Molecular spectroscopy of blood plasma differentiates metabolic dysfunction-associated steatohepatitis from steatosis

Abstract

Background

Vibrational and chiroptical spectroscopy of blood plasma is an advanced experimental diagnostic approach that allows the identification of disease-specific molecular patterns. This study aimed to test its potential to distinguish between different stages of metabolic dysfunction-associated steatotic liver disease (MASLD) including liver fibrosis.

Methods

We analyzed blood plasma samples from 29 patients with metabolic dysfunction-associated steatohepatitis (MASH) and 24 MASLD patients with simple steatosis using Fourier-transform infrared (FTIR) spectroscopy, Raman spectroscopy and electronic circular dichroism (ECD) spectroscopy to distinguish: (i) MASH; (ii) MASH with moderate to advanced fibrosis; and (iii) overall fibrosis.

Results

FTIR spectroscopy distinguished MASH from simple steatosis with a sensitivity of 73% and a specificity of 92%, with an area under the receiver operating characteristic curve (AUROC) of 0.92, p < 0.001. In addition, FTIR spectroscopy identified MASH with moderate to advanced fibrosis (F2–3) from MASH/steatosis with no to mild fibrosis (F0–1) with a sensitivity of 87% and specificity of 78% (AUROC 0.91, p < 0.001). Furthermore, regardless of the MASH/steatosis diagnosis, fibrosis (F1–3) was distinguished from F0 by the combination of data from the three spectroscopic methods with a sensitivity of 79% and a specificity of 88% (AUROC 0.91, p < 0.001). Both FTIR and Raman spectroscopy alone discriminated all target groups with an AUROC > 0.82.

Conclusions

This pilot study revealed the potential of blood plasma spectroscopy to identify and differentiate patients with various stages of MASLD.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12967-025-06885-6.

Background

Metabolic dysfunction-associated steatotic liver disease (MASLD) is currently the most common chronic liver disease [1] and is strongly associated with metabolic syndrome, affecting approximately 38% of the global population [2]. The prevalence of MASLD is expected to continue to increase exponentially by 2030 [3]. The pathophysiological basis is complex, involving mainly the effects of genetic predisposition and the gut microbiome (gut-liver axis), an energy-dense diet and a sedentary lifestyle, all evolving in systemic insulin resistance, lipotoxicity, adaptive immune mechanisms, and inflammation as major drivers of hepatic steatosis [4].

Approximately 30% of patients with MASLD can develop metabolic dysfunction-associated steatohepatitis (MASH) [5], which can progress to liver cirrhosis and liver cancer [6]. Consequently, patients with MASH (and especially with MASH with significant fibrosis ≥ F2 grade, “at-risk MASH”) contribute to the increase in the liver related, as well as non-liver related complications and death [7]. Furthermore, an increasing need for liver transplants for hepatocellular carcinoma and decompensated cirrhosis in MASH is associated with a significant financial burden [5].

The accurate diagnosis of MASH, often needed in clinical pharmacologic trials, is still based on liver biopsy [8, 9], which is unacceptable for screening due to its invasiveness and other limitations [10]. Similarly, the proton density of the fat fraction, the “gold standard” for the evaluation of steatosis, relies on advanced liver imaging and carries substantial financial costs [11]. However, conventional noninvasive fibrosis markers have only moderate diagnostic performance for MASH with fibrosis grade ≥ F2 [12]. Therefore, additional noninvasive blood markers for high-risk MASH are intensively sought.

Among metabolomic approaches [13], spectroscopy-based methods (such as nuclear magnetic resonance spectroscopy) have been successfully tested for metabolic profiling of MASLD [14]. Recently, a relatively new approach of vibrational and chiroptical spectroscopy of blood plasma proteins revealed conformational changes in plasma proteins specific for liver cirrhosis and hepatocellular carcinoma in cirrhosis, including those of the MASH etiology [15]. Furthermore, Ten-Doménech et al. [16] in their recent review focusing on the role of mid-infrared spectroscopy, described infrared spectroscopy-based methods as a rapidly emerging promising real-time diagnostic tool for steatosis. Infrared spectroscopy of blood plasma has also been used successfully to discriminate between patients with MASH and steatosis [17–20].

