Thread by @Dr__Guess

There's a new meta-analysis out in @bmj_latest on low-carb diets on remission of #type2diabetes.

Here are my thoughts in a LONG thread. 🤓

The short version:

1) this adds nothing new 😕
2) long term RCTs on this are so flawed/heterogeneous we can't compare them meaningfully 😴
3) please stop doing meta-analyses on low-carb and glycaemia (with one exception h/t to @DylanMacKayPhD)

This MA isn't too different from previous MA - it just puts the continuous variable: glucose concentration into a dichotomous variable: remission yes/no.

The reported headline results are that low-carb is better than comparator diet at achieving remission at 6 months but no different at 12 months.

(short term yay, long term nay is the story of low-carb meta-analyses so far)

Look at the weight loss though: low carb results in greater wt loss (7.4kg difference) than comparator diet at 6 months; 0.3kg difference at 12 months.

This suggests to me (on the basis of the RCTs compiled) that weight loss is driving remission, not carb restriction *per se*.

This meta-analysis also includes a bizarre definition of remission:

"glucose concentration lower than the threshold for diabetes diagnosis WHETHER PEOPLE ARE ON WHATEVA MEDICATIONS TO **LOWER** GLUCOSE OR NOT".

LOL.

NOTE: if the "remission BUT NO GLUCOSE LOWERING MEDS" definition is used, low carb was no better than the comparator diet at 6 or 12 months.

now, do I think this means that low-carb is no better than a comparator diet at lowering glucose/getting remission of T2D?

No - IMO this type of meta-analysis tells us so little about whether low-carb can/can't help lower glucose/achieve remission that we should stop doing them.

Let's look at some of the individual studies:

Yamada (2014) included people on insulin, was weight neutral, carbs 70-130g/day, protein 25%kcal

Control was high carb matched for energy intake, protein 15%kcal.

No difference in weight loss between groups

Morris (2019) excluded people on insulin, intervention was 800kcal intensive weight loss.

Control was brief written advice and a pep talk from a nurse with no follow-up.

Difference in weight loss between groups =7.5kg

Tay (2014) excluded people on insulin, intervention was <50g carb a day, protein was 28%kcal, weight loss ~12kg.

Control was high carb, protein was 17%kcal but weight loss was also 12kg

Guldbrand (2014) included people on insulin, intervention was 20%kcal carb, 30%kcal protein.

Control, same kcal deficit as intervention, 55-60% kcal carb; 10-15%kcal protein.

Weight loss, amount of carbs/protein, meds (and type!)...

These are just "some" of the critical confounding factors which should preclude comparisons between low-carb/"comparator diet" and they have not been accounted for.

I mean OMG let's look at the "comparator diets".....

11/21 people in the "low carb LOST 10KG group" got remission.

0/12 in the "here's a leaflet about weight loss, lost 2kg not low carb group" got remission.

Gee low carb looks GOOOOOOOOOOOODDDD

In Tay 2014:

36/46 in the low carb group got remission. They lost ~12kg.

30/47 in the high-carb group got remission. They lost ~12kg.

Wait, so low-carb ain't all that.......?

Protein:

Protein potently increases insulin secretion even in long-standing T2D.

IMO low carb NEEDS to be high protein to lower glucose [independent of weight loss].

Not accounting for a ~doubling of protein intake in assessing what low carb does to glycaemia is just CRAZY.

Now meds:

Low carb might reduce post prandial glucose more in someone on metformin vs someone on a sulphonylurea so the overall effect of low-carb on HbA1c would then differ between these two patients.

(If you're still with me here, I salute your endurance)

Most of these RCTs just lump all meds together (ie were they on "meds" or "not on meds'). And then these meta-analyses lump these RCTs together....

At best, an RCT might break down whether someone was on oral hypoglycaemics or on insulin.

But "on insulin" is not just one homogenous group.

(If you're STILL with me go and get yourself a nice alcoholic beverage or snack).

Basal is primarily for fasting glucose homeostasis. Bolus is for post-prandial.

Low carb (with/without added protein!) might primarily lower post-prandial glucose...?

Again, the effect on low-carb on glycaemia might differ between pts on different ratios of basal:bolus.

Then of course, there's the differing amounts of carb that were categorised as "low carb".

(& OMG some studies don't even measure how much carb people are actually eating in the trial)

(It's so dumb we don't account for this I can't even be bothered to go into it).

So please can we stop trying to compare these studies at a "study level".

We need person-level data (this was @DylanMacKayPhD 's great idea).

@JohnArnoldFndtn Please fund Dylan's team to do this so we can solve this once and for all. ❤️