ORIGINAL RESEARCH

Neurotoxicity in breast cancer survivors ≥10 years post-treatmentis dependent on treatment type

Myrle M. Stouten-Kemperman & Michiel B. de Ruiter &

Vincent Koppelmans & Willem Boogerd &

Liesbeth Reneman & Sanne B. Schagen

# Springer Science+Business Media New York 2014

Abstract Adjuvant chemotherapy (CT) for breast cancer(BC) is associated with very late side-effects on brain functionand structure. However, little is known about neurotoxicity ofspecific treatment regimens. To compare neurotoxicity pro-files after different treatment strategies, we usedneurocognitive testing and multimodality MRI in BC survi-vors randomized to high-dose (HI), conventional-dose(CON-) CT or radiotherapy (RT) only and a healthy control(HC) group. BC survivors who received CON-CT (n=20) andHC (n=20) were assessed using a neurocognitive test batteryand multimodality MRI including 3D-T1, Diffusion TensorImaging (DTI) and 1H-MR spectroscopy (1H-MRS) to mea-sure various aspects of cerebral white (WM) and gray matter(GM). Data were compared to previously assessed groups ofBC survivors who received HI-CT (n=17) and RT-only (n=15). Testing took place on average 11.5 years post-CT. 3D-T1showed focal GM volume reductions both for HI-CT andCON-CT compared to RT-only (p<.004). DTI-derived meandiffusivity and 1H-MRS derived N-acetyl aspartate showedWM injury specific to HI-CT but not CON-CT (p<.05).Residual effects were revealed in the RT-only group compared

to HC on MRI and neurocognitive measurements (p<.05).Ten years after adjuvant CT for BC lower cerebral GM vol-ume was found in HI as well as CON-CT BC survivorswhereas injury to WM is restricted to HI-CT. This mightindicate that WM brain changes after BC treatment may showmore pronounced (partial) recovery than GM. Furthermore,our results suggest residual neurotoxicity in the RT-onlygroup, which warrants further investigation.

Key words Breast Cancer .MRI . Chemotherapy .

Cognition . Late effects

Introduction

Advances in the treatment of breast cancer (BC) have led to agrowing number of long-term BC survivors. They frequentlyreport cognitive problems, particularly after adjuvant chemo-therapy (CT) (Pullens, De Vries, and Roukema 2010). Cross-sectional and longitudinal neurocognitive studies confirm the-se cognitive problems in a subgroup of BC survivors (13–70 %), with the domains most commonly affected beingprocessing speed, memory, and executive function (Wefeland Schagen 2012; Wefel, Vardy, Ahles, and Schagen 2011).

Neuroimaging studies have reported deterioration of fron-tal and temporal white matter (WM) microstructure (Deprez,Billiet, Sunaert, and Leemans 2013) and focal and overall graymatter (GM) volume reductions (Pomykala, de Ruiter,Deprez,McDonald, and Silverman 2013) within a fewmonthsto 3 years after completion of modern, anthracycline-basedCT. However, studies on very late effects (≥10y post-treatment) of anthracycline-based CT are non-existent.

In our previous study, we found detrimental effects oncognition and brain structure in BC survivors who receivedadjuvant high-dose CT (HI-CT) 10 years earlier, compared toBC survivors who received radiotherapy (RT-only) (de Ruiter

M. M. Stouten-Kemperman :M. B. de Ruiter :V. Koppelmans :S. B. Schagen (*)Netherlands Cancer Institute/Antoni van Leeuwenhoek Hospital,Plesmanlaan 121, 1066 CX Amsterdam, The Netherlandse-mail: s.schagen@nki.nl

M. M. Stouten-Kemperman :M. B. de Ruiter : L. RenemanDepartment of Radiology Academic Medical Center, University ofAmsterdam, Amsterdam, The Netherlands

V. KoppelmansNeuromotor Behavior Laboratory School of Kinesiology, Universityof Michigan, Ann Arbor, MI, USA

W. BoogerdDepartment of Neuro-OncologyNetherlands, Cancer Institute/Antonivan Leeuwenhoek Hospital, Amsterdam, The Netherlands

Brain Imaging and BehaviorDOI 10.1007/s11682-014-9305-0

et al. 2011, 2012). Here, we compared our previous findings inHi-CT with an additional group of patients exposed to afrequently used anthracycline-based CT (conventional-doseCT; CON-CT). Both patient groups previously participatedin a trial in which they were randomly assigned to eitheradjuvant HI-CT or CON-CT. We chose to compare thesecytotoxic regimens, with the advantage of using homoge-neous BC patient groups in terms of disease grade and diseasestage. In addition, we evaluated the two different CT regimensto a cancer-specific (RT-only) control group.

At last, we included a group of women without a history ofcancer to compare to the RT-only group, as recent studiessuggest some level of impaired cognitive function in breastcancer survivors not exposed to CT. We expected very latedetrimental effects of CT on brain structure and cognitivefunctioning with more severe effects of HI-CT than CON-CT.

Methods

Participants

All BC survivors were recruited from the Netherlands CancerInstitute, VU University Medical Center, Leiden UniversityMedical Center and the Erasmus University Medical Center-Daniel den Hoed Cancer Center. The study was approved bythe review board of the Netherlands Cancer institute, whichserved as the central ethical committee for all participatinghospitals. Written informed consent was obtained from allparticipants.

