Preclinical PK Cheatsheet
Species physiology, IVIVE values, PBPK starter kit, allometric scaling, and Caco-2/MDCK/PAMPA permeability. Researched, cited, free.
Species physiology
Reference physiological parameters for the 5 most-used preclinical species. Davies & Morris (1993) is the canonical source; modern updates and discrepancies are flagged inline with theModern source disagrees with the canonical value; hover the icon for the inline note.icon.
Body weight
Reference adult body weight per species. Used to convert per-kg-normalised parameters back to absolute values for PBPK and dose calculations.
Cardiac output
Total blood pumped per unit body weight per hour. Sets the upper bound on perfusion-limited clearance.
Lidocaine: High-extraction; CL approaches cardiac output
Liver blood flow (total)
Sum of hepatic artery + portal vein flow. Sets the upper bound on hepatic clearance for high-extraction drugs ().
Propranolol: High-extraction; hepatic CL ≈ liver blood flow in human
Kidney blood flow
Both kidneys, total. Upper bound on renal clearance for high-extraction renally cleared drugs.
PAH (para-aminohippurate): PAH clears clinically via near-complete renal extraction. It directly measures renal plasma flow (~600 mL/min in human); renal blood flow is back-calculated as RBF = RPF / (1 − Hct) ≈ 1100-1200 mL/min.
Glomerular filtration rate
Plasma volume filtered per minute per kg. Drugs with are filtered without secretion. implies tubular secretion.
Inulin: Used clinically to measure GFR; cleared exclusively by filtration
Bile flow
Hepatic biliary excretion route. Drugs eliminated via bile rely on this flow rate plus active transport.
Indocyanine green: Cleared almost exclusively in bile; used to assess hepatic function
Total body water
Intracellular + extracellular + plasma water. Drugs distributing into total body water (e.g., antipyrine) have this value.
Antipyrine: total body water; classical probe drug
Plasma volume
Sets the lower bound for . Drugs that bind tightly to albumin sit a few-fold above plasma volume because albumin is also present in interstitial fluid, not because the drug enters cells. The simple estimate overpredicts in this regime.
Warfarin: L/kg, small but ~3× plasma volume because albumin is also extravascular ( would predict ~8.6 L/kg).
Blood volume
Total blood (plasma + erythrocytes). Used for B/P ratio scaling and PBPK initialization.
IVIVE scaling values
Turns in vitro intrinsic clearance into a whole-organism prediction. Drop in any and below — the calculator predicts across all 5 species using , liver weight, and liver blood flow as the conversion factors. The parameter tables that follow show where those numbers come from.
Live IVIVE clearance preview
Predicts hepatic CL across all 5 species using the embedded liver blood flow, MPPGL, and liver weight values.
Strict well-stirred form takes (blood unbound). For , is interchangeable; for high- drugs (cyclosporine, tacrolimus, sirolimus) convert: .
Values for , , and are read live from the data files.
The values plugged in
Critical fact-check: Davies & Morris 1993 does NOT contain MPPGL values (common citation error); the canonical 32 mg/g human value comes from Barter 2007. Human HPGL is genuinely disputed (99 vs 120 vs 139), flagged inline.
MPPGL (microsomal protein per gram liver)
Conversion factor that scales in vitro liver microsomal (μL/min/mg protein) up to in vivo via the well-stirred model. Without this, microsomal data cannot become a clearance prediction.
HPGL (hepatocellularity per gram liver)
Scales in vitro hepatocyte (μL/min/10⁶ cells) up to in vivo. Hepatocyte assays preserve more of the metabolic machinery than microsomes (UGTs, conjugations, transporters), so HPGL-based IVIVE is preferred for non-CYP-dominated drugs.
Liver weight
Combined with MPPGL or HPGL, gives mg microsomal protein per kg body weight (or 10⁶ cells/kg) — the final scaling factor for IVIVE.
