Diagnostic Reference Levels for nuclear medicine imaging in Austria: A nationwide survey of used dose levels for adult patients

Purpose To assess dose levels in routine nuclear medicine (NUC) procedures in Austria as a prior to a legislative update of the National Diagnostic Reference Levels (NDRL). Method As part of a nationwide survey of common NUC-examinations between June 2019 and November 2019, data sets were collected from 33 Austrian hospitals with NUC equipment. All hospitals were asked to report the NUC imaging devices in use (model, type, year of manufacture, detector material, collimators), the standard protocol parameters for selected examinations (standard activity, collimator, average acquisition time, reconstruction type, use of time-of-flight) and to report data from 10 representative examinations (e.g. injected activity, weight), incl. the most common NUC-examinations for planar imaging/SPECT and PET. Median/mean values for injected activity were calculated and compared to current Austrian and international NDRL. A Pearson correlation coefficient was computed comparing different variables. Results In total, all 33 hospitals (100% response rate) reported data for this study for 60 SPECT devices, 21 PET/CT devices and 23 scintigraphy devices. Fixed activity values for scintigraphy/SPECT and PET were employed by about 90% and 56% of the hospitals, respectively. The most widely performed examinations for scintigraphy/SPECT are bone imaging, thyroid imaging, renal imaging (with MAG3/EC) and lung perfusion imaging (in 88% of the hospitals) and F-18 FDG-PET studies for oncology indications (in 100% of the hospitals). Significant correlations were found for patient weight and injected activity (scintigraphy/SPECT), use of iterative reconstruction and injected activity (PET) as well as size of field-of-view and injected activity (PET). Conclusions The reported injected activity levels were comparable to those in other countries. However, for procedures for which NDRL exist, deviations in injected activities of >20% compared to the NDRL were found. These deviations are assumed to result mainly from advances in technology but also from deviations between NDRL and prescribed activities as given in the information leaflets of the radiopharmaceuticals.


