Incremental Utility of First Pass Perfusion-CMR for Prognostic Risk Stratification of Cancer-Associated Cardiac Masses

Abstract

Background:

CMR differentiates C_MET and C_THR based on tissue characteristics stemming from vascularity on late gadolinium enhancement (LGE). Perfusion-CMR can assess magnitude of vascularity; utility for C_MASS is unknown.

Objectives:

To determine if perfusion-CMR provides diagnostic and prognostic utility for cardiac masses (C_MASS) beyond binary differentiation of metastasis (C_MET) and thrombus (C_THR).

Methods:

The population comprised adult cancer patients with C_MASS on CMR; C_MET and C_THR were defined using LGE-CMR: C_MASS+ patients were matched to (C_MASS-) controls for cancer type/stage. First pass perfusion-CMR was interpreted visually and semi-quantitatively for C_MASS vascularity, including contrast enhancement ratio (CER: [plateau vs. baseline]) and uptake rate (CUR [slope]). Follow-up was performed for all-cause mortality.

Results:

462 cancer patients were studied, including patients with (C_MET=173, C_THR=69) and without C_MASS on LGE-CMR: On perfusion-CMR, CER and CUR were higher within C_MET vs. C_THR (p<0.001); CUR yielded better performance (AUC 0.89–0.93) than CER (0.66–0.72; both p<0.001) to differentiate LGE-CMR evidenced C_MET and C_THR, although both CUR (p=0.10) and CER (p=0.01) typically misclassified C_MET with minimal enhancement. During follow-up, mortality among C_MET patients was high but variable; 47% of patients were alive 1-year post-CMR. Patients with semi-quantitative perfusion-CMR evidenced C_MET had higher mortality than controls (HR=1.42 [CI 1.06–1.90], p=0.02), paralleling visual perfusion-CMR (HR=1.47 [1.12–1.94], p=0.006) and LGE-CMR (HR=1.52 [1.16–2.00], p=0.003): Among patients with C_MET on LGE-CMR, mortality was highest among patients (p=0.002) with lesions in the bottom perfusion (CER) tertile, corresponding to low vascularity. Among C_MET and cancer-matched controls, mortality was equivalent (p=NS) among patients with lesions in upper CER tertile (corresponding to higher lesion vascularity). Conversely, patients with CMET in the middle (p=0.03) and lowest CER tertiles (lowest vascularity) had increased mortality (p=0.001).

Conclusions:

Perfusion-CMR yields prognostic utility that complements LGE-CMR: Among cancer patients with LGE-CMR defined C_MET, mortality increases in proportion to magnitude of lesion hypoperfusion.

Introduction

Over 19 million people worldwide are diagnosed with cancer annually,(1) among whom survival has markedly increased - highlighting the importance of strategies to refine risk stratification for cancer complications: Cardiac masses (C_MASS) are a leading indication for imaging cancer patients, among whom differentiation of metastasis (C_MET) from thrombus (C_THR) is key to inform prognosis and therapy. Late gadolinium enhancement (LGE) cardiovascular magnetic resonance (CMR) tissue characterization has been well-validated as a reference for diagnostic differentiation of cardiac neoplasm and thrombus – including prior work by our group(2–7) and others(8–10) in which it has been shown to predict outcomes. Beyond differentiation between broad diagnostic categories, a key rationale for imaging of masses concerns prognostic assessment: LGE-CMR assesses presence of enhancement to differentiate CMET and CTHR, but magnitude of enhancement can be problematic as this method entails static imaging late after contrast infusion (during which contrast has been shown capable of traversing between vascular and avascular regions). This limitation is important, as C_MET can vary with respect to magnitude of vascularity, and inadequate vascular supply to neoplastic tissue (i.e. tumor necrosis) is a known marker of aggressiveness.(11–13) In this context, imaging of contrast kinetics by perfusion-CMR holds the potential to further advance diagnostic and prognostic risk assessment for patients with C_MET.

Perfusion-CMR can assess dynamic contrast enhancement to measure vascularity: The notion that vascular supply to a given region parallels magnitude of enhancement on perfusion imaging is widely accepted in the context of myocardium.(14) Given that CMET and CTHR intrinsically differ with respect to tissue properties stemming from vascularity, perfusion-CMR would be expected to differentiate between these diagnoses. Perfusion could also provide unique insights regarding magnitude of vascularity within tumors, thus enabling risk stratification that extends beyond binary diagnostic categories. Utility of perfusion-CMR for cardiac masses is acknowledged in consensus statements(15,16), but primary literature supporting this premise is limited to case reports or studies in which perfusion was tested together with other CMR approaches(8,9) – prohibiting assessment of its intrinsic utility.

This study tested utility of perfusion-CMR among a cancer registry of patients with cardiac masses and controls matched for cancer etiology/stage: Perfusion-CMR was used to assess lesion vascularity, as was LGE-CMR. Goals were to test 1) diagnostic performance of semi-quantitative and visual perfusion-CMR analyses to differentiate C_MET and C_THR established by the reference of LGE-CMR; 2) whether perfusion-based partitions to distinguish C_MET from C_THR confer differential prognosis; and 3) if mortality risk among cancer patients with C_MET varies in proportion to lesion avascularity on perfusion-CMR.

Methods

Study Population

The population was derived from a bi-institutional (MSKCC, Cornell) registry of adult patients (≥18 years old) with systemic (extra-cardiac) cancer and CMR-evidenced cardiac masses (C_MASS) who underwent a standardized imaging protocol, inclusive of LGE and perfusion. C_MASS+ patients were matched (1:1) to patients without C_MASS on CMR, but equivalent primary cancer type (primary organ) and disease stage. Additional registry details have been previously reported.(4–7) Patients with primary cardiac malignancies (i.e. sarcoma) were not eligible for this study. Systemic cancer registry patients with C_MASS+ were also excluded from this study if perfusion-CMR was not performed as part of the imaging protocol (n=52) or CMASS was not visualized on perfusion-CMR (n=12) due to prominent lesion mobility or inadequate spatial coverage (prohibiting tissue characterization analyses), and/or lack of cancer care at enrolling sites (MSKCC, Cornell) prohibiting adequate clinical characterization (n=11). Patients with C_MASS (on LGE) in whom perfusion-CMR was concomitantly performed as part of the imaging protocol were similar to those without perfusion with respect to age, sex, LVEF, and clinical prognosis (all p=NS).

This study entailed analysis of data acquired for clinical purposes between 2012–22. Ethics approval was provided by the Memorial Sloan Kettering Cancer Center and Weill Cornell Institutional Review Boards, which approved waivers of informed consent for analysis of preexisting clinical data.

