Understanding Diuretics, Part 1: The Site-of-Action Map

Ask a room of physicians where furosemide works and most will say “the loop of Henle,” which is true but not precise enough to be useful. Ask them where thiazides work, or why torsemide behaves more predictably than furosemide in a decompensating patient, or why doubling a stalled dose of furosemide sometimes does nothing at all, and the answers get vaguer fast. That vagueness is understandable, since diuretic pharmacology is taught once, early, and rarely revisited, but it has real consequences at the bedside, where a meaningful share of what gets labeled “diuretic resistance” turns out to be a predictable pharmacokinetic problem rather than a mysterious one. Not all of it, though: hemodynamics and neurohormonal activation, including reduced renal perfusion, RAAS activation, and venous congestion, are just as often the real driver, and that side of the story is where Parts 2 and 3 will spend most of their time. This post is about the pharmacokinetic layer specifically, not a claim that it’s the whole story.

This post lays the mechanistic groundwork the rest of the series will build on: where each class works, why that location determines its potency, and why getting a drug to its site of action is a separate problem from picking the right drug in the first place. Parts 2 through 5 will apply this foundation to diuretic strategy in HFrEF, cirrhosis, HFpEF, and isolated right-sided heart failure — four conditions that use the same five drug classes but for very different physiological reasons.

[Nephron site-of-action diagram to be inserted here]

The Short Version

Furosemide, torsemide, bumetanide, thiazides, potassium-sparing agents, and even SGLT2 inhibitors and acetazolamide all work on the same nephron, but at five different addresses, and where a drug works determines how powerful it is. The proximal tubule reclaims about 65% of filtered sodium before any of it reaches a diuretic-sensitive segment; the thick ascending limb handles the next 25%; the distal tubule about 5–7%; the collecting duct only 2–3%. That’s the entire reason loop diuretics (furosemide, torsemide, bumetanide) are the most powerful class and everything else is, by comparison, working the margins — they’re intercepting the largest remaining slice of the pie.

Getting a loop diuretic to that site is its own problem, separate from picking the right drug. These drugs are protein-bound in the blood and have to be actively pumped into the tubule rather than simply filtered there — which is why hypoalbuminemia or reduced kidney blood flow can blunt the effect no matter the dose. And furosemide’s oral absorption is famously unreliable (roughly 10–100%, worse in decompensated patients), while torsemide and bumetanide are absorbed much more consistently. That’s a real pharmacokinetic advantage — but it’s worth knowing the largest head-to-head trial (TRANSFORM-HF) found no difference in mortality between torsemide and furosemide, so treat “prefer torsemide” as a reasonable pharmacologic default for an individual patient’s absorption problem, not an evidence-backed outcomes recommendation.

Even a dose that works has a built-in expiration date. Loop diuretics have a threshold below which nothing happens and a ceiling above which more drug adds nothing, and once the dose wears off — in as little as 4 hours for bumetanide, a bit longer for furosemide — the kidney swings into a phase of rebound sodium retention that can undo much of what the dose accomplished. That’s part of why twice-daily dosing often outperforms one larger dose, and why “duration of action” claims for any of these drugs deserve a second look rather than taking the package insert at face value.

There are two ways to break through a stalled response, and they work in opposite directions on the nephron. Adding a thiazide — metolazone is the one most often reached for in this combination — doesn’t just add its own small effect: the distal tubule adapts over time to reclaim sodium the loop diuretic already dislodged upstream, and the thiazide’s real job is blocking that reclamation, unmasking the loop diuretic’s full effect. Acetazolamide and SGLT2 inhibitors do the mirror image from the other direction: by blocking proximal reabsorption, they deliver more sodium downstream to the thick ascending limb, priming the loop diuretic to do more work. Neither pairing is “two diuretics stacked” — each is one drug removing the nephron’s own compensation against the other. Both are also real interventions with real costs, not free lunches: the loop-thiazide combination in particular comes with a well-documented increase in worsening renal function and electrolyte disturbance, so it buys more natriuresis at the price of closer monitoring, not just a smarter unlock.

