The 131st Annual Meeting (November 15-19, 2003) of APHA |
Brenda Gillespie, PhD, Center for Statistical Consultation and Research, University of Michigan, 3550 Rackham Building, Ann Arbor, MI 48109-1070, 734 647 4609, bgillesp@umich.edu
In some settings repeated measurements on a person are summarized into a single value for each subject, such as the time that a particular threshold is crossed. In an ongoing study of visual field (VF) loss in glaucoma patients, data are collected at 6-month intervals. A patient is considered "failed" when a 3-point VF score increase over baseline is observed. In a different study of HPV infection in women, monthly samples to detect HPV are collected. For those HPV positive at baseline, the endpoint is time to clearance, defined as two consecutive HPV negative samples. The advantage of the "time to event" definition is the simplicity of having a single value to represent a patient's outcome. There are three major disadvantages: (1) loss of the ability to characterize patterns of values over time; (2) loss of power resulting from the data reduction; and (3) error introduced in estimating the "crossing time" due to errors or other problems in the original values. For example, in the glaucoma study cataract can worsen the VF, sometimes leading to a definition of "failure"; the VF measure may recover after cataract extraction, but data following "failure" are not considered. A woman may be cleared of HPV one month, but may become newly infected the following month, so the clearance would not be captured. The potential difficulty arising from such time-to-event definitions will be statistically assessed. Guidelines will be given on when such data reduction is and is not wise.
Learning Objectives:
Keywords: Clinical Trails, Statistics
Presenting author's disclosure statement:
I do not have any significant financial interest/arrangement or affiliation with any organization/institution whose products or services are being discussed in this session.