Can we well assess the relative efficacy and tolerability of a new drug versus others at the time of marketing authorization using mixed treatment comparisons? A detailed illustration with escitalopram