The value of the case—control approach for the evaluation of the risk of upper gastrointestinal bleeding associated with the previous use of non-steroidal anti-inflammatory drugs