Image Acquisition

CMR was performed on commercial (1.5T [89%], 3.0T [11%]) scanners (General Electric [GE], Waukesha, WI). Exams included ECG-gated cine, LGE, and perfusion components. Cine-CMR utilized a steady-state free precession (SSFP) pulse sequence acquired in contiguous LV short (mitral annulus - apex) and long axis (2, 3, 4 chamber) orientations. LGE-CMR utilized an inversion recovery pulse sequence. Conventional (inversion time [TI] ~300msec) and “long TI” (600msec) LGE-CMR were used as the reference for C_MASS vascularity concordant with prior methods applied and validated by our group.(2,3,5,6,17) Conventional LGE-CMR was acquired in all patients; additional breath holds for supplemental long TI imaging were tolerated in 98% (236/242) of C_MASS + patients (97% C_THR, 98% C_MET). Long TI LGE was typically acquired 5–10 minutes post-contrast administration and conventional LGE thereafter (10–30 minutes).

Perfusion-CMR utilized an ECG-gated gradient echo pulse sequence. To standardize data interpretation, images were obtained in long axis (4-chamber) orientation; contiguous slices (~8 mm thickness) were acquired throughout the LV. Perfusion-CMR (typical TR: 2.0 msec, TE: 1.3 msec, TI: 105, flip angle 20°, FOV: 40 cm, 256 × 256 pixels, in plane spatial resolution 1.6 mm) was acquired in standardized orientations (4 chamber [3–4 slices; thickness 8–10 mm, spacing 8–10 mm]). Imaging was initiated at time of gadolinium injection (0.1 mmol kg, ~2cc/sec); and continued ~30–40 seconds thereafter (total contrast dose 0.2 mmol/kg, inclusive of first pass perfusion and subsequent infusion prior to LGE).

Image Analysis

C_MASS was initially localized on cine-CMR, on which it was defined via established criteria as a discrete tissue prominence independent from normal anatomic structures,(2,3,5,6,17,18) quantified based on size (area, cross-sectional linear dimensions), and further categorized by type (C_MET, C_THR) based on tissue characterization:

LGE-CMR:

C_MET and C_THR were differentiated based on presence of contrast enhancement via the reference standard of LGE-CMR using established criteria previously validated by our group and others:(2,3,5–7,17,18) In brief, C_MET was defined as a discrete mass with vascular tissue properties (enhancement present),(5,6) and C_THR defined based on avascularity (enhancement absent).(2,3,17,18) C_MET were further sub-typed based on visually scored magnitude of enhancement (diffusely hyper-enhancing, mixed, predominantly hypo-enhancing) in accordance with established methods previously used by our group.(7)

Perfusion-CMR:

First pass perfusion was used to measure magnitude of contrast enhancement based on semi-quantitative analysis using the following standardized algorithm:

• Perfusion datasets were initially reviewed to identify C_MASS location, which was confirmed based on co-localization with cine-CMR.

• Regions of interest (ROI) were then placed to encompass lesion borders, automatically propagated, and manually adjusted to optimize lesion tracking throughout contrast infusion to calculate contrast uptake rate (defined as slope of signal intensity curve during initial contrast infusion) and contrast enhancement ratio (defined as [plateau – baseline / baseline] signal intensity), both as absolute and (blood pool) normalized indices: Blood pool data was obtained from ROIs drawn in the systemic blood pool (LV cavity or aortic root), so as to calculate normalized CUR and CER based on lesion specific data (e.g. normalized CER = C_MASS CER / blood CER).

For patients with multiple masses, the largest lesion (maximal diameter) was used for analysis. LGE and perfusion-CMR analyses were performed independently (blinded to results of the other method).

To test relative value of semi-quantitative and qualitative (visual) analyses, perfusion-CMR datasets were independently analyzed using a qualitative approach approximating that of the above semi-quantitative algorithm. To do so, lesions were interpreted by consensus of three readers (ATC, JK, JWW) for diagnostic etiology (CMET or CTHR) based on presence or absence of contrast enhancement. For cases designated as CMET, lesions were scored for magnitude of first pass contrast uptake using three strata (grade III: minimal [<50% of lesion area], grade II: mixed [≥50% of lesion demonstrating contrast uptake; scattered foci of hypoperfusion], grade I: diffuse contrast uptake).

Ancillary analyses included quantification of cardiac chamber size and function, which was measured on cine-CMR using established methods as have been shown to yield high reproducibility by our lab.(19,20)

Clinical Characterization

Electronic medical records were reviewed to ascertain cardiovascular risk factors, cancer etiology and therapeutic indices, and discern all-cause mortality following CMR: Follow-up date was calculated based on interval between baseline CMR and latest institutional follow-up visit. Pathology records and associated samples (in patients with surgical excision or biopsy of cardiac and/or contiguous extra-cardiac tissue) were extracted from data archives and reviewed for exploratory analysis to support CMR diagnostic classification of C_MET and compare perfusion results to histopathology evidenced tissue properties.

Statistical Methods

Comparisons between groups were made using Student’s t test (expressed as mean ± standard deviation) for normally distributed continuous variables. Non-normally distributed variables (median, interquartile range) were compared via the Mann-Whitney U test. Categorical variables were compared using Chi-square or, when fewer than 5 expected outcomes per cell, Fisher’s exact test: Paired testing (t-tests or McNemar’s tests) was employed for matched case-control comparisons. The Kaplan-Meier method was used to estimate survival; follow-up duration was reported as median (IQR). Cox proportional hazards models compared mortality risk between groups. Receiver operating characteristics (ROC) analysis including area under the curve (AUC) estimates were used to evaluate overall diagnostic test performance of perfusion indices (absolute and normalized contrast upslope), and derive cutoffs (maximal sensitivity, specificity) for C_MET as established using the reference of LGE-CMR. Univariable regression analyses were used to identify predictors of all-cause mortality, which were then further tested in multivariable regression analyses. Calculations were performed using SPSS v26 (SPSS Inc. [Chicago, IL]). Two-sided p<0.05 was deemed indicative of statistical significance.

Results

Population Characteristics

The study population comprised 462 cancer patients who underwent CMR, including patients with cardiac masses (C_MASS) and controls matched (1:1) for cancer etiology/stage. Table 1 details population characteristics together with comparisons between patients with each C_MASS type (C_MET, C_THR) and their respective controls. As shown, cancer diagnosis varied: Among patients with C_MASS, one-fifth (21%) of patients had primary non-cardiac sarcoma with an array of primary cancers (genitourinary, gastrointestinal, lymphoma) near evenly represented (13–18%). C_MET and C_THR patients were similar to respective controls with respect to age, clinical indices, and anti-cancer regimen.

Table 1.