Introduction
Over the past century, medical imaging has evolved to an indispensable part of diagnostic procedures. However, some of the most common imaging methods, such as X-ray, computed tomography (CT) as well as nuclear medicine imaging (NUC) procedures, such as scintigraphy, single photon emission computed tomography (SPECT) and positron emission tomography (PET) make use of ionizing radiation, and, thus, inherently imply radiation exposure to patients [1]. Even though the exposure from diagnostic procedures is small (in the range of 0.01 mSv to 20 mSv effective dose) and eventual negative effects of such small exposures are controversially discussed [2][3][4], in sum they substantially contribute to the general radiation exposure of the population [1,5,6].
To limit the general radiation exposure of the population and potential adverse effects to individual patients, diagnostic procedures including ionizing radiation need to follow the ALARA (As Low As Reasonable Achievable) principle. Thus, exposure from medical imaging should be limited to the lowest level necessary to reliably answer the respective diagnostic question. To achieve this, diagnostic procedures should be continuously optimized, among others considering technical advances in imaging technologies and changes in diagnostic practice.
To monitor and facilitate the dose optimization process, the International Commission for Radiologic Protection (ICRP) introduced the concept of diagnostic reference levels (DRL) [7][8][9]. DRL are standard levels of easily measurable quantities, such as the dose length product in CT or injected activity in NUC, for common procedures. DRL set a general guidance for the dose in clinical operation and help to identify if routine doses are unusually high or low. They do not apply directly to individual examinations and patients [7].
According to the recommendations of the ICRP [7] and European legislation [10], DRL are set and updated on national level based on the assessment of observed dose distributions. Therefore, national radiological and NUC procedures need to be regularly surveyed to account for changes in diagnostic practice and technological advances.
In the standard literature, two main approaches for setting national diagnostic reference levels (NDRL) in NUC are discussed: first, as outlined in Radiation Protection Radiation Protection N • 109 [11] of the European Commission, NDRL in NUC are administered doses necessary for a good image during a standard procedure, which follow the concept of "optimal doses". The DRL should be approached as closely as possible and a substantial deviation of these optimal doses is only permitted for special occasions, such as for obese patients. Consequently, following this concept, NDRL for NUC are not set at the third quartile but rather at the median or mean of each examination type and these recommendations should be developed in close cooperation with relevant societies and experts. This concept is implemented in Austria [12] or Switzerland [13] and was recommended by the European Commission in 2014 [14]. Another approach would be to use the same definition and system for DRL as it is valid for CT or radiography. Consequently, NDRL are determined based on the third quartile of a dose distribution. If these NDRL are on average constantly exceeded, justification is necessary. The ICRP 103 [7] recommends using this approach and several publications were recently published [15][16][17]. Also a combination of these two outlined approaches is possible and was described by A Shahzad, S Bashir and A Anwar [18].
In Austria, NDRL are regulated in the Austrian Medical Radiation Protection Ordinance [12,19]. For NUC, this ordinance states "Diagnostic reference levels for nuclear medicine examinations are defined as activities to be administered to standard-sized adult persons." [12] and a deviation from the NDRL by more than 20% must be justified. Consequently, in the course of this study, NDRL were developed based on the median (and mean) values in close cooperation with NUC experts and under involvement of all relevant societies. Additionally, third quartiles of the dose distribution are provided for comparison with other NDRL derived based on the third quartile.
The last evaluation of diagnostic procedures and respective doses of NUC examinations in Austria was performed in 2008 [20,21]; the NDRL were updated in 2010 accordingly [19]. Since then, significant technological advances in imaging technologies have been made and new radiopharmaceuticals have come to use. Technical advances include the introduction of direct photon detection with cadmium-zinc-telluride (CZT) based detectors for scintigraphy and SPECT imaging [22]. For specific organ examinations, such as myocardial perfusion imaging, dedicated SPECT devices and optimized collimators are more commonly available now [23]. In PET imaging, a general trend to extended axial field-of-views (FOV), and, thus, improved sensitivity can be seen [24,25]. Furthermore, the implementation of semiconductor-based PET detectors and improved detection electronics led to significant advances in time-of-flight (TOF) resolution, and, in turn, improvements in signal-to-noise ratios [26]. Finally, progress in image reconstruction has been made within the last decade for SPECT as well as PET imaging [27,28].
In parallel to the technical development, new diagnostic agents and respective procedures were established. For example, in the last decade Tc-99m-and Ga-68 labelled PSMA was introduced for the assessment of prostate cancer [29]. F-18 labelled amyloid tracers for dementia imaging are now commonly used [30]. F-18 labelled choline has been established for parathyroid imaging [31] and cardiac SPECT examinations shifted from mainly based on thalliumchloride to the use of technetium labelled perfusion tracers [32].
The Federal Ministry of Health and the Austrian National Public Health Institute engage in updating the Austrian NDRL for medical imaging procedures [33,34]. To assess these above-mentioned changes in technology and clinical practice and to update the enforced NDRL for NUC, a survey of NUC procedures was performed across Austria. This included the assessment of the used technology, examination protocols and administered activities. Here, we report on the methodology and the outcome of this survey which was the basis of the update of the NDRL for NUC in Austria.

Material and methods
The study design of this nationwide survey was conducted with reference to similar international studies and based on the recommendations by the ICRP 135 [7].

Examination types
To define relevant examination types for the survey, Austrian NDRL for NUC (which were enforced at that time) were used [19]. In addition, the last study on Austrian NDRL for NUC by A Stemberger, T Leitha and A Staudenherz [21], a report by the European Commission [14] on European NDRL, NDRL from Switzerland [13] and Germany [35] were included and discussed with a multidisciplinary expert group consisting of radiologists, medical physicists and radiographers as well as representatives from relevant societies. Frequencies of relevant NUC-examinations were taken from the Austrian-wide routine documentation data held and maintained by the Austrian Federal Ministry of Health [6,36]. Table 1 and Table 2 depict the examination types and indications included into the survey besides obligatory and optional parameters, respectively.

