A century-old marker of filtration function, used as a marker of injury — the mismatch explains most of our diagnostic delay, and all of our failed trials.
Koyner Ch 16 · Ronco Ch 9, 25
Hoste E et al. Intensive Care Med. 2020;46:943–953 (RUBY) · Koyner JL et al. Kidney360. 2022;3:1158–1168 (standardised clinical assay; the cut-off data) · Bagshaw SM et al. Crit Care. 2021;25:185 (external validation) · Koyner Ch 17
Five ways creatinine lies — and the arithmetic that beats the lag
1 · It lags
AKI detected 2–3 days after the insult — a blind window.
2 · It is diluted
A concentration, not a mass. +8 L can hold the number flat while GFR halves.
3 · It tracks muscle, not kidney
Generation falls with sarcopenia, sepsis, liver failure. SCr 0.6 in cachexia can be GFR 40.
4 · It is secreted
10–15% secreted. Transporter blockade raises SCr with no change in GFR.
5 · It reports function, never damage
Damage can be complete while the number stays normal.
Steady state: GFR × SCr = generation — every eGFR equation assumes it; the ICU never provides it
KeGFR = (SCr₁ × eGFR₁) / SCrmean × [ 1 − (24 × ΔSCr) / (Δt × MaxΔSCr/day) ]
Sieving coefficient ≈ 1 — uninterpretable during KRT
eGFRcr-cys (CKD-EPI 2021) = creatinine + cystatin C + age + sex, no race coefficient More accurate than either marker alone.
Proenkephalin (penKid)
Freely filtered and not inflammation-driven — a real-time GFR surrogate. Not FDA-cleared.
Order cystatin C when…
Creatinine and the clinical picture disagree
Narrow-index drug dosing
No usable muscle mass
Secretion blocker on chart
Inker LA et al. N Engl J Med. 2021;385:1737–1749 · Yong Z et al. Sci Rep. 2017;7:41012 · Hollinger A et al. Kidney Int Rep. 2018;3:1424–1433 (Kid-SSS) · Ronco Ch 25 · Koyner Ch 16–17
02
Damage, stress & the 2×2
Three independent axes — filtration, structural damage and cellular stress — collapsed by a single creatinine into one crude number.
Koyner Ch 16–17 · Ronco Ch 24, 26–27 · ADQI 23
The damage × function matrix — four patients, one creatinine
Damage marker negative
Damage marker positive
Creatinine / UO normal
No AKI — reassuring
Subclinical AKI — injury with no functional signal; excess KRT and death
Creatinine / UO abnormal
Haemodynamic AKI — RAAS blockade, SGLT2i, decongestion; tubule intact
Clinical AKI with damage — the only quadrant our therapies target
The left-lower box: creatinine rise ≠ injury
In ADHF diuresis and in SPRINT, creatinine rose while damage markers did not. Biomarkers stop you stopping the beneficial drug.
The right-upper box: injury without a rise
NGAL-positive, creatinine-negative → higher KRT and death. ADQI 23 formalised it.
