miR-34a and miR-223 as diagnostic biomarkers for the severity of non-alcoholic fatty liver disease

Abstract

Non-alcoholic fatty liver disease (NAFLD) ranges from simple steatosis to non-alcoholic steatohepatitis (NASH) and fibrosis, yet current non-invasive tools lack accuracy in distinguishing these stages. This study aimed to evaluate the circulating miR-34a and miR-223 as biomarkers for NAFLD staging. In this cross-sectional study, 200 participants were equally divided into healthy individuals and patients with steatosis, NASH and fibrosis groups. Serum miRNA expression was measured by stem-loop RT-qPCR and miR-16 was used for normalization, while clinical parameters were measured to validate miRNAs. ANOVA and Kruskal-Wallis tests were used for group comparisons, while ROC analysis and multivariate regression were used to assess the diagnostic performance. Both miRNAs showed significant changes in mean expression across disease stages (p < 0.001). miR-34a increased from controls (1.42-fold) to NASH (41.28-fold), followed by a relative decline in fibrosis (12.84-fold), while miR-223 remained low in steatosis before rising sharply in NASH (42.17-fold). Both miRNAs correlated strongly with ALT, AST, platelet count, and lipid parameters (|r| > 0.6, p < 0.01). The combined miRNA panel accurately differentiated steatosis from NASH (AUC = 0.978; 96% sensitivity; 94% specificity). Both were independently associated with advanced disease (miR-34a: OR = 1.32, 95% CI: 1.18–1.48; miR-223: OR = 1.28, 95% CI: 1.15–1.42; p < 0.001), and patients in the highest quartiles had a 116-fold higher risk. The addition of both miRNAs to clinical models improved AUC from 0.812 to 0.912 (NRI = 0.267, p < 0.001). The expression patterns of miR-34a and miR-223 offered a practical non-invasive approach for staging of NAFLD.

Graphical abstract

1. Introduction

Nonalcoholic fatty liver disease (NAFLD), recently renamed metabolic-associated fatty liver disease (MAFLD), is a common cause of chronic liver disease with a global prevalence affecting about a quarter of the world's population.¹ NAFLD is a multistage disease begins with steatosis which is the accumulation of excess fat in liver cells (hepatocytes), progresses to a more severe form accompanied by inflammation and hepatocellular injury known as Nonalcoholic steatohepatitis (NASH), and subsequently to Fibrosis which is characterized by the accumulation of extracellular matrix proteins including collagen and its advanced stage results in cirrhosis that led to liver failure.2, 3, 4 Accurate differentiation among the stages of NAFLD remains clinically important, as early intervention may alter the course of the disease,⁵ yet differentiation between these stages is still considered a diagnostic challenge in routine clinical practice.² Although liver biopsy is the reference standard for diagnosing progressive NASH and fibrosis,⁶ it is limited due to its invasive nature, potential complications such as bleeding, and sampling variability.⁷ Several non-invasive markers are used in liver function tests including liver enzymes such as alanine aminotransferase (ALT) and aspartate aminotransferase (AST), platelet count, fibrosis-4 (FIB-4), and albumin, but unfortunately they suffer from significant limitations in sensitivity (loss of the ability to assess the degree of hepatic fat accumulation), specificity (levels may remain normal even in severe disease stages) and the ability to differentiate between disease stages.⁸ Because of the discrepancy between biochemical tests and histological staging of NAFLD, together with the presence of disadvantages in liver biopsy, it becomes necessary to search for circulating biomarkers capable of reflecting the pathological stages and disease progression.⁹ Molecularly, the progression of NAFLD is influenced by several mechanisms, the most important of which are epigenetic modifications of gene expression.¹⁰ Among these mechanisms, microRNAs (miRNAs) regulate gene expression, and their expression levels have been found to be associated with the development of NAFLD.¹¹ For example, miR-34a has been reported to increase during the transition from simple steatosis to NASH, where it is induced by p53-mediated pathways in response to oxidative stress and lipid accumulation, triggering hepatocyte apoptosis and metabolic dysfunction.12, 13, 14 Another example, miR-223 plays an important role in inflammation when its expression increases as the liver responds to immune signals, helping to drive the shift from simple steatosis to the active inflammation seen in NASH.15, 16, 17 Because these miRNAs are packaged inside extracellular vesicles, they are protected from degradation in the blood. This makes them easier to be detected in serum and highlights their potential as non-invasive biomarkers.¹⁸ Although both miRNAs have been linked individually to liver injury, their combined ability to distinguish between different stages of NAFLD has not been well studied.¹⁹ Despite the progress in molecular profiling and multi-omics research, clinically applicable biomarkers capable of differentiating among various stages of NAFLD are still limited.20, 21 This study aimed to investigate whether the combined assessment of serum miR-34a and miR-223 could improve stage discrimination across a well-characterized cohort including healthy controls, steatosis, NASH, and fibrosis. We proposed that evaluating these miRNAs in parallel may provide a more informative and practical blood-based approach for identifying patients with different stages of NAFLD.

2. Materials and methods

2.1. Study design and participants

This research was designed as a cross-sectional study comprised of 200 participants equally categorized into four groups: healthy controls (26 males/ 24 females, mean age: 30.1 ± 4.2), simple steatosis (26 males/ 24 females, mean age 37.8 ± 7.1), non-alcoholic steatohepatitis (NASH) (23 males/ 27 females, mean age: 46.8 ± 8.9) and fibrosis (24 males/ 26 females, mean age, 55.2 ± 9.3), conducted in government hospitals in Baghdad, Basra, and Mosul between August 2024 and November 2025. Patients had been diagnosed by specialist hepatologists at the participating hospitals. Simple steatosis was diagnosed based on two criteria: abdominal ultrasonography and a Fatty Liver Index (FLI) ≥60. NASH and fibrosis were diagnosed by histopathological examination of liver biopsy specimens according to the Non-Alcoholic Fatty Liver Disease Clinical Research Network (NAFLD CRN) scoring system. NASH was defined by a NAFLD Activity Score (NAS) ≥4, whereas fibrosis was classified according to the NAFLD CRN fibrosis staging system (F1–F4). Healthy controls had no clinical, biochemical, or imaging evidence of the disease.

2.2. Ethical approval

Ethical approval was obtained from the Ethics Committee of the Institute of Genetic Engineering and Biotechnology for Postgraduate Studies, University of Baghdad, Iraq (Approval No. IGB-ETH/2026/I-004). All participants were informed about the study and agreed to take part by providing written informed consent prior to sample collection according to the declaration of Helsinki. Their information was kept confidential, and the data were recorded without personal identifiable information.

