An embarrassment of riches this week as we got not one, but two randomized clinical trials evaluating the timing of dialysis initiation in acute kidney injury. Of course, the results don’t agree at all. Back to the drawing board folks. For the video version of this post, click here.
OK here’s the issue – we nephrologists see hospitalized people with severe acute kidney injury – the abrupt loss of kidney function – all the time. There is always this question – should we start dialysis before things get too bad, in order to get ahead of the metabolic disturbances, or should we hold off, watch carefully, and jump when the time is right? Several people – and, full disclosure, I’m one of them – have examined this question and the answers have been confusing. The question begs for a randomized trial.
And, as I mentioned, we have two. One, appearing in the Journal of the American Medical Association, says yes, earlier dialysis is better with mortality rates of 40% in the early arm versus 55% in the late arm. The other, appearing in the New England Journal of Medicine says there is no difference – mortality rates were 50% no matter what you did.
Figure 1: JAMA Trial - GO TEAM EARLY!
Figure 2: NEJM Trial - D'Oh!
Sometimes, rival movie studies put out very similar movies at the same time to undercut each other’s bottom line. So which of these trials is Deep Impact, and which is Armageddon? Which is Ed TV and which is the Truman show? Which is Jobs and which is Steve Jobs?
Figure 3: It's like looking in a mirror...
In this table I highlight some of the main differences:
The NEJM trial was bigger, and multi-center, so that certainly gives it an edge, but what draws my eye is the difference in definitions of early and late.
The NEJM study only enrolled people with stage 3 AKI – the most severe form of kidney injury. People in the early group got dialysis right away, and the late group got dialysis only if their laboratory parameters crossed certain critical thresholds. The JAMA paper enrolled people with Stage 2 AKI. In that study, early meant dialysis right after enrollment, and late meant dialysis started when you hit stage 3.
OK so definitions matter. The NEJM trial defined early the way the JAMA trial defined late. So putting this together, we might be tempted to say that dialysis at stage 2 AKI is good, but once you get to stage 3, the horse is out of the barn – doesn’t matter when you start at that point.
That facile interpretation is undercut by one main issue: the rate of dialysis in the late group.
See, one of the major reasons you want to hold off on dialysis is to see if people will recover on their own. In the JAMA study, enrolling people at stage 2 AKI, only 10% in the late group made it out without dialysis – and those people were dialyzed, on average, only 20 hours after randomization. In the NEJM study, using the more severe inclusion criterion, roughy 50% of the late group required no dialysis. To my mind, if 91% of the late group got dialysis, you’re not helping anybody – the whole point of not starting is so that you never have to start, not that you can delay the inevitable.
Regardless of your interpretation, these studies remind us not to put too much stock in any one study. They should also remind us that replication – honest to goodness, same protocol replication – is an incredibly valuable thing that should be celebrated and, dare I say, funded.
For now, should you dialyze that person with AKI? Take heart – there’s good evidence to support that you should keep doing whatever you’ve been doing.
Acute kidney injury (AKI) is a major killer in the US. At least, that's what the data shows. But a lot depends onhow you diagnose AKI. Our latest study suggests that natural variation in lab measurement of creatinine is leading to strikingly high false-diagnosis rates, and that the association we see between AKI and bad outcomes like mortality may be an artifact of the fact that sicker people get more blood draws. Take a look at our full paper here.
And if you want to run the simulation for yourself (and you have stata), you can get our code and relevant files here.
The full paper is online, but the quick version is that alerts for AKI did not improve clinical outcomes for AKI patients. Not a bit. What's worse, in at least one group of patients, the alerts increased the rate of renal consult and the rate of dialysis. For me, this was a lesson in unintended consequences, and a reminder of why randomized trials are so important.
See the full story at The Lancet: