Table 3.

Multiple regression analysis for postoperative ICL vault prediction in the development dataset

	Combination of variables	Pearson’s correlation coefficient	P-value for correlation	MAE (µm)	Bayesian information criterion
Development set (self-validation) (N = 696)	Selection from STAAR nomogram - ICL size, WTW, ACD	0.516	< 0.001	109.8	6891.4
	Selection from KS formula – ICL size, ATA (= ACA distance)	0.427	< 0.001	118.3	6961.0
	Forward stepwise selection – age, mean K, ICL size, ICL power, CCT, AQD, ACA distance, LT	0.636	< 0.001	101.7	6789.2
	Backward stepwise selection – age, toric lens, SE, mean K, ICL size, AQD, ACA distance, LT, WTW	0.608	< 0.001	98.3	6810.1
	Partial least squares – age, sex, toric lens, mean K, ICL size, ICL power, CCT, AQD, ACV, LT, ACA distance, pupil diameter, WTW	0.664	< 0.001	96.9	6801.3
	LASSO – ICL size, AQD, ACA distance, LT	0.607	< 0.001	101.8	5769.9
Validation set (N = 196)	Selection from STAAR nomogram - ICL size, WTW, ACD	0.505	< 0.001	112.3	1917.2
	Selection from KS formula – ICL size, ATA (= ACA distance)	0.473	< 0.001	116.9	1920.6
	Forward stepwise selection – age, mean K, ICL size, ICL power, CCT, AQD, ACA distance, LT	0.580	< 0.001	104.9	1917.5
	Backward stepwise selection – age, toric lens, SE, mean K, ICL size, AQD, ACA distance, LT, WTW	0.578	< 0.001	105.7	1927.6
	Partial least squares – age, sex, toric lens, mean K, ICL size, ICL power, CCT, AQD, ACV, LT, ACA distance, pupil diameter, WTW	0.577	< 0.001	105.8	1935.1
	LASSO – ICL size, AQD, ACA distance, LT	0.582	< 0.001	104.7	1894.9

ACA anterior chamber angle, ACV anterior chamber volume, AQD aqueous depth, ATA angle-to-angle distance, CCT central corneal thickness, LASSO least absolute shrinkage and selection operator, LT lens thickness, MAE mean squared error, SE spherical equivalent, WTW white-to-white distance