Background/Objectives: Thyroid cancer spans indolent differentiated tumors to lethal poorly differentiated and anaplastic carcinomas, and histopathology and molecular testing do not capture the spatial organization of tumor, stromal, and immune compartments that underlies invasion, dedifferentiation, and treatment resistance. Spatial transcriptomics profiles gene expression in situ. We systematically reviewed its platforms, applications, and findings in thyroid cancer. Methods: PubMed/MEDLINE, Scopus, and Web of Science were searched, following PRISMA 2020 guidance and supplemented by citation chasing, for English-language primary studies applying spatially resolved transcriptomics to thyroid cancer tissue. Subtype, platform, complementary methods, and principal findings were extracted, and each study was appraised item by item with the design-matched Joanna Briggs Institute (JBI) checklist. Results: Of 104 records, 21 studies were included: 12 case series, four analytical cross-sectional studies, three single-case reports, and two reanalyses of one public dataset. Nineteen used capture-based whole-transcriptome platforms, predominantly 10x Visium, and two used GeoMx digital spatial profiling; papillary thyroid carcinoma predominated, and medullary carcinoma was not examined. Tumor-microenvironment remodeling accompanying dedifferentiation toward anaplastic carcinoma was replicated in five independent cohorts. Cancer-associated fibroblast activation was reported in three cohorts, although the SERPINE1-positive and POSTN-positive programs were each single-cohort; APOE-linked tumor states appeared in two cohorts with opposite directionality; and invasive-front extracellular-matrix dysregulation in follicular tumors was reported in two. Six studies reported functional perturbation and 12 protein-level validation; no biomarker had independent external validation. Methodological concern was low in four studies, moderate in seven, and high in 10: only one of 12 case series reported consecutive and complete inclusion, eight reported no outcomes for the profiled patients, and nine analyzed spots or cells rather than patients. Conclusions: Spatial transcriptomics has resolved reproducible microenvironmental remodeling during dedifferentiation, but its prognostic and mechanistic claims rest on small, selected cohorts without independent validation. Larger, consecutively enrolled, clinically annotated studies are needed.

Background/Objectives: Among endocrine malignancies, thyroid cancer is the most prevalent and encompasses a broad clinical spectrum, ranging from indolent differentiated tumors to highly aggressive poorly differentiated and anaplastic thyroid carcinomas. Histopathology and molecular testing incompletely capture the spatial organization of tumor, stromal, and immune compartments that drives invasion, dedifferentiation, and treatment resistance. Spatial transcriptomics preserves tissue architecture while profiling gene expression, mapping molecular programs onto histologic context. We systematically reviewed its platforms, applications, and biological insights in thyroid cancer. Methods: PubMed/MEDLINE, Scopus, and Web of Science Core Collection were searched for English-language studies combining spatial-transcriptomic and thyroid-cancer concepts, following PRISMA 2020 guidance. Primary studies that applied a spatially resolved transcriptomic method to thyroid cancer tissue and reported spatial gene-expression findings were included; data were extracted on subtype, platform, complementary methods, and principal findings. Results: Of 104 records (47 after de-duplication), 21 studies met the inclusion criteria. Most used capture-based whole-transcriptome profiling, particularly 10x Visium, while GeoMx digital spatial profiling supported FFPE-compatible biomarker studies. Recurrent findings included POSTN+ and SERPINE1+ cancer-associated fibroblast programs at invasive fronts, APOE-linked immunosuppressive states, organized immune-evasion niches, and progressive tumor-microenvironment remodeling accompanying dedifferentiation toward ATC. Spatial transcriptomics also clarified invasion, lymph node and distant metastasis, and cell–cell communication. Conclusions: Spatial transcriptomics is opening new avenues for dissecting thyroid cancer progression, but current evidence is limited by small cohorts, platform concentration, and variable validation. Larger, standardized, clinically annotated studies are needed to define reproducible spatial biomarkers.

Purpose: We introduce the Temporal Index of Divergent Evidence (TIDE), a metric that evaluates how the topic center of gravity of a corresponding corpus of literature aligns with, or shifts relative to, a recommendation’s semantics over time. Methods: TIDE computes annual mean cosine similarity using MedCPT embedding vectors between a recommendation and corresponding PubMed records binned by year (2000-2024), testing for monotonic temporal trend (Kendall τ). Three recommendations with expected divergent, stable, and convergent trajectories were examined: two-stage exchange for periprosthetic joint infection (PJI), emergent fasciotomy for acute compartment syndrome (ACS), and non-operative management of distal radius fractures (DRF) in adults ≥ 65 years. The signal was characterized by an additive per-token attribution, an isolated vs. in-context analysis, and an adversarial minimal-pair test. Results: PJI diverged (τ = −0.84, year-shuffle p = 0.001), DRF converged (τ = +0.42, p = 0.005), and ACS showed no detectable trend (τ = −0.07, p = 0.66). The two significant directions (PJI, DRF) held under matched input [CLS] pooling, while ACS remained near zero. Publication volume baselines did not distinguish the three trajectories (publication count was universally positive, τ = +0.77 to +0.97). A reversed-meaning minimal-pair control showed little cosine sensitivity to reversal. Paper-level TF-IDF recovered the qualitative three-case pattern. Conclusion: Embedding-based drift analysis is a proof-of-concept candidate pre-triage signal that tracks topical alignment, not evidentiary direction; superiority over TF-IDF remains unproven. The present study is hypothesis-generating, showing potential use of embedding representations as an analytical unit rather than simply for retrieval.