Latent Diffusion Models (LDMs) introduce exciting opportunities in medical imaging, from disease progression prediction to interpolation to generate entire datasets of rare data. The stochastic nature of generative models makes it challenging to validate their outputs and assess their robustness across diverse datasets. BrLP, a state-of-the-art LDM for T1-weighted (T1w) images that incorporates auxiliary brain volume information, has been evaluated on Alzheimer’s Disease (AD) progression, and achieves structural similarity index (SSIM) of 0.91 ± 0.03. In this work, we conducted a pilot study of the BrLP model using the Baltimore Longitudinal Study of Aging (BLSA) dataset. Our objectives are to (1) evaluate the model performance on an external dataset using pretrained image-based and brain image-based metrics such as mean squared error (MSE), similarity index, and mean absolute error (MAE) between …