Coming from WinNonlin / NONMEM¶

A quick-reference guide for scientists transitioning from Phoenix WinNonlin and NONMEM to OpenPKFlow.

WinNonlin NCA -> OpenPKFlow NCA¶

WinNonlin	OpenPKFlow	Notes
`NCA -> Model 200 (plasma)`	`NCAStudy.from_csv()` + `study.analyze()`	CSV input, returns `NCASummaryResults`
`AUClast (linear trapezoidal)`	`auc_linear(times, concs)`	Explicit method required
`AUClast (log-linear trapezoidal)`	`auc_log(times, concs)`	Returns `AUCResult` with warnings
`AUC linear-up/log-down`	`auc_linear_up_log_down(times, concs)`	FDA/EMA default
`AUCINF_obs`	`NCAResult.auc_inf_obs`	`AUClast + Clast/lambda_z`
`%AUCextrap`	`NCAResult.auc_percent_extrapolated`	Warns if >20% (FDA criterion)
`Cmax`	`cmax(concs)`	`NCAResult.cmax`
`Tmax`	`tmax(times, concs)`	`NCAResult.tmax`
`lambda_z` (best-fit)	`lambda_z(times, concs, method="auto")`	BAR² auto-selects tail window
`t1/2`	`NCAResult.half_life`	`ln(2) / lambda_z`
`CL_F` (oral)	`NCAResult.cl_f`	Apparent clearance
`Vz_F` (oral)	`NCAResult.vz_f`	Apparent volume
`CL` (IV)	`NCAResult.cl`	Absolute clearance
BLQ handling (set to 0)	`blq_method="zero"`	One of: none/drop/zero/half_lloq/lloq
Steady-state NCA	`NCAStudy(..., steady_state=True, tau=12)`	AUCtau, Cmax_ss, Cmin_ss, fluctuation%
Urinary excretion	`NCAStudy(..., urine_volume_col="vol", urine_conc_col="conc_u")`	Ae, CLr, %excreted
CDISC PP output	`summary.to_cdisc_pp()`	PPTESTCD/PPORRES/PPORRESU format
Dose-normalised params	`NCAResult.dn_auclast`, `.dn_cmax`	Per-mg normalisation

Example¶

from openpkflow.nca import NCAStudy

study = NCAStudy.from_csv("subjects.csv", auc_method="linear_up_log_down")
results = study.analyze()
print(results.summary())
results.report("nca_report.html")

WinNonlin Bioequivalence -> OpenPKFlow BE¶

WinNonlin	OpenPKFlow	Notes
`BE -> 2x2 crossover (TOST)`	`be_tost(reference, test)`	Paired per-subject log-differences
GMR + 90% CI	`BETOSTResult.gmr`, `.gmr_lower_90ci`, `.gmr_upper_90ci`	Default limits 80-125%
Intra-subject CV%	`BETOSTResult.cv_intra_pct`	Back-transformed from log-diff SD
BE verdict (PASS/FAIL)	`BETOSTResult.bioequivalent`	Boolean flag
NTI limits (90-111%)	`be_tost(..., be_lower=0.90, be_upper=1.1111)`
Sample size (power 80%)	`be_sample_size(gmr=0.95, cv=0.20)`	Sequential search
Power for given N	`be_tost_power(gmr=0.95, cv=0.20, n=24)`	Non-central t exact method
BE report	`result.report("be_report.html")`	HTML with CI bar chart
Export to BioEqPy	`study.to_bioeqpy_csv()`	For RSABE/replicate designs

WinNonlin IVIVC -> OpenPKFlow IVIVC¶

WinNonlin	OpenPKFlow	Notes
Wagner-Nelson deconvolution	`wagner_nelson(times, concs)`	Returns F_a fraction absorbed
Loo-Riegelman deconvolution	`loo_riegelman(times, concs, dose, k10, k12, k21)`	2-compartment method
Convolution prediction	`convolution_predict(diss_times, diss_frac, ir_fn)`	Numerical convolution
Levy plot	`levy_plot_data(f_in_vitro, f_in_vivo)`	Linear regression R²
Predictability (%PE)	`ivivc_predictability(...)`	FDA <=15% individual, <=10% mean

Example¶

from openpkflow.ivivc import IVIVCStudy

study = IVIVCStudy(dissolution_csv="dissolution.csv", plasma_csv="plasma.csv",
                   dose=100, method="wagner_nelson")
result = study.run()
result.report("ivivc_report.html")

Phoenix NLME / NONMEM -> OpenPKFlow Pop PK¶

Phoenix NLME / NONMEM	OpenPKFlow	Notes
FOCE-I (METHOD=1)	`run_foce_i(model, data)`	scipy L-BFGS-B, zero extra deps
SAEM (METHOD=3)	`run_saem(model, data)`	PyMC optional, numpy fallback
`$PK` block	`PopPKModel(route="oral", n_cmt=1, omega_type="diagonal")`	Frozen dataclass
`$THETA`	`model.to_theta()`/`model.from_theta()`	Pack/unpack for optimizer
`$OMEGA`	`PopPKModel(omega_type="full")`	Diagonal or full block matrix
`$SIGMA`	Built into proportional error model	Fixed at v2.3.0
`$DATA`	`load_pop_csv(path)`	NONMEM-style CSV (ID, TIME, DV, EVID, MDV)
`$TABLE` output	`result.to_dataframe()`	Parameter estimates + SEs
Objective function (-2LL)	`PopPKResult.minus_2ll`	Also AIC, BIC
EBE shrinkage	`PopPKResult.ebe_shrinkage`	Per-parameter shrinkage
SEs (delta method)	`PopPKResult.se_*` fields	Via inverse Hessian
2-compartment models	`PopPKModel(n_cmt=2)`	Oral or IV bolus
IV bolus route	`PopPKModel(route="iv_bolus")`	Supports 1-cmt and 2-cmt
Covariates	Not available (removed in v2.3.0)	Use Pharmpy/nlmixr2 for covariate selection
GOF diagnostic plots	`result.plot()`	6-panel OBS/IPRED/CWRES/EBE
VPC	`simulate_vpc(model, data)`	Simulation-based percentile bands
Population dataset	`create_nonmem_dataset(...)`	Dose + observation records merged

