Three alternative methods to resolve paradoxical associations of exposures before term