Oct 23, 2020 This Week in Cardiology Podcast

John M. Mandrola, MD


October 23, 2020

Please note that the text below is not a full transcript and has not been copyedited. For more insight and commentary on these stories, subscribe to the This Week in Cardiology podcast.

In This Week’s Podcast

For the week ending October 23, 2020, John Mandrola, MD comments on the following news and features stories.

The Lessons of Tocilizumab and COVID

COVID science offered up a big lesson on critical appraisal of evidence this week. The IL-6 blocker tocilizumab has been proposed as a treatment for COVID-19; it makes sense that an inflammation blocker might work.

On Tuesday, JAMA-IM reported a large multicenter cohort study that included nearly 4000 patients with COVID-19. the study question: Is tocilizumab associated with reduced mortality?

In a nonrandomized observational study, the authors compared 433 patients who received tocilizumab and 3491 patients who did not. To deal with the lack of randomization, the authors did adjustments to account for the substantial differences in baseline differences of the two groups. This was accomplished by inverse probability weighting, and after said IPW, the authors wrote that ... “baseline and severity-of-illness characteristics were well balanced between groups.”

In all, 29% treated with tocilizumab died vs 41% not treated with tocilizumab. The hazard ratio was 0.71 with CI of 0.56-0.92. Hugely significant

But in the same journal issue of JAMA-IM published two small RCTs, one from Italian authors and the other from a French team. The first included 126 patients with COVID pneumonia. The primary endpoint (PEP) was a composite of ICU admissions, death, or clinical worsening of oxygenation. In both groups, 17 patients had clinical worsening. No difference. Trouble here was the confidence intervals of the hazard ratio of 1.05 ranged from 0.59 to 1.86. That means tocilizumab could have reduced the composite by 41% or worsened it by 86%.

The second RCT enrolled 131 patients; half got tociizumab and the other half usual care. They used a composite PEP combining a score > 5 on a 10-point clinical progression score and survival without need for ventilation. They used a Bayesian-type of analysis and reported that the probability of a 20% lowering of this combined endpoint was 83% But there was no difference in mortality; seven vs six patients died in each group.

Now to the NEJM RCT. A Harvard-led group enrolled 243 patients with COVID-19 pneumonia to tocilizumab vs standard of care. It was a 2:1 randomization, so 161 received active treatment vs 82 controls. The PEP was intubation or death. In all, 10.6% of the tociizumab group reached the PEP vs 12.5% of the controls. The hazard ratio was 0.83 but again the confidence interval ranged from 0.38 to 1.81. Thus, tocilizumab could have reduced the occurrence of a bad outcome by 62% or worsened it by 81%. Nine of 161 patients in the tocilizumab arm died vs three of 82 patients on placebo. Hardly a good signal.

The conclusions were thus: Tocilizumab was not effective for preventing intubation or death in moderately ill hospitalized patients with COVID-19. Some benefit or harm cannot be ruled out, however, because the confidence intervals for efficacy comparisons were wide. Yet another underpowered trial.

On Twitter, Professors Sanjay Kaul and Frank Harrell had an enlightening conversation about how big a study needs to be to feel comfortable excluding a benefit.

Kaul wrote that the trial in the NEJM had an unrealistic sample size estimation based on exuberant expected effect size. “...The Trial was powered to detect an implausibly large effect size (50% RRR from 30% placebo event rate). But the actual placebo event rate is only 12.5%!”

Frank Harrell replied that for sample size calculations, effect size should be based on the clinical effect one would not want to miss.

The problem is that if you say you’d like to be able to detect a 20% reduction in events, you would need 4 times more patients. Having more patients in trials is not as easy as it sounds. For cost and practical reasons, trials have to make compromises. One potential solution is to combine small trials into a meta-analysis.

Turns out there is another RCT of tocilizumabpublished on a preprint server, from a group at Baylor This trial randomly assigned 452 patients to tocilizumab or placebo. Clinical status at day 28 was not statistically significantly improved for tocilizumab vs placebo (P=0.36). There was no difference in mortality at day 28 between tocilizumab (19.7%) and placebo (19.4%)

Mike Johansen is an academic family medicine professor from Ohio who published a rough meta-analysis on Twitter. He looked at overall death. Combining the results of the four RCTs, now the hazard ration is 1.05 and the confidence interval tightens to 0.75 to 1.50.

Roche put out a press release saying that their multinational-placebo sponsored trial of tocilizumab (EMPACTA) found no statistical difference in mortality.

Cardiologists can learn a ton from COVID19 science. This is classic. A large observational study shows an association of a drug with improved survival; the effect size is massive. But four RCTs all find no benefit. Each was underpowered and could have missed an effect, but taken together, plus the press release, we can safely conclude that tocilizumab, a drug that costs about $3000 per dose, offers no benefit.

Hypertrophic Cardiomyopathy

The combination of the pandemic and SGLT2 inhibitor trials distracted attention from an intriguing RCT of a novel oral agent for patients with hypertrophic cardiomyopathy (HOCM).

EXPLORER-HCM is a placebo-controlled trial studying the novel oral medication called Mava-Cam-Ten, a selective inhibitor of cardiac myosin-ATP-ase that works by reducing actin-myosin cross-bridge formation. That action reduces contractility and improves myocardial energetics.

