This document presents the results of a study aiming to identify potential noninvasive biomarkers for diagnosing idiopathic portal hypertension (IPH) through metabolomic profiling. The study analyzed plasma samples from 33 IPH patients, 33 cirrhotic patients, and 33 healthy volunteers. Multivariate analysis identified subsets of 28 and 31 metabolites that differentiated IPH patients from cirrhotic patients and healthy volunteers, respectively, with excellent predictive accuracy. While validation in larger studies is still needed, the identified metabolic profiles may provide noninvasive means of diagnosing IPH and gaining insight into its pathogenesis.

1. Presented by:
Nagi Abdalla
엠네기
Metabolomics Discloses Potential Biomarkers for
the
Noninvasive Diagnosis of Idiopathic Portal
Hypertension

2. Contents
 Introduction / Idiopathic Portal Hypertension
 Methods
 Results
 Discussion & Conclusion

3. INTRODUCTION

7. Idiopathic Portal Hypertension
 (Noncirrhotic Portal Fibrosis, Hepatoportal Sclerosis)
 Etiology: (unknown)
 Recurrent infections
 Altered immune response
 Genetic predisposition
 (HLA-DR3 )
 Hypercoagulability
 HIV infection
 Diagnosis :
 based on the presence of unequivocal portal
hypertension and a definite histological exclusion of
cirrhosis and any other specific disorder that is able to

8. Study Aim
 The need for a less invasive diagnostic method
 Liver biopsy might not be helpful to diagnose IPH
 Importance of metabolomics in clinical research..
 Aim of the study:
“to discover a noninvasive metabolomic profile in
plasma allowing differentiating IPH from healthy
individuals and from patients with CH”

9. METHODS

10. Patients
 Diagnosis criteria:
 Signs of portal hypertension
 Exclusion of cirrhosis
 Exclusion of hepatic venous thrombosis
 Other exclusions:
 Patients with other conditions such as thrombosis,
hepatocellular CA, liver biopsy with <6 complete portal
tracts
 HIV patients were only included if IPH diagnosis was
unequivocal
 “only patients with unequivocal IPH were included in
the study”

11.  Blood samples: were collected in citrate-containing tubes
and centrifuged then stored at -80C
 99 samples were collected
 Ethical statement: informed consent was given to all
participants
 Experimental procedures:
(A global metabolite profiling UPLC-MS methodology)
 LC-MS system:
 UPLC-(TOF)MS
 Source: ESI @150◦C
 Column: 1 mm i.d. × 100 mm Acquity 1.7 μm C8 BEH column
(Waters)
 M.P: A:0.05%FA B: CAN (0.05%FA) gradient flow

12.  Data processing:
 LC-MS data processing: Noise reduction  identify relevant
peak intensities  normalization to other peaks in the sample
 inter-assay normalization to reference sample following
linear regression method.
 Pairwise univariate data analysis was performed in IPH vs.
CH samples and IPH vs. HVs, to eliminate biomarkers that do
not discriminate between groups

13.  Multivariate data analysis:
 Missing variables were not considered
 t-test P value corrected by multiple comparisons and VIP
score (estimates the importance of each variable in the
projection PLS model) “VIP ≥1”
 Results are 202 (IPH-CH) and 57 (IPH-HV) significant
markers (P<0.05)
 Markers with higher VIP (2.2/2.1) were selected to build a
PLS-DA model to discriminate IPH from CH and IPH from HV.
 Markers selection is based on strong parameters: (1-0.7) of both:
 R2 (goodness of fit)
 Q2 (goodness of prediction)

14.  Model validation was done by:
 using training (2/3 of data) and test sets (1/3 of data) to predict class
membership and class discrimination [X100 “random” times]
 corresponding random sampling cross-validated AUC measures were
determined for each set as (mean±SD) to check sensitivity and
specificity
 Heatmaps were created to represent the selected models
 A hierarchical clustering algorithm was performed on both
variables and samples.

15. RESULTS

16. Main clinical characteristics of the
patients included in the study
Variables IPH, n=33 Cirrhosis, n=33
Healthy
volunteers, n=33
Age at time of blood sample (years) 42±16**
59±8 39±10
Gender (male), n (%) 21 (64) 27 (82) 19 (58)
Signs of portal hypertension, n (%)
Varices 28 (85) 25 (76) —
Variceal bleeding 13 (39) 8 (24) —
Ascites 10 (30) 11 (33) —
Hepatic encephalopathy 0 2 (6) —
Laboratory data
Hematocrit (%) 39±5.6#
39±6.4 41±3.2
Platelet count (× 109
/l) 114±92##
99±36.9 265±49.5
Creatinine (mg/dl) 0.9±0.2 0.86±0.3 0.89±0.2
AST (UI/l) 36±15*,##
67±44 19±4.7
ALT (UI/l) 41±29*,#
66±49 20±10.2
Albumin (g/l) 41±5.3**
37±3.9 43±3.1
Bilirubin (mg/dl) 1.3±1.2#
1.1±0.4 0.7±0.3
Prothrombin ratio (%) 78±13##
79±11 93±7.9
Child-Pugh class, n (%)
A 27 (82) 27 (82) —
B 6 (18) 6 (18) —
C 0 0 —

17. Metabolites analysis: PLS-DA
(HPI vs. CH)

18. Metabolites analysis: PLS-DA
(HPI vs. HV)

19. DISCUSSION

20.  The PLS-DA models show a clear differentiation of IPH
vs. cirrhotic patients & IPH vs. healthy controls based
on a subset of 28 & 31 metabolites respectively, with an
excellent predictive power (based on R2 and Q2 values)
& AUC.
 The cross-validation showed an excellent performance
of both models with a good sensibility, specificity, and
AUC in the training and testing sets.
 In this study: sub analysis of the metabolomic profile of
IPH patients was unable to cluster patients into different
IPH groups and the author suggested to study larger
population of patients
 Thus this study supports the use of metabolomic
profiling to diagnose the disease rather than identifying
the etiology
• Some of the detected metabolites may reflect some of
the drugs that patients are taking. However, it seems

21. Study limitations
 Patient number:
 however, as IPH is a rare condition, a sample over 30
patients could be considered adequate
 lack of an external validation set:
 since this is a pilot study; such external validation
studies will be more appropriate at a later step, when the
specific metabolites included in the models could be
identified with new technologies.
 However, the existence of metabolites discriminating
IPH from CH and HV opens the interesting possibility
that the identification of these specific metabolites
may disclose some keys for a better understanding of
the pathogenesis of IPH

22. CONCLUSION
 The results from this study disclose a subset of
putative biomarkers of IPH
 patients with IPH could be identified based on
their metabolic profile, obviating the need for
invasive investigations and facilitating the correct
diagnosis of this uncommon disease.

23. Thank you for listening
감사합니다!