The purpose of our study was to evaluate detection of central pulmonary embolism (PE) on non-contrast CT. The unenhanced series (PA localizer, 5 mm slice thickness) of 180 CT pulmonary angiograms (CTPA) [90 consecutive CTPA positive for central and proximal interlobar PE taken from 3,161 consecutive CTPA (2.85 % prevalence) mixed with 90 randomly selected negative controls] were independently reviewed by two radiologists (R1 and R2). This was subsequently followed by concluding validation from a third radiologist with access to contrast-enhanced CTPA images and official reports. Six suboptimal studies were excluded. Patient age, gender, hemoglobin value and clot location were recorded. Interobserver agreement, sensitivity, specificity, positive predictive value (PPV) and negative predictive value (NPV) were determined. Sensitivity, specificity, PPV and NPV were 65 % [95 % confidence interval (CI) 54-75 %, 57/88], 49 % (95 % CI 38-60 %, 42/86), 3.6 % (95 % CI 2.8-4.6 %),97.9 % (95 % CI 97.1-98.6 %) for R1, and 55 % (95 % CI 44-65 %, 48/88), 76 % (95 % CI 65-84 %, 65/86), 6.2 % (95 % CI 4.3-9.3 %), 98.3 % (95 % CI 97.8-98.7 %) for R2. Patient age, sex and hemoglobin value were found to have no impact on PE detection. For strictly central PE, sensitivity was 80 % (45/56) (R1) and 70 % (39/56) (R2), while for bilateral PE, sensitivity was 74 % (48/65) (R1) and 66 % (43/65) (R2). For truly positive PE, interobserver agreement was 67 % (59/88) with a moderate Cohen's kappa of 0.41 (95 % CI 0.24-0.57). In conclusion, non-contrast CT shows high NPV for central PE identification, however it is neither sensitive nor specific enough to accurately detect it.