Weight loss when it's hard 5. Son of diazoxide

Life can be mean to you, just on a random chance basis. Sometimes bad stuff just happens.

Imagine what it's like to be planning a study, let's say on the use of a drug for managing atopic dermatitis in dogs, and you've struggled to recruit 20 participants. After the randomisation process you notice that you've ended up having recruited eight West Highland White Terriers (all with pink toes, sweaty armpits and sore ears, many Westies live with atopy) in to the treatment group (n=9) and a heterogeneous mix of healthy mongrels in to the placebo group (n=11). Oh dear. This is a real problem and can happen with small group-size studies. At this point you should consult a statistician or perhaps your group leader, before staking your career on the outcome of this particular study. Suicide is not usually needed, even if considered.

So poor old Due, the first author of a paper produced by Astrup's group in Denmark, struggled through 280 potential participants for his (or her? Dunno) diazoxide study before eventually finding 47 people who conformed to the entry criteria. Twelve of these said get lost. So 35 people started the study. Randomisation and drop outs gave n=13 in the diazoxide group and n=18 in placebo group by the end of the study.

Due got bitten by his patient groups and doesn't appear to even realise it happened! Whoever did the statistics on the participant characteristics of this study noted that the diazoxide group had an average fasting insulin level of 157 pmol/l where as the placebo group had a level of 119 pmol/l, with p less than 0.001 for the difference. Yes, I counted and recounted the number of zero's in this p value. It's not a typo. Oh dear! This never made it in to the discussion, it never even made it in to the results section, just in to a table, so I assume Due never noticed! ! ! ! It must have gotten past two scruntineers too.

Within one week of starting the drug there was a significantly greater drop in the average insulin level of the diazoxide group. Of course there was, this is what the drug does. Even Due noticed this. Because of the way the results table is laid out we only get the differences, down by 50 pmol/l in the diazoxide group, by 5 pmol/l in the placebo group. Again p less than 0.001 when comparing the changes. What Due didn't do was compare the insulin levels at this point BETWEEN groups. The diazoxide group is going to be in the region of 107 pmol/l and the placebo group at this point would be around 114 pmol/l.

I defy anyone to find a statistically or biologically significant difference between the fasting insulin levels at this point.

By the end of the study the fasting insulin values were about 100 pmol/l on diazoxide plus semi starvation versus about 80 pmol/l on semi starvation alone in the placebo group. I've no idea what the p value for the differences between groups might be at this point and Due certainly isn't going to tell us.

Going to table 2 we find that the diazoxide group, which ran a fasting insulin level of just over 100 pmol/l throughout the study lost 4.9kg and the placebo group, which ran a fasting insulin which started at 115 pmol/l and which probably dropped continuously to 80 pmol/l and so probably averaged just below 100 pmol/l, lost 6.4kg.


An aside:

At this point it is certainly worth noting from table 2 in Alemzadeh's paper that the initial fasting insulin levels were the opposite way round to those in Due's paper, tending to emphasise the diazoxide effect (204 pmol/l in the placebo group vs 168 pmol/l in the diazoxide group). Checking the bottom left hand graph of figure 2 shows that this difference was not statistically significant. We have to decide for ourselves if it was biologically significant.



Back to Due's paper:

Averaged over the bulk of the study the placebo group probably had slightly lower insulin levels that the diazoxide group and lost slightly more weight. All non significant, but with no significant difference between insulin levels, why should you expect any significant difference in weight loss?

Diazoxide is not a slimming drug. It reduces insulin levels, which is what allows the weight loss. Whether you drop insulin levels by starvation, diazoxide and starvation, LC eating or becoming muscle bound, insulin is what matters. No differences in insulin, no differences in weight loss.

Of course, it is possible that Due and Astrup were fully aware of what they achieved and thought long and hard about how they presented their results. I don't think so. I sincerely hope not. Stupidity is much more pleasant to consider than deviousness.

Peter



PS I'm not going to discuss the discussion. Due doesn't really discuss anything there anyway, just re states his results, recapitulates those of the original diazoxide paper without understanding where or how the differences in outcome arise and then states that diazoxide is useless for weight loss. There is no synthesis or understanding of what is going on. Probably due to working under Astrup.

PPS It's worth noting that semi starvation DOES drop insulin levels and does increase insulin sensitivity (Matsudas index from 5.91 up to 8.31 in this study) and so does lowering insulin with diazoxide plus semi starvation (Matsudas index up from 5.18 to 9.27). I wonder what 8 weeks of hypocaloric LC eating would do, or even weight stable LC eating...