Hacker News Reader

Comment by WalterBright | Hacker News Reader

But it does sound like if there's priority and fear, much time could be saved. For example, "6 months to recruit 1000 people". I suspect in this case, a week can work.

> Then you need to analyze all the data you collected, so another 3 months.

Help me understand this. You vaccinate 1000 people. In 6 months, the number of them that get sick is 10% of the rate of the unvaccinated. 0.1% have a bad reaction. How much more analysis do you need?

> the FDA takes 3 months to review and approve

How much time if that is moved to the front of the queue and all hands work on it? Given how enormously disruptive the coronavirus is, we cannot afford to not have all hands working on it.

View on HN

Let's tackle recruitment. I've seen trials recruit really fast and that took a month. It's not like someone says "hey I'd like to take part in this trial" and you give them a shot of drug. You need to screen everyone who comes in. Basic labs, then make sure they don't have some other disease that will mess up the results. That's just one part of it.

Then in terms of the trial, 6 months would be nice, but it wouldn't be ethical to actually expose them to the virus. You'd simply need to recruit a very large number of patients and see if the treatment groups has less disease than the non-treated group. With infection rates pretty low across the entire population, you'd need to recruit thousands, just to capture the handful that actually get exposed.

Then when it comes to analysis, you need to clean the data (yes, there will be errors), then you need to do a statistical analysis of not only efficacy but every possible side effect. You're probably measuring things like viral shedding, so that's another entire dataset. Oh, then you'll have missing data. Do you toss that patient's data out entirely? Or keep them? What impact does that have on your conclusion?

I could go on and on, but suffice to say it's a ton of work. And yes, you can speed up some parts, but other parts you just have to let them run their course.

If tens or hundreds of thousands of people were dying every day from the virus, might we find many volunteers willing to be intentionally exposed to the virus in order to rapidly test the efficacy of a vaccine? I imagine so.

I believe the magnitude of this pandemic will in time push us (rather violently) to think outside the box.

Even if volunteers take the vaccine, this will not significantly change the timeline. Do you really want to cut short the observation time to e.g. 1 month, then greenlight the vaccine and administer it to millions or billions of people, only to find out that after 6 months it causes an aneurysm or whatever in 10% of them?

That's a key issue.

If you're testing a drug for a serious cancer, then there is a lot of tolerance for side effects - the alternative is death, so even if it causes things like nerve damage, it's an acceptable risk.

With vaccines, you're going to treat a huge percentage of the population and they'll be healthy when you give it to them. As such, the risk-benefit analysis is much different since there is a chance for serious harm.

Start by 100 people, monitor for 1 month. If it looks promising, add 1000 people, monitor for 1 month. If it looks promising, add 10000 people, monitor for 1 month. If it looks promising, add 100000 people, monitor for 1 month. If it looks promising, add 1000000 people, monitor for 1 month.

And so on. The exact rate could be set according to the actual infection and death rates in the population. Given the current costs of curfew, we could create an entire government agency just to administer this, and still save trillions.

For inspiration, consider the Manhatten project.

> rapidly test the efficacy of a vaccine

How do you test the efficacy if you don't know who you're injecting? Maybe it works really well on a certain population, and reacts harmfully to people with some pre-existing genetic trait or condition. You wouldn't know, because you took a random set of volunteers.

If you're going to counter and say "we'd still do the pre-analysis on the volunteers"... Well, that's what takes time.

> Let's tackle recruitment. I've seen trials recruit really fast and that took a month. It's not like someone says "hey I'd like to take part in this trial" and you give them a shot of drug. You need to screen everyone who comes in. Basic labs, then make sure they don't have some other disease that will mess up the results.

You could do that time-consuming screening after they get the trial vaccine, not before. After all, I suspect it is highly unlikely that a virus vaccine would give someone a bacterial disease or a genetic disorder.

Also, before the 6 month trial is up, you're still going to have data trickling in that will give a good indication whether it's working or not.

Source: I saw the movie "Contagion" and learned that a vaccine is considered safe and effective based on one monkey not getting sick. I believe everything celebrities tell me.

Ok. So let’s say you give the vaccine and run the tests later. This person’s white blood cell count is way above normal.

Is that caused by the vaccine? Was it something else? Was it always there? You have no idea because you didn’t capture a baseline.

The challenge of clinical trials is you don’t know what to expect, and you’re measuring effects on top all the stuff that’s happening in the background with completely healthy people.

The goal is to capture as much data as possible so you can actually draw conclusions from it. Otherwise you’re left thinking “it’s probably the drug, but I’m not sure”.

> The goal is to capture as much data as possible

Normally, yes. But this isn't normally - here the goal is to get a vaccine as quickly as possible.

For example, white blood cell counts. Draw the blood, set it aside, inject the vaccine. Then get around to testing the blood. You save what, a week that way? That's a big deal.

Based on your descriptions, it seems that a lot of time can be saved by doing things concurrently rather than serially.

We had "if programmers build a plane". Maybe it's time for "if programmers run a vaccine development" video. Lol.

Walter is completely right. Under extreme circumstances (worst pandemic in 100 years) our risk thresholds should obviously shift. You can't cut corners too much and there is an inherent serial component, because if you want to vaccinate potentially hundreds of millions or even billions of people, you need to have some confidence about the risks involved. Even if you accept higher numbers of adverse vaccination reactions than you normally would. This means several months of observation you really cannot avoid. But most of the other things that take months according to refurb seem like baroque excess in the current context. Three months to analyze results and six months to recruit? Seriously? Also you can basically run any number of trials in parallel with different thresholds and then pick where you want to be on the risk/reward scale based on the situation a couple of months down the line.