In this respect, vibrational (Raman and Fourier-transform infrared (FTIR) spectroscopy) and chiroptical spectroscopy (electronic circular dichroism (ECD)) are excellent methods for analyzing disease-derived changes in molecular composition and structure in blood serum/plasma that offer a rapid and non-destructive analysis of small-volume samples without extensive sample preparation [21–23].

Furthermore, a combination of vibrational and chiroptical methods has been shown to improve diagnostic accuracy, overcoming the limitations of single methods (such as water interference in FTIR and fluorescence in Raman spectroscopy) and providing complex information about molecules, including chemical structure and relative abundance, which may be altered in MASLD/MASH [21, 24, 25].

The objectives of this study were to test the potential of vibrational and chiroptical spectroscopy of blood plasma to identify MASLD patients with MASH and, secondarily, to investigate its ability to detect the presence and stage of liver fibrosis.

Materials and methods

Patients

The study population of this retrospective study included patients with MASLD from the 4th Department of Internal Medicine of the General University Hospital in Prague and the Liberec Regional Hospital in the Czech Republic who were followed up between 2009 and 2022. All patients who were available and met the inclusion criteria were consecutively admitted to the study. Some participants were recruited from our previous study [26].

Inclusion criteria were as follows: age 18 to 80 years, MASH or simple steatosis diagnosed by liver biopsy, while exclusion criteria involved secondary causes of fat accumulation in the liver or other liver diseases, such as viral hepatitis, drug-induced liver disease, autoimmune liver disease, biliary tract disease and hereditary metabolic diseases, presence of cirrhosis (fibrosis grade F4) in liver biopsy, history of alcohol abuse (based on high GGT activity and serum carbohydrate deficient-transferrin and/or urinary ethyl glucuronide concentrations, and personal history of alcohol abuse), and insufficient amount of sample material.

The target groups were defined as follows:

• i)MASH to be distinguished from Simple steatosis;
• ii)MASH with moderate to advanced fibrosis grade F2–3 (at-risk MASH; hereafter referred to as “MASH + F2–3”) to be distinguished from MASH or steatosis with no to mild fibrosis grade F0–1 (“F0–1”);
• iii)Overall fibrosis grade, F1–3 (“Fibrosis”) to be distinguished from the absence of fibrosis (“Steatosis (F0)”), regardless of the MASH status.

The study was approved by the Ethics Committee of the General University Hospital in Prague, Prague, Czech Republic on 22nd July 2021 (no. 83/21). All study participants provided their written informed consent at the time of liver biopsy, including their consent for future evaluation of stored blood samples. All research was carried out in accordance with both the Declarations of Helsinki and Istanbul.

The study was registered on ClinicalTrials.gov (NCT06647095) and complied with the STARD statement criteria [27].

Liver biopsy

Liver biopsies were performed according to clinical guidelines, routinely stained, and evaluated by a pathologist. The diagnosis of MASH or steatosis was made according to Kleiner et al. [28]. The stage of liver fibrosis was determined using Brunt scoring [29].

Sampling and routine biochemical analysis.

Fasting peripheral blood samples were collected from all participants and used immediately for routine parameter measurements or stored at − 80 °C until analysis. Routine biochemical parameters were measured using an automatic Cobas 8000 analyzer (Roche Diagnostics, Rotkreuz, Switzerland).

Spectroscopic methods

The frozen blood plasma samples were thawed at room temperature and immediately filtered using spin columns with a polyvinylidene difluoride membrane with a porosity of 0.45 μm (Ciro, Deerfield Beach, FL, USA). The samples were centrifuged for 10 min at 10,000 g at 10 °C.