Tissue volumes (% body weight)
For PBPK model parameterisation. Brown 1997 is the canonical reference; modern mechanistic PBPK (Rodgers/Rowland 2006 for Kp prediction) builds on these. Adipose % and bone definition warrant inline notes — see flags.
| Tissue | Mouse | Rat | Dog (beagle) | Monkey (cyno) | Human | Source |
|---|---|---|---|---|---|---|
| Skeletal muscleAdult human male; elderly populations 25-30%, lean athletes up to 50% | 38.4 | 40.4 | 45.7 | 40 | 40 | Brown RP 1997 ↗Brown RP, Delp MD, Lindstedt SL, Rhomberg LR, Beliles RP. Toxicol Ind Health (1997) |
| Adipose tissueSingle most BMI-variable tissue; always document body composition assumption | 7 | 7.6 | 15 | 6.5 | 21.4Brown 1997 reference adult male; ICRP 89 gives 18% (M)/27% (F); modern PBPK uses 18% lean, 30-45% obese | Brown RP 1997 ↗Brown RP, Delp MD, Lindstedt SL, Rhomberg LR, Beliles RP. Toxicol Ind Health (1997) |
| Liver | 5.5 | 3.6 | 3.29 | 2.5Brown 1997 covers rhesus (~2.16%); cynomolgus literature reports 2.5-3.0%. | 2.57 | Brown RP 1997 ↗Brown RP, Delp MD, Lindstedt SL, Rhomberg LR, Beliles RP. Toxicol Ind Health (1997) |
| Kidney | 1.67 | 0.73 | 0.55 | 0.48 | 0.44 | Brown RP 1997 ↗Brown RP, Delp MD, Lindstedt SL, Rhomberg LR, Beliles RP. Toxicol Ind Health (1997) |
| Heart | 0.5 | 0.33 | 0.78 | 0.27 | 0.47 | Brown RP 1997 ↗Brown RP, Delp MD, Lindstedt SL, Rhomberg LR, Beliles RP. Toxicol Ind Health (1997) |
| Lung | 0.73 | 0.5 | 0.94 | 0.53 | 0.76 | Brown RP 1997 ↗Brown RP, Delp MD, Lindstedt SL, Rhomberg LR, Beliles RP. Toxicol Ind Health (1997) |
| Brain | 1.65 | 0.57 | 0.8 | 1.5 | 2 | Brown RP 1997 ↗Brown RP, Delp MD, Lindstedt SL, Rhomberg LR, Beliles RP. Toxicol Ind Health (1997) |
| Skin (dermis)Human is dermis only (~3.7%); subcutaneous fat is counted in the adipose row. If you need a "whole-skin" parameter (e.g., topical PBPK), use ~16% (dermis 3.7% + subcutaneous ~12%). Rodent/dog/monkey values follow Brown 1997's broader skin definition and already include subcutaneous — do not double-count adipose for those species. | 16.5 | 19 | 9.1 | 8 | 3.7Dermis-only convention (ICRP 89 ~4.5%; Brown 1997 dermis 3.71%). Brown 1997's broader "skin" of 11.1% includes hypodermis/subcutaneous, which double-counts vs the separate adipose row above. | Brown RP 1997 ↗Brown RP, Delp MD, Lindstedt SL, Rhomberg LR, Beliles RP. Toxicol Ind Health (1997) |
| Bone (skeleton)"Skeleton" (incl. marrow) ~14%; "bone mineral" alone ~5% | 10.7 | 7.3 | 8.8 | 6 | 14.3 | Davies B 1993 ↗Davies B, Morris T. Pharm Res (1993) |
| GI tractWall only; if including lumen contents, +2-4% in fed state | 4.2 | 2.7 | 4.6 | 4 | 1.7 | Brown RP 1997 ↗Brown RP, Delp MD, Lindstedt SL, Rhomberg LR, Beliles RP. Toxicol Ind Health (1997) |
| Spleen | 0.35 | 0.2 | 0.27 | 0.25 | 0.26 | Brown RP 1997 ↗Brown RP, Delp MD, Lindstedt SL, Rhomberg LR, Beliles RP. Toxicol Ind Health (1997) |
Plasma free fraction()— drug examples by tier
Verified human values from Obach 2008 + Lombardo 2018. Method-dependent for highly-bound drugs (): equilibrium dialysis vs ultrafiltration give different numbers.
- Gabapentin1Negligible PPB
- Atenolol0.94
- Metformin0.99
- Caffeine0.65
- Theophylline0.6
- Antipyrine0.95Free-distribution probe drug; minimal binding
- Phenobarbital0.45Reported 0.40-0.55 across sources (Obach 2008, Lombardo 2018); sits at the high/moderate boundary.
