Bariatric surgery is associated with improvement in kidney outcomes Alex R. Chang, Yuan Chen, Christopher Still, G. Craig Wood, H. Lester Kirchner, Meredith Lewis, Holly Kramer, James E. Hartle, David Carey, Lawrence J. Appel, Morgan E. Grams Kidney International Volume 90, Issue 1, Pages 164-171 (July 2016) DOI: 10.1016/j.kint.2016.02.039 Copyright © 2016 International Society of Nephrology Terms and Conditions
Figure 1 Trajectories of weight and eGFR in bariatric surgery patients and matched controls. Trajectories of weight and estimated glomerular filtration rate (eGFR) were adjusted for propensity score and baseline weight and eGFR, respectively, in mixed effects models with random intercept and slopes for each individual, allowing unstructured correlation between the random effects. Time was modeled using restricted cubic splines with 5 knots at the 5th, 27.5th, 50th, 72.5th, and 95th percentiles. Shaded areas represent 95% confidence interval bounds. Kidney International 2016 90, 164-171DOI: (10.1016/j.kint.2016.02.039) Copyright © 2016 International Society of Nephrology Terms and Conditions
Figure 2 Kaplan-Meier curves estimating time to kidney outcomes by surgery group and control group. The estimated glomerular filtration rate (eGFR) decline ≥30% outcome was defined as having a follow-up outpatient eGFR ≥30% lower than the baseline eGFR value. ESRD was defined as eGFR <15 ml/min/1.73 m2 or treated ESRD per US Renal Data System registry. Shaded areas represent 95% confidence interval bounds. Kidney International 2016 90, 164-171DOI: (10.1016/j.kint.2016.02.039) Copyright © 2016 International Society of Nephrology Terms and Conditions
Figure 3 Trajectories of eGFR stratified by baseline eGFR. Trajectories of estimated glomerular filtration rate (eGFR) for patients with baseline eGFR ≥90 and <90 were examined using mixed effects models adjusted for propensity score and baseline eGFR, with random intercept and slopes for each individual, allowing unstructured correlation between the random effects. Time was modeled using restricted cubic splines with 5 knots at the 5th, 27.5th, 50th, 72.5th, and 95th percentiles. Shaded areas represent 95% confidence interval bounds. Kidney International 2016 90, 164-171DOI: (10.1016/j.kint.2016.02.039) Copyright © 2016 International Society of Nephrology Terms and Conditions
Figure S1 Distribution of propensity scores in matched surgery and control patients. Controls were selected by matching patients without bariatric surgery with those with bariatric surgery using caliper sizes one-half the SD of the propensity scores. Matched controls were successfully identified for 985 of 1394 bariatric surgery patients. Kidney International 2016 90, 164-171DOI: (10.1016/j.kint.2016.02.039) Copyright © 2016 International Society of Nephrology Terms and Conditions
Figure S2 Trajectories of systolic blood pressure. Trajectories of systolic blood pressure were examined, using mixed-effects models adjusted for propensity score, with random intercept and slopes for each individual, allowing unstructured correlation between the random effects. Time was modeled using restricted cubic splines with 5 knots at the 5th, 27.5th, 50th, 72.5th, and 95th percentiles. Shaded areas represent 95% confidence interval bounds. Kidney International 2016 90, 164-171DOI: (10.1016/j.kint.2016.02.039) Copyright © 2016 International Society of Nephrology Terms and Conditions
Figure S3 Prevalence of ICD-9 diagnosis of hypertension over time. Prevalence of hypertension over time was examined, using mixed-effects models adjusted for propensity score, with random intercept and slopes for each individual, allowing unstructured correlation between the random effects. Time was modeled using restricted cubic splines with 5 knots at the 5th, 27.5th, 50th, 72.5th, and 95th percentiles. Shaded areas represent 95% confidence interval bounds. Kidney International 2016 90, 164-171DOI: (10.1016/j.kint.2016.02.039) Copyright © 2016 International Society of Nephrology Terms and Conditions
Figure S4 Prevalence of ICD-9 diagnosis of diabetes over time. Prevalence of diabetes over time was examined, using mixed-effects models adjusted for propensity score, with random intercept and slopes for each individual, allowing unstructured correlation between the random effects. Time was modeled using restricted cubic splines with 5 knots at the 5th, 27.5th, 50th, 72.5th, and 95th percentiles. Shaded areas represent 95% confidence interval bounds. Kidney International 2016 90, 164-171DOI: (10.1016/j.kint.2016.02.039) Copyright © 2016 International Society of Nephrology Terms and Conditions
Figure S5 Kaplan-Meier curves estimating time to eGFR decline ≥30% requiring 2 qualifying eGFR values. Outcome defined as having 2 follow-up outpatient eGFRs of ≥30% lower than the baseline eGFR value. Shaded areas represent 95% confidence interval bounds. Kidney International 2016 90, 164-171DOI: (10.1016/j.kint.2016.02.039) Copyright © 2016 International Society of Nephrology Terms and Conditions
Figure S6 Trajectories of eGFR stratified by hypertension status. Trajectories of eGFR for patients stratified by baseline hypertension were examined, using mixed-effects models adjusted for propensity score and baseline eGFR, with random intercept and slopes for each individual, allowing unstructured correlation between the random effects. The upper red and blue lines represent individuals without hypertension, and the lower red and blue lines represent individuals with hypertension. Time was modeled using restricted cubic splines with 5 knots at the 5th, 27.5th, 50th, 72.5th, and 95th percentiles. Shaded areas represent 95% confidence interval bounds. Kidney International 2016 90, 164-171DOI: (10.1016/j.kint.2016.02.039) Copyright © 2016 International Society of Nephrology Terms and Conditions
Figure S7 Trajectories of eGFR stratified by diabetes status. Trajectories of eGFR for patients stratified by baseline diabetes were examined, using mixed effects models adjusted for propensity score and baseline eGFR, with random intercept and slopes for each individual, allowing unstructured correlation between the random effects. The upper red and blue lines represent individuals without diabetes, and the lower red and blue lines represent individuals with diabetes. Time was modeled using restricted cubic splines with 5 knots at the 5th, 27.5th, 50th, 72.5th, and 95th percentiles. Shaded areas represent 95% confidence interval bounds. Kidney International 2016 90, 164-171DOI: (10.1016/j.kint.2016.02.039) Copyright © 2016 International Society of Nephrology Terms and Conditions