• Citer

CT dose survey in adults: what sample size for what precision?

par Taylor, Stephen ;Van Muylem, Alain ;Howarth, Nigel;Gevenois, Pierre-Alain ;Tack, Denis
Référence European radiology, 27, 1, page (365-373)
Publication Publié, 2017-01

Article révisé par les pairs

• Détails
• Contenu

Résumé :

Objectives: To determine variability of volume computed tomographic dose index (CTDIvol) and dose–length product (DLP) data, and propose a minimum sample size to achieve an expected precision. Methods: CTDIvol and DLP values of 19,875 consecutive CT acquisitions of abdomen (7268), thorax (3805), lumbar spine (3161), cervical spine (1515) and head (4106) were collected in two centers. Their variabilities were investigated according to sample size (10 to 1000 acquisitions) and patient body weight categories (no weight selection, 67–73 kg and 60–80 kg). The 95 % confidence interval in percentage of their median (CI95/med) value was calculated for increasing sample sizes. We deduced the sample size that set a 95 % CI lower than 10 % of the median (CI95/med ≤ 10 %). Results: Sample size ensuring CI95/med ≤ 10 %, ranged from 15 to 900 depending on the body region and the dose descriptor considered. In sample sizes recommended by regulatory authorities (i.e., from 10–20 patients), mean CTDIvol and DLP of one sample ranged from 0.50 to 2.00 times its actual value extracted from 2000 samples. Conclusions: The sampling error in CTDIvol and DLP means is high in dose surveys based on small samples of patients. Sample size should be increased at least tenfold to decrease this variability. Key Points: • Variability of dose descriptors is high regardless of the body region. • Variability of dose descriptors depends on weight selection and the region scanned. • Larger samples would reduce sampling errors of radiation dose data in surveys. • Totally or partially disabling AEC reduces dose variability and increases patient dose. • Median values of dose descriptors depend on the body weight selection.