- Lidocaine0.3
- Propranolol0.13
- Quinidine0.13
- Verapamil0.1
- Phenytoin0.10.20+ in uremic patients
- Sildenafil0.04
- Diazepam0.02
- Atorvastatin0.02
- Warfarin0.005Equilibrium-dialysis values – (Obach 2008); textbook compilations sometimes round to . Free fraction rises in hypoalbuminemia and on albumin-displacement DDIs.
- Ibuprofen0.005Equilibrium-dialysis values – (Obach 2008 / Lombardo 2018); the lower end is concentration-dependent because albumin sites saturate at therapeutic doses.
- Naproxen0.0017
- Itraconazole0.002
Blood:plasma ratio()— drug examples by tier
Hinderling 1997 is the canonical theory; modern values (Lombardo 2018) confirm tiers. Tacrolimus and cyclosporine are saturable — single B/P numbers oversimplify clinical reality.
- WarfarinB/P = 0.58
- IbuprofenB/P = 0.57
- NaproxenB/P = 0.55
- AtorvastatinB/P = 0.61
- CaffeineB/P = 1.04
- AntipyrineB/P = 1
- AtenololB/P = 1.12
- AcetaminophenB/P = 1.08
- CyclosporineB/P = 1.4Concentration- and hematocrit-dependent saturable binding; range 1.4–2.0
- TacrolimusB/P = 25Range 13-114 in transplant patients; saturable, hematocrit-dependent
- SirolimusB/P = 3695% in RBCs
- ChloroquineB/P = 7Range 5-10
- HydroxychloroquineB/P = 7.2
- AcetazolamideB/P = 5.5Binds carbonic anhydrase in RBCs
Standard scaling exponents()
Hu & Hayton 2001 analysed allometric b-values for 115 xenobiotic datasets (each ≥ 4 species) compiled from the literature; mean , 99% CI - — explicitly excluding . ~81% of individual b-values were statistically indistinguishable from either or . Subset trends: renal-cleared compounds ~, metabolism-dominated ~.
Theory: 0.75: West/Brown/Enquist 1997 fractal vascular network theory (dominant). 0.67: Boxenbaum surface-area theory (relevant for transport-limited compounds).
Hu TM 2001 ↗Hu TM, Hayton WL. AAPS PharmSci (2001)Tissue composition (water, lipid, protein) is roughly mass-proportional across mammals; of free drug () often scales tighter to than total .
Theory: Linear scaling with body mass; tissue partitioning per unit mass is approximately species-invariant.
Mahmood I 2007 ↗Mahmood I. Adv Drug Deliv Rev (2007). Empirically scales less cleanly than CL or because variance compounds. Same quarter-power as gestation length, , and lifespan ("physiological time").
Theory: Mathematical derivation from Vd and CL exponents; reflects allometric "physiological time" (Boxenbaum).
Boxenbaum H 1982 ↗Boxenbaum H. J Pharmacokinet Biopharm (1982)Mahmood Rule of Exponents (CL)
When the empirical animal CL exponent falls in a given band, apply the corresponding correction. Empirical accuracy: ~2-fold for 50-70% of small molecules with ROE applied. ~30-50% of compounds still miss observed CL by >2-fold even with ROE — particularly transporter-mediated and protein-binding-disparate drugs (see Famous Failures).
| Animal CL exponent (b) | Method | Rationale |
|---|---|---|
| 0.55 – 0.70 | Simple allometry (no correction) | Animal CL exponent in this range indicates standard mass-scaling; corrections do not improve prediction. |
| 0.71 – 1.00 | CL × Maximum Life-span Potential (MLP) | Higher exponents in animal data indicate the drug is metabolised faster than mass-scaling predicts; the MLP correction "stretches" predictions toward the human-appropriate scale. Mahmood & Balian 1996 specifies b = 0.71-1.0 (inclusive); exponents > 1.0 use BrW. |
| > 1.00 | CL × Brain Weight (BRW) | Animal CL scaling > 1 indicates extreme species variation in metabolic rate. Brain weight (inversely correlated with CYP-mediated oxidation rates) provides a stronger correction. Mahmood & Balian 1996 gives no upper cap, but ROE accuracy degrades sharply above ~1.3 — prefer IVIVE for transporter- or protein-binding-driven drugs at very high exponents. |
Tang & Mayersohn 2006: predicting allometric failure
Recommendation: apply Tang & Mayersohn FCIM — , where is the simple-allometry intercept ( across animals) and — or use IVIVE / human-relevant in vitro data. The single most actionable predictor of allometric failure (Mahmood 2002, Tang & Mayersohn 2005-2006). When animal-vs-human plasma protein binding diverges by > 5×, simple allometry typically misses observed by an order of magnitude. High lipophilicity () and transporter substrate status are additional risk factors but neither is a standalone failure rule.
