"skipMainNavigation"
closeDrawerMenuopenDrawerMenu
Search
Quick Search in Journals
Quick Search anywhere
Advanced Search
RSNA Journals
Menu
  • Information
    • Previous
    Next
    Original ResearchOpen Access

    Pretherapy Ferumoxytol-enhanced MRI to Predict Response to Liposomal Irinotecan in Metastatic Breast Cancer

    Published Online:Feb 3 2023https://doi.org/10.1148/rycan.220022
    See Invited Commentary byHeike E. Daldrup-Link

    Abstract

    Purpose

    To investigate ferumoxytol (FMX)-enhanced MRI as a pretreatment predictor of response to liposomal irinotecan (nal-IRI) for thoracoabdominal and brain metastases in women with metastatic breast cancer (mBC).

    Materials and Methods

    In this phase 1 expansion trial (ClinicalTrials.gov identifier, NCT01770353; 27 participants), 49 thoracoabdominal (19 participants; mean age, 48 years ± 11 [SD]) and 19 brain (seven participants; mean age, 54 years ± 8) metastases were analyzed on MR images acquired before, 1–4 hours after, and 16–24 hours after FMX administration. In thoracoabdominal metastases, tumor transverse relaxation rate (R*2) was normalized to the mean R*2 in the spleen (rR*2), and the tumor histogram metric rR*2,N, representing the average of rR*2 in voxels above the nth percentile, was computed. In brain metastases, a novel compartmentation index was derived by applying the MRI signal equation to phantom-calibrated coregistered FMX-enhanced MRI brain scans acquired before, 1–4 hours after, and 16–24 hours after FMX administration. The fraction of voxels with an FMX compartmentation index greater than 1 was computed over the whole tumor (FCIGT1) and from voxels above the 90th percentile R*2 (FCIGT1 R*2,90).

    Results

    rR*2,90 computed from pretherapy MRI performed 16–24 hours after FMX administration, without reference to calibration phantoms, predicted response to nal-IRI in thoracoabdominal metastases (accuracy, 74%). rR*2,90 performance was robust to the inclusion of some peritumoral tissue within the tumor region of interest. FCIGT1 R*2,90 provided 79% accuracy on cross-validation in prediction of response in brain metastases.

    Conclusion

    This first in-human study focused on mBC suggests that FMX-enhanced MRI biologic markers can be useful for pretherapy prediction of response to nal-IRI in patients with mBC.

    Keywords: MRI Contrast Agent, MRI, Breast, Head/Neck, Tumor Response, Experimental Investigations, Brain/Brain Stem

    Clinical trial registration no. NCT01770353

    Supplemental material is available for this article.

    Published under a CC BY 4.0 license.

    See also commentary by Daldrup-Link in this issue.

    Download as PowerPoint

    Summary

    In a phase 1 clinical study of participants with metastatic breast cancer, quantitative metrics derived from pretreatment ferumoxytol-enhanced MRI were predictive of tumor response to liposomal irinotecan.

    Key Points

    • ■ In thoracoabdominal metastatic breast cancer metastases, tumor transverse relaxation rate (R*2) relative to spleen relaxation rate (rR*2) computed on a single 16- to 24-hour postferumoxytol (FMX)-enhanced MRI scan predicted response to liposomal irinotecan (nal-IRI) at the tumor level with 74% accuracy.

    • ■ In brain metastases, the fraction of voxels with FMX compartmentation index greater than 1 was computed over the whole tumor (FCIGT1) and from voxels above the 90th percentile R*2 (FCIGT1R*2,90), novel compartmentation index metrics based on the spoiled gradient-recalled echo signal equation, and FCIGT1 R*2,90 predicted response to nal-IRI at the tumor level with 79% accuracy.

    Introduction

    Breast cancer is the most diagnosed cancer (1) and the leading cause of cancer-related death in women worldwide (2). Despite continued improvements in available treatments, the 5-year survival rate in those with metastatic breast cancer (mBC) remains low at 27% (3), and treatment options are typically palliative and are limited to hormonal therapy, radiation therapy, and chemotherapy to control symptoms and prolong survival. Irinotecan (IRI) is a topoisomerase I inhibitor used as a salvage treatment in heavily pretreated patients with mBC (4). Liposomal IRI (ONIVYDE; Ipsen Biopharmaceuticals; historic names include nal-IRI [liposomal irinotecan], MM-398, and PEP02) is a liposome-encapsulated IRI with an individual liposome diameter of 110 nm (5). Preclinical studies in models of spontaneous and brain metastases of triple-negative breast cancer showed the accumulation of nal-IRI in metastases. The treatment provided a significant survival benefit at lower doses compared with standard IRI (6,7). In an early-phase study (ClinicalTrials.gov identifier, NCT01770353) of heavily pretreated participants with mBC, the objective response rate to nal-IRI (50 or 70 mg/m2 free base every 2 weeks) was 30% for those with central nervous system disease and 34.5% for those with non–central nervous system disease (3).

    Nal-IRI increases the nominal plasma half-life of IRI, permitting the slow deposition of the drug into tumor lesions across leaky vasculature. Nal-IRI is trapped in the tumor extracellular space via the enhanced permeability and retention effect (8), where it is phagocytosed by local macrophages that act as drug depots and release IRI locally (9). The enhanced permeability and retention effect depends on factors such as local tumor-associated macrophage (TAM) content and vascular permeability (10) and physiochemical properties of the nanoparticles (11). The TAMs modulate mechanistic parameters, such as vascular permeability, facilitating the deposition of nal-IRI nanoparticles at the tumor site (12). Across the blood-brain barrier, vascular permeability alone could be an important factor in nanoparticle deposition (13,14). Furthermore, 20–200-nm nanoparticles have desirable tumor accumulation and penetration properties (15). Correlates of nal-IRI deposition would be useful when predicting tumor response to treatment.

    While biopsy- or surgery-based pathologic analysis remains the reference standard to monitor tumor response, imaging offers the potential to noninvasively sample the entire volume of multiple tumors in a patient, and imaging using nanoparticle tracers or contrast agents has also been investigated (16,17). Ultrasmall superparamagnetic iron oxides (USPIOs) are nanoparticle contrast agents that induce dose-dependent enhancements in tissue MRI relaxation properties (R1 and R*2). The USPIOs have a unique dynamic distribution process between the vascular and extravascular compartments, resulting in varied MRI enhancement. During the first few hours after infusion, USPIOs reside primarily in the vascular space, with MRI enhancement from the vascular component alone. After approximately 24 hours, as a function of local microvascular permeability, most of the administered USPIOs extravasate (enhanced permeability and retention effect) and are phagocytosed by local macrophages, resulting in an additional delayed MRI enhancement (18,19). Furthermore, the pharmacokinetic and biodistribution characteristics of USPIOs mimic those of nal-IRI (16) and similarly sized nanotherapeutics (17). Therefore, USPIO-enhanced MRI is being investigated as an indicator of tumor microvascular permeability and TAM content.

    Ferumoxytol (FMX) (Feraheme; AMAG Pharmaceuticals) is a USPIO approved by the U.S. Food and Drug Administration to treat iron deficiency anemia, and it is used as an off-label MRI contrast agent (14). It is a neutrally charged USPIO with a colloidal particle size of 23 nm (20,21). Preclinical studies showed that FMX colocalized with nanotherapeutics nal-IRI and poly (d,l-lactic-co-glycolic acid)-b-polyethylene glycol within the tumor microenvironment without affecting the pharmacokinetics, biodistribution, or cellular distribution of the nanotherapeutics (16,17). These studies also showed that R*2 enhancement on FMX-enhanced MR images was predictive of treatment response and accumulation of nanotherapeutics at the tumor site (14,16,17,22).

