Kiyoshi Kubota1. 1. NPO Drug Safety Research Unit, Tokyo, Japan.
Abstract
INTRODUCTION: A postmarketing study without a comparator group has been recognized as a problem as it provides no measure of association. Nevertheless, the design is sometimes used in company postmarketing studies particularly when the study involves the primary data collection. In this report, the "Symmetry Analysis Cohort Design" without a comparator group but with a control period is proposed. METHODS AND RESULTS: In the proposed design, the rate ratio is estimated using the method of prescription sequence symmetry analysis with slight modification so that the rate ratio can be estimated using data on subjects who have started the drug during the study period but no data on other subjects. DISCUSSION: The proposed design has an advantage that it can provide the measure of association. Another advantage common to all self-controlled methods is that the effect of the measured and unmeasured confounders is automatically canceled out when the effect is stable over the study period. Compared with the standard design with a comparator group, the proposed design also has weaknesses. For example, adjustment of confounding by the indication may be difficult when the indication is an acute condition. In addition, the rate ratio is not valid when the probability of the prescription of the drug is dependent on the occurrence of the outcome in the unexposed (pre-dose) period. The design may be used to evaluate the need for further studies although its real usefulness is to be determined in the future.
INTRODUCTION: A postmarketing study without a comparator group has been recognized as a problem as it provides no measure of association. Nevertheless, the design is sometimes used in company postmarketing studies particularly when the study involves the primary data collection. In this report, the "Symmetry Analysis Cohort Design" without a comparator group but with a control period is proposed. METHODS AND RESULTS: In the proposed design, the rate ratio is estimated using the method of prescription sequence symmetry analysis with slight modification so that the rate ratio can be estimated using data on subjects who have started the drug during the study period but no data on other subjects. DISCUSSION: The proposed design has an advantage that it can provide the measure of association. Another advantage common to all self-controlled methods is that the effect of the measured and unmeasured confounders is automatically canceled out when the effect is stable over the study period. Compared with the standard design with a comparator group, the proposed design also has weaknesses. For example, adjustment of confounding by the indication may be difficult when the indication is an acute condition. In addition, the rate ratio is not valid when the probability of the prescription of the drug is dependent on the occurrence of the outcome in the unexposed (pre-dose) period. The design may be used to evaluate the need for further studies although its real usefulness is to be determined in the future.