Moledina DG et al. Am J Kidney Dis. 2017;70:807–816 · Haase M et al. J Am Coll Cardiol. 2011;57:1752–1761 · Ahmad T et al. Circulation. 2018;137:2016–2028 · Ostermann M et al. JAMA Netw Open. 2020;3:e2019209 (ADQI 23) · Koyner Ch 16
The biomarker menu — site of origin, biology, window
Marker
Class
Tubular site
Sample/peak
Evidence
NGAL
Damage
TAL, collecting duct
Urine · 2–3 h
Meta AUC 0.815
KIM-1
Damage
Proximal tubule
Urine · 12 h
Slow for early warning
IL-18
Damage
Proximal tubule
Urine · 6 h
Confounded by inflammation
[TIMP-2]·[IGFBP7]
Stress
Proximal tubule + TAL
Urine · 4–12 h
FDA-cleared; AUC 0.80
CCL14
Persistence
Tubulointerstitial
Urine
RUBY AUC 0.83
Cystatin C
Function
Filtered
Plasma · 12–24 h
Real-time GFR when Cr lies
Haase M et al. Am J Kidney Dis. 2009;54:1012–1024 · Koyner JL et al. J Am Soc Nephrol. 2012;23:905–914 · Hoste E et al. Intensive Care Med. 2020;46:943–953 (RUBY) · Ronco Ch 26–27 · Koyner Ch 16
NGAL — the most studied, the most confounded
Where it earns its keep
Paediatric cardiac surgery: urinary NGAL rose ≥50-fold at 2 h, before creatinine
TRIBE-AKI: highest tertile OR 7.7 for progression
Separates transient from sustained AKI in the ED
Levels at AKI predict 3-year mortality
Why it fails — and the number that proves it
A neutrophil granule protein — sepsis plasma specificity only 0.57
CKD raises baseline; pyuria raises urine NGAL
Plasma and urine assays not interchangeable
No trial shows acting on NGAL improves an outcome
How to use it if your laboratory offers it
Use it as a trend within one patient against a known insult time. A single value in undifferentiated sepsis is nearly information-free.
Mishra J et al. Lancet. 2005;365:1231–1238 · Koyner JL et al. J Am Soc Nephrol. 2012;23:905–914 · Zhang A et al. Crit Care. 2016;20:41 · Nickolas TL et al. J Am Coll Cardiol. 2012;59:246–255 · Ronco Ch 26
Cell-cycle arrest: why [TIMP-2]·[IGFBP7] is a different kind of test
The idea
Damage markers report cells already injured. Arrest markers report cells that stopped dividing to protect themselves.
The mechanism
ROS, PAMPs and DAMPs push tubular epithelium into the cell cycle; IGFBP7 → p53/p21, TIMP-2 → p27.
Why arrest helps
Prevents mitosis with damaged DNA; conserves ATP.
Why it is measurable
Both secreted into the lumen — one sample integrates the whole nephron.
Division of labour
IGFBP7 in surgical patients, TIMP-2 in sepsis; the product works across both.
The honest limit
Reversible and non-specific — it says "threat in 12 h", never why.
Kashani K et al. Crit Care. 2013;17:R25 · Noto A, Cortegiani A, David A. Crit Care. 2019;23:48 (urine osmolality) · Ronco Ch 27 · Koyner Ch 16
Sapphire, Topaz, Opal — the cut-offs and what earned them
SAPPHIRE · Crit Care 2013
Endpoint: moderate-severe AKI in 12 h. AUC 0.80 — superior to NGAL, KIM-1, IL-18 and cystatin C.
TOPAZ · Am J Respir Crit Care Med 2014
At 0.3: sens 92%, spec 46%, RR 7 — the FDA-cleared point.
OPAL · Nephrol Dial Transplant 2014
Sapphire cut-offs on the commercial platform: sens 89% at 0.3; spec 90% at 2.0.
≤ 0.3(ng/mL)²/1000 · sens 92%, spec 46% · rule-OUT
> 0.3 – 2.0Stress present · PPV ≈0.25 at the 0.3 cut-off · monitor, de-risk
> 2.0spec 90–95%, sens ~42% · PPV ≈0.50 · rule-IN
Kashani K et al. Crit Care. 2013;17:R25 · Bihorac A et al. Am J Respir Crit Care Med. 2014;189:932–939 · Hoste EAJ et al. Nephrol Dial Transplant. 2014;29:2054–2061 · Koyner Ch 16 · Ronco Ch 27
Why a good AUC still buys you a bad answer
AUC belongs to the test; predictive value to test and patient.
At 0.3 the panel is a rule-out — relax, don't escalate
Only >2.0 is worth acting on, and even there half of positives never develop the outcome
No early-prediction panel reaches PPV >0.5 — CCL14's 72% is persistence
The same arithmetic, used well — IL-9 for AIN
Biopsied AKI: urine IL-9 + TNF-α gave AIN AUC 0.84.
