Predicting the outcome of antiretroviral therapies for HIV-1 is a pressing clinical challenge, especially when the treatment regimen includes drugs for which limited effectiveness data is available. This scarcity of data can arise either due to the introduction of a new drug to the market or due to limited use in clinical settings. To tackle this issue, we introduce a novel joint fusion model, which combines features from a Fully Connected (FC) Neural Network and a Graph Neural Network (GNN). The FC network employs tabular data with a feature vector made up of viral mutations identified in the most recent genotypic resistance test, along with the drugs used in therapy. Conversely, the GNN leverages knowledge derived from Stanford drug-resistance mutation tables, which serve as benchmark references for deducing in-vivo treatment efficacy based on the viral genetic sequence, to build informative graphs. We evaluated these models' robustness against Out-of-Distribution drugs in the test set, with a specific focus on the GNN's role in handling such scenarios. Our comprehensive analysis demonstrates that the proposed model consistently outperforms the FC model, especially when considering Out-of-Distribution drugs. These results underscore the advantage of integrating Stanford scores in the model, thereby enhancing its generalizability and robustness, but also extending its utility in real-world applications with limited data availability. This research highlights the potential of our approach to inform antiretroviral therapy outcome prediction and contribute to more informed clinical decisions.