Data Monitoring in Clinical Trials Using Prediction
- PDF / 115,220 Bytes
- 10 Pages / 612 x 792 pts (letter) Page_size
- 76 Downloads / 220 Views
733
Data Monitoring in Clinical Trials Using Prediction
Scott R. Evans, PhD Lingling Li, PhD LJ Wei, PhD Department of Biostatistics, Harvard School of Public Health, Boston, Massachusetts
Key Words Data monitoring; Prediction; Predicted intervals; Interim analyses; Clinical trials Correspondence Address Scott Evans, PhD, FXB-513, Harvard School of Public Health, 651 Huntington Avenue, Boston, MA 02115 (e-mail: evans@ sdac.harvard.edu).
Clinical trials (CTs) are often monitored for efficacy or futility. Several methods for interim monitoring of CTs have been developed. Although informative, few of these methods convey information regarding effect sizes (eg, treatment differences), and none use prediction to convey information regarding potential effect size estimates and associated precision, with trial continuation. We propose use of prediction
INTRODUCTION Clinical trials (CTs) are often monitored for efficacy or futility. Several methods for interim monitoring of CTs have been developed. Although informative, few of these methods convey information regarding effect sizes (eg, treatment differences) and none use prediction to convey information regarding potential effect size estimates and associated precision, with trial continuation. We propose use of prediction and specifically “predicted intervals” (PIs) as a flexible and practical tool for quantitative monitoring of CTs. In the second section, we summarize existing methods for interim data monitoring. We then introduce PIs and outline construction of PIs for binary, continuous, and time-to-event endpoints (“Predicted Intervals” section) and present examples. A discussion follows in the last section.
EXISTING METHODS FOR INTERIM D ATA M O N I T O R I N G Several methods for interim data monitoring of efficacy and futility exist. We briefly outline these methods and discuss areas where new methods can contribute additional information. METHODS FOR ASSESSING EFFICACY Sponsors of CTs often strive to get effective and safe medicinal products to market as quickly as possible, frequently employing sequential meth-
and specifically “predicted intervals” (PIs) as a flexible and practical tool for quantitative monitoring of CTs. PIs provide information regarding effect sizes, are invariant to study design, and provide flexibility in the decision-making process. We outline construction of PIs for binary, continuous, and time-to-event endpoints and present examples of their use. PIs provide a valuable tool for Data Monitoring Committees.
ods to evaluate early evidence of efficacy. Statistical tests for efficacy during interim monitoring can inflate the trial-wise α error rates unless appropriate adjustments are made. Group sequential methods [O’Brien-Fleming (1), Pocock (2), Peto et al. (3), Lan and DeMets (4), Wang and Tsiatis (5)] have been developed to identify decision boundaries that preserve the desired α error rate during CTs with interim monitoring. These methods specify how much α will be spent at each interim look and at the final analysis and require
Data Loading...
