How Much Benefit Would Make This Worth It?

At a case conference, I was surprised by how many colleagues thought chemotherapy was a no-brainer. We were discussing whether a woman should have it at all. I thought there was quite a bit to talk through with her before we settled on an answer.

Here’s a hypothetical version of the case. She’s in her late 30s, premenopausal, with stage III, hormone receptor-positive, HER2-negative breast cancer. Her tumour has a luminal A profile and a low-risk genomic result. We’re deciding what to give her before surgery. How much chemotherapy would help is uncertain.

She might well choose it. I could understand that. But I’d want her to know what we thought it would add, how sure we were, what she’d have to go through, and what the hormonal treatment options looked like. She might hear all of that and decide the likely benefit wasn’t enough for her. I thought there was room for either conversation.

We didn’t agree. Something else surprised me too: some colleagues had little experience with genomic tests other than Oncotype DX. I wondered whether we were giving enough attention to evidence that was less familiar to us. I don’t know how much that explained anyone’s recommendation. It did make me want to slow down before calling the choice obvious.

There’s more than one test

RxPONDER used Oncotype DX, MINDACT used MammaPrint, and OPTIMA used Prosigna. So when we say a tumour is low risk, the next question is which test told us that. The studies enrolled different patients, used different endocrine treatment and followed people for different lengths of time. Knowing one of them well doesn’t do the work of reading the others.

OPTIMA is a good example of why that matters. Its first results, presented at ASCO 2026, support leaving out chemotherapy for patients with low Prosigna scores, including premenopausal women receiving ovarian suppression. At five years, the trial met its prespecified non-inferiority criterion for invasive breast cancer-free survival: outcomes without chemotherapy were close enough to those with it to fall within the difference the trial had set as acceptable. The patients were 40 or older and had already had surgery. We still have to work out how far those results take us with a younger woman deciding on treatment before surgery.

And that’s only part of the reading. There’s SOFT and TEXT on endocrine treatment and ovarian suppression after surgery. There’s STAGE on endocrine treatment before surgery in premenopausal women, and NBRST on tumour subtype and response to treatment before surgery. The Early Breast Cancer Trialists’ Collaborative Group compared chemotherapy given before and after surgery. Further down the road, monarchE and NATALEE help us consider adding a CDK4/6 inhibitor to endocrine treatment for particular patients. None of these studies tests all the choices this woman faces.

I can’t put that together reliably in my head while someone waits for an answer. I can remember findings and have a view about what they mean. Getting from there to a number I can defend takes work. Which patients resemble her? What treatment were they already getting? Are we talking about shrinking the tumour, preventing recurrence or living longer? Over how many years? Adding a few trial results together won’t give us her answer.

We need calculations whose assumptions we can check, including what happens when we change them. There will still be gaps. If I think chemotherapy is likely to add little, I need to explain how little, how sure I am, and what might make that estimate wrong. Where we can’t make a reliable estimate, I need to say so.

Suppose she’d accept treatment at the upper end of the plausible benefit range, but wouldn’t at the lower end. That uncertainty matters to her choice. She needs more from me than “the evidence is uncertain, but I recommend it.”

I can see why we’d disagree

Look at it from the colleague’s side. She’s young. She has stage III disease. There are data suggesting chemotherapy might help, even though they don’t directly establish its benefit in her precise circumstances. It isn’t hard to understand wanting to give her that chance.

I can also understand starting with “first, do no harm” and wanting stronger evidence before putting her through chemotherapy. Neither doctor needs to have missed a paper. They may have read the same papers and come away with different ideas about how much evidence is enough to act.

One worries more about giving treatment that turns out to have done more harm than good. The other worries more about withholding something that could have helped. Both have to take the full evidence seriously. Both have to reckon with the harm of getting it wrong. “Do no harm” doesn’t let us off either hook.

What I keep coming back to is that we’re deciding how much uncertainty to accept in somebody else’s life. If I make that call without talking it through with her, she gets my tolerance for risk along with my medical opinion.

And there’s a lot tucked into the word “toxicity.” Depending on the treatment, it can mean fatigue, pain, nerve symptoms, trouble concentrating, or effects on sexual health and fertility. It can get tangled up with work, caring for family and being able to manage on your own. How long it lasts matters. So does whether it might stay.

Hair loss is one example. As a bald man, I can certainly empathise (I still miss my hair, and many of my friends have gone to great lengths, travelling to other countries to get hair transplants). I’ve been struck in adjuvant breast cancer conversations by how traumatic the prospect can be for some patients. I still need to ask what it means to her. My own feelings about it won’t get us very far.

There’s evidence that we don’t always hear the whole story. A May 2026 analysis of PANTHER compared clinicians’ toxicity reports with patients’ symptom reports during adjuvant breast cancer chemotherapy. It included 1,566 people and six symptom domains. Agreement was poor; clinicians generally assigned lower grades. They were using different instruments, so the scores weren’t directly interchangeable. Even allowing for that, I’d want to know what my assessment was missing.

Calling a side effect “manageable” tells her something about what we can do when it happens. It doesn’t tell her whether going through it is worth the benefit. That’s the part we still have to talk about.

The benefit needs that conversation too. In my clinical conversations, avoiding recurrence has value in itself. People want to keep the cancer from coming back. They don’t have to explain that wish in terms of a survival gain before I take it seriously. And someone can feel very strongly about avoiding recurrence while also dreading what treatment might do to her day-to-day life.

Her age won’t tell me how she’ll weigh those things. A 2026 study by Dhakal and colleagues found younger participants more willing to accept more effective treatment despite significant pain. But the age-and-sex groups didn’t differ significantly in how much extra life they required to accept harsh treatment. Even within that study, the answer depended on what people were being asked to trade.

Postmus and colleagues found another split in a survey of people with multiple myeloma. Some put more weight on reducing chronic mild-to-moderate toxicity than severe or life-threatening toxicity. Others put them the other way round. The answers depended on the choices and ranges offered, and they don’t tell us what a woman with breast cancer would choose. They do make me wary of assuming that the way I rank harms is the way a patient will rank them.

So what do I tell her?

I owe her a recommendation. She shouldn’t have to read all those papers and settle the argument herself. I need to explain which options make clinical sense, why I lean towards one, and where her priorities might change my mind. Writing “risks and benefits discussed” in the chart doesn’t tell anyone what we actually worked through.

It would be easier to make the call myself. I could give her a plan, explain the side effects and get on with it. But if she understands the evidence and would make a different choice, I need to leave room for that. Finding out takes time. I think it’s part of the job.

It’s also part of why I’m building Kesis applications. Kesis Clinical is in clinical testing to help clinicians review evidence, assumptions, benefits and tradeoffs. I want the calculations behind a decision like this available for us to examine, including where they depend on assumptions and where we’re still unsure. We need to test how much that helps the conversation and the decision. A number on a screen won’t tell her what she ought to be willing to go through.

There’s a long road ahead for the woman in this hypothetical: local treatment, years of endocrine therapy, and consideration of other treatments appropriate to her risk. I want her to know which parts we expect to help most. When treatment gets difficult, I want us to help her through it and be willing to rethink the plan, so she has the best chance of completing the parts that offer her the most value.

She might end up choosing chemotherapy. She might choose an approach centred on hormonal therapy. We might talk it all through and still disagree about what I’d recommend. But she’d understand why I recommended it, and I’d understand what mattered to her. That’s why I couldn’t see it as a no-brainer.