Incorporating published univariable associations in diagnostic and prognostic modeling

Table 1 Overview of approaches

		No meta-analysis	Greenland/Steyerberg	Improved adaptation method
			adaptation method	Variant 1	Variant 2
Step 1	Estimate associations in IPD
	Implemented	Yes	Yes	Yes	Yes
	Association type	m	u+m	u+m	u+m
	Prior distribution	none	none	none	weakly informative
Step 2	Summarize univariable associations
	Implemented	No	Yes	Yes	Yes
	Source	-	I+L	I+L	I+L
	Pooling Method	-	random effects	random effects	random effects
Step 3	Estimate adaptation from univariable to multivariable association
	Implemented	No	Yes	Yes	Yes
	Assumptions	-	(1)+(2)	(1)	(1)
	Estimation procedure	-	analytic	bootstrap	bootstrap
	Prior distributions	-	none	none	weakly informative
Step 4	Apply adaptation to summary estimate from the literature and estimate β _m\|L
	Implemented	No	Yes	Yes	Yes

This overview illustrates the characteristics of the approaches discussed and used in the simulation study. In the first step, univariable (u) and multivariable (m) associations are estimated in the IPD. In the second step, the univariable associations from the literature (L) and data at hand (I) are summarized. Afterwards, the adaptation from univariable to multivariable association is estimated in step 3. The assumptions about the variance component here are as follows: (1) estimated associations in the individual participant data (IPD) are independent from estimated associations in the literature, and (2) $Cov ({\hat{β}}_{m | I}, {\hat{β}}_{u | I}) = Var ({\hat{β}}_{u | I})$ . Finally, step 4 estimates a multivariable association by applying the adaptation to the univariable summary estimate from the literature.

ISSN: 1471-2288