Estimating the Relative Risk in Cohort Studies and Clinical Trials of Common Outcomes

Abstract

Logistic regression yields an adjusted odds ratio that approximates the adjusted relative risk when disease incidence is rare (<10%), while adjusting for potential confounders. For more common outcomes, the odds ratio always overstates the relative risk, sometimes dramatically. The purpose of this paper is to discuss the incorrect application of a proposed method to estimate an adjusted relative risk from an adjusted odds ratio, which has quickly gained popularity in medical and public health research, and to describe alternative statistical methods for estimating an adjusted relative risk when the outcome is common. Hypothetical data are used to illustrate statistical methods with readily accessible computer software.

Incarceration and High-Risk Sex Partnerships among Men in the United States

“…55,56 We included an incarceration by illicit drug use product-interaction term to estimate associations among illicit drug users and nonusers of illicit drugs separately and to evaluate whether associations differed significantly (pG0.15 level). Adjusted models included age, race, educational attainment, poverty status, marital status, cohabitation status, and age at first sex.…”

Diabetes Confers Little to No Increased Risk of Postoperative Complications After Hip Fracture Surgery in Geriatric Patients

“…Although odds ratios often are used in cohort and crosssectional studies, it has been established that the use of relative risk is more appropriate when the outcome being tested is not rare. When this is the case, the ''rare disease assumption'' that the odds ratio approximates relative risk no longer is valid [11,20]. The use of relative risk enables a more accurate statistical comparison of risks in such cases.…”

Risk of Unintentional Overdose with Non-Prescription Acetaminophen Products

“…Generalized linear models with a Poisson distribution and log link function were used to estimate risk ratios for included covariates compared to each referent condition. 18,19 A Generalized Estimating Equation (GEE) approach was used to adjust model coefficients and standard errors for within-patient correlation as tasks were repeated using multiple medications. 20,21 Both models for overdosing and double-dipping included the potential confounding variables and risk factors of age, gender, race (black, white, other), literacy (limited (< 9 th grade or low + marginal) vs. adequate (≥ 9 th grade)), household income, acetaminophen use (non, moderate, heavy), and clinic type (academic vs. community).…”