Is the superiority of apoB over non-HDL-C as a marker of cardiovascular risk in the INTERHEART study due to confounding by related variables?