|Thursday, February 21|
|PS1 Poster Session 1 & Opening Mixer||
Thu, Feb 21, 6:30 PM - 8:00 PM
Statistical Methods Used in Clinical Trials to Identify Biomarkers That Can Help Fight Cardiovascular Disease (302470)
David DeMicco, Pfizer, Inc.
Sarah Young, Pfizer, Inc.
Keywords: Biomarkers, Clinical Trials, Cardiovascular Disease, Non-parametric methods
Over the past few decades, significant achievements have been made in reducing the morbidity and mortality rates due to cardiovascular disease, the number one killer in America. The foundation for such a success included statistically identifying and validating important risk factors as therapeutic targets. For example, the National Cholesterol Education Panel (NCEP) have used LDL cholesterol as the primary target of therapy for the prevention and management of cardiovascular disease. LDL cholesterol is now part of routine visits to the doctor’s office; and statins are prescribed by physicians as the most effective medicine to lower LDL cholesterol and reduce the risk for future cardiovascular events.
However, the fight against cardiovascular disease continues, since the currently recommended strategies to estimate and reduce the risk of disease cannot fully account for all future cardiovascular events. As recently as 2009, CDC estimated 600,000 thousand patients died of heart disease annually and another 130,000 thousand died of stroke. Hence other biological parameters or biomarkers, which often relate to subclinical inflammation and pathophysiological processes such as the progression of atherosclerosis, are being examined for significant impact on cardiovascular risk. Examination of these biomarkers not only helps us to gain new insights into the disease processes, but also helps us better predict and reduce the risk of future coronary heart disease (CHD) and stroke. However, since many of these biomarkers are novel and their biological functions are not very well understood, the statistical design and analysis methods used must be well understood, and with highly interpretable results.
Recent large scale clinical trials have used cardiovascular events reduction as a study outcome and have showed that pharmaceutical treatment can significantly reduce event rates of CHD or stroke. These studies have provided unique opportunities to assess biomarkers in a well-controlled clinical trial setting: to directly access the association between biomarkers and clinical outcome, between biomarkers and pharmaceutical treatment, and between biomarkers and other established cardiovascular risk factors. We will present some of these statistical methods that we have used to assess and interpret these associations, ranging from non-parametric analyses based on ranks to COX proportional hazard analyses based on log-transformed data, while taking into account the attributes of the biomarkers, the type of clinical data, and most importantly, the acceptance of people who read and use our statistical results: the clinical investigators, the regulatory authorities, the patients, and the general public.