    To build on these preclinical studies, we investigated the use of FMX-enhanced MRI as a companion diagnostic tool in a pilot phase 1 study of nal-IRI in advanced solid tumors. In a preliminary study, we observed the colocalization of FMX with macrophages in biopsy tissue from tumors and found that FMX concentration levels computed using post-FMX R*2 were highly correlated with lesion size change in solid tumors (23). A correlation between tumor R*2 measured after FMX and TAMs has also been reported in the setting of pediatric lymphoma and osteosarcoma by Aghighi and colleagues (24) and in the setting of adult glioma by Iv et al (25).

    Systemically injected nanoparticles such as nal-IRI and FMX that extravasate into solid tumors may remain extravascular and extracellular in the tumor microenvironment, or they may be compartmentalized inside phagocytic cells, such as TAMs. The success of nanoparticle-based drug therapy depends on individual lesion vascular permeability to the formulation to allow the deposition of efficacious drug concentrations to the patient’s tumor lesions (26–28). Heterogeneity in permeability is observed within and between the lesions (23,28,29). In this phase 1 expansion study in participants with advanced breast cancer treated with nal-IRI, we investigated tumor FMX-enhanced MRI parameters as pretreatment predictors of individual tumor response to nal-IRI. We hypothesized that FMX-enhanced MRI parameters capture individual lesion permeability characteristics for nanoparticles and specifically measure FMX deposition in each tumor lesion; thus, these parameters may be biologic markers of nal-IRI uptake into that tumor that can be used to predict tumor response to nal-IRI. In this report, we apply novel data analysis and fitting strategies to improve the lesion-specific FMX-enhanced MRI signal estimates and describe pretherapy predictors of mBC tumor response to nal-IRI.

    Materials and Methods

    Clinical Study

    Ipsen sponsored this prospective study (November 2012–October 2018), and one of the authors (S.G.K.) is an Ipsen employee. All authors, regardless of employment, had control of data inclusion and analysis. This Health Insurance Portability and Accountability Act–compliant study was approved by the institutional review board of all participating institutions, and all study participants provided written informed consent before participating in the FMX-enhanced MRI and treatment portions of the study. Of the 30 participants enrolled, 29 participants with mBC received at least one dose of nal-IRI in the expansion phase of the parent phase 1 open-label nonrandomized study, of whom 27 also had FMX-enhanced MR images that were suitable for analysis. The safety and efficacy results of this expansion study have been published separately (3).

    Study Inclusion and Exclusion Criteria

    Women aged 18 years or older were enrolled in the study if they had advanced or metastatic breast cancer that had progressed on at least one but no more than five prior regimens in the metastatic setting; Eastern Cooperative Oncology Group performance status of 0, 1, or 2; and acceptable kidney, bone marrow, and liver function. Participants were enrolled in three cohorts (cohort 1, hormone receptor positive and human epidermal growth factor receptor 2 negative; cohort 2, triple-negative breast cancer; cohort 3, breast cancer with active brain metastasis). Notable exclusion criteria included prior IRI or bevacizumab therapy within the preceding 6 months, known hypersensitivity to IRI or FMX, inability to undergo MRI, and administration of parenteral iron in the preceding 4 weeks.

    Study Schema

    Participants underwent thoracoabdominal and/or brain MRI based on metastasis sites (Table 1). The study schema with timing of FMX-enhanced MRI and standard diagnostic scans is shown in Figure 1. As previously reported, non–central nervous system tumors were assessed using Response Evaluation Criteria in Solid Tumors, version 1.1 (30), with CT or MRI, while central nervous system disease was assessed using modified Response Evaluation Criteria in Solid Tumors (3). MRI scans were acquired before FMX (day 1) and 16–24 hours after FMX (day 2) administration in all participants and at 1–4 hours after FMX (day 1) administration in a subset of participants (n = 18). In six participants, the pre-FMX MRI scans were replicated twice in the same session to assess reproducibility. Ferumoxytol was administered to participants at a dose of 5 mg of iron per kilogram of body weight as an intravenous infusion over 15 minutes, not exceeding 510 mg of iron total. Participants were observed for signs or symptoms of hypersensitivity reactions; this included monitoring vital signs (blood pressure, respiratory rate, and pulse rate) during and for at least 30 minutes after FMX infusion. Within 7 days after FMX infusion, participants received their first dose of intravenous nal-IRI (70 mg/m2 free base), infused over 90 minutes. The protocol was amended on November 3, 2016, to reduce the starting dose to 50 mg/m2 free base with allowance for escalation to 70 mg/m2 free base.

    Table 1: Participant Characteristics, Ferumoxytol-enhanced MRI Scan Sessions, and Tumor Response Classes

    Table 1:
    Timing of ferumoxytol (FMX)-enhanced MRI, diagnostic imaging, and liposomal irinotecan (nal-IRI) treatment in a phase 1 study of nal-IRI in an expansion cohort of participants with metastatic breast cancer. Pretreatment FMX-enhanced MRI scans were acquired before FMX and 16–24 hours after FMX administration in 27 participants. In 18 participants, FMX-enhanced MRI scans were also acquired 1–4 hours after FMX administration. The nal-IRI treatment regimen and standard-of-care diagnostic imaging were initiated approximately 1 week after the end of FMX-enhanced MRI sessions.

    Figure 1: Timing of ferumoxytol (FMX)-enhanced MRI, diagnostic imaging, and liposomal irinotecan (nal-IRI) treatment in a phase 1 study of nal-IRI in an expansion cohort of participants with metastatic breast cancer. Pretreatment FMX-enhanced MRI scans were acquired before FMX and 16–24 hours after FMX administration in 27 participants. In 18 participants, FMX-enhanced MRI scans were also acquired 1–4 hours after FMX administration. The nal-IRI treatment regimen and standard-of-care diagnostic imaging were initiated approximately 1 week after the end of FMX-enhanced MRI sessions.

    Download as PowerPoint

    FMX-enhanced MRI Acquisition

    Ferumoxytol-MRI data were acquired with 1.5-T General Electric (26 participants) and 3.0-T Siemens (one participant) scanners at six different clinical sites. A series of six fat-suppressed fast spoiled gradient-echo images were acquired with the 1.5- or 3.0-T MRI scanners, respectively, with the following echo times: 1.5 and 2.2, 3.0 and 4.5, 4.5 and 6.8, 6.0 and 9.0, 9.0 and 11.3, and 13.2 and 13.5 msec, as previously described (23). Briefly, the section thickness was 6 mm (1-mm section gap) for thoracoabdominal metastases and 5 mm (no section gap) for brain metastases, and matrix size was 256 × 256 with a field of view to match the size of the imaging section with phantom tubes containing known concentrations of FMX (Fig S1).

    Tumor Delineation

    Tumors were identified and manually contoured on all applicable sections using mint Lesion, version 3.2 (Mint Medical) by an expert oncoradiologist (R.L.K.) with more than 25 years of experience; tumors were contoured without foreknowledge of eventual responder or nonresponder status of individual lesions. Regions of interest (ROIs) were also manually defined within multiple normal tissues and organs by an expert imaging scientist (N.R.) with more than 20 years of experience. Reference tissues analyzed included muscle, fat, spleen, renal cortex, white matter, cerebrospinal fluid, gray matter, and major blood vessels.

