Sanjay Kaul Grades Four Big Trials From ACC 2018

Barbershop BP Study, ODYSSEY, VEST, CANTOS-Diabetes

Sanjay Kaul, MD


March 16, 2018

Sanjay Kaul, MD

Here is my Consumer Reports–like star grading (one star represents a low grade and five the highest grade) for four of the late-breaking clinical trials from the 2018 American College of Cardiology (ACC) Scientific Sessions. The grading is based on three key attributes: the quantity of evidence as assessed by the effect size and statistical persuasiveness (robustness); the quality of evidence (well-designed, adequately powered trial with excellent trial conduct, including low amount of missing data and prespecified statistical plan); and highly desirable benefit-risk balance. Cost is a secondary consideration in this calculus.

Black Barbershop Study (★★★★)

Description: Barbershop-based healthcare and BP reduction in black men

In a community-based, open-label, cluster-randomized trial conducted in black-owned barbershops in the Los Angeles area, pharmacist-guided BP-lowering intervention was more effective in reducing systolic BP (SBP), the primary endpoint of the study, compared with the control intervention, which comprised encouraging lifestyle modification and doctor appointments by the barbers.[1] Even though the study was relatively small (N = 319), the effect size is impressively large (21.6 mm Hg and 14.9 mm Hg reductions in SBP and diastolic BP, respectively) and statistically persuasive (P < .001), with a highly favorable benefit-risk balance.

Nearly 64% of participants in the intervention arm compared with 12% in the control arm achieved the 2017 ACC/AHA BP guideline-recommended target of < 130/80 mm Hg.[2] Given the open-label design, the possibility of confounding leading to an overestimate of treatment effect cannot be completely ruled out.

Bottom line: Overall, this trial is an exemplar of implementation science at its best and arguably the most important and impactful trial presented at the ACC this year!

ODYSSEY (★★★★)

Description: Alirocumab (PCSK9 inhibitor) versus placebo in patients with elevated lipids after acute coronary syndrome (ACS) on maximally tolerated statin therapy

Although the treatment effect is modest (1.6% absolute risk reduction [ARR] and 15% relative risk reduction [RRR] in major adverse cardiac events [MACE]) and primarily driven by nonfatal endpoints (myocardial infarction [MI], ischemic stroke, and unstable angina), the P value is quite robust (P = .0003) and is consistent with the "substantial evidence" criterion to support a cardiovascular indication based on a single trial. The trial execution was excellent, as reflected in the high retention rate, with over 99% ascertainment for the primary endpoint and 99.8% ascertainment for vital status.

The lack of increased risk for incident neurocognitive and diabetes adverse events is reassuring; however, the follow-up might not be sufficiently long to completely allay potential concerns. Nonetheless, the overall benefit-risk profile is desirable enough to justify approval for the cardiovascular indication by the regulatory bodies.

ODYSSEY vs FOURIER. How does ODYSSEY[3] compare with FOURIER,[4] the other major PCSK9 inhibitor (evolocumab) outcomes trial? Despite some key differences in trial design (treat-to-target approach utilized in ODYSSEY), demographics (higher-risk post-ACS cohort in ODYSSEY), primary endpoint (relatively more robust composite endpoint in ODYSSEY), and the duration of median follow-up (longer at 34 versus 26 months in FOURIER), the magnitude of LDL-C lowering (mean reduction of approximately 60%) and reduction in the primary endpoint (15% RRR) are similar in the two trials.

However, there are two key differences. First, ODYSSEY reported a 0.6% ARR and 15% RRR in all-cause deaths (nominal P = .026). In contrast, FOURIER reported a nonsignificant 0.1% absolute risk and a 4% relative risk increase in all-cause deaths. Second, prespecified subgroup analysis by baseline LDL-C in ODYSSEY appeared to suggest a greater treatment benefit in MACE reduction (ARR of 3.4%, number needed to treat [NNT] = 29) in those with LDL-C ≥ 100 mg/dL compared with two other tertiles (interaction P = .09). A post-hoc analysis showed 1.7% ARR in all-cause deaths (NNT = 60) in the group with baseline LDL-C ≥ 100 mg/dL (interaction P = .12), comprising nearly 30% of the overall trial population. In contrast, no heterogeneity of treatment effect was observed in FOURIER across baseline LDL-C quartiles (interaction P = .69). Similar lack of heterogeneity in treatment effect has also been reported in IMPROVE-IT[5] (ezetimibe plus statin versus statin) and in other statin trials.

Clinical, Regulatory or Payer Implications

Because of the testing hierarchy used and the lack of control for type 1 (false-positive) error, the reduction in all-cause death in ODYSSEY should, strictly speaking, be viewed as exploratory. Such hypothesis-generating findings ideally require confirmation in future studies. While it is tempting to view mortality as the "ultimate" endpoint, it must respect the laws of probability. The likelihood that the mortality benefit in ODYSSEY represents a spurious (false-positive) finding due to a play of chance cannot be ruled out.

