Table 6.

Summary of the TTC models published since 2009.

Reference	Data input	No. of chemicals	Grouping- discriminate chemicals based on toxicity potency	Discriminate local/systemic	TTC
Carthew et al., 2009	Local NOAEC Systemic NOAEL (EPA, BfR, TNO, ECETOC)	92	Cramer Classes (CC)	Yes	Local:	CC1: 1400 μg/d CC3: 470 μg/d
					Systemic:	CC1: 980 μg/d CC3: 170 μg/d
Escher et al., 2010	NOEC RepDose, local/sys, organic compounds	203	Cramer Classes	CC1 only	CC1: 71 μg/d CC3: 4 μg/d
Tluczkiewicz et al., 2016	NOEC RepDose	296	Structure factors identified for high vs low NOEC; machine learning; 28 groups: 19 high, 9 low	No	Low toxic: 4260 μg/d High toxic: 2 μg/d
Schüürmann et al., 2016	NOEC RepDose	296	Structural alerts identified for high vs low NOEC, machine learning. Physicochemical properties, bioavailability, metabolism, MoA, to explore	No	Low Tox NOEC >12 ppm High Tox NOEC <0.75 ppm
Hoersch et al., 2018	IFA GESTIS DNEL list	1876	Statistical DNEL distribution (99th percentile, 8 h occupational exposure)	Yes	50 μg/m3 corresponding to 500 μg/worker/d
Nelms and Patlewicz, 2020	ToxVal database (subacute, subchronic, chronic, reproductive, developmental, multigeneration toxicity studies)	4703 (of which 613 used in TTC)	Identified chemical structure, process through Kroes and Patlewicz profilers (OECD Toolbox, ToxTree). Filter for relevant studies (species, duration), remove statistical outliers, taking minimum NOAEL/NOAEC MoA sub-categories, Bootstrapping to explain uncertainty of 95thpercentile CC3, MoA profiling scheme for aquatic toxicity → reactive and baseline	No (inconclusive)	Comparable to Escher et al.: CC3 CC1: 8.23 μg/d CC3: 4.28 μg/d Baseline: 22.4 μg/d Reactive: 4.3 μg/d
RIFM/P&G	Carthew, RepDos, P&G, RIFM, ECHA	246	Hierarchical clustering with machine learning 5 clusters, 4 features	Yes