Current therapies for treating pancreatic ductal adenocarcinoma (PDAC) are largely ineffective, with the desmoplastic environment established within these tumors being considered a central issue. We established a 3D spheroid co-culture in vitro model using a PDAC cell line (either PANC-1 or Capan-2), combined with stellate cells freshly isolated from pancreatic tumors (PSC) or hepatic lesions (HSC), and human type I collagen to analyze the efficiency of the chemotherapeutic gemcitabine (GEM) as well as two novel drug candidates derived from natural products: pseudopterosin (PsA-D) and O-methyltylophorinidine (TYLO). Traditional 2D in vitro testing of these agents for cytotoxicity on PANC-1 demonstrated IC50 values of 4.6 (±0.47) nM, 34.02 (±1.35) µM, and 1.99 (±0.13) µM for Tylo, PsA-D, and GEM, respectively; these values were comparable for Capan-2: 5.58 (±1.74) nM, 33.94 (±1.02) µM, and 0.41 (±0.06) µM for Tylo, PsA-D, and GEM, respectively. Importantly, by assessing the extent of viable cells within 3D co-culture spheroids of PANC-1 with PSC or HSC, we could demonstrate a significant lack of efficacy for GEM, while TYLO remained active and PsA-D showed slightly reduced efficacy: GEM in PANC-1/PSC (IC50 = >100 µM) or PANC-1/HSC (IC50 = >100 µM) spheroids, TYLO in PANC-1/PSC (IC50 = 3.57 ± 1.30 nM) or PANC-1/HSC (IC50 = 6.39 ± 2.28 nM) spheroids, and to PsA-D in PANC-1/PSC (IC50 = 54.42 ± 12.79 µM) or PANC-1/HSC (IC50 = 51.75 ± 0.60 µM). Microscopic 3D rendering supported these cytotoxicity outcomes, showing little or no morphological spheroid structure change during this period of rapid cell death. Our results support the use of this 3D spheroid co-culture in vitro model having a desmoplastic microenvironment for the identification of possible novel chemotherapeutic drug candidates for PDAC, such as TYLO and PsA-D.