Radiation exposure and mortality risk from CT and PET imaging of patients with malignant lymphoma

Malignant lymphoma comprises a heterogeneous group of diseases, differing with regard to histology, treatment and outcome. It is beyond the scope of this study to encompass all the different entities. Instead we focus on the most common types of malignant lymphoma; i.e. HD in children and diffuse large B-cell lymphoma (DLBCL) in adults [27‐30]. The most typical imaging strategy for children with HD (age <18 years) and for adults with DLBCL was used in this study. In children with HD, the imaging strategy was based on the protocol according to the Children’s Oncology Group [COG]; in adults with DLBCL, the imaging strategy was based on the HOVON 84 international multicentre trial currently running in The Netherlands [www.​hovon.​nl, EudraCTnr. 2006-005174-42] (Table 1).

Imaging with ultrasound and chest radiography were not considered in this study, as ultrasound is not associated with radiation exposure, and the radiation exposure from chest radiography is negligible compared with CT and ¹⁸F-FDG PET.

There was no information available that allowed for assessment of organ doses in paediatric patients undergoing CT. Therefore we created Medical Internal Radiation Dose (MIRD) paediatric patient models in five general categories according to age and weight: newborn (3.6 kg), 1 year old (9.7 kg), 5 years old (19.8 kg), 10 years old (33.2 kg) and 15 years old (56.8 kg). The adult MIRD patient model represents an average-sized patient of 74 kg. All these hermaphrodite patient models (MIRD V) are described by spheres, ellipsoids and cones, as illustrated in Fig. 1 [31]. For assessment of radiation exposure from CT, we used these patient models in combination with an algorithm for Monte Carlo dose calculations (ImpactMC software, version 1.0, VAMP, Erlangen, Germany) [31, 32]. The Monte Carlo dose calculations are based on a virtual model of the CT system with respect to geometry, X-ray spectrum, filtration and CT parameters. The simulation is performed on 3D voxelised versions of the MIRD V phantoms. To each voxel in the volume during the simulation, a density value and a mass attenuation coefficient are assigned corresponding to five different materials: air, lung, soft tissue, fat and bone. During the simulation of a CT examination, the energy deposited in each voxel (absorbed dose) is accumulated and saved in an additional volume. The manufacturer of the Aquilion CT system (Toshiba Medical Systems) disclosed two measured X-ray spectra (100 kV and 120 kV) and information about the design of the CT system. This information was implemented in the Monte Carlo simulation. A MatLab script (The MathWorks, Natick, MA, USA) was used to extract organ doses from the calculated dose distributions within the MIRD mathematical phantoms.

In CT it is common practice to adjust acquisition parameters to the size of the patient, and to the clinical application [33]. The acquisition protocol we used was based on local and optimised practices, and was checked against general recommendations, particularly with regard to the optimisation of the paediatric acquisitions. We thus derived the following clinical CT acquisition parameters. For small children, a tube voltage of 100 kV was used (weight <30 kg), and for larger children and adults a tube voltage of 120 kV (≥ 30 kg), both in combination with a pitch factor of 0.83. Radiation output of the CT system is expressed as the volume computed tomography dose index (CTDIvol), but also the tube charge (mAs) is provided for scans that are performed with an Aquilion 64 CT system (Toshiba Medical Systems, Japan) [34]. The clinically applied tube current, rotation time and pitch factor are associated with CTDIvols of 2.1 mGy (newborn, 3.6 kg, 27 mAs), 2.9 mGy (1 year old, 9.7 kg, 38 mAs), 4.1 mGy (5 years old, 19.8 kg, 53 mAs), 4.6 mGy (10 years old, 33.2 kg. 37 mAs), 7.1 mGy (15 years old, 56.8 kg, 58 mAs), and 9.5 mGy (adult, 74 kg, 77 mAs).