Population Characteristics

	Overall (n=462)	Cardiac Metastasis		p†	Cardiac Thrombus		p†
		MET + (n=173)	MET – (n=157)		THR + (n=69)	THR – (n=63)
Clinical Characteristics
Age (years)	58 ± 16	60 ± 15	59 ± 16	0.97	52 ± 16	57 ± 16	0.06
Sex (male)	54% (250)	54% (94)	55% (86)	0.91	48% (33)	59% (37)	0.30
BSA (kg/m2)	1.9 ± 0.3	1.8 ± 0.3	1.9 ± 0.3	0.31	1.9 ± 0.2	1.9 ± 0.3	0.85
Cancer Etiologies *
Sarcoma	21% (98)	24% (42)	24% (38)	-	15% (10)	13% (8)	-
Gastrointestinal	18% (82)	13% (22)	14% (22)	-	28% (19)	30% (19)	-
Genitourinary	13% (62)	14% (24)	13% (21)	-	15% (10)	11% (7)	-
Lymphoma	13% (59)	10% (17)	11% (17)	-	19% (13)	19% (12)	-
Lung	10% (48)	13% (22)	13% (20)	-	4% (3)	5% (3)	-
Skin/Melanoma	7% (32)	10% (17)	10% (15)	-	0% (0)	0% (0)	-
Stage IV Cancer	93% (431)	100%(173)	100% (157)	-	77% (53)	76% (48)	-
Anti-Cancer Regimen
Chemotherapy
Alkylating Agents	48% (222)	41% (71)	43% (68)	0.69	70% (48)	56% (35)	0.19
Antimetabolites	36% (165)	31% (54)	31% (48)	0.89	51% (35)	44% (28)	0.23
Antimicrotubular Agents	46% (213)	37% (64)	41% (64)	0.52	70% (48)	59% (37)	0.21
Anthracycline	26% (122)	20% (35)	26% (40)	0.24	41% (28)	30% (19)	0.24
Kinase Inhibitors	9% (42)	10% (18)	11% (17)	1.00	7% (5)	3% (2)	0.38
Immunotherapy	16% (73)	16% (28)	22% (35)	0.08	9% (6)	6% (4)	0.69
Radiation Therapy	42% (192)	49% (85)	38% (59)	0.03	33% (23)	40% (25)	0.46
Mediastinal RT	10% (48)	17% (29)	8% (13)	0.01	1% (1)	8% (5)	-
Coronary Artery Disease	11% (50)	15% (25)	6% (10)	0.06	9% (6)	14% (9)	0.58
Atherosclerosis Risk
Factors
Hypertension	36% (168)	39% (67)	39% (61)	1.00	25% (17)	37% (23)	0.08
Hypercholesterolemia	29% (132)	30% (52)	31% (48)	0.90	22% (15)	27% (17)	0.68
Diabetes Mellitus	13% (59)	12% (21)	15% (24)	0.50	7% (5)	14% (9)	0.39
Tobacco Use	33% (154)	34% (59)	34% (54)	1.00	36% (25)	25% (16)	0.17
Cardiac Morphology and Function
Left Ventricle  Ejection fraction (%)  End-diastolic volume (mL)  End-systolic volume (mL)  Stroke volume (mL)  Myocardial mass (gm)  End-diastolic diameter (cm) Right Ventricle  Ejection fraction (%)  End-diastolic volume (mL)  End-systolic volume (mL)  Stroke volume (mL)	60 ± 12 125 ± 44 53 ± 33 73 ± 21 114 ± 43 4.8 ± 0.7 54 ± 9 132 ± 43 63 ± 29 69 ± 21	62 ± 10 113 ± 33 44 ± 20 69 ± 20 108 ± 39 4.6 ± 0.5 54 ± 9 124 ± 36 58 ± 23 66 ± 20	60 ± 13 132 ± 50 57 ± 40 75 ± 22 117 ± 45 4.9 ± 0.8 53 ± 9 136 ± 50 67 ± 36 69 ± 23	0.07 <0.001 <0.001 0.03 0.13 <0.001 0.47 0.03 0.02 0.24	61 ± 10 128 ± 41 53 ± 33 76 ± 22 105 ± 41 4.9 ± 0.6 56 ± 8 133 ± 41 60 ± 24 73 ± 22	55 ± 15 139 ± 47 66 ± 39 73 ± 23 133 ± 47 5.0 ± 0.8 52 ± 11 139 ± 40 69 ± 28 70 ± 21	0.02 0.19 0.03 0.51 0.001 0.13 0.01 0.43 0.04 0.27
Atria  Left atrial area (cm2)  Right atrial area (cm2)	20 ± 7 18 ± 6	19 ± 6 18 ± 5	20 ± 7 20 ± 7	0.03 0.68	21 ± 7 17 ± 5	21 ± 7 19 ± 7	0.93 0.15

^*Other cancer etiologies: endocrine (6% [n=26]), breast (6% [n=26]), leukemia (4% [n=17]), and head/neck (3% [n=12]).

^†Matching not possible in 22 patients (16 C_MET, 6 CT_HR) due to non-equivalent cancer diagnosis/stage among controls undergoing CMR.

Regarding anatomic distribution, C^MET distribution varied widely (30% LV | 27% LA | 32% RV | 36% RA | 30% pericardium); 28% of patients had lesions in multiple cardiac chambers whereas nearly all (90%) C_THR localized to the right atrium. Notably, while LV and RV chamber size were smaller in patients with C_MET compared to controls (p<0.05) – a finding potentially attributable to metastatic chamber invasion, systolic function (EF) was similar between groups (p=NS).

Perfusion-CMR Contrast Kinetics

Table 2 compares perfusion-CMR measured contrast kinetics between C_MET and C_THR as defined using the reference of LGE-CMR. As shown, contrast uptake rate during first pass gadolinium infusion was higher within C_MET irrespective of whether analyzed based on absolute or blood pool normalized data (both p<0.001). Contrast enhancement ratio also yielded differences between C_MET compared to C_THR (p<0.001), although magnitude of difference was less (~2-fold) than that of contrast uptake (> 5-fold). Of note, Table 2 also demonstrates that differential tissue properties of C_MET and C_THR were not attributable to differential contrast dosing, hemodynamic indices, or LV systolic function between groups (all p=NS). Figure 1 provides representative examples of cardiac masses assessed by perfusion-CMR, as well as corresponding semi-quantitative data for which dynamic contrast enhancement differentiated C_MET from C_THR.

Table 2.