Bihorac A et al. Am J Respir Crit Care Med. 2014;189:932–939 (sens 0.92 / spec 0.46 at 0.3) · Moledina DG et al. JCI Insight. 2019;4:e127456 · Koyner Ch 16 (PPV 0.25 at 0.3, 0.50 at 2.0) · Ostermann M et al. JAMA Netw Open. 2020;3:e2019209
Timing and stewardship — when to draw, and what you promised to do
Marker
When
Answers
Avoid when
[TIMP-2]·[IGFBP7]
0–12 h after insult
Stage 2–3 in 12 h?
AKI present; low pre-test
NGAL
2–6 h
Tubular injury now?
Sepsis
CCL14
Established AKI
Persist ≥72 h?
Pre-AKI
Cystatin C
12–24 h
Real GFR?
CRRT
01
Pre-testName the insult and time zero.
→
02
Pre-commitWhat each result changes.
→
03
Pair the axesFunctional + damage.
→
04
Close the loopRecord risk and actions.
The missed window
Damage markers run on the insult clock — 48 h late, they read negative.
Krawczeski CD et al. J Am Coll Cardiol. 2011 · Ostermann M et al. JAMA Netw Open. 2020;3:e2019209 (ADQI 23) · Ronco Ch 26–27 · Koyner Ch 8, 16
CCL14 — the only marker aimed at a decision you actually make
RUBY
Enrolled within 36 h of KDIGO stage 2; endpoint persistent stage 3 ≥72 h. CCL14 AUC 0.83 — better than KIM-1, NGAL and cystatin C.
Low cut-off ≈ 1.3 ng/mL — rule out
Sensitivity 91%, NPV 92%
Reassure: this stage-3 AKI is likely transient; don't rush access
High cut-off ≈ 13 ng/mL — rule in
Specificity 93%, PPV 72%
Plan durable access; involve the family early
Why the question matters more than the AUC
Adding CCL14 moved AUC only 0.86 → 0.88. "Will this AKI persist?" is a question we act on daily — a modest AUC aimed at a real decision beats an excellent one at none.
Biomarker-guided care: what actually moved an outcome
A test improves outcomes only when welded to an intervention that works.
PrevAKI · Intensive Care Med 2017
276 cardiac surgery patients randomised to a KDIGO bundle. AKI at 72 h 55.1% vs 71.7% — NNT ≈6.
BigpAK-2 · Lancet 2025
1,176 patients: moderate-severe AKI 14.4% vs 22.3%, NNT 12.
The equipoise you must state out loud
Every trial was unblinded, with an endpoint a haemodynamic bundle can move
The benefit belongs to the bundle; the biomarker only selects who gets it
No biomarker strategy has moved mortality or KRT
Meersch M et al. Intensive Care Med. 2017;43:1551–1561 · Zarbock A et al. Lancet. 2025;406:2782–2791 (BigpAK-2) · Göcze I et al. Ann Surg. 2018;267:1013–1020 · Zarbock A et al. Anesth Analg. 2021;133:292–302 · Koyner Ch 8
Where the field is going — sub-phenotypes and the 2026 definition
Two reproducible AKI sub-phenotypes
Latent class analysis (Bhatraju) separated AKI-SP2 (high sTNFR-1, angiopoietin ratio, ICAM-1; worse non-recovery and mortality) from AKI-SP1
In VASST, vasopressin cut 90-day mortality only in SP1 (27% vs 46%, p = 0.02)
Every neutral AKI trial may have lumped two diseases together — enrichment is how the next generation gets designed
The KDIGO 2026 AKI/AKD guideline (public-review draft) brings structural biomarkers and cystatin C into the diagnostic framework
Bhatraju PK et al. Am J Respir Crit Care Med. 2019;199:863–872 · KDIGO 2026 AKI/AKD public review draft · Koyner Ch 8
03
Stress tests, risk scores & alerts
Cardiology stresses the heart before it fails. We can stress the glomerulus with protein and the tubule with furosemide — and the cheap test out-predicts the expensive panel.
