The art of diuresis

Most ICU patients initially receive a lot of fluid. Later, they need to get rid of a lot of fluid.

This arc (a rough reciprocal of the course of critical illness itself) is older than you or me, and in days of yore gray-bearded sages would opine that patients need to “swell to get well, then pee to be free.” Nowadays, question has increasingly been raised as to whether patients benefit from quite so much fluid resuscitation, as hypervolemia clearly has significant sequelae—promoting pulmonary edema, degrading the vascular glycocalyx, even worsening organ perfusion via venous congestion—and it’s probably true that extracting bad fluid later is not an excuse for giving too much up front.

Nevertheless, even when clinicians maintain a light hand on the IV bag, it’s inevitable that many critically ill patients will wind up with many excess liters of total body water. How can we fix this?

By dialysis, if necessary; occasionally by draining third spaces (paracentesis, thoracentesis); but mostly, it’s by diuresis. And oddly for a practice so important, it isn’t taught well in school. Here’s how to do it.

When should you diurese?

Some would say when a patient is off vasopressors, no longer in acute kidney injury, and otherwise, basically, stable.

This may be too conservative. Remember your Frank-Starling relationship, which tells you that a fluid bolus will significantly increase a patient’s cardiac output only if they are in a fluid responsive state of their preload curve. If you invert this concept, you realize its diuretic implication: removing fluid will only decrease cardiac output if a patient is fluid responsive as well, and patients you’re planning to diurese are typically hypervolemic and the furthest thing from fluid responsive. (The exception is the patient who is “whole body overloaded, but intravascularly dry,” as discussed below.) In these cases, even some degree of shock may not be a contraindication to judicious diuresis. Neither is kidney injury, which in such patients is presumably not pre-renal and therefore not fluid-dependent either—ATN doesn’t benefit from edema, friends.

So suffice to say that it’s time to diurese when the patient’s current fluid status becomes a hindrance more than a help. Note that this has little to do with their absolute balance, and some patients may actually be healthiest while either more positive or more negative than they were admitted. Look at them and use your brain.

How should you diurese?

Start with furosemide (Lasix). A common starting dose is 20 mg IV.

Remember, however, that renal insufficiency will directly obstruct your efforts, so adjust as needed. A good starting dose is therefore 20 times the serum creatinine; a patient with a creatinine of 2.0 should therefore receive 40 mg.

Then follow the urine output. (A Foley catheter is invaluable for this, although not absolutely mandatory.) The patient should vigorously diurese within a couple hours. Some texts suggest a satisfactory cutoff for adequate diuretic response; for our purposes, let’s just say it should impress you and be conducive to your ultimate goals of fluid removal.

If not… you must give a higher dose.

Some people seem to believe repeating the same ineffective dose, or making some piddling increase, will achieve their ends. Not so. Loop diuretics have a threshold effect; you need to adequately saturate the receptors to achieve diuresis. If 20 mg didn’t do it now, 20 mg given more frequently won’t do it later. Instead, double it.

20 mg becomes 40 mg. If that doesn’t answer, 40 mg becomes 80 mg, then 160 mg, and this doubling process can continue until you lose your nerve. Maximum doses of Lasix are determined by clinician comfort more often than by pharmacology, and swaggering gun-totin’ nephrologists casually give doses that would make your ears ring. In general, don’t even start to worry about adverse effects until the total daily dose approaches the 1,000 mg ballpark, and single doses up to 200 mg or so are almost always acceptable (although the pharmacy may ask that you run them in a mini-bag instead of giving it by push).

So: start at a decent dose, double it until you’re making progress, then repeat that dose as needed when urine output tapers back down. Giving furosemide every 6–12 hours is usually reasonable; there is an old legend that the trade name “Lasix” was coined since it “lasts six hours,” but it really depends on the patient. Repeat doses can be given manually, but only if someone will pay attention throughout the day and night to order them; otherwise, schedule them around the clock. In outpatient and floor settings, diuresis is often held at night to let patients sleep; in the ICU, particularly in catheterized patients, it’s a 24 hour business.

If diuresis is vigorous, watch the blood pressure. If total urine output is anything more than trivial, monitor the electrolytes, particularly serum potassium and magnesium. Both of these will be lost in urine and can drop precipitously, so check them twice a day at least.