Perfusion Contrast Kinetics among Patients with Cardiac Metastasis and Thrombus

	Cardiac Metastasis (n=173)	Cardiac Thrombus (n=69)	p
Perfusion CMR  Contrast uptake rate (CUR)   Absolute   Normalized (blood pool, %)    Blood Pool Contrast Enhancement Ratio (CER)  Absolute  Normalized (blood pool, %)    Blood Pool	9.4 [5.5–18.1] 9.0 [4.9–16.7] 120.0 [64.7–207.5] 1.3 [0.8–2.4] 15.7 [10.3–24.2] 9.5 [5.6–15.2]	1.5 [0.3–3.0] 1.0 [0.3–3.2] 114.5 [57.4–210.6] 0.7 [0.3–1.8] 8.7 [4.2–17.2] 8.3 [5.9–15.2]	<0.001 <0.001 0.74 <0.001 <0.001 0.94
Gadolinium Dose (ml)	28.7 ± 7.7	29.4 ± 6.9	0.48
Heart Rate	85.9 ± 17.7	81.2 ± 16.6	0.07
Stroke Volume (LV [ml])	69.1 ± 20.0	75.5 ± 21.7	0.03
Blood Pressure †
Systolic Blood Pressure (mm Hg)	120.2 ± 17.2	119.8 ± 16.6	0.88
Diastolic Blood Pressure (mm Hg)	75.2 ± 10.1	75.4 ± 13.9	0.94
Mean Arterial Pressure (mm Hg)	90.3 ± 10.8	90.2 ± 12.2	0.97

^*Cardiac metastasis and thrombus defined using the reference standard of LGE-CMR

^†Measured non-invasively via brachial cuff within 24 hours of CMR (available in 91% [91% C_MET, 91% C_THR]).

Figure 1. Representative Examples.

Representative examples of left (1A) and right (1B) sided C_MET and C_THR as respectively assessed by LGE and perfusion-CMR (top), as well as corresponding perfusion analyses to measure contrast kinetics (bottom). In brief, regions of interest (ROI) were placed over lesions and tracked throughout the cardiac cycle so as to measure slope (contrast uptake rate [CUR]) and plateau enhancement (contrast enhancement ratio [CER]).

Note that for both examples, C_MET (blue) demonstrated higher contrast enhancement than did C_THR (purple), consistent with presence of vascular supply. However, a range of vascularity was evident among C_MET cases for which lesions with minimal enhancement on LGE-CMR (center; dark blue outline) demonstrated lower magnitude of semi-quantitative perfusion than did diffusely enhancing lesions (left; light blue outline), corresponding to lesser differentiation from C_THR.

Diagnostic Performance of Perfusion-CMR

Perfusion-CMR indices were tested with respect to differential approaches and optimal diagnostic cutoffs to delineate between C_MET and C_THR. Figure 2 provides ROC curves for semi-quantitative perfusion cutoffs in relation to the reference of LGE-CMR evidenced C_MET. As shown, absolute and blood pool normalized CUR provided higher overall diagnostic performance (AUC: 0.89-0.93) than did CER (AUC: 0.66–0.72; both p<0.001). Table 3 reports individual diagnostic parameters as derived using cutoffs for maximal sensitivity and specificity: Note, absolute CUR yielded highest overall accuracy (87%), but sensitivity (84%) and negative predictive value (71%) were suboptimal, attributable to substantial false negatives (n=27). Table 3 also demonstrates that qualitative (visual) perfusion-CMR interpretation yielded slightly lower overall diagnostic performance to that of semi-quantitative analysis, as evidenced by decreased accuracy (83%) to that of semi-quantitative CMR (maximal accuracy [absolute CUR] 87%).

Figure 2. Perfusion-CMR for Differentiation of Cardiac Metastasis and Thrombus.

Receiver operating characteristics (ROC) curves for perfusion-CMR cutoffs in relation to C_MET as established by the reference of LGE-CMR (red = contrast uptake rate [CUR], blue = enhancement ratio [CER]; solid and dashed lines correspond to absolute and blood pool normalized data for each).

Note good overall diagnostic performance for semi-quantitative CUR (AUC 0.89–0.93), which was higher than that of CER (0.66–0.72). Corresponding diagnostic test parameters derived from ROC curves shown in Table 3.

Table 3.

Diagnostic Performance of Perfusion-CMR to Differentiate Metastasis vs. Thrombus

	Sensitivity	Specificity	Accuracy	Positive Predictive Value	Negative Predictive Value
Semi-Quantitative Analysis
Contrast Uptake Rate
CMASS	84% (146/173)	94% (65/69)	87% (211/242)	97% (146/150)	71% (65/92)
CMASS/blood	93% (160/173)	73% (50/69)	87% (210/242)	89% (160/179)	79% (50/63)
Contrast Enhancement Ratio
CMASS	76% (131/173)	54% (37/69)	69% (168/242)	80% (131/163)	47% (37/79)
CMASS/blood	73% (126/173)	65% (45/69)	71% (171/242)	84% (126/150)	49% (45/92)
Qualitative (visual) Analysis
CMASS	88% (152/173)	70% (48/69)	83% (200/242)	88% (152/173)	70% (48/69)

^*Diagnostic indices (e.g. sensitivity) calculated in relation to presence of cardiac metastasis as defined using the reference standard of LGE-CMR cut off CUR (C_MASS = 4.34, C_MASS/blood = 2.64%), CER (C_MASS = 0.80, C_MASS/blood = 11.56%)

Among the overall cohort of patients with LGE-CMR evidenced C_MET, 32 had pathology analyses either via surgical excision or biopsy of cardiac and/or contiguous extra-cardiac tissue (median interval from CMR was 0.05 [IQR −0.86–0.77] months): Of these, all had pathology evidenced neoplasms (97% [31/32] had pathological confirmation of primary cancer type). Among this subgroup of patients with pathology, semi-quantitative perfusion correctly classified lesions as neoplastic in 75% (n=24/32) and visual perfusion in 88% (n=28/32). Figure 3 provides representative examples of perfusion-CMR enhancement patterns in relation to pathology evidenced tissue characterization, for which results demonstrated C_MET with minimal enhancement to correspond to lesions with tumor necrosis or prominent cellularity on pathology, and prominent enhancement on CMR to reflect prominent vascularity on pathology.

Figure 3. C_MET Perfusion on CMR in Relation to Pathology-Evidenced Lesion Vascularity.

Representative examples of C_MET with variable magnitude of perfusion (top) together with semi-quantitative perfusion-CMR analysis (middle) and pathology-evidenced lesion vascularity (bottom). As shown, variability in contrast kinetics on perfusion-CMR paralleled differential vascular and cell density on pathology.

3A: Two cases of minimally enhancing CMET on perfusion-CMR, including thymoma (left) for which pathology demonstrated high tumor cell burden with minimal vascular content and melanoma (right) for which pathology demonstrated scattered vascular content (melanoma) with prominent tumor necrosis.

3B: Two cases of prominently enhancing CMET on perfusion-CMR (left: neuroendocrine, right: spindle cell sarcoma), both of which had high vascular density on pathology.

Legend: asterisk = tumor necrosis, dagger = myocardium; inserts = high power views of tumor components (pathology images obtained using hematoxylin and eosin staining).

To explore whether diagnostic performance varied in relation to magnitude of enhancement, perfusion-CMR performance was tested in relation to lesion sub-type on LGE-CMR: Two thirds (58%) of predominantly hypo-enhancing lesions (on LGE) were misclassified as CTHR based on CER, a rate nearly 3-fold higher compared to mixed or enhancing lesions (22%; p=0.01). Similarly, one third (33%) of CMET lesions misclassified by perfusion-CMR derived CUR were scored as predominantly hypo-enhancing on LGE-CMR, a rate > 2-fold higher compared to mixed or enhancing lesions (14%; p=0.10). Similarly, Figure 1 illustrates CMET enhancement patterns on LGE-CMR, for which predominantly hypo-enhancing lesions (middle) demonstrated lesser differentiation from CTHR than did those with diffuse enhancement (left columns).

