Neuroimaging and Neurocognitive Outcomes in Older Patients with Multiple Myeloma Treated with Chemotherapy and Autologous Stem Cell Transplantation

Background Many patients with hematological malignancies treated with stem cell transplantation (SCT) experience cognitive dysfunction. However, few studies have investigated treatment-related neurotoxicity in older adults with multiple myeloma (MM) treated with high dose chemotherapy (HDC) and autologous SCT (HDC/ASCT). In this study, we examined gray matter (GM) volume, resting state functional connectivity (RSFC), neurocognitive function (NF), and proinflammatory cytokines (PCy) in older patients with MM pre- and post-HDC/ASCT. Methods Eighteen MM patients underwent magnetic resonance imaging, neurocognitive tests, and serum PCy measurement prior to HDC/ASCT, and fifteen patients completed follow ups an average of five months post-HDC/ASCT. Results There were significant decreases in RSFC from pre- to post-HDC/ASCT in (1) the central executive network (CEN) involving the left dorsolateral prefrontal cortex and right posterior parietal cortex (p = 0.022), and (2) the CEN involving the right posterior parietal cortex and the salience network involving the right dorsal anterior cingulate cortex (p = 0.029); these comparisons were no longer significant after multiple comparisons correction. There were no significant changes in GM volumes or NF, except for improvement in attention (Digit Span Backward, p = 0.03). There were significant increases in several PCy post-HDC/ASCT (p ≤ 0.05). Conclusions This pilot study showed decreased RSFC involving the left frontal, right posterior parietal and right anterior cingulate cortices in MM patients post-HDC/ASCT, relatively stable NF, and increases in PCy. These findings are congruent with studies in patients with hematological malignancies and other cancers and provide supporting evidence for the vulnerability of frontoparietal regions to chemotherapy adverse effects.


Introduction
There is compelling evidence that chemotherapy is associated with neurotoxicity ( Reductions in prefrontal gray matter (GM) volume (McDonald, Conroy et al. 2012) and changes in resting state functional connectivity (RSFC) (Kesler 2014, Vachha, Gohel et al. 2022) have been reported postchemotherapy in patients with breast and ovarian cancers.We reported reductions in prefrontal GM volume and changes in white matter (WM) integrity in patients with hematological malignancies treated with high dose chemotherapy (HDC) ± total body irradiation (TBI) and stem cell transplantation (SCT) (Correa, Root et al. 2013, Correa, Wang et al. 2016).Systematic reviews indicated that neurocognitive dysfunction is prevalent pre-and post-SCT (Buchbinder, Kelly et al. 2018).
Neurocognitive dysfunction has been documented in multiple myeloma (MM) patients after HDC and preautologous SCT (ASCT) with declines post-HDC/ASCT (Jones, Vichaya et al. 2013, Root, Campbell et al. 2020); however, other studies reported improvement in neurocognitive function (NF) months postchemotherapy (Jacobs, Small et al. 2007, Bury-Kaminska, Szudy-Szczyrek et al. 2021).Pro-in ammatory cytokines (PCy) play a critical role in tumor growth and progression in MM, with high levels identi ed in many patients (Musolino, Allegra et al. 2017).Aging can be associated with PCy elevations (Weiskopf, Weinberger et al. 2009), suggesting that older MM patients may be even more susceptible to cytokine dysregulation related to disease and treatment, which may contribute to neuroin ammation.
There is a paucity of research investigating neurotoxicity in older MM patients undergoing HDC/ASCT, even though this intervention has been used more often in the elderly (Lemieux, Hulin et al. 2012), and NF has been recognized as a critical dimension of survivorship in older cancer patients (Mandelblatt, Jacobsen et al. 2014).In this pilot study, we assessed GM volume, RSFC, NF, and PCy in older MM patients prior to HDC/ASCT and an average of ve months post-HDC/ASCT.

Methods
Patients MM patients scheduled for conditioning HDC/ASCT were recruited through the Adult Bone Marrow Transplant Service at Memorial Sloan Kettering Cancer Center (MSK).Eligibility criteria: (1) MM diagnosis, (2) complete, partial, or very good partial disease remission at enrollment, as per standard International Myeloma Working Group Criteria, (3) age 60-75 at enrollment, (4) uent in English.

