Christos Argyropoulos MD, PhD, FASN 🇺🇸 0kale/accThe PISCES trial creates some ...
The PISCES trial creates some a interesting conundrum for D&I. On the one hand we have a very solid result found in a well executed trial. On the other hand we have an intervention that has not… | Christos Argyropoulos
The PISCES trial creates some a interesting conundrum for D&I. On the one hand we have a very solid result found in a well executed trial. On the other hand we have an intervention that has not dazzled the world in non-dialysis settings and despite the credible arguments about dose and prevailing "n3 deficiency in ESRD) these previous underwhelming results make nephrologists skeptical. Skepticism demands replication, but how can one replicate? Here are my thoughts: First, the intervention: the fish oils used in replication should be of the same amount (4g/day) and composition (EPA/DHA ratio of 2:1 to deliver at least 2.4 of the former and 0.8 g of the latter) Second, the outcome: it makes little sense to power against a precise relative risk since this would inflate the sample size. Instead, I would power against a directional outcome: the intervention should deliver a benefit at least as great as things as we adopted in the past. For example high flux dialyzers were eventually associated with reductions in cardiovascular and all cause mortality of around 15 to 20%. Third the design. I presume everyone will like to run this on the cheap by adopting a cluster randomized design but this would be a non-smart design in my opinion for a number of reasons: there are still seasonal and covid / flu related variations in morbidity and mortality, so unless you can get the right units enrolled and delivery synchronized, you will end up confounding results with these trends. Even in the absence of such trends, variability in mortality among dialysis units is HUGE , so why shoot yourself in the foot by basing the results on a design that is open to this confounding? Finally variation in diet and intake of n3 by region would also unnecessarily increase variation and noise in the results. For these reasons, a traditional design that randomizes in strata of units would be preferable. Having done the tedious work of armchair thinking, when can we start the replication?
Christos Argyropoulos MD PhD
VoidedMike
