For those lovely people who support what I do, this will be my Monday note (22nd April). I’m sharing the punchline early (spoiler alert), as this needs to be countered asap… (there’s still plenty more to tell you about this study…)
Yesterday’s headlines variously reported “Statins ‘don’t work well for one in two people’” (BBC) (Ref 1); and “Half of people prescribed statins don’t reach cholesterol goals after two years, study says” (CNN) (Ref 2).
The headlines came from this study, which is not on open view (I’ve got a copy), but you can see the summary here (Ref 3).
The researchers looked back at a UK database to identify 165,411 people who didn’t have cardiovascular disease (CVD), but who were put on statins between September 1990 and June 2016. These patients also had a measure for their LDL-Cholesterol in their files from the year before going on statins and another LDL-Cholesterol result some time during the two years after going on statins.
The researchers used national cholesterol-lowering guidelines to say that – if people had at least a 40% reduction in their LDL-Cholesterol, this would be called ‘optimal’ and if they had less than 40%, it would be called ‘sub-optimal’. The 165,411 people were therefore split into two groups. The findings were presented as:
1) 51% of patients were in the ‘sub-optimal group’ and 49% of patients were in the ‘optimal’ group. (The most likely reason for half of people being in the ‘sub-optimal’ group is that they stopped taking statins, or didn’t take them often – this is the ‘naughty patient’ confounder. These people tend to respect authority less, do as they’re told less and ignore health/medication messages more.)
2) The optimal group went on to have fewer CVD events over the next few (average of six) years. The relative risk difference was presented as 17%.
I spotted an issue in the characteristics table (Table 1 in the paper) – the top part of which is shown below:
I emailed the authors and they kindly replied overnight.
The issue was that Table 1 shows the sub-optimal group to have significantly more “alcohol misuse” (1.43 times that of the optimal group). Geez – just how bad does drinking need to be before it becomes misuse? We don’t have detailed information on alcohol consumption beyond this – were the alcohol misusers also far less likely to be non or moderate drinkers and far more likely to be heavy and frequent drinkers?
Then look at the smoking information. We can see, from the very limited information we do have, that the ‘sub-optimal group’ were 25% more likely to be smokers. However, we don’t have smoking information for 96% of patients.
Drinkers are more likely to be smokers and we already have enough information to see that our naughty sub-optimal patients are drinking, smoking disasters. Do they exercise less? Probably – we don’t have that information. Are they more likely to be obese? Maybe – we don’t have that information. Table 1 also tells us that they were more likely to be poorly-controlled diabetics and less likely to have hypertension treated.
The correspondence that I had with the researchers confirmed that they had not adjusted for anything other than age and baseline LDL-Cholesterol to get the risk ratios (adjusting for the latter would have favoured the optimal group, by the way). They had not adjusted for alcohol misuse, or smoking, or any other lifestyle factors that were known to be different between the two groups – even with vast amounts of missing information.
The way in which this omission was presented in the reply tells us a great deal. Smoking and alcohol misuse were not seen as relevant to the relationship between LDL-Cholesterol and CVD and thus they were not adjusted for. This show the mindset of people wedded to the cholesterol hypothesis: the belief is that LDL-C causes CVD and thus, if smoking and alcohol don’t impact LDL-C, then why should they impact CVD? Because we know that smoking and alcohol misuse cause CVD directly: Duh!
This study should have been reported as, “Smokers and alcohol misusers have more CVD events – oh and they tend to be less compliant with health advice generally.”
Ref 1: https://www.bbc.co.uk/news/health-47933345
Ref 2: https://edition.cnn.com/2019/04/15/health/statin-cholesterol-goals-study/index.html
Ref 3: https://heart.bmj.com/content/early/2019/03/30/heartjnl-2018-314253