On the assessment of adverse drug reactions from spontaneous reporting systems: The influence of under-reporting on odds ratios