Long-term experience and analysis of data on diagnostic reference levels: the good, the bad, and the ugly

To analyze 11-year data of France for temporal trends in dose indices and dose optimization and draw lessons for those who are willing to work on creation and update of diagnostic reference levels (DRLs).

The data from about 3000 radiology departments leading to about 750,000 imaging exams between 2004 and 2015 was analyzed, and patterns of reductions in dose for those below and above the DRLs were estimated and correlated with technology change.

Dose optimization achieved was important and significant in departments which were above or just below the DRL (p = .006) but not in those which were around half of the DRL values. The decrease in 75th percentile value of Kerma air product (KAP) for chest radiography by 27.4% between 2004 and 2015 was observed with the number of flat panel detectors increase from 6 to 43%. A good correlation between the detector type distribution and the level of patient radiation exposure is observed. Otherwise, setting DRLs for standard-sized patient excludes patients lower and higher weighted than "standard."

The concept of DRL may become obsolete unless lessons drawn from the experience of users are taken into account. While establishing DRLs should be part of the regulations, setting up and updating values should be governed by bodies whose decision-making cycle is short, at the most 1 year. A local rather than national approach, taking into account body habitus and image quality, needs to be organized.

• The technology changes faster than regulations. Requirement of DRL establishment should be part of the regulations; however, setting and updating values should be the role of professional societies. • The concept of DRL, highlighting the 75th percentile values and dedicated to standard-sized adult, misses optimization opportunities in the majority of patients who are below the 75th percentile value and outside the range of standard-sized adult. • The ugly aspects of the DRL concept include its non-applicability to individuals, no customization to clinical indications, and lack of consideration of image quality.