FTIR spectra were acquired on a Nicolet iS50 FTIR spectrometer (Thermo Fisher Scientific, Waltham, MA, USA) using an attenuated total reflection technique with a zinc selenide crystal. Spectra were collected in the spectral region 4000–650 cm^− 1 with a resolution of 4 cm^− 1 and 256 co-added scans. Additionally, the water and water vapor spectra were acquired under the same experimental conditions and subsequently subtracted from the blood plasma sample spectra. The spectra were normalized to the unit vector, and a linear baseline correction was applied in the region of 2000–1800 cm^–1.

Raman spectra were acquired on a DXR SmartRaman spectrometer (Thermo Fisher Scientific, Waltham, MA, USA) equipped with a 780 nm diode laser. Ten accumulations with an acquisition time of 60 s and a laser power of 90 mW were acquired in the spectral region of 300–2000 cm^− 1 with a spectral resolution of < 4 cm^− 1 for each sample. The spectra were baseline corrected using the open source BubbleFill algorithm (minimum bubble width 250 cm^–1), smoothed using Savitzky-Golay (window length of 15, third order polynomial), and unit vector normalized.

ECD spectra were recorded using a J-815 spectrometer (Jasco, Tokyo, Japan) with a Peltier cell holder to control the sample temperature (23 °C). Samples were diluted with sterile phosphate-buffered saline (pH 7.4) in a 1:3 volume ratio and placed in a cell with an optical path length of 0.01 mm. Spectra were acquired in the spectral region of 280–185 nm, and six accumulations with a scanning speed of 50 nm/min, a response time of 2 s, a data pitch of 0.1 nm and a sensitivity of 100 mdeg were collected. Data were processed using standard normal variate.

Statistical analyses

Statistical differences in characteristics between the groups of patients were analyzed using Statistica 13.5 software (TIBCO Software, Palo Alto, CA, USA). Cases with incomplete data were discarded from the corresponding analyzes. The Mann-Whitney U-test was used to compare continuous and ordinal variables between the MASH and steatosis groups. The binary variables were compared using Fisher’s exact test. The acquired p-values (two-tailed) were Bonferroni corrected and only p-values < 0.05 were considered significant.

The classification model creation and feature selection were performed in the R software as previously described [30]. Briefly, the partial least squares-discrimination analysis algorithm with repeated k-fold cross-validation (5-fold, ten repetitions) was used. The models were created with an emphasis on the highest area under the receiver operating characteristics curve (AUROC), and the optimal number of latent variables was determined using the “oneSE” selection function. Initially, the full spectral range was used to build the classification model and only the variables with the highest impact on classification were selected to create the final classification model for each method. The data were always mean-centered and when using only the selected variables also autoscaled. To ensure that the models were not overfitted, permutation tests with thousand repetitions were performed (p-value < 0.05 indicates a statistically significant model).

Results

Patients and cohorts

Of the 99 initially considered patients who underwent liver biopsy, we excluded 46 patients according to exclusion criteria. Finally, 53 patients were included in the study (29 MASH, 24 steatosis patients, all biopsy-proven; see Fig. 1 Flowchart).

Fig. 1.

Study flowchart

Abbreviations: MASLD, metabolic dysfunction-associated steatotic liver disease; MASH, metabolic dysfunction-associated steatohepatitis

Participants in our study were predominantly men (63–83%), with no significant age difference between the compared groups. In the MASH study group, a more severe grade of liver fibrosis prevailed (p = 0.011), as moderate to advanced fibrosis (grade F2–3) was present in 62% of MASH patients compared to 25% in cases of simple steatosis. Patients with MASH also had approximately 2.5 times higher values of alanine aminotransferase (ALT) activities (p = 0.004). The clinical, biochemical and histological characteristics of the cohorts are summarized in Table 1, and further details are provided in Tables S1a–c in the Supplementary Material. All patients met the definition of MASLD as recently defined [1].

Table 1.