Clinical Prognosis

Mortality was assessed after CMR to test whether perfusion-based diagnostic classification of C_MET and C_THR conferred differential prognosis, and if magnitude of lesion perfusion further differentiated mortality risk. 92% of patients had follow-up ≥1 year or mortality within this interval, including an aggregate 1-year mortality in patients with LGE-defined C_MET of 53% (43% in controls; p=0.09). Median duration of post-CMR follow-up was 29 months [IQR 12.6 – 48.5] among survivors.

Figure 4 provides Kaplan Meier survival curves of patients with LGE (left) and perfusion (middle, right) CMR defined C_MET and C_THR, as well as controls matched for primary cancer type and stage. As shown, each tissue characterization method demonstrated mortality to be increased among patients with C_MET, although risk for death was higher using LGE (HR=1.52 [1.16–2.00], p=0.003) than perfusion (semi-quantitative HR=1.42 [1.06–1.90], p=0.02; visual HR=1.47 [1.12– 1.94], p=0.006). Regarding C_THR, LGE yielded similar mortality risk between cases and controls (HR=1.43 [0.88–2.32], p=0.15). Conversely, patients with semi-quantitative perfusion-CMR classified C_THR had higher mortality than controls (HR=1.61 [1.07–2.41], p=0.02), consistent with above noted misclassification of low vascularity (predominantly hypo-enhancing) tumors. Of note, differential mortality between respective controls for C_MET and C_THR paralleled differences in cancer type and stage between these diagnostic groups.

Figure 4. Mortality Based on Diagnostic Categorization by CMR Tissue Characterization Methods.

Kaplan-Meier analyses for patients with C_MET and C_THR as established by LGE (left), semi-quantitative perfusion-CMR (middle) and visual perfusion-CMR (right), as well as among respective cancer-matched controls. As shown, both CMR tissue characterization methods demonstrated mortality to be higher among patients with C_MET, although adverse prognosis conferred by LGE was higher (HR=1.52 [1.16–2.00], p=0.003) than that conferred by perfusion, irrespective of whether interpreted using semi-quantitative (HR=1.42 [1.06–1.90], p=0.02) or visual (HR=1.47 [1.12–1.94], p=0.006) based diagnostic classifications. Conversely, LGE-CMR demonstrated patients with C_THR to have similar mortality to controls, whereas mortality was higher when C_THR was established based on perfusion, consistent with study analysis showing tumors with minimal enhancement (on LGE) to often be misclassified as C_THR based on perfusion cutoffs.

Perfusion-Based Strata of Mortality Risk

Given variable magnitude of difference in perfusion indices between C_MET and C_THR (paralleled by differential diagnostic performance), mortality was tested in relation to perfusion pattern to discern if mortality among cancer patients with C_MET varied in relation to magnitude of contrast enhancement perfusion-CMR.

Figure 5 stratifies patients with CMET (defined by the reference of LGE-CMR) into population-based tertiles of lesion vascularity based on contrast enhancement ratio, as well as follow-up data among C_MET patients (compared to controls) within respective tertiles. As shown, mortality risk among the overall cohort (5A) increased in proportion to lesion hypoperfusion (HR=1.40 [1.12–1.74], p=0.003) and was highest among patients with lesions in the bottom perfusion (CER) tertile. One year survival in the top tertile was 57%, ~1.5 times higher than that of the lowest tertile (40%, p=0.07). Consistent with above noted limits in diagnostic performance of semi-quantitative perfusion-CMR, qualitative analysis provided lesser prognostic value than did semi-quantitative assessments as evidenced by non-significant differences in mortality risk among patients with visually assigned perfusion tertiles (HR=1.07 [0.89–1.29, p=0.47).

Figure 5. Mortality Among Patients with Cardiac Metastases in Relation to Lesion Hypoperfusion.

5A: Kaplan-Meier curve encompassing patients with CMET (defined by the reference of LGE-CMR) stratified based on population-based tertiles of lesion vascularity (CER) as measured on semi-quantitative perfusion-CMR. As shown, there was a stepwise association between C_MET hypoperfusion and risk for death: Patients in the lowest tertile had higher mortality (p=0.002) than those in the upper-most tertile, paralleling a similar trend for patients in the middle tertile (p=0.09). Primary data from which tertile partitions derived shown in upper right.

5B: Comparisons between CMET cases and cancer-matched controls within each respective semi-quantitative perfusion tertile. As shown, patients within lowest tertile of perfusion-CMR evidenced vascularity (CER) had higher risk for death than did controls (p=0.001), paralleling a similar pattern but lesser magnitude of difference in the middle perfusion tertile (p=0.03). Conversely, mortality for patients with lesions in the highest perfusion tertile (corresponding to highest vascularity) was equivalent to that of cancer-matched controls (p=0.67).

Regarding case-control comparisons, data shown in 5B demonstrates that patients with CMET in the lowest tertile of semi-quantitative perfusion-CMR evidenced vascularity (based on CER) had higher risk for death than did cancer-matched controls (HR = 2.20 [1.37–3.53], p<0.001), paralleling a similar pattern but lesser magnitude of risk for patients in the middle tertile (HR = 1.66 [1.04–2.64], p=0.03). Conversely, prognosis was equivalent to controls among patients with lesions in the highest tertile of perfusion-CMR evidenced contrast enhancement ratio (HR = 0.90 [0.54–1.48], p=0.67).

Despite differential prognosis in relation to lesion perfusion pattern, Table 4 demonstrates that lesion location and cancer etiology did not differ between perfusion-CMR derived contrast enhancement partitions (all p=NS) whereas lesion size (area, length) increased in relation to lesion hypoperfusion (p<0.05). Regarding anti-cancer treatments, mediastinal radiation was associated with lesion hypoperfusion, as evidenced by increased frequency of use among patients stratified based on tertile-based decrements in CER (p<0.05), whereas frequency of chemotherapy and other anti-cancer therapeutic classes did not differ between perfusion-based partitions (p=NS).

Table 4.

