Diagnostic Reference Levels in Medical Imaging

This is a welcome document.  Thank you for the hard work in putting it together.

Below are comments from the Medical Sector Committee of the Society for Radiological Protection in the United Kingdom.

• Consider adding “Alert levels” to the document.  Alert levels can be the “Alert Values” referred to in Computed Tomography Dose Check (NEMA Standards Publication XR 25-2010, October 2010, http://www.nema.org/stds/xr25.cfm).  It can also be the alert levels used in interventional radiology when the cumulative air kerma has reached a pre-defined level set by the user.
• Where it mentions that DRL is a useful tool for optimisation, consider this amendment: “…useful tool for harmonisation of practice and optimisation of radiological protection in medical imaging …” (e.g. in page 21 (8), page 27 (39), (41) and page 125 line 5, etc.)
• Page 35 (56) Table 2.1: Add paediatric CT, paediatric fluoro and radiotherapy radiological imaging (CT, kV imaging and CBCT) to examination selection for patient dose surveys
• Understanding of the clinical task at hand is very important in the process of optimisation and in the assessment of image quality.  Therefore in page 38 (73): Consider adding “an understanding of the clinical task”, e.g. “Since an understanding of the clinical task as well as imaging and radiation performance of the equipment is required for optimisation, periodic constancy testing……”
• Emphasise the importance of establishing standard examination codes for the different types and variants of radiological procedures.  This is particularly important in automated data extraction methodologies.  For example, in CT and fluoroscopic procedures, there can be a variety of protocols for the same procedure type.  Would somewhere near page 41 (86) be a suitable place to add this recommendation?
• Also page 41 (86) line 5: suggestion for re-phrase: “Data should be examined to avoid gross errors and if appropriate, removal of the top 5% and bottom 5% of the data may be considered in order to remove outliers”
• Page 102 line 3: added “very much” to “When the median value of a DRL quantity at a facility is very much lower than the median value of the benchmark national or regional DRL survey distribution, image quality (or diagnostic information, when multiple images are used) should be examined as a priority in the review.”  This is because if the median value is lower than the local DRL, this may be just a result of on-going successful optimisation work.  It is only when the median value is very much lower than the DRL, then one might begin to wonder if image quality is adequate for the clinical task.  Now, one might then have to define how much is “very much”.  Also, it is recommended that “should be examined” be changed to “may need to be examined”.
• Page 103 Figure 7.1: Recommend that the flow chart be moved nearer to the front of the document, rather than having it sitting close to the back of the document.  If needs be, the flow chart can be repeated.

Section 1.6, paras 33 & 34, page 26

The concept of ‘Achievable Dose’ is much more readily understood than DRLs, so is to be welcomed in this document. However, after explaining the concept in paras 33 & 34, para 35 then reverts to the much more complex phrase “median value of the distribution of a DRL quantity observed in a survey of healthcare facilities”. It has been suggested that using ‘Achievable Dose’ here and throughout the document would improve and simplify the understanding of the text.

Section 1.6, para 35, page 26

Paragraphs 33 & 34 define ‘Achievable Dose’ as the “median value of the distribution of a DRL quantity observed in a survey of healthcare facilities”, whereas paragraph 35 appears to use the same definition to describe “an additional value specified that would serve as a simple test to identify situations where levels of patient dose are low and investigation of image quality should be the first priority”. One can’t have an achievable dose and a trigger for situations where the dose might be too low set at the same value!

The idea of combining an Achievable Dose (a dose for a group of patients that one might expect to be able to achieve), coupled with a DRL (a dose that one would not be expected to exceed), might be a way to proceed, but one couldn''t see this explained as simply as this anywhere in the text.

Section 2.1, para 49, page 34

The document refers to comparison of local data with a DRL as being the ''first step'' in optimisation. It has been commented that this ought to be the ''last step''. Here''s the main issue that some people have with DRLs - they are derived using an arbitrary cut- off in the ranking of mean (or median) doses from a wide range of practices, over a wide geographical area, using a broad range of equipment types. However, one does not know which, if any, of these doses are ''optimised''. It has ben argued that ideally, the ''first step'' in optimisation should be to set up the system optimally before use and subsequently keep it there. In this case, one should be using an internal reference point, not an external one. A statistical process control approach, similar to the way one monitored processor temperature in the old days, would be much more sensitive to local change. Essentially, in an ideal situation, one should get it right to start with, balancing adequate image quality with lowest reasonably achievable dose, then monitor the dose continuously, or periodically, to make sure it doesn''t wander too far off. The statistical criteria for determining the threshold for triggering action will use data generated from the same equipment. Once we are all getting it right, using the most dose-efficient equipment in the most dose-effective way, there should be quite a narrow distribution of median doses across facilities, in which case it would not matter very much precisely where a facility sits in that distribution. The data reported by Sutton, referenced as ‘Sutton 2014’ appears to demonstrate that this has indeed happened in this particular situation.