Measures
Structural & Functional Imaging.Patients were imaged in Tesla scanners (GE, Discovery 750W, USA with a GEM HNU 24-channel head coil) at MSK; ve patients were imaged in two different Tesla scanners at each timepoint using the same parameters.Structural Imaging: T1-weighted anatomical images with whole brain coverage were obtained with spoiled gradient-recalled and high-resolution three-dimensional magnetization-prepared rapid acquisition with gradient-echo sequences.Functional Imaging: For rsfMRI, T2*-weighted images were acquired with a single-shot gradient echo-planar imaging (EPI) sequence (TR/TE = 2500ms/30 ms, FA = 80°, slice thickness = 4 mm, matrix = 64 × 64).For the rsfMRI, patients were instructed to keep the eyes open and xated on a crosshair.Image Processing.For structural image processing, VBM analysis was performed using the longitudinal processing stream in the VBM8 toolbox (http://dbm.neuro.uni-jena.de/vbm/)under the SPM8 software package (Version 8, Wellcome Department of Imaging Neuroscience, London, UK) within MATLAB (Version 7, Mathworks, Inc., Natick, MA).Following reconstruction, follow-up MPRAGE structural images were registered to baseline MPRAGE images for each subject, bias corrected, segmented into GM, WM, and cerebrospinal uid compartments using the Montreal Neurologic Institute (MNI) T1 weighted template and tissue probability maps, linear and non-linear registered to MNI space, and the resulting GM tissue class smoothed using an isotropic Gaussian spatial lter (FWHM=8 mm).For rsfMRI preprocessing, a data pre-processing scheme was implemented according to published methods (Yang, Gohel et al. 2021, Vachha, Gohel et al. 2022).Brie y, the rst ve timepoints of fMRI data were removed to allow for T1 relaxation effects followed by head-motion correction, co-registration, segmentation, normalization to MNI standard space, temporal regression of 24 head-motion parameters (Friston, Williams et al. 1996), and ve principal components of WM and cerebral spinal uid time series (Behzadi, Restom et al. 2007), temporal ltering between 0.01 to 0.1 Hz, and spatial-smoothing with 6 mm fullwidth-at-half-maximum Gaussian lter.Using the head motion parameters, we calculated subject-speci c measures of mean frame-wise displacement (Jenkinson, Bannister et al. 2002).

Statistical and Imaging Analyses
Voxel-Based Morphometry (VBM).Following omnibus testing, pairwise t-tests were performed at the group level to analyze within group changes from pre-to post-HDC/ASCT.For the structural contrast, initial uncorrected voxel-wise threshold was p≤ 0.001 with resulting maps family-wise error corrected over the whole brain at p≤0.05.
Resting State Functional Connectivity Analysis (RSFC) was performed using region-of-interest based correlation as described previously (Biswal, Yetkin et al. 1995).Three resting state networks (RSNs) were extracted for prioritized analyses: the central executive network (CEN), the salience network (SN), and the default mode network (DMN).Spherical regions of interest (ROIs) were created surrounding the ROI coordinates appropriate to the RSNs of interest.The CEN and SN ROIs were created using the coordinates de ned by Uddin et al. ( respectively.The DMN ROIs were created using NeuroSynth (Yarkoni, Poldrack et al. 2011) with the functional connectivity and co-activation map derived in 02/2022 using the term "default mode".For each of the coordinates, a 6mm spherical ROI was created.Table 1 lists the MNI coordinates for each ROI.
Correlation matrices were produced by extracting the time course from each of the ROIs and computing the Pearson correlation coe cient (r) between each ROI pair in the CEN, SN, and DMN.Each of the pairwise ROI correlations were Fisher z-transformed for further statistical analysis.Changes in RSFC z-scores from pre-to post-HDC/ASCT were assessed using linear mixed models, adjusting for scanner (1,2).
Neurocognitive Analysis: Raw neurocognitive test scores were transformed into z-scores based on ageadjusted normative values.Neurocognitive test z-scores and self-report scale scores were summarized at each timepoint using descriptive statistics, and differences in scores from pre-to post-HDC/ASCT were compared using Wilcoxon signed rank tests.Standardized effect sizes (i.e., Cohen's d) were calculated to quantify the magnitude of score changes over time.False discovery rate (FDR) was used to adjust pvalues for multiple comparisons.The Reliable Change Index (RCI), which represents the change in scores divided by the standard error of measurement, was used to identify patients whose raw scores improved or declined beyond expected due to practice effects and measurement error.For each test score, we calculated the proportion of patients with RCI-indicated reliable decline.

