...not sure where you came up with those p-values, but your m/m number is too high by a factor of at least 10.

The null hypothesis in clinical trials is that the tested drug gives no benefit. In your example, you say that 19.6% of placebo patients deteriorate and no leronlimab patients do. That would mean 19.6%/3 = 0.065 of all patients deteriorate. In that case, it would be VERY unlikely that zero leronlimab patients deteriorated while 4 placebo patients did. Formally, you can calculate the probability that, under the null hypothesis, you'd see 4 more placebo patients than leronlimab patients deteriorate, which would give a p-value of 0.0032. BUT even this is too high, because the clinical score is ordinal, which carries more information than a simple binomial, which would make for an even smaller p-value.

Your s/c example doesn't match the primary outcome metric, which is mortality, not progression. Also, assuming no "progression" in the leronlimab arm is extremely optimistic given that there were very non-trivial mortality rates in the EINDs.