    Preprocessing

    For each participant and imaging session, multiecho FMX-enhanced MR images were preprocessed using in-house Matlab scripts as follows. First, all images were checked for quality and adherence to the study imaging protocol. Next, the multiecho images were spatially coregistered to the reference echo image on which the radiologist drew tumor contours (usually the longest echo time). Globally registered images were then locally registered in regions that contained the contoured tumors.

    Postprocessing

    R*2 mapping.— The signal (S) of a spoiled gradient-echo sequence was modeled using the following equation:

    Here, M0, TR, TE, and α are the proton density, repetition time, echo time, and flip angle, respectively. R1 and R*2 are longitudinal and transverse relaxation rates, respectively. Maps of R*2 were computed from pixel intensities of registered multiple-echo-time data fitted using log-linear regression shown by Equation 2:

    where S and S0 = are the signal at echo time ≠ 0 msec and echo time = 0 msec, respectively.

    Estimation of FMXc.— The pixelwise concentration of FMX in brain metastases, FMXc,brain, at the post-FMX imaging session was computed using Equation 3 since these images were coregistered between the pre- and post-FMX sessions:

    Quantification of FMX compartmentation (brain metastases only).— The longitudinal relaxivity, r1, of USPIOs would be expected to either not change or decrease due to compartmentation within cells, while the transverse relaxivity, r*2, would be expected to increase (31). Estimation of pixelwise apparent concentration FMXc using coregistered R1 and R*2 maps acquired before and after FMX would therefore theoretically allow for identification of pixels with high concentrations of compartmentalized (phagocytosed) FMX. In our study, we did not acquire R1 maps to limit the total scanning duration for reasons of participant comfort. Instead, for each participant imaging data set, we used a modified spoiled gradient-recalled echo signal equation to calculate the R1-weighted post-FMX image that would be expected for a given FMXc,brain and compared that with the R1-weighted component that was computed from the shortest-echo-time post-FMX image acquired. We did this because pixels with substantial compartmentation of FMX would be brighter on the expected R1-weighted image relative to the computed R1-weighted image. This approach is described in Equations 4–10 and Figure S2, as follows:

    Here, r1,FMX is 19 mM−1 sec−1 at 1.5 T and 9.5 mM−1 sec−1 at 3.0 T (32), and R1,0 is the tissue longitudinal relaxation rate in the absence of a contrast agent. The R1-weighted component of the spoiled gradient-recalled echo signal in Equation 4 was computed in Equation 6:

    where the weighting factor exp(−TE · R*2) · sin (α) accounts for R*2 contribution to the computed signal from field strength, field inhomogeneity, or FMX accumulation. The subscript “i” in Equations 4–6 corresponds to the imaging sessions (see Fig S2), where i = 0, 1, and 2 represents before FMX, 1–4 hours after FMX, and 16–24 hours after FMX, respectively. R*2,i is transverse relaxation rate of the imaging session i. By combining Equations 4, 5, and 6, the signal equation can be written in terms of R1 as in Equation 7:

    M0 and r1,FMX represent the pixelwise apparent proton density and longitudinal relaxivity of FMX, respectively.

    This model required one pre-FMX scan, an early-phase post-FMX scan, and a delayed-phase post-FMX MRI scan to identify voxels with FMX compartmentation within the cells. A nonlinear regression was used to fit for the parameters R1,0 and M0 in Equation 7. The fitted parameters were then used to calculate fitted R1, and model-expected R1 weighted signal, as shown in Equations 8 and 9:

    SExpected,i and SComputed,i represent the model-expected and model-estimated R1 signal, respectively. A compartmentation index was computed as the ratio of SExpected,i and SComputed,i, as shown in Equation 10: values above unity would indicate some degree of extravascular intracellular compartmentation:

    Scanner receiver gain differences between the imaging sessions were accounted for by linear scaling of pixel intensities to the mean intensities of the two phantom tubes containing 20 and 80 mg/mL iron that were imaged with each participant. The scaling was applied before the nonlinear regression step.

    Tumor Classes

    Responding tumor (RT) and nonresponding tumor (NRT) classes were defined on the basis of maximum percentage diameter change at follow-up relative to baseline. Radiologic diameter change plots for all tumors are shown in Figures 2 and 3. A thoracoabdominal metastasis was labeled RT if it exhibited a diameter decrease greater than 30% (33) and was labeled NRT otherwise. We stratified thoracoabdominal tumors by diameter at baseline and manually pair-matched tumors to create two balanced classes (Table 2). A brain metastasis was classified as RT if it exhibited a diameter decrease greater than 30% and did not progress (an increase from nadir greater than 20% [33]) and was classified as NRT otherwise (Table 3). In Figures 2 and 3 and Tables 2 and 3, the maximum tumor diameter change is marked as zero if a tumor either did not change in diameter or grew beyond baseline diameter.

    Radiologic follow-up of individual tumors. Tumor diameter change for 49 thoracoabdominal tumors from 19 participants. Notation at the top of each plot indicates participant (P) number and corresponding lesion (L) number with response status (1 for responding tumors and 0 for nonresponding tumors).

    Figure 2: Radiologic follow-up of individual tumors. Tumor diameter change for 49 thoracoabdominal tumors from 19 participants. Notation at the top of each plot indicates participant (P) number and corresponding lesion (L) number with response status (1 for responding tumors and 0 for nonresponding tumors).

    Download as PowerPoint
    Radiologic follow-up of individual tumors. Tumor diameter change for 19 brain tumors from seven participants. Notation at the top of each plot indicates participant (P) number and corresponding lesion (L) number with response status (1 for responding tumors, 0 for nonresponding tumors).

    Figure 3: Radiologic follow-up of individual tumors. Tumor diameter change for 19 brain tumors from seven participants. Notation at the top of each plot indicates participant (P) number and corresponding lesion (L) number with response status (1 for responding tumors, 0 for nonresponding tumors).

    Download as PowerPoint

    Table 2: Pretreatment Tumor Diameters for Thoracoabdominal Metastases

    Table 2:

    Table 3: Pretreatment Tumor Diameters for Brain Metastases

    Table 3:

    Image Analysis

    FMX-enhanced MRI discriminators.— For thoracoabdominal tumors, relative R*2 (rR*2) per pixel was computed by dividing R*2 by the corresponding spleen ROI average values. Histograms of rR*2 for each tumor were analyzed (details in Fig S3), and the average values of all pixels above the nth percentile were labeled as rR*2,N. For example, the mean value of rR*2 in pixels above the 90th percentile is denoted rR*2,90.

    For brain metastases, compartmentation index maps were computed using Equation 10. Participants who underwent brain MRI did not always have accompanying spleen images. Therefore, we only computed unnormalized R*2,90 for brain metastases. From the compartmentation index maps, we computed two metrics for each brain metastasis: (a) the fraction of pixels within a tumor volume of interest with compartmentation index greater than 1 (FCIGT1) and (b) FCIGT1 and R*2 in the 90th and higher percentile (FCIGT1R* 2,90). The area under the receiver operating characteristic (ROC) curve (AUC) for discrimination of RT and NRT tumors was computed for both metrics.

    Optimizing the FMX-enhanced MRI Discriminators for Classifying Thoracoabdominal Tumors

    To classify thoracoabdominal tumors as RT or NRT, we optimized the threshold value for rR* 2. We investigated both the optimal percentile (nth bin) and rR* 2,N as follows.