2.3. Biochemical analysis

Venous blood samples were collected from all participants after an overnight fast and serum was separated for biochemical measurements, including standard liver function tests Alanine Aminotransferase (ALT) and Aspartate Aminotransferase (AST), and a comprehensive lipid profile (Triglycerides, Total Cholesterol, HDL—C, and LDL-C), all were measured using the i-chroma automated clinical chemistry analyzer (Boditech Med Inc., South Korea). Platelet counts were extracted from the clinical data to calculate the FIB-4 index, which serves as a non-invasive marker for liver fibrosis using the following formula:

$$\mathrm{FIB}-4=\left(\mathrm{Age}\left[\text{years}\right]\mathrm{x}\mathrm{AST}\left[\mathrm{U}/\mathrm{L}\right]\right)/\left(\text{Platelet Count}\left[10^9/\mathrm{L}\right]\mathrm{x}\text{square root of}\mathrm{ALT}\left[\mathrm{U}/\mathrm{L}\right]\right)$$ (1)

2.4. miRNA extraction

Total RNA, including miRNA, was extracted from serum using the miRNeasy Serum/Plasma Kit (Qiagen, Hilden, Germany) according to the manufacturer's instructions.²² 200 μL of serum was mixed with 1000 μL of QIAzol and incubated for 5 min at room temperature (to ensure the complete dissociation of nucleoprotein complexes), then, 200 μL of chloroform was added, and the mixture was centrifuged at 12,000 xg for 15 min at 4 °C (to induce phase separation). The aqueous phase was recovered and mixed with 1.5 volumes of 100% ethanol before being loaded onto an RNeasy MinElute spin column, followed by sequential washing with RWT and RPE buffers, the RNA was eluted in 14 μL of RNase-free water. RNA concentration and purity were determined using the Qubit 4.0 Fluorometer with the Qubit RNA HS Assay Kit (Thermo Fisher Scientific, USA).²³ The extracted RNA was stored at −80 °C until downstream analysis.

2.5. cDNA synthesis

cDNA was synthesized from all miRNAs (miR-34a, miR-223 and miR-16) using the miRNA 1st Strand cDNA Synthesis Kit (by stem-loop) (Vazyme, Nanjing, China). Based on the manufacturer's instructions, a total volume of 20 μL was prepared containing: 5 μL RNA template, 4 μL of 5× Reaction Buffer, 2 μL dNTP Mix, 1 μL RT enzyme, 1 μL RNase inhibitor, 1 μL of the specific stem-loop RT primer (Table 1) and the final volume was adjusted to 20 μL using nuclease-free water.²⁴ The thermal profile was optimized as follows: 16 °C for 30 min for annealing and extension, 42 °C for 60 min for reverse transcription, and finally, 85 °C for 5 min to inactivate the reverse transcriptase enzyme and the obtained cDNA was stored at −20 °C for subsequent quantitative PCR analysis. A no-reverse transcriptase (No-RT) control was included to exclude genomic DNA contamination.

Table 1.

Primer sequences and characteristics for Stem-Loop RT-qPCR analysis of targeted miRNAs.

Target miRNA	Primer Type	Sequence (5′ to 3′)	Tm (°C)
miR-34a	Stem-loop RT	GTCGTATCCAGTGCAGGGTCCGAGGTATTCGCACTGGATACGACACAACC	ــــــــــ*
	Forward	TGGCAGTGTTCTTAGCTGGT	60.3 °C
miR-223	Stem-loop RT	GTCGTATCCAGTGCAGGGTCCGAGGTATTCGCACTGGATACGACGGGTAT	ــــــــــ*
	Forward	TGTCAGTTTGTCAAATACCCC	59.6 °C
miR-16	Stem-loop RT	GTCGTATCCAGTGCAGGGTCCGAGGTATTCGCACTGGATACGACATAACC	ــــــــــ*
	Forward	TAGCAGCACGTAAATAATTGGCG	60.1 °C
Universal	Reverse	GTGCAGGGTCCGAGGT	59.0°C

Footnotes: * Tm is not applicable for stem-loop primers due to their hairpin structure and short 3′ target-specific binding sequence.

2.6. Quantitative real-time PCR (RT-qPCR)

Expression levels of miR-34a, miR-223, and miR-16 (used as the internal reference gene for normalization) were quantified by RT-qPCR using the QuantiNova SYBR Green PCR Kit (Qiagen, Hilden, Germany) according to the manufacturer's instructions.²⁵ The reaction was performed in a 20 μL volume containing 10 μL of 2× SYBR Green Master Mix, 1 μL of each primer (general forward and reverse primer) at 10 μM, 2 μL of complementary cDNA template, and 6 μL of nuclease-free water. The thermal conditions were 95 °C for 2 min (initial dissociation), followed by 40 cycles of 95 °C for 15 s (dissociation) and 60 °C for 60 s (annealing/elongation). Finally, the melting curve was analyzed by gradually increasing the temperature from 60 °C to 95 °C to verify the specificity of the primers (Supplementary Fig. S1). qPCR efficiency was evaluated using five-point serial dilution to construct standard curves (Supplementary Tables S1-S3 and Fig. S2).

A negative control sample (without template) was included in each cycle to detect contamination or nonspecific amplification.

2.7. Data normalization and relative quantification

Relative expression levels of miR-34a and miR-223 were determined using the 2^−ΔΔCt method. The cycle threshold (Ct) values for each target miRNA were first normalized to the internal control (miR-16) to calculate the ΔCt as follows:

$$\mathrm{\Delta Ct}=\mathrm{Ct}\text{target}-\mathrm{Ct}\mathrm{miR}-16$$ (2)

The 2^−ΔΔCt was then calculated by subtracting the average ΔCt of the healthy control group from the ΔCt of each individual sample. Final results were expressed as the fold change in gene expression relative to the control group.

2.8. Primer sequences

The oligonucleotide sequences used for the reverse transcription (Stem-loop primers) which were designed based on the method described by Chen et al.,²⁶ and the subsequent quantitative PCR (Forward and Universal Reverse primers) were designed for the current study, all primers were provided by Macrogen (Seoul, South Korea) (Table 1). The Melting Temperature (Tm) values are provided for the qPCR primers to support the choice of the 60 °C annealing temperature.

