The purpose of this study was to create a predicting tool for UGIB event in NSAID users. The patients of this case-control study were NSAID users who had received NSAIDs for at least 3 days and were gastroscoped. The patients with a history of gastrointestinal varices, gastrointestinal cancer, chronic renal failure, coagulopathy, or Mallory-Weiss tear were excluded. The data was collected between July 2001 and January 2002 by patient interviewing and medical record reviewing. One hundred and fifty four NSAID users were identified (89 in the UGIB group, 65 in the non-bleeding group).
Most patients were elderly (mean age + SD: 60.9 + 12.6 years). Age and the number of current NSAID users were significantly higher in UGIB patients than in non-bleeding patients (p < 0.05 and p < 0.01, respectively). The number of antiulceration drug users in non-bleeding patients was higher than in UGIB patients (p < 0.01).
An equation for prediction of UGIB probability in NSAID users was generated by using enter logistic regression. The best model of predicting the risk of UGIB event in NSAID users was logit (UGIB) = 0.33 + 2.09 Multiple NSAID use + 1.43 H. pylori infection + 0.34 Current NSAID use + 0.12 (Age x Sex) - 8.53
Sex - 2.41 Antiulceration drugs - 0.000048 Age. The model had 80.2% of the overall rate of correct classification. The positive and negative predictive values were 80.8% and 78.9% respectively. The probability of UGIB = elogit(UGIB)/1 + elogit(UGIB). If the value of the probability of UGIB is more than 0.5, the patient has a high risk of UGIB.
Multiple NSAID use is the strongest factor that affects the probability of UGIB in NSAID users. H. pylori infection is another strong risk factor of NSAID-related UGIB. Antiulceration drug usage reduced the risk of UGIB in this group of patients. The developed model can be used as a guide for pharmacotherapeutic planning in clinical practices.

Keywords: Nonsteroidal anti-inflammatory drugs, Upper gastrointestinal bleeding, Prediction