Biostatistics journal club: Adjusting for Treatment Effects in Studies of Quantitative Traits – Anti-Hypertensive Therapy and Systolic Blood Pressure.
Discussion of the above paper led by Bernard Rosner, PhD, Harvard Medical School, Harvard T.H. Chan School of Public Health, Brigham and Women’s Hospital, on its proposal to use censored regression methods to address the issue of data analysis in the use of anti-hypertensive medication (MEDS) in adults and the important biases that can occur in commonly used methods of accounting for use of MEDS. Registration is required. This event will take place via Zoom. Register here.
Bernard Rosner, PhD
Professor of Medicine (Biostatistics), Harvard Medical School; Professor in the Department of Biostatistics, Harvard T.H. Chan School of Public Health; Senior Biostatistician, Channing Division of Network Medicine, Brigham and Women’s Hospital
Adjusting for Treatment Effects in Studies of Quantitative Traits: Anti-hypertensive Therapy and Systolic Blood Pressure
A common problem in observational studies is how to adjust for medication use, when use of the medication may affect the main outcome variable. Specifically, in studies where systolic blood pressure (SBP) is the outcome, one has to account for use of anti-hypertensive medications (MEDS). MEDS are commonly prescribed in adults, especially with increasing age. The purpose of this paper is to assess different methods of accounting for use of MEDS including: (a) fitting a conventional model with treatment as a covariate, (b) ignoring the problem and treating the observed SBP as if it were the true “underlying” SBP, (c) excluding subjects on MEDS from the analysis, (d) adding a fixed constant to SBP for subjects on MEDS, or (e) using censored regression methods for subjects who use MEDS to estimate their underlying SBP. Similar issues arise when BP level is used as an exposure in CVD risk models. We also will discuss the related issue of accounting for use of MEDS while developing norms for SBP in adults.
Use of anti-hypertensive medication (MEDS) is common in adults. However, this complicates the analysis of data where quantitative levels of blood pressure (BP) is either the outcome, or is an exposure in CVD risk models. Important biases can occur in commonly used methods of accounting for use of MEDS. This paper proposes using censored regression methods to address the above issue. We also discuss accounting for medication use when developing norms for BP in adults.
Tobin, M.J., Sheehan, N.A., Scurrah, K.J., and Burton, P.R. Adjusting for treatment effects in studies of quantitative traits: antihypertensive therapy and systolic blood pressure. Statistics in Medicine 2005; 24: 2911–2935.