That’s the whole mental model this series leans on: which site, how much sodium is available there, and what’s helping or fighting the drug working on it. Parts 2 through 5 apply it to HFrEF, cirrhosis, HFpEF, and isolated right-sided heart failure — four conditions that share these same five drug classes but fail to respond to them for very different reasons.

---

The Nitty-Gritty

Everything below is the same material in full mechanistic detail — transporter names, receptor pharmacology, and the primary literature behind each claim. Skip it on a first read; come back to it when a patient’s response doesn’t match what the short version predicts.

Where Sodium Escapes: A Nephron Site-of-Action Map

Bottom Line: Roughly 65% of filtered sodium is reclaimed before it ever reaches a diuretic-sensitive segment, which is exactly why loop diuretics — acting on the segment handling most of what’s left — are the most potent class, and why every other class is, by comparison, working at the margins.

The nephron reabsorbs filtered sodium sequentially, and each segment recovers a characteristic share of the load:

• Proximal tubule (~65%): dominated by Na-H exchange and, in the earliest segment, SGLT2-linked glucose-sodium cotransport. Diuretics targeting this segment (carbonic anhydrase inhibitors, SGLT2 inhibitors) intercept only a small fraction of what’s reabsorbed here, and downstream segments compensate for most of what they block.

• Descending limb: water-permeable, essentially sodium-inert. There is no clinically useful diuretic target here.

• Thick ascending limb (~25%): the site of the Na-K-2Cl cotransporter (NKCC2) — the target of loop diuretics, and the largest fraction of reabsorption available to any diuretic class.

• Distal convoluted tubule (~5-7%): the Na-Cl cotransporter (NCC) — the thiazide target. A smaller absolute contribution, but one with outsized clinical relevance once loop diuretics are already in use (more on that below).

• Collecting duct (~2-3%): ENaC and the mineralocorticoid receptor — the site of amiloride, triamterene, spironolactone, and eplerenone. Low natriuretic capacity on its own, but the segment where potassium handling is decided.

The arithmetic explains the hierarchy of potency clinicians already sense intuitively — loop diuretics work best because they’re intercepting the largest remaining fraction of filtered sodium — but it also explains something less intuitive: why blocking the thick ascending limb alone is self-limiting, since the distal nephron can adapt to reclaim more of what escapes.

ENaC is worth a closer look because it behaves differently from the cotransporters upstream. NKCC2 and NCC are co-transporters — they physically couple sodium movement to chloride and potassium. ENaC, by contrast, is a channel: sodium simply diffuses through it, driven by the electrochemical gradient the basolateral Na-K-ATPase maintains in collecting duct principal cells. That inward flow of positive charge leaves the tubular lumen electrically negative, and it’s this negative potential that drives potassium out into the urine through neighboring ROMK channels — which is why ENaC is the hinge point for potassium handling as well as sodium. Amiloride and triamterene block the channel directly. Spironolactone and eplerenone don’t touch ENaC at all; they block the mineralocorticoid receptor in the same principal cells, which is what aldosterone normally uses to increase both the number of ENaC channels inserted in the membrane and how often each one stays open — so the aldosterone antagonists get to the same potassium-sparing, sodium-losing endpoint by removing the signal that upregulates the channel, rather than blocking the channel itself. Finerenone works the same receptor but isn’t really part of this natriuretic conversation the way the other four classes are: it’s a nonsteroidal MR antagonist, more receptor-selective than spironolactone or eplerenone and without their antiandrogenic side effects, and it’s approved for chronic kidney disease with type 2 diabetes and, as of 2025, for heart failure with mildly reduced or preserved ejection fraction — not as a volume-management diuretic, but for its anti-fibrotic and anti-inflammatory effects on the heart and kidney. It’s worth keeping in mind here because it will come up again in Part 4, on a different basis than the ENaC/aldosterone framework in this section.

Getting Into the Tubule Is a Separate Problem From Blocking the Transporter

Bottom Line: Loop diuretics don’t reach NKCC2 by glomerular filtration — they’re actively secreted into the tubule lumen from the blood, which means anything that impairs that secretory step (hypoalbuminemia, reduced renal blood flow, competing organic acids) blunts the drug’s effect independent of the dose given.

