The drug that's great and still feels wrong.

Value has more than one dimension.

Cost-per-QALY compresses everything a technology is worth into a single ratio: money per unit of health. That compression is its power (one number, comparable across everything) and also its blindness. Because the worth of funding a drug genuinely has more than one dimension, and the QALY captures only some of them.

Think about what actually moves people when they judge whether a drug should be funded:

• Health gain: how much extra healthy life it delivers. (This one the QALY does capture.)
• Severity: how sick the patients are to begin with. Most people feel a year of health matters more for the desperately ill than the mildly unwell, but a QALY is a QALY regardless.
• Unmet need: whether any alternative exists. A drug that's the only option feels weightier than one more entry in a crowded field.
• Equity: whether funding it helps close, or widen, gaps between groups.
• Innovation: whether it opens a new mechanism that might lead somewhere.
• Impact on carers, uncertainty, wider social value, and more.

Cost-per-QALY sees the first of these clearly and the rest barely or not at all. So when a decision feels wrong on a pure ICER basis, it's usually because the ICER is silent on a dimension you care about. The orphan-drug modifiers of Module 12 were exactly this: severity and rarity, forced back into a calculation that had squeezed them out. The question is whether to keep smuggling these dimensions in through the back door, or to put them all on the table at once. That's what multi-criteria decision analysis does.

MCDA: make the dimensions explicit.

Multi-criteria decision analysis (MCDA) is, at heart, startlingly simple. If value has many dimensions, then instead of collapsing them into one ratio, list them, score them, weight them, and add them up. Four steps:

1. Criteria. Decide the dimensions of value that matter: health gain, severity, unmet need, equity, innovation, whatever the decision should reflect.
2. Scores. Rate each option on each criterion: say 0 to 100. Drug B might score low on health gain but very high on severity and unmet need.
3. Weights. Decide how much each criterion matters relative to the others. Is severity half as important as health gain? Twice? This is where the values live.
4. Aggregate. Multiply each score by its weight, sum across criteria, and you get one overall value score per option, and a ranking.

That's it. A drug that loses on cost-per-QALY can win on MCDA if it scores high on criteria that the QALY ignored and that you've chosen to weight heavily. Laid out as a grid (options down one side, criteria across the top, scores in the cells, weights along the bottom) it makes visible exactly what you value and how much. Which sounds like it's just adding structure. But it's doing something far more subversive to the "objective" ICER you started with, as the next screen shows.

The buried weights.

Here's the move that turns MCDA from a technique into an insight. You might think the choice is between an objective cost-per-QALY and a subjective MCDA full of hand-picked weights. That framing is wrong, and seeing why is the heart of this lesson.

Cost-per-QALY is already an MCDA. It's an MCDA with the weight on health gain set to everything, and the weight on every other criterion (severity, unmet need, equity, innovation) set to zero. When you say "we just fund by cost per QALY," you have not escaped weighting; you have chosen a very specific, very extreme set of weights and then stopped talking about them. The famous principle that "a QALY is a QALY, whoever receives it" sounds like neutral objectivity. It isn't. It's a value judgement: the decision to weight equity, severity, and everyone's circumstances at zero, and pure health at one. Defensible, maybe. But a choice, not a fact of nature.

This is the reframe: MCDA does not introduce subjectivity into an objective process. It exposes the subjectivity that was always there, buried inside the QALY's methodology. Cost-per-QALY doesn't lack weights: it hides them. So the real choice was never "objective ICER versus subjective MCDA." It was always "weights you can see and argue about, versus weights buried where no one can touch them." And once you know that, you can do something rather satisfying: take an MCDA and turn the buried weights back into visible dials.

Turn the weights up and down.

Here are three drugs, each scored on five dimensions of value. You set how much each dimension matters, the weights, and watch the ranking rearrange. Then try the two settings that reveal what's really going on.

Watch what the sliders proved. With every weight but health at zero, you got cost-per-QALY exactly: Drug A on top, the dying children at the bottom. Nudge severity, unmet need, and equity up from zero, and Drug B climbs, because you've finally let the things you care about count. The drugs never changed; only your weights did. This is the whole argument made mechanical: the ICER isn't the opposite of MCDA, it's a single, extreme setting of MCDA, and there is no "correct" setting waiting to be found, only the setting you can defend. The value was always in the weights. MCDA just puts your hands on the dials.

Now you.

Transparency isn't objectivity.

A warning, because MCDA's greatest strength is also its most seductive danger. MCDA looks scientific: a grid, scores out of 100, weights, a weighted sum, a final value with a decimal point. It has the visual grammar of objectivity. And it is emphatically not objective.

Every number in an MCDA came from a judgement. Where did the weight of 0.3 on innovation come from? A committee decided, and the same committee, on a different day, or a different committee, might have said 0.5, and the ranking would have moved. The scores, too, are judgements dressed as data. So MCDA doesn't convert soft judgement into hard fact. What it does, and this is genuinely valuable, is convert hidden judgement into visible judgement. The subjectivity that cost-per-QALY buried, MCDA lays on the table where it can be seen, questioned, and argued.