Characteristics of study cohorts

	MASH	Simple steatosis	MASH + F2–3	F0–1	Fibrosis	Steatosis (F0)
Study subjects [n]	29	24	15	29	39	14
Age [years]	36.6 (32.0–57.0)	51.4 (42.9–61.0)	48.0 (31.3–57.4)	46.7 (33.0–60.3)	47.1 (33.6–58.5)	47.5 (33.6–58.5)
Men [%]	82.8	62.5	73.3	72.4	76.9	64.3
BMI [kg/m2]	30.1 (27.8–33.4) b	30.8 (25.4–33.9) a	31.9 (27.8–35.0) b	29.7 (27.4–33.4) a	30.9 (27.4–34.4) b	29.6 (26.7–31.2) b
Liver fibrosis gradea [n]	*
F0	3	11	-	14	-	14
F1	8	7	-	15	15	-
F2	6	5	5	-	11	-
F3	12	1	10	-	13	-
Bilirubin [µmol/L]	10.0 (7.9–15.3) b	14.5 (10.9–17.9) d	9.2 (7.3–11.2) c	11.2 (10.2–17.8) c	10.6 (8.6–17.3) b	11.2 (10.0–16.2) d
ALT [µkat/L]	1.8 (0.9–2.6) c **	0.7 (0.5–1.0)	2.0 (1.3–3.3) c	0.9 (0.5–1.8)	1.1 (0.7–2.2) b	0.8 (0.5–1.5)
AST [µkat/L]	1.1 (0.7–1.5) b	0.6 (0.5–0.9)	1.2 (0.9–1.5) c	0.8 (0.5–1.1)	0.9 (0.6–1.3) b	0.6 (0.5–1.0)
GGT [µkat/L]	2.0 (1.0–5.3) b	1.5 (0.5–3.3)	3.9 (2.0–5.6) c	1.2 (0.6–2.3)	1.9 (0.7–5.0) b	1.4 (0.6–2.7)
ALP [µkat/L]	1.2 (1.0–1.5) c	1.0 (0.5–3.3)	1.1 (1.1–1.7) d	1.1 (0.9–1.3)	1.1 (1.0–1.5) b	1.0 (0.8–1.3)
s-TAG [mmol/L]	1.6 (1.3–2.0) b	1.1 (0.8–1.7) a	1.7 (1.3–1.9) b	1.4 (0.8–1.7) a	1.6 (1.1–2.0) b	1.5 (0.8–1.7) b
T2DM [n]	7	5	4	6	9	3

* Bonferroni corrected p < 0.05, ** <0.01 between the MASH and Steatosis groups

^{a, b, c, d} Incomplete data: Missing values ≤ 5% ^a, ≤ 10% ^b, ≤ 20% ^c, ≤ 30% ^d

Data are presented as median and interquartile range

ALP, alkaline phosphatase; ALT, alanine aminotransferase; AST, aspartate aminotransferase; BMI, body mass index; GGT, gamma-glutamyl transferase; T2DM, type 2 diabetes mellitus; s-TAG, serum triacyclglycerols

The results of the individual classification models for each spectroscopic method and their combinations are summarized in Table 2 (for the corresponding confusion matrices, see Tables S2–13 in Suppl. Material). Of the individual methods, FTIR and Raman spectroscopy achieved diagnostic accuracy superior to that of ECD spectroscopy for all types of models, with AUROC > 0.82. FTIR spectroscopy performed the best in identifying MASH (AUROC 0.92, sensitivity 73%, specificity 92%) and MASH + F2–3 (AUROC 0.91, sensitivity 87%, specificity 78%), while Raman spectroscopy best detected the general presence of fibrosis (AUROC 0.84, sensitivity 73%, specificity 81%).

Table 2.