Dose calculations that were performed for the MIRD phantoms with the Monte Carlo software yielded organ doses, total body dose and effective dose according to ICRP 103. Dose assessment was performed for 12 CT acquisitions, i.e. both for whole-body CT (including neck, chest and abdomen), and for CT of the neck and chest only, in all six age categories. These doses were incorporated into the risk model, as described under “Risk assessment”. From these results, appropriate organ doses and the effective dose were derived for the year of diagnosis and treatment and for the following years of surveillance. If necessary, linear interpolation of dose values was performed to yield dose estimations at ages not included in the table.

For assessment of organ dose and effective dose from ¹⁸F-FDG PET, published tables were used that provide information about organ dose and effective dose per MBq of administered ¹⁸F-FDG activity. The ICRP provides information for 1-, 5-, 10- and 15-year-old children, and for adults [35]. Ruotsalainen et al. [36] published tables for estimation of radiation dose to the newborn in ¹⁸F-FDG PET studies. For PET a dose of 3 MBq of ¹⁸F-FDG per kg body weight was assumed to be administered, based on the current state-of art imaging with integrated PET/CT systems (f.i. Biograph 40 TruePoint PET-CT, Siemens Medical Systems, Knoxville, TN, USA) . The five different age categories in the children being analysed in this study correspond to ¹⁸F-FDG doses of 10 MBq (newborn), 30 MBq (1 year old), 60 MBq (5 years old), 100 MBq (10 years old) and 170 MBq (15 years old). In adults, administration of 220 MBq ¹⁸F-FDG was assumed.

We performed risk assessment for five age categories of male and female paediatric patients diagnosed with HD (newborns, and children at the age of 1, 5, 10 and 15 years) and in three age categories for adult male and female patients diagnosed with DLBCL at the ages of 55, 65 and 75 years.

The BEIR VII excess relative risk (ERR) model was used for calculating radiation risk [37]. Chapter 12 of this BEIR report provides the equations and parameters for estimating organ-specific solid cancer mortality and leukaemia mortality. The ERR is expressed as a function of gender, absorbed organ dose, age at exposure and the attained age, and includes three organ-specific fit parameters. The organs are stomach, colon, liver, lung, breast, prostate, uterus, ovary, bladder, other organs and thyroid.

In the BEIR VII ERR model, the organ-specific excess risk for solid cancer mortality is expressed, relative to the gender- and age-dependent risk of the background cancer mortality for specific organs. This background is the naturally occurring mortality, not the mortality that is induced by radiation exposure during medical imaging. Data on the risk of naturally occurring cancer mortality depending on organ, gender and attained age were derived from ICRP Publication 103 (Euro-American cancer mortality rates by age and site) [38]. Subsequently, according to the BEIR VII model, an overall ERR function depending on organ dose, gender, age at exposure and attained age was calculated for male and female patients. Organ dose was incorporated into the risk model as a function of attained age; this was required because patients have different weights (and ages) and thus receive different associated radiation exposures during diagnosis, treatment and surveillance of malignant lymphoma. In addition to radiation risk, mortality rates that are typical for the young patient group with HD and the adult patient group with DLBCL could be taken into account, i.e. the overall 10-year survival for HD (94%) and 5-year survival for DLBCL (58%) [27‐30].

Risk calculations were performed using life tables, and they were done with and without taking into account the disease-related mortality. Life tables (also called mortality tables) are used in demography for measuring and modelling population processes. A life table shows, for each age, what the probability is that a person of that age will die before his or her next birthday. They allow for the calculation of the fraction of radiation-induced deaths, the reduction of life expectancy and the survival rate. An essential component of the life table is age- and gender-dependent mortality; in this study three different sources of mortality were integrated in the life tables: the background, the radiation induced, and the disease-related mortality. The background mortality that is typical for the asymptomatic population (gender- and age-specific probability of dying) was derived for the European population at large based on the Eurostat database [39]. The radiation-induced, age-, gender- and dose-dependent overall mortality that was calculated with the BEIR VII model as described in the previous section, was also incorporated into the life tables. Finally, disease-related mortality, as mentioned in the previous section, can also be integrated in the life tables. Calculations were carried out with the life tables with and without taking into account the mortality risks that are typical for patients with malignant lymphoma (HD- and DLBCL-related mortality rates).