Loop diuretics travel through the blood bound tightly to albumin. Only the small, free (unbound) fraction of a drug can cross into the glomerular filtrate — and because almost none of a loop diuretic circulates free, almost none of it gets into the tubule this way. To reach NKCC2 on the luminal side of the thick ascending limb, the kidney instead has to actively pump the drug there from the blood side: it’s transported from peritubular blood into the proximal tubule cell via organic anion transporters (OAT1 and OAT3) at the basolateral membrane, then secreted into the lumen via multidrug resistance–associated protein 4 (MRP4) and related transporters at the apical membrane — where it travels downstream to compete with chloride for its binding site on NKCC2.¹ This secretory pathway is why oral bioavailability and protein binding matter as much as dose: a drug that never reaches the lumen in adequate concentration won’t work no matter how much is given intravenously, and this is one of the mechanisms underlying diuretic resistance in advanced kidney disease and decompensated heart failure, where proximal secretion is impaired independent of glomerular filtration rate.²

The Oral Bioavailability Problem Nobody Titrates For

Bottom Line: Furosemide’s oral bioavailability ranges from roughly 10% to 100% (averaging near 50%), while torsemide and bumetanide are reliably 80–100% absorbed regardless of gut congestion — so a patient who seems “furosemide resistant” on an oral regimen may simply be absorbing an unpredictable fraction of each dose, not failing the drug class.

This variability isn’t a minor pharmacokinetic footnote. Furosemide’s absorption is genuinely inconsistent between patients and, in the same patient, between health and decompensation. In a study comparing the same patients decompensated versus after diuresis to their dry weight, decompensation significantly delayed absorption — longer lag time, later and lower peak serum concentration — consistent with gut wall edema slowing the rate at which oral furosemide crosses into the blood; total drug absorbed over time (area under the curve) was not significantly different between the two states, so the practical effect is a blunted, delayed peak rather than a smaller overall dose getting through.^3,4 Torsemide and bumetanide don’t share this problem: both are absorbed reliably at 80–100% bioavailability, and torsemide’s absorption is largely unaffected by food intake, where furosemide’s and bumetanide’s are not.³ This is the pharmacologic basis for a preference some clinicians already have for torsemide in patients transitioning to chronic oral therapy — not because it’s a fundamentally different drug, but because its behavior is predictable in a way furosemide’s isn’t. That pharmacokinetic case is real, but it hasn’t translated into a demonstrated outcomes advantage: TRANSFORM-HF, the largest randomized trial directly comparing the two drugs after a heart failure hospitalization, found no significant difference in 12-month all-cause mortality between torsemide and furosemide.⁵ The bioavailability argument above is a reasonable basis for an individual patient’s absorption problem — it isn’t evidence that switching a stable patient to torsemide changes outcomes, and the two shouldn’t be conflated.

The Ceiling Effect and the Braking Phenomenon

Bottom Line: Loop diuretics have a threshold dose below which nothing happens and a ceiling above which more drug adds nothing — but even an effective dose is followed by a rebound phase of sodium retention, so a single well-dosed dose can still leave net sodium balance unchanged by the next day if nothing covers that window.

Loop diuretic dose-response is not linear — it’s sigmoidal, with a threshold below which urinary sodium excretion doesn’t meaningfully rise, and a ceiling above which further dose increases add toxicity without added natriuresis. Chronic kidney disease and heart failure both shift this curve rightward (a higher threshold dose is needed) without necessarily changing the ceiling’s shape, which is the pharmacologic argument for dose escalation up to a class-appropriate ceiling before adding a second agent.¹

Separately, once a loop diuretic’s effect wears off — typically within 6 hours of an oral dose — the kidney enters a phase of avid sodium retention that can offset much of the sodium lost during the drug’s active window, a phenomenon originally described as “post-diuretic sodium retention.”^1,6 In a sodium-replete outpatient this matters less; in a patient who isn’t restricting sodium intake between doses, it can mean the net 24-hour sodium balance is far smaller than the dose would suggest — one reason twice-daily dosing, rather than a single larger dose, is often the better strategy for a patient who isn’t responding as expected.

