Pipelines can be exposed to a wide variety of loads, depending on the environments and the area of application. These loads may impose large longitudinal plastic strain on pipelines, which could constitute a significant threat to the structural capacity of the pipeline. Reliable calibration of the strain capacity of pipelines plays an important role in the strain-based design (SBD) method. In this paper, a tensile strain capacity (TSC) predictive model (an equation) for welded X42 vintage pipes has been developed by conducting nonlinear parametric analysis followed by nonlinear regression analysis. First, our previously validated extended finite element method (XFEM) model was used to demonstrate the applicability of the XFEM in simulating full-scale ductile fracture response of pipelines subjected to biaxial loading, using pressurized American Petroleum Institute (API) 5L X42 vintage pipes subjected to four-point bending. Second, a parametric study investigating the effects of pipe and defect geometries as well as loading on the pipe TSC is presented. The nonlinear parameterization using XFEM was conducted in abaqus/standard. The TSC trends obtained for the various parameters considered were examined to derive appropriate individual variable functions for each parameter while taking any significant interactions between the parameters into consideration. Also, a nonlinear regression analysis is employed to develop a nonlinear semi-empirical model for predicting the TSC. The results obtained from the developed TSC predictive model (TSCvin.) was compared with those evaluated using the validated XFEM models. The results showed good agreement. Finally, statistical analysis was conducted to ensure the model is unbiased and predicts conservative TSCs by modifying the model using probabilistic error analysis. The modified model is capable of increasing the confidence level in the predicted TSC hence becoming a practical tool for reliable prediction of TSC of X42 vintage pipes needed for conducting pipeline integrity assessment. This modified predictive model is useful in practical applications because it provides a quantifiable degree of conservatism and reliability to the predicted TSCs.