Macroangiopathy is a positive predictive factor for response to immunotherapy

Immunotherapies demand for predictive biomarkers to avoid unnecessary adverse effects and costs. Analytic morphomics is the technique to use body composition measures as imaging biomarkers for underlying pathophysiology to predict prognosis or outcome to therapy. We investigated different body composition measures to predict response to immunotherapy. This IRB approved retrospective analysis encompassed 147 patients with ipilimumab therapy. Degree of macroangiopathy was quantified with the newly defined total plaque index (TPI), i.e. the body height corrected sum of the soft and hard plaque volume of the infrarenal aorta on portalvenous CT scans. Furthermore, mean psoas density (MPD), different adipose tissue parameters as well as degree of cerebral microangiopathy were extracted from the imaging data. Subsequent multivariate Cox regression analysis encompassed TPI, MPD, serum LDH, S100B, age, gender, number of immunotherapy cycles as well as extent of distant metastases. TPI and MPD correlated positively with PFS in multivariate analysis (p = 0.03 and p = 0.001, respectively). Furthermore, single visceral organ and/or soft tissue involvement significantly decreased progression risk (p = 0.01), whereas increased S100B level showed a trend towards PFS shortening (p = 0.05). In conclusion, degree of macroangiopathy and sarcopenia were independent predictors for outcome to immunotherapy and of equivalent significance compared to other clinical biomarkers.

Until the end of the study, 128 patients (87.1%) had shown progressive disease, while 19 (12.9%) were stable. Median time to progress (IQR) was 124.9 (14.8-164.0) days. Detailed patient characteristics are shown in Table 1.
Macroangiopathy. Body height correction of the total plaque volume defined as the sum of hard and soft plaque volume was performed via calculating the ratio of the total plaque volume and the anterior height of the os sacrum (hereinafter referred to as total plaque index (TPI)). Median TPI (IQR) was 2.3 (0.2-13.7) mm² and TPI correlated significantly with PFS in univariate Cox regression analysis with a reduction of progression risk by 9% every 10 mm² (p univ = 0.04, HR 0.91, CI 0.84-1.00). Patients with a TPI in the upper quartile had approximately www.nature.com/scientificreports www.nature.com/scientificreports/ half of the progression risk at any time point compared to patients with a TPI in the first quartile (p univ = 0.02, HR 0.56, CI 0.35-0.92, Fig. 1A). Median volume (IQR) of hard and soft plaques was 2.4 (0.1-14.9) cm³/10 and 52.7 (0.0-218.9) mm³, respectively. Neither hard nor soft plaque volume alone correlated significantly with PFS in univariate analysis (p univ = 0.06 and p = 0.12, respectively). Detailed results including HRs and CIs of both uniand multivariate analysis are presented in Table 2. Muscle mass. Median mean psoas density (MPD, (IQR)) was 49.9 (45.0-52.9) Hounsfield Units (HU). MPD correlated significantly with PFS in univariate regression analysis with a reduction of progression risk of around 30% every 10 HU (p univ = 0.01, HR 0.71, CI 0.55-0.93). Furthermore, patients with an MPD greater than the lower quartile Q1 had a 0.6-fold reduced progression risk at any time point compared to patients in the lowest quartile of the psoas density, i.e. patients with the highest fatty infiltration of the psoas muscle (p univ = 0.01, HR 0.6, CI 0.40-0.88, Fig. 1B).
Blood-based parameters. In 131 patients (89.1%) serum LDH had been quantified within three months prior to immunotherapy with 51 of the patients revealing an increased serum LDH level (38.9%). Univariate Cox regression analysis did show a trend to increase of progression risk in patients with an elevated serum LDH prior to immunotherapy (p univ = 0.13, HR 1.34, CI 0.92-1.95). S100B value was available in 94 patients (63.9%) with 47 (50.0%) patients showing an increased S100B value before initiation of ipilimumab therapy. S100B value was significantly associated with PFS in univariate analysis (p univ = 0.03, HR 1.63, CI 1.05-2.55).  www.nature.com/scientificreports www.nature.com/scientificreports/ Multivariate Cox regression analysis. Finally, multivariate analysis was conducted including all metrics with p uni <0.2, i.e. MPD, TPI, age, gender, number of immunotherapy cycles and extent of distant metastases. Multivariate analysis confirmed the significant impact of the MPD and TPI on PFS (MPD: p multi = 0.001, HR 0.66, CI 0.51-0.85; TPI: p multi = 0.03, HR 0.99, CI 0.98-1.00). Furthermore, patients with distant metastases in more than one visceral organ had a significantly shorter PFS than patients with metastases in only one visceral organ ± cutaneous or nodal metastases (p multi = 0.02, HR 1.56, CI 1.06-2.30).
After adjustment for age, gender, distant metastases and immunotherapy cycles, increased S100B level revealed a close to significant trend to decreased PFS (p multi = 0.05, HR 1.57, CI 1.00-2.47, n = 94), while increased serum LDH was not significantly associated with PFS (p multi = 0.39, HR 1.18, CI 0.81-1.74, n = 131). Results of multivariate analysis are presented in Table 2.