Scale a value across species
. Drag the exponent to see the prediction shift. Default exponents (0.75 CL, 1.0 Vd, 0.25 t½) come from the meta-analyses cited above.
Famous allometric scaling failures
Where the equations break, and why. The single feature most differentiating this cheatsheet from PDF competitors. Each fold-error is verified against primary literature.
UCN-01 (7-hydroxystaurosporine)
5800× overpredictionMechanism: Extreme species difference in α1-acid glycoprotein (AAG) binding. Human AAG binds UCN-01 with (Fuse 1998); animal AAG binds weakly. Simple allometry overpredicts UCN-01 human by ~5800× because it ignores the AAG-binding mismatch. Tang & Mayersohn 2005 FCIM corrects for this: , where is the simple-allometry intercept from the animal fit and . With , FCIM brings the prediction to within ~5-fold of observed.
Why it matters: The textbook extreme case. Demonstrates that interspecies differences in plasma protein binding can dominate allometric prediction by orders of magnitude.
Fuse E 1998Diazepam
33× overpredictionMechanism: Hepatic intrinsic clearance is much higher in mouse than human; CYP2C19 + CYP3A4 species differences are the dominant driver. Chimeric humanised-liver mouse work (PXB / TK-NOG models, Kakuni / Yoshizato / Tateno groups) has consistently shown that "humanising" the hepatocyte pool brings rodent CL closer to human, consistent with the species-difference mechanism. Diazepam is mostly albumin-bound, so albumin affinity differences are a secondary contributor; AAG involvement is minor. Simple allometry predicted ~860 mL/min; observed human CL ~26 mL/min.
Why it matters: The original "vertical allometry" case (Boxenbaum & Ronfeld 1983). Recognisable drug; ideal for explaining why CL scaling can fail.
Tang H 2006 ↗Tang H, Mayersohn M. J Pharm Sci (2006)Valproic acid
29× overpredictionMechanism: Plasma protein binding ~90% in humans, lower in rodents. Hepatic glucuronidation rate also species-divergent. Free CL collapse + UGT differences compound.
Why it matters: Widely used drug, illustrates protein-binding-driven failure. Tang & Mayersohn 2006 dataset benchmark case.
Tang H 2006 ↗Tang H, Mayersohn M. J Pharm Sci (2006)Tamsulosin
16× overpredictionMechanism: Strong AAG binding in humans (>99% bound); CYP3A4/CYP2D6 metabolism with marked species differences.
Why it matters: Marketed BPH drug; modern allometric failure exemplar.
Tang H 2006 ↗Tang H, Mayersohn M. J Pharm Sci (2006)Rosuvastatin
3× underpredictionMechanism: Hepatic uptake rate-limited by human OATP1B1 (~77% of hepatic uptake) and OATP1B3 with marked species differences. Oatp1a/1b knockout mice show 8× higher systemic exposure. Without transporter-specific corrections, simple allometry underpredicts CL.
Why it matters: Modern, current-relevance failure case (statin class). Demonstrates that transporter-mediated uptake breaks classical allometric scaling — increasingly relevant for OATP/BCRP/P-gp substrates.
Bowman CM 2019 ↗Bowman CM, Okochi H, Benet LZ. Drug Metab Dispos (2019)Methotrexate
2.5× underpredictionMechanism: Renal OAT3-mediated tubular secretion species differences. OAT3 activity varies 3-5× per kg kidney across species. Knockout mouse data confirm OAT3 is essential for MTX clearance.
Why it matters: Renal allometric failure (most published failures are hepatic). Important for renally-cleared OAT/OCT substrates.
