MOTIVATION: Spatial omics studies compare cell-cell organization across samples, but most methods model between-sample variability while treating sample-level spatial estimates as error-free. Overlooking within-sample uncertainty can distort inference in heterogeneous cohorts, motivating methods that explicitly quantify and propagate this uncertainty into cohort-level analyses. RESULTS: We present Pooled ANalysis Of VaRiance-Aware Modeling and Inference of Colocalization (PANORAMIC), a hierarchical framework for spatial colocalization analysis that uses edge-corrected neighborhood enrichment to estimate local cell-type colocalization, spatial bootstrapping to quantify within-sample uncertainty, and multilevel random-effects meta-analysis to propagate this uncertainty across samples, patients, and conditions. In simulations, PANORAMIC improved recovery of within-sample uncertainty and between-sample heterogeneity relative to naive estimators across diverse spatial settings and progressive data degradation. Applied to a colorectal cancer tissue microarray profiled by multiplexed immunofluorescence imaging, PANORAMIC identified stronger B- and T-cell colocalization in tumors with Crohn's-like reaction than in tumors with diffuse inflammatory infiltration, together with tighter higher-order immune organization consistent with immune aggregates. These findings were missed using standard methods, showing that propagating within-sample spatial uncertainty can improve cohort-level inference in spatial omics studies. AVAILABILITY AND IMPLEMENTATION: PANORAMIC is released as an open-source R package at https://github.com/plevritis-lab/panoramic and archived on Zenodo at https://doi.org/10.5281/zenodo.19927197.