All BC survivors who received CTwere diagnosed as high-risk patients and participated in a multicenter randomized trialcomparing the efficacy of HI-CT to CON-CT (Rodenhuiset al. 2003) (details on disease stage and CT regimen areprovided in Table 1). BC survivors who did not require CThad undergone locoregional surgery and RT. All participantswere evaluated in an earlier neuropsychological study fromour group (Schagen et al. 2006). At that time HCs wererecruited among female friends and family of BC survivors.The study inclusion/exclusion criteria have been publishedpreviously (de Ruiter et al. 2011). The total sample consistedof 19 HI-CT, 24 CON-CT, 15 RT-only BC survivors and 27HC. Figure 1 gives an overview of participant inclusion. Fourof the CON-CT and 7 of the HC group declined MRI assess-ment. MRI data was also not available for 2 patients of the HI-CT group (see De Ruiter et al. 2012 for detailed information).Therefore,MRI data were available in 17HI-CT, 20 CON-CT,15 RT-only survivors and 20 HCs.

Procedure of the assessment

To assess symptoms of anxiety and depression, the HopkinsSymptoms Checklist-25 (HSCL) (Hesbacher, Rickels,Morris,

Newman, and Rosenfeld 1980) was used. The EuropeanOrganisation for Research and Treatment of Cancer(EORTC) Quality of Life Questionnaire-C30 (Aaronsonet al. 1993) was used to assess health-related quality of life.Furthermore, we used the same battery of 7 neuropsycholog-ical tests (yielding 16 test indices) as was used in our earlierstudy (de Ruiter et al. 2011). After the neuropsychologicalassessment the MRI scanning session took place. The totalexperimental procedure lasted ~2.5 h per participant.

MR Imaging and data processing

All participants were scanned on a 3.0 Tesla Intera MRIscanner (Philips Medical Systems, Best, The Netherlands).Whole brain Diffusion Tensor Imaging (DTI), Fluid Attenu-ated Inversion Recovery (FLAIR), T1-weighted 3D spoiledgradient echo and single voxel 1H-MRS point-resolved

Table 1 Information on disease stage and cytotoxic regimen of studypopulation

HI-CT CON-CT RT-only HC(n=19) (n=24) (n=15) (n=27)

Disease stage > Stage 1a > Stage 1a Stage 1 N/A

FEC regimen

5-Fluorouracil 500 mg/m2 500 mg/m2 N/A N/A

Epirubicin 90 mg/m2 90 mg/m2 N/A N/A

Cyclophosphamide 500 mg/m2 500 mg/m2 N/A N/A

Number of cycles 4 5 N/A N/A

CTC regimen

Cyclophosphamide 6 g/m2 N/A N/A N/A

Thiotepa 480 mg/m2 N/A N/A N/A

Carboplatin 1.6 g/m2 N/A N/A N/A

Number of cycles 1 N/A N/A N/A

Tamoxifen treatment yes yes Nob N/A

HI-CT: high-dose chemotherapy (CT); CON-CT: conventional-dose CT;RT-only: radiotherapy-only, HC: healthy controls. Both CON-CTand HI-CTwere followed by radiotherapy and 2–5 years of tamoxifen treatment.aWith at least four axillary lymph nodes with metastases but withoutdistant metastases. b One patient from the RT-only group was treated withtamoxifen for 5 years

Participants Eligible

Total participating

Could not bereached/declined

to participate

MRI dataavailable

RT-onlyHI-CT CON-CT HC

Fig. 1 Flowchart of participant attrition. HI-CT: high-dose chemothera-py (CT); CON-CT: conventional-dose CT; RT-only: radiotherapy-only,HC: healthy controls

Brain Imaging and Behavior

spectroscopy (PRESS) were acquired using the same param-eters as described in our previous study (de Ruiter et al. 2011,2012).

DTI was analyzed with a standard processing pipelinewithin the FMRIB Diffusion Toolbox (FDT), part of FSL4.1 (Jenkinson, Beckmann, Behrens, Woolrich, and Smith2012). First, eddy-current induced morphing was correctedby affine registration of the diffusion-weighted images to theaverage b0 image. Then a diffusion tensor model was fit to thedata to generate fractional anisotropy (FA) and mean diffusiv-ity (MD) maps. Subsequently, tract-based spatial statistics(TBSS) was performed to warp all FA images to a study-specific template and create individual FA skeletons and co-registered MDmaps. For VBM, all T1-weighted images werebrain-extracted using BET (Smith 2002). Next, tissue-typesegmentation was carried out using FAST (Zhang, Brady,and Smith 2001). GM segmentations were used to create astudy-specific GM template to which all GM images werenon-linearly normalized. All warped images were modulatedby dividing each voxel by the Jacobian of the warp field andthen smoothed with an isotropic Gaussian kernel with a sigmaof 3 mm. Spectra were extracted using LCModel (Provencher1993). Due to time constraints, spectra were not acquired for 1patient of the RT group, 3 patients of the CON-CT group and1 participant of the HC group.

Statistical analysis

Demographic variables, self report measures, neuropsycho-logical test scores and MR spectra were analyzed with IBMSPSS Statistics 20 (IBM, Armonk, NY), by means ofANOVA, Fazekas ratings by means of a χ2 test. We evaluatedcognitive status using two approaches. First, we identifiedcognitive impairment based on frequently used cut-off scores(see previous studies for detailed description of this method,(Schagen et al. 2006) and tested for differences in proportionof impaired patients using logistic regression with age and IQas covariates.

Second, we calculated a distance score for each patient: theMahanalobis (MH) distance, based on means and variances ofthe HC group. An advantage of theMH distance is that it takescorrelations between tests into account and captures smallercognitive deviations that would be missed with dichotomizedcut-off scores (DeCarlo 1997; Koppelmans et al. 2012a,2012b). MH distance was log2 transformed and subsequentlybetween group differences were tested withMann–Whitney Unon-parametric t-tests. For all group analyses our contrastswere: 1) HI-CT vs. CON-CT, 2) HI-CT vs. RT-only, 3) CON-CT vs. RT-only and 4) RT-only vs. HC.

