G E Zellner | PubMed Health

PubMed2001

A SAS macro for sample size re-estimation.

The assessment of sample size in clinical trials comparing means requires a variance estimate of the main efficacy variable. If no reliable information about the variance of the key response is available at the beginning of a clinical trial, the use of data from the first 'few' patients entered in the trial ('internal pilot') may be appropriate to estimate the variance and thus to recalculate the required sample size. A SAS macro that implements the EM algorithm for carrying out and simulating such interim power evaluations without unblinding the treatment status is presented.

Algorithms↗

PubMed1998

Dependence of vancomycin clearance on renal function via regression and bootstrap methods.

BACKGROUND: Frequently, the estimation of vancomycin on the basis of renal function is too rough because the unknown parameters of a regression function between the vancomycin clearance (CL) and the creatinine clearance (ClCR) are based on small sample sizes. OBJECTIVE: In this study we aim to compare linear and nonlinear regression, spline interpolation and nonlinear kernel estimation for defining the relationship between measured Cl and ClCR. METHOD: We used data from published papers and appropriate numerical methods. The variability and accuracy of the estimated regression functions were determined from bootstrap methods and kernel density estimators. Tests to prove the usually assumed linearity of the regression were carried out and the influence of patient age and weight on Cl was determined. RESULTS: A linear relationship reported by several authors earlier has been determined as ClVAN = 0.763 ClCR + 2.715, (ml/min) (Cl = 0.011 ClCR + 0.055, (ml/min/ kg)). CONCLUSION: Nonparametric regression analysis shows that a nonlinear approximately parabolic function could fit the relationship between Cl and ClCR in the present case somewhat better than a linear function.

Anti-Bacterial Agents↗