Results of all classification models for individual spectroscopic methods and their combination

MASH (n = 29) x Simple steatosis (n = 24)
Method	AUROC	Sensitivity	Specificity	Permutation test (p-value)
FTIR spectroscopy	0.922 (0.896–0.949)	0.733 (0.673–0.792)	0.920 (0.883–0.957)	< 0.001
Raman spectroscopy	0.847 (0.809–0.885)	0.786 (0.735–0.838)	0.726 (0.667–0.785)	< 0.001
ECD spectroscopy	0.752 (0.705–0.798)	0.659 (0.598–0.719)	0.663 (0.604–0.722)	< 0.001
Combination a	0.908 (0.881–0.936)	0.790 (0.737–0.843)	0.869 (0.820–0.918)	< 0.001
MASH + F2–3b(n = 15) x F0–1c(n = 29)
Method	AUROC	Sensitivity	Specificity	Permutation test (p-value)
FTIR spectroscopy	0.905 (0.878–0.933)	0.873 (0.824–0.922)	0.783 (0.743–0.824)	< 0.001
Raman spectroscopy	0.848 (0.811–0.885)	0.707 (0.622–0.792)	0.759 (0.705–0.813)	< 0.001
ECD spectroscopy	0.783 (0.740–0.825)	0.7533 (0.679–0.828)	0.665 (0.618–0.712)	< 0.001
Combination a	0.899 (0.870–0.928)	0.813 (0.748–0.878)	0.811 (0.766–0.856)	< 0.001
Fibrosisd(n = 39) x Steatosis (F0), e(n = 14)
Method	AUROC	Sensitivity	Specificity	Permutation test (p-value)
FTIR spectroscopy	0.824 (0.782–0.867)	0.706 (0.654–0.757)	0.770 (0.694–0.846)	< 0.001
Raman spectroscopy	0.837 (0.799–0.875)	0.726 (0.678–0.775)	0.8067 (0.742–0.871)	< 0.001
ECD spectroscopy	0.777 (0.732–0.822)	0.665 (0.615–0.716)	0.810 (0.740–0.881)	< 0.001
Combination a	0.909 (0.881–0.937)	0.792 (0.747–0.837)	0.880 (0.820–0.941)	< 0.001

^a Combination of FTIR, Raman, and ECD spectroscopic methods; ^b Patients with MASH and fibrosis grades F2–3; ^c Patients with MASH or steatosis with no/F1 fibrosis; ^d Fibrosis of any grade (F1–3); ^e Absence of fibrosis, regardless of MASH status

Values are presented as average and 95% confidence interval

AUROC, area under the receiver operating characteristic curve; ECD, electronic circular dichroism; FTIR, Fourier-transform infrared

Combined statistical models were created using only the variables selected from the individual method models. The model distinguishing MASH from Simple steatosis reached an AUROC of 0.91, with a sensitivity and specificity of 79% and 87%, respectively. The identification of MASH + F2–3 was achieved with a balanced sensitivity and specificity of 81% (AUROC 0.90). When discriminating fibrosis of any grade from steatosis (F0), the model had a sensitivity of 79% and a specificity of 88% with an AUROC of 0.91.

The permutation tests for all created statistical models reached p-values < 0.001, confirming that the statistical models were not overfitted (see Suppl. Material, Fig. S1 for MASH; Fig. S2 for MASH + F2–3; Fig. S3 for Fibrosis).

The average FTIR, Raman and ECD spectra of blood plasma from patients with MASH compared to steatosis, together with the difference spectra identifying the regions with the highest spectral differences between the two groups of samples, are presented in Figs. 2–4 (see Suppl. Material for the comparison of the other groups of samples, Figs. S4–9).

Fig. 2.

Average and difference spectra of blood plasma of patients with MASH (red) and steatosis (blue) analyzed by FTIR spectroscopy

The black points in the average spectra depict spectral variables included in the final classification models, while the shaded areas represent the standard deviation of intensity

FTIR, Fourier-transform infrared

Fig. 4.

Average and difference spectra of blood plasma of patients with MASH (red) and steatosis (blue) analyzed by ECD spectroscopy

The black points in the average spectra depict spectral variables included in the final classification models, while the shaded areas represent the standard deviation of intensity

ECD, electronic circular dichroism

In general, the spectra of blood plasma predominantly contained spectral bands originating from vibrations or electronic transitions within proteins, amino acids, and (phospho)lipids. A more detailed assignment of the bands in the spectra of the blood plasma for each method can be found in Table 3.