For exploratory purposes, we converted each raw neuro-psychological test score into a standard z-score by using meantest scores of the HC group as references. These Z-scores wereaveraged for six different neuropsychological domains and

analyzed with MANCOVA with age and IQ as covariates.Although time between surgery and assessment differed sig-nificantly between patient groups, we did not incorporate thisin our analyses because of non-overlapping ranges of thisvariable between groups (HI-CT: 8.8–11.4 years; CON-CT12.1–14.9 years; RT-only 8.3–9.8 years).

The mean of all WM DTI values (FA, MD) across theskeleton, and GM and WM as percentage of intracranialvolume were calculated. Group differences were tested inSPSS with ANCOVA, with age as covariate. Focal groupdifferences in WM microstructure (DTI) and GM volume(VBM) were analyzed with voxel-wise t-tests (corrected at acluster level threshold p<.05) with age (and intracranial vol-ume for VBM) entered as covariate(s). These analyses arebased on a permutation-based inference method for nonpara-metric statistical thresholding that corrects for multiple com-parisons by using the null distribution of the maximum (acrossthe image) cluster size (Nichols and Holmes 2002).

For correlation analyses we chose to limit the number ofstatistical tests, thereby restricting our analyses to measuresthat were more sensitive in detecting group differences. Voxel-based correlations between focal MD and MH distance/1H-MRS were calculated in FSL. Non voxel-based correlationsacross groups were calculated in SPSS. We correlated neuro-psychological test performance (classical cut-off score, MHdistance, domain scores) with self-reported measures (self-reported cognitive functioning, anxiety, depression) andwhole brain measures (percentage GM/WM, whole-brainMD, Fazekas ratings). Further, we correlated self-reportedmeasures with each other. An additional correlational analysiswithin groups was performed between time since treatmentand whole-brain MD, GM/WM percentage, Fazekas score,1H-MRS and MH distance score. For all analyses only sig-nificant associations larger than .25 were considered mean-ingful. Alpha levels were set at p=0.05 for all analyses. For allgroup analyses our contrasts were: 1) HI-CT vs. CON-CT, 2)HI-CT vs. RT-only, 3) CON-CT vs. RT-only and 4) RT-onlyvs. HC.

Results

Demographic and clinical data

Table 2 presents the characteristics of all participants, patient-related outcomes and neurocognitive assessment scores. Nosignificant differences were found between groups on age andestimated premorbid IQ score. There was a significant overalldifference between groups on time interval between surgeryand assessment (F3,85=222.84, p<.001), reflecting earlier re-cruitment of the HI-CT and RT-only group for our previousstudies (de Ruiter et al. 2011, 2012). No significant differenceswere found between groups on measures of quality of life,

Brain Imaging and Behavior

depression and anxiety. Analyses of self-reported cognitivefunctioning (as measured with the EORTC subscale ‘cogni-tive functioning’) showed that the RT-only group scored sig-nificantly lower on this measure than the HC group (F1,42=7.10, p=.011).

Neuropsychological tests

Group differences in the proportion of impaired participants(using predefined cut-off scores) were not significant. How-ever, the HI-CT group comprised the highest number ofimpaired patients, followed by the CON-CT group and theRT-only group. This pattern was also found in the MH dis-tance score. Pairwise comparisons showed that the MH dis-tance score was significantly higher for the RT-only groupthan for the HC (U =.113, p=0.019), indicating worse overallcognitive performance for the RT-only group. At the cognitivedomain level, verbal memory was significantly lower for theHI-CT group than the CON-CT group (F1,32=11.22, p=.002)and lower for the RT group than the CON-CT group (F1,39=6.52, p=.006) and the HC (F1,42=14.08, p<.001).

VBM

No differences were found in global GM and WM percentagebetween groups (Table 3). VBM showed lower GMvolume inthe HI-CT group compared to the RT-only group in posteriorbrain areas, including cerebellum, occipital cortex, posteriorparietal cortex and precuneus. In the CON-CT group com-pared to the RT-only group, GM was lower in the occipitalcortex and cerebellum. Higher GM volume was found in theRT-only group compared to the HC group in the cerebellum,occipital cortex, cingulum, calcarine sulcus and precuneus(Table 4, Fig. 2).

1H-MRS and FLAIR white matter hyperintensities

The HI-CT group showed a significantly lower (NAA +NAAG)/Cr ratio compared to the CON-CT (F1,36=6.02,p=.020) and RT-only group (F1,31=7.46, p=.011) (Table 3,Fig. 3). The Fazekas ratings (all between 0–2) of WMhyperintensities did not differ significantly between groups(Table 3).

Table 2 Demographic and clinical characteristics of the study population and summary of neurocognitive assessment

HI-CT CON-CT RT-only HC

Age 56.3 (5.5) 59.8 (6.3) 58.2 (5.8) 60.31 (4.8)

Estimated IQ (NART) 101.1 (17.9) 100.6 (13.1) 100.7 (17.3) 108.6 (14.1)

Years since surgery*a 9.9 (0.5) 13.51 (0.7) 9.2 (0.5) N/A

Years since chemotherapy*b 9.5 (0.8) 13.42 (0.7) N/A N/A

EORTC QLQ-C30

Global quality of life 82.0 (12.1) 84.7 (13.4) 81.1 (16.2) 88.3 (13.3)

Cognitive functioning*c 77.2 (19.4) 80.6 (25.9) 72.2 (21.5) 86.4 (13.1)

Physical functioning 83.5 (12.2) 87.5 (12.2) 88.4 (11.7) 91.6 (9.0)

Fatigue 25.7 (14.4) 20.4 (20.5) 23.0 (19.4) 16.0 (15.5)

HSCL-25 total score 11.3 (6.4) 11.8 (8.8) 14.8 (16.3) 12.6 (16.1)

HSCL depression 11.7 (6.4) 11.4 (9.9) 15.7 (19.2) 15.7 (24.5)