2.9. Statistical analysis

Statistical analyses were performed using SPSS version 28 (IBM Corporation, Armonk, NY, USA). Quantitative real-time PCR data were processed using the comparative Ct (2^−ΔΔCt) method, with fold change calculated as 2^−ΔΔCt normalized to healthy controls. Normality was assessed using the Shapiro-Wilk test. Normally distributed clinical variables were presented as mean ± SD, while categorical variables were expressed as frequencies (%). The FIB-4 index was calculated as: (Age × AST) / (Platelet count × √ALT). Group comparisons were performed using one-way ANOVA with Tukey's HSD post-hoc test for clinical variables and the Kruskal-Wallis test with Dunn's post-hoc test for miRNA levels. No additional correction beyond the post-hoc test was applied. The stability of miR-16 was evaluated before normalization by comparing its mean Ct values across all study groups. Correlations were assessed using Spearman's rank correlation coefficient. Diagnostic performance was evaluated by ROC curve analysis, with AUC, sensitivity, specificity, and Youden index reported. Optimal cut-off values were selected using the maximum Youden's Index and were data-driven not pre-specified. Bootstrap internal validation (1000 resamples) was conducted to assess both the internal stability and overfitting of the combined diagnostic models, and to obtain bootstrap-corrected AUC with 95% confidence intervals. Univariate and multivariate logistic regression identified independent variables associated with advanced disease (NASH and Fibrosis), with results expressed as odds ratios and 95% confidence intervals. Multinomial logistic regression assessed associations with each disease stage using healthy controls as the reference. Quartile analysis evaluated dose-response relationships. A two-tailed p-value < 0.05 was considered statistically significant.27, 28

3. Results

3.1. Demographic and clinical data

To confirm the progressive metabolic and liver injury across the pathological stages of NAFLD, the demographic and clinical data were analyzed using one-way ANOVA for continuous variables presented as mean ± SD and the Chi-square test for sex. As shown in (Table 2), age increased significantly with disease severity from 30.1 ± 4.2 years in the healthy group to 55.2 ± 9.3 years in the fibrosis group, BMI and liver enzymes (ALT and AST) were higher in all patient groups, with the highest values observed in the NASH group: BMI 33.1 ± 4.2 kg/m², ALT 94.5 ± 8.2 U/L, and AST 88.9 ± 7.1 U/L. In contrast, platelet counts decreased with disease progression from 289.4 ± 41.2 × 10⁹/L in healthy to 135.8 ± 28.4 × 10⁹/L in fibrosis, which led to the subsequent increase in FIB-4 index from 0.89 ± 0.21 in healthy to 4.56 ± 1.34 in fibrosis. Lipid profiles showed significant alterations: triglycerides and total cholesterol were higher in the NASH group, while HDL-C decreased with disease progression. LDL-C and LDL/HDL ratio had a similar pattern, in which they increased from healthy to NASH and decreased in fibrosis. All comparisons were statistically significant at (p < 0.001) except for gender (p = 0.921).

Table 2.

Baseline characteristics of study groups.

Variable	Healthy (n=50)	Steatosis (n=50)	NASH (n=50)	Fibrosis (n=50)	p-value
Age (years)	30.1 ± 4.2	37.8 ± 7.1	46.8 ± 8.9	55.2 ± 9.3	<0.001
Sex (M/F)	26/24	26/24	23/27	24/26	0.921
BMI (kg/m2)	23.1 ± 2.3	29.2 ± 3.1	33.1 ± 4.2	30.2 ± 4.8	<0.001
ALT (U/L)	24.2 ± 5.1	44.8 ± 5.9	94.5 ± 8.2	69.8 ± 7.3	<0.001
AST (U/L)	21.3 ± 5.3	40.1 ± 5.1	88.9 ± 7.1	74.2 ± 6.2	<0.001
Platelets (109/L)	289.4 ± 41.2	262.8 ± 38.9	239.7 ± 35.6	135.8 ± 28.4	<0.001
FIB-4 Index	0.89 ± 0.21	1.42 ± 0.38	2.89 ± 0.92	4.56 ± 1.34	<0.001
Triglycerides (mg/dL)	112.4 ± 16.8	182.5 ± 21.3	221.7 ± 24.1	181.2 ± 23.7	<0.001
Total Cholesterol (mg/dL)	186.26 ± 12.67	212.55 ± 14.45	231.82 ± 17.01	191.15 ± 16.30	<0.001
LDL-C (mg/dL)	101.02 ± 11.95	144.89 ± 10.75	161.07 ± 11.64	134.38 ± 15.44	<0.001
HDL-C (mg/dL)	60.2 ± 5.8	48.9 ± 5.1	42.3 ± 4.8	37.8 ± 4.5	<0.001
LDL/HDL Ratio	1.68 ± 0.22	2.98 ± 0.41	3.78 ± 0.52	3.61 ± 0.63	<0.001

3.2. Expression of miR-34a and miR-223

The expression levels of miR-34a and miR-223 across all study groups were calculated using the 2^-ΔΔCt method after normalization with miR-16 as an internal control. The mean Ct values of miR-16 were stable across all groups with no significant differences (p = 0.350) and a low global CV (2.36%), reflecting its suitability for normalization. Statistically significant differences in the expression of both miRNAs were observed at (p < 0.001, using Kruskal-Wallis test due to skewed distribution) as shown in (Table 3). Post-hoc analysis showed significant differences indicated by superscript letters (a, b, c and d). The expression of miR-34a was lowest in healthy controls (1.42 ± 1.89) and increased gradually in steatosis (3.85 ± 3.21) reaching the highest level in NASH (41.28 ± 32.45), then followed by a relative decline in fibrosis (12.84 ± 18.76), while remaining elevated in comparison to healthy controls. For miR-223, the expression was low in healthy and steatosis (1.51 ± 1.78, 1.68 ± 1.54) respectively, then reached the highest level in NASH (42.17 ± 33.89) and it decreased in fibrosis (26.38 ± 28.92) but still higher than in healthy. The ΔCt values confirmed these patterns in which lower ΔCt values indicate higher expression. A boxplot (Fig. 1) illustrates fold-change expression value in all groups.

Table 3.

miRNA expression Levels across Disease groups.

miRNA	Healthy (n=50)	Steatosis (n=50)	NASH (n=50)	Fibrosis (n=50)	p-value*
miR-34a (Fold Change)	1.42 ± 1.89a	3.85 ± 3.21b	41.28 ± 32.45c	12.84 ± 18.76d	< 0.001
miR-223 (Fold Change)	1.51 ± 1.78a	1.68 ± 1.54a	42.17 ± 33.89b	26.38 ± 28.92c	< 0.001
ΔCt miR-34a	5.39 ± 1.42a	3.78 ± 0.89b	0.42 ± 1.21c	3.52 ± 1.34b	< 0.001
ΔCt miR-223	5.21 ± 1.56a	5.08 ± 1.34a	0.38 ± 1.42b	1.28 ± 1.51c	< 0.001
Ct miR-16	22.45 ± 0.44	22.48 ± 0.54	22.63 ± 0.52	22.47 ± 0.60	0.350

Footnotes: Data presented as Mean ± SD; data analyzed using Kruskal-Wallis test followed by post-hoc Dunn's test, Superscript letters (a,b,c and d) indicate significant pairwise differences (p<0.05)*. The global coefficient of variation (CV) for miR-16 Ct values was 2.36%.