Example¶

from openpkflow.pop import PopPKModel, run_foce_i

model = PopPKModel(route="oral", n_cmt=1, omega_type="diagonal")
result = run_foce_i(model, "theoph.csv")
print(result.summary())
result.report("pop_report.html")

NONMEM VPC -> OpenPKFlow Pop Diagnostics¶

NONMEM (PsN `vpc`)	OpenPKFlow	Notes
Simulation-based VPC	`simulate_vpc(model, data, n_sim=500)`	Users `sim/` analytical engine
5th/50th/95th percentile bands	`VPCResult.observed_bands`, `.simulated_bands`	Time-binned
VPC plot	`vpc_result.plot()` or `vpc_result.report()`	scatter + band overlay

WinNonlin Dissolution -> OpenPKFlow Dissolution¶

WinNonlin	OpenPKFlow	Notes
f1 (difference factor)	`f1(reference, test)`	0 = identical
f2 (similarity factor)	`f2(reference, test)`	100 = identical, >=50 passes
Bootstrap f2	`study.bootstrap_compare(ref, test)`	CI for samples <12 vessels
Max deviation	`max_deviation(reference, test)`	FDA alternative when f2 cannot be used
MSD (Mahalanobis)	`msd(reference, test_matrix)`	Chi-squared test
Model fitting (Weibull)	`study.fit_models(formulation, models=["weibull"])`	AICc ranking
Model fitting (KP)	`models=["korsmeyer_peppas"]`	60% rule check
Model fitting (Higuchi)	`models=["higuchi"]`
Multi-media (pH panel)	`MultiMediaStudy({...})`	ICH M13A/B
Alcohol dose-dumping	Supported in multi-media	Separate medium with ethanol
Report generation	`result.report("out.html")`	HTML/PDF/DOCX/MD

Example¶

from openpkflow.dissolution import DissolutionStudy

study = DissolutionStudy.from_csv("dissolution.csv")
result = study.compare("Reference", "Test")
print(f"f2 = {result.f2_value:.1f}")  # >= 50: similar
result.report("dissolution_report.html")

PK Simulation -> OpenPKFlow Sim¶

Phoenix WinNonlin PKS	OpenPKFlow	Notes
1-cmt IV bolus	`c_1cmt_iv_bolus(times, dose, CL, Vz)`	Gibaldi & Perrier 2nd ed.
1-cmt IV infusion	`c_1cmt_iv_infusion(times, dose, CL, Vz, duration)`	Constant-rate infusion
1-cmt oral	`c_1cmt_oral(times, dose, CL_F, Vz_F, ka)`	Bateman equation
2-cmt IV bolus	`c_2cmt_iv_bolus(times, dose, CL, V1, Q, V2)`	Biexponential
2-cmt oral	`c_2cmt_oral(times, dose, CL_F, Vz_F, ka, Q, V2)`	3-exponential
Repeated dosing	`DoseRegimen.from_repeated(...)` + `simulate()`	Superposition
Model object	`OneCompartmentModel(...)` or `TwoCompartmentModel(...)`	CL/V parameterization

Key Differences¶

Parameterization: OpenPKFlow uses CL/V (or CL_F/Vz_F), not rate constants (k10, k12, k21). Clearance and volume are more interpretable and align with NCA output naming.

No GUI: OpenPKFlow is a Python library + CLI. All outputs are scriptable, versionable, and CI-compatible.

No covariates in Pop PK: Covariate estimation code was removed in v2.3.0. Use Pharmpy or nlmixr2 for covariate selection, then OpenPKFlow for base model estimation.

Reports are static HTML/PDF/DOCX/MD: No interactive dashboards. All report templates are Jinja2 and can be customized.

RSABE / replicate BE: Not in OpenPKFlow. Use the companion BioEqPy package with to_bioeqpy_csv() export.

Environment R -> OpenPKFlow¶

R (PKNCA/nlmixr2)	OpenPKFlow	Notes
`PKNCA::pk.nca()`	`NCAStudy.from_csv().analyze()`	Cross-validated against PKNCA 0.12.1
`nlmixr2::nlmixr()` FOCE-I	`run_foce_i()`	Cross-validated against nlme (Pinheiro & Bates 2000, Table 8.1)
`nlmixr2::nlmixr()` SAEM	`run_saem()`	PyMC Metropolis + Robbins-Monro
`PowerTOST::power.TOST()`	`be_tost_power()`	Exact non-central t method
`PowerTOST::sampleN.TOST()`	`be_sample_size()`	Cross-validated against PowerTOST 1.5-7
`bootf2::calcf2()`	`f2(method="all_points")`	Point estimate validated
`lm()` / `optim()` model fits	`fit_dissolution_models()`	All 5 models cross-validated against base R