The latest salvo in the hyponatremia wars takes off where Seethapathy left us (NEJM Evidence | NephJC). Last week, Juan Carlos Ayus published his latest study (JAMA Intezrnal Medicine) on hyponatremia, a meta analysis of patient outcomes in the management of hyponatremia.
If you are not aware of Ayus’ prior hits, you should take a look.
Retrospective data that looks at speed of correction and uses that to look at outcomes has been a staple in hyponatremia research for decades. It is the basis for the current, decades-old, hyponatremia guidelines from the US and Europe. However, until recently the only outcome of interest was osmotic demyelinating syndrome (ODS, and boy is that a loaded term that probably should be re-litigated, see Seethapathy’s grand rounds in Ottawa, YouTube). When looking at ODS, a pattern emerges that slower correction is associated with fewer cases of hyponatremia, however it is unusual to look at a case series and only care about one outcome. If we pull back a bit and look at all the outcomes that matter, things like length of stay and mortality, the situation changes. In Ayus’ meta-analysis, the primary outcome was mortality and just like in Seethapathi, slower correction was associated with increased mortality. Not a great look. And I think the orders of magnitude are important here.
Patients that experience faster correction of sodium consistently have better clinical outcomes in these sorts of analysis. And the rate of ODS is vanishingly low, expecially with the looming specter of death. The focus on ODS makes us ignore what an ominous sign hyponatremia really is. I can’t think of an electrolyte with such a frightening association with mortality.
The obvious weakness with this type of analysis is they do not provide any insight as to why the correction was slow. While we like to think that the slow correctors are populated by patients with bespeckled nephrologists carefully calculating sodium and water prescriptions according to the Edelman formula, but the reality is that the patients with slow correction is populated by people with liver and heart disease that are not easy to fix. And their prolonged length of stay and poor outcomes are driven by these dismal diagnosis.
And what of the rapid correctors? This cohort gets better quickly, not because cowboy nephrologists are slinging 3% saline to rapidly bring the sodium to heal, no, rather these are patients whose body wants a normal sodium and as soon as it is able to reject the excess water it will return the sodium to normal. And this is often despite the best intention of the treatment team. In Sterns’ study on the DDAVP clamp, 25% of patients in the control group (historic controls for their retrospective case series) made more than 1200 ml of urine an hour.
The slow correctors is populated by patients who are so ill that their bodies have rejected sodium and water homeostasis in the name of perfusion. Not a good place to be, hence the bad outcomes. While the fast correctors is populated by people whose physiology remains committed to sodium and water homeostasis. It is not hard to see why disease so severe that it rewrites the laws of homeostasis would have increased rates of devastating outcomes.
More simply the results are confounded.
Does the confounding explain all of the excess mortality?
Is it that far fetched to believe that the low sodium itself could contribute at least a little to the excess mortality seen with hyponatremia? It seems likely that something the body spends so much energy trying to keep regulated would be important and have an effect on outcomes. And given the rarity of CPM/ODS, even a small residual effect would swamp the concern for CPM/ODS because not only is ODS, rarer than we were lead to believe, the outcomes of ODS are not as bad as we were taught.
In Jason George’s study of nearly 1,500 people with sodiums less than 120, they had 9 patients with ODS on imaging. In terms of neurologic outcomes among these 9, “five patients with documented osmotic demyelination had recovery with no neurologic deficits, two patients died from unrelated causes, and two were lost to follow-up.”
When you look at Ayus’ meta-analysis of hyponatremia are you sure there is no residual signal? Because the people promoting ever slower rates of correction to avoid CPM feel that there is nothing on the other side of the balance pan. That slowing the rates of correction will always be justified. That it is okay to prolong hospitalization by any length of time. That we should adopt interventions like DDAVP clamps without any prospective data because the risk of ODS is so important that concern for this complication should drive the therapeutics in hyponatremia.
And while we are adopting DDAVP, we should restrict and avoid tolvaptan, not because it causes ODS, there were no cases in Shrier’s Phase three SALT1 and SALT2 trials, but because it may correct the sodium faster than guidelines suggest. Now we are not even worried about ODS, but rather the purported risk factor for ODS, speed of correction.
We adopted the risk factor rather than the outcome in the case of hypertension. We want to avoid the stroke and CV death, so we treat the blood pressure to a target BP. But the hypertension guys didn’t just look at retrospective data. They went out and did the work to see if treating blood pressure avoided the outcome, and it did. We should demand similar certainty for the treatment of hyponatremia.
We no longer should accept retrospective observational data. The Hyponatremia Intervention Trial (HIT, Protocol and rational for design publication in PubMed) showed that we could do prospective, randomized trials, in the treatment of hyponatremia. The study was not positive, but it was important. Here are my tweets from the Late Breaking and High Impact Clinical Trial Session at Kidney Week where the results were announced. We are still awaiting publication.
It is time to demand this. Hyponatremia is too common and for us to trust that the mortality signal is entirely a statistical mirage.