Cytokine Panel Analysis
Cytokine data was analyzed using the MSD Discovery Workbench Ò software to measure levels of a ten-PCy panel at each timepoint.A four-parameter logistic (4PL) t calibration curve was generated for each analyte using the standards to calculate the concentration of each analyte.Upper and lower limits of quantitation for each PCy were established as the highest and lowest points of the standard curve on each plate whose back calculated values were within 80-120% of the expected values of the 4PL regression t and exhibited less than 20% CV across the duplicate standard wells.
Correlations.Spearman correlations were calculated to assess the association of RSN z-scores, neurocognitive test z-scores, self-report scale scores, and PCy levels separately for each timepoint.

Results
Eighteen MM patients completed a neurocognitive assessment and a brain MRI pre-HDC/ASCT, and fteen patients were available for follow up an average of ve months (median = 5.82, range = 3.50 -7.00) post-HDC/ASCT.One patient was excluded from the imaging analysis due to scan misregistration at follow up.Thirteen patients provided blood samples for PCy analysis at each timepoint.Descriptive statistics for demographic and disease variables are presented in Table 2.All patients received conditioning HDC the day before or the same day as the ASCT.Five patients were treated with Siltuximab, an IL-6 blocker (median half-life elimination: ~21 days, range:14 to 30 days) (Kurzrock, Voorhees et al. 2013, Chen, Teachey et al. 2016), infused seven days pre-HDC/ASCT and twenty-one days post-HDC/ASCT as part of a separate MSK protocol, which did not require high IL-6 levels for enrollment.In this study, data collection occurred prior to the rst Siltuximab infusion and at least 3-4 months after the second infusion for all patients.Among the three patients who did not return for follow-up, one had disease progression, one was deceased, and one relocated.

Structural & Functional Imaging
The results showed signi cant decreases in RSFC in (1) CEN ROIs involving the left dorsolateral prefrontal cortex (L-DLPFC) and right posterior parietal cortex (R-PPC) (p=0.022), and (2) the CEN ROI involving the R-PPC and the SN ROI involving the right dorsal anterior cingulate cortex (R-dACC) (p=0.029)from pre-to post-HDC/ASCT (Figure 1); these comparisons were no longer signi cant after correction for multiple comparisons.There were no signi cant changes in RSFC in the DMN.VBM analysis results showed no signi cant changes in regional GM volumes (FWE corrected, p>0.05).

Neurocognitive Function & Self-Report Scales
The neurocognitive tests z-scores and self-report scales scores are presented in Table 3. Mean z-scores were within the average range for all tests, except for ACT-perseverations.There were no signi cant changes in from pre-to post-HDC/ASCT, except for a signi cant improvement in attention (Longest Digit Span Backward, p=0.03); this comparison was no longer signi cant after multiple comparisons correction.There were no signi cant changes in the CES-D and FACIT-FS scores, and scores were within normal limits at each timepoint (Radloff 1977, Webster, Cella et al. 2003).There were no signi cant differences on the neurocognitive tests between the fteen patients who completed both timepoints and the three patients who performed the pre-HDC/ASCT assessment only.
Correlations.There were no signi cant correlations among RSN z-scores, neurocognitive test z-scores, self-report scale raw scores, and PCy levels either pre-or post-HDC/ASCT.

