The cinacalcet randomized controlled trial: EVOLVE.
Speaking is the lead author Glen Chertow.
Disclosure: I (Joel Topf) was a primary investigator for EVOLVE.
The second trial to be presented at late breaking and high-impact trials is EVOLVE.
- Largest dialysis study
- Longest dialysis study
- Most international dialysis study
- Only double-blind dialysis study
- If you complain that nephrology does not have enough high quality studies you should pay attention to EVOLVE.
Cinacalcet was approved based on its ability to lower the PTH, the next question was, “does lowering the PTH with cinacalcet reduce mortality?” This was an investigator originated study that was sponsored by Amgen.
Dr. Chertow showed a slide with the hypothesis: “EVOLVE was designed to test the hypothesis that treatment with cinacalcet would reduce the risks of death and nonfatal cardiovascular events among patients with secondary hyperparathyroidism who were undergoing hemogdialysis.”
Patients were randomized 1:1 to cinacalcet or matching placebo. The trial was event driven, i.e., they ran the trial until they had 1,882 end-points. This was projected to take 3 years. It took 5.
The primary endpoint was a composite of all-cause mortality, or non-fatal cardiovascular event (MI, hospitalization for unstable angina, heart failure, or peripheral vascular event).
Secondary endpoints included: fracture, parathyroidectomy, CV death, stroke, and all the individual components of the primary end-point.
The next slide was a summary of the statistical approach. Note: when the authors present the statistical approach in a 12 minute presentation, buckle your seatbelt for a negative trial that, if you squint your eyes just the right way, looks positive.
Baseline characteristics looked well balanced; however, the cinacalcet group was older, 55 versus 54. Note: at a subsequent investigator meeting, the study designers shared that there was only an 8% chance that a 4,000 person study would find that large of a difference in ages. If they had to do it all over again they would put some age restrictions in so the potential age range would be smaller, decreasing the chance of the study being unbalanced. Foreshadowing: this will become more important as the story unfolds.
The first result they show are the biochemical parameters.
- Decrease in PTH (about 200 pg/ml lower)
- Decrease in calcium (about 0.4 mg/dl lower)
- Decrease in Ca x P product
- No decrease in phosphorous. Phosphorous has been about 10% lower with cinacalcet other trials. This did not materialize here.
Next slide primary endpoint: HR 0.93 P=0.112. Not significant. All-cause mortality HR 0.94 P=0.249. Not significant.
They then re-ran the data to balance the ages. Age adjusted relative hazard: 0.88 P=0.007. Multivariable (best fit) HR 0.88 P=0.008 and multivariable-adjusted (all included) 0.88 P=0.02. Wow. EVOLVE runs into an unbalanced age distribution and once the data is age adjusted they reach significance. Brutal bad luck. In terms of statistical-juke-and-jive, this is pretty mild. All they had to do was adjust the data for age and, boom, significant P-value.
Then they looked into why people stopped the drug. Time to first discontinuation of study drug for protocol-specified reason (transplant, parathyroidectomy, etc.) was balanced across both groups but for non-protocol reasons drop out was quicker and more common for placebo. This is unusual; usually side effects cause patients to drop out of the experimental group faster than the placebo group. Presumably this was due to patients dropping out of placebo due to uncontrolled PTH and doctors switching them to commercial cinacalcet. Problems only seen when doing an RCT of an approved drug.
Dr. Chertow then discussed lag-censored analysis. This is a modified approach to intension-to-treat analysis (ITT). ITT is the most rigorous approach but the designers of EVOLVE pre-specified a lag-censored approach. The idea here is that if a patient stops taking study drug, at some point in the future the effect of the drug could not explain an outcome. For example if a patient took the drug for a week, it is hard to accept that a parathyroidectomy 5 years later was due to the drug. The authors pre-specified a lag of 6 months. This means that any outcomes that occurred more than six-months after the last pill was consumed were ignored. The lag-censored analysis only censored people who stopped the drug for protocol specified reasons (transplant or parathyroidectomy were mentioned, there may be others). They found a HR for the primary outcome of 0.85 P=0.003. They found a HR for mortality of 0.83 P=0.009.
On to bone-mineral outcomes. Time to parathyroidectomy was longer with cinacalcet, HR 0.44 P<0.001. No difference in fracture risk.
Adverse events: there were a lot. 273 per 100 patient years. Wow. Serious adverse events were also found 53.3 per 100 patient years. The only ones that were significantly (P<0.001) more common with cinacalcet were unsurprisingly: All AE (273 vs 217), hypocalcemia (6.7 vs 0.9), nausea (18.3 vs 9.1), vomiting (15.4 vs 2.5).
Final thoughts, the drug failed its primary endpoint and it looks like one of the main reasons is they ran into an unbalanced distribution of ages. This was a larger, long and in terms of double blinding, unprecedented study in dialysis, there were a hundred reasons it could have failed, the fact that it was something as simple as unbalanced age distribution is kind of tragic.
Dr. Chertow mentions that EVOLVE is going to be published online in the New England Journal of Medicine and described the discussions with the editors as being “spirited” on more than one occasion. Sounds like there were some fireworks pounding the final manuscript together.
Post written by Dr. Joel Topf, eAJKD Advisory Board Member, and edited by Dr. Vinay Nair, eAKD Advisory Board member.