Abstract Background: Transcriptional profiling of pancreatic cancers (PC) has defined classical and basal subtypes; basal tumors have worse outcomes. Initial data suggest worse outcomes to FOLFIRINOX (FFX) compared with gemcitabine nab-paclitaxel (GnP) in basal tumors. In addition, it is hypothesized that classical to basal conversion may lead to chemotherapy resistance, although clinical evidence for subtype conversion and subtype specific chemotherapy sensitivity are sparse. Here, we examined the temporal stability and clinical implications of PC transcriptional subtypes in a large, real-world dataset. Methods: We examined patients with metastatic PC, who had outcomes data and RNA sequencing (n=5,335) performed at Caris Life Sciences (Phoenix, AZ). Classical and basal subtypes were identified along a six subtype continuum from strong classical (SC) to strong basal (SB) on either pole using PurIST algorithm. Overall survival (OS) was defined from date of biopsy of metastatic site to date of death or last known contact and estimated using Kaplan-Meier method. Survival was compared between groups using log-rank test and cox proportional hazards regression. Predictors of subtype discordance were assessed in patients who underwent RNA-sequencing of two tumor biopsies using multivariable logistic regression. Results: In patients with metastatic PC, SB tumors (n = 1,579) had significantly worse OS compared to SC tumors (n = 2,118), with a median OS of 5.7 vs. 10.0 months, respectively (hazard ratio [HR] 1.49, 95% confidence interval [CI], 1.39-1.60, p = 2.77e-27) independent of impact of age, gender and mutations in KRAS, TP53, CDKN2A or SMAD4 in cox proportional hazard models. OS was worse in patients with SB tumors who received either 1L GnP [SB (n = 112) vs. SC (n = 178), median OS: 6.0 vs. 10.4 months (HR 1.35 [95% CI, 1.05-1.75], p = 0.02)], or 1L FFX [SB (n = 155) vs. SC (n = 215), median OS: 8.9 vs. 12.5 months (HR 1.35 (95% CI, 1.06-1.72), p = 0.013)]. The prognostic impact of transcriptional subtype within different KRAS-mutant alleles is not well defined. OS was short and not significantly different between the three most frequent KRAS alleles (G12 V/D/R) in SB tumors (median OS, KRAS G12R/V/D: 5.4 vs. 5.4 vs. 5.5 months). In SC tumors, KRAS G12D was associated with a significantly worse OS compared to KRAS G12R (median OS: 8.6 vs. 11.9 months, HR 1.23 (95% CI, 1.04-1.47), p = 0.018). KRAS G12V had a numerically worse survival compared with KRAS G12R in SC tumors (median OS: 9.4 vs. 11.9 months, HR 1.15 (95% CI, 0.96-1.37), p = 0.12). Sixty of 77 cases (78%) maintained their transcriptional subtype between two temporally &/or spatially disparate lesions. The likelihood of transcriptional subtype change was independent of the duration between (p = 0.79) or site of (p = 0.37) the two biopsies. Conclusions: SB subtype is a strong independent predictor of worse outcomes, irrespective of mutant KRAS allele or upfront chemotherapy regimen, and tends to remain stable between biopsies in individual patients Citation Format: Harshabad Singh, Joanne Xiu, Kevin Kapner, Chen Yuan, Raja R Narayan, Matthew Oberley, Alex Farrell, Rishi Surana, Brandon M Huffmann, Kimberly J Perez, James M Cleary, Alexander C Jordan, Andressa Dias Costa, Hannah L Williams, Srivatsan Raghavan, Benjamin Weinberg, Michael J Pishvaian, Rachna T Shroff, Sanjay Goel, Stephanie K Dougan, Jonathan Nowak, David Spetzler, George Sledge, Brian M Wolpin, Andrew J Aguirre. Temporal stability and chemotherapy responsiveness of classical and basal transcriptional subtypes of pancreatic cancer [abstract]. In: Proceedings of the AACR Special Conference in Cancer Research: Advances in Pancreatic Cancer Research; 2024 Sep 15-18; Boston, MA. Philadelphia (PA): AACR; Cancer Res 2024;84(17 Suppl_2):Abstract nr PR-21.