HSCL anxiety 10.7 (6.3) 11.8 (9.1) 13.6 (13.5) 7.9 (7.0)

Neurocognitive domains (z-scores)

Verbal memory*d −1.1 (1.2) −0.3 (0.9) −1.1 (1.1) 0.02 (0.7)

Verbal fluency −0.1 (0.8) −0.03 (0.7) 0.2 (0.8) 0 (0.9)

Visual memory −0.2 (1.1) −0.3 (1.2) 0.1 (0.7) 0 (0.9)

Attention −0.2 (0.7) −0.2 (0.9) −0.1 (0.8) 0 (0.8)

Executive functioning −0.5 (1.1) −0.4 (0.9) −0.1 (0.7) 0 (0.8)

Motor speed −0.4 (0.8) −0.4 (1.0) −0.3 (0.8) 0 (1.0)

Cognitive impairment, 5 (26.3 %) 3 (12.5 %) 0 1 (3.7 %)

number of patients impaired

Mahanalobis distance score (MhD)*e 41.7 (36.9) 37.2 (45.4) 30.4 (25.0) 10.6 (6.1)

HI-CT: high-dose chemotherapy (CT); CON-CT: conventional-dose CT; RT-only: radiotherapy-only, HC: healthy controls. Values indicate mean (SD)unless specified otherwise. * p<.05. EORTC QLQ-C30, European Organization for Research and Treatment of Cancer health-related Quality-of-LifeQuestionnaire (a higher score indicates better functioning, except for fatigue); HSCL-25, Hopkins Symptom Checklista HI-CT<CON-CT; HI-CT>RT; CON-CT>RT b HI-CT<CON-CT c RT<HC d HI-CT<CON-CT; CON-CT>RT; RT<HC e RT>HC

Brain Imaging and Behavior

DTI

A significantly higher MD across the WM skeleton in the HI-CTcompared to the CON-CT group (F1,36=10.69, p=.003) andRT-only group (F1,32=7.86, p=.009) was found. No significantdifferences in mean FA values were found between groups(Table 3). TBSS revealed significantly higher MD values inthe HI-CT group compared to the CON-CT group in the bodyand genu of the corpus callosum, anterior and superior coronaradiata, external and internal capsule, sagittal striatum and su-perior longitudinal fasciculus (Table 4, Fig. 4). Further, the HI-CT group had significantly larger MD values than the RT-onlygroup in the body and genu of the corpus callosum, cingulum,posterior and superior corona radiata, external and internalcapsule, superior fronto-occipital/longitudinal fasciculus. Final-ly, a higher MDwas also found in the RT-only group comparedto the HC group in the superior longitudinal fasciculus.

Relation between neuropsychological test performanceand self-reported measures

All correlations were calculated across groups. No significantassociations were observed between the proportion of im-paired participants/MH distance score and reports of anxiety,depression and self-reported cognitive functioning. However,there was a significant negative association between the atten-tion domain and self-reported cognitive functioning (r=−.30,p=.006). Finally, HSCL total score was negatively associatedwith self reported cognitive functioning (r=−.413, p<.001).

Relation between neuropsychological test performanceand MRI measures

No associations were found between the proportion of impairedparticipants/cognitive domains/MH distance score and total per-centage GM,WM and whole-brain MD across groups. Howev-er, there was a significant negative association between Fazekasscore and executive functioning (r=−.305 p=.009) and verbal

fluency (r=−.278, p=.018). Voxel-based tests did not show anassociation between MH distance and focal MD.

Relation between MD and NAA+NAAG/Cr

To correlate MD values with 1H-MRS data, mean skeleton-ized MD values within the 1H-MRS region of interest (ROI)were calculated. A significant negative correlation was foundbetween MD and (NAA+NAAG)/Cr (r=−.27, p=.026).Voxel-based analyses across groups revealed a significantnegative association between (NAA + NAAG)/CR and MDvalues in the right superior and posterior corona radiata.

Relation between time since treatmentand neuropsychological and MRI measures

No associations were found within groups between time sincetreatment and (NAA+NAAG)/Cr, total percentage of GM/WM andMH distance score.Within the HI-CT group positiveassociations were found between time since CT and whole-brain MD (r=−.27, p=.026) and Fazekas score (r=−.27,p=.026). Adjusting for the effect of age, we found that withinthe HI-CT group, time since CT was a significant positivepredictor for MD (β=.54, t (16)=2.18, p=.05), and Fazekasscore (β=.49, t (16)=2.29, p=.04).

Discussion

This is the first study to show that very late neurotoxicity(≥10 years) of adjuvant CT for BC depends on the specificcytotoxic regimen administered, as indicated byneurocognitive testing and multimodality MRI. Further, ourdata suggest residual cognitive and MRI differences betweenRT-only BC survivors and HC.

Our results suggest that HI-CT is more neurotoxic thanCON-CT. First, the two compound scores for overall cogni-tive impairment indicated numerical worse cognitive

Table 3 Gray and white matter and 1H-MRS MRI measures

HI-CT CON-CT RT-only HC

(NAA+NAAG)/Cr*a 1.66 (0.13) 1.79 (0.15) 1.80 (0.15) 1.76 (0.11)

Gray matter volume (%) 38.10 (1.57) 37.35 (1.68) 38.79 (1.60) 38.11 (1.26)

White matter volume (%) 38.68 (0.83) 38.85 (0.95) 38.11 (1.41) 38.32 (0.98)

Mean diffusivity (MD; μm/s2)*b 0.750 (0.018) 0.738 (0.024) 0.738 (0.019) 0.723 (0.023)

Fractional anisotropy (FA) 0.467 (0.017) 0.467 (0.018) 0.472 (0.014) 0.475 (0.014)

