Quote:“The new protocol for Severely Ill COVID-19 patients is for 342 patients, double blinded with 2:1 ratio. Patients enrolled in this trial are expected to be administered leronlimab for two weeks with the primary endpoint being the mortality rate at 14 days”
Some might be asking what is behind these numbers, why 2:1 ratio ?? Why 342 patients ??
First of all, trials are kept to a minimum of patients for expedience and economic reasons (well, not always). To ensure that neither too few nor too many patients are included in the study, the sample size is planned in advance. To calculate how many patients will be needed there are two main factors:
The level of significance: This is the probability of obtaining a statistically significant test result, even in the case that there are no differences. This is normally taken as 2.5% for so called one-tailed tests or 5% for two-tailed tests.
The statistical power: this is the probability of identifying a real difference with the statistical test. Normally taken as 80% or 90%.
Now, why a 2:1 ratio? Normal trials are a 1:1 ratio and a 2:1 complicates analysis and extends the number of patients needed. There are several possible reasons for this:
- Patients are more likely to join two arm trials with placebo when their odds of receiving active treatment are greater.
- A high drop-out rate is expected in one arm (allocating more patients to the arm with a high drop-out rate allows greater power for a “per-protocol” analysis.
- Gathering additional safety information: A larger sample size in the active group gives more power to detect adverse events
- Learning curves: Some new technologies have a learning curve. Allocating more patients to the new technology reduces the effect of the learning curve on the final trial result
- Humanitarian reasons: If the product is thought to help patients more will be exposed to the beneficial drug than the placebo arm (I think this is the reason in this case).
The good news is that these trials (unequal allocation ratios) are normally aimed at vindicating treatments (“confirmatory trials”) !!!. That is, the philosophy is to confirm something.
Now, why 342 patients? First of all, with 2:1 ratio 12% more patients are needed. This would indicate a 1:1 design for 306 patients (is possible to work with a 2:1 but complicates things unnecessarily). I will attempt to explain the number simply, but unfortunately, will need to use some mathematical jargon. Just go to the conclusion if you don’t like math (nothing wrong with it
) If we used a two-sided design with a power of 90% and 5% significance. The question is: how many people will be needed? To answer we need to know what we want to obtain and whether the information we have already is firm or not. In simple words, let me " affirm " the following for the sake of explanation (I am making up the numbers): In USA the death rate for COVID patients going into ventilators is 80%. We (FDA) will approve a drug that reduces this rate to 70%. Or, I could say: In the USA the rate of death of persons going into ventilators is in average 80% with a normal distribution and a standard deviation of so and so, I will approve the drug only if the mean of the resulting death rate is such with so and so deviation. This will produce a different calculation for the patients needed (roughly double).
Not knowing if FDA already has a solid number (for example China and other countries statistics, or USA so far) of just assumed a distribution. Let’s work with the latter (distribution is known, not the number.) We have then from a formula (or tables) that for approximately 305 patients the difference of means is 0.15 (266 from 0.9:0.75, 322 for 0.85:0.7, 350 for 0.8:0.65).
For those interested a very useful document is: Sample Size Tables for Clinical Studies David Machin Michael Campbell. Can download PDF (warning, don’t download the WinZip Driver but use the green bar that say “Download PDF” down below).
https://epdf.pub/sample-size-tables-for-clini...udies.html
Conclusion : If my assumptions are correct (5% significance and 0.9 power), FDA is expecting that Leronimab will reduce the death rate in a factor of approximately 0.15 for approval.
I am making lots of assumptions here like type of statistical distribution and parameters and the results are just approximations for illustrative purposes for what to expect as we start to see the results of Leronlimab coming.
14
