Modern statistical methods for handling missing repeated measurements in obesity trial data: beyond LOCF