Switching from linear to macrocyclic gadolinium‐based contrast agents halts the relative T1‐Weighted signal increase in deep gray matter of children with brain tumors: A retrospective study

Studies have shown signal intensity (SI) changes in the brains of children exposed to repeated doses of a gadolinium‐based contrast agent (GBCA).

T HE FIRST INTRAVENOUS gadolinium-based contrast agent (GBCA) for magnetic resonance imaging (MRI) was licensed globally for clinical use in 1988. 1 Apparent deposition of gadolinium in the brain following repeated doses of GBCA was recognized in 2014 by Kanda et al. 2 The authors reported a correlation between increased signal intensity (SI) within the dentate nucleus (DN) and globus pallidus (GP) and the number of doses of GBCA received.
Several studies have shown SI changes in the brains of children exposed to repeated doses of GBCAs. [3][4][5][6][7][8][9][10][11][12][13] This is important because pediatric patients (assuming a normal life expectancy) have a longer period over which neurotoxicity can act or become manifest and be more likely to receive further doses of gadolinium throughout their lifetimes.
Relative T 1 hyperintensity in the DN has also been attributed to the treatment effects of chemotherapy or radiation therapy. 13 A recent study reported changes in SI in the DN and GP in children with brain tumors undergoing brain irradiation that are independent of the administration of GBCA, 14 and another suggests increased R1 values in adults undergoing brain irradiation may increase susceptibility to gadolinium deposition. 15 The primary hypothesis for this study is that the trajectory of progressive changes in relative DN and GP SI in children receiving multiple doses of GBCA for brain tumor follow-up will change when the child is switched from a linear to a macrocyclic GBCA. Testing of this hypothesis by retrospective analysis was enabled by a change in institutional GBCA administration policy. Until 2009, pediatric patients at our institution were administered the linear GBCA gadopentetate dimeglumine (Magnevist, Bayer Healthcare, Berlin, Germany) because this was the only contrast agent licensed for use in young children in the UK. In 2009, the macrocyclic agent gadobutrol (Gadovist, Bayer Healthcare) was licensed in the UK for children over 7 years old. In 2012, it was licensed for children of 2 years old and above and in 2015 it was licensed for children less than 2 years of age. 16 Our institution followed these timelines so that where contrast agent was administered the patient would have received either linear or macrocyclic GBCA, depending on the date of the scan and the age of the child. Over the period of the retrospective analysis the MR imagers used for scanning pediatric patients and the basic MRI sequence protocol for pediatric brain tumor evaluation remained unchanged.
Given the uncertainty about the relationship between treatment effects on deep gray matter SI in children receiving multiple doses of GBCA, we tested a secondary hypothesis that children receiving chemotherapy and/or radiation therapy will show a steeper trajectory of relative T 1 -weighted (T 1 W) signal increases in the DN and GP than children receiving no active treatment or surgery alone. An exploratory analysis of the effect of GBCA type (linear or macrocyclic) on relative T 1 W signal trajectories for these two patient groups was also conducted.

Materials and Methods
Consent was obtained from the parent/guardian of all participants included in this study for their clinical data to be collated and analyzed as part of an ongoing research study approved by the UK National Health Service Research Ethics Committee (04/MRE04/41).
Participants were pediatric brain tumor patients treated at the Nottingham University NHS Trust and imaged between 2006 and 2017. Participants were included in the analysis if they had undergone MRI with six doses or more of either linear GBCA, macrocyclic GBCA, or both, with an MRI protocol that included an unenhanced axial T 1 WI sequence and were under 18 years of age at the time of their first scan. Potential participants were excluded if they had a condition known to be associated with abnormal DN or GP signal on MRI; if they had brain lesions or surgical resection involving the DN or GP; or if they had previously undergone MRI outside our institution with a GBCA that could not be identified. Individual MR scans were excluded from the analysis if they originated from other institutions, did not include an unenhanced axial T 1 WI, or if degraded by either patient motion or flow artifacts obscuring the anatomical structure of interest. A history of chemotherapy or radiation therapy administration was obtained from the clinical notes. Screening of the patients prior to their MRI scan should have ensured that none of the patients had renal impairment; however, nuclear medicine and blood test results were also checked. Medical records were checked to identify any other reasons for exclusion as were their series of MRI scans.
Thirty-five patients were analyzed in total. For the purposes of analysis, three participant subgroups were defined based on GBCA administration: 1) The "linear-only" group included all those who received six or more doses of gadopentetate dimeglumine, but the analyses referring to this group included only the SI ratios prior to their first dose of gadobutrol (if given); 2) the "macrocyclic-only group" included participants who only received gadobutrol (ie, no exposure to gadopentetate dimeglumine) and received six or more doses of gadobutrol; 3) the "switchover group" included those who received six or more doses of gadopentetate dimeglumine followed by three or more doses of gadobutrol (and could therefore include patients already in the linear-only group). Evolution of the different groups is described in Fig. 1.