    Optimal percentile bin selection.— For rR*2, mean values of all pixels in the nth bin were calculated for the 5th–99th percentile, and AUC was computed for rR*2,5 through rR*2,99. The robustness of the AUC for rR*2,N at each N to variability in the tumor ROI definition was assessed using Reproducibilityindex,N (details in Appendix S1). An optimal value of N (Nopt) was chosen that provided both high AUC for discrimination between RT and NRT tumor classes and a high reproducibility index (Equation S4 in Appendix S1) with respect to ROI dilation and erosion. The AUC was reported as AUC ± SE·Zcrit (details in Appendix S2).

    Optimal cutoff value rR*2,N opt.— For the identified optimal percentile bin, Nopt, ROC analysis was performed to identify optimal cutoff value for the discriminator rR*2,N opt using the distance-to-corner criterion (34). The ROC plots were generated using R software, version 4.1.3 (R Project for Statistical Computing).

    Repeatability and Generalizability

    Bland-Altman analysis was used to assess the repeatability of R*2 measurements in a reference tissue, namely muscle. For each participant, the R*2 average across multiple repeated sessions is plotted against the R*2 difference. The coefficient of repeatability (35), which measures the maximum difference likely to occur between repeat measurements, was computed using Equation 11:

    where n is the number of participants and di,1 and di,2 are repeated measurements of participant i.

    The coefficient of variation was calculated using Equation 12:

    To test the local generalizability of the discriminators, we performed leave-one-out cross validation (LOOCV). The performance of the LOOCV test was quantified by calculating the number of correctly identified tumors divided by the number of possible permutations of LOOCV runs.

    Statistical Analysis

    The significance of observed differences between pre-FMX and post-FMX values of R*2 in different tissues was assessed using a two-sample t test in Matlab, version 2018b (MathWorks). In addition, changes in R*2 after FMX injection were compared across thoracoabdominal and brain metastasis. Bonferroni correction for multiple comparisons was applied, resulting in a significance level of P < .001. Mean R*2 (rR*2) between RT and NRT in thoracoabdominal metastases was compared using the t test, with P < .01 considered to indicate a significant difference. Violin plots (36) were used to visualize the differences in R*2 (rR*2) between RT and NRT.

    Results

    FMX-enhanced MRI Participant Characteristics

    Participant demographics are summarized in Table 1. FMX-enhanced MRI studies from a total of 27 (mean age, 52 years ± 10 [SD]) participants were analyzable: 17 participants (mean age, 54 years ± 8) underwent only thoracoabdominal FMX-enhanced MRI, six participants (mean age, 47 years ± 11) underwent only brain FMX-enhanced MRI, and the remaining four participants (mean age, 49 years ± 7) underwent both brain and thoracoabdominal FMX-enhanced MRI. Two participants each from the thoracoabdominal and brain metastasis arms were excluded from the analysis because of a lack of clinical follow-up information on tumor size. In addition, one participant from the brain metastasis arm was excluded from analysis due to poor image quality of acquired multiecho data. A total of 49 thoracoabdominal (19 participants; mean age, 48 years ± 11) (liver, n = 37; lymph node, n = 3; peritoneum, n = 3; pleura, n = 2; chest wall, n = 2; pancreas, n = 1; retroperitoneum, n = 1) and 19 brain (seven participants; mean age, 54 years ± 8) metastases were analyzed.

    Analysis of Tissue R*2 Changes 16–24 Hours after FMX versus before FMX

    Figure 4A and 4B shows example R*2 maps computed before FMX and 24 hours after FMX in a participant with liver metastases. Qualitatively, a significant increase in R*2 can be observed in organs of the mononuclear phagocyte system (liver and spleen), blood vessels, and kidney. Tumor R*2 also appeared heterogeneously higher after FMX relative to before FMX. Figures 4E and 4F depict R*2 maps computed before FMX and 24 hours after FMX in a participant with brain metastases. R*2 in tumor and blood vessels was higher after FMX compared with before FMX, while R*2 changes in normal brain parenchyma were relatively small. Quantitative R*2 values in various tissue types are presented in Figure 5A. As expected, R*2 of tumors, mononuclear phagocyte system, blood vessels, and the kidney was higher (P < .001) after FMX than before FMX. The post-FMX versus pre-FMX increase in R*2 of thoracoabdominal metastases (ΔR*2 = 83 sec−1) was more than twice (P < .001) the increase observed in brain metastases (ΔR*2 = 33 sec−1).

    Transverse relaxation rate (R*2) maps computed before ferumoxytol (FMX) and 24 hours after FMX administration in a representative participant with (A, B) liver and (E, F) brain metastases (arrows). R*2-weighted anatomic images on which radiologists drew tumor regions of interest (arrows) are shown for (C, D) liver and (G, H) brain. The color scale, from blue to red, indicates increasing values of R*2 (in seconds−1).

    Figure 4: Transverse relaxation rate (R*2) maps computed before ferumoxytol (FMX) and 24 hours after FMX administration in a representative participant with (A, B) liver and (E, F) brain metastases (arrows). R*2-weighted anatomic images on which radiologists drew tumor regions of interest (arrows) are shown for (C, D) liver and (G, H) brain. The color scale, from blue to red, indicates increasing values of R*2 (in seconds−1).

    Download as PowerPoint
    Summary of transverse relaxation rate (R*2) measurements before and after ferumoxytol (FMX) administration. (A) R*2 values measured from pre-FMX and 16- to 24-hour post-FMX MRI scans in tumor and normal tissues (P < .001). (B) Bland-Altman plot of repeatability of R*2 measurements in reference muscle tissue. CSF = cerebrospinal fluid, IVC = inferior vena cava.

    Figure 5: Summary of transverse relaxation rate (R*2) measurements before and after ferumoxytol (FMX) administration. (A) R*2 values measured from pre-FMX and 16- to 24-hour post-FMX MRI scans in tumor and normal tissues (P < .001). (B) Bland-Altman plot of repeatability of R*2 measurements in reference muscle tissue. CSF = cerebrospinal fluid, IVC = inferior vena cava.

    Download as PowerPoint

    In six participants, we obtained repeat studies in the pre-FMX session to assess the repeatability of R*2 measurements. The Bland-Altman repeatability test on the pre-FMX reference muscle tissue in these six participants is shown in Figure 5B. All calculated difference points are within the 95% CI line. The coefficient of repeatability was 6.36 sec−1, and the coefficient of variation was 6.5%.

    Tumor R*2 Normalization Improves Classification of Thoracoabdominal Metastases

    Of the systemic tumors assessed with Response Evaluation Criteria in Solid Tumors, most lesions (28 of 49 [57%]) experienced a decrease in diameter after therapy (Fig 2). To control for the effect of pretreatment tumor diameter in response to nal-IRI treatment, we manually matched the pretreatment diameters of tumors and created two balanced classes of RT (21) and NRT (21) for our analysis (Table 2). While the 16- to 24-hour post-FMX mean R*2 of the systemic whole tumors showed no evidence of a difference between the two classes (Fig 6A), the spleen-normalized mean rR* 2 of whole tumors at 16–24 hours after FMX was different between the two classes (P < .01, Fig 6B), with an AUC of 70% (Fig S5).

    Discriminator classification performance for response assessment in abdominal metastases. Violin plots of responding tumor (RT) versus nonresponding tumor (NRT) classes for (A) mean unnormalized transverse relaxation rate (R*2) and (B) mean spleen normalized transverse relaxation rate (rR*2). Nineteen participants’ average spleen R*2 are shown 24 hours after FMX (C). The error bar represents R*2 SD within the regions of interest. The area under the receiver operating characteristic curve (AUC) of rR*2 (D) shows that discriminator performance increases as the percentile threshold (N) increases. (E) However, Reproducibilityindex decreases with increasing percentile. A compromise threshold percentile (Nopt) is indicated by the dashed gray line, which was selected to provide good AUC for rR*2 and acceptable Reproducibilityindex.