Duration of Action: Why Some Diuretics Leave a Bigger Retention Window Than Others

Bottom Line: Furosemide and bumetanide’s effect lasts only about 4–6 hours, leaving most of the day exposed to post-diuretic retention if dosed once — and torsemide’s real advantage here is more modest than commonly claimed, since rigorous pharmacokinetic studies put its duration at roughly 6–8 hours, not the 12–16 hours often cited. Among thiazides, by contrast, the duration difference between hydrochlorothiazide and chlorthalidone is both large and clinically real.

Duration of action compounds the braking phenomenon described above: a shorter-acting drug means a longer daily window of unopposed sodium retention. Bumetanide has the shortest duration of the three loop diuretics — roughly 4 hours, occasionally extending to 6 at higher doses — while furosemide lasts somewhat longer, roughly 6–8 hours. Torsemide is frequently described in clinical summaries as offering 12–16 hours of coverage, but the primary pharmacokinetic literature is considerably more modest — controlled dosing studies put torsemide’s duration of action at approximately 6 hours as well, independent of renal function, putting it much closer to furosemide than the 12–16 hour figure suggests.⁷ A rigorous crossover study in healthy subjects found that immediate-release torsemide, like furosemide and bumetanide, produced no net 24-hour sodium or fluid loss despite a large initial natriuretic effect — the same post-diuretic retention problem, just with a different drug.⁸ An extended-release torsemide formulation, built specifically to close that gap, roughly doubled the time patients spent at therapeutic drug concentrations and did produce measurable net sodium loss — evidence that duration of action, not potency, was the actual limiting factor.⁸ The practical takeaway is that torsemide’s more reliable bioavailability (discussed above) is a stronger argument for its use than any dramatic duration-of-action advantage, which the primary literature doesn’t clearly support.

Among thiazides, the duration difference is real and substantial. Hydrochlorothiazide’s duration of action is roughly 6–12 hours, while chlorthalidone’s is 24–72 hours — a difference driven less by a simple half-life gap than by chlorthalidone’s avid uptake into red blood cells and high-affinity binding to erythrocyte carbonic anhydrase, which creates a slow-release drug reservoir rather than a conventional elimination curve.⁹ This is a meaningful part of why chlorthalidone achieves better 24-hour blood pressure control than an equivalent dose of hydrochlorothiazide despite similar peak potency, and why a missed dose of chlorthalidone is far more forgiving than a missed dose of HCTZ.

Two other members of this class are worth knowing individually, because both will come up again later in this series. Metolazone isn’t quite a classical thiazide mechanistically — alongside its DCT/NCC effect, it has a meaningful proximal tubule action as well, which is the traditional explanation for why it’s long been favored over hydrochlorothiazide as the loop-diuretic partner of choice in patients with significantly reduced GFR, and why it’s the agent most often reached for in this combination in HFrEF specifically. That reputation is probably somewhat overstated, though: hydrochlorothiazide and chlorthalidone have both been shown to retain meaningful antihypertensive and natriuretic efficacy in stage 4 CKD in randomized trials, and head-to-head comparisons of metolazone against other thiazide-type agents haven’t consistently shown it to be superior.¹⁰ Its long duration of action — roughly 12–24 hours, supporting once-daily dosing — is real and clinically useful; the idea that it’s the only thiazide that works in advanced kidney disease is weaker than the teaching implies. Chlorothiazide (IV, brand name Diuril) is the other one worth knowing: it’s the only thiazide-type agent available as an intravenous formulation, which matters for hospitalized patients who can’t take oral medication or whose gut edema makes oral absorption unreliable — it allows sequential nephron blockade to be delivered entirely parenterally alongside an IV loop diuretic. Supply has been inconsistent at various points, so it’s worth confirming current formulary availability before counting on it.

Why the Distal Nephron Fights Back

Bottom Line: Adding a thiazide to a loop diuretic isn’t really “adding a second diuretic” — the distal nephron adapts to reclaim much of the extra sodium a loop diuretic delivers to it, so the thiazide’s real job is blocking that reclamation and letting the loop diuretic’s own effect finally reach the urine. That’s sequential nephron blockade: not two drugs stacked additively, but one drug removing the nephron’s escape route from the other.