Discussion
In the current study, we investigated the value of different body composition measures to predict response to immunotherapy with ipilimumab. Apart from confirming mean psoas density, a surrogate for sarcopenia, as a positive predictive factor for outcome to immunotherapy, we report a surprising positive correlation between total plaque burden of the abdominal aorta and progression free survival.
Even though the immune system is generally capable not just to identify, but also to eradicate neoplastic cells, it is physiologically impeded by immune check points in order to prevent autoimmunity. Antibodies that block the checkpoint CTLA-4 or PD1/PD-L1 have demonstrated efficacy in a number of malignancies. However, response rates are variable and these novel therapies, especially those relying on CTLA-blockade and thereby targeting an early point in anti-tumoral immune response, bear a significant risk of immune-related adverse events (irAE) with clinically detectable irAEs in up to 72% and grade 3-4 irAEs in 24% of patients treated with ipilimumab 24,25 .
Beside the potential irAE, therapy with checkpoint inhibitors is very cost-intensive with an ipilimumab monotherapy amounting to approximately $160,000 per patient and combined ipilimumab and nivolumab therapy to $300,000 26 . Both potential irAE and high costs result in an urgent demand for predictive biomarkers that allow for reliable identification of patients that are likely to benefit from immunotherapy.
Our results confirm the validity of the sarcopenia-frailty concept, that predicts a correlation between sarcopenia and disease outcome, even in patients undergoing immunotherapy 17 . The prognostic or predictive value of the psoas density was shown in a variety of other pathological conditions before, such as trauma, liver transplantation as well as outcome after tumor resection or cardiovascular surgery [27][28][29][30][31] .
Furthermore, we assessed the potential of vessel health in predicting response to immunotherapy and thus defined the total plaque index, i.e. the body height corrected sum of the hard and soft plaque volume, to quantify extent of macroangiopathy. Body height correction was performed to rule out any influence of the body height on the plaque volume, as it seemed reasonable that greater vessel wall surface might correlate with greater plaques burden.
As sacrococcygeal dimensions should be used for body height estimation when long bones are not available, body height correction was performed via normalizing the plaque volume by the total anterior height of the sacral and coccygeal vertebrae as previously described 32,33 . This measurement had not just given the most accurate stature estimate in forensic studies, but is also easy to perform and routinely included in the examination area.
The shown positive correlation between TPI and PFS might reveal a surprising hint towards pathophysiology of immune checkpoint blockade: While the positive relationship between macroangiopathy and immune status may appear paradoxical in the first instance, it seems conceivable that systemic inflammatory diseases such as atherosclerosis enhance the effect of immunotherapy due to a generally increased immune activity. This would be in accordance with the finding that patients with a tendency to autoimmunity and irAE are more likely to benefit from immune checkpoint blockade and that premature interruption of therapy due to irAEs is known to not having a negative effect on progression free survival [34][35][36][37] .
Notably, the isolated assessment of the hard plaque volume did not reveal a significant correlation with response to immunotherapy, whereas consideration of the soft plaques volume, which represents only around one tenth of the total plaque burden, revealed a significant impact on PFS. This might be due to the pronounced inflammatory component of noncalcified atheroma whose progress is known to be induced by immune cells and T-cell derived cytokines and whose stability negatively correlates with levels of acute-phase proteins such as c-reactive protein or serum amyloid A 38,39 .
However, even though cerebral small vessel disease as well is said to be associated with chronic inflammation to some degree 40 , Fazekas Score for deep white matter lesions, i.e. a surrogate of microangiopathy, did not correlate with progression free survival in our study.
Furthermore, fat tissue volume has previously been discussed to have an either negative effect on outcome after immunotherapy due to higher comorbidity or otherwise a positive effect due to its known endocrine function. However, our study did not reveal any impact of adipose tissue volume on PFS and therewith contradicts both the results of Sabel et al. who reported on a negative impact of increasing visceral fat distance on overall survival after treatment with ipilimumab in 44 patients 17 and the findings of McQuade et al. who reported on a positive correlation between BMI and therapy outcome in male patients treated with either targeted or immunotherapy 19 .
Compared to the multiple tissue and blood-based biomarkers that have been under investigation so far such as tumor PD-L1 expression, neutrophil to lymphocyte ratio, absolute lymphocyte count, level of lactate dehydrogenase, S100B or density of tumor infiltrating lymphocytes 15,[41][42][43] , the imaging derived biomarkers investigated in this study have the practical advantage that they occur in routine clinical practice and do not require any additional examinations or blood sampling. Moreover, morphometric assessment demands very little time with results being available instantly, prior to a potential start of therapy and without the need of follow-ups for kinetic analysis.
www.nature.com/scientificreports www.nature.com/scientificreports/ Furthermore, prognostic equivalency of the morphometric parameters compared to the assessed blood-based markers was proven with total plaque index and mean psoas density correlating significantly with PFS, while increased S100B level prior to start of immunotherapy did only show a borderline effect on PFS and elevated serum LDH was not associated with PFS at all. -Not to forget that hazard ratios depend on the subdivision and range of the explanatory variable and therefore must not be compared between the different metrics one-by-one… Our study has limitations: CT data analysis was conducted semi-automatically with consecutive increase of expenditure of time as well as reduction of reproducibility. Furthermore, our analysis was limited to patients with malignant melanoma and treatment with ipilimumab even though inclusion of patients with different cancer entities and therapy with other checkpoint inhibitors or different immunotherapy regimes would have been desirable.