    Figure 6: Discriminator classification performance for response assessment in abdominal metastases. Violin plots of responding tumor (RT) versus nonresponding tumor (NRT) classes for (A) mean unnormalized transverse relaxation rate (R*2) and (B) mean spleen normalized transverse relaxation rate (rR*2). Nineteen participants’ average spleen R*2 are shown 24 hours after FMX (C). The error bar represents R*2 SD within the regions of interest. The area under the receiver operating characteristic curve (AUC) of rR*2 (D) shows that discriminator performance increases as the percentile threshold (N) increases. (E) However, Reproducibilityindex decreases with increasing percentile. A compromise threshold percentile (Nopt) is indicated by the dashed gray line, which was selected to provide good AUC for rR*2 and acceptable Reproducibilityindex.

    Download as PowerPoint

    Ninetieth Percentile for Classification of Thoracoabdominal Metastases

    The AUC of rR*2,N for discrimination between NRT and RT was found to increase with N (Fig 6D and Fig S5). However, the Reproducibility index in tumor rR* 2,N with erosion and dilation of the tumor ROI was lower for higher N (Fig 6E). Thus, accuracy and reproducibility moved in opposite directions as N was increased from the 80th to the 99th percentile. We selected a compromise value of Nopt = 90th percentile, which provided good classification performance (AUC) and acceptably low sensitivity to variability in the tumor contours (higher Reproducibility index).

    Thoracoabdominal Metastases Response Prediction Using rR*2,90

    Complete ROC plots of rR*2,90 and R*2,90 are shown in Figure S6. Normalization of R*2 values to the respective mean values in the spleen improved the AUC. To discriminate between RT and NRT, the AUC was 0.77 ± 0.1 (95% CI) for rR*2,90 (Table 4, Fig S6). The optimal threshold value of rR* 2,90 was 1.07. For the optimum threshold, training accuracy and LOOCV accuracy were 76% and 74%, respectively (Table 4).

    Table 4: Summary of AUC Results for Discriminators Used to Predict RT versus NRT in Thoracoabdominal and Brain Metastases

    Table 4:

    Thoracoabdominal Metastases: Classification Accuracy Is Sensitive to ROI Erosion

    Radiologist-drawn tumor contours were eroded and dilated by 1 to 5 pixels in all directions in plane to investigate the impact of tumor contouring variability on performance of the tumor classifier rR*2,90. As observed in Figure 7A, the AUC of the rR*2,90 classifier decreases monotonically as the radiologist-drawn tumor contours are eroded by 1 to 5 pixels. In contrast, dilation of the radiologist-drawn contours by 1 to 5 pixels did not affect the AUC of rR*2,90 (Fig 7B).

    Sensitivity of the average of spleen normalized transverse relaxation rate voxels in the 90th percentile bin in regions of interest (rR*2,90) discriminator performance to variability in the tumor regions of interest. Area under the receiver operating characteristic curve (AUC) for discriminating responding tumor versus nonresponding tumor thoracoabdominal metastases was calculated after (A) erosion and (B) dilation of radiologist-drawn regions of interest by 1 to 5 voxels in all directions in plane.

    Figure 7: Sensitivity of the average of spleen normalized transverse relaxation rate voxels in the 90th percentile bin in regions of interest (rR* 2,90) discriminator performance to variability in the tumor regions of interest. Area under the receiver operating characteristic curve (AUC) for discriminating responding tumor versus nonresponding tumor thoracoabdominal metastases was calculated after (A) erosion and (B) dilation of radiologist-drawn regions of interest by 1 to 5 voxels in all directions in plane.

    Download as PowerPoint

    Brain Metastases: Distribution of Voxels with Compartmentation Index of 1 or Greater Correlates with Response

    We computed compartmentation index maps as surrogates for mismatch between the expected R1 enhancement for a given R*2 enhancement. Figure 8A shows the computed and expected R1-weighted maps, compartmentation index map, and R*2 maps for three example tumors. Qualitatively, on the compartmentation index maps, the responding tumor had a higher distribution of FCIGT1 voxels than a tumor that initially responded but regrew and a progressing tumor. The difference was less clear on R*2 maps, with a responding tumor and a tumor that first responded then progressed showing high values compared with the tumor that progressed immediately after treatment. The ROC curves for the three putative discriminators are plotted in Figure 8B, and the corresponding AUC values are tabulated in Table 4. R*2,90 provided an AUC below 0.5, indicating poor classification power. However, the compartmentation index metrics enabled good discrimination of RT versus NRT tumors, with AUCs of 0.75 and 0.86 for FCIGT1 and FCIGT1R*2,90, respectively. FCIGT1 R*2,90 had good accuracy on both training (84%) and LOOCV (79%).

    Performance of discriminators for classification of responding tumor (RT) versus nonresponding tumor (NRT) brain metastases. (A) Qualitative illustration of maps generated for three tumors with different responses to liposomal irinotecan treatment. Images depict a responding tumor (top row), a tumor that first responded then progressed (middle row), and a tumor that progressed (bottom row). For each tumor, computed and model-expected R1-weighted post-FMX MR images, their ratio, and reference transverse relaxation rate (R*2) maps are shown. Radiologist-drawn tumor contours are indicated by arrows. (B) The receiver operating characteristic plots of three putative discriminators of RT (nine lesions) versus NRT (10 lesions). FCIGT1 = fraction of pixels within a tumor volume of interest with compartmentation index greater than 1, FCIGT1R*2,90 = fraction of pixels within a tumor volume of interest with compartmentation index greater than 1 and R*2 in the 90th and higher percentile.

    Figure 8: Performance of discriminators for classification of responding tumor (RT) versus nonresponding tumor (NRT) brain metastases. (A) Qualitative illustration of maps generated for three tumors with different responses to liposomal irinotecan treatment. Images depict a responding tumor (top row), a tumor that first responded then progressed (middle row), and a tumor that progressed (bottom row). For each tumor, computed and model-expected R1-weighted post-FMX MR images, their ratio, and reference transverse relaxation rate (R*2) maps are shown. Radiologist-drawn tumor contours are indicated by arrows. (B) The receiver operating characteristic plots of three putative discriminators of RT (nine lesions) versus NRT (10 lesions). FCIGT1 = fraction of pixels within a tumor volume of interest with compartmentation index greater than 1, FCIGT1R*2,90 = fraction of pixels within a tumor volume of interest with compartmentation index greater than 1 and R*2 in the 90th and higher percentile.