Ronco Ch 9 · Koyner Ch 7–8, 16
Renal functional reserve — a normal GFR is not a safe GFR
Protein load 1–1.2 g/kg raises GFR. RFR = maxGFR − baseline, typically 20–35
Filtered amino acids resensitise the macula densa → afferent vasodilation
RFR falls with age, nephron loss, hyperfiltration, pregnancy
Cr stays normal until ~50% of nephron mass is lost
Losing reserve means losing autoregulatory range — NSAID sensitivity in CKD
Why it is not routine
Cardiologists stress-test because a positive result has a therapy attached; we have no equivalent.
Bosch JP et al. Am J Med. 1983;75:943 · Fliser D et al. Kidney Int. 1993;44:1082 · Ronco Ch 9
The furosemide stress test — the protocol, exactly
1
Select — KDIGO stage 1–2, not volume-depleted.
2
Dose — furosemide 1.0 mg/kg IV, or 1.5 after a prior loop.
3
Replace — match losses mL-for-mL. Withhold if congested.
4
Measure — hourly UO for 2 h; flush the catheter first.
5
Read — <200 mL over 2 h = failed FST.
Furosemide is not filtered — albumin-bound, secreted by OAT1/OAT3, then blocks Na-K-2Cl. A failed FST means the tubule cannot secrete, or the loop cannot respond.
Chawla LS, Davison DL, Brasha-Mitchell E, et al. Crit Care. 2013;17:R207 · Ronco Ch 9 · Koyner Ch 12
The FST out-predicts the panel — and sequencing them beats either
0.87AUC for stage-3 progression (n=77)
87/84%sens/spec at 200 mL/2 h
0.86AUC for KRT
0.90AUC in biomarker-positive patients
Meta-analysis · Crit Care 2020
AKI progression: sens 0.81, spec 0.88, LR⁺ 5.45, LR⁻ 0.26.
Put the two slides together
The FST's LR⁺ 5.45 vs NephroCheck's 1.7 is the whole argument. At 25% pre-test, a failed FST moves you to ~64%.
Pitfall 2 — diagnostic, not therapeutic
Passing is not being treated; failing must not trigger infusions to chase urine.
Koyner JL et al. J Am Soc Nephrol. 2015;26:2023–2031 · Chen J-J et al. Crit Care. 2020;24:202 · Ronco Ch 26 · Koyner Ch 16
Risk scores: excellent for populations, mediocre for individuals
Score
Population
Inputs
Output
Cleveland Clinic (Thakar)
Cardiac surgery
13 items, including CHF, IABP, diabetes
Dialysis-requiring AKI: 0.5% → >20%
Simplified Renal Index
Cardiac surgery
Six weighted items
Risk of KRT; externally validated
Mehran
PCI
Hypotension, IABP, CHF, anaemia, contrast volume
CA-AKI 7.5% → 57%; contrast the only modifiable input
Nephrotoxin exposure (NINJA)
Hospitalised children
Nephrotoxin count + automatic Cr surveillance
A QI trigger
Renal angina index
Paediatric ICU
Risk × injury strata
AUROC 0.74–0.81 — who gets a biomarker
Thakar CV et al. J Am Soc Nephrol. 2005;16:162–168 · Mehta RH et al. Circulation. 2006;114:2208–2216 · Wijeysundera DN et al. JAMA. 2007;297:1801–1809 · Mehran R et al. J Am Coll Cardiol. 2004;44:1393–1399 · Basu RK et al. Clin J Am Soc Nephrol. 2014;9:654–662 · Koyner Ch 7
ICU severity scores and electronic alerts — the limits of detection
What the scores are, and are not, for
APACHE II — 12 variables plus age and chronic health
SAPS III — 20 variables; calibration degrades over time
SOFA — six systems daily; a rise ≥2 defines organ dysfunction
Renal SOFA inherits every Cr and UO limitation
AKI mortality models underperform general ICU models
Tackling AKI · stepped-wedge RCT
E-alert + bundle: no mortality change; better recognition, shorter stay.