That distinction, transparent, not objective, is the one experts most often blur, usually to their cost. Someone builds an elaborate MCDA, gets a precise-looking score of 68.4, and defends the decision with "the model says so", as if the decimal made it true. But the 68.4 is only as solid as the weights poured into it, and those were a choice. The illusion of precision is real and dangerous: soft inputs, run through arithmetic, come out looking hard. The value of MCDA is not the number at the end. It's the conversation the method forces: naming the criteria, arguing the weights, seeing whose values are in the mix. Treat the output as a fact and you've taken the one genuinely good thing about MCDA, its honesty about values, and thrown it away.

The traps: double-counting and compensation.

Beyond mistaking transparency for objectivity, MCDA has two technical traps that catch even careful analysts.

Double-counting. The QALY already reflects some of the things you might list as separate criteria. Disease severity, for instance, is partly baked into the QALY gain: sicker patients often have more to gain. So if you use "health gain (QALYs)" and add "severity" as its own criterion, you may be counting severity twice: once inside the QALY, once beside it. The drug for severe disease then gets double credit, and the analysis is quietly rigged. Criteria must be genuinely independent, each capturing something the others don't, or the weighted sum lies. Designing a non-overlapping criteria set is harder than it looks, precisely because value dimensions bleed into each other.

Compensation. A simple weighted sum is compensatory: a very high score on one criterion can offset a very low score on another, because they just add up. Usually that's fine, but sometimes a low score should be disqualifying, not compensable. A drug that scores brilliantly on health gain and innovation but catastrophically on equity or safety might still post a high total, and the ranking would wave it through, hiding a dealbreaker inside an average. If some criteria have floors that mustn't be breached, a naïve weighted sum won't protect them; you need non-compensatory rules on top.

And underneath both traps sits the unanswerable question: whose weights? Someone must supply them (a committee, a citizens' panel, elicited public preferences) and each source carries its own claim to legitimacy and its own biases. MCDA never resolves this. What it does is make the question askable: it drags "whose values, weighted how much?" into the open, where a democracy can at least argue about it, instead of leaving it buried in a methodology no one thinks to question. That's the honest claim for MCDA. Not that it finds the right answer, but that it makes the real argument visible.

What's the strongest rebuttal?

Why this matters for HTA

MCDA is one of the most misunderstood tools in HTA: treated by fans as a scientific upgrade and by critics as unscientific noise, when it's really neither:

• Use MCDA for its transparency, never for a false objectivity. Its genuine value is forcing the criteria and weights into the open, so a decision's value judgements can be seen and argued. The moment someone defends a choice with "the MCDA score says 68.4," that value has been squandered: the number is only as sound as the weights inside it, and those were a judgement. Present MCDA as a structured argument about values, not as a machine that outputs the answer.
• Remember cost-per-QALY is a point in the same space. When you report an ICER, you are already running an MCDA with all weight on health gain. That's a legitimate choice, but name it as one, and be ready to explain why severity, equity, and unmet need are being weighted at zero, especially when the decision feels wrong (orphan drugs, end-of-life, the worst-off). The instinct that "the ICER misses something" is usually correct, and MCDA is how you say what.
• Police the traps. Before trusting any multi-criteria result, check that the criteria don't double-count what the QALY already contains, and that no disqualifying weakness is being compensated away inside a tidy weighted sum. And always ask the question MCDA can't answer but does make askable: whose weights are these, and by what legitimacy? The method organises the values debate; it never ends it.

Every HTA decision weighs many values into one verdict: the only question is whether it does so where you can see it. Cost-per-QALY answers by hiding the weights inside a ratio; MCDA answers by laying them on the table. Neither is objective, because the question was never objective. One is just honest about it.

MCDA, in one breath.

• Cost-per-QALY compresses value onto one axis (health per pound), but real value has many dimensions (severity, unmet need, equity, innovation, carer impact) that the QALY misses. The orphan-drug "modifiers" of Module 12 were these dimensions smuggled back in.
• MCDA makes them explicit: list the criteria, score each option on each, assign weights for how much each matters, and aggregate into one value score and ranking. A drug that loses on cost-per-QALY can win if it scores high on criteria you weight heavily.
• The deep point: cost-per-QALY is already an MCDA with buried weights: health gain weighted at everything, everything else at zero. "A QALY is a QALY, whoever gets it" is a value choice, not a fact. MCDA doesn't add subjectivity; it reveals the subjectivity that was always hidden. Zero out every weight but health gain and MCDA collapses back into the ICER, proof they're the same tool at different settings.
• But transparency isn't objectivity: the weights are still judgements, so a precise-looking MCDA score is only as solid as its weights. Watch the traps: double-counting (a value counted inside the QALY and as its own criterion) and compensation (a great score masking a disqualifying one), and always ask whose weights? MCDA organises the values argument; it never settles it.

There is no number that isn't a choice about what counts. Cost-per-QALY makes the choice and hides it; MCDA makes the choice and shows it. The honest question in every appraisal isn't "what does the model say?" but "whose values are in the weights, and can we defend them?"

One criterion kept forcing its way back onto the table: equity, the fairness of who gains and who loses. MCDA lets you weight it, but that only sharpens the question it can't answer: what is equity in health, exactly, and how could you ever measure it well enough to put a number on? Fairness turns out to be far deeper and stranger than a single criterion, and it's the subject of the next lesson.