Cost-effectiveness thresholds selected a posteriori to analysis?