The regional ethics committee approved this retrospective study using pseudonymised data from patients referred for a clinical examination, waiving the need for individual consent. Paediatric non-gated cardiovascular PCCT (Naeotom Alpha, Siemens Healthineers, Erlangen, Germany) examinations with intravenous contrast agent completed between 30th September 2021 and 1st March 2023 in children with suspected or confirmed congenital heart defects were eligible. Initially, only 90 kV protocols were available as this was an early installation. After 70 kV was made available by the vendor, all clinical examinations were performed at 70 kV, which also limited the number of 90 kV examinations in the current study. The aim was to include the same number of examinations using 70 kV as had been completed using 90 kV.

Neonates and infants were positioned in a vacuum pillow and imaged during free breathing. Small children between 1 year and 3 years were sedated with Propofol (Sandoz AS, Novartis, Stockholm, Sweden) according to clinical routine, and imaged during free breathing. Children above 3 years were generally not sedated, and images were acquired during breathhold when possible.

Tube voltage was 70 kV or 90 kV. Table 1 shows PCCT parameters. Different monoenergetic levels were tested before the current study as part of clinical optimisation, and 55 keV was considered the optimal monoenergetic level for visualisation of the cardiac chambers and the thoracic vessels. Therefore, the current study used 55 keV throughout.

During the initial phase of clinical acquisitions, different reconstruction algorithms were tested to find the optimal quantum iterative reconstruction (QIR) level as this was the first release of the system. Furthermore, reconstruction algorithms also changed with software upgrades. This explains why different QIR levels are present for 90 kV in the current study (Table 1), whereas 70 kV protocols applied the same QIR level throughout. The clinical optimisation process for choosing QIR 4 as the superior level was by consensus discussions among all paediatric cardiovascular radiologists in the department (including those not acting as observers in the current study), in a side-by-side comparison of QIR 2, QIR 3, and QIR 4 reconstructed images.

Finally, the selected image quality level 70 means that the system’s dose modulation compensated for the lower voltage, leading to less impact of noise that would otherwise increase with lower voltage.

The radiation dose parameters CT dose index (CTDIvol), dose-length product (DLP), and size-specific dose estimate (SSDE) were extracted from the PCCT system, and Eeff was calculated based on DLP using age-dependent conversion factors [9].

Iodixanol 270 mg I/ml (GE Healthcare, Stockholm, Sweden) was administered using a MEDRAD Centargo injector (Bayer Pharmaceuticals, Leverkusen, Germany) with 15 s bolus length. For bodyweight ≤ 10 kg, a lower extremity peripheral vein was used for contrast agent administration, and for bodyweight > 10 kg an upper extremity peripheral vein was used. Preloading of contrast agent was applied in patients with bodyweight ≤ 10 kg for the programmed injection of contrast agent to be delivered directly without intermediate saline.

For 70 kV, an optimal dilute concentration of 189 mg I/ml was determined as part of clinical optimisation, with constant injected volume and rate. For 90 kV, the clinical routine dose of 270 mg I/ml was used.

Before the evaluation of study cases, the observers participated in a session with other PCCT congenital heart defect cases to obtain a common assessment basis.

All study cases were fully anonymised, including the removal of all personal identifying information, date and time of acquisition, kV level, QIR level, contrast density, and all other scan information. Also, 70 kV and 90 kV cases were randomised, although perceived image differences related to the essentially noise-free images at 70 kV could not be overcome. Four blinded paediatric cardiobascular radiologists (F.S., P.W., M.W., and S.M. with 30, 28, 24, and 20 years’ experience, respectively) independently assessed the PCCT examinations with regard to overall and specific diagnostic quality.

Overall diagnostic quality was defined as how well the examination in total could answer the clinical questions. Specific diagnostic quality was defined as how well the anatomical structures of clinical value to the individual congenital heart defect case were visualised for diagnosis. The examinations were scored 1–4 where 4 corresponded to “high diagnostic quality”, 3 “acceptable diagnostic quality”, 2 “low diagnostic quality”, and 1 “insufficient diagnostic quality”.

All examinations were evaluated as per clinical routine with four image stacks prepared for the observers’ convenience: 0.6 mm transverse images, and 2 mm images in the transverse, coronal, and sagittal planes. Multiplanar reconstruction and volume rendering based on the 0.6 mm images were used as in clinical routine.

Statistical analyses were performed in Prism 9.5.1 (GraphPad Software, San Diego, CA) and in R [10, 11]. Values are reported as median (range) or median [interquartile range; IQR]. Due to the low score variability, standard interrater reliability measures for ordinal data are misleading. As observers were aligned in scoring before the study, chance is not the main driver for agreement. Therefore, percent rater agreement is presented. However, for comparison with other studies, intraclass correlation is also reported, assessed using a two-way, consistency, average-measures model. Mann–Whitney’s and Jonckheere-Terpstra’s tests were applied to test for score differences and for trends between Eeff and scores, respectively [12]. Differences between groups for 70 kV and 90 kV were assessed, and radiation dose differences were also assessed after correction for confounders. P < 0.05 was considered to show statistically significant differences.