Alert-only RCTs
Alerts in isolation did not improve outcomes. An alert without an owner is noise.
For your hospital project
Build the response first — who is paged, what they do — then attach the trigger.
Knaus WA et al. Crit Care Med. 1985;13:818 · Vincent JL et al. Intensive Care Med. 1996;22:707 · Selby NM et al. J Am Soc Nephrol. 2019;30:505–515 · Wilson FP et al. Lancet. 2015;385:1966–1974 and BMJ. 2021;372:m4786 · Koyner Ch 7–8 · Ronco Ch 6
04
Imaging the kidney and the venous side
Ultrasound answers binary questions superbly and continuous questions badly — the skill is knowing which kind of question you are asking.
Koyner Ch 18 · Ronco Ch 33–35
Renal ultrasound: five questions, in this order — then leave
Obstruction? Confluent anechoic spaces in the sinus fat
Chronically diseased? Length <9 cm, thin cortex
Parenchyma abnormal? Echogenicity vs liver — describe, don't score
Koyner Ch 18 · Faubel S et al. Clin J Am Soc Nephrol. 2014;9:382–394 · O'Neill WC. Clin J Am Soc Nephrol. 2014;9:373–381 · Ronco Ch 34–35
Renal resistive index — a real signal buried in confounders
RRI = (peak systolic velocity − end-diastolic velocity) / peak systolic velocity Interlobar artery; average three cycles. RRI tracks vascular compliance as much as renal resistance.
Normal <0.70; ATN, vasculitis and HRS run >0.78
Small series: baseline <0.68 identified those who did not develop AKI
The multicentre test failed: admission RRI predicted persistent AKI with AUROC 0.58
Everything that moves the number
Age, atherosclerosis, chronic hypertension — all raise it
Heart rate, arrhythmia, AR, low pulse pressure
IAH, high PEEP, vasopressor dose
Operator angle, depth, gain, sample volume
A trend within one patient, never a threshold across patients
Darmon M et al. Intensive Care Med. 2018;44:1904–1913 (multicentre, AUROC 0.58) · Darmon M et al. Intensive Care Med. 2011;37:68–76 · Ronco Ch 33 · Koyner Ch 18
VExUS and the IVC — reading congestion as a pattern
Vein
Normal
Mild
Severe
Hepatic
S > D
S < D
S reversed
Portal
<30%
30–49%
≥50%
Intrarenal
Continuous
Biphasic
Monophasic
0IVC <2 cm
1IVC ≥2 cm, mild
2+ one severe
3+ ≥2 severe
Evidence, and its edges
Grade 3 predicts AKI in cardiac cohorts (HR 3.69), not the general ICU.
Where it lies
The IVC gates it: <2 cm forces grade 0 — invalid with raised IAP
S-wave reversal occurs with TR and AF; portal pulsatility fails in cirrhosis
Beaubien-Souligny W et al. Ultrasound J. 2020;12:16 · Longino A et al. Chest. 2024;165:590–600 (right-heart catheterisation validation) · Klompmaker P et al. Crit Care Med. 2026 (systematic review and meta-analysis, 32 studies) · Lichtenstein DA, Mezière GA. Chest. 2008;134:117–125 (BLUE) · Zoccali C et al. J Am Soc Nephrol. 2013;24:639–646 · Koyner Ch 18 · Ronco Ch 33
05
Contrast, and the consults you will be asked to give
The most over-feared diagnosis in the hospital — and the fear itself now causes measurable harm. Then three consults where today's tests change the plan.
