Table 5.

Results of simulations assessing main/internal validation study-based analysis under case-control sampling^a,b

Non-differential Case: SE1 = SP1 = SE0= SP0 = 0.8
Model	${\widehat{\beta }}_1$ (SD)	95% CI coverage	${\widehat{\beta }}_2$ (SD)	95% CI coverage
“Naïve”	0.30 (0.06)	0.0%	0.07 (0.16)	22.7%
Main/internal validationb	2.04 (0.20)	95.3%	0.51 (0.27)	95.1%

Differential Case: SE1=0.8, SP1=0.7, SE0=0.6, SP0=0.9
Model	${\widehat{\beta }}_1$ (SD)	95% CI coverage	${\widehat{\beta }}_2$ (SD)	95% CI coverage
“Naïve”	0.30 (0.07)	0.0%	1.28 (0.16)	0.0%
Main/internal validationb	2.02 (0.20)	95.2%	−0.07 (0.30)	48.0%

^a1000 simulations in each case; 500 internal validation and roughly 1,000 main study observations per simulation based on 100% and 5% sampling of cases and controls, respectively. True model (1) parameters: (β₀,β₁β₂) = (−0.4, 2.0, 0.5)

^bMain/internal validation study likelihood defined in Eq. (7)–(8)