In 68 centers in 13 countries, about 250 patients with HOCM with a left ventricular outflow tract (LVOT) gradient of ≥ 50 and class 2-3 symptoms received Mava-Cam-Ten or placebo for 30 weeks. The PEP was a 1.5 ml/kg/min increases in VO2 max (maximum rate of O2 uptake during exercise) and an improvement in NY Heart Association class or a 3 ml/kg/min increase without a functional class improvement. The results were positive:

  • 37% of patients on Mava-Cam-Ten vs 17% of patients on placebo met the PEP.

After 30 weeks, the post-exercise LVOT gradient decreased more in the Mava-Cam-Ten arm. The absolute difference between groups was about 36 mmHg.

Science like this makes cardiology fun. Unlike beta-blockers or calcium channel-blockers, which act to reduce heart rate and contractility, here is a novel drug pinpointing a more central pathophysiological issue—the myosin-actin interaction. And it seems to work. If blinding was good, the drug had modest effects on exercise tolerance. More patients achieved an improvement in exercise capacity and/or symptom relief. And you don’t see a signal of adverse events.

But...huge caveats. They screened more than 420 to find 250 eligible. Patients were mostly white. Average age nearly 60; so, an older group. Most had NYHA class 2 symptoms, so not that symptomatic. A mean improvement of 1.4 on VO2 max is very modest. NYHA class did not improve in about 1 in 3 patients on Mava-Cam-Ten; nearly half still had gradients > 30.

As pointed out in a column from surgeons Patrick Myers and Edward Quintana, septal myectomy is much more effective in reducing gradients and has been associated with long-term symptom reduction.

As for long-term safety of MavaCamTen, we cannot say much from a 30-week study of a highly-selected cohort. This drug is a negative inotrope. That should give us pause about long-term use.

Brain Microbleeds

I visit our neuroradiologists a lot. As an atrial fibrillation specialist, I am often asked to decide on anti-thrombotic strategies in patients who have had a stroke. One of the things we see often is an MRI that shows cerebral microbleeds, markers of cerebral small vessel disease and present in one-third of patients with ischemic stroke

The risk of future intracranial hemorrhage (ICH) increases as the number of cerebral microbleeds increase. This raises questions about the safety of oral anticoagulation (AC). In absolute terms, the risks of recurrent ischemic strokes is much higher, but the asymmetry of an ICH is the issue. Bleeding in the brain while taking oral AC is terrible.

JAMA-Neurology published a sub-study of the NAVIGATE-ESUS trial, which compared rivaroxaban vs aspirin in patients who had an embolic stroke of unknown source. There were 3500 patients per group. The trial was stopped early for futility—rivaroxaban was no better than aspirin.

But researchers looked at a subgroup of patients with baseline MRI data (roughly 3700 patients). Microbleeds were noted in 11% or nearly 400 of these 3700 patients. Having cerebral microbleeds (CMB) increased the odds of future ischemic stroke, ICH, and all-cause mortality. That is not news. The finding that drove positive comments from investigators—who should know better—was this: “there was no notable trend for greater risk of ICH with rivaroxaban (vs ASA) in patients with CMBs compared with those without CMBs (P > .99 for interaction.” Here is the problem: in a trial of 7000 there were only 15 ICH events. And in this non-random substudy, there were only 8 ICH events. Eight. You can’t tell anything from that. This observational analysis of 8 events gives no reliable information on the risk of ICH and direct oral AC use in the presence of CMB.

Surrogate Endpoints and the Slender Stent

At the Transcatheter Cardiovascular Therapeutics meeting recently, I was drawn to a study comparing a novel, super low-profile stent. What drew me to this non-inferiority (NI) study was how we interpret evidence of whether something is as good as an established comparator. If a stent is just as good as another stent, you want to prove it is superior to those used now; if it offers some other advantage, say, crossing difficult lesions or an easier implant process, then you go for noninferiority.

In a trial called OPTIMIZE the slender stent was compared with standard stents. The endpoint was TLF – target lesion failure, and this was a composite of cardiac death, target-vessel MI, creatine kinase-MB or troponin > 3 times the upper limit of normal within 48 hours, and clinically driven TL revascularization.

The rate of events was 10.3% for the slender stent and 9.6% for the standard. The upper bound of the confidence interval exceeded the NI of 3.5% and the slender stent did not make NI.

The problem: there were many more events than predicted. And most of the events in both arms were due to the biomarker-sensitive definition of target vessel MI. If you have more events than predicted, and you are using fixed NI margin, it is a harder to reach NI, unless the trial has massive numbers.

The principle investigator (PI) and all the discussants, each agreed that biomarkers post procedure are not clinically relevant. The PI showed an analysis looking at events that excluded the biomarker-defined endpoint, one that included, events that were felt more clinically relevant and that one easily showed the slender stent was non-inferior.

Why is this important? Because this is exactly how the EXCEL trialists showed that left main stenting was equivalent to surgery. They a) included periprocedural MI, which the NOBLE investigators did not, and b) they used a biomarker-heavy definition of clearly biased against surgery—as surgery will always cause more enzyme release.

Surrogate endpoints are necessary, but, my friends, you have to make sure they are clinically relevant and consistent.


Comments on Medscape are moderated and should be professional in tone and on topic. You must declare any conflicts of interest related to your comments and responses. Please see our Commenting Guide for further information. We reserve the right to remove posts at our sole discretion.
Post as: