From glomerular filtration to bedside dosing — how the nephron clears drugs, how eGFR equations estimate that clearance, and how declining kidney function reshapes safe prescribing
Every renally-cleared drug begins its exit from the body at the glomerulus, a tuft of roughly 200 fenestrated capillary loops per nephron wrapped by podocyte foot processes. Blood is delivered under a favorable hydrostatic pressure gradient (afferent arteriole wider than efferent), forcing roughly 20% of renal plasma flow across the three-layer filtration barrier into Bowman's space. What crosses is governed almost entirely by two variables: how big the molecule is, and — for the drug fraction not bound to albumin or α1-acid glycoprotein — how negatively charged it is, since the basement membrane itself carries a net negative charge that repels anionic solutes.
Three anatomic layers make up the glomerular filtration barrier, each contributing a distinct filter property:
1. Fenestrated capillary endothelium: • Pores ~70–100 nm in diameter, freely permeable to water, ions, and small/medium solutes • Excludes blood cells and large plasma proteins by pore size alone
2. Glomerular basement membrane (GBM): • Type IV collagen + laminin + heparan sulfate proteoglycans • Heparan sulfate carries fixed negative charges → repels anionic macromolecules more than cationic ones of the same size • Effective pore size ~8 nm; this is the primary size-selective layer
3. Podocyte slit diaphragm: • Interdigitating foot processes bridged by nephrin/podocin protein complexes • Final size-selective checkpoint, slit pores ~4–14 nm
Consequence for drug pharmacokinetics: • Free (unbound) drug with molecular weight well under the ~66 kDa albumin cutoff (essentially all small-molecule drugs, MW 150–1000 Da) is filtered essentially completely — its concentration in the ultrafiltrate equals its free plasma concentration • Protein-bound drug is NOT filtered: a highly protein-bound drug (e.g., warfarin, 99% bound) has only ~1% of total plasma concentration available for glomerular filtration at any instant • As free drug is filtered and removed, bound drug dissociates from albumin to re-establish equilibrium — binding acts as a reservoir that continuously "feeds" the free fraction, but overall renal clearance of highly bound drugs is still limited by how much is unbound at any moment • Highly charged anionic drugs (e.g., some beta-lactams) are filtered somewhat less efficiently than their size alone would predict, because the negatively charged GBM electrostatically repels them
GFR is the whole-kidney sum of single-nephron filtration rates (~1 million nephrons per kidney, each filtering ~60–70 nL/min at rest), and it is the single most important physiologic variable in renal drug dosing — which is precisely why every renal dose-adjustment protocol starts by asking: what is this patient's GFR right now?
A drug's filtration clearance can never exceed GFR × free fraction. This is the ceiling on passive renal elimination — anything cleared faster than that ceiling (like penicillin, cleared 5–10× faster than filtration alone would allow) must be getting an assist from active tubular secretion, the subject of Stage 2.
Filtration alone cannot explain the renal clearance of many important drugs. The proximal tubule houses a battery of polyspecific transporters that actively pump drugs from peritubular blood into the tubular lumen — a process that can clear even protein-bound, non-filtered drug and push total renal clearance well above GFR. Further downstream, the distal tubule and collecting duct run the process partly in reverse: lipophilic, un-ionized drug diffuses passively back out of the concentrating filtrate into blood, a phenomenon acutely sensitive to urine pH and flow rate.
Active secretion is a two-step, energy-dependent relay across the proximal tubule epithelial cell:
Basolateral (blood-facing) uptake: • OAT1 and OAT3 (organic anion transporters, SLC22 family) exchange intracellular α-ketoglutarate for extracellular anionic drugs — penicillins, cephalosporins, loop and thiazide diuretics, methotrexate, NSAIDs • OCT2 (organic cation transporter 2) takes up cationic drugs down their electrochemical gradient — metformin, cimetidine, amantadine, cisplatin • Because these transporters pull drug directly from peritubular blood, they can secrete even a drug that is highly protein-bound and poorly filtered
Apical (lumen-facing) efflux: • MRP2/MRP4 extrude anions into the lumen after basolateral uptake • MATE1/MATE2-K extrude the cations taken up by OCT2, using the lumen-to-cell H+ gradient • P-glycoprotein (P-gp/MDR1/ABCB1) is a broad apical efflux pump handling digoxin, many tyrosine-kinase inhibitors, and — critically — the direct oral anticoagulants (dabigatran is a P-gp substrate)
Net effect on clearance: • Renal clearance of penicillin (~500 mL/min) far exceeds GFR (~120 mL/min) — proof that secretion, not filtration, dominates its elimination • Because secretion is transporter- (not filtration-) limited, competing substrates cause clinically important interactions: probenecid blocks OAT1/3 and was historically co-administered with penicillin to prolong its half-life; cimetidine blocking OCT2/MATE can raise metformin levels
Distal reabsorption and pH-trapping: • As filtrate concentrates along the loop of Henle and distal tubule, the concentration gradient favors passive back-diffusion of lipophilic, un-ionized drug across the tubular epithelium • Weak acids (e.g., aspirin, phenobarbital) are reabsorbed more in acidic urine (more un-ionized) and excreted faster in alkaline urine — the basis for urinary alkalinization (IV sodium bicarbonate) as an antidote strategy in salicylate or phenobarbital overdose • Weak bases (e.g., amphetamines) show the opposite pattern — acidifying the urine accelerates their excretion • Ionized, polar drug conjugates (glucuronides, sulfates) are essentially trapped in the lumen and excreted regardless of pH — "ion trapping"
A timed 24-hour urine collection for measured creatinine clearance is the gold standard for GFR, but it is slow and error-prone in practice. Instead, clinicians estimate GFR from a single serum creatinine value combined with age, sex, and (historically) body size or race, using regression equations validated against measured GFR in large cohorts. Which equation is used matters: it changes the calculated dose for a narrow-therapeutic-index drug.
Cockcroft-Gault (1976) — estimates creatinine CLEARANCE, not true GFR: CrCl (mL/min) = [(140 − age) × weight (kg) × (0.85 if female)] / (72 × SCr (mg/dL)) • Uses actual (or adjusted) body weight, not body-surface-area normalized • Still the equation referenced in many drug package inserts (e.g., DOAC dosing cutoffs), so it remains clinically load-bearing despite being the oldest and least precise
MDRD (Modification of Diet in Renal Disease, 1999/2006): eGFR = 175 × SCr^−1.154 × age^−0.203 × (0.742 if female) × (1.212 if Black, original version) • Normalized to 1.73m² body-surface area • Systematically underestimates GFR when true GFR is above ~60 mL/min/1.73m² — poor discrimination in mild disease
CKD-EPI 2021 creatinine equation (race-free) — current standard of care: eGFR = 142 × min(SCr/κ, 1)^α × max(SCr/κ, 1)^−1.200 × 0.9938^age × (1.012 if female) • κ = 0.7 (female) or 0.9 (male); α = −0.241 (female) or −0.302 (male) • The min()/max() split creates two different exponents depending on whether SCr is below or above κ — this piecewise form fits the true (nonlinear) relationship between creatinine and GFR far better than a single power law • The pre-2021 CKD-EPI equation included a race coefficient (1.159 if Black); the 2021 revision (Inker et al., NEJM) removed it after evidence that race-based adjustment was not biologically justified and could systematically over-estimate kidney function in Black patients, delaying transplant referral and specialist care • This live simulation implements the 2021 race-free equation for a female reference patient — move the creatinine and age sliders and watch eGFR, CKD stage, and recommended dose update in real time
CKD staging by eGFR (KDIGO): G1 ≥90 (normal/high) · G2 60–89 (mildly reduced) · G3a 45–59 · G3b 30–44 · G4 15–29 · G5 <15 (kidney failure)
Limitations shared by all creatinine-based equations: • Serum creatinine reflects muscle mass as well as filtration — a frail, low-muscle-mass patient can have a "normal" creatinine despite genuinely reduced GFR • None of these equations are validated for rapidly changing (acute) kidney function — they assume creatinine is at steady state • Cystatin C-based or combined creatinine-cystatin C equations are recommended by KDIGO to confirm CKD staging when accuracy is critical (transplant, chemotherapy dosing)
Once eGFR is known, the pharmacokinetic question becomes practical: by how much should this drug's dose or interval change? Renal dosing nomograms translate eGFR (or CrCl) bands into percentage-of-normal-dose or extended-interval recommendations, but the single most important conceptual split is between the loading dose — set by volume of distribution and largely renal-function-independent — and the maintenance dose, set by clearance and directly proportional to remaining renal function.