    Download as PowerPoint

    Discussion

    FMX-enhanced MRI has been shown to correlate with macrophage content in tumors (23–25), which may be mechanistically relevant, as TAMs can affect vascular permeability, and thus, deposition of nal-IRI (37); may alter its phagocytosis; may serve as a drug depot; and may contribute to payload release with subsequent conversion of IRI to its active metabolite, SN-38 (9). In a phase 1 expansion study of participants with mBC, we investigated FMX-enhanced MRI as a pretreatment predictor of response to nal-IRI in both thoracoabdominal and brain metastases. The post-FMX versus pre-FMX increase in R*2 of thoracoabdominal tumors (ΔR*2 = 83 sec−1) was more than twice (P < .001) the increase observed in brain metastases (ΔR*2 = 33 sec−1), indicative of lower microvascular permeability in brain metastases compared with thoracoabdominal metastases in participants with mBC. Therefore, we assessed brain and extracranial metastases with different methods. From MRI performed 16–24 hours after FMX, we computed tumor R*2 normalized to spleen R*2 in thoracoabdominal metastases. A mean pretreatment spleen-normalized R*2 of tumor voxels in the 90th percentile or higher (rR*2,90) higher than 1.07 was predictive of response to nal-IRI with a mean AUC of 0.77 ± 0.1 (95% CI) and an LOOCV of 74%. In brain metastases, we computed a measure of intracellular compartmentation of FMX, which may result from TAM phagocytosis of the FMX, which was predictive of the response of brain metastases to nal-IRI. The FCIGT1R*2,90 metric provided 79% accuracy on LOOCV for discriminating between RT and NRT brain metastases, with RT having higher pretreatment values of FCFGT1R*2,90 than NRT.

    In thoracoabdominal metastases, unnormalized R*2 was not useful in the prediction of RT versus NRT in this study. Normalization of tumor R*2 to the mean R*2 in the spleen (rR*2) increased the predictive power for distinguishing between RT and NRT. Spleen R*2 was chosen as a reference for normalization because together with the liver, the spleen is a major reservoir of phagocytic cells and accumulates nanoparticles and liposomes, thus affecting relevant pharmacokinetics of FMX and nal-IRI in the systemic circulation and deposition characteristics into other tissues, including tumors. Although all participants received a fixed dose of 5 mg of iron per kilogram of body weight FMX (up to 510 mg of iron as a maximum dose), normalization of tumor R*2 to spleen R*2 measured in the same patient and imaging session will compensate for small patient-to-patient differences in FMX pharmacokinetics due to differences in blood volume (38) and activity of the mononuclear phagocyte system (39). A previously reported approach for quantification of FMX uptake into tissue requires both a calibration phantom and baseline pre-FMX scans for reference (23). In this study, we have investigated the feasibility of obviating the need for a pre-FMX scan on the hypothesis that R*2 values above some threshold are likely to reflect FMX accumulation rather than intrinsic magnetic susceptibility. Importantly, measurement of rR*2,90 requires only one FMX-enhanced MRI scan, performed 16–24 hours after administration of FMX, and does not require calibration of R*2 values against an FMX phantom.

    rR*2,90 is a reporter of combined passive and phagocytic accumulation of FMX nanoparticles in the tumor. An FMX-enhanced MRI biologic marker of pretreatment TAM content may be mechanistically predictive of a given response of the tumor to nal-IRI. We have sought to compute an index of FMX compartmentation by exploiting the fact that compartmentation within cells will have opposing effects on R1 and R*2 FMX MRI enhancement. In a recent clinical glioblastoma study, Barajas et al (40) used information from susceptibility weighted images acquired at three different phases of FMX distribution to compute segregation and extravascular localization of FMX imaging maps. The positive and negative segregation and extravascular localization of FMX imaging values were correlated with increasing TAM infiltration and increasing microvascular density regions, respectively. The segregation and extravascular localization of FMX imaging–positive metric correlated with the overall survival of the participants with glioblastoma. Computation of the compartmentation metrics requires scans from a pre-FMX and two post-FMX sessions within 24 hours. Additionally, linear scaling against FMX phantom tubes is also required to correct for receiver gain and coil sensitivity differences between the multiple MRI sessions. It is conceivable that the need for additional FMX-enhanced MRI and inclusion of FMX phantoms will be obviated if both R1 and R*2 maps are acquired during the 16- to 24-hour post-FMX session, though this remains to be investigated.

    We investigated intrasession repeatability of the R*2 measurement in a subset of six participants. The coefficient of variation was 6.5% for repeat measurements of R*2 in muscle, with excellent agreement over a range of R*2 values with Bland-Altman analysis. Our measurements of R*2 in different tissues before and after FMX showed good agreement with values reported by Stirrat et al (41). We also investigated the sensitivity of the predictive performance of rR*2,90 to perturbations of the radiologist-drawn tumor contours. Interestingly, the metric was relatively insensitive to dilation of the ROIs by 1 to 5 pixels in plane, but its performance was degraded by erosion of the manual ROIs by 1 to 5 pixels in plane. These results may reflect the abundance of FMX nanoparticles located at the tumor periphery. Thus, erring on the side of including more peritumoral tissue within the tumor contours is preferable for tumor response assessment with FMX-enhanced MRI.

    This study had some limitations. First, the study sample was relatively small, at 27 participants with 49 thoracoabdominal and 19 brain tumor lesions that were evaluable. Second, reproducibility was tested only for pre-FMX scans, and an initial assessment of generalizability of the pretreatment predictors of response was performed with LOOCV, but prospective testing is required to establish reproducibility and generalizability more rigorously. Third, there were minor participant-to-participant variations in voxel dimensions for a given anatomic site (brain or thoracoabdominal). Spatial resolution can affect R*2 contrast enhancement due to dependence of susceptibility on the underlying geometry. The echo times used in this study were less than 13.5 msec, and published studies indicate that susceptibility contrast changes are relatively insensitive to spatial resolution variations at short echo times (<20 msec) (42). While the spatial resolutions of brain and thoracoabdominal images were significantly different, the impracticality of obtaining spleen images for normalization in participants with only brain metastases meant that independent predictors were developed for metastatic lesions in the two anatomic locations.

    In conclusion, we showed pretreatment FMX-enhanced MRI as a marker of response to nal-IRI in participants with mBC with thoracoabdominal and brain metastases. In thoracoabdominal metastases, we showed that rR*2 computed from one 16–24-hour post-FMX-enhanced MRI scan can be used as a pretherapy response predictor; this could simplify clinical implementation and logistical concerns, as calibration phantoms or pre-FMX scans may not be needed. In brain metastases, we computed a novel compartmentation index based on the spoiled gradient-recalled echo signal equation applied to phantom-calibrated FMX-enhanced MRI scans that were acquired before FMX, 1–4 hours after FMX, and 16–24 hours after FMX, which had high accuracy on LOOCV for predicting response in brain metastases. We analyzed only pretherapy response prediction of individual tumor lesions, which may not necessarily equate to overall participant-level prediction, as we did not account for heterogeneity of response within the same participant, nontarget lesions, or appearance of new lesions. Any bearing to patient-level response prediction must be tested in future studies. Furthermore, follow-up testing should include acquisition of both R1 and R*2 maps at the 16–24-hour post-FMX time point to potentially prevent the need for phantom tubes and additional FMX-enhanced MRI time points.