Survey design and data entry form
In Austria, approximately 39% of the NUC examinations are performed in private imaging centres (= ambulatory care sector) and 61% in hospitals (only including examinations which are fully covered by the social insurance system) [30]. The predominant examination in the ambulatory care sector, approx. 83%, are thyroid imaging examinations. The inclusion of private imaging centres would have caused practical issues by means of availability of contact data, thus, the decision was made that this survey will only include hospitals. We expect that the data collected is representative also for private imaging centres as the majority of examinations are performed in hospitals and thyroid imaging is a well standardised examination.
All 33 hospitals with NUC equipment were invited via e-mail, followed up by e-mail reminders and, if necessary, telephone calls to non-responding hospitals to increase response rates. The data collection period lasted from June 2019 to October 2019. The survey was carried out using a HTML based online data entry form. Consequently, this offered the possibility to identify individual data entries, create standardized reports and retrospectively trace and analyse each data entry.  The survey consisted of three parts; (a) the assessment of the imaging devices used, (b) the definition the local standard protocols and (c) the collection of individual patient data.
(a) First, each hospital was asked to declare the NUC imaging devices in use. This included the manufacturer, model and year of manufacture of the NUC-imaging devices, detector material and the axial FOV (for PET) ( Table 1 and Table 2).
(b) For each device, the hospitals were asked to enter the standard protocol parameters for the selected examinations. This included the standard activity (fixed or per kg), the colli-mator as well as (for PET) the average acquisition time per bed position, use of iterative reconstruction (IR) and TOF (Table 1  and Table 2).
(c) For each examination type, the hospitals were asked to report data from 10 representative examinations. The required information was the injected activity and the weight of the patient. On an optional basis, the form allowed to report the height, sex and year of birth of the patient.

Data quality
Measures were taken to increase response rates and improve data quality. To incentivise study participation, analyses comparing the median activity values of an individual participating hospital to the new NDRL recommendations were advertised and made available to the participating hospitals. During the data collection phase, a questions and answers hotline was installed to allow participating hospitals to directly call and talk to experts in order to assist with and answer questions regarding data entry. Automated pop-up warning messages were integrated into the data entry form to directly inform the participants on entries with highly implausible values. In such cases the software would prompt the user to check the dose entered. To further improve data quality, data entries were centrally monitored during the data collection period and hospitals entering implausible dose values were contacted in short order to clarify eventual faulty entries. After the data entry period ended, data were again checked for plausibility.
After this plausibility check, data cleaning was performed especially for the Tc-99m DMSA examination type as several patient datasets were entered including children. Consequently all datasets with patients younger than 18 years were excluded (for planar/SPECT 123 datasets and for PET 0 datasets). For datasets without the year of birth provided, patient weight was used as a proxy and all datasets with patient weight equal or lower 35 kg were excluded (for planar/SPECT 21 datasets and for PET 0 datasets). The number of patient datasets with weight between 35 kg and 70 kg was 1,403 (40% of all datasets) for planar/SPECT and 353 (34% of all datasets) for PET. Furthermore, for myocardial imaging, values for stress and rest examinations had to be reported. If only stress examination were reported for myocardial imaging and no rest value was available, these datasets were excluded (in total 5 protocol datasets and 62 patient datasets). 13 patient datasets for the examination type thyroid whole body imaging were excluded as they were performed for therapy purpose.

Statistical analysis
All analyses were conducted with the statistical software package R (Version 3.3.1; The R Foundation, Free Software Foundation) and Microsoft Excel (Office 365, Microsoft). Median, third quartiles, mean, standard deviation (SD), maximum and minimum were calculated for each examination type. Pearson's correlation coefficient was calculated to anal-yse the association between different variables such as injected activity and year of manufacture of the devices. Significance was set at a P value of <0.01.