MRI Protocol
Scanning was performed on three scanners: an Intera 1.5T (Philips Healthcare, Best, The Netherlands); an Achieva 3T (Philips Healthcare), and a Signa Excite 1.5T (GE Healthcare, Chicago, IL). Our institution uses a set protocol which was employed to scan all pediatric brain tumor patients throughout the study period to maintain consistent parameters. Parameters for the axial T 1 WI spin echo sequence used for analysis are shown in Table 1. The GBCA dose was calculated strictly as per the manufacturers' recommendation: 0.2 ml (0.002 mmol/ml) per kg of weight for gadopentetate dimeglumine and 0.1 ml (1.0 mmol/ml) per kg for gadobutrol. The GBCA type and dose at each timepoint were recorded in and collected from our institution's radiology information system.

Image Analysis
Image viewing and analysis was performed using the GE Centricity Universal Viewer V6.0 PACS system (GE Healthcare). The axial T 1 WI sets were scored for quality using a scoring range of 0-3. Images that were scored 0 were rejected from analysis. A second independent reviewer (neuroradiologist with 14 years' experience, R.D.) reviewed all included image slices to confirm that the images were appropriate for inclusion in the analysis (see Supplementary Material for examples of image quality grading). Regions of interest (ROIs) were drawn manually by S.R. on the unenhanced axial T 1 WI, supplemented by reference to other T 2 WI to aid structure identification if required. The middle cerebellar peduncle (MCP) and the cerebral white matter (CWM) were selected as control regions for ratio calculations based on the relative lack of evidence of gadolinium deposition in white matter compared with deep gray matter. [17][18][19] An elliptical ROI was placed on the DN and GP. ROIs were standardized to between 0.3 and 0.6 cm in size depending on the size of the anatomy. Training and supervision by a neuroradiologist with 30 years' experience, T.J., assisted in the identification of the individual structures and ensured accurate placement of ROIs. An example of the ROIs drawn can be seen in Fig. 2.  Ratios of mean SI were calculated for the DN relative to the MCP (DN/MCP) and for the GP relative to the CWM (GP/CWM). Intensity ratios were used to reduce confounding effects from the variation in data acquisition across different scanners and field strengths. For each subject, the ratios were normalized to their respective ratio at the first scan, and the gradient of the relative SI ratio slope as the number of doses increased was calculated from linear regression (Microsoft Excel 2010, Redmond, WA).
For the switchover group, ratios were calculated for before and after contrast agent switch. Gadopentetate dimeglumine ratios at each timepoint were normalized by their first scan values, and gadobutrol ratios by the last unenhanced T 1 WI, which was acquired immediately prior to the first dose of gadobutrol.

Statistical Analysis
Statistical analysis was performed using SPSS v. 21 (IBM SPSS Statistics for Windows, Armonk, NY). A one-sample t-test or Mann-Whitney U-test, depending on normality, were performed to test for slope gradients differing from zero. Group comparisons of SI ratio slope gradients were made using independent samples t-tests. In the switchover group, differences between slopes before and after changing the contrast agent were compared using paired sample ttests. P < 0.05 was considered statistically significant.