    Disclosures of conflicts of interest: H.R. No relevant relationships. A.M.A.L. No relevant relationships. J.R.C. No relevant relationships. H.S.H. Research funding to institution from AbbVie, Arvinas, GSK, G1 Therapeutics, Quantum Leap Healthcare Collaborative, Marker, Pfizer, Zymeworks, Celcuity, and Department of Defense; payment for speakers bureaus from Lilly; advisory board for Novartis, AstraZeneca, and Gilead. D.K.J. R37 NCI/NIH grant 1R37CA229810-01A1 (Using Radiogenomics to Noninvasively Predict the Malignant Potential of Intraductal Papillary Mucinous Neoplasms of the Pancreas and Uncover Hidden Biology). S.G.K. Employee of Merrimack Pharmaceuticals (to March 2017) and Ipsen Bioscience, Merrimack was the initial clinical study sponsor; patent applications for treatment of breast cancer with liposomal irinotecan with Ipsen Biopharm; Ipsen stock and stock grants. J.C.S. Funding and drug support to institution to conduct the trial: Merrimack (and later Ipsen); funding and drug support for conduct of clinical trial to institution from Merck, Pfizer, Tesaro/GSK, Plexxikon, Corcept, AbbVie, Bolt Biotherapeutics, Immune Sensor Therapeutics, Syros Pharmaceuticals, Agenus, Sermonix, Ipsen, Arcus, Endocyte, Five Prime, and Baxter; honoraria for advisory board participation from Pfizer, Immunomedics, AstraZeneca, Novartis, Tempus, and Ipsen; stock options in Biosplice Therapeutics; employer-employee relationship with Biosplice Therapeutics. R.L.K. No relevant relationships. N.R. Funding was received from Ipsen Pharmaceuticals and Imaging Endpoints to institution to defray study-related costs. Some of the reported work was performed by the Quantitative Imaging Shared Service (Core) of the Moffitt Cancer Center, which is partly supported by funding from the NIH (P30 CA076292); grant funding that was unrelated to the work reported in this article: NIH: R01 CA249016 (MPI), U01 CA200464(co-I), U01 CA143062 (co-I), U54 CA193489 (co-I), State of Florida: “Live Like Bella Foundation” (co-I), Halozyme Therapeutics: 2-year research grant (PI), HealthMyne: 1-year research grant (co-I); Gillies RJ, Gatenby RA, Raghunand N, Arrington J, Stringfield O, Balagurunathan Y, Goldgof DB, Hall LO. Radiologically Identified Tumor Habitats. US Patent No. 10,827,945 (subject matter of the patent is unrelated to the work reported in this article).

    Acknowledgments

    The authors thank all participants involved in the study as well as their caregivers, care teams, investigators, and research staff at participating institutions. We thank Daniel D. Von Hoff, MD, for valuable consultations and advice throughout this study.

    Author Contributions

    Author contributions: Guarantor of integrity of entire study, N.R.; study concepts/study design or data acquisition or data analysis/interpretation, all authors; manuscript drafting or manuscript revision for important intellectual content, all authors; approval of final version of submitted manuscript, all authors; agrees to ensure any questions related to the work are appropriately resolved, all authors; literature research, H.R., J.R.C., S.G.K., N.R.; clinical studies, J.R.C., H.S.H., D.K.J., J.C.S., R.L.K., N.R.; experimental studies, N.R.; statistical analysis, H.R., A.M.A.L., N.R.; and manuscript editing, H.R., A.M.A.L., J.R.C., D.K.J., S.G.K., J.C.S., R.L.K., N.R.

    1 Current address: Biosplice Therapeutics, San Diego, Calif.

    This study was sponsored by Ipsen and supported in part by research funding From the National Institutes of Health (P30 CA076292, QI Core).

    Data sharing: Where patient data can be anonymized, Ipsen will share all individual participant data that underlie the results reported in this article with qualified researchers who provide a valid research question. Study documents, such as the study protocol and clinical study report, are not always available. Proposals should be submitted to [email protected] and will be assessed by a scientific review board. Data are available beginning 6 months and ending 5 years after publication; after this time, only raw data may be available.

