The Chloride Case
It's only a ghost tale. But you should read it before having popcorn.
Critical care is a specialty haunted by ghosts. We have created a safe home for them to grow and thrive in our ambivalent minds. On the one side, we are probably the most objective, precise, obsessive, controlling type of physicians. On the other side, our syndromes are vaguely described. The Chloride Case, as I dubbed the ghostly proposition that mild hyperchloremia kills patients, found a home in this ambivalence.
Chloride is among the most abundant ions in the extracellular space. As you know, chloride is integral to the study of metabolic acidosis. Kidneys exert tight control of Chloride levels. However, kidneys can concentrate the anion in urine only to a certain amount. If you are worried about hyperchloremia, give your patient some water.
Pre-clinical studies showed that fluids high in chloride like normal saline (NS, 0.9% NaCl) are associated with reduced renal perfusion, enhanced inflammatory mediators, and worse hemodynamics, compared to fluids not high in chloride (balanced solutions). An elegant study by Chowdhury et al. included 12 volunteers to receive either NS or a balanced solution in a double-blind, cross-over trial and measured renal artery flow velocity and renal cortical perfusion. The study found a significant reduction in both parameters in the NS group. However, they found no differences in serum levels of NGAL, a biomarker of kidney injury, a finding ignored by the chloride community.
Biological plausibility was achieved by the aforementioned and several other preliminary studies. Retrospective and even small prospective clinical studies popped up in the medical literature, with inconsistent findings. However, the “additional grams of chloride cause death” hypothesis gained momentum and credulous intensivists merrily embarked on the train, blind to the marginal utility of the intervention. In a complex clinical environment such as critical care, the causes of death are plenty. How many ICU deaths are attributable to hyperchloremia? Is hyperchloremia a cause of death, or do patients die because of the disease that took them to the ED or ICU in the first place?
The hypothesis fails to identify a dominant cause of death amenable to prevention by avoiding hyperchloremia. Furthermore, hyperchloremia is not even a dominant cause of renal failure. It is a marginal cause of kidney hypoperfusion, which in turn may or may not contribute to renal failure. Treating (or preventing) a marginal cause yields marginal, insignificant effects. Researchers mistook biological plausibility for clinical relevance. But the train had already departed. And the large RCTs arrived.
The BaSICS trial concerns the perils of modest levels of hyperchloremia. It randomized patients in circulatory shock to receive emergency resuscitation with NS or a balanced solution. The intervention, so to speak, translates into two additional grams of chloride per liter of crystalloid. Amazingly, BaSICS somehow assumes that many people could die up to 90 days after receiving a few extra grams of chloride!
The total chloride exposition on the first day was around 3 g higher in the saline group and then reduced in the following days. I could not unknow it since. Every time I see my Brazilian terrier drinking from the swimming pool, I start mental calculations of how much chlorine he swallowed. But I admit I never get it. It's too much math for me! Anyway, next time you have popcorn, remember a teaspoon of table salt has 3 g of chloride.
The idea of giving the patients the amount of chloride in a teaspoon of salt and measuring mortality in 90 days will make any list of the most laughable study designs ever conceived. Of course, the BaSICS trial showed no mortality difference. But it was not the only, and not the first study. PLUS and SMART studies make good companions. Both randomized patients to NS or a balanced solution. SMART showed no difference in mortality as well as in two renal outcomes.1 PLUS yielded outcomes that are so evenly distributed that the outcomes table looks like Table 1.
Do you, my innocent reader, think that three large, high-quality RCTs showing no mortality benefit linked to a ridiculous intervention are enough to bury the question?
NEVER EVER underestimate the human capacity of milking a database. As you may know, a meta-analysis quickly emerged in the Chloride Case. The meta-analysis included several RCTs besides the triad BaSICS PLUS SMART, summoning data from more than 35,000 patients. Considering the combined data, the 95% confidence interval for the relative risk for mortality spanned from 0.91 to 1.01, which does not refute the null hypothesis. There is no compelling evidence of benefit or harm. Finished? Can the ghostbusters declare there is no ghost here?
Not yet! Besides presenting the traditional frequentist approach, which considers the binomial nature of the outcome (dead or alive) and reports incidences, the meta-analysis also presents a Bayesian spin that suggests a high probability of mortality reduction. It has to be so ethereal a reduction that it remained invisible after randomizing tens of thousands of patients. Ghosts are invisible, aren't they?
And this is the strange end of our ghost tale. The ghostbusters captured no ghosts, frustrating the audience. But ghostbusters know there is something out there! So, at the end of the expedition, they present some noise produced by their extravagant devices as evidence of things they only imagined. ICU nights are long and dark, a prominent surgeon once told me. The precise kind of place where ghosts like to hide…
Subscribe and share!
The SMART trial did show a significant difference in the primary outcome, a composite outcome comprising death and two renal outcomes, although there was no significant difference in any of the three components separately. Ok, that's what composite outcomes are for. That was SMARTER indeed, however, the difference is ethereal: it won't appear on a plain contingency table. It requires fancy statistical models to materialize.