From randomization to tissue sampling
The strongest result of the AER002 trial is its prespecified comparison. 36 people with Long COVID were randomized under double blinding: 24 received a single antibody infusion and 12 received placebo. In the peer-reviewed phase 2a paper, the day-90 physical-health endpoint did not significantly favor the antibody. That endpoint was chosen in advance. An interesting subgroup identified later cannot replace it. The tested regimen did not establish an overall benefit in this enrolled population. The sample was powered to detect only a large difference; smaller effects could not be definitively excluded.[1]
The sample narrows further between recruitment and tissue measurement. Of 460 people expressing interest, 43 consented and 36 were randomized. Intestinal biopsy was optional: 17 provided samples and only one baseline biopsy contained detectable viral RNA. 10 joined the PET substudy. Complete participant follow-up is a strength, while the small biological substudies impose a separate limit. Complete follow-up does not mean the proposed target was measured throughout relevant tissues.[1]
Enrollment did not begin with people selected for confirmed viral persistence. A validated marker of the proposed biological target was not required. The design therefore answers two questions with different precision: benefit was not demonstrated in the enrolled group, while an outcome in a future marker-selected group remains undetermined. A straightforward alternative is that the drug is ineffective. The absence of target-based selection is not evidence in favor of efficacy.[1]
How much an exploratory response can carry
The exploratory findings need to be read within that boundary. Among participants with lower baseline antibody levels, greater drug exposure was associated with self-reported improvement. The joint exposure-and-baseline-antibody analysis was post hoc. It can generate questions about selection and dosing for another trial. Its association offers weaker evidence than the prespecified randomized comparison; testing many relationships in a small group creates opportunities for chance patterns.[1]
Baseline antibody levels were imbalanced between the two arms. Excluding four reinfected participants did not change the primary result, providing a concrete robustness check on the negative finding. Severely affected people unable to attend visits were not represented, and volunteers for biopsy were not a randomized tissue-sampling subset. The exploratory response association could reflect biological variation, selection, baseline differences or measurement variability. A design capable of distinguishing those possibilities must be specified in advance.[1]
The research implication extends beyond searching the same data for a more attractive subgroup. A persistence-selected trial would need to define its enrollment marker, tissue-sampling plan, drug-exposure assessment and primary clinical endpoint together in advance. Such a design could ask whether a clinical difference emerges among people in whom the target is actually measured. The present study retains its contribution: it did not support a general benefit claim for the tested regimen and made the missing measurements in a narrower mechanistic question explicit.[1]