claim: 67/arm detects only 14.5-19.4 point differences (two-proportion, alpha=0.05 two-sided, power=0.80)
 baseline p2  simple MDE pp  two-term MDE pp
        0.05           10.5             16.2
        0.10           14.5             19.0
        0.15           17.3             20.9
        0.20           19.4             22.2
        0.25           21.0             23.1
        0.30           22.2             23.7
        0.35           23.1             23.9
        0.40           23.7             24.0
        0.45           24.1             23.8
        0.50           24.2             23.3

A) simplified formula reproduces the cited 14.5-19.4pp band at baseline rates 0.10-0.20 -> matches the claim exactly (10-20% event rate).
B) fuller two-term calculation gives MDE ~16-24pp across baselines 0.05-0.50 — the claim is at the optimistic edge but correct in order of magnitude.
VERDICT: CONFIRMED WITH REFINEMENT. Under the simplified formula the band is exact (baselines 10-20%); under the two-term formula the true MDE is slightly larger (~16-24pp). Substantive conclusion unchanged: a 67/arm cohort can only detect very large absolute differences and is underpowered for moderate effects.