IVIVC Level A
Upload an in vitro dissolution profile and the matching in vivo PK for two or more oral formulations. The engine deconvolutes %absorbed via Wagner-Nelson, correlates it against %dissolved, and predicts a held-out formulation's Cmax/AUC from its dissolution alone β the classical %PE check (FDA 1997 Guidance V.B, threshold only, not an automatic pass/fail).
- Oral formulations only, no IV arm. This form does not accept an IV dataset. If your study has an IV arm, this report does not apply to it β contact us instead of forcing this route.
- ke is estimated per formulation from its own terminal phase (log-linear regression), same method this platform uses for NCA Ξ»z.
- With few formulations, the correlation and its internal predictability check are far below the sample a real Level A submission needs β the report states N explicitly, it never hides it.
- If a formulation's dissolution doesn't fit any classified model above RΒ²=0.90, the report marks that correlation NOT CONCLUSIVE instead of forcing a number.
- The report also includes a fractional (Mittag-Leffler) variant, side by side β research-only, NOT FDA-recognized, with a known quadrature-singularity limitation on sparse time grids, printed next to its own numbers.
1 Β· Upload dissolution + PK data
βΆ Accepted CSV format Β· click to expand
Dissolution CSV β exact column names required:
formulation, time_h, fdiss_pct
(cumulative % dissolved, 0-100).
formulation,time_h,fdiss_pct REF,0,0 REF,0.5,22.4 REF,1,41.0 REF,2,68.5
PK CSV β exact column names required:
formulation, time_h, conc_ng_ml
(plasma concentration, ng/mL).
formulation,time_h,conc_ng_ml REF,0,0 REF,0.25,238.8 REF,0.5,1033.9 REF,1,1850.2
The formulation labels must match between both
files. At least 2 formulations are needed to fit the correlation,
plus 1 more held out to test it (3 minimum, in total).
Do NOT submit datasets with direct patient identifiers. Use pseudonymised/formulation-level data only.
2 Β· Choose fit vs. held-out formulations
Formulations found in BOTH files. Pick 2+ to fit the correlation, and exactly 1 different one to hold out and predict.