Fig. 1.

Fold change values of miR-34a and miR-223 expression across all study groups: Healthy, n=50; Steatosis, n=50; NASH, n=50; Fibrosis, n=50 (The median values are represented by horizontal lines inside the squares. The boxes indicate ranges between quartiles, the whiskers represent data dispersion, and the dots indicate outliers).

3.3. Comparisons of miRNAs expression between disease groups

To determine which specific groups differ significantly in miRNA expression levels, Dunn's test with Bonferroni correction was used to perform the pairwise comparisons as shown in (Table 4). For miR-34a, significant differences were observed in all comparisons (p < 0.05), with the largest differences between healthy versus NASH (Z = 8.91, adj. p < 0.001) and steatosis versus NASH (Z = 7.89, adj. p < 0.001), while the lowest difference was between steatosis and fibrosis (Z = 3.45, adj. p = 0.002). In contrast, miR-223 showed no significant difference between healthy and steatosis (Z = 1.24, adj. p = 0.645), whereas the largest differences were observed between healthy versus NASH (Z = 8.76, adj. p < 0.001) and steatosis versus NASH (Z = 8.45, adj. p < 0.001).

Table 4.

Pairwise comparisons between groups (Post-hoc Analysis).

Comparison	miR-34a		miR-223
	Z-statistic	Adj. p-value	Z-statistic	Adj. p-value
Healthy vs. Steatosis	4.82	<0.001	1.24	0.645
Healthy vs. NASH	8.91	<0.001	8.76	<0.001
Healthy vs. Fibrosis	6.34	<0.001	8.12	<0.001
Steatosis vs. NASH	7.89	<0.001	8.45	<0.001
Steatosis vs. Fibrosis	3.45	0.002	7.89	<0.001
NASH vs. Fibrosis	4.56	<0.001	2.89	0.012

3.4. Correlations between miRNA expression and clinical parameters

Spearman's rank correlation was performed to assess the relationship between miRNAs and clinical parameters. Both miRNAs were significantly correlated with all clinical markers (p < 0.01 or lower), although the strength of these correlations varied (Table 5). Strong positive correlations (r > 0.6) were observed between both miRNAs and ALT, AST, total cholesterol, LDL-C, and LDL/HDL ratio, indicating that higher miRNA levels are strongly linked to increased liver enzymes and worse lipid profiles. Strong negative correlations (r < −0.6) were found between miR-34a and platelets and HDL—C, meaning higher miRNA levels are associated with lower platelet counts and lower good cholesterol. Moderate positive correlations (r = 0.5–0.6) were observed between both miRNAs and age, BMI, and triglycerides, suggesting these factors increase with miRNA expression. Moderate negative correlations (r = −0.5 to −0.6) were observed between miR-223 with platelets and HDL—C, showing similar but slightly weaker associations compared to miR-34a.

Table 5.

Correlation between miRNAs and clinical parameters.

Parameter	miR-34a (Fold Change)	miR-223 (Fold Change)
Age	0.52**	0.48**
BMI	0.58**	0.53**
ALT	0.71	0.68
AST	0.69	0.65
Platelets	−0.62	−0.59**
Triglycerides	0.55**	0.51**
Total Cholesterol	0.64	0.60
LDL-C	0.65	0.61
HDL-C	−0.61	−0.57**
LDL/HDL Ratio	0.67	0.63

Footnotes: ** p < 0.01. Bold indicates strong correlation (|r| > 0.6).

3.5. miRNAs as diagnostic markers to distinguish NAFLD stages

Receiver operating characteristic (ROC) analysis was used to assess the diagnostic performance of miRNAs (individually or combined) in distinguishing between different stages of liver disease. Both miRNAs performed excellently (AUC = 0.934 for miR-34a, AUC = 0.956 for miR-223) to distinguish steatosis from NASH, with the combined model achieving near the perfect discrimination (AUC = 0.978; 96% sensitivity, 94% specificity). The combined model also performed best for detecting any disease versus healthy (AUC = 0.921) and fibrosis versus others (AUC = 0.911) as presented in (Table 6) and (Fig. 2). miR-223 excelled at fibrosis detection (AUC = 0.882), while miR-34a performed better for NASH vs. fibrosis (AUC = 0.812). Additionally, bootstrap internal validation was conducted using 1000 resamples to evaluate the internal stability of the diagnostic models. The analysis demonstrated minimal optimism (ranging from 0.013 to 0.021), and bootstrap-corrected AUC values remained high across all combined models, indicating good internal stability with a low risk of overfitting.

Table 6.

Diagnostic performance of miRNAs for liver Disease detection (ROC analysis).

Comparison	miRNA	AUC (95% CI)	Cut-off	Sensitivity	Specificity	Youden's Index	Optimism	Corrected AUC (95% CI)
Healthy vs. Any Disease	miR-34a	0.892 (0.84–0.94)	>2.1 FC	86%	84%	0.7
	miR-223	0.754 (0.69–0.82)	>2.3 FC	72%	70%	0.42
	Combined	0.921 (0.88–0.96)	–	89%	88%	0.77	0.013	0.908 (0.83–0.98)
Steatosis vs. NASH	miR-34a	0.934 (0.89–0.97)	>8.5 FC	90%	88%	0.78
	miR-223	0.956 (0.92–0.98)	>5.2 FC	94%	92%	0.86
	Combined	0.978 (0.95–0.99)	–	96%	94%	0.9	0.021	0.956 (0.89–0.99)
NASH vs. Fibrosis	miR-34a	0.812 (0.73–0.88)	<18.5 FC	78%	76%	0.54
	miR-223	0.694 (0.60–0.78)	>15.0 FC	68%	64%	0.32
	Combined	0.846 (0.77–0.91)	–	82%	80%	0.62	0.018	0.827 (0.74–0.90)
Fibrosis vs. Others	miR-34a	0.768 (0.70–0.83)	<5.5 FC	74%	72%	0.46
	miR-223	0.882 (0.83–0.93)	>8.2 FC	86%	84%	0.7
	Combined	0.911 (0.87–0.95)	–	88%	86%	0.74	0.013	0.898 (0.81–0.96)

Footnotes: AUC, Area Under the Curve; CI, Confidence Interval; FC, Fold Change.

