FOURIER: Very Low LDL-C Post Hoc Analysis Doesn't Move the Needle

This one (almost) changed my mind about the proprotein convertase subtilisin-kexin type 9 (PCSK9) inhibitor drugs and the titrate-to-lower-LDL-C-level strategy.

Before Dr Robert Giugliano (Brigham & Women's Hospital, Boston, MA) presented this post hoc analysis of the FOURIER trial^[1] at the European Society of Cardiology (ESC) 2017 Congress , I was skeptical on both the clinical utility of the PCSK9 inhibitors and the titrate-to-lower-LDL-C strategy.

My thinking was this: Yes, the monoclonal antibody injections impressively lower LDL-C, but in the main FOURIER trial^[2], it took more than 27K patients to show a 1.5% reduction in nonfatal events. There was no signal for improved overall mortality.

The ultimate goal of preventive drugs for heart disease is to prolong life—not produce statistically significant reductions in composite end points. At the current price of these drugs, or really, at even a fraction of the current cost, they are way too expensive.

What's more, the FOURIER trial truncated at slightly more than 2 years. The  EBBINGHAUS safety study provided reassuring results on adverse events, but people don't take preventive therapy for just 2 years. Unknown is the only fair way to describe the long-term safety of these drugs.

The Post Hoc Analysis

FOURIER trialists aimed to explore the relationship between progressively lower LDL-C levels and efficacy and safety. The analysis was prespecified, but as pointed out by three Dutch editorialists^[3], the authors may have used a randomized study data set, but they evaluated the association between achieved LDL-cholesterol concentrations and risk using an observational study design (emphasis mine).

The authors split the subjects in the main trial into five groups based on LDL-C levels, from the lowest category of less than 0.5 mmol/L (19 mg/dL) to the highest category of greater than 2.6 mmol/L (100 mg/dL). Then they looked at efficacy and safety events in these categories, using the highest LDL-C category as the reference.

For efficacy, they observed a steady decrease in primary-outcome events with declining LDL-C levels. Basically, lower LDL-C produced fewer events, and this included levels down to 0.2 mmol/L (8 mg/dL). The main figure from the paper shows a steeply declining line with LDL-C levels on the x-axis and adjusted event rates on the y-axis.

For safety, they found no signal of adverse outcomes (after adjustment)—even at the lowest levels of LDL-C.

In the discussion, the authors make the case that these findings align well with many previous studies that suggest lower is better for LDL cholesterol.

What is provocative about these observations is the lack of a lower limit for LDL-C. As the LDL-C plummets, benefits keep accruing without a safety signal.

Reasons to Wait

Why do I remain skeptical? One is that PCSK9-inhibitor safety cannot be reliably assessed at 2.2 years. The authors state this in the paper. They also point out that there are two ongoing observational extension studies of FOURIER that will collect data up to 5 years. Good. We need that data.

The other reason not to overinterpret the benefits from the lower LDL-C quintiles is found in the appendix. On Table S3, the benefits in the primary end point don't translate to better survival. The adjusted hazard ratio for cardiovascular death and all-cause death are not significantly reduced—even in the lower two quintiles of superlow LDL-C levels.

Similar to the main FOURIER trial, PCSK9-inhibitor efficacy comes only in reduction of nonfatal events or the softer end point of coronary revascularization.

This finding raises doubt on the lower-is-better-LDL-C idea. Here we have large numbers of patients with extremely low LDL-C levels, and these patients are experiencing fewer cardiovascular events, but they're not living longer. Not even a signal of better survival. If serum LDL-C is so toxic, reducing it to these levels should prolong life.

Finally, a point about data display in that figure depicting a steep decline (negative slope) in event rates as LDL-C levels drop. The problem is the x-axis: the right side of the graph, the group with LDL-C >2.6 mmol/L [100 mg/dL] accounts for only 17% of the population but a high number of events.

The downward slope of the line for the other 83% of trial participants (those in the quintiles below 2.6 mmol/L [100 mg/dL]) is not as steep. Event rates (nonfatal ones) in these categories look to be within a few percentage points of each other, and the confidence intervals overlap.

Final Conclusion

Before embracing the expensive PCSK9 inhibitors or the aggressive strategy of using increasing doses of medications to achieve the lowest possible LDL-C, I think we need longer-term safety data and clear survival benefits. We are not there yet.

JMM

Cite this: FOURIER: Very Low LDL-C Post Hoc Analysis Doesn't Move the Needle - Medscape - Sep 01, 2017.