Abstract Background: A growing proportion of women diagnosed with atypical hyperplasia (AH) and early-stage endometrioid endometrial cancer (EEC) require nonsurgical treatment options. Our previously published phase II clinical trial of the levonorgestrel intrauterine device (LIUD) for AH and grade 1 EEC (G1EEC) demonstrated 83% response rate at 1 year; however, prospective long-term follow-up data and determinants for duration of response to LIUD are limited. Methods: Follow-up data from patients that participated in our prospective phase II trial of LIUD as a nonsurgical approach for AH and G1EEC were collected from medical records. Duration of response and time to relapse were evaluated. Spatial transcriptomic profiling from longitudinal biopsy samples from pre-treatment (n=5), on-treatment (n=5), and disease relapse lesions (n=3) were evaluated using digital spatial profiling (DSP, Nanostring GeoMX Whole Transcriptome Atlas). In silico cell type deconvolution was performed and pathway analyses was conducted using gene set variation analysis enrichment. Results: Of 43 participants exhibiting initial response to LIUD, 41 had follow-up data. Sixteen experienced disease relapse (39%) with a median time to relapse of 17 months after initial response. Clinical factors associated with shorter duration of response included age (HR 0.95, p=0.021), initial diagnosis of G1EEC (HR 3.51, p=0.014), premenopausal status (HR 3.85, p=0.039), and Hispanic ethnicity (HR 3.45, p=0.017). Pre-treatment IGF1 and IGF2 expression levels were positive predictors of duration of response (HR 0.44 and 0.45, respectively, p<0.05). DSP results revealed that LIUD increased NK cells (log2FC=46.13, FDR=0.004), and lymphocyte cytotoxicity signaling markers GZMB (log2FC=1.06, padj=0.023) and PRF1 (log2FC=1.62, padj=0.004), but NK cells were reduced in relapse lesions (log2FC=-55.96, FDR=0.02) as was PRF1 expression (log2FC=-1.87, padj=0.038). Multiple immune-related pathways (IFNα and IFNγ response, TGFβ signaling) were significantly upregulated in relapse lesions compared to matched pre-treatment lesions (FDR<0.05). IDO1 expression, a marker for immune exhaustion, was upregulated in relapse lesions compared to matched pre-treatment and on-treatment lesions (log2FC=2.56, FDR=0.01). Conclusions: While LIUD for treatment of AH or G1EEC has excellent initial response rates, upfront resistance, and relapse after initial response to LIUD impacts nearly half of patients. Our study shows progestin-related immunomodulatory effects, including increased NK cell infiltration and lymphocyte cytotoxicity. Immune mechanisms for relapse are implicated here, including reversal of progestin-related immunomodulatory effects (decreased NK cell infiltration and lymphocyte cytotoxicity) and increased immune exhaustion (prolonged IFNα and IFNγ response and increased IDO1 expression in relapse lesions). Thus, relieving immune exhaustion may increase durability of response to LIUD for treatment of AH or G1EEC. Citation Format: Mikayla B. Bowen, Brenda Melendez, Qian Zhang, Richard Yang, Bryan Fellman, Barrett Lawson, Naomi Adjei, Joseph Celestino, Tri Nguyen, Khalida Wani, Bhavana Singh, Christof Straub, Liang Zhang, Cheryl Tan, Felicia New, Diana Urbauer, Alexander Lazar, Karen Lu, Jennifer Wargo, Shannon N. Westin, Melinda S. Yates. Immune exhaustion and reversal of progestin-related immune modulation in adaptive resistance to levonorgestrel intrauterine device for treatment of atypical hyperplasia and early endometrial cancer [abstract]. In: Proceedings of the AACR Special Conference on Endometrial Cancer: Transforming Care through Science; 2023 Nov 16-18; Boston, Massachusetts. Philadelphia (PA): AACR; Clin Cancer Res 2024;30(5_Suppl):Abstract nr B024.