Combined = Logistic regression model including both miRNAs.

Youden's Index = Sensitivity + Specificity – 1.

Bootstrap-corrected AUCs and the corresponding 95% CIs were obtained using 1000 bootstrap resamples.

Fig. 2.

ROC curves for miR-34a, miR-223, and combined model in discriminating Steatosis from NASH (n=50 per group).

3.6. Comparison with traditional clinical markers

ROC analysis was also used to assess the diagnostic performance of miRNAs against traditional markers in distinguishing steatosis from NASH and for fibrosis detection. P-values were calculated using DeLong's test by comparing each marker to miR-34a that was selected as a reference. For steatosis vs. NASH, both miRNAs (AUC > 0.93) significantly outperformed all traditional markers (ALT, AST, platelets, FIB-4; p < 0.05 for all) as shown in (Fig. 3a). For Fibrosis detection (Fig. 3b), miR-223 (AUC = 0.882) performed best among miRNAs, significantly exceeding miR-34a (p = 0.012). The FIB-4 index showed comparable accuracy (AUC = 0.901), while ALT and AST performed poorly (AUC < 0.73). These findings indicate that the diagnostic accuracy of miRNAs was superior or comparable to traditional markers.

Fig. 3.

ROC curves: A- distinguishing steatosis from NASH (n=50 per group), B- Fibrosis detection (Fibrosis, n=50; Others, n=150).

3.7. Factors independently associated with advanced disease

To identify variables associated with advanced liver disease (NASH + fibrosis vs. healthy + steatosis), multivariate logistic regression analysis was performed to estimate the unadjusted and adjusted odds ratios as shown in (Table 8). All variables including both miRNAs, age, BMI, ALT, platelets and LDL/HDL ratio were significantly associated with advanced stages of the disease. LDL/HDL ratio and BMI showed the strongest associations in unadjusted and adjusted models. Both miRNAs also remained significant with miR-34a showing (Unadjusted OR =1.45, p<0.001) and (Adjusted OR =1.32, p<0.001), whereas miR-223 showing (Unadjusted OR =1.38, p<0.001) and (Adjusted OR =1.28, p<0.001).

Table 8.

Multivariate logistic regression analysis for advanced disease.

Variable	Unadjusted OR (95% CI)	p-value	Adjusted OR (95% CI)	p-value
miR-34a (per 1-unit increase)	1.45 (1.28–1.64)	<0.001	1.32 (1.18–1.48)	<0.001
miR-223 (per 1-unit increase)	1.38 (1.22–1.56)	<0.001	1.28 (1.15–1.42)	<0.001
Age (per 10-year increase)	1.52 (1.18–1.96)	0.001	1.45 (1.12–1.88)	0.005
BMI (per 5-unit increase)	1.78 (1.32–2.41)	<0.001	1.62 (1.21–2.16)	0.001
ALT (per 10 U/L increase)	1.24 (1.08–1.42)	0.002	1.18 (1.04–1.34)	0.012
Platelets (per 50-unit decrease)	1.68 (1.28–2.21)	<0.001	1.52 (1.18–1.96)	0.002
LDL/HDL Ratio (per 1-unit increase)	1.92 (1.42–2.59)	<0.001	1.71 (1.24–2.35)	<0.001

3.8. Multinomial logistic regression analysis by disease group

A multinomial logistic regression analysis including both miRNAs as well as age and BMI, selected based on their clinical importance, was performed to identify variables associated with each disease group relative to the healthy controls. For steatosis, higher miR-34a (RRR = 1.28), increasing age (RRR = 1.12), and higher BMI (RRR = 1.24) were significant associated factors (p < 0.05), while miR-223 was not associated with this early stage (p = 0.345). For NASH, all factors were significant, with both miRNAs showing strong effects in which miR-34a increased risk more than twofold (RRR = 2.34) and miR-223 showed a similar pattern (RRR = 2.18), along with age and BMI (p < 0.001 for all). For fibrosis, all variables remained significant, but here miR-223 (RRR = 1.89) showed a stronger association than miR-34a (RRR = 1.45), while age and BMI also contributed significantly. Overall, these findings demonstrate that miR-34a is linked to all disease stages, whereas miR-223 becomes significantly associated only at the more advanced stages of NASH and fibrosis (Table 9).

Table 9.

Multinomial logistic regression analysis for disease stages relative to Healthy Controls.

Variables	Steatosis vs. Healthy		NASH vs. Healthy		Fibrosis vs. Healthy
	RRR (95% CI)	p-value	RRR (95% CI)	p-value	RRR (95% CI)	p-value
miR-34a	1.28 (1.08–1.52)	0.004	2.34 (1.89–2.89)	<0.001	1.45 (1.18–1.78)	<0.001
miR-223	1.08 (0.92–1.27)	0.345	2.18 (1.78–2.67)	<0.001	1.89 (1.52–2.35)	<0.001
Age	1.12 (1.02–1.23)	0.018	1.24 (1.12–1.37)	<0.001	1.34 (1.18–1.52)	<0.001
BMI	1.24 (1.08–1.42)	0.002	1.38 (1.18–1.61)	<0.001	1.28 (1.08–1.52)	0.004

Footnotes: RRR, Relative Risk Ratio; RRR > 1 indicates increased risk of being in that disease category.

3.9. Quartile-based risk classification for advanced liver disease

A quartile-based risk classification for advanced liver disease (NASH and Fibrosis), (Table 10) with corresponding trends illustrated in (Fig. 4). A clear dose-response relationship was observed for both miRNAs, as the proportion of patients with advanced disease increased progressively across quartiles. For miR-34a, advanced disease rates rose from 12% in Q1 to 88% in Q4, with patients in the highest quartile showing a 53.7-fold increased risk (p for trend < 0.001). Similarly, for miR-223, rates increased from 16% in Q1 to 92% in Q4, corresponding to a 60.3-fold increased risk in Q4 (p for trend < 0.001). The combined risk analysis demonstrated even stronger discrimination: patients with both miRNAs in low quartiles (Q1–Q2) had only 8% advanced disease, whereas those with both in high quartiles (Q3–Q4) had 91% advanced disease, yielding an OR of 116.3 (95% CI: 34.5–392.1). These findings confirm that the higher expression of miRNAs (Q3-Q4) is associated with a higher risk of advanced disease stages compared to their lower expression (Q1-Q2).

