Application of Survival Analysis
Techniques to Evaluation of Factors
Affecting Compliance in a Clinical Trial
of Hypertension Control
E. O. Smith, R. J. Hardy, G. R. Cutter,
J. D. Curb, and C. M. Hawkins
From the School of Public Health, The University of Texas, Houston, Texas
ABSTRACT: Survival analysis techniques including follow-up life tables and regression
methods that utilize concomitant information are shown to be applicable to the
analysis of medical follow-up study dropout data. These techniques have been
applied to data from the Hypertension Detection and Follow-up Program in order to
gain knowledge about factors associated with a study participant's tendency to
either remain active in the Program or drop out.
INTRODUCTION
Increasing attention has been focused on patient compliance with therapeu-
tic regimens in recent years. The contribution of patient compliance to the
success or failure of therapeutic regimens and public health programs has
been shown to be surprisingly large. In addition, patient compliance is
important to the design, analysis, and interpretation of clinical trials.
This type of problem is, however, dealt with in many elementary statistics
text under the heading of missing data. These methods are useful in small
studies where the number of missing data are very small. Feinstein [1]
addresses the issue of compliance and gives an analytical review of the
subject. He points out the impact of noncompliance or partial compliance
on data analysis as well as some of the pitfalls involved in interpreting data
from a selected compliant population. Interest in the problem has increased
in recent years with the advent of large multicenter trials, often depending
on behavioral change. The potential for bias and inference problems due to
noncompliance in these essentially nonrepeatable studies with their enor-
mous impact on the scientific community and their enormous cost is not
small. Although the cost of attaining the cooperation of those "reluctant
Address requests for reprints to Dr. E. O'Brian Smith, Beta Blocker Heart Attack Trial, Coordinating
Center, The University of Texas, School of Public Health, Houston, TX 77030.
Received February 6, 1979; revised and accepted October 3, 1979.
ControlledClinical Trials1, 59-69 (1980) 59
© Elsevier NorthHolland,Inc.,1980 0197-2456/80/01005911502.25