Part XV · Supportive care, survivorship, and patient experience · Chapter 87

Communication and shared decision-making

A large relative benefit and a small absolute one are the same number described twice.

1 · Communicating prognosis and uncertainty

Patients want prognostic information more often than clinicians offer it, and they want it in a form clinicians rarely use. A median survival is a property of a distribution. A patient is a single draw from it.

Three framings carry different information and are not interchangeable. A median says where the middle is. A range says how wide the distribution is. A best-case and worst-case pair says what the tails look like. Giving only the median hides exactly the part that matters to someone deciding whether to take a trip this year.

Uncertainty should be attributed rather than merely admitted. Saying that the estimate is uncertain is less useful than saying what would make it more certain, which is often a characterisation that has not been done or a response that has not yet been observed.

2 · Absolute and relative benefit, and why the difference matters

This is the section the rest of the chapter depends on.

A hazard ratio is a ratio. It says nothing about how much disease there is to prevent. The same hazard ratio produces a large absolute benefit in a high-risk population and a negligible one in a low-risk population, and the trial that generated it was run in one of them.

Adjuvant abemaciclib in high-risk node-positive disease is the worked example. Comparing abemaciclib with endocrine therapy against endocrine therapy alone, seven-year invasive disease-free survival was 77.4% against 70.9%, with a hazard ratio of 0.73. Seven-year overall survival was 86.8% against 85.0%, with a hazard ratio for death of 0.84¹. Both numbers come from the same trial. The first is a 6.5 point difference in disease events. The second is a 1.8 point difference in deaths, at seven years, in the highest-risk population enrolled.

A patient told that the drug reduces the risk of recurrence by about a quarter has been given a true statement that will not support her decision. Now take the same evidence framed differently. Out of a hundred women like her, roughly seven will avoid a recurrence and roughly two will avoid a death by seven years. All one hundred will take a drug with diarrhoea and fatigue for two years. A patient told that has been given the evidence in a form she can use.

The same arithmetic governs extended endocrine therapy and genomic-assay-guided chemotherapy decisions. The assays in Prognostic and predictive assays and somatic profiling exist precisely because relative benefit is roughly constant while absolute benefit is not.

3 · Decision aids and preference-sensitive choices

A decision is preference-sensitive when the options differ in the kind of outcome they produce rather than in how much of one outcome. Mastectomy against breast conservation, chemotherapy against omission at intermediate genomic risk, and extended endocrine therapy are all of this kind.

Decision aids improve knowledge and reduce decisional conflict, and they do not reliably change which option is chosen². That is the correct result. An aid that shifted choices in one direction would be a persuasion tool.

The failure mode is deploying an aid where the decision is not actually preference-sensitive. Most adjuvant trastuzumab decisions are not close calls, and presenting them as balanced misrepresents the evidence.

4 · Explaining genomic results and variants of uncertain significance

A variant of uncertain significance is not a weak positive. It is an absence of information, and patients and non-specialist clinicians both tend to read it as a small positive.

The two sentences that do most of the work are that the laboratory found a change it cannot classify, and that the correct action is the one that would be taken if no test had been done. Reclassification over time should be raised at disclosure rather than later, because it explains why the result may change without the patient changing.

Somatic reports carry a different confusion. An alteration listed as actionable in one tumour type may have no evidence in breast cancer, which is what tiering systems such as the ESMO scale exist to encode³. A report that lists a drug is not a report that recommends it.

5 · Cultural humility, language access, and health literacy

Professional interpretation is a clinical intervention with an evidence base, and ad hoc interpretation by family members produces omissions in both directions. Using a child as an interpreter is a safety event.

Health literacy is not general literacy, and it is poorly predicted by education. Numeracy is worse still, which is why the absolute-benefit framing in Absolute and relative benefit, and why the difference matters should use natural frequencies rather than percentages wherever possible.

Cultural humility differs from cultural competence in a way that matters practically. Competence implies a body of knowledge about a group. Humility implies asking this patient what she wants, which is more accurate and takes less time.

6 · Misinformation and how to respond to it

Correcting a false claim by repeating it tends to reinforce it. The more reliable approach is to state the accurate claim, explain the mechanism, and address the underlying concern that made the false claim attractive.

Most breast cancer misinformation attaches to real gaps. Claims about sugar, dairy or deodorant persist partly because patients are given no account of why they developed the disease. An honest answer, that most breast cancer has no identifiable cause in the individual, is more useful than a correction alone.

The clinician's authority is limited here and should be spent carefully. Dismissing a patient's source tends to end the conversation rather than the belief.

References

All 3 references

1. Johnston S, Martin M, O'Shaughnessy J, et al. Overall survival with abemaciclib in early breast cancer. Ann Oncol 2026 37:155-165. PMID 41110697
2. Stacey D, Lewis KB, Smith M, et al. Decision aids for people facing health treatment or screening decisions. Cochrane Database Syst Rev 2024 1:CD001431. PMID 38284415
3. Condorelli R, Mosele F, Verret B, et al. Genomic alterations in breast cancer: level of evidence for actionability according to ESMO Scale for Clinical Actionability of molecular Targets. Ann Oncol 2019 30:365-373. PMID 30715161