Common fluid goals are to end a 24-hour period 500ml, 1000ml, or 2000ml net negative. However, in a severely overloaded patient, you may be limited only by tolerance. If blood pressure and electrolytes aren’t a problem, keep cracking on.

What if it’s not working?

If horrifying doses of furosemide are not producing diuresis, you need to change up your game.

Adding a thiazide diuretic is often helpful. These are not potent diuretics on their own, but alongside a loop agent they create synergy. Metolazone works better than most others at low GFRs and is therefore usually our preference; consider 10 mg orally, given before furosemide to catalyze its effect. Most thiazides are enteral only, the main exception being IV chlorothiazide, which not all hospitals carry.

If you’re seeing diuresis, but it’s not lasting—i.e. it drops off after only a few hours—you can consider a furosemide infusion. In general, continuous drips have not been shown to be more effective than intermittent pushes, which makes sense given the threshold effect noted (if a certain dose isn’t yielding diuresis, giving it slower and longer won’t work any better). But it does spread the effect across the day. This is most useful in hemodynamically labile patients who may not tolerate a drop in preload after a big Lasix push.

Some clinicians are fond of bumetanide (Bumex), which is also a loop diuretic and fairly interchangeable with furosemide. However, occasionally it may produce a better response, particularly in hypoalbuminemic patients; if you’re struggling it can be worth a shot. 1 mg of Bumex (IV and oral dosing are equivalent) is roughly equivalent to 20–40 mg IV furosemide.

At the end of the day, severe diuretic resistance may simply indicate worse renal function than you realized (a “furosemide stress test” has been shown to correlate with renal outcomes), and might be an indication that it’s time to consider dialysis.

What about hypovolemic patients?

Some patients are so “leaky”—i.e. have such poor capillary integrity due to sepsis, hypoalbuminemia, etc—that they may have 10 liters of extra fluid in their interstitial space, yet remain intravascularly hypovolemic. All of their fluid is just stubbornly stuck in their lungs, thighs, and shins.

Attempting to diurese these patients tends to result in poor output, a rapidly-diminishing curve, or simply hypotension. What to do?

One answer is to simply wait; they may not be ready. However, some of these people will never get much better.

You can try albumin. In theory, increasing the oncotic pressure may draw some fluid back into the intravascular space, allowing it to be diuresed away: try administering hypertonic albumin (e.g. 25%), followed an hour or two later by your diuretic. This practice remains controversial, has almost no evidence, and is very possibly voodoo; however, the old “push/pull” refuses to die and occasionally seems to work.

What if the numbers start to get weird?

Loop diuresis will tend to produce hypokalemia, hypernatremia, and metabolic alkalosis.

Hypokalemia (and hypomagnesemia) are simple enough: just replace electrolytes aggressively. In severe cases you can consider adding a potassium-sparing diuretic like spironolactone.

Hypernatremia arises from the dilute urine produced by loop agents—patients urinate more free water than salt—and is probably the cause of metabolic alkalosis. Henderson-Hasselbalch fans refer to this as a “contraction alkalosis” and wave their hands to explain it; strong ion or “Stewart” adherents say it’s merely the result of hypernatremia increasing the strong ion difference.

Whatever the cause, it can be frustrating, as you may have to give the patient back some free water, resulting in a fluid yo-yo. One answer can be introducing a thiazide (as above), which provides more natriuresis than a loop diuretic and helps balance the urine tonicity.

If significantly alkalemic, acetazolamide (Diamox) can also steady the ship by acidifying the urine. Give it cautiously—a single dose, or perhaps a day or two’s worth at a time—to avoid overshooting. Happily, it also has a diuretic effect, which may yield additional synergy.

(The late, legendary John Hinds would balance these patients by giving pure free water and hydrochloric acid via central line, but that’s a bit daring for most of us; your pharmacist would probably have a stroke at the mere suggestion.)

When are you finished?

When they look clinically euvolemic. There should be no hypervolemic complications such as pulmonary edema, yet also no signs of hypovolemia, such as prerenal azotemia or hypotension.


Like most deescalation, deresuscitation, and ICU liberation, diuresis can be a bit of a dull topic compared to the glitz and glamor of acute resuscitation. However, after a few years of staring at enormously edematous patients for whom 75% of their problems seem to be fluid-related, it gets awfully appealing. A skilled, aggressive hand with diuretics can speed vent weaning, normalize organ function, and virtually eliminate ARDS. That’s pretty good medicine.