Koyner Ch 8, 16, 18, 44 · ACR/NKF 2020
What the control groups did to contrast nephropathy
Non-contrast CT causes the same "contrast nephropathy" rate: 8.8% vs 9.7%
21 propensity-matched cohorts, 169,455 patients: OR 0.97 for AKI; no excess dialysis or death
What survives is a gradient: OR 1.00 at eGFR ≥60, 1.40 at 30–44, 2.96 below 30
CI-AKI
Contrast-induced — causality presumed
CA-AKI
Contrast-associated — after contrast, causality unproven
The association that does remain
When CA-AKI occurs it predicts in-hospital death and faster CKD progression.
Causality is unproven — it marks the patient with least reserve.
Obed M et al. Eur Radiol. 2022;32:8432–8442 (21 propensity-matched cohorts) · Aycock RD et al. Ann Emerg Med. 2018;71:44–53 · Davenport MS et al. Radiology. 2013;268:719–728 (eGFR gradient) · McDonald JS et al. Intensive Care Med. 2017;43:774–784 (ICU cohort) · Bruce RJ et al. AJR. 2009;192:711–718 · Koyner Ch 44
Who is genuinely at risk — the 2020 ACR/NKF thresholds
He is above the 2.0 threshold. At a post-cardiotomy prevalence near 30%, what is his post-test probability — and what changes in the next 12 hours?
Think 45 seconds · answer on the next slide
Meersch M et al. Intensive Care Med. 2017;43:1551–1561 (PrevAKI) · Zarbock A et al. Lancet. 2025;406:2782–2791 (BigpAK-2) · Koyner Ch 8, 39
Case 2 — what the evidence supports doing
01
Confirm the window4 h post-bypass, no established stage 2–3 AKI
→
02
Deploy the bundleHold ACEi/ARB, remove nephrotoxins, optimise haemodynamics
→
03
Fix the gradientMAP 68 − CVP 14 = RPP 54. Decongest or raise MAP
→
04
Re-measureUO hourly; Cr q12h
The one-line summary
A high [TIMP-2]·[IGFBP7] earns this patient an hour of attention, not a catheter.
Say the counter-argument as well
Every element of the bundle is basic post-operative care — the biomarker may simply have made a busy unit do what it should already have done.
Meersch M et al. Intensive Care Med. 2017;43:1551–1561 · Zarbock A et al. Anesth Analg. 2021;133:292–302 · Koyner Ch 8, 39
Case 3 · Septic shock, day 3 — the creatinine has stalled
58 F, urosepsis, off vasopressors, balance +6.5 L. Creatinine rose 0.8 → 2.1 over 48 h and has been flat for 24 h. Albumin 2.1. Recovering, or heading for dialysis?
≥200 mL → tubule intact: loop infusion, tolerate a modest Cr rise
<200 mL → 25% pre-test becomes ~64%: secure access, brief the family
The trap in this case
A Cr flat for 24 h feels like recovery — but it is equally consistent with a new steady state at a much lower GFR.
Chawla LS et al. Crit Care. 2013;17:R207 · Chen J-J et al. Crit Care. 2020;24:202 · Beaubien-Souligny W et al. Ultrasound J. 2020;12:16
Final poll — interpret the result, not the number
63 y, admitted 30 h ago with pneumonia. Creatinine 1.1 (baseline 1.0), UO 0.9 mL/kg/h, no vasopressors, no contrast planned. The night registrar sent a [TIMP-2]·[IGFBP7]: it returns at 0.9. A plasma NGAL is also elevated.
A. Both positive — start KRT planning · B. Bolus and repeat both in 6 h · C. Full KDIGO bundle, repeat in 12 h · D. Neither test should have been sent — treat the pneumonia
Hands up — then name the specific test property that makes your answer correct
Kashani K et al. Crit Care. 2013;17:R25 (Sapphire) · Koyner JL et al. J Am Soc Nephrol. 2015;26:2023–2031 (NephroCheck) · Koyner Ch 16
Key takeaways
Cr lags 2–3 days, is diluted, and tracks muscle loss — cystatin C settles it.
Three axes: function, damage, stress. Subclinical AKI carries excess KRT and death; haemodynamic AKI does not.