Risk stratification in upper gastrointestinal bleeding; prediction, prevention and prognosis

Abstract

In the first part of this thesis we developed a novel prediction score for predicting upper gastrointestinal (GI) bleeding in both NSAID and low-dose aspirin users. Both for NSAIDs and low-dose aspirin use risk scores were developed by identifying the five most dominant predictors. The risk of upper GI bleeding increased with a higher risk score in both groups. The predictive power of the scores was only moderate, but sensitivity and specificity were higher compared to scores recommended by international guidelines. It is now recommended to add gastroprotection to NSAID/ low-dose aspirin users at increased risk. To assess the cost-effectiveness of this recommendation we performed two cost-utility analyses. NSAID+PPI co-therapy was found to be the most cost-effective strategy in all chronic arthritis patients irrespective of their risk for GI complications. For patients with increased GI risk, the NSAID/PPI single tablet formulation was also a cost-effective option. In the low-dose aspirin model we found that PPI co-therapy was cost-effective in all patients taking low dose aspirin for primary and secondary prevention of acute coronary syndrome. In secondary prevention, a single tablet formulation of low-dose aspirin and PPI was another cost-effective option in patients with increased risk for upper GI bleeding or in those with moderate PPI compliance. In the second part of this thesis we identified all published prediction scores for the outcome of upper GI bleeding. Substantial heterogeneity in endpoints and results was seen in the 16 identified scores. Moreover, the methodological quality was suboptimal in most studies. We suggested that clinicians should use the “best available” scores identified. The recommendation to use these scores was further underlined in the study in which we assessed how the clinical intuition (or gut feeling) of experienced gastroenterologists performs in predicting the prognosis of patients with gastrointestinal bleeding. We found that the recommended prediction scores had a higher predictive power than the gut feeling. As the various prediction scores were mostly limited to inclusion of clinical or patient-related factors to predict the outcome of upper GI bleeding, we hypothesized that hospital- or procedure-related factors may also play an important role in the outcomes after a bleeding. In a large multicenter cohort study we found that patients admitted to the Emergency Unit during the weekend were at increased risk of mortality compared to those admitted during the week. Time of admission was not associated with an adverse outcome. An often used predictor for the outcome after upper GI bleeding due to peptic ulcers is the Forrest classification, in which ulcers are categorized based on the presence of endoscopically visible bleeding stigmata. However, this classification was only designed to classify ulcers. We found that the Forrest classification was still a good univariate predictor for rebleeding. Additionally, we concluded that the Forrest classification could be simplified. In the last chapter we combined all knowledge that is currently available so far and developed a novel prediction score, the RASTA score, to predict the need for a clinical intervention. We also included the gut feeling of experienced gastroenterologists.