The Problem with Washout Periods

What is the "default" diet?

dietary washout

A washout period, also called a run-in-period, is common in diet studies. It’s a diet that participants are put on before or between the actual experiments.

Washout to homogenize participants

It helps because each participant comes in eating a random diet, and that could influence what our experimental diet does to them. For example, if you’re studying a keto diet, you would expect different results from participants who are currently eating a high-carb diet vs. those who maybe are already doing a relatively low-carb or even keto diet.

If you blindly put all of them on a keto diet, you might find that some have adaptation issues and some don’t. Some might experience improvements in glucose control, some don’t. If you were to find out that some of them were already doing keto, that could help explain this.

A washout or run-in diet allows you to somewhat control this by placing everyone on the same diet. This could take the form of just advising people to eat a certain way, or you could provide them with pre-packaged meals for the given period.

Of course, if you’re doing n=1 experiments on yourself, it’s easy: just eat a certain way :)

Washout between experiments

We know that many dietary effects and processes take quite some time to appear, so we want to allow ample time for the participants’ bodies to “reset” between experiments.

For example, if you put someone on a ketogenic diet it might take a few days for fat adaptation to revert and blood ketones to go down again.

If you were to put somebody on a 30 day keto diet, and then immediately a 30 day high-carb diet, they will still be in ketosis for parts of the first few days of the latter. You’d certainly still have fatty acids in the blood from the last keto dinner on day 1 of the HCLF diet.

A washout period between experiments helps neutralize this problem.

The difficult problem of “Neutrality”

But here we start getting into issues. What is a “neutral” washout period? You could argue that a non-ketogenic diet is “neutral” when it comes to ketosis.

But as many people going on HCLF diets experience, including yours truly on the rice diet and the honey diet, you also need to adapt to high-carb diets for days to weeks.

In that case, do you consider a low-carb, low-fiber, ketogenic diet the “neutral” one?

Do you just pick a middle-ground diet? But is there even an objective, absolute middle ground?

Is it 33% carbs, 33% fat, 33% protein? That would be considered extremely high protein and very uncommon if seen from the ancestral perspective of evolutionary diets we know of.

And what about sodium content? How many seed oils should there be in the “neutral” diet? How much fiber?

Should we just put everybody on the Standard American Diet to be “neutral” when we know it’s pretty bad for almost everybody?

What about a fast? Surely nothing is a more neutral diet than not eating at all? But apart from the difficulty for many people, fasting is extremely low-fiber and extremely ketogenic - not exactly neutral if we’re going to study either of those or their effects.

In short, I believe this problem cannot be solved. There is no objectively “neutral” diet. Pretty much all diets require some adaptation if you’re coming from enough of a dietary “distance” and are considered normal by some people, and weird/strange by other people.

Why does washout matter?

We could just throw up our hands and exclaim that it’s impossible to find an optimal washout period for every conceivable dietary experiment.

But it would be pretty bad for finding out anything useful in our experiments.

And it is, in fact, pretty bad. If you look at most diet studies, if they have any washout periods, they’re horrendously designed.

Because fat loss works so slowly in many cases, even a moderate effect like bloating or water weight gain can completely mask any true fat loss/gain effect a diet might have.

For example, I have gained 4.4lbs of lean mass in 24h by DEXA before. If whatever caused this change (high protein intake, in my case) isn’t washed out, it’ll probably significantly skew any results of a diet study.

If you put me on a diet and I lost 4lbs of pure fat in a month, it would be completely negated in terms of weight gain by the one 24h period.

Ok, but what if we measured body fat? Then that would still be significantly skewed. If you put me on a diet and I lost 0 fat, but my lean mass went up by 4.4lbs due to this non-washed out change, that would look like about half a percent less in body fat %.

Now that might not sound like much, but most diet studies barely find 4lbs fat lost in A YEAR of dieting.

In one study, semaglutide (which is BETTER than ozempic) produced 7lbs of fat loss in 52 weeks.

This is considered the 2nd generation “miracle drug” because most diet studies do much worse. Yet I could skew this by over 60% in just 24h of slight dietary intake changes.

By the way, the people in that study also lost on average 5lbs of lean mass, but they still count this as a victory because body fat % slightly decreased, lol. Was the 5lbs lean mass lost mostly bloat, or water weight, or muscle, or other required tissue? We can’t know since they didn’t do a proper washout.

In short, fat loss is hard enough to study without any of these effects. Unless your effect size is insanely good (like losing 20lbs in the first month, as I did when I started ex150) it is going to be completely overshadowed by the non-washed out side effects.

What to do if no neutral ground exists?

I struggled with this myself in the beginning of experimental journey: I noticed quickly that my high-protein refeeds would cause insane weight rebounds, but those would come off again within days.

The conclusion was that these weren’t changes in body fat, but some sort of temporary weight, “water weight” in short (although it’s not all water).

At this point I’m pretty good at estimating what sort of water weight change a given diet will produce. For example, I predicted & observed that going from ex150 on the Honey Diet would increase my weight by ~5-7lbs very rapidly, but that it would plateau there.

That was after observing that a pure rice diet made me “keep on” the water weight from a protein refeed, when it would’ve normally come off on ex150.

It is maybe not possible to find an objectively neutral washout period for ALL diets, but you can predict & do one depending on your “departure diet” and your “destination diet.”

For example:

1. If I’m currently on ex150 and want to try ex500, I don’t have to wash out much. The diets are the same except for the amount of ground beef (150g vs. 500g). I’m already low-fiber and ketogenic. No washout period needed.

2. If I’m currently on ex150 and am planning another HCLFLP diet like a pure rice or bread diet, I should expect a certain plateau effect of water weight. If I wanted to “wash out” this plateau effect, I could do another sort of high-carb diet in the meantime to preempt the plateau. Personally I don’t think this is very helpful in n=1 experiments; it’ll just delay the start of the experiment via a known effect; we could’ve just used the washout period for the experiment itself.

3. If I’m currently on a HCLFLP diet (e.g. rice diet) and going back on an ex150-style low-fiber ketogenic diet, I would expect to lose a bunch of water weight quickly in the beginning. I can either just accept this and take it into account when reviewing the experiment, or I can try a washout period of e.g. fasting for 3 days.

Personally I’ve used 2 day dry-fasting periods for washout recently, because it provides a nice reset from the routine and is just about the quickest way to washout carb or protein water weight I’ve seen. It’ll also clean out your digestive system and all the bloat from carb diets.

Practicalities

For larger n=many studies, this is obviously somewhat problematic. For example you might not know what each of the participants are eating before you put them on your diet. It might not be feasible to design a washout period for each departure diet.

Another problem is that the washout period itself might be causing effects. For example, if your experiment intends to study a ketogenic diet and you wash your participants out with another ketogenic diet, your experiment began (in some sense) a few days early.

That might be fine if you’re studying something that is supposed to be happening from ketogenic diet B but not ketogenic diet A, but 1. that might not be the case or 2. you might not even know that, which is why you’re trying to study it.

As n=1 or n=small experimenters we have the advantage that we can treat each individual specifically and make mental corrections.

For example, when I ran the n=10 trial of ex150 I quickly noticed the pattern of water weight loss people were experiencing. It was quite predictable and I included it in every participant’s report.

With larger sample sizes that can’t all be interpreted individually, one could use a technique like discarding any unreasonably rapid weight loss in the first 3-10 days. Once a plateau has been reached, what follows could be counted as “real” fat loss.

This is how I like to judge diets in myself, but it does require experience with both departure & destination diets, and a lot more attention than just subtracting 2 numbers per participant.