Experimental Event Rate & Control Event Rate

The Control Event Rate (CER) is the proportion who suffer the outcome in the control arm; the Experimental Event Rate (EER) is the proportion who suffer it in the treatment arm. Given just these two numbers you can derive every common effect measure: ARR = CER − EER, RR = EER ÷ CER, RRR = (CER − EER) ÷ CER, and NNT = 1 ÷ ARR.

100 control-arm patients; the highlighted 16 are the CER — those who died on placebo. The experimental arm runs the same picture but with only ~14–15 highlighted (the EER). The whole effect of the drug is the thin gap between the two rates — here just over one figure in a hundred.

CER sets the baseline — how bad things are without treatment. EER tells you how bad they are with it. Their difference is the absolute benefit (ARR); their ratio is the relative effect (RR). A big-sounding RR can sit on a tiny absolute gap, which is exactly why you read both rates, not just the relative headline.

EER and CER are the foundation every effect measure stands on. Quote only relative figures (RR, RRR) and the reader literally cannot judge clinical importance — the same 9% relative reduction is life-changing when CER is 16% and meaningless when CER is 0.2%. Reporting both event rates is what lets the exam (and the bedside) tell a real benefit from a statistical mirage.

• ARR = CER − EER · RR = EER ÷ CER
• RRR = (CER − EER) ÷ CER · NNT = 1 ÷ ARR
• Always report the absolute rates, not just the relative measure

CRASH-2 (Lancet 2010) — 20,211 bleeding trauma patients, tranexamic acid vs placebo. 28-day all-cause mortality: CER (placebo) = 16.0% (1613/10060) and EER (TXA) = 14.5% (1463/10067). From those two rates alone: ARR = 1.5%, RR = 0.91, RRR ≈ 9%, NNT ≈ 67 — every measure falls out of the two event rates. Benefit is time-critical: TXA given ≤3 h reduces death; given >3 h the subgroup had more bleeding deaths — so these pooled rates hide an important interaction with time-to-treatment.

• Quoting relative measures (RR / RRR) with no absolute event rates — the reader can’t gauge clinical importance or back out the NNT.
• Ignoring the baseline (control) risk — the same RR means very different things at high vs low CER.
• Mixing intention-to-treat and per-protocol event rates between arms, so EER and CER aren’t measured on comparable populations.