. 2021 Apr 19;13:82. doi: 10.1186/s13148-021-01055-z

Table 3.

Bootstrapped differences in R², SE, and MAD between different clocks and GA estimation methods

Dataset * (count)	Comparison between clocks	R² (95% CI)	SE (95% CI)	MAD (95% CI)
START non-ART (n = 200)	450 K/EPIC overlap – EPIC GA	−0.0001 (−0.021, 0.018)	0.001 (−0.142, 0.175)	0.162 (−0.375, 0.794)
	ETD-based – EPIC GA	0.048 (−0.041, 0.123)	−0.409 (−1.00, 0.335)	0.645 (−0.181, 1.209)
	ETD-based – 450 K/EPIC overlap	0.048 (−0.039, 0.119)	−0.410 (−1.03, 0.308)	0.483 (−0.409, 0.984)
PREDO Non-ART (n = 148)	Bohlin – EPIC GA	−0.062 (−0.117, −0.014)	0.528 (0.095, 0.994)	3.27 (1.87, 3.92)
	Knight – EPIC GA	−0.247 (−0.342, −0.161)	1.89 (1.97, 2.69)	1.13 (0.196, 2.40)
	Knight – Bohlin	−0.185 (−0.273, −0.102)	1.36 (0.698, 1.97)	−2.15 (−3.11, −0.382)

Dataset * (count)	Comparison between GA estimation methods	R² (95% CI)	SE (95% CI)	MAD (95% CI)
START ART (n = 838)	ETD – ultrasound	0.015 (−0.003, 0.033)	−0.284 (−0.544, −0.037)	−0.102 (−0.465, 0.174)

Dataset * (count)

Comparison between GA estimation methods

R² (95% CI)

SE (95% CI)

MAD (95% CI)

START

ART (n = 838)

ETD – ultrasound

0.015 (−0.003, 0.033)

−0.284 (−0.544, −0.037)

−0.102 (−0.465, 0.174)

*See Table 1 and Fig. 1 for further details on these datasets

GA gestational age, SE standard error, MAD median absolute deviation, ETD embryo transfer date