TABLE 1

[I_1T]/IC₅₀ and [I₂]/IC₅₀ threshold calibrations in different laboratories/companies and regulatory agencies

The in vitro tool and substrate, as well as the output parameter used to estimate the IC₅₀ are listed in the second column. The fourth column contains the number of clinical studies used in the training set to establish the threshold in each group.

Group	In Vitro Tool (Equation)/Substrate	Thresholds	No. of Clinical Studies (Training Set)	Comments	References
EMA	Unknown	[I_1u]/IC₅₀ < 0.02	Unknown	No rationale presented	(EMA^a)
EMA	Unknown	[I₂]/IC₅₀ < 10	Unknown	No rationale presented	(EMA^a)
FDA	Diverse and unknown	[I_1T]/IC₅₀ < 0.1	14	IC₅₀s from different origin no statistical analysis to establish thresholds	(Zhang et al., 2008)
FDA	Diverse and unknown	[I₂]/IC₅₀ < 10	14		(Zhang et al., 2008)
FDA	Diverse and unknown	[I_1T]/IC₅₀ < 0.1	11	11 clinical DDIs not included in Zhang et al., 2008	(Agarwal et al., 2013)
FDA	Diverse and unknown	[I₂]/IC₅₀ < 10	11	IC₅₀s from different origin no statistical analysis to establish thresholds	(Agarwal et al., 2013)
P-gp Inhibition Working Group	Diverse (P_appBA)/diverse	[I_1T]/IC₅₀ < 0.03	30	Testing set (n = 73) – IC₅₀ laboratory-to-laboratory variability included statistical analysis to refine thresholds	(Ellens et al., 2013)
P-gp Inhibition Working Group	Diverse (P_appBA)/diverse	[I₂]/IC₅₀ < 45	30		(Ellens et al., 2013)
Pfizer	Caco-2 (NSF)/digoxin	[I_1T]/IC₅₀ < 0.1	26	IC₅₀s from same origin statistical analysis to refine thresholds	(Fenner et al., 2009; Cook et al., 2010)
Pfizer	Caco-2 (NSF)/digoxin	[I₂]/IC₅₀ < 5	26		(Fenner et al., 2009; Cook et al., 2010)
AstraZeneca	Caco-2 (P_appBA)/digoxin	[I_1T]/IC₅₀ < 0.1	1^b	1 IC₅₀ / 1 drug / 3 doses accurate dose-dependent prediction	(Elsby et al., 2011)
AstraZeneca	Caco-2 (P_appBA)/digoxin	[I₂]/IC₅₀ < 10	1^b		(Elsby et al., 2011)
Takeda	LLC-PK1-MDR1 (ER)/digoxin	[I₁]/IC₅₀ < 0.1	36	IC₅₀s from same origin statistical analysis to refine [I₂]/IC₅₀ threshold only	(Sugimoto et al., 2011b)
		[I₂]/IC₅₀ < 30
		IC₅₀ < 2 µM
Chugai	Diverse	DIN < 10.8 L	21	IC₅₀s from different origin focus on intestinal DDI (digoxin/fexofenadine / talinolol)	(Tachibana et al., 2009)
Chugai	Diverse	[I₂]/IC₅₀ < 43.2	21		(Tachibana et al., 2009)
Merck/ GLSynthesis	Liposomal system (hamster P-gp) = fluorosome-trans-pgp	[I_1T]/IC₅₀ < 0.03	16	Nonhuman P-gp / no drug-like molecule as substrate/no real analysis of the cut-off no statistical analysis to establish thresholds	(Melchior et al., 2012)
Merck/ GLSynthesis	Liposomal system (hamster P-gp) = fluorosome-trans-pgp	[I₂]/IC₅₀ < 30	16		(Melchior et al., 2012)
Takeda	LLC-PK1-MDR1 (MCFR)/substrate matching clinical observation	[I_1u]/K_i < 1	12	IC₅₀s from same origin assessment of P-gp inhibition threshold at the BBB	(Sugimoto et al., 2013)
Roche	LLC-PK1-MDR1 (ER)/digoxin	[I₂]/IC₅₀ < 6.5	68	Testing set (n = 12 + 16) statistical analysis to refine thresholds and select IC₅₀ equation

BBB, blood-brain barrier; MCFR, modified corrected flux ratio; NSF, net secretory flux.
↵^a http://www.ema.europa.eu/docs/en_GB/document_library/Scientific_guideline/2010/05/WC500090112.pdf
↵^b Single study with three doses.