Routine head CT dose varies threefold between protocols in common use

Design

retrospective analysis of an international CT dose registry using k-means clustering, stratified by age group and acquisition approach

Population

904,209 head CT series — 864,182 in adults and 40,027 in children, 2015 to 2021

Primary outcome

size-adjusted dose-length product and volume CT dose index across clusters of acquisition technique

Effect

more than twofold and up to 3.4-fold variation; adult helical range 372 to 998 mGy·cm (relative dose 2.7)

Nearly a million head CT series — 864,182 in adults and 40,027 in children, acquired between 2015 and 2021 and recorded in an international dose registry — were grouped by k-means clustering on tube current, voltage, collimation, scan length and pitch. The point was not to design an ideal protocol but to find which of the protocols already in routine use deliver the least radiation.

Adult helical scans fell into seven clusters, adult axial into six, paediatric helical into six and paediatric axial into four. In every stratum, mean size-adjusted dose-length product varied more than twofold across clusters, and up to 3.4-fold. For adult helical scans, the largest group, size-adjusted dose-length product ranged from 372 to 998 mGy·cm — a relative dose of 2.7 between the lowest and highest clusters, for the same examination. Lower-dose clusters generally used lower tube current and voltage, but there was no single route to a low dose; different protocols got there differently.

This is a benchmarking finding rather than a discovery. It says that somewhere in the registry are units doing routine head CT at a third of the dose others use, with protocols that are clinically established and widely used rather than experimental. The implication for any department is to find out which cluster its own protocol resembles.

The practical caveat is image quality, which the study did not measure. A dose reduction that costs diagnostic confidence in posterior fossa detail is not a saving. But a 2.7-fold spread across established protocols is too large to be explained by quality alone, and it is checkable locally: pull the size-adjusted dose-length product from your own dose-monitoring data and compare it with the range reported here.

• Pull size-adjusted dose-length product for your routine head CT from dose-monitoring data and compare it with the 372 to 998 mGy·cm adult helical range
• Review tube current and voltage first — those distinguished the low-dose clusters most consistently
• Check paediatric protocols separately; the variation was as wide there and the consequences are larger
• Do not change a protocol without a parallel read of image quality on the studies that need the most
• Repeat the audit after any scanner or software upgrade, which silently resets defaults

The statistics, in plain English

Cluster analysis describes how practice groups itself; it does not establish that any cluster is optimal, because no diagnostic outcome was measured. Size-adjusted dose-length product is the right comparator since it accounts for patient diameter, so the spread is not explained by body habitus. With 904,209 scans, the ranges are precisely estimated; what remains unknown is the image quality delivered at 372 mGy·cm compared with 998.

Read the rest in the app

You have read your two free briefings this month. The app carries all 27 specialties, every morning, free — and this finding is waiting in it.

Get Daily Doctor — free

Scan to keep reading on your phone. No account needed to start.