time, presenter, trial ID, why it matters and to whom internally. If this is not done in advance it will not be done at all.
- Pre-write the interpretation frames. For each priority readout, write down in advance what result would be good, bad or ambiguous for you, and what each would imply. Doing this before you see the number is the only defence against reading the data to suit the plan.
- Assign human coverage against conflicts. Priority sessions collide. Decide the trade-offs in advance rather than in a corridor.
- Leave slack for late-breakers. A plan with every analyst committed cannot absorb the one readout most likely to matter.
- Agree the turnaround before you go. Same-day flash on priority readouts, fuller synthesis within 72 hours. After about a week, congress coverage stops being intelligence and becomes archive.
The reason this is hard to staff
A major congress needs people who can sit in an oncology session, understand the statistics, recognise which subgroup analysis is post-hoc, and know enough about the competitive context to see the implication — all at once, at volume, for four days. That is a narrow skill set, and it is why congress coverage is one of the most commonly outsourced CI activities.
If you are buying it, the questions worth asking are about coverage depth rather than breadth: how many analysts are physically present, whether they cover the discussant and Q&A rather than just the primary slides, and what the turnaround actually is. Published post-congress write-ups are the easiest way to judge that before committing — Beyond the Abstracts: AACR 2026 is an example of the form.
After the meeting
Two things almost always get skipped, and both are cheap.
Reconcile against your pre-written frames. Where the result landed outside what you expected, ask why your model was wrong. That is the only mechanism by which a CI function's judgment improves.
Update the record, not just the deck. New data cut, revised expected readouts, changed standard of care. If congress output lands only as a PDF, next quarter's landscape refresh starts from scratch — the failure described in indication-level tracking.