1. The Apple Watch Experiment
“It had to start somewhere,” said Gopi Dandamudi, MD, from CHI-Franciscan Health Pacific Northwest. On the basis of two unpublished studies, the US Food and Drug Administration (FDA) cleared Apple to enter the healthcare space. The release of the ECG app for the Apple Watch begins a grand experiment in screening for disease.
Perhaps the most remarkable part of this experiment is not its size but its leadership. Patients, not doctors, are in charge. Is this the beginning of what Medscape’s editor-in-chief, Eric Topol, called The Creative Destruction of Medicine?
I love Apple products, but I am pessimistic about the ECG app. While brilliant advances have made this a great time to be sick, it’s a bad time for people without disease to interact with healthcare. I predict that the mixture of fear, fee-for-service models, overburdened clinicians, and a poor understanding of the normal variations of the heart rhythm will lead to a tsunami of iatrogenic harm in the short run.
2. Treating Hypertension in the Community
Speaking of healthcare led by nonphysicians, here is the conclusion of the late Ron Victor’s cluster randomized trial of blood pressure reduction: “Among black male barbershop patrons with uncontrolled hypertension, health promotion by barbers resulted in larger blood-pressure reduction when coupled with medication management in barbershops by specialty-trained pharmacists.”
This conclusion understates the effect size. The mean systolic blood pressure fell by a massive 27.0 mm Hg (to 125.8 mm Hg) in the intervention group and by 9.3 mm Hg (to 145.4 mm Hg) in the control group; the mean reduction was 21.6 mm Hg greater with the intervention (95% confidence interval, 14.7 to 28.4 mm Hg; P < .001).
Results of the FAITH trial, another cluster-randomized trial—of black churches—confirmed the efficacy of a community-centered approach to health. In FAITH, motivational interviewing plus lifestyle interventions by lay health advisors resulted in significant blood pressure reductions compared with usual health education at 6 months.
For me, the worst aspect of US healthcare is the injustice. While overtreatment proliferates in wealthy enclaves, underserved populations, such as black men in urban areas, suffer needless morbidity from something as easily treated as high blood pressure. The genius of the barbershop approach is that it shows the feasibility of new models of care. Ron Victor and Anthony Reid leave a great legacy; we need to continue this work.
3. Aspirin Flops for Primary Prevention
This year, three large randomized controlled trials (RCTs) comparing aspirin at 100 mg to placebo for the prevention of cardiac events in people without heart disease told a clear story.
The 7-year ASCEND trial enrolled more than 15,000 patients with diabetes and found that aspirin reduced cardiac events by 1.1% but increased major bleeding by 0.9%. The 5-year ARRIVE trial enrolled more than 12,500 patients with moderate cardiac risk and observed no significant difference in cardiac events but a twofold greater risk for gastrointestinal bleeding in the aspirin arm. In the more than 19,000 elderly people followed for nearly 5 years in the ASPREE trial, aspirin did not prolong disability-free survival and was associated with a higher rate of bleeding and a statistically significant 1.5% higher rate of death.[7,8,9]
This medical reversal teaches two lessons. One is simple: Don’t use aspirin in patients without heart disease. The more important lesson pertains to the importance of contemporary clinical trials. The combination of better baseline therapies (statins) plus the secular decline in cardiac event rates makes it much harder for any therapy to show benefits. When applying evidence at the bedside, favor newer trials.
4. MitraClip Uncertainty
Two trials published this year reported utterly divergent results for percutaneous repair of secondary mitral regurgitation (MR) with the MitraClip device. The 12-month-long MITRA-FR trial showed no difference in a composite primary outcome of death or heart failure hospitalization. The 24-month-long COAPT trial showed a massive absolute risk reduction of 32% for the primary endpoint of heart failure hospitalization and an absolute reduction in death (secondary endpoint) of 17%.
Proponents of MitraClip explain these differences with four arguments: Medical management in COAPT was more aggressive, which meant it enrolled truly refractory patients; COAPT enrolled patients with more severe MR and less dilated ventricles, which means their left ventricle had not yet passed the point of no return; procedural complications were lower in COAPT (8.5% vs 14.6% in MITRA-FR); and the operators were more effective in reducing MR, as seen in the number of patients with residual MR of 3 or greater (5% vs 17%).
I am not convinced patients in these two trials were that different. A comparison of baseline characteristics suggests that the average patient was similar; grading of MR and left ventricle dimensions are hardly an exact science, and the 1-year death rates of the control groups of both trials are close: 22.4% in MITRA-FR and 23.2% in COAPT. I see equipoise.
