Very high conflict of interest RCT with design optimized for a
null result: very low risk patients, high existing immunity, post-hoc change
to exclude patients more likely to benefit. There was no significant
difference in viral clearance among low risk patients with high viral load at
baseline. All 3 progression events occured in the control arm (one
hospitalization, two COVID-19 related rhabdomyolysis). Patients in both arms
cleared the virus quickly with a viral clearance half-life of 21.1 hours vs.
19.2 hours, which may be in part due to prior immunity.
Hypothetically, if we want to design a trial to produce a null
effect on viral clearance we can:
•Choose very low risk patients that
typically recover quickly without any treatment.
•Choose patients with high existing
levels of immunity from prior infection or vaccination.
•Make sure we do not test culture
viability and we do not distinguish between live/inactive virus.
While we do not comment on the reason for the design, this
trial does all of these and more with a post-hoc change to exclude patients
from being treated early prior to high viral load:
•Very low risk patients: median age 27,
range 18-45, all patients recovered, no major comorbidities, control arm viral
clearance half-life 0.8 days. This leaves little room for improvement,
especially with an oral treatment. Results with this population have minimal
relevance to real-world usage in patients at higher risk.
•Patients had high existing immunity
with very high vaccination levels and very high baseline antibody positive
results (notably favoring the control arm with 2.2 times as many baseline
antibody negative patients in the ivermectin arm).
•Authors do not test culture viability,
using PCR which does not distinguish between live and disabled virus.
Authors have made an additional
post-hoc change in favor of finding null results. Specifically, inclusion
criteria were changed from the pre-specified criteria
to only include
patients with very high viral load. Authors added a restriction to require PCR
Ct <25 or a specific antigen test positive within 2 minutes at baseline. This
minimizes the chance of including patients that may be caught early before
peak viral load, i.e., patients more likely to have benefit from antiviral
mechanisms of action. As shown in Figure S7, almost all patients including
control patients were enrolled at peak viral load and had declining viral load
from baseline. i.e., the selection criteria and population largely prevented
enrolling patients earlier than the point where their immune system was
already efficiently handling the virus.
Note that these methods are synergistic - for example
restricting to high viral load implies greater chance of the virus spreading
to more tissues, requiring longer treatment for oral ivermectin to reach
therapeutic levels in those tissues, thereby reducing the chance of reaching
therapeutic levels before recovery with the population of very fast recovering
Notably, authors are aware that the post-hoc change favors a
null result — in the discussion they note that results do not apply to
prophylaxis because less potent viral suppression is needed. Similarly, the
required therapeutic level may be much lower as treatment occurs earlier in
the viral cycle when viral load is lower and spread to other tissues is
Authors provide very little baseline information, however very
large differences are seen - 10, 22, and 50% antibody negative for each arm,
maximum age 31, 45, and 43. Baseline mean viral load was 1.6 times higher in
the ivermectin arm vs. control arm before log conversion.
Figure 2b shows that control patients were enrolled at peak
viral load, while ivermectin patients had peak load one day later. From Figure
S7, this is driven by a small percentage of patients, and may be related to
the 2.2x difference in baseline antibody negative results. These differences
suggest a significant randomization failure in favor of the control group.
Authors provide results for only 10 of the 40
casirivimab/imdevimab patients, which do not match the other arms in terms of
variants, have a much lower maximum age (31 vs. 45), and much lower antibody
negative at baseline (50% versus 22%/10%).
This study is excluded in the after exclusion results of meta
post-hoc change to exclude patients treated before high viral load, population very low risk, recovering quickly without treatment, high baseline immunity, 2.2x greater baseline antibody negative for the treatment arm.
Schilling et al., 7/19/2022, Randomized Controlled Trial, Thailand, preprint, median age 27.0, 38 authors, study period 30 September, 2021 - 18 April, 2022, average treatment delay 2.0 days, dosage 600μg/kg days 1-7, trial NCT05041907 (history)