Contrast Enhancement Partitions in Relation to Semi-Quantitative Perfusion Data, Anatomic Features, and Cancer Type

	Contrast Enhancement Ratio (population-based tertiles)			p
	<33th (III)	33–66th (II)	>66th (I)
Semi-Quantitative Perfusion  CER   CMET   CMET/Blood (%)  CUR   CMET   CMET/Blood (%) Lesion Location   Right atrium   Right ventricle   Left atrium   Left ventricle   Multi-chamber   Pericardium Lesion Size   Area (cm2)   Perimeter (cm)   Maximal Length   Orthogonal Length   Perimeter/Min Length Cancer Etiology   Sarcoma   Gastrointestinal   Genitourinary   Lymphoma   Lung   Skin/Melanoma Anti-Cancer Regimen   Chemotherapy   Alkylating Agents   Antimetabolites   Antimicrotubular Agents   Anthracycline   Kinase Inhibitors   Immunotherapy   Radiation Therapy (RT)   Mediastinal RT	0.6 [0.5 – 0.8] 105 [8.6 – 15.2] 6.7 [3.5 – 13.0] 8.3 [4.6 – 15.4] 30% (17) 39% (22) 25% (14) 19% (11) 19% (11) 39% (22) 7.6 [2.2 – 18.2] 11.9 [6.0 – 17.9] 4.0 [2.0 – 6.6] 2.3 [1.3 – 3.8] 4.5 [4.0 – 5.3] 25% (14) 9% (5) 19% (11) 12% (7) 9% (5) 9% (5) 44% (25) 30% (17) 37% (21) 26% (15) 14% (8) 18% (10) 60% (34) 28% (16)	1.3 [1.2 – 1.6] 16.6 [11.8 – 25.1] 9.3 [5.8 – 15.1] 7.8 [4.7 – 15.8] 43% (25) 28% (16) 26% (15) 35% (20) 31% (18) 29% (17) 6.0 [2.9 – 17.0] 9.7 [7.1 – 17.7] 3.3 [2.3 – 5.7] 2.0 [1.5 – 3.4] 4.7 [3.9 – 5.7] 22% (13) 21% (12) 7% (4) 12% (7) 16% (9) 10% (6) 47% (27) 36% (21) 38% (22) 19% (11) 12% (7) 16% (9) 40% (23) 12% (7)	2.6 [2.3 – 3.4] 21.3 [15.2 – 31.9] 14.1 [7.5 – 38.5] 9.8 [6.1 – 20.4] 35% (20) 31% (18) 29% (17) 36% (21) 33% (19) 21% (12) 3.3 [1.9 – 8.3] 7.2 [5.5 – 12.7] 2.6 [1.9 – 4.0] 1.8 [1.2 – 2.7] 4.1 [3.6 – 5.2] 26% (15) 9% (5) 16% (9) 5% (3) 14% (8) 10% (6) 33% (19) 27% (16) 36% (21) 16% (9) 5% (3) 16% (9) 48% (28) 10% (6)	<0.001 <0.001 <0.001 0.42 0.33 0.44 0.84 0.10 0.22 0.11 0.04 0.03 0.02 0.28 0.06 0.91 0.08 0.14 0.35 0.54 0.95 0.28 0.59 0.98 0.34 0.27 0.95 0.10 0.02

Table 5 reports regression analyses testing clinical, therapeutic, and imaging predictors of mortality in the study cohort. Regarding cancer related indices, among cancer diagnostic categories, patients with lymphoma were at decreased risk for death (HR= 0.44 [0.26 – 0.75], p=0.002), whereas mortality risk was increased in patients with gastrointestinal malignancies (1.63 [1.12 – 2.37], p=0.01) or those treated with antimetabolites or kinase inhibitors (both p<0.05): Conversely, neither age, cardiovascular risk factors, or conventional indices of cardiac chamber geometry or systolic function were associated with differential prognosis (all p=NS). Regarding CMET tissue characteristics, decreased lesion perfusion (contrast enhancement ratio) conferred increased mortality risk – irrespective of whether assessed as a continuous variable or based on graded tertiles of hypoperfusion (both p<0.05); results paralleled a similar trend for graded severity of hypo-enhancement on LGE-CMR (p=0.09). In multivariable analysis for which the model included all variables associated with differential prognosis in univariable regression, decreased lesion perfusion (CER) remained independently associated with increased risk for death (1.30 [1.03 – 1.64], p=0.03).

Table 5.

Mortality Predictors Among Patients with LGE-CMR Evidenced Cardiac Metastases

	Hazard Ratio	p
Univariable Regression *
Clinical Characteristics   Age (per year)   Gender (male)   Cancer Etiology     Sarcoma     Gastrointestinal     Genitourinary     Lymphoma     Lung     Skin/Melanoma   Anti-Cancer Regimen     Chemotherapy     Alkylating Agents     Antimetabolites     Antimicrotubular Agents     Anthracycline     Kinase Inhibitors     Immunotherapy     Radiation Therapy     Mediastinal RT   Coronary Artery Disease   Cardiovascular Risk Factors     Hypertension     Hypercholesterolemia     Diabetes Mellitus     Tobacco Use  Cardiac Function and Geometry   Left Ventricle (LV)     LVEF (per %)     LVEF (<50%)     LVEDV (per ml)   Right Ventricle (RV)     RVEF (per %)     RVEF (<50%)     RVEDV (per ml)  CMET Lesion Characteristics   Anatomic Properties     Maximal diameter (per 10 cm)     Area (per 10 cm2)   Tissue Properties    Perfusion CMR     Contrast Enhancement Ratio     Contrast Enhancement Ratio Grade (I-III) ‡     Contrast Uptake Rate     Visual Perfusion Grade (I-III) §   Late Gadolinium Enhancement CMR     SNR     Visual LGE Grade (I-III) ¶	1.00 [0.99 – 1.01] 1.07 [0.82 – 1.40] 1.21 [0.89 – 1.64] 1.63 [1.12 – 2.37] 1.11 [0.75 – 1.63] 0.44 [0.26 – 0.75] 1.46 [0.99 – 2.14] 0.79 [0.50 – 1.24] 1.23 [0.94 – 1.61] 1.81 [1.36 – 2.40] 1.23 [0.94 – 1.62] 0.87 [0.63 – 1.21] 1.68 [1.12 – 2.52] 0.95 [0.67 – 1.35] 1.07 [0.81 – 1.40] 1.10 [0.74 – 1.64] 0.83 [0.53 – 1.33] 0.80 [0.60 – 1.05] 0.92 [0.68 – 1.23] 1.06 [0.73 – 1.53] 0.96 [0.73 – 1.28] 1.01 [0.99 – 1.02] 0.81 [0.54 – 1.23] 1.00 [0.99 – 1.00] 1.01 [0.99 – 1.03] 0.79 [0.57 – 1.08] 1.00 [0.99 – 1.00] 1.82 [0.84 – 3.95] 1.10 [0.97 – 1.26] 0.77 [0.64 – 0.93] 1.32 [1.05 – 1.66] 1.00 [0.99 – 1.01] 1.07 [0.94 – 1.22] 1.00 [1.00 – 1.01] 1.21 [0.97 – 1.52]	0.62 0.63 0.22 0.01 0.61 0.002 0.06 0.30 0.14 <0.001 0.14 0.40 0.01 0.78 0.64 0.65 0.44 0.11 0.56 0.77 0.80 0.35 0.32 0.15 0.27 0.13 0.19 0.13 0.14 0.005 0.02 0.81 0.33 0.28 0.09
Multivariable Regression †
Contrast Enhancement Ratio Grade (perfusion)	1.30 [1.03 – 1.64]	0.03
GI (cancer etiology)	1.21 [0.70 – 2.08]	0.50
Lymphoma (cancer etiology)	0.79 [0.41 – 1.53]	0.48
Antimetabolite (therapy)	1.71 [1.13 – 2.58]	0.01
Kinase Inhibitors (therapy)	1.52 [0.84 – 2.74]	0.16

^*Univariable comparisons between cancer patients with CMR-evidenced cardiac metastases and cancer-matched controls.