Results
Of the 65 patients identified as suitable subjects, whose parents or caretakers had consented to data collection, 35 individuals fulfilled the inclusion criteria for this analysis and 28 were excluded based on the exclusion criteria. A further two potential participants had individual scans without unenhanced T 1 WIs, which put them below the required number of scans for inclusion. Out of the 35 patients selected, it was possible to analyze both the DN and GP in 18 subjects, only DN in eight subjects and only GP in nine subjects. This was because some had lesions/surgery in one area but not the other. Details of the 35 included participants are shown in Table 2.
A subgroup 18 patients were scanned three or more times each at 1.5T and 3T using the same contrast agent. The agreement of ratios calculated from the 1.5T images compared with those calculated from 3T images only can be seen in the Bland-Altman plot of Fig. 3.
For the whole group, positive slope gradients were observed with the distribution of slope gradients varying significantly from "no slope" for both DN/MCP (mean AE standard deviation [SD] = 0.002 AE 0.005, P = 0.026, t = 2.360) and GP/CWM, (mean AE SD = 0.004 AE 0.005 P < 0.005, t = 4.271. The mean percentage increase in SI per dose across the group was 0.013% AE 0.043 (SD) for the DN/MCP ratio and 0.027% AE 0.041 for the GP/CWM ratio.
The data were then split into the "linear only" and "macrocyclic only" groups. Analysis of the linear group showed positive slope gradients with the distribution of slope gradients varying significantly from "no slope" for both DN/MCP (mean AE SD = 0.007 AE 0.004, P < 0.005, t = 8.618) and GP/CWM (P < 0.005, t = 6.07) (see Fig. 4a for an example of a   linear group patient). The mean percentage increase per dose in SI across the group was 0.063% AE 0.044 for the DN/MCP ratio and 0.078% AE 0.082 for the GP/CWM ratio. In contrast, the macrocyclic group showed no clear slope, confirmed by the lack of a significant difference from the "noslope" condition in the one-sample t-test for either DN/MCP (P = 0.104, t = -1.870) or GP/CWM (P = 0.546, t = 0.628) (see Fig. 4b for an example of a macrocyclic group patient). The mean percentage increase per dose in SI across the group was -0.034% AE 0.051 for the DN/MCP ratio and 0.004% AE 0.040 for the GP/CWM ratio.
For the whole group, an independent samples t-test demonstrated no significant difference in the T 1 WI SI ratio slope gradient between those who did and those who did not receive chemotherapy or radiation therapy, for either DN/MCP or GP/CWM ratios (DN/MCP P = 0.740, t = -0.337; GP/CWM P = 0.694, t = 0.399). The percentage increase per dose for the whole group was 0.011% AE 0.043 for the DN/CWM ratio and 0.024% AE 0.043 for the GP/CWM ratio. When the analysis was repeated for the linear group separately, the independent Mann-Whitney U-test again showed no significant difference between those who did and those who did not receive chemotherapy and/or radiation therapy (linear group: DN/MCP P = 0.246, GP/CWM P = 0.733; macrocyclic group: DN/MCP P = 0.385, GP/CWM P = 0.703). For the patients who received no therapy and were in the linear group, the percentage increase per dose was 0.092% AE 0.035 for the DN/CWM and 0.137% AE 0.113 for the GP/CWM ratio. Those who had received therapy and were in the linear group showed a percentage increase per dose of 0.044% AE 0.039 for the DN/CWM ratio and 0.045% AE 0.030 for the GP/CWM ratios.
For the switchover group, the paired samples t-test showed a significant difference in the distribution of slope gradients before and after the switch from linear to macrocyclic for DN/MCP (P = 0.014), with the plot demonstrating an apparent reversal from increasing to decreasing trajectory of ratios at the point of switch (Fig. 5a). No significant difference was seen in the distribution of slope gradients before and after the switch for GP/CWM (P = 0.680); however, inspection of the plot reveals a plateauing of the trajectory of ratios at the point of the switch (Fig. 5b) but a noticeably larger standard deviation.