The truth about cardiac devices is that once they are approved and reimbursed, optimistic cardiologists will use them. Look at Watchman uptake despite dubious data. Being wrong about MitraClip could be a massive blunder. Before expanding the indications for this device, FDA regulators should wait for two more pieces of data: 2-year results of MITRA-FR and the results of RESHAPE-HF, the third RCT of MitraClip in patients with secondary MR.
5. Stop and Think About Methods Uncertainty
Many variables affect the results of a study. The research question posed, inclusion/exclusion criteria, comparator, and endpoints are obvious factors. I had always thought the analysis of the data was rote and that one set of data led to one result. I was wrong.
Brian Nosek, PhD, and his team at the University of Virginia have shown that choices made in the way a set of data is analyzed can lead to substantial variation in effect sizes.
In the “Many Analysts, One Data Set” paper they recruited 29 teams of experienced researchers to analyze one large data set in an effort to answer one simple question: Are soccer referees more likely to give red cards for foul play to dark-skin–toned players? They found that each team made a unique choice in how they would analyze the data, and, crucially, these decisions led to statistically positive results two thirds of the time and negative results the rest of the time.
Given the coming revolution in “big data” and the fact that most medical science turns on frequentist techniques (eg, the P value), this previously under-recognized variance has great importance for clinicians.
Perhaps we are on the cusp of change in data analysis. In April, John Ioannidis, MD, DSc, from Stanford University, wrote in support of lowering the P value cutoff for significance in medical studies from .05 to .005. He called this a temporary fix until more durable solutions to data analysis are adopted.
The next time you see a small effect size, a P value close to .05, or a change in a trial’s endpoints, go back and revisit Nosek and colleagues’ paper and their discussion.
6. The Ischemia Trial, Faith Healing, and Subtraction Anxiety
The debate earlier this year over changing endpoints of the still-ongoing ISCHEMIA trial gifted cardiology a trove of lessons. The ISCHEMIA trial will compare the initial strategy of percutaneous coronary intervention (PCI) plus medical therapy vs medical therapy alone in patients with stable coronary artery disease. The original trial plan boasted a true test of efficacy by measuring two hard endpoints: cardiovascular death and myocardial infarction (MI).
The design and size (more than 5000 patients) of ISCHEMIA make it the final test of PCI in patients with stable coronary disease. The key feature of this trial addresses persisting criticism of COURAGE: namely, all patients had to have documented moderate (or greater) ischemia on stress testing.
In January, as the trial was nearly done recruiting, ISCHEMIA researchers amended the primary endpoint to a five-component composite of cardiovascular death, nonfatal MI, resuscitated cardiac arrest, or hospitalization for unstable angina or heart failure. They explained that this change was described in the original protocol, done at the direction of an independent data committee and necessary to prevent underpowered results from low event rates.
This news stimulated one of the most clinically relevant editorials I have ever read. In “Faith Healing and Subtraction Anxiety,” Christopher Rajkumar, MBBS, and colleagues at the Imperial College London explain how human belief systems affect decision-making. And why these beliefs make it vital to have bias-proof endpoints in unblinded trials.
Using the example of two studies on fractional flow reserve,[17,18] the editorialists explained that the simple act of telling patients that their physiology test indicated no artery obstruction dramatically improved symptoms. In the DEFER trial, for instance, the number of patients reporting chest pain was reduced from 88% to 54% after patients heard their lesion was not obstructive. Take-home: Our beliefs (faith) can help people feel better.
Subtraction anxiety exerts the opposite effect. When doctors believe that PCI “fixes” people, there is anxiety over not having it. Rajkumar and colleagues point to the positive FAME-2 trial, in which the composite endpoint was driven not by MI or death but by higher rates of revascularization in the medical (or stent-subtracted) arm. Take-home: Our beliefs can trump evidence and can impair our ability to conduct unbiased trials.
7. Fish Oil Finally Wins
But In REDUCE-IT, a high dose (4 g daily) of purified eicosapentaenoic acid used in patients with high triglyceride levels reduced a composite of major cardiac events by almost 5% in absolute terms. All components of the composite endpoint were significantly reduced. Also remarkable was that these patients were receiving statin therapy and had median low-density lipoprotein cholesterol levels of 74 to 76 mg/dL.
Three factors caused skepticism over REDUCE-IT: One was its outlier nature; another was the unknown mechanism of benefit; and the third was concerns about potential harm from the mineral oil placebo, which was associated with significantly higher levels of non-high-density lipoprotein cholesterol, high-sensitivity C-reactive protein, and apolipoprotein B.