Results
In total, all 33 hospitals (100% response rate) reported data for this study. 64% (N = 21) of the hospitals have a dedicated scintigraphy device (a scintigraphy device without a SPECT option, e.g. dedicated thyroid cameras) in operation, 94% (N = 31) of the responding hospitals are equipped with a SPECT or SPECT/CT device, and 55% (N = 17) of those hospitals with a SPECT or SPECT/CT device additionally operate a PET/CT. 58% (N = 18) of the hospitals with SPECT or SPECT/CT have more than one SPECT or SPECT/CT device in operation, whereas only 24% (N = 4) of the hospitals operating a PET/CT have more than one PET/CT device. Table 3 and 4 summarizes the number of hospitals per examination type. a) Devices: In total, data was reported for 60 SPECT or SPECT/CT devices and for 21 PET/CT devices in operation. In addition, data were submitted for 23 scintigraphy devices.  Table 3 and Table 4 the number of devices per examination type is illustrated. b) Examination protocols: In total, data for 463 examination protocols for planar imaging/SPECT and 112 examination protocols for PET were collected. The administered activity was based on a fixed activity value for 87% (N = 401) of the planar/SPECT-examination protocols, whereas for 13% (N = 64) of the protocols administered activity was based on activity per body weight. For PET fixed administered activities were used in 56% (N = 63) of the protocols, while for 44% (N = 49) an activity per body weight approach was used. Table 3 and Table 4 summarize the examination protocols for planar imaging, SPECT and PET/CT examinations together with the mean and median values of the fixed activities and activity per body weight as defined in the protocols for the two activity determination methods, respectively. c) Patients: In total, data for 3,505 individual patients were submitted for planar imaging/SPECT and for 1,045 individual patients for PET. The number (N) of reported values together with the mean weight in kg, mean and median injected activity (in MBq), standard deviation as well as minimum and maximum activity values are reported in Table 3  Table 3 Planar imaging/SPECT -Number (N) of hospitals/devices, data for examination protocols and collected patient data.     Table 5 the mean and median activity values for patient data from this study (planar/SPECT) are compared to the old NDRL for Austria, NDRL for Switzerland, new NDRL from Germany [37], Croatia and EU. In Table 6 the same is illustrated for PET. The most widely executed examination types for planar imaging/SPECT are bone, thyroid, renal imaging with MAG3/EC and lung perfusion imaging (Figure 1). Data for these examinations was submitted from 88% (N = 29) of all hospitals. The least widely used examination is renal imaging with DTPA receiving data from only 6% (N = 2) of the hospitals. For PET, tumour imaging with FDG is by far the most extensively performed examination (100% of all hospitals with PET/CT; N = 17) whereas tumour imaging with Cu-64 PSMA was documented in only 18% (N = 3) of all hospitals.
ICRP 135 [7] recommends using a weight interval for average sized patients from 60 to 80 kg. To investigate the effects of light and heavy patients on the calculation of the median activities, a new analysis was performed excluding those datasets related to patients weighing less than 60 kg or more than 80 kg (indicated here as "weight-restricted dataset"). Percentage differences between the analysis with all patients and the weight-restricted dataset are shown in Figure 2. With imposing weight restrictions, the overall mean weight decreased from 77 kg to 70 kg for planar imaging/SPECT and from 78 to 71 kg for PET. For planar imaging/SPECT the most impact on mean weight (-16%) was found for Tc-99m PSMA examinations, inflammation imaging (-14%) and myocardial perfusion imaging (-13%). Weight restriction did not impact the findings for "bleeding". Regarding the differences between median activity, highest difference was found for adrenal imaging with -29%. However, for most of the examination types only minor or no differences were present in the median injected activities after restriction of body weight. Similar could be observed for PET imaging with highest weight differences for myocardial imaging and lowest for brain imaging, however, with associated median activity differences of only 1% to 3%.
To investigate the relationship between different variables a Pearson correlation coefficient was computed. No significant correlation was found between the following variables:

