Table 5. Odds ratios between prostate cancer death and study arm (screening vs control), by four estimation methods.

	Estimation method
Country	Empiricala	Empirical, corrected using overall estimates of adjudicator accuracyb	Empirical, corrected using differential estimates of adjudicator accuracy by study armc	Directly from latent class modeld
Netherlands	0.342	0.35	0.337	0.328
Belgium	0.759	0.904	0.866	0.902
Sweden	0.355	0.381	0.395	0.368
Finland	0.52	0.575	0.568	0.556
Switzerland	0.625	0.5	0.259	0.437

^aEstimated from cross-tabulation of adjudication consensus by study arm.

^bEstimated proportions of prostate cancer deaths in each study arm were corrected using estimated false positive and false negative adjudication rates in LCM 2. Odds ratio is then calculated from these corrected proportions.

^cSimilar to approach (b), except that adjudicator accuracy was estimated from LCM 3.

^dBased on LCM 2 estimates of the association of study arm with the latent variable (prostate cancer death).