By Maryanne Demasi, PhD, Paula Byrne, PhD, Mark Jones, PhD, Robert DuBroff, MD.

On several occasions, we have been asked to respond to a critique of our 2022 systematic review and meta-analysis on statin trials.

The critique was written by Peter Attia in April 2022, a physician and popular podcaster whose interests lie at the intersection of longevity, lipids and heart disease.

Many of the criticisms in Attia’s article were outside the scope of our research, but given his large social media following, and the many requests for our response, we thought we’d address his main points of concern.

A quick recap

Our study, published in JAMA Internal Medicine, examined 21 statin trials involving 143,532 participants and found:

• No consistent relationship between lowering LDL-Cholesterol (LDL-C) and death, heart attack or stroke, following statin therapy.
• After statin therapy, the relative risk reductions for death, heart attack and stroke were 9%, 29%, and 14% respectively.
• The corresponding absolute risk reductions were 0.8%, 1.3% and 0.4% (see graph).
• The benefits of statins were minimal, and most of the trial participants who took statins, derived no clinical benefit.

Our response to Attia

First, Attia suggests our study doesn’t justify a more judicious approach to statin prescribing, and raises concerns that the coverage in the press could’ve prompted people to stop taking their statins.

… the data from this study do not provide justification for such revision, despite rampant – and potentially deadly – insinuations to the contrary in popular press and social media.

While we cannot control press coverage, the central point of our study was clear — too often, statin advocates will communicate a drug’s relative risk reduction (impressive number), without quoting the absolute risk reduction (trivial number), a practise we consider unethical.

Patients need transparent communication of the relative and absolute risk reduction when being informed about the benefits and harms of statins. This is crucial for patient autonomy, informed consent, and shared decision-making in the doctor-patient relationship.

Second, Attia says we “chose” LDL-C as the focus of our analysis.

Byrne et al. chose to use LDL-C as the relevant variable for mechanistically linking statin efficacy to clinical outcomes, yet LDL-C is not an ideal metric for determining the atherogenic risk. 

….either apoB or non-HDL-C would do so more accurately, yet neither of these preferred metrics was included in the analysis conducted by Byrne and her colleagues

However, LDL-C was the metric used in the design of the 21 statin trials we analysed.  Attia suggests that apoB or non-HDL-C are better indicators of heart disease, but few of the original statin trials reported apoB or non-HDL, so that analysis would not have been possible.

It’s useful to note that LDL-C is the metric usually used in cardiovascular “risk calculators” and to justify statin treatment in professional guidelines for doctors. 

For example, the American Heart Association guidelines say that “LDL-C is the primary cause of atherosclerosis” and promote “the general principle that the lower the better for LDL-C” and that “under certain circumstances the measurement of apo-B may have advantages….nevertheless… carries extra expense, and it’s measurement in some laboratories may not be reliable.”

Third, Attia’s critique pointed out that our study analysed the benefits of statins for an average duration of 4.4 years.

This duration is almost certainly too short to show the full potential effect of LDL-C reductions on CV risk and mortality.

We agree that the duration was short, but the original trials themselves were short.  While we excluded trials in our analysis that were shorter than one year, most of the statin trials, particularly in the later decades, were less than 5 years in duration.

Interestingly, Attia points to studies using Mendelian randomisation as a way of determining the long-term benefits of statin therapy. These complex studies focus on genetic differences in markers of LDL-C that are apparent at birth, rather than treatment induced reductions in actual LDL-C that begin in later life.

If an individual born with a genetic metabolic defect is identified and enrolled into such a study as an adult, we’d have no means of identifying other individuals with the same genetic defect who didn’t survive into adulthood.  This is a clear example of selection bias.  Moreover, we’d have no idea what clinical events took place before enrolment. It’s like conducting a randomised controlled trial lasting 50 years but ignoring the clinical data acquired during the first 45 years of the trial.

Attia argues that if the treatment duration is too short, then the only way to detect a significant effect on mortality is to examine studies with a large magnitude in LDL-C reduction. However, we found some statin trials, with the greatest LDL-C reduction (AURORA and CORONA), reported no clinical or survival benefit.

He cites the FOURIER trial as evidence.

(See for example the FOURIER trial on combined evolocumab and statin treatment), and they show strong causal associations between LDL-lowering treatment and CV events and mortality.

However, the FOURIER trial did not show a mortality benefit. In fact, there were more cardiovascular deaths and total deaths in the group taking the PCSK9 inhibitors compared to placebo, albeit not statistically significant. 

Notably, the FOURIER trial is now being questioned by researchers who “reanalysed” the data, and found some deaths may have been misclassified. These authors concluded that the potential harm from the treatment is higher than initially reported. 

Additionally, a systematic review and meta-analyses of 54 trials using PCSK9 inhibitors, published in BMJ Heart, reported no reduction in total mortality or cardiac deaths despite dramatic reductions in LDL-C.

The problem with assuming that the benefits of statins continue to accumulate over decades, is that it ignores the fact that statin harms also accumulate — especially if the harms take years to manifest e.g. cognitive decline and diabetes mellitus.

Fourth, Attia points to a study to support the “very strong associations between LDL-C and risk of CV [cardiovascular] events.”

He cites a study that assessed CV events using a “composite of cardiovascular death, nonfatal myocardial infarction, or coronary revascularization.” However, this overlooks the main strength of our analysis. 

We avoided using a composite endpoint because it mixes objective outcomes (death, heart attack and stroke), with subjective outcomes (hospital admissions, angina, and revascularisations). 

We focused on hard outcomes only, making our study robust and less prone to bias.

Recently, our own findings were confirmed by a meta-analysis of 60 randomised controlled trials of statins, PCSK9 inhibitors and ezetimibe, which found no association between the degree of LDL-C reduction and cardiovascular or total mortality.

Finally, Attia says that the article “hasn’t shaken my faith in statins” but we have put faith aside and focused on the data.

The vast majority of these trials are funded by the drug industry, and have been carried out during an era with little regulatory oversight and accountability.

Further, statin manufacturers and a group of researchers at Oxford University remain guardians of the individual participant data, and refuse access to outside researchers for independent scrutiny.