Giovanni Filardo1,2, Benjamin D Pollock1, James Edgerton3,4. 1. Department of Epidemiology, Office of the Chief Quality Officer, Baylor Scott & White Health, Dallas, TX, USA. 2. Department of Statistical Science, Southern Methodist University, Dallas, TX, USA. 3. Department of Cardiothoracic Surgery, The Heart Hospital Baylor Plano, Plano, TX, USA. 4. Texas Quality Initiative, Irving, TX, USA.
Abstract
OBJECTIVES: The high prevalence of obesity makes accurately estimating the impact of anthropometric measures on cardiac surgery outcomes critical. The Society of Thoracic Surgeons coronary artery bypass graft (CABG) surgery risk model includes body surface area (as a continuous variable, using spline functions), but most studies apply various categorizations of body mass index (BMI)-contributing to the contradictory published findings. We assessed the association between BMI (modelled as a continuous variable without assumptions of linearity) and CABG operative mortality and examined the impact of applying previous studies' BMI modelling strategies. METHODS: We identified 25 studies investigating the BMI-operative mortality association: 22 categorized BMI, 2 as a linear continuous variable,1 used spline functions. Our cohort of 12 715 consecutive patients underwent isolated CABG at 32 cardiac surgery programmes in North Texas from 1 January 2008-31 December 2012. BMI was modelled using restricted cubic spline functions in a propensity-adjusted model (controlling for Society of Thoracic Surgeons risk factors) estimating operative mortality. The analysis was repeated using each categorization identified and modelling BMI as a linear continuous variable. RESULTS: BMI (modelled with a restricted cubic spline) was significantly associated with operative mortality (P < 0.0001). Risk was lowest for BMI near 30 kg/m2 and highest below 20 kg/m2 and above 40 kg/m2. No categorization, nor the linear continuous model, fully captured this association. CONCLUSIONS: BMI is strongly associated with CABG operative mortality. Categorizing BMI (or assuming a linear relationship) heavily biases estimates of its association with post-CABG mortality. In general, smoothing techniques should be used for all continuous risk factors to avoid bias.
OBJECTIVES: The high prevalence of obesity makes accurately estimating the impact of anthropometric measures on cardiac surgery outcomes critical. The Society of Thoracic Surgeons coronary artery bypass graft (CABG) surgery risk model includes body surface area (as a continuous variable, using spline functions), but most studies apply various categorizations of body mass index (BMI)-contributing to the contradictory published findings. We assessed the association between BMI (modelled as a continuous variable without assumptions of linearity) and CABG operative mortality and examined the impact of applying previous studies' BMI modelling strategies. METHODS: We identified 25 studies investigating the BMI-operative mortality association: 22 categorized BMI, 2 as a linear continuous variable,1 used spline functions. Our cohort of 12 715 consecutive patients underwent isolated CABG at 32 cardiac surgery programmes in North Texas from 1 January 2008-31 December 2012. BMI was modelled using restricted cubic spline functions in a propensity-adjusted model (controlling for Society of Thoracic Surgeons risk factors) estimating operative mortality. The analysis was repeated using each categorization identified and modelling BMI as a linear continuous variable. RESULTS: BMI (modelled with a restricted cubic spline) was significantly associated with operative mortality (P < 0.0001). Risk was lowest for BMI near 30 kg/m2 and highest below 20 kg/m2 and above 40 kg/m2. No categorization, nor the linear continuous model, fully captured this association. CONCLUSIONS: BMI is strongly associated with CABG operative mortality. Categorizing BMI (or assuming a linear relationship) heavily biases estimates of its association with post-CABG mortality. In general, smoothing techniques should be used for all continuous risk factors to avoid bias.
Authors: Benjamin D Pollock; Jeph Herrin; Matthew R Neville; Sean C Dowdy; Pablo Moreno Franco; Nilay D Shah; Henry H Ting Journal: JAMA Netw Open Date: 2020-07-01
Authors: Giovanni Filardo; Benjamin D Pollock; Briget da Graca; Danielle M Sass; Teresa K Phan; Debbie E Montenegro; Gorav Ailawadi; Vinod H Thourani; Ralph J Damiano Journal: Mayo Clin Proc Innov Qual Outcomes Date: 2020-12-10