Metabolic Imaging to Improve Patient-Specific Therapy Outcome

Non-communicable diseases are the cause of a majority of the global death, with
cancer expected to rank as leading cause in the upcoming years. The incidence
of gastrointestinal cancers is largely attributing to the global increased
cancer incidence, with colon cancer ranking as the fourth most common site of
newly diagnosed cancer in 2018 (colon cancer (6.1%), oesophageal cancer (3.2%)
and pancreatic cancer (2.5%))[1]. In 2018, gastrointestinal cancers made-up
15.5% of all global cancer related deaths in 2018. Although treatment
strategies have improved, patients suffering from metastasized gastrointestinal
cancer often receive ineffective treatments or undergo periods of therapy
resistance because tumor non-response is hard to detect, especially in early
stages of therapy. For example, 50% of patients will experience progression
within six cycles (54 weeks) of first line treatment [2]. One of the reasons is
that, currently, the primary ineffectiveness of a treatment cannot be
determined earlier than nine weeks by assessment of morphological changes via
RECIST following start of therapy [3,4].

Metabolomics studies body functions through the measurement of metabolites,
chemical signatures of cellular processes. The capacity of metabolomics to
identify specific pathways to characterize tissues has been demonstrated by
recent studies [5], not only via cultured organoids from biopsies [6], but also
intra-operatively via a mass-spectrometry pen [7], demonstrating 96% accuracy.
MRI (Magnetic Resonance Imaging), is able to provide 3D images of the body. In
conventional MRI the resonance of only hydrogen nuclei (1H) are registered.
Therefore, MRI is often limited to morphological imaging, essentially mapping
water and fat, providing little information on chemical composition and
biological processes [8]. Initial studies focusing on a single nucleus to
examine biological processes have shown potential (e.g., brain scans), but
additional research is required to develop a chemical imaging technology for
the full human body and to include a broad spectrum of nuclei such as
phosphorus (31P) and sodium (23Na) but also deuterium (2H), fluorine(19F),
oxygen (1O) and carbon (13C) [9-11]. Technological breakthroughs in detecting
all biologically relevant nuclei opened up the possibility to provide dynamic
3D maps of metabolic and physiologic activity, which may be linked to clinical
parameters, such as response to therapy [12-15]. Full body metabolic imaging
with MR, however, is not yet clinically feasible as a result of the following
challenges: i.) The RF wavelength of the proton (1H) imaging signals is much
shorter (<=11 cm) than the dimensions of the human body (non-uniformities) [8];
ii.) Signals from metabolites are often obscured by other metabolites and
highly abundant signals (e.g., water, lipids and surrounding tissues such as
muscles) [9,16]; iii) Frequency shifts are caused by decreased B0-field
homogeneity due to susceptibility differences of tissues or (non-magnetic)
metals such as clips used in surgery; iv.) Many organs are moving due to
respiration and pulsating blood flow, resulting in ghosting artefacts during
relatively long metabolic scan sessions; v) The relatively low abundant
X-nuclei require numerous high energetic RF pulses to regain signal,
consequently increasing specific absorption rate (SAR) which limits the maximum
imaging volume.

However, recent studies reported full organ coverage 31P MRSI at 7T is feasible
in the liver, heart and lungs using a 31P whole body birdcage coil.[12,17-19]
This hardware improvement allows fast 31P MRSI over large volumes of interest
without a necessary increase in SAR. The frequencies of 31P (120.6Mhz) and 23Na
(78.8Mhz) at 7T are lower than 1H and results in less destructive RF
interference due to the increased wavelengths, improving the 31P B1+-field
homogeneity compared to conventional 1H 7T MRI or conventional 31P surface
coils used only for local and surface MRS applications. In addition, the
signals of all 31P metabolites are well separated due to the increased spectral
resolution at 7T, improving the detection of individual metabolites without any
necessary post-processing steps. However, signal contamination from
neighbouring tissue is still present from high abundant signals such as
surrounding muscles and need to be assessed and minimized using specific
imaging and processing strategies [10,16] Inhomogeneity of the main magnetic
field (B0) can be overcome using conventional shimming techniques, however,
large volumes are difficult to shim, with increasing difficulty at higher field
strengths. Current advances demonstrate the use of a local shim coil arrays
which increases the degrees of freedom available to improve B0 field
homogeneity [20,21]. The fast repetitions and large number of averages which
result in a long scan protocol, make it impractical to perform breath-hold
triggered MRSI. Strategies to correct for bulk motion induced artefacts use 1H
navigator images or 31P navigator spectra, however, have not been explored in
MR spectroscopic imaging.
Unlocking the potential of the available 7T MR technologies allows development
of new methodologies to image a full scale of chemical processes in the human
body, via the detection of a wide spectrum of relevant nuclei.