In conclusion, our study reveals a positive correlation between manifestation of macroangiopathy and response to immunotherapy, which might be exploited as an easily performed predictive imaging biomarker for outcome to immunotherapy. Furthermore, our study results include valuable hints towards pathophysiology and demand for a deeper understanding of the association of chronic inflammatory disease and immune status. www.nature.com/scientificreports www.nature.com/scientificreports/ patients and Methods patients and ethics. The ethics committee of the University of Heidelberg approved this study and written informed consent was waived due to its retrospective character. The study was conducted in accordance with all relevant guidelines and regulations.
We reviewed the medical record and imaging studies of all patients who were referred to our imaging center between August 2010 and April 2017 with a diagnosis of malignant melanoma. Patients with ipilimumab therapy were included in the study and abdominal CT scans and cerebral MRI performed up to 3 months prior to or after start of immunotherapy were analyzed. Patients without available CT of the abdomen and patients who had received more than four cycles of immunotherapy were excluded (n = 8). Patients without available FLAIR sequence in the aforementioned period were not included in the analysis of deep white matter hyperintensities (n = 3).
Image and data analysis. CT scans were performed at our institution on a dual-energy Siemens system The majority of cranial FLAIR MRI was performed at our institution on a 1.5 Tesla MRI system (Magnetom Symphony, Siemens) with the following sequence parameters: echo time 98 ms, repetition time 8000 ms, inversion time 2340 ms, slice thickness 5 mm, field of view 230, number of slices 30, flip angle 150°, acquisition matrix 0/320/205/0, number of averages 2 and acquisition time 5:38 min. If no cMRI with a FLAIR sequence of our institution, but of another institution was available for the corresponding period, external imaging was consulted for image analysis. www.nature.com/scientificreports www.nature.com/scientificreports/ Image analysis was carried out with the reader blinded to the patients' data and according to the following description (for an example of the applied image post processing see Fig. 2): • Degree of macroangiopathy was determined for the abdominal aorta inferior to the renal arteries and superior to the bifurcations of the common iliac arteries into the external and internal iliac arteries. The aorta was semi-automatically segmented on the coronal plane and the volume of hard and soft plaques was determined via thresholding with a median threshold of 185 (range 180-220) HU for calcifications and 100 (range 60-100) HU for soft plaques 44 , respectively. Segmentations were corrected manually if necessary. • Psoas muscle contours were drawn on the axial plane by using the live wire tool on the fourth lumbar vertebrae (L4) level, and total psoas areas as well as MPD were calculated. • The ventral length of the sacrum was measured from the promontorium to the inferior coccyx with the help of the curve tool on the median sagittal plane, which was reconstructed by MITK from the axial plane. • For assessing the fatty tissue area, fat density thresholds of −190 to −30 HU were used. The visceral adipose tissue area was defined as the total area inside the abdominal fascia meeting fat-density thresholds, whereas the subcutaneous adipose tissue area was defined as the total area between the abdominal fascia and skin on L4 level. Total body fat was defined as the sum of subcutaneous and visceral adipose tissue area. • Finally, the FLAIR sequence was used to determine the degree of deep white matter hyperintensitie, which are attributed to chronic small vessel disease, by applying the Fazekas Scoring System (Fig. 3) 45,46 . As the aetiology of deep and periventricular white matter hyperintensities differ, the reader was instructed to disregard hyperintensities associated with the ventricles.
Additionally, patient height and weight were extracted from the medical records to calculate their BMI. Furthermore, serum LDH and S100B values quantified within three months prior to immunotherapy as well as number of sites with distant metastases were determined. Progression free survival was calculated as the time interval between the last immunotherapy cycle and change in therapy due to disease progression or date of death. However, for patients who did not experience disease progression in the study period, PFS was calculated with the date of the last medical report with documented stable disease. statistical analysis. Descriptive statistics were carried out using Microsoft Excel 2016 and statistical testing was performed with the software package R (R Foundation for Statistical Computing, version 3.4.4, 2018-03-15). Significance was set at the p < 0.05 level. To control for multiple testing, a hierarchal hypothesis testing approach was performed: Firstly, univariate Cox regression analysis was performed as screening for potential predictive factors. Subsequently, multivariate analysis was conducted including known potential confounders such as gender and age as well as all parameters with p univ < 0.02 to confirm their independent predictive value and to compare the degree of impact of the individual parameters.

Data Availability
All data is available from the corresponding author on reasonable request in pseudonymised form.