Discussion
We hypothesized that the trajectory of progressive changes in relative DN and GP SI in children receiving multiple doses of GBCA will change when switching from a linear to a macrocyclic GBCA. This hypothesis is supported with regard to the DN/MCP ratios, where a significant difference in slope gradients was identified in children before and after the switchover. Two other studies in adults have evaluated serial linear and subsequent macrocyclic doses in the same patients with similar findings. 20,21 Although no statistically significant change in trajectory was found for the GP/CWM ratios before and after the switchover, the plot of mean values shows a plateauing of the trajectory after the switch to macrocyclic GBCA.
Our data for the separate linear and macrocyclic groups also demonstrates different trajectories of signal change; the linear group showed positive slope gradients, in keeping with presumed gadolinium deposition at both sites (DN and GP), whereas the macrocyclic group did not. The findings are in line with the majority of the recent literature identifying an increase in T 1 W hyperintensity in the GP and DN after serial doses of linear GBCAs but not macrocyclic GBCAs. 2,[4][5][6][7]11,13,22,23 A notable finding was the negative slope gradient for DN/MCP ratios in the switchover group after the change to macrocyclic GBCA. This finding agrees with the two studies in adults that made use of a change in the type of contrast agent. 20,21 It is difficult to specify the reason for this decline. It has been demonstrated that macrocyclic GBCAs also deposit in the brain but to a lesser extent than linear agents, 24 and a few studies have reliably demonstrated SI ratio increases in the macrocyclic group. 25 Studies in rats demonstrated that gadolinium found in the rat brain after linear and macrocyclic contrast administration was in three distinct forms: intact GBCA, soluble macromolecules, and in insoluble form, the intact GBCA being identified as macrocyclic contrast. 26,27 This suggests that the reason for the decline in SI ratio therefore could relate to a steady washout of the linear gadolinium, which has been demonstrated in rats. 28 Changes to the binding of the gadolinium, dechelation and precipitation of the gadolinium is another consideration. 29 Alternatively, this decline in ratio values could be due to a disproportionate increase in SI of the control ROI used in the ratio calculation, the MCP. This latter point highlights a problem with much of the existing literature that uses SI ratios; ie, the assumption that the control ROI is not subject to gadolinium deposition. For example, several previous studies use the pons as a control ROI for the DN, 2,7,8,10,11,17,23 but it is recognized that the pons, which contains many gray matter nuclei, is a site of gadolinium deposition itself. 13 Our data do not allow isolation of the separate effects of potential gadolinium clearance from the DN and gadolinium deposition in the MCP to be disentangled, for which quantitative T 1 -mapping would be required.
The values for DN/MCP and GP/CWM increased with GBCA dose number for our pediatric cohort as a whole, which broadly confirms the findings of other groups studying pediatric populations. 3,4,6,[8][9][10][11][12][13][14] However, we found no significant difference in the trajectory of SI ratio increase between children treated with chemotherapy and radiation therapy and those who received no treatment or surgery alone. These findings are in line with another study in adults and children that found that radiation therapy had no impact on SI change, 5 but conflict with a recent study showing that SI changes in the DN and GP occur in patients with brain tumors undergoing brain irradiation independent of GBCA administration. 14 Our analysis is limited by a relatively small sample size, particularly when the sample is split into groups for analysis. It would not be ethical to conduct a prospective trial testing the relative gadolinium deposition rates in children following multiple GBCA doses, and hence we were limited to conducting a retrospective analysis. One strength of our study is that during the study period (2006 to 2017) the MRI scanners used at our institution for pediatric brain tumor evaluation, and the basic MRI sequence protocol, did not change, which provides continuity in the dataset. Although patients were scanned on both 1.5T and 3T scanners, the use of SI ratios removes confounding effects.
To date, brain deposition of gadolinium has not been shown to be harmful in the majority of patients and the longterm consequences are unknown. Our data do not allow us to explore the consequences of presumed gadolinium deposition for the children involved. Apart from the fact that we do not have quantitative cognitive or neurological function scores for these children, this group is likely to carry a significant neurological and cognitive burden due to their complex neurosurgical and treatment histories. In addition, maturation of brain structure and function occurs throughout childhood, and hence gadolinium deposition may affect the developing brain differently from a mature adult brain. However, it is important that the consequences of gadolinium deposition are studied prospectively, as detrimental effects could be greater in the developing brain that is undergoing rapid FIGURE 5: Group mean values (n = 18) for each dose and fitted slopes from the switchover group for the relative (a) dentate nucleus (DN) / middle cerebellar peduncle (MCP) and (b) globus pallidus (GP) / cerebral white matter (CWM) T 1 signal intensity ratios before (dashed line) and after (dash-dot line) the switch from linear to macrocyclic. For each patient, the intensity ratios of the linear group are normalized by the ratio of the last linear contrast agent scan, and macrocyclic ratios by the ratio of their first macrocyclic scan, so that both those scans are plotted at (0,1). The gray area represents the standard deviation around the mean values. structural and functional maturation compared with the mature adult brain.
In conclusion, our analyses confirm previous reports that a T 1 signal increase occurs in the deep gray matter following repeated doses of linear GBCA but not macrocyclic agents. Furthermore, we demonstrate that switching from linear to macrocyclic GBCAs halts the relative T 1 signal increase in deep gray matter of children with brain tumors. We found no evidence of an independent effect of chemotherapy or radiation therapy on the trajectory of T 1 signal increase.