^†Multivariable regression model comprised of factors significantly associated with mortality (p<0.05) in univariable analysis.

^‡Semi-quantitative perfusion tertiles as reported in Figure 5 (I: highest tertile, II: middle tertile, III lowest tertile); HR reported inversely in relation to perfusion tertile (lower 2 tertiles vs. highest tertile)

^§Qualitative perfusion grades as in text (grade III: <50% of lesion area, grade II: mixed: >50% of lesion demonstrating contrast uptake; scattered foci of hypoperfusion, grade I: diffuse contrast uptake).

^¶LGE grades as in text (diffusely hyper-enhancing, mixed, predominantly hypo-enhancing)

Discussion

This is the first study to test perfusion-CMR tissue characterization for cardiac masses. Key findings are as follows: First, among a broad cancer cohort, perfusion contrast kinetics differed between vascular and avascular masses, as evidenced by higher upslope and peak enhancement for LGE-CMR defined C_MET vs. C_THR (p<0.001). Second, while perfusion-CMR yielded good overall performance (maximum AUC 0.89–0.93) for CMET, sensitivity (84%) and negative predictive value (71%) were suboptimal and were predominantly attributable to misclassification of tumors with minimal enhancement on LGE-CMR. Third, although both methods showed mortality to be increased among patients with CMET, risk for death was higher using LGE (HR=1.52 [1.16–2.00], p=0.003) than perfusion (semi-quantitative HR=1.42 [1.06–1.90], p=0.02; visual HR=1.47 [1.12–1.94], p=0.006). Last, whereas CMET strongly impacted prognosis, mortality risk varied – nearly half (47%) of affected patients were alive 1 year post CMR: Among patients with CMET on LGE-CMR, mortality varied in relation to perfusion-CMR – as evidenced by an inverse association between lesion enhancement (CER) and mortality (HR=1.40 [1.12–1.74], p=0.003), and increased mortality among patients with lowest perfusing lesions compared to cancer-matched controls (HR=2.20 [1.37–3.53], p<0.001). Results also demonstrated that key lesion perfusion characteristics (CER) were similarly distributed among patients within major cancer categories within our study population, suggesting that perfusion characteristics are not linked to cancer type. Taken together, findings (Central Illustration) demonstrate added prognostic value of perfusion for patients with CMET among whom lesion hypovascularity is a novel marker of adverse prognosis.

Central Illustration. Perfusion CMR for Cancer-Associated Cardiac Masses.

Multicenter data from two cancer centers demonstrated perfusion-CMR evidenced contrast kinetics to differ between vascular and avascular masses, as evidenced by higher uptake (slope) and plateau enhancement for cardiac metastasis vs. thrombus. While perfusion performed well in relation to LGE, performance as a primary method to differentiate mass types and stratify prognosis was compromised by misclassification of metastases with minimal enhancement: Among patients with CMET on LGE-CMR, perfusion-CMR demonstrated mortality to increase with lesion hypo-perfusion and to be augmented for patients with low vascularity lesions and equivalent (to that of controls) for patients with high vascularity lesions. Findings support utility of integrated CMR tissue characterization for cardiac masses, in which initial diagnosis is established by LGE and prognosis for CMET further stratified based on perfusion-CMR evidenced lesion hypo-vascularity.

Applied clinically, our results support an algorithm by which LGE-CMR is used for initial diagnostic differentiation between C_MET and C_THR, and mortality risk among patients with LGE-evidenced C_MET is further stratified based on perfusion-CMR. Whereas LGE-CMR is well validated for differentiating between neoplasm and thrombus based on presence or absence of enhancement, perfusion-CMR performed poorly for this purpose in our dataset (58% of patients with predominantly hypo-enhancing lesions on LGE were misclassified as thrombus based on CER and even 14% of patients with mixed or diffusely enhancing lesions on LGE were misclassified). In this context, it is important to recognize that when LGE-CMR demonstrates enhancement within a cardiac mass (which may be minimal in some cases), perfusion characteristics should not alter its diagnostic classification as neoplastic in etiology. Notably, our analyses also demonstrated that qualitative (visual) perfusion yielded slightly lower diagnostic accuracy for differentiating between C_MET and C_THR to that of semi-quantitative perfusion (both of which were sub-optimal in relation to LGE-CMR). Semi-quantitative analysis provided added value for stratifying prognosis among patients with LGE defined CMET, whereas visual perfusion analysis lacked sufficient precision to stratify prognosis in relation to reader assigned strata of lesion vascularity. These data add to a growing body of literature highlighting the utility of semi-quantitative perfusion-CMR for myocardial perfusion and highlight the need for future research testing utility of fully quantitative tissue characterization CMR for prognostic assessment of cardiac masses.

Regarding mechanism, we speculate that perfusion-CMR evidenced enhancement within C_MET reflects two phenomena - tumor-associated vascularity and increased capillary perfusion. The notion that vascular supply is intrinsic to tumorigenesis has been demonstrated on a cellular basis; tumor neovascularization has been shown to have increased capillary permeability.(21) Regarding prognosis, it is known that aggressive neoplastic cell proliferation can decouple from vascular supply such that tumor growth outstrips oxygen supply and intrinsic cell death (tumor necrosis) occurs: Necrotic cells release proinflammatory and tumor-promoting cytokines, which in turn recruit immune and inflammatory cells, thereby exerting tumor-promoting activity by inducing angiogenesis and proliferation.(22,23) This would be expected to manifest with decreased vascularity in poor prognosis tumors as was observed in our study. It is also possible that decreased tumor vascularity can impede drug delivery, resulting in therapeutic non-responsiveness as a mechanism for our observed association of hypoperfusion with mortality. Whereas the notion that impaired vascularity (tumor necrosis) is an adverse marker is consistent with established concepts in tumor biology,(24,25) we are unaware of any prior study in which this tenet has been used to stratify clinical risk for cancers involving the heart.