Fazekas 0 6 (35.3 %) 9 (45 %) 7 (46.7 %) 9 (45 %)

Fazekas 1 9 (52.9 %) 9 (45 %) 6 (40.0 %) 9 (45 %)

Fazekas 2 2 (11.8 %) 2 (10 %) 2 (13.3 %) 2 (10 %)

HI-CT: high-dose chemotherapy (CT); CON-CT: conventional-dose CT; RT-only: radiotherapy-only, HC: healthy controls. Values indicate mean (SD)unless specified otherwise. * p<.05.a HI-CT<Con-CT; HI-CT<RTb HI-CT>Con-CT; HI-CT>RT

Brain Imaging and Behavior

functioning after HI-CT than after CON-CT (26.3 % vs.12.5 %), although this was not statistically significant at thep<.05 level. Second, most MRI measures revealed moresevere neurotoxicity after HI-CT than after CON-CT. In

accordance with Brown et al. (Brown et al. 1998), HI-CTshowed a relative reduction in NAA in left cerebral WM.Further, lower global WM integrity was found in the HI-CTgroup compared to the CON-CT group and occurred focally

Table 4 MRI between group analyses for voxel-based morphometry (VBM) and mean diffusivity (MD) values

MNI coordinates x y z Cluster (vox) t-value

Between group analyses

VBM results

HI-CT<CON-CT No significant clusters

CON-CT<HI-CT No significant clusters

HI-CT<RT-only −12 −60 −68 39016 1.74

−24 −96 −24 29638 2.52

−22 −62 −10 8833 1.71

RT-only<HI-CT

CON-CT<RT-only 16 −80 −56 41684 1.8

−10 −52 −50 468 1.76

−42 −64 −28 385 1.72

−28 −36 −48 281 1.7

RT-only<CON-CT No significant clusters

RT-only<HC No significant clusters

HC<RT-only 18 −72 −60 36791 1.83

−28 −92 −24 2935 1.75

−16 −62 28 469 2.83

−16 −102 −8 37 1.7

24 −62 −34 15 1.77

MD results

HI-CT<CON-CT No significant clusters

CON-CT<HI-CT −40 −7 −29 20690 3.34

−35 −55 13 863 3.75

18 −36 33 14 2.23

−32 −47 21 12 3.09

−20 −35 39 11 2.29

HI-CT<RT-only No significant clusters

RT-only<HI-CT 60 −25 −11 2158 3.98

−45 −13 −24 1937 3.94

−23 −80 11 1583 3.9

−19 32 19 1474 3.33

20 15 −21 1326 4.54

15 24 20 976 4.1

22 38 23 20 2.72

−17 −92 5 13 2.21

CON-CT<RT-only No significant clusters

RT-only<CON-CT No significant clusters

RT-only<HC No significant clusters

HC<RT-only −54 −8 20 1352 4.31

−7 31 9 743 4.27

−24 15 −17 732 3.59

−36 28 27 15 2.97

HI-CT: high-dose chemotherapy (CT); CON-CT: conventional-dose CT; RT-only: radiotherapy-only, HC: healthy controls

Brain Imaging and Behavior

in predominantly frontal brain areas. These findings are con-sistent with a study in CON-CT BC patients 4 months post-treatment (Deprez et al. 2011). In our study, HI-CT hadwidespread effects on association fibers, as well as projectionfibers and commissural fibers, which play a major role incognitive functioning (Schmahmann, Smith, Eichler, andFilley 2008). In contrast to our previous study (de Ruiteret al. 2012), voxel-based tests on FA only showed groupdifferences below our stringent statistical threshold, potential-ly because we used a different eddy-current correction methodthan before (Mangin, Poupon, Clark, Le Bihan, and Bloch2002).

GM findings were less dependent on type of CT. LowerGM volume in posterior brain areas was found after HI-CT aswell as after CON-CT versus RT-only. Further, the CON-CT

group showed lower GM volume in occipital and cerebellarbrain areas when compared to the RT-only group. Thesefindings concur with other functional and structural neuroim-aging studies (de Ruiter and Schagen 2013; Pomykala et al.2013) and may be associated with impairments in cognitivefunctioning (Strick et al. 2009).

Both substance-dependent and dose-dependent factors mayexplain more severe late neurotoxicity after HI-CT than afterCON-CT. In the HI-CT group the fifth cycle of conventionaldose FEC (5-Fluorouracil, Epirubicin, Cyclophosphamide)was replaced by high-dose CTC (Cyclophosphamide, Thiote-pa, Carboplatin) (see Table 1). The twelve-fold increase incyclophosphamide in this last cycle may have caused anincrease in neurotoxicity in the HI-CT vs. CON-CT group.This is in concordance with preclinical studies that haveshown dose-dependency of cyclophosphamide on central ner-vous system measures (Dietrich, Monje, Wefel, and Meyers2008). Additionally, more severe neurotoxicity after HI-CTmight also be due to the (high) dosages of carboplatin and/orthiotepa that were not incorporated in CON-CT. For bothagents preclinical studies have found strong indications forcentral neurotoxicity (Husain, Whitworth, Hazelrigg, andRybak 2003).