Table 10.

miRNA quartiles and risk of advanced disease.

miRNA Quartile	Range (FC)	n	% Advanced Disease*	OR (95% CI)	p for trend
miR-34a
Q1	<1.2	50	12%	1.00 (ref)	<0.001
Q2	1.2–3.8	50	28%	2.85 (1.12–7.25)
Q3	3.8–18.5	50	64%	13.09 (5.12–33.45)
Q4	>18.5	50	88%	53.67 (18.34–157.2)
miR-223
Q1	<1.1	50	16%	1.00 (ref)	<0.001
Q2	1.1–2.8	50	22%	1.48 (0.58–3.78)
Q3	2.8–18.5	50	58%	7.26 (2.98–17.68)
Q4	>18.5	50	92%	60.33 (19.45–187.2)
Combined Risk
Low	Both in Q1-Q2	75	8%	1.00 (ref)	<0.001
High	Both in Q3-Q4	55	91%	116.3 (34.5–392.1)

Footnotes: Combined Risk: Low = both miRNAs in Q1–Q2; High = both miRNAs in Q3–Q4*.

Fig. 4.

Proportion of patients with advanced disease according to miR-34a and miR-223 expression quartiles.

3.10. Diagnostic modeling

Four diagnostic models were constructed to assess the contribution of each miRNA (individually and in combination) in identifying advanced liver disease as shown in (Table 11, Fig. 5). Model A included age, BMI, and ALT, as these are routinely measured in patients with suspected or confirmed NAFLD. This baseline model demonstrated good discrimination (AUC = 0.812; AIC = 212.4) and served as the reference for comparing the added value of miRNAs. The addition of miR-34a alone (Model B) significantly improved model performance (AUC = 0.878; AIC = 189.2; NRI = 0.184; p = 0.002), as did the addition of miR-223 alone (Model C: AUC = 0.869; AIC = 192.6; NRI = 0.156; p = 0.008). Model D, which included both miRNAs with clinical variables, showed the best performance, yielding an AUC of 0.912, lowest AIC (168.4), and the highest net reclassification improvement (NRI = 0.267; p < 0.001) suggesting that combining both miRNAs with routine clinical markers led to better risk classification than using clinical markers alone.

Table 11.

Diagnostic performance of the four different models.

Model	Variables Included	AIC	AUC	NRI	p-value
Model A	Age + BMI + ALT	212.4	0.812	Reference	–
Model B	+ miR-34a	189.2	0.878	0.184	0.002
Model C	+ miR-223	192.6	0.869	0.156	0.008
Model D	+ both miRNAs	168.4	0.912	0.267	<0.001

Footnotes: AIC, Akaike Information Criterion (lower = better fit); NRI, Net Reclassification Improvement vs. Model A.

Fig. 5.

AUC values of the four diagnostic models: Model A (traditional clinical variables), Model B (miR-34a), Model C (miR-223), and Model D (both miRNAs).

4. Discussion

As presented in Table 2, both age and BMI increased across the study groups from the healthy group to the fibrosis group (p < 0.001), which aligns with the established understanding that advancing age and obesity represent primary contributors to NAFLD progression. These results match worldwide patterns, where NAFLD becomes more common with age and is closely tied to features of metabolic syndrome.²⁹ Liver enzyme levels (ALT and AST) followed a unique pattern: they were highest in the NASH group and then dropped slightly in the fibrosis group. This pattern reflects known natural history of NAFLD, in which the fibrosis can be accompanied by burned out steatosis and subsequently leads to decrease in the release of liver enzymes.23, 30 Platelet counts declined steadily across the groups, falling from 289.4 to 135.8 × 10⁹/L (p < 0.001), a finding that underscores the development of portal hypertension and progressive hepatic fibrosis. This decline serves as an important marker confirming the clinical validity of the study cohorts, given that thrombocytopenia in this context reflects both diminished thrombopoietin synthesis and increased splenic platelet sequestration which changes characteristic of advanced liver disease and cirrhosis.³⁰

The analysis of gene expression revealed that both miRNAs were significantly upregulated in patients compared to healthy controls (p < 0.001). For miR-34a, its expression level increased across the study groups, reaching the highest level in NASH (41.28-fold change), followed by a relative decrease in the fibrosis group but remained elevated compared to healthy controls (Table 3). This pattern is supported by existing research, with systematic review have identified miR-34a as one of the most frequently studied miRNAs in MASLD, appearing in 14.44% of clinical studies.²³ Ragab et al. also found miR-34a was more active in NAFLD patients with early fibrosis compared to controls.³¹ The relative decrease seen in fibrosis of this study is consistent with their additional finding that miR-34a expression is lower in patients with advanced fibrosis. This suggests that the regulatory role of this miRNA varies according to the stage of the disease. Whereas the expression level of miR-223 remained stable in healthy and steatosis groups, followed by an increase in the NASH group (42.17-fold change) (Table 4). This upregulation aligns with a bibliometric review of Wang et al. who found that miR-223 is specifically associated with obesity-related MASH.³² This significant result is supported by mechanistic studies that include extracellular vesicles derived from neutrophils, which are rich in miR-223, are absorbed by hepatocytes, thereby contributing to the attenuation of NASH progression. This internalization process appears to be facilitated by LDLR expression on hepatocytes and APOE on neutrophil-derived vesicles, with free fatty acids that promote vesicular uptake.³³

The distinct expression pattern of each miRNA reflects its pathophysiological role across different stages of NAFLD. The increase in miR-34a expression in steatosis stage is consistent with its participation in lipid metabolism and insulin resistance, which modulates the production of very low density lipoprotein (VLDL) and synthesizing fatty acids from scratch.³⁴ This effect is attributed to the ability of miR-34a to inhibit PPARα and SIRT1, which are key regulators of fatty acid β-oxidation and hepatocyte survival.¹⁴ The suppression of these regulators leads to lipid accumulation and hepatocyte apoptosis, providing a biological explanation for the progressive increase in the expression of miR-34a from steatosis to its peak in NASH.¹¹ While the elevation of miR-223 expression reflects its regulatory role in immune cells and supports the hypothesis that miR-223 is upregulated in response to the active hepatic inflammation which is a characteristic of NASH.³¹ miR-223 contributes to the communication between immune cells and hepatocytes through neutrophil-derived extracellular vesicles, thus modulating inflammatory signaling during active steatohepatitis.³³ This mechanism provides a biological explanation for the higher level of miR-223 expression in the NASH group.³⁵ A mechanistic study on mouse models has given an additional evidence that the dysregulation of miR-223 increases the hepatic inflammation and promotes fibrosis.¹⁶ Furthermore, miR-223 has been implicated in the progression from steatosis to NASH by downregulating hepatic Cxcl10 and Taz expression, indicating its potential as a therapeutic target in NASH.³⁵ The relative decline in the expression of both miRNAs from their peak levels in NASH to the fibrosis stage may be explained by the progressive replacement of functional hepatocytes with extracellular matrix, reducing the number of viable hepatocytes capable of producing these miRNAs. This change is consistent with the reduced hepatocellular activity that characterizes advanced fibrosis.³⁶

