Module 3 · Relative vs Absolute

Same drug. Two headlines. Both true.

A health magazine runs the story two ways on the same day:

One sounds like a breakthrough. The other sounds like barely anything. Yet they describe the identical trial — not a rounding difference, not a spin, the very same result.

Two frames, one truth

Take the trial from last lesson: risk fell from 2% (control) to 1% (treatment). There are two honest ways to say what happened.

Both come from the same two numbers. The relative version always sounds larger, because it's a proportion of a proportion — it floats free of how common the event actually was. The absolute version stays anchored to reality: how many people, per hundred, are actually spared.

Watch how differently they land.

Flip the frame

Here is one fixed result — risk falling from 2% to 1%. Don't change the data. Just flip how it's reported, and watch the same number transform.

Try this: flip it back and forth a few times, watching only the headline change while the underlying data sits still.

Why the frame changes everything

The reason relative figures are so seductive is that they hide the baseline. "50% reduction" sounds equally impressive whether the risk went from 80% to 40% or from 0.002% to 0.001% — but those are wildly different drugs.

Absolute figures refuse to hide it. "20 in 100 fewer" and "1 in 100,000 fewer" are both honest, and they sound as different as they truly are.

A relative number on its own is unmoored — it tells you the shape of the effect but not its size. To know the size, you need the absolute figure, and to compute that, you need the one number marketing most often omits: the baseline risk. Let's see exactly how much it matters.

The baseline is everything

Here's the proof. We'll hold the relative reduction fixed at an impressive-sounding 50%. All you'll change is the baseline risk — how common the event was to begin with. Watch the actual benefit.

Try this: slide the baseline from low to high. Watch "50% reduction" stay frozen while the NNT swings from hundreds down to a handful.

The asymmetry trick

Now the move you'll see again and again, once you know to look. The framing isn't chosen at random — it's chosen to flatter:

• The drug's benefit is reported relatively — "cuts risk by 50%!" — to sound as large as possible.
• The drug's side effects are reported absolutely — "only 2 in 1,000 had a serious reaction" — to sound as small as possible.

Same document, two different frames, each picked to push the same conclusion: take the drug. It's not lying — every number is true. It's selective framing, and it's everywhere in promotional material, patient leaflets, even press releases of trials.

The tell is inconsistency. When benefits and harms are described in different frames, someone is steering you. Insist on seeing both in the same frame before you compare them.

Spot the framing

Read each claim as an assessor. Is it a complete, honest presentation — or is the framing doing the persuading?

The reflex to build

By now the habit should be forming. Whenever you meet an effect, run three quick questions:

• What's the baseline risk? Without it, no relative figure can be judged.
• What's the absolute change (ARR / NNT)? This is the real, human-scale benefit.
• Are benefits and harms in the same frame? If not, re-express them until they are.

A relative figure alone is never enough to support a decision. The moment you see one, your reflex should be to ask for the baseline and the absolute number — because that's where the actual value of the drug is hiding.

Why this matters for HTA

This is one of the most practical instincts in the whole appraisal toolkit, because relative framing is the default language of every submission's headline.

• A submission leading with a relative reduction ("45% lower risk") is incomplete until you've found the baseline and the absolute figure. Demand them.
• Payers pay for absolute benefit, not ratios. A 50% relative reduction is worthless to budget holders until it becomes "X events prevented per 1,000 treated" — that's what the money buys, and what feeds the cost-per-event and, later, the cost-per-QALY.
• Watch for mismatched frames — benefit relative, harm absolute — in the same dossier. Re-express both consistently before weighing them.
• A small baseline risk quietly shrinks every relative claim — a dramatic relative reduction on a rare event may prevent almost nothing in practice.

Manufacturers speak in relative terms because relative terms sound large. Your job is to translate every claim into the absolute, baseline-anchored figure — because that, not the ratio, is what a health system actually pays for.

Relative vs absolute, in one breath

• The same result has a relative frame (RRR, e.g. "50% lower") and an absolute frame (ARR/RD, e.g. "1 in 100 fewer") — both true, wildly different in impact.
• Relative figures hide the baseline and always sound larger; absolute figures stay anchored to how many people are really affected.
• The same relative reduction can be a huge or trivial benefit depending entirely on the baseline risk.
• The classic trick: benefits framed relatively (to look big), harms framed absolutely (to look small) — true, but steered.
• Always ask for the baseline and the absolute figure; payers fund absolute benefit, not ratios.

"Cuts risk by 50%" is not a result until you know 50% of what. The honest number is always the absolute one — and it's the one most often left out.

You've now seen how the frame around a number can mislead. Next, something subtler: a measure that misleads not through framing but through its own arithmetic — the odds ratio. It looks like relative risk, is often reported as if it were, and quietly exaggerates the more common the event becomes. It's the last great trap in reading effect sizes.