Wednesday, 19 October 2016

Epidemiology - the "science " of deceit.

Today a link on Twitter pointed to a study that claimed to show that "Eating breakfast was associated with significantly lower CHD risk in this cohort of male health professionals".

So I took a quick look at the study and weeded out some of the raw data. It's an analysis of 27,000 ageing male healthcare professionals of whom about 3,400 said they skipped breakfast. This subgroup were generally a bit younger (5 year difference in average age) and three times more of them were smokers - 15% vs 5%.
Over the course of about 15 years there were a total of 1527 cases of diagnosed coronary heart disease, 171 of which were in the small breakfast skipping group :-

Cases 1356 171
N 23616 3386

Cases per 1k 57.4 50.5

1.14 1.00

Person years 338074 49880

Average years 14.3 14.7

Ratios 7.9


So there were 8 times more cases of CHD in 7 times the number of participants in the breakfast eating group. That's 14% more cases per 1,000 study participants amongs the breakfast eaters.

The study goes on to "adjust" the data in line with invisble relationships until out of the other end of the computational sausage machine merges the conclusion that "Men who skipped breakfast had a 27% higher risk of CHD compared with men who did not".

I don't know how this works, or what they assume, or how valid it is. If less people die or get ill by skipping breakfast and smoking more, as per the data, then I am disinclined to believe that skipping breakfast can be harmful. Obviously the age advantage helps, but it almost looks like skipping breakfast can counter the impact of smoking which ordinarily we would expect to increased CHD markedly.

I am indebted to Prof Richard Feinman for his teachings on the importance of looking at the raw data.

No comments:

Post a Comment