Discussion
This study summarizes the outcome of a recent survey of the employed activity levels in nuclear medicine procedures in Austria with the aim to provide a well-founded basis for the update of the NDRL. The study design was tailored to achieve a maximum response rate to enable a valid assessment of the used dosages given the comparatively low absolute number of nuclear medicine departments in Austria. This was successfully achieved for in hospital nuclear medicine departments with a response rate of 100%.
The collected dose data are in general comparable to those found in other recent studies (Table 5 and Table 6) [13][14][15]35,38]. This is reasonable as today several international guidelines and recommendation for most of the standard examinations exist, which were developed in multilateral consensus between the national professional societies (e.g. [39][40][41]). However, compared to the, at the time of the survey valid NDRL in Austria, deviations of >20% in average activity were found for nearly half (8/17) of the surveyed procedures (Table 5 and 6). For most of the examinations with deviation of >20%, injected activities were found to be lower than the NDRL. For example, the average injected dose for tumour imaging with F-18-FDG was found to be 30% below the NDRL. This is expected to result from technological improvements, such as 3D PET data acquisition, IR and increased sensitivities.
During the data analysis also different reasons for deviating injected activities where found. For some of the radiopharmaceuticals the NDRL valid at the time of the survey where not in agreement with the dose recommendation for the radiopharmaceuticals. For example, Thyroid imaging using Tc-99m pertechnetate was done on (median) injecting 76 MBq and parathyroid imaging using Tc-99m isonitrile using (median) 400 MBq (Table 5). Both median activities were below the NDRL of 110 MBq and 740 MBq, respectively. However, the injected activities were in agreement with the recommended dosage as defined in the pharmaceutical information leaflet of 20-80 MBq and 200-700 MBq, respectively. Such inconsistencies in prescribed doses and NDRL should be avoided.  Thus, prescribed doses need to be reviewed and taken into account during updates of NDRL.
In general, the results nicely show the necessity to update DRL on a regular basis to account for changes in clinical practice and technological advances. As seen in the correlation analysis, especially IR and increased sensitivity due to extended FOVs in PET devices seemed to contribute to a general dose reduction. While being reasonable as such, it is also in line with findings for the use of IR in CT [34]. Another interesting finding pertains to the correlation of administered activities with weight. Although, a general significant correlation of patient weight with dose was found, a re-evaluation of the data in a subset of standard patients with weights of 70 ±10 kg revealed no practically considerable changes in injected activity levels for most of the protocols (Figure 2). The reason therefore is expected to be the relatively high number of protocols with fixed activity levels (87%). This facilitates the derivation of DRL as deviations of data reported for individuals from a standard patient do not influence the general average doses. However, rigor data inspections and weight restrictions might be necessary to derive reasonable DRL in cases where patient weight-based dosage is used in the majority of sites.

Limitations
To keep the workload for participating hospitals reasonable and to achieve a high participation rate, we limited the minimum number of patient datasets per examination type to 10. This approach was also used in previous studies [33,34] and is regarded as appropriate, if the participation rate is high. As NUC examinations are more standardised than radiological examinations and all hospitals with NUC equipment participated, we believe the data quality in this work represents a Table 7 Austrian NDRL for nuclear medicine as enforced in 2020. good compromise between theoretical optimum and practicability of the data collection.

Conclusion
This work presents data on injected activities used in clinical practice for diagnostic nuclear medicine procedures in Austria to provide a base for the update of the NDRL. The injected activities found are comparable with those reported in other countries. However, for procedures for which NDRL exist, deviations in injected activities of >20% compared to the NDRL were found. Reasons therefore are assumed to result mainly from advances in technology but also from inconsistencies between NDRL and prescribed dosages as given in the information leaflets of the radiopharmaceuticals.

Funding
This study has received funding by the Austrian Federal Ministry of Health.

Conflict of interest
Michael Hinterreiter is an employee of GE Healthcare Handels GmbH (Technologiestraße 10, 1120 Vienna, AUT). The other authors declare that they have no conflict of interest.