Table 2.

Comparison of the impact of relaxing different eligibility criteria in patients taking drug D1

Re-included subpopulations		Natural augmentation (number of patients re-included when opening the trial to the specific “real-world population” subgroup)	Prediction bias with natural augmentation	Mean squared error (MSE) of prediction with natural augmentation
Relaxed eligibility criteria	Illness duration between 1 and 3 years	188	0.033	0.818
	1 past suicide attempt	190	0.057	0.820
	Private practice	159	0.041	0.830
	Alcohol abuse	73	0.054	0.836
	Drug abuse	65	0.053	0.833
	Illness duration between 1 and 3 years + 1 past suicide attempt	339	0.024	0.803
	Illness duration between 1 and 3 years + private practice	321	0.024	0.814
	Illness duration between 1 and 3 years + alcohol abuse	248	0.037	0.812
RCT population		not applicable	0.054	0.852
SOHO “real-world population”		not applicable	0.000	0.000

Results for the “RCT population” and SOHO “real-world populations” are displayed as benchmark