    References

    • 1. Sung H, Ferlay J, Siegel RL, et al. Global cancer statistics 2020: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries. CA Cancer J Clin 2021;71(3):209–249. Crossref, Medline, Google Scholar
    • 2. Siegel RL, Miller KD, Jemal A. Cancer statistics, 2016. CA Cancer J Clin 2016;66(1):7–30. Crossref, Medline, Google Scholar
    • 3. Sachdev JC, Munster P, Northfelt DW, et al. Phase I study of liposomal irinotecan in patients with metastatic breast cancer: findings from the expansion phase. Breast Cancer Res Treat 2021;185(3):759–771. Crossref, Medline, Google Scholar
    • 4. Shigeoka Y, Itoh K, Igarashi T, et al. Clinical effect of irinotecan in advanced and metastatic breast cancer patients previously treated with doxorubicin- and docetaxel-containing regimens. Jpn J Clin Oncol 2001;31(8):370–374. Crossref, Medline, Google Scholar
    • 5. Noble CO, Krauze MT, Drummond DC, et al. Pharmacokinetics, tumor accumulation and antitumor activity of nanoliposomal irinotecan following systemic treatment of intracranial tumors. Nanomedicine (Lond) 2014;9(14):2099–2108. Crossref, Medline, Google Scholar
    • 6. Bernards N, Ventura M, Fricke IB, et al. Liposomal Irinotecan Achieves Significant Survival and Tumor Burden Control in a Triple Negative Breast Cancer Model of Spontaneous Metastasis. Mol Pharm 2018;15(9):4132–4138. Crossref, Medline, Google Scholar
    • 7. Mohammad AS, Griffith JI, Adkins CE, et al. Liposomal Irinotecan Accumulates in Metastatic Lesions, Crosses the Blood-Tumor Barrier (BTB), and Prolongs Survival in an Experimental Model of Brain Metastases of Triple Negative Breast Cancer. Pharm Res 2018;35(2):31. Crossref, Medline, Google Scholar
    • 8. Golombek SK, May JN, Theek B, et al. Tumor targeting via EPR: Strategies to enhance patient responses. Adv Drug Deliv Rev 2018;130:17–38. Crossref, Medline, Google Scholar
    • 9. Kalra AV, Kim J, Klinz SG, et al. Preclinical activity of nanoliposomal irinotecan is governed by tumor deposition and intratumor prodrug conversion. Cancer Res 2014;74(23):7003–7013. Crossref, Medline, Google Scholar
    • 10. Prabhakar U, Maeda H, Jain RK, et al. Challenges and key considerations of the enhanced permeability and retention effect for nanomedicine drug delivery in oncology. Cancer Res 2013;73(8):2412–2417. Crossref, Medline, Google Scholar
    • 11. Ejigah V, Owoseni O, Bataille-Backer P, Ogundipe OD, Fisusi FA, Adesina SK. Approaches to Improve Macromolecule and Nanoparticle Accumulation in the Tumor Microenvironment by the Enhanced Permeability and Retention Effect. Polymers (Basel) 2022;14(13):2601. Crossref, Medline, Google Scholar
    • 12. Miller MA, Zheng YR, Gadde S, et al. Tumour-associated macrophages act as a slow-release reservoir of nano-therapeutic Pt(IV) pro-drug. Nat Commun 2015;6(1):8692. Crossref, Medline, Google Scholar
    • 13. Karathanasis E, Park J, Agarwal A, et al. MRI mediated, non-invasive tracking of intratumoral distribution of nanocarriers in rat glioma. Nanotechnology 2008;19(31):315101. Crossref, Medline, Google Scholar
    • 14. Toth GB, Varallyay CG, Horvath A, et al. Current and potential imaging applications of ferumoxytol for magnetic resonance imaging. Kidney Int 2017;92(1):47–66. Crossref, Medline, Google Scholar
    • 15. Zhang YR, Lin R, Li HJ, He WL, Du JZ, Wang J. Strategies to improve tumor penetration of nanomedicines through nanoparticle design. Wiley Interdiscip Rev Nanomed Nanobiotechnol 2019;11(1):e1519. Crossref, Medline, Google Scholar
    • 16. Kalra AV, Spernyak J, Kim J, et al. Magnetic resonance imaging with an iron oxide nanoparticle demonstrates the preclinical feasibility of predicting intratumoral uptake and activity of MM-398, a nanoliposomal irinotecan (nal-IRI). Cancer Res 2014;74(19_Supplement):2065. Crossref, Google Scholar
    • 17. Miller MA, Gadde S, Pfirschke C, et al. Predicting therapeutic nanomedicine efficacy using a companion magnetic resonance imaging nanoparticle. Sci Transl Med 2015;7(314):314ra183. Crossref, Medline, Google Scholar
    • 18. Kordbacheh H, Baliyan V, Parakh A, Wojtkiewicz GR, Hedgire S, Harisinghani MG. Pictorial review on abdominal applications of ferumoxytol in MR imaging. Abdom Radiol (NY) 2019;44(10):3273–3284. Crossref, Medline, Google Scholar
    • 19. Dousset V, Ballarino L, Delalande C, et al. Comparison of ultrasmall particles of iron oxide (USPIO)-enhanced T2-weighted, conventional T2-weighted, and gadolinium-enhanced T1-weighted MR images in rats with experimental autoimmune encephalomyelitis. AJNR Am J Neuroradiol 1999;20(2):223–227. Medline, Google Scholar
    • 20. Pohlmann A, Karczewski P, Ku MC, et al. Cerebral blood volume estimation by ferumoxytol-enhanced steady-state MRI at 9.4 T reveals microvascular impact of α1 -adrenergic receptor antibodies. NMR Biomed 2014;27(9):1085–1093. Crossref, Medline, Google Scholar
    • 21. Petronek MS, Spitz DR, Buettner GR, Allen BG. Oxidation of ferumoxytol by ionizing radiation releases iron. An electron paramagnetic resonance study. J Radiat Res (Tokyo) 2022;63(3):378–384. Crossref, Medline, Google Scholar
    • 22. McConnell HL, Schwartz DL, Richardson BE, Woltjer RL, Muldoon LL, Neuwelt EA. Ferumoxytol nanoparticle uptake in brain during acute neuroinflammation is cell-specific. Nanomedicine 2016;12(6):1535–1542. Crossref, Medline, Google Scholar
    • 23. Ramanathan RK, Korn RL, Raghunand N, et al. Correlation between Ferumoxytol Uptake in Tumor Lesions by MRI and Response to Nanoliposomal Irinotecan in Patients with Advanced Solid Tumors: A Pilot Study. Clin Cancer Res 2017;23(14):3638–3648. Crossref, Medline, Google Scholar
    • 24. Aghighi M, Theruvath AJ, Pareek A, et al. Magnetic Resonance Imaging of Tumor-Associated Macrophages: Clinical Translation. Clin Cancer Res 2018;24(17):4110–4118. Crossref, Medline, Google Scholar
    • 25. Iv M, Samghabadi P, Holdsworth S, et al. Quantification of Macrophages in High-Grade Gliomas by Using Ferumoxytol-enhanced MRI: A Pilot Study. Radiology 2019;290(1):198–206. Link, Google Scholar
    • 26. Karathanasis E, Chan L, Balusu SR, et al. Multifunctional nanocarriers for mammographic quantification of tumor dosing and prognosis of breast cancer therapy. Biomaterials 2008;29(36):4815–4822. Crossref, Medline, Google Scholar
    • 27. Kim M, Kizilbash SH, Laramy JK, et al. Barriers to Effective Drug Treatment for Brain Metastases: A Multifactorial Problem in the Delivery of Precision Medicine. Pharm Res 2018;35(9):177. Crossref, Medline, Google Scholar
    • 28. Blethen KE, Arsiwala TA, Fladeland RA, et al. Modulation of the blood-tumor barrier to enhance drug delivery and efficacy for brain metastases. Neurooncol Adv 2021;3(Suppl 5):v133–v143. Medline, Google Scholar
    • 29. Harney AS, Arwert EN, Entenberg D, et al. Real-Time Imaging Reveals Local, Transient Vascular Permeability, and Tumor Cell Intravasation Stimulated by TIE2hi Macrophage-Derived VEGFA. Cancer Discov 2015;5(9):932–943. Crossref, Medline, Google Scholar
    • 30. Eisenhauer EA, Therasse P, Bogaerts J, et al. New response evaluation criteria in solid tumours: revised RECIST guideline (version 1.1). Eur J Cancer 2009;45(2):228–247. Crossref, Medline, Google Scholar
    • 31. Brisset JC, Desestret V, Marcellino S, et al. Quantitative effects of cell internalization of two types of ultrasmall superparamagnetic iron oxide nanoparticles at 4.7 T and 7 T. Eur Radiol 2010;20(2):275–285. Crossref, Medline, Google Scholar
    • 32. Knobloch G, Colgan T, Wiens CN, et al. Relaxivity of Ferumoxytol at 1.5 T and 3.0 T. Invest Radiol 2018;53(5):257–263. Crossref, Medline, Google Scholar
    • 33. Ruchalski K, Braschi-Amirfarzan M, Douek M, et al. A Primer on RECIST 1.1 for Oncologic Imaging in Clinical Drug Trials. Radiol Imaging Cancer 2021;3(3):e210008. Link, Google Scholar
    • 34. Unal I. Defining an Optimal Cut-Point Value in ROC Analysis: An Alternative Approach. Comput Math Methods Med 2017;2017:3762651. Crossref, Medline, Google Scholar
    • 35. Bland JM, Altman DG. Statistical methods for assessing agreement between two methods of clinical measurement. Lancet 1986;1(8476):307–310. Crossref, Medline, Google Scholar
    • 36. Violin Plots for Matlab. https://github.com/bastibe/Violinplot-Matlab. Published 2016. Accessed November 2022. Google Scholar
    • 37. Miller MA, Chandra R, Cuccarese MF, et al. Radiation therapy primes tumors for nanotherapeutic delivery via macrophage-mediated vascular bursts. Sci Transl Med 2017;9(392):eaal0225. Crossref, Medline, Google Scholar
    • 38. Nadler SB, Hidalgo JH, Bloch T. Prediction of blood volume in normal human adults. Surgery 1962;51(2):224–232. Medline, Google Scholar
    • 39. Tavares AJ, Poon W, Zhang YN, et al. Effect of removing Kupffer cells on nanoparticle tumor delivery. Proc Natl Acad Sci USA 2017;114(51):E10871–E10880. Crossref, Medline, Google Scholar
    • 40. Barajas RF Jr, Schwartz D, McConnell HL, et al. Distinguishing Extravascular from Intravascular Ferumoxytol Pools within the Brain: Proof of Concept in Patients with Treated Glioblastoma. AJNR Am J Neuroradiol 2020;41(7):1193–1200. Crossref, Medline, Google Scholar
    • 41. Stirrat CG, Alam SR, MacGillivray TJ, et al. Ferumoxytol-enhanced magnetic resonance imaging methodology and normal values at 1.5 and 3T. J Cardiovasc Magn Reson 2016;18(1):46. Crossref, Medline, Google Scholar
    • 42. Deistung A, Rauscher A, Sedlacik J, Stadler J, Witoszynskyj S, Reichenbach JR. Susceptibility weighted imaging at ultra high magnetic field strengths: theoretical considerations and experimental results. Magn Reson Med 2008;60(5):1155–1168. Crossref, Medline, Google Scholar

    Article History

    Received: Feb 18 2022
    Revision requested: Apr 6 2022
    Revision received: Nov 10 2022
    Accepted: Dec 20 2022
    Published online: Feb 03 2023