When a loop diuretic delivers more sodium than usual to the distal nephron, that segment adapts over time by increasing its NCC-mediated reabsorptive capacity — in effect, reclaiming a growing share of the sodium the loop diuretic already succeeded in dislodging upstream. Left unblocked, this is a major reason loop diuretic monotherapy plateaus: the drug is still working at the thick ascending limb, but its output is being quietly recaptured one segment downstream. A thiazide blocks NCC and closes that escape route, which is why the combination so often produces natriuresis well beyond what either drug achieves alone — the thiazide isn’t just contributing its own modest 5-7% of filtered sodium, it’s unmasking the loop diuretic’s full effect. In a randomized crossover study of patients with advanced renal failure (GFR <30 mL/min/1.73m²) — a population where thiazides are often assumed to be ineffective on their own — adding a thiazide to a loop diuretic substantially increased natriuretic and chloruretic response compared with the loop diuretic alone, demonstrating that this synergy persists even when thiazide monotherapy would be expected to fail.¹¹ That efficacy comes with a real cost, not just a theoretical one: in CLOROTIC, a placebo-controlled trial of adding hydrochlorothiazide to intravenous furosemide in acute heart failure, the combination produced significantly more weight loss and better diuretic response, but also significantly more worsening renal function than placebo, with no difference in 30- or 90-day mortality or rehospitalization.¹² Sequential nephron blockade is genuine physiology, not a clinical pearl to reach for reflexively — it buys more natriuresis at the price of closer electrolyte and renal function monitoring, and the mechanistic case for it doesn’t by itself settle whether that trade is worth making in a given patient.

Proximal Blockade Can Prime the Thick Ascending Limb, Too

Bottom Line: The same logic that makes a thiazide unmask a loop diuretic’s effect from downstream also works from upstream — blocking proximal tubule sodium reabsorption with acetazolamide or an SGLT2 inhibitor delivers more sodium to the thick ascending limb, where a loop diuretic is waiting for it, making the combination more than additive rather than simply redundant.

Acetazolamide inhibits carbonic anhydrase in the proximal tubule, blocking a meaningful share of proximal sodium reabsorption and increasing the sodium and chloride load delivered downstream to the thick ascending limb. In the ADVOR trial, adding intravenous acetazolamide to standard loop diuretic therapy in patients hospitalized for acute decompensated heart failure produced a significantly higher rate of successful decongestion by day 3 than loop diuretic therapy with placebo.¹³ SGLT2 inhibitors act on a nearby stretch of the same segment — blocking SGLT2-linked glucose-sodium cotransport in the S1 segment and, through an off-target effect, the sodium-hydrogen exchanger (NHE3) that handles a large share of proximal sodium reabsorption more broadly. In a placebo-controlled crossover study of patients with stable heart failure and type 2 diabetes, empagliflozin produced a clinically meaningful, sustained increase in fractional sodium excretion when combined with intravenous bumetanide, an effect that persisted without attenuation over 14 days.¹⁴

This is sequential nephron blockade applied one segment earlier than the loop-thiazide pairing described above: proximal blockade doesn’t just contribute its own small natriuretic effect, it increases the sodium substrate arriving at the loop diuretic’s own site of action — and, further downstream, primes the thiazide-sensitive segment in turn. Worth being precise about what these two trials actually tested, though: ADVOR enrolled patients on a chronic stable loop diuretic dose at admission, not patients defined as diuretic-resistant, and Griffin’s empagliflozin study was in stable outpatients rather than decompensated or treatment-resistant ones. The mechanism — more sodium delivered downstream when the proximal tubule is blocked — is well supported by both studies; the idea of proximal blockade as a specific rescue strategy for a truly stalled, resistant response is a reasonable mechanistic extrapolation from that evidence, not something either trial was designed to test directly. Site of action, in other words, isn’t only about where a drug works in isolation; it’s about how much sodium a given segment is handed by everything working upstream of it, which is exactly the framework the next four parts will use to explain why the same five drug classes behave so differently across HFrEF, cirrhosis, HFpEF, and right-sided heart failure.