It should also be recognized that perfusion-CMR frequently yielded discordant results with LGE, and that it often misclassified low vascularity lesions deemed to be C_MET based on the reference standard of LGE. Whereas perfusion-CMR typically misclassified lesions that were predominantly hypo-enhancing on LGE, discordances did occasionally occur in the context of mixed or diffusely enhancing C_MET – with similar shortcomings when analysis was performed using CER and CUR (22% and 14% misclassification rate in relation to the reference of LGE, respectively). Additionally, our data demonstrated that the prognostic value of perfusion was complementary to that of LGE, as evidenced by mortality data showing that among patients with LGE classified C_MET, prognosis was worse in those with decreased lesion vascularity on perfusion. While these findings might be taken to indicate shortcomings of a given pulse sequence, they could also be taken to support the notion that each provides distinct information regarding lesion vascularity: LGE provides a static window regarding contrast accumulation over a given interval, whereas perfusion provides a snapshot of dynamic contrast uptake during contrast infusion. It is also well known that regions of decreased vascularity can accumulate contrast over time, as evidenced by studies of post-MI microvascular obstruction, in which discordances between perfusion and LGE have been shown to be common(26) and microvascular obstruction size to decrease over time when imaged serially via LGE.(27) Given that 1) neoplasms can contain both vascular and avascular (tissue necrosis) components, and that 2) gadolinium can diffuse between vascular and avascular regions with time (i.e. the interval between contrast infusion and LGE acquisition) – first pass perfusion CMR-based assessment of dynamic contrast kinetics holds appeal as a means to complement LGE-based diagnostic differentiation by further characterizing magnitude of vascularity within LGE defined cardiac neoplasms. To our knowledge, this is the first study to test these physiologic concepts with respect to CMR assessment of cardiac masses.

Regarding significance, it is important to interpret our findings in context of literature indicating that C_MET are greatly under-recognized in patients with advanced cancer – highlighting the need to refine screening and prognostic risk stratification as a rationale for this study. Pathology studies have reported cardiac metastases to occur in approximately 10–20% of patients with advanced cancer,(28–31) many of which were not diagnosed prior to autopsy. Regarding imaging, the notion that tissue characterization improves diagnostic and prognostic assessment of C_MET is well-established, as evidenced by data from our group and others which have used the technique of LGE-CMR to demonstrate that presence of enhancement within lesions correlates with tumor vascularity and stratifies prognosis.(4–7,10) It is also known that vascularity of C_MET can vary as can outcomes – the latter of which have markedly improved with advances in anticancer therapy. As evidence of this, whereas presence of C_MET associated with increased mortality in our cohort, nearly 50% of all C_MET patients were alive at 1 year follow-up: In this context, our finding that lesion hypo-enhancement on perfusion-CMR stratified mortality among patients with C_MET is important to a growing cancer population for whom it extends risk stratification beyond anatomic assessment or binary tissue characterization based on presence or absence of enhancement.

Several limitations should be noted. First, our population encompassed an array of cancer types, thus prohibiting robust assessment of individual cancer types and modifiers. On the other hand, our study design matched patients with cardiac masses to controls with equivalent cancer etiology and extent. In this context, our finding that perfusion results paralleled LGE and refined risk stratification even among patients with C_MET and matched controls adds to robustness of our results. Second, CMR methods used in the current study are applicable to lesions of sufficient size to perform reliable tissue characterization, and that our approach is not feasible in patients with highly mobile lesions. As evidence of this, 3% of patients with C_MASS were excluded due to inadequate spatial coverage or prominent lesion mobility. These data highlight need for future research to test whether high spatial resolution and/or temporally accelerated CMR methods provide incremental utility for diagnostic and prognostic assessment of cardiac masses. It is also possible that lesions deemed to be C_MET may have represented primary cardiac malignancies, given that uniform pathology was not available. While we believe this unlikely (in context of the detailed nature of clinical evaluations, active treatment for systemic cancers, and relative rarity of primary [compared to metastatic] cardiac malignancies), this possibility cannot be fully excluded. Finally, data was derived from a registry that did not include serial CMR to test whether perfusion characteristics are static or dynamic within an individual patient, whether perfusion (or LGE) predicts lesion growth, or whether anti-cancer treatment has modifying effects on lesion tissue characteristics. Whereas all patients with and without C_MET had stage IV disease (matching was done based on systemic cancer etiology) and our results showed similar distribution of major cancer categories among perfusion tertiles, lack of longitudinal data in patients with uniform cancer type impeded our ability to fully test if some cancer patients have hypo-vascular metastases and poor survival, whereas others have more vascular metastases and better prognosis (i.e. vascularity is a static marker). Future studies are necessary to explore these issues.

In conclusion, this study adds to a growing body of literature demonstrating utility of CMR tissue characterization in patients with cardiac masses and provides new insights regarding additive prognostic value of perfusion. Larger-scale research is warranted to test whether perfusion-CMR predicts non-mortality related adverse outcomes or response to specific anti-cancer therapies in patients with C_MET, towards the goal of tissue characterization imaging guided tailored therapy and improved clinical outcomes for advanced cancer patients.

Supplementary Material

1. Supplemental Figure. Semi-quantitative Analysis of C_MASS Vascularity using First Pass Perfusion-CMR.

A region of interest (ROI) was drawn over CMASS on perfusion images as shown in bottom image panel. The measured contrast signal intensity (SI) was plotted over time (t): (1) Contrast uptake rate (CUR) was calculated from the upslope of SI vs time. (2) Contrast enhancement ratio (CER) was calculated by normalizing plateau SI from baseline SI.

Perspectives: Core Clinical Competencies and Translational Implications

Competency in Medical Knowledge 1:

Contrast kinetics differ between vascular and avascular masses, as evidenced by higher upslope and peak enhancement for cardiac metastasis vs. thrombus as assessed using first pass perfusion-CMR.

Competency in Medical Knowledge 2:

While late gadolinium enhancement (LGE) CMR is superior to perfusion as a primary method to differentiate cardiac metastases from thrombus, mortality among patients with LGE evidenced cardiac metastases is heterogeneous and varies in relation to perfusion-CMR evidenced lesion hypoperfusion: Findings support utility of an integrated CMR tissue characterization algorithm for cardiac masses, in which initial diagnostic categorization of cardiac metastasis vs. thrombus is established by LGE and prognosis among patients with CMET is further stratified based on perfusion-CMR evidenced lesion hypo-vascularity.

Translational Outlook:

Future research is warranted to test whether perfusion-CMR predicts non-mortality related adverse outcomes or response to specific anti-cancer therapies in patients with cardiac metastases, towards the goal of tissue characterization imaging guided tailored therapy and improved clinical outcomes for advanced cancer patients.

Abbreviations

CER

Contrast enhancement ratio

C_MASS

Cardiac mass

C_MET

Cardiac metastasis

C_THR

Cardiac thrombus

CUR

Contrast uptake rate

LGE

Late gadolinium enhancement

LA

Left atrium

LV

Left ventricle

RA

Right atrium

RV

Right ventricle

SI

Signal intensity

TI

Inversion time