In regard to the subgroup analysis based on baseline LDL-C, it is important to note that the interactions reported in the trial were not adjusted for multiple comparisons, thereby inflating the likelihood of false-positive results. Lack of positive interaction reported in previous lipid-lowering trials would appear to support this. Conversely, subgroup analyses lack the statistical power to capture true positive interactions and are thus also prone to false-negative results.

Thus, given these caveats, there is insufficient evidence to use these results to reliably inform and guide regulatory, clinical, or insurance/payer decision-making.

Cost Implications

The 1.6% ARR in MACE in ODYSSEY translates to an NNT of 64. So, at a retail annual cost of $14.5K, the cost per MACE avoided is $2.56M: 64 (NNT) x 2.8 (median follow-up in years) x 14,500. With a 50% or 75% discount, the price comes down to $1.28M or $0.64M. Even if we limit the use to those with LDL-C ≥ 100 mg/dL (which, as I argue above, is not statistically or clinically justifiable) and discount the price by 50%, the cost per MACE avoided of $0.59M might still be deemed cost-prohibitive. If the reduction in mortality were a robust and credible (statistically reliable) finding, one might consider the cost-effectiveness more favorably.

Bottom line: Does ODYSSEY move the needle for PCSK9 inhibitors? Not really! I find it hard to see how these results will persuade the payers to be more favorably disposed to this drug, unless they are convinced of the mortality benefit and the price is substantially discounted to amounts recently suggested by formal cost-effectiveness analysis by the Institute for Clinical and Economic Review ie, $2306-$3441 (assuming observed reduction in major CHD events in all eligible patients, which provides the most reliable and statistically valid estimate). The announcement by the sponsors to consider such a price discount is a positive first step in this direction, although the devil will be in the details. Only time will tell if such a prudent move will materialize to resurrect the potential of these drugs in lipid management.


CANTOS Diabetes (★)

Description: IL-1-beta inhibition with canakinumab and prevention of incident type 2 diabetes in a prediabetic population

I had previously given the principal results of the CANTOS trial a three-star grading because of the modest effect size in reducing MACE, which barely met the multiplicity-adjusted threshold of statistical significance, an increased risk for fatal infection, and the prohibitive cost. This study addressed a prespecified key secondary hypothesis: whether IL-1-beta inhibition with canakinumab would prevent incident type 2 diabetes in a prediabetic population.[6] Despite significant lowering of inflammatory biomarkers (hsCRP and IL-6), IL-1-beta inhibition using canakinumab over a 5-year period did not limit the development of new-onset diabetes among 4960 patients with protocol-defined prediabetes at trial entry (49.3% of overall CANTOS trial).

These null findings challenge the inflammation hypothesis of diabetes. They also fail to confirm previous findings of improved glycemic control observed with anakinra, an IL-1 receptor antagonist, in a randomized study of 70 patients with type 2 diabetes.[5] Of note, baseline concentrations of hsCRP and IL-6 predicted the onset of type 2 diabetes in CANTOS, but lowering these inflammatory biomarkers via canakinumab failed to reduce incident diabetes, suggesting that the association between these biomarkers and diabetes is not causal. Alternatively, non-IL-1-beta-mediated inflammation might play a more important role.

Bottom line: In my opinion, these results weaken the "common soil" hypothesis linking type 2 diabetes and cardiovascular disease via inflammation.

VEST (★)

Description: Wearable cardioverter defibrillator (WCD) and sudden death in the immediate post-MI period (< 90 days) in patients with reduced LVEF, as a bridge to evaluation for ICD

A nonsignificant 0.8% absolute risk decrease in the primary endpoint of sudden death in an open-label RCT that was not sham controlled is decidedly a null result, in terms of both the quality and the quantity of the evidence. Of the nearly 1500 patients using the device, there were only 20 appropriate shocks and 10 inappropriate shocks, with two- to fourfold greater incidence of rash and itching discomfort compared with control. The investigators and sponsors have emphasized the results on total mortality, a secondary outcome, which favored the WCD arm (3.1% vs 4.9%; P = .04). Is this justified?

Previous experience has demonstrated that when a trial fails to win on the primary endpoint but wins on a secondary endpoint, including mortality, the latter is likely to represent a spurious finding because of lack of control of type 1 (false-positive) error ("all the alpha spent on the negative primary endpoint"). Thus, a conservative approach is warranted even when the observed benefit is a reduction in a seemingly unbiased endpoint (ie, mortality). Furthermore, the most striking difference in mortality advantage for WCD in VEST is due to stroke-related death (0 vs 4; P = .01), which is difficult to explain. In fact, omitting stroke-related death renders the total mortality advantage no longer statistically significant, indicating the fragility of the results.

Bottom line: The benefit-risk ratio is clearly undesirable, thereby earning VEST[7] a low grading. Cost is an important consideration that further dampens my overall enthusiasm for this device. The statement by the principal investigator that "Despite a negative result for sudden death, based on the associated lowered total mortality, it is reasonable to prescribe the WCD in patients who are post-MI and have reduced ejection fraction, until further evaluation for an ICD] at 40 to 90 days" is unjustified and clearly represents a case of enthusiasm exceeding the evidence.


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: