select search filters
roundups & rapid reactions
before the headlines
Fiona fox's blog

expert reaction to study showing half of patients on statins do not have ‘healthy’ cholesterol after 2 years

Research published in HEART demonstrates that half of patients prescribed statins in primary care fail to reach ‘healthy’ cholesterol levels after two years of treatment.

Prof Kevin McConway, Emeritus Professor of Applied Statistics, The Open University, said:

“This is a good, careful study, though it does have some inevitable limitations stemming from the kind of data that it used.

“It’s an observational study, so it can’t show that the difference in cardiovascular disease rates, between people who respond well to statins and people who don’t, is actually caused by the difference in response to statins. There are other differences between the two groups, and it’s possible that these differences are the cause of the difference in disease risks, and not the response to statins at all. The researchers made considerable efforts to allow statistically for these other differences, but, as they point out, those adjustments can never be perfect and they can’t take account of factors that the researchers have no data on. They had data on smoking for only 4% of the people they studied, and data on body mass index for only 55%, and their data on deprivation applies to the small area (population about 1500) where the people lived, so it’s really about the average level of deprivation of their neighbours, not the participants individually. And there are other possibly important factors on which they had no data at all. So, while it’s certainly possible that the difference in disease risk was caused by differences in response to statins, at least in part, it’s also possible that it might be caused by something else entirely.

“Even if the difference in risks is caused by the difference in response to statins, this research can’t make it clear what should be done about that. Current guidelines give a target for the size of the reduction in LDL (‘bad’) cholesterol that doctors should aim for, but it’s less clear what they should do if that target isn’t reached. As a statistician, I can’t comment on what might be done about that.

“But I can comment on some of the numbers. It’s worth saying that the caveat in the press release, that the study can’t establish cause, really only applies to the conclusions about differences in risks. It doesn’t apply to the finding that about half of the people didn’t respond on target to their statins. The accompanying editorial points out that this high rate of sub-optimal response might be an over-estimate of the rate now, since the data were collected over a long period of time going back to 1990, and many participants were recruited before the current guidelines came in. But the rate of sub-optimal response is likely still to be high.

“Putting exact figures on the risks, for a study like this, is a little complicated, because the risks can be calculated in several ways. The press release says that those who didn’t respond optimally to statins were 22% more likely to develop cardiovascular disease that were those who did respond optimally. That’s true as far as it goes, but the 22% reduction is in something called the hazard, which doesn’t add up over time in a very obvious way. From other numbers in the paper, the following can be found. Imagine that there is a group of 100 people, whose pattern of ages and pattern on ‘bad’ cholesterol levels before taking statins match those of the people in this study. Imagine they take statins for 10 years, and that all of them respond well. Then about 14 or 15 of the hundred would have been diagnosed with cardiovascular disease. That already sounds like quite a lot – but remember these people would be middle aged or elderly, and that they probably would not have been prescribed statins unless their risk of cardiovascular disease was relatively high. But now suppose there’s a different group of 100 people, with the same pattern of age and ‘bad’ cholesterol levels before they took statins as the first group – except that they respond sub-optimally to their statins. After 10 years about 16 or 17 would have been diagnosed with cardiovascular disease. Still distinctly more than with the people who responded well, though perhaps the difference isn’t so big as the impression given by a 22% figure might indicate. (These numbers, like all those in the study, omit any diagnoses of cardiovascular disease in the first two years after starting on statins, on the grounds that the statins probably wouldn’t have their full effect on cardiovascular disease until 2 years have passed after starting on them.)”

Prof Kausik Ray, Professor of Public Heath, Imperial College London, said:

“This is an observational study and reflects so called “real world practice and real world behaviours”. It is not a randomised trial and therefore cannot provide evidence about effectiveness of therapies. The study and its conclusions are prone to some methodological issues as stated below.

  1. The % reduction in LDL-c with statins that is reported in trials is a mean. So if the median (or mean) reduction is 40% by definition half will be above half below. Here its useful to look at interquartile range so the range of effects from the 25-75 percentile for instance.
  1. Khunti Ray et al JAMA NETWORK OPEN using the same data source showed that % reduction is related to both adherence and the potency of the drug prescribed. So those that get < 40% reduction are less likely to be adherent to medications or prescribed more potent (higher doses)
  1. We know from CTT that its absolute reduction in LDL-C that determines the relative benefit from lipid lowering therapies. So the observation that a 1 mmol/L lowering of LDL among those with LDL-C lowering < 40% vs > 40% is likely confounded by the fact that those with LDL lowering of < 40% would moe likely have a lower LDL-C so a smaller percentage reduction translates into a lower absolute reduction in LDL say if the starting LDL were 2.5 vs 4.5 and hence a smaller absolute reduction in LDL which translates into a smaller relative benefit.
  1. For the < 40% grove why are only strokes mentioned not deaths from vascular disease non fatal heart attacks and non fatal strokes?
  1. Bias by indication. Not sure if this study was propensity matched. i.e. a doctor would mostly chose a more potent dose of statin in those they perceived to be at higher risk and patients might be more adherent if they perceived themselves to be at higher risk.
  1. The only conclusion is that a large portion of people do not achieve guideline based recommendations. I do not agree this is genetic variation and tests are needed. What is needed is for Drs to prescribe the right doses in the first place for patients to be educated about adherence. Then monitoring in those with suboptimal responses a shared discussion between patient and doctor about the reasons for the suboptimal response and the next steps to ensure the LDL-C reduction is appropriate for the level of risk in the patient.”

‘Sub-optimal cholesterol response to initiation of statins and future risk of cardiovascular disease’ by Akeya et al. was published in HEART at 23:30 UK time on Monday 15th April 2019.

Declared interests

Prof. McConway: Is aTrustee of the Science Media Centre.

Prof. Ray: “I have consulted for companies developing therapies for cholesterol management, diabetes and anti-platelet therapies to prevent cardiovascular disease.”

in this section

filter RoundUps by year

search by tag