Table 3.

Position and assignment of the bands in the FTIR, Raman, and ECD spectra of blood plasma

Molecule	FTIR spectroscopy (cm− 1)	Raman spectroscopy (cm− 1)	ECD spectroscopy (nm)
proteins	1645; 1548; 1454; 1400; 1315	900; 940; 1128; 1230–1350; 1450; 1656	192; 209; 220
(phospho)lipids	1740; 1243	700; 718; 1083; 1127; 1450
amino acids		508; 573; 621; 645; 664; 760; 830; 850; 880; 1004; 1175; 1210; 1450; 1620
nucleic acids	1243	718; 745; 1175; 1450
cholesterol		540; 700; 957
carotenoids		1004; 1156; 1524

ECD, electronic circular dichroism; FTIR, Fourier-transform infrared

Regarding specific molecule alterations, our study identified a decrease in normalized FTIR absorbance in the amide I region (1645 cm^–1) and an increase in amide II (1548 cm^–1) in the MASH (Fig. 2) and MASH + F2–3 groups (Fig. S4). We also detected a decrease in the region corresponding to cholesterol (FTIR 1740 cm^–1) in MASH (Fig. 2). Furthermore, we identified an increase in the intensity of phenylalanine (Raman band at 1004 cm^− 1) in MASH (Fig. 3), and a slight increase in the intensity of phenylalanine in fibrosis (Fig. S8).

Fig. 3.

Average and difference spectra of blood plasma of patients with MASH (red) and steatosis (blue) analyzed by Raman spectroscopy

The black points in the average spectra depict spectral variables included in the final classification models, while the shaded areas represent the standard deviation of intensity

Key outcomes – vibrational and chiroptical spectroscopy determined:

1. MASH: FTIR spectroscopy discriminated MASH from Simple steatosis with a sensitivity of 73% and a specificity of 92% (AUROC 0.92, Table 2).

2. MASH + F2–3: Similarly, FTIR spectroscopy distinguished MASH + F2–3 from F0–1 with a sensitivity of 87% and a specificity of 78% (AUROC 0.91, Table 2).

3. Fibrosis: The combination of FTIR, Raman and ECD spectroscopy differentiated Fibrosis (F1-3) from Steatosis (F0) with a sensitivity of 79% and a specificity of 88% (AUROC 0.91), Table 2).

Discussion

In this study, we demonstrate that the examination of patient blood plasma using molecular spectroscopic methods (FTIR, Raman, and ECD spectroscopy) significantly differentiated: (i) MASH from simple steatosis; (ii) MASH + F2–3 from F0–1 fibrosis; and (iii) fibrosis of any grade (F1-3) from steatosis without fibrosis (F0).

Focusing on the individual methods, FTIR spectroscopy achieved a high AUROC, ranging from 0.82 to 0.92 in all models, suggesting a high sensitivity of the method to changes in blood plasma composition across the various conditions studied. Comparably, Raman spectroscopy provided AUROC values between 0.84 and 0.85 for all models, also showing its potential in the field of MASLD and fibrosis diagnostics. On the contrary, ECD spectroscopy showed considerably lower AUROC values for all models (ranging 0.75 to 0.78), which is presumably related to the substantially limited spectral information compared to both FTIR and Raman spectroscopic methods arising from the different physical principle of ECD.

In general, statistical models based on the combination of all three spectroscopic methods have achieved diagnostic accuracy comparable to or higher than that obtained with the individual methods. Specifically, in identifying liver fibrosis, the combined model provided a significantly higher AUROC than the individual methods. Furthermore, the performance characteristics of the sensitivity and specificity of the combined models seem more balanced in most cases. Therefore, the benefits of a multiparametric approach in the spectroscopic analysis of blood plasma should always be considered when designing the diagnostic study.

