November 15, 2020

expert reaction to study looking at statins, placebo and side effects

A study published in the NEJM and presented at the American Heart Association Conference looks at statins, placebo and side effects.

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

“It’s very well known that some medical treatments can improve a patient’s condition even if they don’t actually involve anything that is specifically aimed to improve the condition.  This is called the ‘placebo effect’.  In trials of new drugs, for instance, just taking a pill can sometimes have a positive effect on a patient’s health, even if that pill contains no active ingredients.  If some patients are given a pill containing a new drug, and others are given no pills at all, those who have the active pills might do better on average, simply because they have taken a pill.  So in practice all the patients would take pills that looked identical, but only some of them would include the active ingredient, and the others would be what are known as placebos.  (In practice, most drug trials would be more complicated than that, but they would certainly very often involve some sort of placebo.)

“What’s probably not so well known is that there’s an effect that could work the other way.  Sometimes taking a pill with no active ingredients can make people’s condition worse in some way, rather than better.  That’s an example of what’s called a ‘nocebo effect’.  One type of example of nocebo effects, that has been observed with some treatments, is that patients who anticipate that a treatment may have adverse side effects can experience such effects, even if they are taking a placebo treatment that does not contain any active ingredients.

“There is a great deal of evidence that statins can lower the risk of serious cardiovascular diseases in many types of patient, by lowering cholesterol levels in the patients’ blood.  Statin treatments have, however, been occasionally controversial because of concerns about adverse effects.  This is the case even though several clinical trials of statins, in which some patients had an active statin and others a placebo, showed no difference in the rate of adverse effects between patients who had the active treatment and those who had the placebo instead.  The general evidence is that the rate of adverse effects for people taking statins is actually low.  But nevertheless some patients do experience adverse side effects, and because of that many do discontinue taking the statins, and thus miss out on the beneficial effects.  The researchers on this new study suggest that in fact many of the perceived side effects may arise because of nocebo effects, rather than any effect of the actual statins in the tablets they take.  They therefore recruited 60 patients who had previously stopped taking statins because of side effects.  Over a year, each patient took different treatments for a month at a time.  There were three possible treatments: a statin, a placebo that looked just like the statin tablets, and no tablet.  Each day the patients reported how intense their adverse symptoms were, on a scale from 0 to 100.  At the end of the trial, the average (mean) symptom intensity when patients were taking neither tablet was 8.0.  When they were taking the placebo, the mean intensity was 15.4, and when they were taking the tablets that did contain statins, it was 16.3.  So the adverse effects were on average slightly worse on the statins than on the placebo, but both were considerable worse than when neither pill was being taken.  Allowing for the ‘baseline’ level of adverse effects when the patients knew they weren’t taking statins (because they were taking nothing), that makes the average level of side effects on the dummy tables 90% of the average level on the statin tablets.  In fact, the difference between average symptom intensity on the placebos and on the statins is within the range that could occur just because of random variability of symptoms and patients, so it’s not clear that the experience on placebos and on statins is really any different at all.  The researchers conclude that the nocebo effect, rather than the actual statin content, is responsible for a very high proportion of the adverse symptoms.  After the trial, half the patients involved had successfully restarted statin treatment.  (Of course that means that the other half had not restarted, and when asked, 18 of them still said this was because of side effects.)

“This is an “n-of-1” trial, which in this context means that each patient’s symptom levels on each of the three different treatments is most directly being compared with the same 1 patient’s symptom intensity levels on the other treatments.  That’s important, because the general level of adverse symptoms differs a lot from one patient to another, and essentially comparing each patient with him- or herself makes the comparisons statistically much clearer.  It also means that each patient individually can see how his or her symptoms on the placebo compare with those on the statin, and that may well have contributed to a persuasive effect that persuaded many of them to restart statins.

“I find these results quite convincing, but there are some issues to consider.  First, something that is in my view a non-issue.  The researchers report that they had to change the statistical method of analysing the results, because their original proposed method clearly did not work appropriately with the data they had gathered.  A change of plan can always lead to suspicion about the motive.  But in this case they explain clearly what went wrong, and I can see that it was necessary to change.  (My only small concern is that they did not anticipate the problem that arose, and did not in the original planning propose to use the method that they did eventually use, or something similar.)  The details of the analysis method that was used cannot be made very clear in the brief text of the main research report, but many more details are given in the supplementary appendix, and the method looks appropriate to me as a professional statistician.

“Second, there does perhaps remain a question of how general these findings might be.  The patients in this trial were not typical of all patients who are prescribed statins, because they had all discontinued statins previously, after side effects that started within two weeks after they started on statins.  Also, it’s possible that the patients who eventually took part in the trial were not even totally typical of patients who have discontinued statins.  In fact, 284 people got as far as being invited to the screening stage of the study, in that they were apparently eligible to be included in the trial, but of those, 222 did not in fact attend the screening.  Half of those gave no reason; the others gave a wide variety of reasons, including in some cases not being willing ever to restart on statins, even for just one tablet.  So there is some question about how typical the 62 were who did attend the screening.  (Two of those did turn out to be ineligible for the study, but the remaining 60 did at least start the trial.)  The trial does, in my view, give good evidence that the nocebo effect plays an important role in adverse effects attributed to statins, but exactly how this applies in a wide range of different types of patient is rather less clear.

“One point about the press release; it says that “90 per cent of symptoms experienced by patients on statins were present when they took placebo tablets”.  That’s potentially misleading – it sounds as if it might be based on a count of different symptom types or individual symptoms, but that’s not what was found.  Instead the average reported intensity of side effects when taking a placebo was 90% of the intensity when taking statins, after allowing for the level of side effects when no tablets were being taken.  I know that’s more complicated – but I wanted to point out that it isn’t the case that certain identifiable side effects, 90% of them, were shown to be due to the nocebo effect in this study, and so the other 10% may be due to something else.”

Prof Liam Smeeth, Professor of Clinical Epidemiology, and Dean of the Faculty of Epidemiology and Population Health, London School of Hygiene and Tropical Medicine, said:

“Many people stop taking statins or are reluctant to start them because of worries about side effects.  This means many people are missing out on the proven ability of statins to reduce people’s risks of heart attacks and strokes.  The study used innovative methods whereby individual people had their symptoms assessed during periods on a statin, on a dummy pill, and during no treatment.  The study convincingly showed that while people had various symptoms that they might attribute to statin use, symptom scores were no higher during periods of statin use compared to periods of no statin.  This demonstrates that while troublesome symptoms often occurred, they were not caused by the patient’s statin.”

‘N-of-1 trial of a statin, placebo or no treatment to assess side effects’ by Frances A. Wood et al. was presented at the American Heart Association conference and published in the NEJM at 15:28 UK time on Sunday 15 November 2020.

DOI: 10.1056/NEJMc2031173

Declared interests

Prof Kevin McConway: “I am a Trustee of the SMC and a member of the Advisory Committee, but my quote above is in my capacity as a professional statistician.”

Prof Liam Smeeth: “Liam Smeeth is the principal investigator of a similar study called StatinWise, the results of which are currently under peer review.”