Phosphorus (31P) MRSI allows monitoring tissue metabolism by measuring specific
energy-and phospholipid-metabolites. Phosphocreatine (PCr), Adenosine
triphosphate (ATP, with α-, ß- and γ- resonances) and inorganic phosphate (Pi)
give insight into cell energy metabolism and the ratios between these
metabolites are already used as diagnostic indicators in systemic diseases [22-
24]. Inorganic phosphate (Pi) can also serve as a diagnostic marker for tissue
pH as its resonance frequency changes with the acidity of the environment and
distinguish between extracellular and mitochondrial species [25]. In addition,
31P MRSI at 7T is able to detect cell membrane precursors, the phosphomonoester
(PME) and cell membrane degradation products, the phosphodiesters (PDE). The
increased SNR and increased spectral resolution at 7T allow discrimination of
individual PME metabolites namely phosphocholine (PC) and phosphoethanolamine
(PE), but also individual PDE metabolites glycerophosphocholine (GPC) and
glycerophosphoethanolamine (GPE). Previous studies show alterations of PME to
PDE ratios (PC to GPC, PE to GPE) suggest proliferation and are often found in
tumor tissue [4, 11, 26-30]. Changes in these ratios during therapy are markers
of therapy response and take place well before morphological changes can be
observed [31-35].

However, to the best of our knowledge, up until now no clinical studies using
whole body multinuclear MR spectroscopic imaging have been performed. In order
to lay the foundation to the study of human biology using non-invasive
chemistry imaging, unique nuclei at 7T and corresponding acquisition strategies
for patient friendly scan times will be used. To demonstrate the feasibility of
dynamic 3D whole body chemical imaging within patient* tolerable scan times and
its potential for early response monitoring and therapy prediction, MAESTRO
will research a stratification strategy for patients with liver metastasis of
gastro-oesophageal cancer receiving first line palliative chemotherapy and
immunotherapy when applicable, in a 70 patient observational study.

With the MAESTRO approach, the consortium aims at reducing the 9-weeks period
before therapy efficacy evaluation to less than three weeks (> 66% reduction).
Moreover, MAESTRO aims to develop a new stratification strategy to dynamically
differentiate between responders and non-responders two months earlier than
current clinical standards by assessing differentiating biochemical markers by
metabolic imaging throughout the first-line treatment course of individu

Primary objective
• In this study we will investigate whether biochemical imaging of change (Δ,
figure 1) in the metabolic phospholipid ratios of PME and PDE between baseline
and after 2 weeks of therapy are predictive for progression free survival (PFS)
and/or overall survival (OS) in gastro-oesophageal cancer patients within 27
weeks of treatment.

Secondary objectives:
• In this study we will investigate whether biochemical imaging of change (Δ,
figure 1) in the metabolic phospholipid ratios of PME and PDE between baseline
and after 2 weeks of therapy are predictive for RECIST progression in
gastro-oesophageal cancer patients after the first 9 week treatment period.
• Investigate whether biochemical imaging of the metabolic phospholipid ratios
of PME and PDE at baseline of therapy are predictive for RECIST progression
after the first 9-week treatment period, and for PFS and OS in
gastro-oesophageal cancer patients.
• Investigate whether biochemical imaging of change (δ, figure 1) in the
metabolic phospholipid ratios of PME and PDE after a 9-week treatment period
are predictive for RECIST progression following that treatment period, and for
PFS and OS in gastro-oesophageal cancer patients.
• Exploratory multi variable analysis for the development of a prediction model
to predict resistance to treatment within 3 weeks after the start of
chemotherapy with the use of all chemistry imaging data including all MR
detectable nuclei and clinical parameters.

Figure 1: C1 - research protocol, chapter 3; study design.

In this multi-centre observational cohort study, patients with liver metastasis
of gastro-esophagogastric cancer will be asked for participation in the study.
After eligible patients have signed informed consent, patients will continue to
receive standard treatment as scheduled. Standard CT-scans at baseline, t = 9
and 18 weeks will be supplemented with MAESTRO-scans. In addition, after the
first chemotherapy cycle at t = 2 weeks, a MAESTRO-scan will be made to assess
early response monitoring. Patients are followed until disease progression
according to RECIST 1.1 progression criteria is detected by CT scan or until t
= 27 weeks. The scan time of MAESTRO-scans will approximately be one hour.

To meet the primary objective the baseline and 2 weeks after start therapy
chemical imaging measurements (Δ, figure 1) are used to predict the RECIST
progression 27 weeks after the start of the treatment. In addition, solely the
baseline measurement will be used to predict the RECIST progression, OS and
PFS, after the 9-week treatment period to meet the secondary objectives. All
other imaging moments every 9 weeks (δ, figure 1) of therapy are used to
predict RECIST progression, OS and PFS following the 9-weeks period of
treatment.



(NB. See chapter 3; study design of C1, the research protocol for figure 1.)

Patients will be asked for three extra hospital visits to undergo 7T MRI of
approximately one hour per session (3 x 1 hour). MRI is a safe non-invasive
technique without use of ionizing radiation and so far, extensive research has
not shown any side-effects of the high magnetic field used in 7T MRI, resulting
in low inherent risks for the participants. Patients' therapy is not delayed by
participation in this study and patients with MRI contraindications are
excluded from participation.

A potential subject who meets any of the following criteria will be excluded
from participation in this study:
• Any psychological, familial, sociological or geographical condition
potentially hampering adequate informed consent or compliance with the study
protocol.
• Contra-indications for MR scanning, including patients with a pacemaker,
cochlear implant or neurostimulator; patients with non-MR compatible metallic
implants in their eye, spine, thorax or abdomen; or a non-MR compatible
aneurysm clip in their brain; patients with claustrophobia; patients with an
abdominal circumference which exceeds MRI-bore circumference.