Regarding spectral change identification/interpretation, the observed differences between spectra of individual study groups (Figs. 2–4, Figs. S4–9 in Suppl. Material) are minor, which is expected considering the similarity of the pathological conditions being differentiated. However, some differences visible mainly in difference spectra and/or detected as significant by classification models can be tentatively attributed to the biochemical processes associated with the progression of MASLD and fibrogenesis, such as a decrease in amide I/II ratio, previously linked to oxidative stress in insulin resistance [31]. Furthermore, we found changes in the region corresponding to cholesterol, which could be due to the perturbed formation and hydrolysis of cholesterol esters in MASH reported previously [32–35]. Additionally, we also identified changes in amino acids in MASH such as phenylalanine described recently [36, 37].

Only a few other studies on the discrimination of MASLD using a spectroscopic approach have been reported so far, mostly limited to infrared spectroscopy [17–20]. Two pilot studies by Le Corvec et al. [17] and Bensaid et al. [18] have shown promising results of FTIR spectroscopy of blood serum in partial discrimination of patients with MASH from those with steatosis. The first study achieved an accuracy of 81% with a sensitivity and specificity of 83% and 78%, respectively, while the latter study reported an AUROC of 0.89, sensitivity of 74% and specificity of 97%. Another clinical study by Anty et al. [19], performing a FTIR spectroscopic analysis of blood serum from morbidly obese patients referred for bariatric surgery, differentiated MASH from steatosis with an AUROC of 0.84 with a sensitivity of 67% and specificity of 80%. Furthermore, Anty et al. [19] have demonstrated that FTIR spectroscopy can significantly discriminate between different grades of inflammation, ballooning, and nonalcoholic fatty liver disease activity scores. Gao et al. [38] used surface-enhanced Raman scattering spectroscopy to distinguish between MASH and MASLD using silver nanoparticles and blood serum. The created classification model yielded an AUROC of 0.83 for the validation cohort. A similar study was conducted by Gurian et al. [39], who used silver nanoparticles to enhance the Raman signals in blood plasma of subjects with MASH and MASLD. Using this approach, they were able to discriminate between the studied groups with an AUROC of 0.81.

Our results are generally consistent with those of the studies. In addition to previously published data, we explored the potential of two other spectroscopic methods (ECD and conventional Raman spectroscopy), as well as a combination of these methods with FTIR spectroscopy, and demonstrated the ability of blood plasma spectroscopy to identify the presence and grade of fibrosis.

Together, blood plasma spectroscopy was highly effective in stratifying MASLD and fibrosis. However, our study has some limitations. This is primarily a pilot study involving a relatively small sample of patients, which may affect the generalizability of the results. However, appropriate statistical methods were employed to address this limitation, including repeated k-fold cross-validation with down-sampling to improve model stability and permutation tests to assess statistical significance. Our findings are also limited by the retrospective design, hence there is a need for prospective validation. In addition, patients with cirrhosis were excluded from this study because cirrhosis can be assessed easily using non-invasive methods. Furthermore, it complicates the classification of MASH/steatosis when using liver biopsy. Although this may limit the applicability of the method in late disease stages, we consider the group of “at-risk MASH” with advanced F2-3 fibrosis and concurrent MASH to be the most important target group in terms of potential therapy.”

The general availability of spectroscopy is limited by the need for specific measurement skills and equipment. Nevertheless, the current trend calls for the development of portable, easy-to-use spectroscopic devices that are operable by non-scientists [40]. Such a sensor (based on Raman spectroscopy) has already been successfully tested for diagnosing MASH [39]. Therefore, future developments in spectroscopy might help bring MASH diagnostics from bench to bedside.

Furthermore, the translation of spectroscopic methods into clinical settings may be facilitated by the introduction of automation throughout the entire diagnostics process. Spectrometers equipped with autosamplers are already available [41, 42], enabling continuous sample analysis. The potential complications briefly mentioned earlier, i.e., water interference in FTIR spectroscopy and fluorescence in Raman spectroscopy, may be mitigated through various established strategies. The former may be realized by analyzing dried samples [41, 43]; however, this approach results in the analysis of biomolecules in their non-native state. An alternative approach used in this study involves analyzing samples in a liquid state and subsequently subtracting the water spectrum from the sample spectrum, a process that may be automated [44]. In the case of fluorescence in Raman spectroscopy, its intensity is related to the instrumentation. With longer excitation wavelength, fluorescence typically decreases. Nevertheless, mathematical correction is often needed, and it can also be performed algorithmically. Thus, the whole pipeline of spectral data processing and subsequent evaluation may be automated, and such tools are available [44, 45].

