Examining the BMI-mortality relationship using fractional polynomials

Table 2 Logistic regression coefficient estimates and standard errors in parentheses for the final adjustment model including smoking status, fractionally transformed BMI¹ and age² and interactions identified as significant

¹ For the male sample, $B M I (p 1) = {(\frac{B M I}{10})}^{- 2}$ - 0.137, $B M I (p 2) = {(\frac{B M I}{10})}^{- 2} * ln (\frac{B M I}{10})$ - 0.376 and for the female sample, $B M I (p 1) = {(\frac{B M I}{10})}^{- 2}$ - 0.142, $B M I (p 2) = {(\frac{B M I}{10})}^{- 1}$ - 0.376.
² For the male sample, $Age (q 1) = {(\frac{Age}{10})}^{2}$ - 20.692 and for the female sample, $Age (q 1) = {(\frac{Age}{10})}^{2}$ - 22.216.

ISSN: 1471-2288