The positive correlation observed between miRNA expression and ALT and AST levels (miR-34a: r = 0.71 and 0.69; miR-223: r = 0.68 and 0.65), as presented in Table 5, provided biological significance by illustrating the close association of these miRNAs with hepatocellular damage. miR-34a directly targets PPARα and SIRT1, which are crucial regulators of lipid metabolism and hepatocyte survival; its upregulation subsequently promotes hepatocyte apoptosis and exacerbates liver injury in NAFLD.³⁷ These correlations are congruent with prior findings indicating a positive correlation between miR-34a and ALT (r = 0.427, p = 0.002) in MASLD patients.³⁸ Furthermore, the same study reported a correlation between miR-223 and ALT (r = 0.429, p = 0.002), as well as with fibrosis markers such as APRI (r = 0.386, p = 0.035) and FIB-4 (r = 0.408, p = 0.025), which confirms the correlations we observed.

Both miRNAs also correlated strongly with lipid profiles (positive with LDL/HDL ratio, negative with HDL—C), confirming their integration into the metabolic dysfunction of NAFLD. A systematic review focusing on cholesterol metabolism found that miR-34a levels correlated with VLDL-C and triglycerides, and that miR-34a targets include key cholesterol regulators such as ABCA1 and ABCG1.³⁹ The strong negative correlation with platelets (miR-34a: r = −0.62; miR-223: r = −0.59) further validates their utility as markers of hepatic fibrotic. Ragab et al. additionally reported that miRNA-34a was positively correlated with metabolic disorders associated with NAFLD such as hypertension and fasting blood sugar,³¹ supporting the broader metabolic significance of this miRNA.

The diagnostic performance of the combined model (miR-34a and miR-223) is consistent with the “multiple hits” hypothesis in liver diseases (liver injury results from several processes rather than a single one). The best performance of the combined model was in distinguishing steatosis from NASH (AUC = 0.978) as shown in Table 6. Individually, miR-34a had the highest performance in distinguishing between healthy and those at any stage of the disease (AUC = 0.892) suggesting the role of this miRNA as a primary factor responding to hepatocyte stress. This finding is consistent with several studies who noted the elevation of miR-34a is usually with the first molecular alterations in NAFLD.40, 41 On the other hand, the strong accuracy of miRNA 223 was in detecting fibrosis from other stages (AUC = 0.882), while its significant contribution was observed in differentiating steatosis from NASH (AUC = 0.956). This result suggests that miR-223 is an inflammatory marker associated with steatohepatitis distinguishing it from early fat accumulation which is aligned with a study by Calventi et al.,¹⁶ who found the elevation of miR-223 in hepatitis. These findings overcome the certain limitations that accompany the use of single marker in diagnosis and the risk of biopsy. Additionally, the bootstrap finding supported the robustness of the combined diagnostic models, as the optimism-corrected AUC values remained close to the apparent AUC values, indicating good internal stability.

Although ALT and AST are widely used in routine practice, they have known limitations due to they are markers of cell death rather than detection of disease stages. The AUC values for distinguishing steatosis from NASH were 0.821 and 0.812 for ALT and AST respectively. This is because the ALT enzyme is present in many organs (including the liver), its elevation level might be a false positive that is not related to liver disease. As for ALT, although it is more specific to the liver, it still fluctuates due to non-pathological factors such as weight change or recent physical activity.⁴² Therefore, miR-34a and miR-223 outperformed these enzymes (p < 0.01) in differentiation steatosis from NASH (Table 7), suggesting that miRNA levels reflect the inflammatory and metabolic signals underlying NASH more closely than enzyme-based markers alone. For fibrosis detection, each miRNA had a different behavior. The expression of miR-34a had a relative decline in fibrosis compared to NASH, but remained elevated compared to healthy controls (AUC = 0.768), while the AUC of miR-223 (0.882) was near to the FIB-4 index (AUC = 0.901; p = 0.412), which is currently used to estimate the amount of scarring (fibrosis) in the liver.⁴³ This finding is clinically significant, as it positions miR-223 as a molecular marker for identifying advanced scarring suggesting it could help to overcome the clinical indeterminate ranges often associated with the FIB-4 index.

Table 7.

Comparison of diagnostic accuracy between miRNAs and traditional clinical markers.

Marker	Steatosis vs. NASH		Fibrosis Detection
	AUC (95% CI)	p-value*	AUC (95% CI)	p-value*
miR-34a	0.934 (0.89–0.97)	Reference	0.768 (0.70–0.83)	Reference
miR-223	0.956 (0.92–0.98)	0.234	0.882 (0.83–0.93)	0.012
ALT	0.821 (0.76–0.88)	0.008	0.712 (0.64–0.78)	0.089
AST	0.812 (0.75–0.87)	0.006	0.724 (0.65–0.79)	0.124
Platelets	0.712 (0.64–0.78)	<0.001	0.845 (0.79–0.90)	0.178
FIB-4 Index	0.856 (0.80–0.91)	0.042	0.901 (0.86–0.94)	0.412

Several non-invasive tools are used for the assessment of liver fibrosis, including FibroScan, FIB-4, APRI, the NAFLD Fibrosis Score, and the ELF score.6, 9 Although each method has limitations related to diagnostic performance in specific clinical settings, availability, or cost,²¹ the diagnostic performance of miR-34a and miR-223 observed in the present study suggests that these biomarkers may complement existing non-invasive tools rather than replace them. Future studies are needed to compare miRNAs as diagnostic non-invasive tools with fibrosis assessment methods in the same patient population to determine their potential clinical utility and diagnostic performance.40, 41

