Role of Blinding in N-of-1 Trials

Recently, there has been much interest in several blinded crossover N-of-1 trials of statin side effects.^1–4 These personalized trials allow patients with previous statin side effects to repeatedly crossover between periods of taking a statin and periods of taking a placebo to explore their own symptoms. The trials seem to indicate that the symptoms patients often associate with taking statins are much rarer than routine clinical practice would indicate.

See Article by Tudor et al

However, we knew this long before these elegant trials. Statins, like any other drug, are only granted a license if they can show reasonable efficacy and tolerability against placebo in large randomized controlled trials. Meta-analyses of tens of thousands of patients across these trials have shown patients more commonly stopped taking their placebo tablets due to side effects than their statins.⁵

However, when faced in a patient in the clinic with severe myalgia that started after they started their statin, this information is of little reassurance. Telling a patient that the statin drug itself caused myalgia in <1 in 1000 patients in placebo-controlled trials does not make them feel better—at best, they will assume they are the unlucky 0.1%. At worst, they will think you do not believe them. Instead, what is needed, is an answer for that patient—what is causing their side effects? This form of personalized medicine allows patients to explore their own symptoms, validates their concerns, and can even get 50% of patients previously intolerant to statins back on these life savings drugs.^2,3

This is the role of the N-of-1 trial. However, they are difficult to run. Any randomized trial is expensive and time-consuming, drug trials are particularly fraught with bureaucracy, and the 3 most high profile N-of-1 trials for statin side effects to date have been blinded,^1,2,4 which introduces further costs. For blinding to be successful, investigators must ensure the statin and the placebo look identical. The easiest way to do this is with over-encapsulation, where off-the-shelf statin and placebo tablets are covered with an identical outer coating. One trial used this,⁴ but there remains the theoretical risk that unblinding can occur if the capsule is broken. The more expensive alternative is to have a bespoke placebo manufactured to look identical to an existing statin.²

In this issue, Tudor et al⁶ investigate whether unblinded N-of-1 trials could be used as a more practical alternative, avoiding the need for expensive placebo and rigorous blinding practices. They recruited patients with prior statin intolerance and randomized 93 individuals to one of 3 arms: behavioral counseling followed by a blinded N-of-1 trial, behavioral counseling followed by an unblinded N-of-1 trial, or usual care (without counseling). The N-of-1 component comprised six 1-month-long blocks of statin or placebo. The investigators found that 71% of the patients who embarked upon the unblinded arm completed the full 6 months, versus 82% of the blinded arm. Similar proportions of patients restarted statins after the trial in the blinded (46%) and unblinded (44%) N-of-1 trials, but fewer in the usual care arm (20%). They, therefore, argue that an unblinded N-of-1 trial approach with behavioral counseling can improve medication compliance when compared with usual care, with an efficacy similar to a blinded N-of-1 trial approach.

The study is pragmatic. However, as with all the previous trials, there are questions about generalizability. Just one in 8 patients invited ended up participating, and an obvious concern is whether these trials inherently select a subset of patients who are relatively unsure about the veracity of their statin side effects, and who are particularly likely to restart them. Regardless, the N-of-1 strategy clearly works for some.

So, do we need blinded N-of-1 trials? It is important to remember how we got here. In blinded randomized placebo-controlled trials, side effects appear no more common in those taking statins than in those taking placebo. And yet, up to 50% of patients taking a statin have stopped it within 2 years.⁷ There are 3 possible explanations for this discrepancy.

The first explanation is the randomized trials of tens of thousands of patients are wrong. Trials are infamous for tending to recruit unrepresentative patients with fewer comorbidities, but this still seems unlikely to account for the huge discrepant burden of reported statin side effects we see in clinical practice.

The second explanation is that reported statin side effects are predominantly ever-present background symptoms that are being falsely blamed on the statins. For example, a patient may suffer either continuously or intermittently from myalgia of another cause. However, when the patient is taking a statin, they may tend to unfairly blame it on the drug.

The third explanation is the nocebo effect. Here, it is not the statin molecule itself that is causing the symptoms, but the act of taking the tablet, and the negative beliefs around it. This is the evil twin of the placebo effect, where taking an inactive tablet makes you feel better.

Different N-of-1 trial designs can tease out these different phenomena.

All the N-of-1 trials for statin side effects to date have recruited patients who have suffered symptoms while taking statins previously, and so tackle the first explanation by specifically targeting the group that the prior RCTs might be criticized for missing.

The second explanation, whether symptoms are ever present, can be tackled by an unblinded N-of-1 trial approach, which allows patients to explore their symptom burden while not taking any tablets (or an open label placebo) and see if they are any worse while taking statins.

However, helping a patient explore the nocebo effect is more difficult. If a patient does get more symptoms when taking a statin in and unblinded N-of-1 trial, this might represent a genuine side effect caused the statin molecule, or the nocebo effect. To separate these phenomena, the trial must have a control arm where the nocebo effect is also present, that is, a blinded placebo arm. Only then can any difference between arms be attributed to the drug, rather than the nocebo effect.

Given this, Tudor et al’s results are interesting: a similar proportion of patients in the unblinded and blinded arms were sufficiently reassured to recommence the statins that they previously did not tolerate. They therefore hypothesize that the nocebo effect probably is not the predominant phenomenon in these patients. Perhaps, although the individual patient data from the 3-arm blinded SAMSON (Self-Assessment Method for Statin Side-effects Or Nocebo) suggests otherwise.² An alternative explanation is the nocebo effect is a significant problem, but we have not yet succeeded in helping patients overcome it, even when a blinded N-of-1 trial supports its existence.

We applaud Tudor et al for showing that we can help many patients who have stopped statins due to side effects with an unblinded approach. This is certainly easier and cheaper to run than the trials that preceded it. However, this is at the price of being able to tease out the true etiology of what our patient is experiencing.

It is a shame these N-of-1 studies need to be run formally as trials at all. Atorvastatin tablets are available in 10, 20, 30, 40, 60, and 80 mg tablets. Imagine if future licensing laws were to require a small number of 0 mg tablets to be created? These tablets could then be prescribed by doctors, have their labels removed, and serve as cheap and perfect placebos for patient-run N-of-1 trials using a simple symptom diary. Until then, maybe unblinded N-of-1 trials are the next best solution.

Article Information

Footnotes