Our results also show differences between patients whoonly received RT and HC. On a neuropsychological summarymeasure (MH distance), the RT only group performed worsethan the HC group. This concurs with previous studies thatreported poorer attention and worse executive functioning inRT-only patients compared to HC (Jim et al. 2009; Phillipset al. 2012). Lower white matter integrity in the RT-only groupcompared to the HC group is consistent with the lower thanexpected cognitive performance observed in this group. Un-expectedly, significantly higher local GM volume than in theHC group was found. The observed pattern of MRI andneuropsychological findings of the RT-only group compared

Fig. 2 Group differences for voxel-based morphometry (VBM, P<.05,corrected for multiple comparisons) Red areas indicate clusters wheregray matter volume was significantly reduced. See text for further de-scription of these clusters. MNI Coordinates for cross-sections are indi-cated. HI-CT: high-dose chemotherapy (CT); CON-CT: conventional-dose CT; RT-only: radiotherapy-only, HC: healthy controls

Fig. 3 Group differences for (NAA+NAAG)/Cr values and Mean Diffusivity (MD) values (*P<.05) HI-CT: high-dose chemotherapy (CT); CON-CT:conventional-dose CT; RT-only: radiotherapy-only, HC: healthy controls

Brain Imaging and Behavior

to HCs is not easy to interpret and warrants replication in alarger sample. Biological mechanisms related to cancer historyand locoregional treatments (e.g. dysregulated cytokine re-lease) may play a role (Collado-Hidalgo et al. 2006). However,RT-only patients declined more often (44.8 %) than partici-pants from other groups. Although age and pre-morbid IQ didnot differ between participants and non-participants, it mightbe possible that selection bias or other pre-existing groupcharacteristics are partially reflected in our findings.

In line with previous research, we found few associationsbetween cognitive test results and self-reported measures(Biglia et al. 2012; Shilling and Jenkins 2007). Further, noneof the self-reported measures appeared to be associated withMRI measures. This might partly be due to the cross-sectionaldesign of the present study. A recent prospective study didshow an association between lower focal FA and self-reportedcognitive dysfunction (Deprez et al. 2011). Global and focalDTI-derived MD values across groups correlated negativelywith (NAA+NAAG)/Cr, indicating that a lower integrity ofWM microstructure is associated with lower axonal metabo-lism. More specifically, this reduction in (NAA+NAAG)/Crmay reflect a defect in the myelin maintenance infrastructure,possibly leading to less restricted diffusion within the axonalbundles (Tang et al. 2007).

A limitation of the present study is that all BC survivorswho received CT also received tamoxifen, which has beenassociated with negative effects on cognition and brain struc-ture and function (Eberling, Wu, Tong-Turnbeaugh, andJagust 2004; Schilder et al. 2010). Therefore, we cannotcompletely rule out that the effects in the CT groups areconfounded by the contributory role of tamoxifen. Anotherlimitation is the cross-sectional nature of the study, especiallyregarding MRI data. ‘Finally, some studies have shown aneffect of time since treatment on cognitive measures and brainmeasures (Conroy et al. 2013; Schilder et al. 2009). Althoughtime between surgery/treatment and assessment differed sig-nificantly between patient groups, we could not meaningfullyadjust for this in the analyses because of non-overlappingranges of this variable between groups. We now show thatin the HI-CT group, longer time since treatment is related tolower white matter integrity and more white matter lesions(for comparable findings see Koppelmans et al. 2014). Sincethe HI-CT group was measured shorter after treatment thanthe CON-CT group, this might indicate that our results in factrepresent an underestimation of the true effect of HI-CT onwhite matter measures.

Strength of this study is that BC survivors had been ran-domly assigned to different CT regimens, allowing us todirectly compare cytotoxic regimens unconfounded bypremorbid group differences. Another strength lies in theinclusion of an RT-only group and a HC group, which allowedus to evaluate residual cognitive and MRI effects in BCsurvivors who were not exposed to CT. Further, the use ofcomplementary computations of cognitive impairment and thestringent criteria for cognitive impairment can be consideredas strengths.

To conclude, our findings suggest an association betweenHI-CTand worse cognitive functioning, long-term lower localGM volume, lower WM integrity and lower axonal function,whereas the effects of CON-CTwere less pronounced and lesswidespread. This suggests that even ≥10 years post-treatmentHI-CT is associated with worse cognitive and structural brainoutcomes than CON-CT, which is more comparable to out-comes in RT-only. Further, residual effects were present in RT-only survivors, which might be due to specific patient-relatedor disease effects in this group that warrant further investiga-tion. WMmight be more prone to (partial) recovery than GM,as suggested by findings from our group of global GM reduc-tions in the absence of WM differences in BC survivors>20 years post-treatment (Koppelmans et al. 2014;Koppelmans, de Ruiter, et al. 2012). This emphasizes theimportance of prospective multimodality studies that includecancer patients exposed to CT as well as unexposed patientsand HCs.

Acknowledgments We thank all the participants and colleagues whocontributed to our studies.

Fig. 4 Group differences for mean diffusivity (MD) values analyzedwith tract-based spatial statistics (TBSS, P<.05, corrected for multiplecomparisons). Group differences are overlaid on a fractional anisotropy(FA) map, derived from the study sample. To facilitate visualization,significant clusters are thickened using the tbss_fill script implementedin FSL. Red areas indicate clusters where MD values were significantlyincreased. Transversal slices with z coordinates in MNI 152 space areshown. HI-CT: high-dose chemotherapy (CT); CON-CT: conventional-dose CT; RT-only: radiotherapy-only, HC: healthy controls

Brain Imaging and Behavior

Funding This is a non-industry-sponsored study. This research wassupported by the AMC Medical Research, grant 09.25.229 I 09.03.270.

Conflicts of Interest Myrle M. Stouten-Kemperman, Michiel B. deRuiter, Vincent Koppelmans, Willem Boogerd, Liesbeth Reneman, andSanne B. Schagen report no conflicts of interest.

Informed consent statement All procedures followed were in accor-dance with the ethical standards of the responsible committee on humanexperimentation (institutional and national) and with the Helsinki Decla-ration of 1975, and the applicable revisions at the time of the investiga-tion. Informed consent was obtained from all patients for being includedin the study.