I probably wouldn't take a vaccine that was validated just by several months of observation on test subjects, especially if it's a novel[1] mRNA vaccine. Maybe a year later, after release, when the potential for harm is better known.

Other people may have a different risk/reward calculation.

[1] https://www.mdpi.com/2076-393X/7/2/37/htm

I wouldn't suggest anyone being forced to take the vaccine against their will.

I'm not sure how this would contradict what I wrote, I totally agree that the observation of the patients is the one thing where you can't really speed things up much. It's inherently serial, no matter how much money and manpower you shove at it, but I don't see how this is true of many of the other stages mentioned upthread. Do you?

I wonder if the downvoters fully understand that Covid-19 is a big deal. If we don't manage to eradicate it and it becomes as common as the (various strains of) cold, we will likely take a big hit to global life expectancy. It may also evolve to become more virulent and we may yet discover various unpleasant things about it. Even if it stays exactly as it is now and we just let it run it's course and get 70-80% of world population infected, before eradicating it in say two years time with a vaccine, we will eat a huge, multi-year dip in life expectancy in the developed world.

And clearly, we are not accelerating this yet in ways that are completely safe (e.g. would we have more and better volunteers faster and more regularly if we paid them more than $100 for each visit?).

I'm not trying to contradict you. Just saying that the kind of vaccine that can be perhaps ready relatively quickly is the kind that's not very common yet, and is in the stage of "new and exciting technology", atm. I'll exercise caution.

There were news today that first subjects are getting the experimental vaccines already. I don't know if it's fast enough or not. But I doubt people will not try to speed things up, at all levels.

What can be done without bureaucracy:

"Although far behind, the United States did work hard to get a jet into the air during the war. Lockheed was engaged to build a jet fighter, using the British De Havilland Goblin engine. Speed was called for, and Lockheed gave Clarence L. “Kelly” Johnson full authority for the project. He set up his special unit, the famous “Skunk Works,” with 23 engineers and 105 shop men, and he did 96 percent of his own fabrication and all inspection. Lockheed's normal development procedure of engineering conferences was by-passed; only 700 drawings were made; sometimes parts were made from a drawing, and sometimes the reverse was done: simplicity was Johnson's keynote. The XP-80 cost only 63 percent of the XP-38 prototype, which had been made by the customary system of development. It was accepted by the AAF only 143 days after the start of development, even though the engine had not been on hand until the 132nd day. After General Electric had designed an improved Whittle-type engine, the 1-40 or J33, the XP-80A was designed around it: the United States had a jet fighter and one of the best aeronautical designs in its history by the end of 1945."

"The most telling criticism of massed-manpower engineering practices is the record of the “Skunk Works.” Starting with the XF-80, this technique has been used periodically by Lockheed's Kelly Johnson to produce the U-2, the JetStar business jet, and the A-11 (F-12). The XF-80, it will be recalled, used 23 engineers and was designed in 143 days; the U-2 also used 23 engineers and took 80 days; the JetStar used 41 engineers and took 241 days; Johnson used 135 engineers to build the A-11, which was designed to go over Mach 3 at 20 miles high. This contrasts with the 3,500 engineers used to build North American's XB-70, which went Mach 3 at 15 miles high. Although it must be acknowledged that none of these aircraft are large, the XF-80, the U-2, and the A-11 all pushed beyond the state of the art, and the repeated success of the “Skunk Works” shows that effectiveness of the system is no fluke."

"The Jetmakers" by Charles Bright (yes, my dad)

Think about it - 23 engineers and 80 days for the U2. Kelly Johnson knew how to get things done.

I accept the ingenuity of people.

Though, if 1% of your planes have issue that causes them to stop working midflight after 1 year, at worst you'll lose a few pilots and fix the issue by inspecting rest of the planes.

Backtracking and fixing issues after giving a vaccine to millions of people is a bit more problematic. Not just technically, but also in gaining the trust back, etc.

You balance the risk of the vaccine against the certainty of people dying without it.

Take your samples before giving them the vaccine, but do the actual screening later.

Is there a type of screening that requires you to do tests over a period of more than a week, because you need the results from some slow tests to order other slow tests? If you just exclude all the people that need unexpected additional tests, is that so many it will ruin the data?

>After all, I suspect it is highly unlikely that a virus vaccine would give someone a bacterial disease or a genetic disorder.

A vaccine could compromise one's immune system and indirectly cause that bacterial disease or genetic disorder. That would be its side effect too.

> I suspect it is highly unlikely that a virus vaccine would give someone a bacterial disease or a genetic disorder.

Besides allergic reactions and contamination, the oral polio vaccine can cause polio:

https://en.wikipedia.org/wiki/Polio

Low infection rates will not be your problem with this one I'm afraid.

Yes, it’s an emergency, but many parts of the process just take as long as they take. Throwing more people and money at the problem can ensure that we have many parallel solutions in the pipeline, but won’t get us there faster.

I don't see why. Can you spell out what's inherently slow apart from the observation phases?

Not who you're replying to, but I imagine bad, immediate reactions will be easy to spot, but more subtle, chronic problems would be found only after analysis. A part of this work would be figuring out what is attributable to the vaccine and what is not.

Reply on news.ycombinator.com