The result of multivariate logistic regression analysis confirmed that both miRNAs are independent markers of advanced liver disease (NASH and fibrosis) in both the unadjusted and adjusted models (p < 0.001), as shown in Table 8. While the study groups exhibited significant variations in age and BMI, reflecting the clinical differences among the NAFLD stages, the continued significant associations even after adjustment suggest that the diagnostic value of both miRNAs was not only determined by these clinical variables. Yet, given the cross-sectional study design, the residual confounding cannot be completely ruled out. Although the strongest associations were identified between LDL/HDL ratio, BMI, and advanced liver disease (aOR = 1.71 and 1.62 respectively), the consistency of the miRNAs was prominent. This is because the standard metabolic indicators are influenced by systemic factors; therefore, the consistent significance of miRNAs suggests that they reflect molecular pathophysiology occurring within liver tissues, such as hepatocyte apoptosis and chronic inflammation.⁴⁴ The results also confirmed that combining molecular biomarkers (miRNAs) with clinical data, including ALT (aOR = 1.18) and platelet count (aOR = 1.52), could significantly enhance the accuracy of non-invasive screening. These findings indicate that miR-34a and miR-223 may act as independent diagnostic biomarkers for detecting NAFLD and distinguishing its stages in addition to their ability to detect the disease, which is supported by recent molecular evaluation of liver injury.¹⁹The diagnostic power of each miRNA at specific stages of the disease was assessed using a multinomial logistic regression analysis through relative risk ratios (RRR) as shown in Table 9. The analysis reveals that miR-34a was significantly associate with all stages of disease, which reflects its role as an early indicator of hepatocyte stress. Its discriminative power peaked in the NASH stage (RRR = 2.34, p < 0.001), indicating every increase in miR-34a expression has a doubled risk compared to healthy. The increased risk in NASH in this study is aligned with the work of López-Pastor et al.,⁴⁴ who confirmed that miR-34a levels are closely related with the severity of inflammation and insulin resistance. The analysis confirmed the strong association of miRNA-223 with NASH (RRR = 2.18) and fibrosis (RRR = 1.89) supporting the concept of this miRNA is not triggered with the accumulation of fat but rather it is the specific marker for the inflammation. On the opposite with regard to age and BMI in which their RRRs (1.12–1.38) were lower than both miRNAs. These findings revealed that the molecular markers may provide additional value for detecting the stages of liver disease as discussed in the previous review (19).

The clear relationship between the gene expression and the severity of disease was achieved by the quartile analysis (Table 10). Higher expression quartiles were significantly associated with greater odds of having advanced disease, with statistically significant association observed for both miRNAs (p < 0.001). The odds ratio for patients with miR-34a in the fourth quartile (Q4) was more than 53 times higher than for the patients with miR-34a in the first quartile (Q1), while the odds ratio for the patients with miR-223 was close to 60 times higher. Notably, the combined high expression levels of both miRNAs were associated with an even higher odds ratio: OR = 116.3 (95% CI: 34.5–392.1), indicating a stronger association than miRNA alone, as reported in previous studies.41, 44 Although each miRNA has its own biological role, it seems that their joint effect reflects their involvement in various complementary disease-associated processes like inflammatory processes, communication between cells, and signaling pathways linked to fibrosis.⁴⁵ These findings suggest that both miRNAs may serve as biologically relevant markers of disease severity in addition to their diagnostic value.⁴⁶ Clinically, patients with high expression levels of both markers may represent a subgroup that could benefit from closer clinical monitoring.⁴⁷The accuracy of diagnosis is improved by the integration of miRNA expression levels (individually and in combination) with traditional clinical variables, including age, BMI, and ALT levels (Table 11). The baseline clinical model achieved a statistically acceptable value (AUC = 0.812), but its ability to determine disease stage remained limited. miR-34a alone was better in diagnosing the disease than miR-223 in terms of accuracy (Model B: AUC = 0.878 vs Model C: AUC = 0.869) and significantly improved the risk classification (NRI = 0.184, p = 0.002), suggesting that miR-34a plays a role in all stages of the disease, as demonstrated in the regression analysis.⁴⁸ Furthermore, the combined model was the most accurate (Model D: AUC = 0.912, NRI = 0.267, p < 0.001) and performed best among all models (AIC = 168.4). These findings show that each miRNA reflects different aspects of the disease, while using both together gives a more accurate description of the patient's condition which was better than using one alone or other clinical factors.²²

Finally, in addition to the diagnostic ability of these miRNAs, a study found that extracellular vesicles carrying miR-223 could reduce markers of fat buildup, inflammation, and scarring.⁴⁷ Another study found that adipose-derived mesenchymal stem cell-secreted extracellular vesicles carrying miR-223-3p are suppressing lipid accumulation and liver fibrosis through E2F1 targeting.⁴⁹ These results enabled researchers to adopt miR-223 as a therapeutic target in liver disease.⁵⁰

5. Conclusion

The circulating miR-34a and miR-223 have emerged as promising tools for the diagnosis and differentiation of NAFLD stages. Individually, the expression pattern of miR-34a rises at early stage (Steatosis) and persists through all disease stages, while miR-223 increases at the NASH and fibrosis stages. Together, these miRNAs significantly outperform routine clinical markers (ALT, AST, platelets, and FIB-4) in distinguishing the disease stages. The diagnostic accuracy of their combined model was excellent in which patients in the highest quartiles (Q3 and Q4) showing a 116-fold increased risk of advanced disease compared to lowest quartiles (Q1 and Q2). This dual-miRNA panel offers a practical, blood-based approach for identifying patients with advanced liver damage.

Limitations

This study has several limitations. First, the cross-sectional design gives useful information about miRNA expression, but it does not show the changes over time. Second, although bootstrap internal validation was performed, the findings were based on a single population and lacked external validation in an independent cohort. Therefore, future studies should validate these findings in larger, multi-center cohorts with longitudinal follow-up.

CRediT authorship contribution statement

Ammar Kadhim Alzubaidi: Writing – review & editing, Writing – original draft, Project administration, Methodology, Formal analysis. Farah Abdulwahab Al-Saadi: Writing – review & editing, Validation, Investigation, Data curation. Yasir B. Fadhil: Writing – review & editing, Visualization, Supervision.

Ethical approval

Full ethical approval was obtained from the Ethics Committee of the Institute of Genetic Engineering and Biotechnology for Postgraduate Studies, University of Baghdad, Iraq (Approval No. IGB-ETH/I-004), as cross-referenced in the Materials and Methods Section 2.2 of the Main Manuscript.

Funding

This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Appendix A. Supplementary data

Supplementary material: The Supplementary Material contains qPCR validation data, including melting curve analysis (Figure S1), the serial dilution protocol (Table S1), average Ct values (Table S2), qPCR validation parameters (Table S3), and standard curves for amplification efficiency validation (Figure S2).

Data availability

All data supporting the study's findings are available from the corresponding author upon reasonable request.