Where This Leaves Us

Site of action, active tubular secretion, oral bioavailability, the ceiling effect, post-diuretic retention, and distal nephron adaptation — this is the toolkit for understanding why a diuretic regimen that should work on paper sometimes doesn’t, and why “resistance” is often several distinguishable problems wearing the same name. Parts 2 through 5 apply this toolkit to four conditions that share these five drug classes but diverge sharply in why diuretic response fails: the low-output, RAAS-activated physiology of HFrEF; the splanchnic vasodilation and secondary hyperaldosteronism of cirrhosis; the heterogeneous, often-overstated congestion of HFpEF; and the absorption problem created when right-sided failure congests the gut itself.

---

References

1. Ellison DH, Felker GM. Diuretic Treatment in Heart Failure. N Engl J Med. 2017;377:1964-1975. doi: 10.1056/NEJMra1703100

2. Felker GM, Ellison DH, Mullens W, Cox ZL, Testani JM. Diuretic Therapy for Patients With Heart Failure: JACC State-of-the-Art Review. J Am Coll Cardiol. 2020;75(10):1178-1195. doi: 10.1016/j.jacc.2019.12.059

3. Pham D, Grodin JL. Dilemmas in the Dosing of Heart Failure Drugs: Titrating Diuretics in Chronic Heart Failure. Card Fail Rev. 2017;3(2):108-112. doi: 10.15420/cfr.2017:10:1

4. Vasko MR, Brown-Cartwright D, Knochel JP, Nixon JV, Brater DC. Furosemide Absorption Altered in Decompensated Congestive Heart Failure. Ann Intern Med. 1985;102:314-318. doi: 10.7326/0003-4819-102-3-314

5. Mentz RJ, Anstrom KJ, Eisenstein EL, et al. Effect of Torsemide vs Furosemide After Discharge on All-Cause Mortality in Patients Hospitalized With Heart Failure: The TRANSFORM-HF Randomized Clinical Trial. JAMA. 2023;329(3):214-223. doi: 10.1001/jama.2022.23924

6. Ellison DH. Mechanistic Insights into Loop Diuretic Responsiveness in Heart Failure. Clin J Am Soc Nephrol. 2019;14:650-652.

7. Knauf H, Mutschler E. Saluretic effect of the loop diuretic torasemide in chronic renal failure. Eur J Clin Pharmacol. 1990;39(4):337-343. doi: 10.1007/BF00315406

8. Wilcox CS, et al. Sodium and Fluid Excretion With Torsemide in Healthy Subjects Is Limited by the Short Duration of Diuretic Action. J Am Heart Assoc. 2018;7:e006135. doi: 10.1161/JAHA.117.006135

9. Carter BL, Ernst ME, Cohen JD. Hydrochlorothiazide Versus Chlorthalidone: Evidence Supporting Their Interchangeability. Hypertension. 2004;43(1):4-9. doi: 10.1161/01.HYP.0000103632.19915.0e

10. Agarwal R, Rahman M, Middleton JP, et al; CLICK Trial Investigators. Chlorthalidone for Hypertension in Advanced Chronic Kidney Disease. N Engl J Med. 2021;385(27):2507-2519. doi: 10.1056/NEJMoa2110730

11. Fliser D, Schröter M, Neubeck M, Ritz E. Coadministration of thiazides increases the efficacy of loop diuretics even in patients with advanced renal failure. Kidney Int. 1994;46(2):482-488. doi: 10.1038/ki.1994.298

12. Trullàs JC, Morales-Rull JL, Casado J, et al; CLOROTIC trial investigators. Combining loop with thiazide diuretics for decompensated heart failure: the CLOROTIC trial. Eur Heart J. 2023;44(5):411-421. doi: 10.1093/eurheartj/ehac689

13. Mullens W, Dauw J, Martens P, et al; ADVOR Study Group. Acetazolamide in Acute Decompensated Heart Failure with Volume Overload. N Engl J Med. 2022;387(13):1185-1195. doi: 10.1056/NEJMoa2203094

14. Griffin M, Rao VS, Ivey-Miranda J, et al. Empagliflozin in Heart Failure: Diuretic and Cardiorenal Effects. Circulation. 2020;142(11):1028-1039. doi: 10.1161/CIRCULATIONAHA.120.045691