My take is that the (purified) brand and dose of fish oil plus the selection of patients with high triglycerides were enough to explain the results. During an interview, Professor Jane Armitage from Oxford thought that REDUCE-IT would apply to about 30% of patients with high triglyceride levels. She reminded me of something doctors often forget: The best way to reduce one’s triglyceride levels is with diet and exercise. Prescription-strength fish oil will likely be expensive. We will see how this influences uptake.
8. Turn Down the Oxygen
When patients are sick and struggling to get enough nutrients to their organs, I have always cranked up the oxygen. This seemed like a no-brainer; hypoxemia is bad.
Researchers from McMaster University proved me wrong. In their meta-analysis and systematic review of 25 RCTs, which included more than 16,000 patients with sepsis, critical illness, stroke, trauma, MI, or emergency surgery, a liberal oxygen strategy was associated with an increase in mortality compared with conservative use. Low heterogeneity of the included RCTs and a meta-regression showing that trials using higher doses of oxygen had higher mortality rates bolstered confidence in the results.
Two messages: Whether the mechanism of harm from too much oxygen is due to free radicals, inflammation, or perhaps delayed recognition of deteriorating patients does not matter. What matters is that higher doses of oxygen cause harm and we should change our practice. This stunning reversal also confirms the value of using evidence in the care of patients. It reminds us of the wisdom of the late physicist Richard Feynman: “The first principle is that you must not fool yourself, and you are the easiest person to fool.”
9. The Good News About the PREDIMED Retraction
In the grand scheme of heart health, the food we eat todos los dias has greater importance than any drug or device. Nutrition evidence, however, is plagued by weak science. In the mass of confounded observational studies, the RCT called PREDIMED stood out as good nutrition science.
In 2013, PREDIMED authors reported that eating a Mediterranean diet supplemented with nuts or olive oil vs a regular Spanish diet led to a 30% lower rate of cardiovascular events. PREDIMED’s more than 2100 citations confirmed its influence.
When the New England Journal of Medicine (NEJM) retracted and republished a revised PREDIMED study this year because of irregularities of randomization, discovered by the British anesthesiologist John Carlisle, it was easy to be cynical. But I agree with outgoing NEJM editor-in-chief, Jeff Drazen, MD, who wrote that their “[PREDIMED] review did not alter any conclusions and should raise public trust in science, not erode it.”
Here’s why: John Carlisle’s work on analyzing baseline characteristics to assess for adequate randomization taught researchers and journal editors an important lesson. Also, the PREDIMED authors handling of the issues was exemplary; they investigated the problem, wrote clearly about it, and did vigorous statistical analysis to account for the irregular randomization. While the new analysis may weaken our confidence in the benefits of a Mediterranean diet, this story advances confidence in science.
10. AF Ablation Questions Remain
In 2018, two large RCTs addressed hard outcomes after atrial fibrillation (AF) ablation. You would think this much evidence would have provided clarity about a common procedure. I don’t think so.
The CASTLE-AF trial assigned about 360 patients with heart failure and AF to ablation or medical therapy. Ablation was associated with a 38% relative reduction in the primary outcome of death from any cause or hospitalization for worsening heart failure. This translated to a 16% reduction in absolute terms, with a number needed to treat of 6. Ablation lowered the overall death rate by 47% (13.4% vs 25%).
While this looks impressive, reasons for caution include that the trial was terminated before it reached its prespecified targets, the number of patients lost to follow-up was large, and the primary analysis was based on a handful of events. External validity of CASTLE-AF is also limited. Enrolled patients were 64 years old on average, were mostly male, and had a median left ventricular ejection fraction of 32%. For every patient like this, I see 20 who are older, sicker, and too frail for an invasive procedure.
Douglas Packer, MD, from the Mayo Clinic presented results of the CABANA Trial of AF ablation vs medical therapy in May, but the paper is not yet published. In this RCT of 2200 patients with symptomatic AF and at least one risk factor, the 14% lower rate of the composite endpoint of death, stroke, bleeding, or cardiac arrest in the ablation arm did not meet statistical significance.
The debate on CABANA turns on the analysis by treatment received, which markedly favored ablation. Because many patients in the medical arm crossed over to ablation, the question is how to square the nonsignificant intention-to-treat findings with the super-positive as-treated analysis.
I had oodles of fun this year. I’ve always loved being a doctor, but reading, writing, and thinking about medical science has made it even more gratifying. I feel blessed to work with the team of professionals here at theheart.org