From a clinical perspective, the highly accurate results of our pilot study, together with previous findings on blood plasma spectroscopy in MASLD [17–19], suggest that this approach may provide a complementary way to identify individuals who are likely to have advanced disease and possibly meet the entry criteria for clinical trials.

The FIB-4 score, the recommended method for determining advanced fibrosis in MASLD, has been reported to have limited accuracy in determining lower grades of fibrosis, as well as in determining MASH [12]. Here we demonstrate the potential of spectroscopic methods to reliably detect fibrosis of any grade, as well as F2–3 fibrosis in MASH.

Following future developments, plasma spectroscopy as a blood-based method could potentially become an alternative to the most accurate non-invasive, liver elastography-based tests for MASH or MASH + F2–3, such as the FibroScan-AST score [12, 46].

In conclusion, spectroscopic analysis of blood plasma has the potential to differentiate patients with MASH from those with steatosis, patients with MASH + moderate to severe fibrosis (F2–3) from those with no to mild fibrosis (F0–1) (MASH/steatosis) and patients with any grade of fibrosis (F1–3) from those without fibrosis. However, our results will need to be confirmed in a prospective study with larger independent cohorts.

Electronic supplementary material

Below is the link to the electronic supplementary material.

Abbreviations

ALP

Alkaline phosphatase

ALT

Alanine aminotransferase

AUROC

Area under the receiver operating characteristic curve

AST

Aspartate aminotransferase

BMI

Body mass index

ECD

Electronic circular dichroism

FTIR

Fourier-transform infrared

GGT

Gamma-glutamyl transferase

MASH

Metabolic dysfunction-associated steatohepatitis

MASLD

Metabolic dysfunction-associated steatotic liver disease

T2DM

Type 2 diabetes mellitus

s-TAG

Serum triacylglycerols

Author contributions

BN: Writing – Original Draft Preparation, Project Administration; OV: Methodology, Software, Validation, Formal analysis, Investigation, Data Curation, Writing – Review & Editing, Visualization; KK: Methodology, Formal analysis, Investigation, Data Curation, Writing – Review & Editing; LH: Conceptualization, Methodology, Writing – Review & Editing; VŠ: Investigation, Writing – Review & Editing; KD: Investigation, Writing – Review & Editing; JP: Investigation, Writing – Review & Editing; KŽ: Investigation, Writing – Review & Editing; JS: Investigation, Writing – Review & Editing; LV: Conceptualization, Supervision, Writing – Review & Editing; ML: Conceptualization, Supervision, Writing – Review & Editing; VS: Methodology, Resources, Supervision, Writing – Review & Editing; RB: Conceptualization, Funding Acquisition, Project Administration, Supervision, Writing – Review & Editing.

Funding

This work was supported by a grant MH CZ–DRO–VFN00064165 and AZV MHCZ NU23-01-00288 given by the Ministry of Health of the Czech Republic and by the Charles University Research program “Cooperatio” and by the Foundation of the Czech Society of Hepatology.

Data availability

The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request.

Declarations

Ethics approval and consent to participate

The study was approved by the Ethics Committee of the General University Hospital in Prague, Prague, Czech Republic on 22nd July 2021 (no. 83/21). All study participants provided their written informed consent at the time of liver biopsy, including their consent for future evaluation of stored blood samples. All research was carried out in accordance with both the Declarations of Helsinki and Istanbul. The study was registered on ClinicalTrials.gov (NCT06647095).

Consent for publication

Not applicable.

Competing interests

The authors declare that they have no competing interests.