Insulin, are you hungry?

An apology. This is a dry post, I had to edit the zombies out as it was getting way too long, maybe another day. It's a bit difficult to know where to start on quite how bad this paper is. Obviously, having read the abstract, we can flick down pretty well immediately to Fig 1 in the full text.



There are a few oddities. First is the flat line in weight gain on days 1, 2 and 3. This is the suppression of hunger by insulin, maybe. There was a full seven days on insulin. This I will return to in the next post.

Next is the sudden increase in weight gain through days 4, 5, 6 and 7 in the insulin infused groups, giving a final set of weight gains on day 7 which are not statistically distinguishable from controls. Except in the group on 2iu/24h of course. The group receiving 2iu/24h is special.

Then there are the data from days 11, 12, 13 and 14. By this time the insulin infusion had stopped (which occurred around day 7ish). Look at the 2iu/24h group. Waaaay after the insulin infusion had stopped their weight gain was still much slower per day than the other three groups. Oddly this didn't reach p < 0.05, despite standard errors which were far from overlapping those of the other three groups. But trying to see what the final weights gains were is difficult because these "post pump" weight gains have been, err, umm, sort of, err. I'm not sure what the word I need is...

You see the data from these last four time points are slightly moved. Each plot has been pulled down, and by a different amount each. No one is going to say by how much. It's pretty obvious that the control line can simply be moved back up to show a linear increase in weight from the insulin infusion period as these rats never got any insulin. But all lines have been shifted down so their day 11 values are set to their day 7 values, whatever the intermediate weight gain on days 8, 9 and 10 was. It is quite likely that the 6iu/24h and the 1iu/24h rats gained weight fairly linearly and so possibly ended up on day 14 at exactly the same weight as the control group. Or heavier.

It's also very likely that the 2iu/24h group also gained weight fairly linearly but slowly, ie their "pulling down" of day 11 values to those of day 7 didn't involve much of a drop compared to the other three groups. Who knows outside the lab?

Here are the data from Fig 1 in tabular form:




Anyhoo, the 2iu/24h rats, however much they did or didn't eat/gain on days 8, 9 and 10, only gained 1.39g/d on days 11, 12, 13,and 14. Food intake per day was down significantly through this later period, 27.7g/d vs at least 30g/d in all other groups. This is very important. The implication is that if you get yourself set up with just the right insulin infusion for a week, then you still won't be hungry a week later! Wow. Insulin is a satiety hormone blah blah blah.

But if you under-dose at 1iu/24h then it's, oh-oh, back up to pre-infusion weight gain rate, or possibly slightly more. Ditto if you over-dose at 6iu/24h, just the same thing happens. Fascinating. Do you think there might be something odd about this 2iu/24h group? Perhaps someone should repeat the experiment at this infusion rate? Then we might see if the result for these rats, on which the whole concept of suppression of weight gain over 7 days rests, was a quirk. No stats were done on the zero weight gain days, ie days 1-3 on insulin. The only p< 0.05, on which the title of the paper rests, was the 2iu/24h group at day seven.

If we lose the 2iu/24h group all we can say is that an insulin infusion reduces weight gain for three days, with complete restoration of any lost weight gain by the seventh day of a continuing infusion.


So, has the experiment been repeated? Luckily it has. By this very group. And the results are in this very same paper! But well buried. You have to be a dissonant pedant to find it. It's all in Figure 4.





This not quite the same experiment as Fig 1, the timings are slightly changed, but the basic design with insulin at 2iu/24h for seven days is identical.

In the main experiment time "on pump" was 7 days and they looked at all of these days, averaging everything over this time.

In Fig 4 they did the same 2iu/24h pump for seven days but only analysed days 3, 4 and 5 as time "on pump". Go figure. They also chose days 8, 9 and 10 as their "post pump" days vs days 11-14 in the first part of the study. Again, go figure. But eyeballing the graphical weight changes in Fig 1, I doubt this matters.

The data in Fig 4 look at meal size and meal frequency because that's how you bury data. But we can reverse engineer Fig 4 to get total food intake per day. Take a ruler to the graph. Multiply meal size by meal frequency and you get food intake per day, neat huh?

The rats on 2iu/24h ate 25.5g/d during "on pump" days 3, 4 and 5. This is pretty much the same as the total 7 day value from Fig 1 and Table 1. Happy researchers? Well done for correct choice of days. But...

Does the depressed food intake continue even after insulin has finished? Do you get sustained appetite control if you get the insulin infusion "just right" for a week? Eyeballing Fig 4's "post pump" values, these are about 3.4g/meal, 9.8 meals/day giving over 33g/d food intake...........

My, those are bloody hungry rats! This is the highest food intake per day in any group in the whole paper. It's the direct opposite of the findings presented in Fig 1 "off pump" section. The sustained depression of food intake shown in both Fig 1 and Table 1 could not be repeated in the Fig 4 experiment.

It doesn't happen.

The 2iu/24h group are no different to any other infusion rate when you look at Fig 4 "post pump" section. Quite why the rats on 2iu/24h used to generate Fig 1 data showed depressed weight gain long term is a complete mystery. Personally I'd want to have had a pathologist check out the pumps in the lowest food intake rats in this group, looking for low grade peritonitis. The pumps are in the abdominal cavity. Maybe some surgeon dribbled in to the wound during implantation. I've worked with surgeons. Ultimately we'll never know.

But ANYONE quoting the data presented of Fig 1 to you WITHOUT even mentioning the results of Fig 4 to you is, well, hmmmmm..... probably in obesity research.

I was going to go on to discuss the flat line of weight gain on days 1, 2 and 3 (at all insulin infusion rates) next but I'll leave that to another post as it has nothing to do with the "insulin at 2iu/24h causes sustained decreased food intake" claim.

Which is complete bollocks.

Peter