“…we should be very cautious extending trial results to patients unlike those enrolled in trials.” I fought similar battles some years ago during my time as a TPA/PPO Medical Director. A screening test had been developed in persons older than 49 years. Not long after this test came on the market, it was being offered to persons less than 50 years old. I said no… not covered.
COMPASS may have shown relative risk reduction of MACE but its absolute risk reduction was overridden by large increase in bleeding. One has to look beyond what the authors are paid to sell you.
Kudos to the investigators for being realistic about control group event rate (they actually underestimated it, a seemingly rare occurrence these days) to ensure trial was adequately powered.
This trial result is very interesting in light of the conceit from Compass being that certain subgroups (CKD, but also PAD, diabetics, and HF) accrued the most benefit. Voyager validates the PAD assertion. But this trial negates the CKD subgroup result.
I can’t recall if the subgroups in Compass were prespecified. But this serves as a good reminder that subgroup results ought only to be hypothesis generating in most cases.
But what is more is that so many preventive strategies have been proven to work in trial populations that look nothing like my average patient who is almost always higher risk of death from many causes.
An important reminder that treatment effects are not universally transferable across populations. TRACK showed that despite a higher baseline cardiovascular risk, low dose rivaroxaban provided no reduction in ischemic events and increased major bleeding in advanced CKD. This underscores the importance of treatment effect heterogeneity and validating benefit through dedicated randomized trials.
A great post about a well done study. The take-home point that the results of RCTs apply to only those patients who are most like patients in the trial is the most important.
Two other points that stick out in this well-done and depressingly negative trial:
1. No difference between CKD 4-5 and dialysis. The point estimate for the primary MACE outcome is identical in the subgroup analysis. It's not just that dialysis patients are different.
2. The study did show one miniscule benefit: fewer venous blood clots (4 in rivaroxaban vs 14 in placebo). And yet despite that finding, there was no benefit at all in dialysis access thrombosis (22 events in rivaroxaban group vs 16 in the placebo group). The blood thinner didn't even keep the dialysis access from clotting!
First, let's just point out that remarkable fact that 24% of the patients died in this trial that had a median followup of 1.7 years (and only 20% still at risk/followed at 3.0 years).
There were 354 deaths out of the 1458 people in the study. Ranked causes:
1. Sudden cardiac death (149, 42%)
2. Infection (64, 18%)
3(t). Other CV causes (60, 17%)
3(t). Other non-CV causes (60, 17%)
4. Unknown (21, 6%)
Once you get past the (presumed) arrhythmias and infections, you really only have about 1/3 of the causes of death left. Narrow that to (non-sudden) CV deaths (MI, stroke, CHF), and it's maybe 1/6 of the population. It's almost like trying to demonstrate a benefit in reducing CV events in patients with lung cancer.
Great analysis, as always
I would just note that the CKD subgroup where statins have not been effective is "CKD requiring dialysis" and not just "advanced kidney disease" :)
“…we should be very cautious extending trial results to patients unlike those enrolled in trials.” I fought similar battles some years ago during my time as a TPA/PPO Medical Director. A screening test had been developed in persons older than 49 years. Not long after this test came on the market, it was being offered to persons less than 50 years old. I said no… not covered.
COMPASS may have shown relative risk reduction of MACE but its absolute risk reduction was overridden by large increase in bleeding. One has to look beyond what the authors are paid to sell you.
Kudos to the investigators for being realistic about control group event rate (they actually underestimated it, a seemingly rare occurrence these days) to ensure trial was adequately powered.
This trial result is very interesting in light of the conceit from Compass being that certain subgroups (CKD, but also PAD, diabetics, and HF) accrued the most benefit. Voyager validates the PAD assertion. But this trial negates the CKD subgroup result.
I can’t recall if the subgroups in Compass were prespecified. But this serves as a good reminder that subgroup results ought only to be hypothesis generating in most cases.
So important!!!
But what is more is that so many preventive strategies have been proven to work in trial populations that look nothing like my average patient who is almost always higher risk of death from many causes.
What are we doing here folks????
An important reminder that treatment effects are not universally transferable across populations. TRACK showed that despite a higher baseline cardiovascular risk, low dose rivaroxaban provided no reduction in ischemic events and increased major bleeding in advanced CKD. This underscores the importance of treatment effect heterogeneity and validating benefit through dedicated randomized trials.
A great post about a well done study. The take-home point that the results of RCTs apply to only those patients who are most like patients in the trial is the most important.
Two other points that stick out in this well-done and depressingly negative trial:
1. No difference between CKD 4-5 and dialysis. The point estimate for the primary MACE outcome is identical in the subgroup analysis. It's not just that dialysis patients are different.
2. The study did show one miniscule benefit: fewer venous blood clots (4 in rivaroxaban vs 14 in placebo). And yet despite that finding, there was no benefit at all in dialysis access thrombosis (22 events in rivaroxaban group vs 16 in the placebo group). The blood thinner didn't even keep the dialysis access from clotting!
Thanks for writing up this study!
"competing risks of death"
First, let's just point out that remarkable fact that 24% of the patients died in this trial that had a median followup of 1.7 years (and only 20% still at risk/followed at 3.0 years).
There were 354 deaths out of the 1458 people in the study. Ranked causes:
1. Sudden cardiac death (149, 42%)
2. Infection (64, 18%)
3(t). Other CV causes (60, 17%)
3(t). Other non-CV causes (60, 17%)
4. Unknown (21, 6%)
Once you get past the (presumed) arrhythmias and infections, you really only have about 1/3 of the causes of death left. Narrow that to (non-sudden) CV deaths (MI, stroke, CHF), and it's maybe 1/6 of the population. It's almost like trying to demonstrate a benefit in reducing CV events in patients with lung cancer.