Figure 2.

Body-mass index vs cardiac mortality at ages 35–69 years in Chennai, India, unadjusted for systolic blood pressure (A), and adjusted for systolic blood pressure (B)

Rate ratios were adjusted for major confounders (age, sex, education, socioeconomic status, smoking, and alcohol), and omitted the first 2 years of follow-up. Exclusions are as per table 1. For each body-mass index category, the area of the square is inversely proportional to the variance of the category-specific log risk, which also determines the 95% CI (represented by error bars).