Read on Twitter
This is not an important study. Neither its findings nor its methods are clear. It is flawed to the point where it has no use.
Iâm sympathetic to David Davis not understanding cluster randomisation, bias in trials, t-tests and Kaplan Meier analyses, but... https://twitter.com/DavidDavisMP/status/1360647462197878791
Iâm sympathetic to David Davis not understanding cluster randomisation, bias in trials, t-tests and Kaplan Meier analyses, but... https://twitter.com/DavidDavisMP/status/1360647462197878791
...Iâm not clear why he would share something he didnât understand and follow it with such dogmatic assertions about policy?
The damage that sharing this does isnât that people will take too much vitamin d - a small regular supplement wonât be harmful. https://www.nhs.uk/conditions/vitamins-and-minerals/vitamin-d/
The damage is that itâs a distraction from meaningful interventions which actually save lives. It promotes really bad science.
The paper is here if you want to read it but I wouldnât bother. https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3771318
The study claims to be randomised - so from the abstract you might think that patients had been randomly assorted into different groups for treatment with vitamin D - which would be a good way of doing this study.
But this doesnât seem to be what was done.
But this doesnât seem to be what was done.
I say seem because what exactly was done isnât clear - always a huge red flag.
It seems like the *wards* are randomised not the patients. This is quite a big problem. Different wards have different types of patients. Obviously.
We group patients in hospital very often by how sick they are. So wards have different patients. Some wards have different equipment. Different levels of staffing.
If the group of patients given the treatment in your study is different to the group given the placebo then youâre not measuring your intervention youâre measuring the difference between the groups.
If this is discussed in detail or controlled for in the paper I canât find it.
This sort of randomisation (of identical clusters) can be legitimate in some studies (though not in this instance) but you need to analyse the data using appropriate methods...which they donât.
The other flaws arenât really necessary to point out since one giant flaw is all you need to totally disregard this. But the rest is a hot pointless mess.
More detailed excellent threads on this here
https://twitter.com/gidmk/status/1361063430745022467?s=21
And here https://twitter.com/fperrywilson/status/1360944814271979523?s=21
https://twitter.com/gidmk/status/1361063430745022467?s=21
And here https://twitter.com/fperrywilson/status/1360944814271979523?s=21
I do hope @DavidDavisMP will consider taking his widely shared post down.
