Would they need to have 2 placebo arms to make this an optimal study design then? One for glucose (as sodium ascorbate was diluted using a glucose solution), and another placebo group for sodium chloride?

Honestly if they had balanced the sodium via almost any mechanism, I wouldn't have had much to say about it. The difficulty with 2 placebo arms is that you have to expand your study population which I think poses an ethical issue, but otherwise yes.

Considering that saline drips a the norm for IV hydration, it might be that a saline infusion added to a saline IV line would be superfluous. Interested to get a medically qualified person's take on this.

It would not be superfluous because they are talking about nearly a gram of additional sodium in one arm vs none in the other.

They describe the carrier solution, which is a 5% glucose solution (D5W). D5W is isotonic and does not include sodium.

Glucose is a common carrier fluid, and in this particular case is given at a very low concentration (D5%). You need to infuse glucose as part of your parental nutrition regimen. This is standard practice across intensive care units, and they probably wouldn't have gotten approval for the protocol without it.

As an aside, my company [1] is working on solving this and other problems by focusing on glucose control. When you keep the blood sugar under control, sepsis develops at much lower rates and mortality decreases across the board by 20-30%.

[1] - https://idealmedtech.com

D5W is common, sure. My post makes no complaint about D5W being uncommon.

The problem here, again, is that this choice led to both sodium and ascorbate administration being different between the two arms. They think they are testing the ascorbate, but they can’t exclude the possibility that they are getting results due to sodium.

Since you used this as an opportunity to plug your company, I will say that the company’s claims about reducing mortality from glucose management seem very poorly supported. Your site only lists retrospective studies, but you will need multicenter prospective RCTs to have any credibility with such bombastic claims. Saying that “mortality decreases” is easily read as causal language, but there is no data supporting such language. If the effect size is anywhere near as large as you claim, these trials will be very easy to run and won’t need a large sample size.

There are several multicenter RCTs showing the efficacy of glucose control, but none as yet that do it well in a fully automated way. We don't have these studies listed on the website, and that's one thing we could definitely change, but they're pretty easy to find in the literature.

You're correct though, that the number of patients needed to demonstrate efficacy in a pivotal trial will be somewhere in the 100-300 range, which is something we're working up to! As with all things in medicine, these things take lots time and lots of capital, but the need is clear and for people who understand that need, there are very few who wouldn't want to incorporate a fully automated glucose control system into their unit.

Regarding mortality reduction claims, I'd encourage you to read the work of Van DeBerghe, Krinsley, Kovatchev, Hovorka, Chase, and Umpierrez, to name but a few of the academics who have run larger studies

Van DeBerghe is cited in the editorial discussing the pivotal practice-changing RCT that showed increased mortality from tight glucose control in the ICU (Finfer NEJM 2009). I think a lot of us intuitively believe that there is a mechanism of glucose control that should be helpful, but like I said, an RCT (of the specific device) will be necessary.

Absolutely! Yeah, NICE-SUGAR definitely has its flaws, the main one being that it has a lot more hypoglycemia than it should - a combination of the open loop protocol, infrequent sample time, no patient model, and static dosing protocol with only rescue bolusing. We did a series of simulation studies a few years ago comparing various state of the art ICU controllers, and while NICE-SUGAR has excellent time in range, it has poor performance in the hypo range, which has been linked to increased mortality.

The key is to get low variability and lots of time in range, but absolutely no hypoglycemia of any kind. That's what it's going to take to crack this problem. We hope our second human trial will continue to show those characteristics, and in a few years, an RCT pivotal trial.

Here's the NICE-SUGAR protocol for your reference: https://studies.thegeorgeinstitute.org/nice/docs/ALGORITHM.p...

You realize the placebo isn’t supposed to contain any active ingredient, right?

You realize that the authors think that the ascorbate molecule is their active ingredient, right?

Not really relevant. Let’s say the authors modify the study as you suggest and use sodium chloride as the active ingredient. Now let’s say no difference is observed between the placebo group and the test group. Does that mean sodium is effective or ineffective? You can’t say without having a true control group to compare against. You’re basically saying “I would be more ok with this study if it was totally useless.”

> Does that mean sodium is effective or ineffective?

This comment misidentifies what the authors see as the active ingredient. They are not trying to study sodium. They are trying to study ascorbate. If you want to study ascorbate, then also giving sodium to one group and not the other will confound your interpretation of the results, as it does here.

> Let’s say the authors modify the study as you suggest and use sodium chloride as the active ingredient

That is not, at all, my suggestion. The authors have identified the active ingredient they want to test. Their test, however, was imbalanced in other ways. I am pointing out that imbalance.

In a placebo trial, it’s very important that the placebo be statistically independent from the treatment group. You’re basically saying you would change the study so the placebo and control were not independent distributions. That would make the study worthless. I don’t know what else to say other than that you have no idea what you’re talking about. No where did the authors say they’re only studying ascorbate. They explicitly say they’re studying a treatment regime of sodium ascorbate. If you want to study only sodium, then run a RCT with only sodium chloride. Comparing sodium chloride to sodium ascorbate is not the way to study the effect of sodium.

> In a placebo trial, it’s very important that the placebo be statistically independent from the treatment group.

Statistical independence has nothing to do with our conversation, which is about designing an appropriate set of comparator groups to enable valid inference about the hypothesis.

I’m afraid I don’t have time to continue a discussion where we have to pretend that the authors - who have identified sodium as a confounder for their hypothesis - actually wanted to study sodium.

Statistic independence has everything to do with it. If your control and treatment groups are not independent, you can’t differentiate a case where both groups experience an effect vs. both groups experiencing no effect. Please explain how that would work in your experiment.