Background. Stroke volume index (SVI) is an important prognostic parameter in pulmonary arterial hypertension (PAH). The direct Fick (DF) method represents the gold standard for measuring it. Indirect Fick (IF) and thermodilution (TD) are simpler and widely used alternatives. However, data on the accuracy of these methods in estimating SVI in PAH are scant. We aimed to compare these different invasive methods, and in a subgroup of patients, to a non-invasive method using MRI. Methods. We enrolled 103 PAH patients undergoing a diagnostic or follow-up right heart catheterization at our centre (mean age 56 years, 56% female). The Bland–Altman analysis was used to assess agreement between methods. Potential demographic, clinical, and hemodynamic biases were explored. The accuracy of cardiac magnetic resonance (CMR)-derived SVI was assessed in a subset of patients. Results. The mean bias for IF-SVI vs. DF-SVI was −5.53 mL/min/m2 with a median percentage error (PE) of 15%. The mean bias was lower, 0.09 mL/min/m2, for TD-SVI vs. DF-SVI with a median PE of 10%. Low cardiac index and severe tricuspid regurgitation (TR) were associated with a greater bias between TD and DF. CMR-SVI showed good accuracy and precision even in patients with severe TR, compared to DF. Conclusions. The indirect Fick is the less reliable method to assess SVI also in PAH patients. Thermodilution is a valid alternative to direct Fick, but it should be used with caution in patients with severe TR or low cardiac index. SVI measured by cardiac MRI is a promising non-invasive alternative, especially in patients with severe TR. Our observation needs to be confirmed by other series and larger studies.