Table 4.

Table showing the average linear predictor estimates, together with associated standard errors, and estimates of $\mathbb{P}({\text{ACR}}^{(24)}20=1)$ for the different treatment groups using both the additive and interaction estimation models

Treatment group	Imputation method
	MLE	LI	DIRECT	TRUE Data
Additive Model: average linear predictor estimate (standard error)
None	$-0.510$ (0.209)	$-0.586$ (0.221)	$-0.525$ (0.226)	$-0.514$ (0.196)
CSP	$-0.410$ (0.210)	$-0.400$ (0.221)	$-0.341$ (0.229)	$-0.322$ (0.199)
PDN	$-0.395$ (0.202)	$-0.520$ (0.208)	$-0.471$ (0.217)	$-0.458$ (0.193)
Both	$-0.294$ (0.198)	$-0.334$ (0.205)	$-0.288$ (0.209)	$-0.265$ (0.187)
Additive Model: average estimate of $\mathbb{P}({\text{ACR}}^{(24)}20=1)$
None	0.375	0.358	0.372	0.374
CSP	0.399	0.401	0.416	0.420
PDN	0.403	0.373	0.384	0.388
Both	0.427	0.417	0.429	0.434
Interaction Model: average linear predictor estimate (standard error)
None	$-0.329$ (0.239)	$-0.411$ (0.247)	$-0.315$ (0.253)	$-0.316$ (0.222)
CSP	$-0.614$ (0.254)	$-0.593$ (0.263)	$-0.574$ (0.276)	$-0.539$ (0.238)
PDN	$-0.574$ (0.236)	$-0.697$ (0.247)	$-0.685$ (0.253)	$-0.658$ (0.229)
Both	$-0.133$ (0.223)	$-0.179$ (0.23)	$-0.101$ (0.266)	$-0.089$ (0.211)
Interaction Model: average estimate of $\mathbb{P}({\text{ACR}}^{(24)}20=1)$
None	0.419	0.399	0.422	0.422
CSP	0.352	0.356	0.361	0.368
PDN	0.360	0.333	0.335	0.341
Both	0.467	0.455	0.475	0.478

Results are shown where multiple imputation was performed for all outcomes using maximum likelihood estimation (MLE), for all outcomes using linear increments (LI), direct imputation of ${\text{ACR}}^{(24)}20$ using ML (DIRECT) and estimates produced using data prior to the application of a missingness structure (TRUE Data). Missing data occurred at 12, 18 and 24 months