References

Aaronson, N. K., Ahmedzai, S., Bergman, B., Bullinger, M., Cull, A.,Duez, N. J., & de Haes, J. C. (1993). The European Organization forResearch and Treatment of Cancer QLQ-C30: a quality-of-life in-strument for use in international clinical trials in oncology. J NatlCancer Inst, 85(5), 365–76.

Biglia, N., Bounous, V. E., Malabaila, A., Palmisano, D., Torta, D. M. E.,D’Alonzo, M., & Torta, R. (2012). Objective and self-reportedcognitive dysfunction in breast cancer women treated with chemo-therapy: a prospective study. Euro j of cancer care, 21(4), 485–92.doi:10.1111/j.1365-2354.2011.01320.x.

Brown, M. S., Stemmer, S. M., Simon, J. H., Stears, J. C., Jones, R. B.,Cagnoni, P. J., & Sheeder, J. L. (1998).Whitematter disease inducedby high-dose chemotherapy: longitudinal study with MR imagingand proton spectroscopy. AJNR American journal of neuroradiolo-gy, 19(2), 217–21.

Collado-Hidalgo, A., Bower, J. E., Ganz, P. A., Cole, S. W., & Irwin, M.R. (2006). Inflammatory biomarkers for persistent fatigue in breastcancer survivors. Clinical cancer research. j of the AmericanAssociation for Cancer Research, 12(9), 2759–66. doi:10.1158/1078-0432.CCR-05-2398.

Conroy, S. K., McDonald, B. C., Smith, D. J., Moser, L. R., West, J. D.,Kamendulis, L. M., & Saykin, A. J. (2013). Alterations in brainstructure and function in breast cancer survivors: effect of post-chemotherapy interval and relation to oxidative DNA damage.Breast Cancer Res Treat, 137(2), 493–502. doi:10.1007/s10549-012-2385-x.

De Ruiter, M. B., & Schagen, S. B. (2013). Functional MRI studies innon-CNS cancers. Brain imaging and behavior. doi:10.1007/s11682-013-9249-9.

De Ruiter, M. B., Reneman, L., Boogerd,W., Veltman, D. J., van Dam, F.S., Schagen, S. B., & Nederveen, A. J. (2011). Cerebral hypore-sponsiveness and cognitive impairment 10 years after chemotherapyfor breast cancer. Hum Brain Mapp, 32(8), 1206–19. doi:10.1002/hbm.21102.

De Ruiter, M. B., Reneman, L., Boogerd, W., Veltman, D. J., Caan, M.,Douaud, G., & Schagen, S. B. (2012). Late effects of high-doseadjuvant chemotherapy on white and gray matter in breast cancersurvivors: converging results from multimodal magnetic resonanceimaging. Hum Brain Mapp, 33(12), 2971–83. doi:10.1002/hbm.21422.

DeCarlo, L. T. (1997). On the meaning and use of kurtosis. PsycholMethods, 2(3), 292–307. doi:10.1037//1082-989X.2.3.292.

Deprez, S., Amant, F., Yigit, R., Porke, K., Verhoeven, J., Van den Stock,J., & Sunaert, S. (2011). Chemotherapy-induced structural changesin cerebral white matter and its correlation with impaired cognitivefunctioning in breast cancer patients.Hum Brain Mapp, 32(3), 480–93. doi:10.1002/hbm.21033.

Deprez, S., Billiet, T., Sunaert, S., & Leemans, A. (2013). Diffusiontensor MRI of chemotherapy-induced cognitive impairment innon-CNS cancer patients: a review. Brain imaging and behavior.doi:10.1007/s11682-012-9220-1.

Dietrich, J., Monje, M., Wefel, J., & Meyers, C. (2008). Clinical patternsand biological correlates of cognitive dysfunction associated withcancer therapy. Oncologist, 13(12), 1285–95. doi:10.1634/theoncologist.2008-0130.

Eberling, J. L., Wu, C., Tong-Turnbeaugh, R., & Jagust, W. J. (2004).Estrogen- and tamoxifen-associated effects on brain structure andfunction. NeuroImage, 21(1), 364–371. doi:10.1016/j.neuroimage.2003.08.037.

Hesbacher, P. T., Rickels, K., Morris, R. J., Newman, H., & Rosenfeld, H.(1980). Psychiatric illness in family practice. The J of clinicalpsychiatry, 41(1), 6–10.

Husain, K., Whitworth, C., Hazelrigg, S., & Rybak, L. (2003).Carboplatin-Induced Oxidative Injury in Rat Inferior Colliculus.Int J Toxicol, 22(5), 335–342. doi:10.1177/109158180302200502.

Jenkinson, M., Beckmann, C. F., Behrens, T. E. J., Woolrich, M. W., &Smith, S. M. (2012). FSL NeuroImage, 62(2), 782–90. doi:10.1016/j.neuroimage.2011.09.015.

Jim, H. S. L., Donovan, K. A., Small, B. J., Andrykowski, M. A.,Munster, P. N., & Jacobsen, P. B. (2009). Cognitive functioning inbreast cancer survivors: a controlled comparison. Cancer, 115(8),1776–83. doi:10.1002/cncr.24192.

Koppelmans, V., Breteler, M. M. B., Boogerd, W., Seynaeve, C., Gundy,C., & Schagen, S. B. (2012a). Neuropsychological performance insurvivors of breast cancer more than 20 years after adjuvant chemo-therapy. Journal of clinical oncology official journal of theAmerican Society of Clinical Oncology, 30(10), 1080–6. doi:10.1200/JCO.2011.37.0189.

Koppelmans, V., de Ruiter, M. B., van der Lijn, F., Boogerd, W.,Seynaeve, C., van der Lugt, A., & Schagen, S. B. (2012b). Globaland focal brain volume in long-term breast cancer survivors exposedto adjuvant chemotherapy. Breast Cancer Res Treat, 132(3), 1099–106. doi:10.1007/s10549-011-1888-1.