Why loading dose usually stays the same: • Loading dose = target concentration × volume of distribution (Vd) • Vd depends on body water/fat compartments and tissue binding, not on how fast the kidney clears the drug — so a patient with eGFR 15 needs essentially the same loading dose as one with eGFR 100 to reach a given peak concentration quickly (e.g., vancomycin loading dose 25–30 mg/kg regardless of renal function) • Skipping or under-dosing the loading dose in CKD is a common and dangerous error — it delays reaching therapeutic levels in serious infection
Why maintenance dose (or interval) must fall: • Maintenance dose (or dosing rate) = target average concentration × clearance • Clearance falls roughly in proportion to eGFR for drugs that are predominantly renally eliminated — so maintenance dosing rate should fall proportionally too, either by giving a smaller dose at the same interval or the same dose at a longer interval • Extended-interval dosing is often preferred for antibiotics with concentration-dependent killing (aminoglycosides); dose-reduction is often preferred for drugs needing steady trough levels
Worked nomogram logic (as implemented in this simulation, vancomycin-like renally-cleared drug): eGFR ≥60 → 100% of standard maintenance dose eGFR 45–59 → ~75% of standard dose eGFR 30–44 → ~50% of standard dose eGFR 15–29 → ~25% of standard dose, extend interval, check trough/level eGFR <15 (or dialysis) → ~10% of standard dose or switch to level-guided, post-dialysis dosing
Narrow therapeutic index (NTI) drugs need tighter monitoring, not just a nomogram: • Vancomycin: trough or AUC/MIC-guided monitoring; accumulation risks nephrotoxicity and ototoxicity • Digoxin: renal clearance ~60–80%; toxic levels cause life-threatening arrhythmia — levels must be checked whenever eGFR changes • DOACs: dabigatran (~80% renal) is the most renally dependent — many guidelines avoid it below CrCl 30; apixaban (~27% renal) and rivaroxaban (~35% renal) have more headroom but still need dose review below CrCl 15–30 • Gabapentin: 100% renal elimination with zero hepatic metabolism makes it entirely dependent on the dosing nomogram — a standard 300 mg TID regimen given unchanged at eGFR 15 can produce sedation, myoclonus, and encephalopathy from simple accumulation
Bringing the pipeline together: a real patient's eGFR rarely stays fixed, and every renally-cleared drug they are on needs re-review each time it changes materially. Unadjusted dosing in progressive CKD is one of the most common preventable causes of adverse drug events in hospitalized and elderly patients — and the risk compounds when multiple renally-cleared or nephrotoxic drugs are combined, or when the patient eventually starts dialysis.
Case A — Gabapentin in a 78-year-old with eGFR 22 (CKD G4): • Standard regimen 300 mg TID assumes eGFR ≥60; at eGFR 22 the recommended dose is roughly 300 mg once daily to BID, and initiation should start lower and titrate slowly • Unadjusted dosing accumulates gabapentin over days because it has no metabolic escape route (100% renal elimination) — presenting as sedation, gait instability, or myoclonus that can be misread as new neurologic disease instead of a dosing error
Case B — Digoxin in an 85-year-old whose eGFR falls from 55 to 28 after an AKI episode: • Digoxin's narrow therapeutic index (target 0.5–0.9 ng/mL) combined with ~70% renal clearance means a dose that was safe at eGFR 55 can become toxic within days as eGFR drops — nausea, visual halos, and life-threatening arrhythmias are classic presentations • Management: reduce maintenance dose or extend interval in proportion to the eGFR fall, recheck a trough level, and screen for interacting drugs (amiodarone, verapamil) that also raise digoxin levels independent of renal function
Case C — DOAC selection in a patient progressing through CKD stages: • At eGFR 50 (G3a): all DOACs generally usable at labeled doses • At eGFR 25 (G4): dabigatran (~80% renal) is often avoided; apixaban (~27% renal) or reduced-dose rivaroxaban preferred, per label criteria (age, weight, creatinine thresholds) • At eGFR <15 or dialysis-dependent (G5): DOAC use becomes highly individualized; some centers use apixaban off-label with caution, others revert to warfarin, which is not renally cleared but requires INR monitoring
Dialysis dosing preview: • For patients on hemodialysis, the drug-dosing question splits further: does the drug get removed by the dialysis membrane itself? • Small molecular weight, low protein binding, and small volume of distribution predict good dialyzability (e.g., many aminoglycosides, some β-lactams) — these often need a supplemental post-dialysis dose • Large Vd or highly protein-bound drugs (e.g., digoxin, despite renal clearance) are poorly removed by conventional hemodialysis, so dosing follows residual native clearance and drug levels rather than dialysis scheduling • This dialysis-specific layer is the next pharmacokinetic frontier once a patient reaches CKD G5 — a preview of a dosing problem distinct from, but built on, everything in Stages 1–4
A widely cited patient-safety statistic: renal dose-adjustment errors are among the most common preventable prescribing errors in hospitalized patients with CKD, and drugs like gabapentin, digoxin, and vancomycin recur again and again in adverse-event reports specifically because their toxicity is silent and cumulative — the patient looks fine until the drug has been quietly accumulating for days. The single highest-leverage habit in nephrology-aware prescribing is simple: recheck eGFR, and recheck every renally-cleared drug's dose, every time renal function changes.