Research methods · Study designs

Placebo, blinding & controls

Run this experiment in your head. Take a hundred people with chronic back pain, give them a pill that is pure sugar, and tell them it's a promising new analgesic. A few weeks later, ask how they feel. A solid chunk will report real, measurable relief — not because they're lying, but because expectation, attention, and the simple passage of time genuinely move the needle on pain. That's the placebo effect, and it is not a trick. It's a real change in a real outcome, caused by everything except the active ingredient. Which means "people on the treatment got better" is, by itself, almost worthless. People on nothing get better too.

Why the treated group improves even when nothing works

Two forces inflate the treatment arm before any active ingredient does a thing. The first is the placebo effect: believing you're being treated changes subjective experience, and it's strongest for outcomes the patient reports themselves — pain, energy, mood, sleep quality, nausea. The second is regression to the mean: people sign up for a trial when they feel their worst, and "worst" tends to drift back toward normal on its own. Add ordinary natural recovery, and a do-nothing pill posts an impressive-looking improvement.

That's why blinding matters. In a single-blind trial the patient doesn't know which arm they're in, so expectation lands equally on both. In a double-blind trial neither the patient nor the staff assessing them knows — which also stops the people scoring "energy" or "pain" from unconsciously nudging the numbers toward the result they're hoping for.

Work one, then finish one

Worked: A stress-supplement brand runs an actual trial — good for them — and reports that the treatment group's stress score improved 30%. Impressive, until you read the control line: the placebo arm improved 25%. Subtract, and the supplement's real effect is 5 points, not 30. The other 25 came from expectation, from being looked after, and from regression to the mean — people enroll when stress is peaking, and peaks subside on their own. The honest claim isn't "cut stress 30%." It's "beat a sugar pill by 5 points," and even that needs a check on whether 5 points is statistically real or just noise.

Your turn: An "energy" supplement advertises that 78% of users felt more energetic after eight weeks — unblinded, no placebo arm, energy self-reported. How much of that 78% can you attribute to the active ingredient? (None of it, with confidence — expectation, the attention of being in a program, and regression to the mean could each produce the entire result. With no placebo arm there's no gap to measure, so the number is uninterpretable.)

Why this matters

Almost every glowing supplement story you meet — the collagen that "fixed" someone's joints, the adaptogen that "transformed" their sleep, the Reddit thread swearing by a peptide — is an uncontrolled treatment arm with no placebo to subtract. The person genuinely felt better. That's not in dispute. What's missing is the other arm: how they'd have felt taking a convincing blank. Before you spend money on a subjective benefit like energy, calm, focus, or pain relief, ask for the gap. If a product can only show within-group improvement, or no comparison at all, it hasn't shown that it did anything — only that people who took it changed, which they would have done anyway. The missing placebo arm is precisely why the anecdote tells you almost nothing.