Variolation Test Design
Okay, what the variolation concept needs most now is a trial/test/experiment ASAP. So to help get the ball rolling, let me sketch a tentative plan. I’m NOT saying this plan is now good enough. I’m saying let’s talk together about how to make it better. (Not so interested here in those ever popular “this can never work” comments.)
As with most projects, the obvious first top issue is staffing, especially leaders. This needs leaders who not only have the ability and expertise to execute it, but who can also inspire confidence in its other staff, subjects, patrons, sponsors, and audiences. (The most I’ve ever led is an assistant, so alas I don’t seem a good candidate.) The main point here is to inspire audiences to action, and that won’t happen if audiences don’t believe the project’s purported results, nor if they find its people too odious to associate with.
So the main purpose of this post is to try to attract participants, especially leaders, to pick up this ball and run with it. I’ll run with you, but I can’t run it by myself. When someone makes a good suggestion, such as in the comments, I’m likely to edit this post to include it. You are warned.
Innovation – Time is short here. And it will no doubt be hard to attack funds, staff, subjects, and needed permissions. So we seek the smallest simplest fastest least-risky test plan that could plausibly inspire further action, including bigger follow-on trials. This is not the time to invent whole new procedures; this is a time to copy and minimally-modify successful existing procedures.
Dimensions – Unless we have a pretty good idea of the ideal procedure, we can’t just directly test that against a control group. Also, it seems hard to set up a control group to be infected fast “naturally”. But in a small experiment we can’t explore very many dimensions either. An obvious compromise is to just look at variation on a single axis, or at most two. Make a plausible standard choice on all the other dimensions, and then see how outcomes depend on that one variable (or two). And an obvious candidate for that one variable is dose, though delivery method seems a strong contender. (Virus strain would be third candidate dimension if we had evidence suggesting strain mortality differences.) One hundred subjects could be okay for this, and we might even make do with twenty if necessary.
Subjects – The more homogenous the subject pool, the better chance to see dose and other effects clearly, but the less well the results generalize to larger populations. We’d prefer to focus on subjects who have the least chance of getting hurt by infection. But to let us compare our results to others, we need our subject selection criteria to correspond to variables often used to estimate hazards in other populations. Selecting by age and comorbidity may be sufficient. Selecting the young and healthy also puts a priority on finding outcome measures that can distinguish well between low levels of symptom severity.
Outcomes – In a small trial, we just can’t use death as the outcome measure; not enough will die. So we need good ways to measure final symptom severity. Viral load sounds good, but how expensive is that? Count coughs? Breathing strength (vs. previous)? Chest x-ray? CT scan? Ultrasound? Temperature? Functional assessment?
Recovery – We also have to decide how long to wait in the disease progress of each subject until we report preliminary results. When are they recovered? Of course we should continue to monitor subjects after this point to see if further developments change our initial results. A week after symptoms end? A low virus load? Fail a covid test?
Dosing – If dosing is the main variable, then we need a very standard way to dose subjects. An obvious candidate is to use water dilution. For example, take N face masks from a sick (high risk but low symptoms) covid19 patient, swirl them in a big tub of (clean!) water, add whatever non-toxic materials viruses like, mix it all well, and then freeze it. (Be careful not to let the virus concentrate in the last pockets to freeze.) Frozen material probably changes little over the period of the study. Use that same batch for the entire study. Chip off parts to dose subjects, and use volume to measure dose. Thaw and add water at the last minute to produce very small doses. Save some of the batch for studies that compare it with future batches.
Spacing – At one extreme we could just have three different doses, with a third of subjects getting each. At another extreme, each subject could get a different dose. This latter approach gives the most data to estimate ideal dosage, but it also makes the data harder to analyze. Whether we have just three doses or a wide range, we need good ways to pick the ends of the dose range.
Adaption – We could either just randomly assign subjects doses from our chosen set, or use an adaptive learning process to respond to previous results, searching for the lowest doses that give a sufficient high chance of infection. An interesting option is to start everyone on a low dose, and then escalate their dose until infection is achieved. The longer we must wait to verify a lack of infection, the longer this approach would take.
Delivery – Dosing via liquid volume fits with several ways to deliver this dose to a subject. Subjects could drink the water, drip it on their eyes, inject it, use a nasal spray, squirt it into their rectum, or swallow a pill. We want a delivery method that we guess will produce a less severe infection, but is also easy and cheap for others to reliably apply later. If we have enough subjects, we might also vary delivery along with dose in an initial trial.
Infection – When we use the lowest doses, subjects are likely to fail to get infected at all. We’d like a fast reliable way to check this, so that if we waited long enough and saw no infection, we could try them again with a higher dose. Their results will be available a bit later than others, but are still valuable. So we’ll need access to fast-enough reliable-enough covid tests, and maybe ~50% more of them than we have subjects. We need even more tests if we use them to measure recovery.
Jurisdiction – It is a big world, so in principle one could shop around for the place with the most favorable legal liability, professional licensing rules, and other regulatory treatment. Alas, this choice interacts strongly with choices of patrons, sponsors, subjects, and staff. Does a medical cruise ship help?
Location – Compared to many other medical trials, in this one there is a priority on isolating subjects so they don’t infect others. Can we trust subjects told to stay isolated at home to do so? Or is it better to rent a hotel, dorms, or other place to better enforce isolation? This place doesn’t have to be especially medical, but we need to make frequent medical measurements, and we’d need subjects to have access to advanced medical care for worse case scenarios. Do organizations who provide such care need to approve and join this project?
Synergies – There may be big synergies from combining this trial with others, such as trials of particular anti-virals or vaccine candidates. As long as we are doing to infect people anyway, first give them a vaccine candidate to see if that helps. But it may take longer to find a willing trial partner and to coordinate with them, and the data analysis may get more complex.
Okay, that’s what I have so far. If you mention issues I should have included, I’ll add them.