Supplemental data:

Body mass index modifies the risk of cardiovascular death in proteinuric chronic kidney disease

RP. Obermayr¹, C. Temml², G. Gutjahr³, A. Kainz⁴, R. Klauser-Braun¹, R. Függer⁵, R. Oberbauer⁴

TABLES		FIGURES
	Webtable 1 Tests for proportional hazards assumption concerning cardiovascular mortality		Webfigure 1 Scaled Schoenfeld residuals for the factor smoke concerning cardiovascular mortality. Legend: Under the proportional hazards assumption, the residuals are constant over time. For the factor smoke, residuals are decreasing. Therefore a time dependent term was used for the factor smoke in the final model.
	Webtable 2 Effect contrasts for cardiovascular death (alternative model 1) Legend: To validate the model of the main text (table 3a) we fitted three additional models: In alternative model 1, variables were included as confounders in a multivariable Cox model if the change in the Chi-Square statistics of the BMI*GFR interaction term exceeded 10%. Contrasts were calculated identically to Table 3a of the main paper. Wald test for the effect size of the BMI*GFR interaction term: P = 0.0001.
	Webtable 3 Effect contrasts for cardiovascular death (alternative model 2) Legend: To validate the model of the main text (table 3a) we fitted three additional models: For alternative model 2 a propensity score was created by logistic regression of cardiovascular death on all potential confounders. Alternative model 2 uses this propensity score in a Cox regression model stratified by quintiles of the propensity score. Contrasts were calculated identically to Table 3a of the main paper. Wald test for the effect size of the BMI*GFR interaction term: P = 0.0070.
	Webtable 4 Effect contrasts for cardiovascular death (alternative model 3) Legend: To validate the model of the main text (table 3a) we fitted three additional models: For alternative model 2 a propensity score was created by logistic regression of cardiovascular death on all potential confounders. Alternative model 3 uses this propensity score as time dependent covariable in a Cox regression model. Contrasts were calculated identically to Table 3a of the main paper. Wald test for the effect size of the BMI*GFR interaction term: P = 0.0001.
	Webtable 5 Effect contrasts for cancer death (alternative model 1) Legend: To validate the model of the main text (table 3a) we fitted three additional models: In alternative model 1, variables were included as confounders in a multivariable Cox model if the change in the Chi-Square statistics of the BMI*GFR interaction term exceeded 10%. Contrasts were calculated identically to Table 3a of the main paper. Wald test for the effect size of the BMI*GFR interaction term: P = 0.0808.
	Webtable 6 Effect contrasts for cancer death (alternative model 2) Legend: To validate the model of the main text (table 3a) we fitted three additional models: For alternative model 2 a propensity score was created by logistic regression of cardiovascular death on all potential confounders. Alternative model 2 uses this propensity score in a Cox regression model stratified by quintiles of the propensity score. Contrasts were calculated identically to Table 3a of the main paper. Wald test for the effect size of the BMI*GFR interaction term: P = 0.2609.
	Webtable 7 Effect contrasts for cancer death (alternative model 3) Legend: To validate the model of the main text (table 3a) we fitted three additional models: For alternative model 2 a propensity score was created by logistic regression of cardiovascular death on all potential confounders. Alternative model 3 uses this propensity score as time dependent covariable in a Cox regression model. Contrasts were calculated identically to Table 3a of the main paper. Wald test for the effect size of the BMI*GFR interaction term: P = 0.0829.