This paper introduces a novel approach to clinical trial outcome prediction by integrating uncertainty quantification through selective classification with the Hierarchical Interaction Network (HINT). The primary goal is to enhance the model's ability to discern nuanced differences, thereby significantly improving its overall performance. Selective classification allows the model to withhold decision-making in cases of ambiguity or low confidence, thereby improving prediction accuracy for instances it chooses to classify. Comprehensive experiments demonstrate that incorporating selective classification into clinical trial predictions significantly enhances model performance, as evidenced by improvements in key metrics such as PR-AUC, F1, ROC-AUC, and overall accuracy. Specifically, the proposed method achieved 32.37%, 21.43%, and 13.27% relative improvements in PR-AUC over the base model (HINT) in phase I, II, and III trial outcome prediction, respectively. When predicting phase III, the method reached a PR-AUC score of 0.9022. These results highlight the robustness and potential utility of this strategy in clinical trial predictions, potentially setting a new benchmark in the field. The study also emphasizes the importance of integrating uncertainty quantification into predictive models, particularly in medical contexts where reliable predictions directly impact patient health and research outcomes. The methodology is applied to clinical trial outcome prediction, demonstrating its potential impact in this critical area of medical research.This paper introduces a novel approach to clinical trial outcome prediction by integrating uncertainty quantification through selective classification with the Hierarchical Interaction Network (HINT). The primary goal is to enhance the model's ability to discern nuanced differences, thereby significantly improving its overall performance. Selective classification allows the model to withhold decision-making in cases of ambiguity or low confidence, thereby improving prediction accuracy for instances it chooses to classify. Comprehensive experiments demonstrate that incorporating selective classification into clinical trial predictions significantly enhances model performance, as evidenced by improvements in key metrics such as PR-AUC, F1, ROC-AUC, and overall accuracy. Specifically, the proposed method achieved 32.37%, 21.43%, and 13.27% relative improvements in PR-AUC over the base model (HINT) in phase I, II, and III trial outcome prediction, respectively. When predicting phase III, the method reached a PR-AUC score of 0.9022. These results highlight the robustness and potential utility of this strategy in clinical trial predictions, potentially setting a new benchmark in the field. The study also emphasizes the importance of integrating uncertainty quantification into predictive models, particularly in medical contexts where reliable predictions directly impact patient health and research outcomes. The methodology is applied to clinical trial outcome prediction, demonstrating its potential impact in this critical area of medical research.