Discussion
This is the rst study describing alterations in RSFC in older MM patients treated with HDC/ASCT.The results showed decreased connectivity in the CEN involving the L-DLPF and R-PPC and in the CEN R-PPC and the SN R-dACC from pre-to an average of ve months post-HDC/ASCT, suggesting that adverse effects of HDC may be of concern in this population.The CEN has been described as a frontoparietal network involved in attention control, working memory, and processing speed ( A structural neuroimaging literature review in non-CNS cancer patients (McDonald 2021) described consistent ndings of reduced GM volume mostly in frontal, temporal and parietal areas, and diffuse alterations in WM integrity, mostly in patients treated with chemotherapy.Our study in patients with hematologic malignancies treated with HDC ± TBI and SCT (Correa, Root et al. 2013) showed GM volume reductions in the bilateral middle frontal gyrus from pre-to one-year post-SCT.However, the current results showed no signi cant changes in regional GM volume post-HDC/ASCT, suggesting no signi cant adverse effects on brain structure.
It is estimated that at least 50% of patients with hematological malignancies experience neurocognitive dysfunction prior to SCT, with either stable performance or declines months to years post-SCT (Syrjala, Dikmen et al. 2004, Jacobs, Small et al. 2007, Harrison, Sharafeldin et al. 2021).In this study, NF was within the average range, although below expected levels considering the mean education of the cohort, with impairment on a test of susceptibility to interference/working memory.This could be in part related to the disease and residual adverse effects of proteasome inhibitors, chemotherapy, and immunotherapy, (Rollin-Sillaire, Delbeuck et al. 2013) and in some patients, to the side effects of dexamethasone (Prado and Crowe 2019) and lenalidomide (Calvi, Marchetti et al. 2019).NF remained relatively stable, with improvement in attention, suggesting no signi cant adverse effects from HDC/ASCT; however, it has been suggested that stable NF post-SCT may represent lack of improvement expected due to practice effects (Phillips, McGinty et al. 2013).Self-report scales indicated no depression or fatigue, and no changes from pre-to post-HDC/ASCT.Some studies reported improvement in NF following chemotherapy in older MM patients (Bury-Kaminska, 2021), and one-year post-SCT in a sample of mostly MM patients.(Jacobs et al. 2007) However, neurocognitive dysfunction has been reported in MM patients after HDC and pre-ASCT, with declines one and three months post-HDC/ASCT (Jones, Vichaya et al. 2013).
Cytokine dysregulation is common in MM and may in uence the development of adverse effects (Musolino, Allegra et al. 2017).High PCy levels may interfere the blood brain barrier integrity ( ).However, the mechanisms of chemotherapy-related PCy dysregulation and NF are not well understood, with con icting results on the strength and direction of these associations (Janelsins, Mustian et al. 2012, Harrison, Sharafeldin et al. 2021).We observed low PCy levels at both timepoints, with signi cant increases post-HDC/ASCT, possibly in uenced by HDC adverse effects (Condomines, Veyrune et al. 2010).In some patients, the effects of dexamethasone and lenalidomide may have also in uence PCy levels.In the patients treated with Siltuximab, IL-6 and other PCy levels increased about ve months post-HDC/ASCT.There is evidence that IL-6 blockade may result in a paradoxical increase in IL-6 and systemic in ammation.(Speake, Habib et al. 2022) (Kostek, Nagaraju et al. 2012).However, additional research would be required to reconcile our ndings with the role of MM and its treatment on PCy dysregulation and neurotoxicity.
In this pilot study, changes in RSFC were more pronounced than in NF, possibly re ecting the use of compensatory mechanisms to maintain cognitive performance in the context of decreased CEN and SE connectivity (McDonald, Conroy et al. 2012), and the greater sensitivity of advanced neuroimaging tools to detect subtle alterations in functional connectivity in vulnerable regions.However, the RSFC results were no longer signi cant after multiple comparisons correction, and there were no signi cant associations between NF, RSFC, and PCy levels.These ndings may be in part related to reduced power to detect small changes and associations beyond what was reported, and to MM patients receiving fewer lines of prior chemotherapy and less neurotoxic agents and conditioning regimens, compared to other hematologic malignancies (Phillips, McGinty et al. 2013).

Conclusion
Decreased RSFC in older MM patients following HDC/ASCT provides further evidence for the prevailing notion that frontal-parietal regions may be vulnerable to chemotherapy adverse effects.Longitudinal studies with larger sample sizes are needed to further investigate the neural correlates of chemotherapyrelated neurotoxicity and the role of PCy in older MM patients, with the goal of developing targeted therapeutic interventions.

Declarations Ethical approval
All procedures performed in studies involving human participants were in accordance with the ethical standards of the institutional and/or national research committee and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards.The research protocol was approved by the Institutional Review Board at MSK.

Consent to Participate
Written informed consent was obtained from all study participants.
Uddin, Supekar et al. 2011) and Yang et al. (Yang, Gohel et al. 2021), Seeley, Menon et al. 2007, Vincent, Kahn et al. 2008, Nee and D'Esposito 2017).The SN includes the dACC and anterior insula (AI) cortex and is involved in monitoring and processing of errors and con ict (Seeley, Menon et al. 2007, Vincent, Kahn et al. 2008, Menon and Uddin 2010) (Heilbronner and Hayden 2016), and attention control in the presence of distraction (Shenhav, Botvinick et al. 2013).Although we found no signi cant changes in the DMN, the CEN and SE results are overall consistent with breast and ovarian cancer studies suggesting that frontoparietal regions are susceptible to the adverse effects of chemotherapy (Kesler 2014, Feng, Wang et al. 2020, Shen, Tsai et al. 2021) (Correa, Root et al. 2017) (Vachha, Gohel et al. 2022).

Figures
Figures