PURPOSE: The goals of this study were to compare the Hoffstein and Loredo predictive formulas for CPAP titration and to design our own formula. The Health Index (HI) a statistical equation that we developed adjusts BMI for gender, age and predicted lean body mass was included in our predictive model.
METHODS: This was a retrospective data collection of all patients with polysomnography/in-lab CPAP titration from 1997-2009.
RESULTS: This is interim data: n=310 patients. Hoffstein’s equation is better correlated to actual CPAP than Loredo’s equation (Correlation coefficient r: 0.49 v. 0.36, p < 0.001), although Loredo’s equation is more accurate. An independent t-test comparison of Loredo or Hoffstein models to actual CPAP values shows that they both underestimate required pressures and their distribution of values is very different than the actual. A multiple regression analysis of the variables produces a more accurate and precise equation than either equation. Using the Health Index, rather than the BMI, in the regression model improved the predictive power by approximately 4%. This is due to greater correlation between the HI and CPAP than BMI with CPAP (0.26 v. 0.20, p < 0.05). We call this equation the Cleveland Clinic Predictor (CCP): CPAP Predicted = 5.55 + 0.05327 (Health Index) + 0.03276 (Neck Circumference) + 0.03422 (AHI Crude) + 0.0005568 (AHI Supine) + 0.001110 (AHI REM) + 0.01301 (RDI) The CCP is very similar to actual CPAP values. The CPP predicts 30% of the variability while being accurate 74% of the time. Assuming +/- 2 points in CPAP estimation as a margin of error, CCP accurately predicts CPAP in 74% of patients (100/135), compared to 54% (105/195) using Loredo and 38% (72/189) using Hoffstein. These differences are significant.
CONCLUSIONS: Both the Hoffstein and Loredo predictive equations underestimate CPAP titration pressures. Our CCP formula is more accurate, but further data must be gathered.
CLINICAL IMPLICATIONS: Our formula may obviate the need for overnight CPAP titration.
DISCLOSURE: The following authors have nothing to disclose: Ryu Tofts, Jordan Dozier, Jonothan Daco, Timur Urakov, Marlow Hernandez, Ndubuisi Okafor, Franck Rahaghi, Gustavo Ferrer, Eduardo Oliveira, Laurence Smolley, Jose Ramirez, Anas Hadeh
No Product/Research Disclosure Information