How do I love the Ministry of Health? Let me count the ways.... [updated]

Ok, not very much.

Readers will recall that I jumped up and down a bit last year about the MoH pulling a fast one in its Nutritional Guidelines for Older People. In short, they cited Rimm and Moats, 2007, as providing evidence that uncontrolled confounds eliminate measured coronary health benefits of moderate alcohol use when in fact Rimm and Moats prove the exact opposite: they show that controlling for those confounds only mildly reduces the J-curve effect.

I emailed MoH, I sent a sharply worded letter to Tony Ryall's office. I didn't hear anything back other than a perfunctory note saying that Ryall had received my letter.

Today I went to check the nutritional guidelines. My old link didn't work for some reason, but I was able to find it here. And lo and behold, they've made a slight change. Did they get rid of the section saying there's no J-Curve? Did they note the new evidence showing the J-Curve to be robust to the prior confounds? Oh, loyal readers, you know our Ministry of Health better than that. They just deleted the reference to Rimm and Moats 2007 from that paragraph while keeping all of the shouting about how there isn't a J-Curve; they instead referenced Middleton Fillmore et al 2007.

Ok. So what does Middleton Fillmore et al 2007 say? They've a metastudy where they claim that high quality studies that separate out abstainers from sick quitters shows no J-curve. But all of their results hinge on the two studies that they deem as being of high quality for coronary heart disease and seven for all-source mortality. By contrast, Di Castelnuovo and Donati's 2006 metastudy compared results in 21 papers results that lumped light and former drinkers with abstainers against those from 27 papers results that specifically excluded light and former drinkers from the reference category. Sorry, but the latter result's just more robust.

Bottom line, Ministry of Health, when they figured out that they were citing a paper that provided contrary results instead of the results they liked, nixed the cite and threw in one that gave them better results, utterly ignoring contrary evidence from more robust studies. Again, the kind of thing that would have me chastising an undergrad or hauling a grad student into the HoD's office for academic dishonesty. But academic dishonesty is the norm at MoH.

Oh, and there's still a reference to Rimm and Moats (2007) in your paper, MoH. It's at page 104:

Methodological issues that suggest the protective effects of alcohol on cardiovascular health may have been overestimated include:

• difficulties categorising amounts and patterns of drinking, and in choosing an appropriate
  reference category to measure drinking against (Emberson and Bennett 2006)
• the inclusion of people reducing or stopping their alcohol intake (often associated with
  ageing or illness) in the ‘abstainer’ category for data analysis, which may have overestimated
  the protective effect of alcohol, as it is not the absence of alcohol that increases the risk of
  cardiovascular disease but ill health ([Middleton] Fillmore et al 2007)
• uncontrolled confounding of healthy lifestyle factors among light to moderate drinkers, such as healthy eating and physical activity (Jackson et al 2005; Rimm and Moats 2007).

MoH pulled the quote I'd highlighted, pulled Rimm and Moats from the bibliography, but forgot to run a full text search to see whether they'd cited Rimm and Moats elsewhere in the paper. They can't even cheat right. [Note, I've saved the PDF of the current version and probably have the old one saved somewhere too. It pays to be a pack-rat when the Ministry of Health runs a memory hole.]

Why oh why can't we get a better Ministry of Health?

Update: I sent the following OIA request and will let folks know what comes of it.

In terms of the Official Information Act, please provide all internal and external memorandums, emails, file notes and written notes regarding the revision of the 2010 Nutritional Guidelines for Older People to delete the reference to Rimm and Moats (2007) from section 4.15 of those guidelines. The reference was present in the Guidelines in August 2010 but has since been replaced with an alternative.