Koppelmans, V., de Groot, M., de Ruiter, M. B., Boogerd, W., Seynaeve,C., Vernooij, M. W., & Breteler, M. M. B. (2014). Global and focalwhite matter integrity in breast cancer survivors 20 years afteradjuvant chemotherapy. Hum Brain Mapp, 35(3), 889–99. doi:10.1002/hbm.22221.

Mangin, J.-F., Poupon, C., Clark, C., Le Bihan, D., & Bloch, I. (2002).Distortion correction and robust tensor estimation for MR diffusionimaging. Med Image Anal, 6(3), 191–8.

Nichols, T. E., & Holmes, A. P. (2002). Nonparametric permutation testsfor functional neuroimaging: a primer with examples. Hum BrainMapp, 15(1), 1–25.

Phillips, K. M., Jim, H. S., Small, B. J., Laronga, C., Andrykowski, M.A., & Jacobsen, P. B. (2012). Cognitive functioning after cancertreatment: a 3-year longitudinal comparison of breast cancer survi-vors treated with chemotherapy or radiation and noncancer controls.Cancer, 118(7), 1925–32. doi:10.1002/cncr.26432.

Pomykala, K. L., de Ruiter, M. B., Deprez, S., McDonald, B. C., &Silverman, D. H. S. (2013). Integrating imaging findings in evalu-ating the post-chemotherapy brain. Brain imaging and behavior.doi:10.1007/s11682-013-9239-y.

Provencher, S. W. (1993). Estimation of metabolite concentrations fromlocalized in vivo proton NMR spectra Magnetic resonance inmedicine. official journal of the Society of Magnetic Resonancein Medicine/Society of Magnetic Resonance in Medicine,30(6), 672–9.

Pullens, M. J. J., De Vries, J., & Roukema, J. A. (2010). Subjectivecognitive dysfunction in breast cancer patients: a systematic review.Psycho-Oncology, 19(11), 1127–38. doi:10.1002/pon.1673.

Rodenhuis, S., Bontenbal, M., Beex, L. V. A. M., Wagstaff, J., Richel, D.J., Nooij, M. A., & de Vries, E. G. E. (2003). High-dose

Brain Imaging and Behavior

chemotherapy with hematopoietic stem-cell rescue for high-riskbreast cancer. N Engl J Med, 349(1), 7–16. doi:10.1056/NEJMoa022794.

Schagen, S. B., Muller, M. J., Boogerd, W., Mellenbergh, G. J., & vanDam, F. S. A. M. (2006). Change in cognitive function after che-motherapy: a prospective longitudinal study in breast cancer pa-tients. J Natl Cancer Inst, 98(23), 1742–5. doi:10.1093/jnci/djj470.

Schilder, C. M., Eggens, P. C., Seynaeve, C., Linn, S. C., Boogerd, W.,Gundy, C. M., & Schagen, S. B. (2009). Neuropsychological func-tioning in postmenopausal breast cancer patients treated with tamox-ifen or exemestane after AC-chemotherapy: cross-sectional findingsfrom the neuropsychological TEAM-side study. Acta Oncol, 48(1),76–85. doi:10.1080/02841860802314738.

Schilder, C. M., Seynaeve, C., Beex, L. V., Boogerd, W., Linn, S. C.,Gundy, C. M., & Schagen, S. B. (2010). Effects of tamoxifen andexemestane on cognitive functioning of postmenopausal patientswith breast cancer: results from the neuropsychological sidestudy of the tamoxifen and exemestane adjuvant multinationaltrial. J of clinical oncology official journal of the AmericanSociety of Clinical Oncology, 28(8), 1294–300. doi:10.1200/JCO.2008.21.3553.

Schmahmann, J. D., Smith, E. E., Eichler, F. S., & Filley, C. M. (2008).Cerebral white matter: neuroanatomy, clinical neurology, and neu-robehavioral correlates. Ann N Y Acad Sci, 1142, 266–309. doi:10.1196/annals.1444.017.

Shilling, V., & Jenkins, V. (2007). Self-reported cognitive problems inwomen receiving adjuvant therapy for breast cancer. Euro j ofoncology nursing the official j of Euro Oncology Nursing Society,11(1), 6–15. doi:10.1016/j.ejon.2006.02.005.

Smith, S. M. (2002). Fast robust automated brain extraction. Hum BrainMapp, 17(3), 143–55. doi:10.1002/hbm.10062.

Strick, P. L., Dum, R. P., & Fiez, J. A. (2009). Cerebellum and nonmotorfunction. Annu Rev Neurosci, 32, 413–34. doi:10.1146/annurev.neuro.31.060407.125606.

Tang, C. Y., Friedman, J., Shungu, D., Chang, L., Ernst, T., Stewart, D., &Gorman, J. M. (2007). Correlations between Diffusion TensorImaging (DTI) and Magnetic Resonance Spectroscopy (1H MRS)in schizophrenic patients and normal controls. BMC psychiatry. doi:10.1186/1471-244X-7-25.

Wefel, J. S., & Schagen, S. B. (2012). Chemotherapy-related cognitivedysfunction. Current neurology and neuroscience reports, 12(3),267–75. doi:10.1007/s11910-012-0264-9.

Wefel, J. S., Vardy, J., Ahles, T., & Schagen, S. B. (2011). InternationalCognition and Cancer Task Force recommendations to harmonisestudies of cognitive function in patients with cancer. The lancetoncology, 12(7), 703–8. doi:10.1016/S1470-2045(10)70294-1.

Zhang, Y., Brady, M., & Smith, S. (2001). Segmentation of brain MRimages through a hidden Markov random field model and theexpectation-maximization algorithm. IEEE Trans Med Imaging,20(1), 45–57. doi:10.1109/42.906424.

Brain Imaging and Behavior