Tang H 2006 ↗Tang H, Mayersohn M. J Pharm Sci (2006)Caco-2 / MDCK / PAMPA permeability tiers
Standard cutoffs (Papp ×10⁻⁶ cm/s). PAMPA values typically run lower than Caco-2 for transcellular drugs (commonly 2–15×) because PAMPA lacks the paracellular pathway. The gap shrinks toward ~1× for highly lipophilic compounds (carbamazepine, naproxen) where transcellular flux dominates in both assays. For paracellular markers (mannitol, atenolol), PAMPA Pe ≈ 0 while Caco-2 Papp captures the leak.
Poor passive permeability; bioavailability often limited; carrier-mediated absorption may rescue.
Adequate for oral absorption; combined paracellular + transcellular routes typical.
Rapid transcellular permeability; absorption rarely the rate-limiting step.
Drug benchmarks (verified Papp values)
Values are cross-lab medians/consensus (Kus 2023 Caco-2 standardisation review + Teksin 2010 head-to-head Caco-2/PAMPA, with Hubatsch 2007 for the mannitol QC threshold). Antipyrine Caco-2 omitted because reported values vary 30–267 ×10⁻⁶ depending on stirring/UWL conditions. PAMPA absolute numbers are protocol-specific (lipid composition, sink conditions, pH) — metoprolol 1.5 here reads 5–15 in sink-condition DS-PAMPA / vendor panels. Use the rank order, not the magnitudes, when comparing to internal data.
| Drug | Caco-2 | PAMPA | Tier | Mechanism |
|---|---|---|---|---|
| Mannitol | 0.2 | 0.02 | low | Paracellular integrity QC marker; Caco-2 cm/s indicates a leaky monolayer (Hubatsch 2007). The ~10× drop in PAMPA vs Caco-2 reflects PAMPA's lack of paracellular pathway — the key conceptual difference between the two assays. |
| Ranitidine | 0.4 | 0.7 | low | BCS Class III (low fa); paracellular |
| Hydrochlorothiazide | 0.7 | 0.5 | low | BCS III/IV; poor passive permeability |
| Atenolol | 0.7 | 0.5 | low | Canonical low-permeability paracellular marker (BCS Class III, in human). PAMPA typically in standard assays. |
| Furosemide | 1 | 0.6 | moderate | BCS IV; paracellular contribution dominates. Highly subclone-dependent (range 0.1-3 across labs). |
| Metoprolol | 40 | 1.5 | high | High transcellular passive permeability; BCS I high-perm reference |
| Propranolol | 50 | 4.3 | high | Transcellular; benchmark high-permeability beta-blocker |
| Carbamazepine | 42 | 27 | high | Highly lipophilic; rapid transcellular |
| Naproxen | 45 | 42 | high | Weak acid ( 4.2); value reported at apical pH 6.5 (gradient assay). Symmetric pH 7.4 lowers ~5-10×; BSA receiver further halves it (in vivo sink). |
| Verapamil | 33 | 9.4 | high | High passive permeability + P-gp substrate (efflux ratio matters in vivo) |
P-gp efflux probes (efflux ratio panel)
Efflux ratio . FDA/EMA threshold for P-gp substrate classification: with significant attenuation by P-gp inhibitor.
ER ~10-20 in MDR1-MDCK, ~15-50 in Caco-2 (subclone-dependent); classical P-gp probe — net efflux requires basolateral uptake
Very high ER; CNS-restricted in vivo because of P-gp efflux at the blood-brain barrier
Range 3-12 across labs; standard P-gp positive control for oral absorption studies
ER ~28-37 in Caco-2; clinically relevant P-gp + OATP substrate
Caco-2 Papp → human jejunal Peff
Units: enter in (e.g. 40 for metoprolol); predicted comes out in .
Coefficients from Sun 2002's all-drug fit (, , pH 7.4). Same paper reports on the passively-absorbed subset alone. Units matter: enter in cm/s and the predicted comes out in cm/s. Carrier-mediated drugs (cephalexin, valacyclovir, levodopa, gabapentin) deviate 3-35× ABOVE the line because Caco-2 underexpresses uptake transporters (PEPT1, OATP) by 2-595× vs human duodenum.