Irisin levels in patients with prediabetes mellitus: A case-control study and meta-analysis

Abstract

BACKGROUND

Circulating irisin, a myokine, has been inconsistently associated with prediabetes mellitus (PreDM), and its relationship with glucagon is unexplored. Methodological variability, particularly between different enzyme-linked immunosorbent assay (ELISA) kits, is hypothesized to be a primary source of these discrepancies. Therefore, we hypothesized that the reported irisin-PreDM association is critically confounded by assay methodology and inadequate adjustment for key metabolic factors.

AIM

To investigate circulating irisin in PreDM via a dual-study approach, evaluating methodological variability and metabolic confounders.

METHODS

A dual-study design was employed: (1) A systematic review and meta-analysis of observational studies; and (2) A single-center, exploratory case-control study of 35 participants (9 PreDM, 26 controls) from a tertiary hospital. Irisin was measured with two commercial ELISA kits (ELK and Elabscience). Key analyses included random-effects meta-analysis, multiple linear regression, Spearman correlation, and Bland-Altman analysis for inter-assay agreement.

RESULTS

The meta-analysis (8 studies, n = 788) found significantly lower irisin in PreDM (standardized mean difference = -0.524, 95% confidence interval: -0.897 to -0.152, P = 0.006) with high heterogeneity (I² = 86.7%). In the case-control study, unadjusted analysis showed higher irisin in PreDM with the Elabscience kit (median: 7386.3 pg/mL vs 6332.0 pg/mL, P = 0.045) but not the ELK kit. After adjusting for gender, body mass index, and fatty liver, the association was null for both kits. Bland-Altman analysis revealed poor inter-assay agreement (mean bias: 5837.25 pg/mL). Irisin correlated with fatty liver (ρ = 0.350, P = 0.039) and 0.5-hour glucose (ρ = 0.421, P = 0.012), but not with glucagon.

CONCLUSION

The irisin-PreDM relationship appears to be influenced by assay methodology and metabolic factors. Standardizing irisin measurement and adjusting for key confounders are important considerations for future research.

Key Words: Irisin; Prediabetes mellitus; Enzyme-linked immunosorbent assay; Methodological variability; Standardization; Case-control study; Meta-analysis

Core Tip: This study suggests that methodological variability between commercial enzyme-linked immunosorbent assay kits may contribute to the conflicting literature on circulating irisin in prediabetes, supported by poor inter-assay agreement in this cohort. After rigorously controlling for metabolic confounders, the observed association between prediabetes and irisin disappears, shifting the focus from irisin as a simple biomarker to the critical need for assay standardization and robust study design in future research.

INTRODUCTION

Prediabetes mellitus (PreDM) is a metabolic state marked by blood glucose levels above the normal threshold but below the criteria for a formal diabetes diagnosis, and it has been well-documented as a strong predictor of type 2 diabetes mellitus (T2DM)[1,2]. Accumulating evidence from earlier research indicates that individuals diagnosed with PreDM exhibit a substantially increased likelihood of progressing to overt diabetes[3,4]. Dysfunctions in both insulin secretion and peripheral action are responsible for driving this pathological risk[5], while the etiological basis of these abnormalities remains a subject of ongoing investigation. Clinical evidence has confirmed that regular physical exercise can serve as an effective intervention to prevent PreDM from progressing to T2DM[6].

As a robust physiological trigger, physical activity (PA) regulates substrate oxidation and hormone secretion profiles, with its effects varying according to the mode, intensity, and duration of exercise[7]. Clinical investigations have demonstrated that lifestyle modification regimens encompassing 150 minutes of weekly moderate-intensity PA, with brisk walking as the primary modality, can cut the incidence of T2DM by as much as 58%[8]. Mechanistically, signaling cascades triggered by PA can activate multiple cellular mediators, including the energy-sensing enzyme AMPK, nitric oxide synthase, calcium/calmodulin-dependent protein kinase, and the transcriptional coactivator PGC-1α. Activation of these factors further elevates insulin sensitivity, thereby ameliorating a series of pivotal metabolic parameters such as enhanced GLUT4 expression, elevated aerobic enzyme activity, and augmented mitochondrial biogenesis[9,10].

In the course of the last decade, mounting evidence has revealed a more intricate dimension underlying the regulatory mechanisms of glucose and lipid metabolism. Skeletal muscle tissue, when undergoing contraction, releases a broad spectrum of bioactive peptides known as “myokines”; these factors not only coordinate the energy supply necessary for sustaining muscle activity, but also participate in the modulation of systemic metabolic balance[11]. PGC-1α, a pivotal transcriptional cofactor, is known to mediate multiple adaptive metabolic responses including mitochondrial biogenesis and/or adipose tissue browning[12,13]. Intriguingly, recent investigations have uncovered a functional link between myokines and PGC-1α, which is embodied by irisin-a novel and highly noteworthy PGC-1α-driven myokine. The principal cellular target of irisin is white adipose tissue cells; its core biological function involves reprogramming white adipose tissue-an organ associated with energy storage and the pathogenesis of insulin resistance-into brown adipose tissue, a process that facilitates lipolytic breakdown, enhances energy expenditure, and induces heat generation[14].

It has been postulated that irisin promotes the browning of subcutaneous adipocytes and augments thermogenic capacity by increasing the expression level of UCP1. By virtue of this mechanism, irisin mediates the favorable effects of exercise on systemic energy metabolism, leading to increased energy expenditure, prolonged lifespan, reduced body weight, and alleviated insulin resistance[15]. These observations highlight the indispensable role of PA in regulating the secretion of myokines such as irisin. Moreover, recent experimental evidence has demonstrated that adipose tissue also serves as a source of irisin secretion[16]. However, the nature of the associations between circulating irisin concentrations and glucose metabolic parameters remains elusive in human subjects.

Separately, glucagon is secreted by pancreatic α-cells in response to multiple regulatory factors, including various nutrient signals, neural modulators, hormonal stimuli, and the onset of hypoglycemia[17,18]. The biological functions exerted by glucagon are functionally antagonistic to those of insulin. Historically, diabetes-related research has been predominantly centered on the role of insulin; in recent years, however, growing attention has been directed toward the pivotal contribution of glucagon to the pathogenesis of both overt diabetes and its prediabetic state. Accordingly, aberrant glucagon secretion has been identified as a core focus for dissecting the pathogenic mechanisms underlying diabetes. Furthermore, to date, no research has yet explored the potential interplay between glucagon and myokines.

Despite numerous studies investigating irisin levels in prediabetes, the findings remain highly inconsistent and contradictory. Some studies report significantly lower circulating irisin concentrations in individuals with prediabetes compared to healthy controls[19-22], suggesting a potential protective role that is diminished in the pre-diabetic state. Conversely, other studies have found no significant difference or even elevated levels[23-26]. This substantial heterogeneity in the literature complicates the interpretation of irisin's role in early dysglycemia. A critical, yet often overlooked, factor contributing to this inconsistency may be the methodological variability in irisin quantification, particularly the use of different enzyme-linked immunosorbent assay (ELISA) kits from various manufacturers, which may yield discordant results due to differences in antibody specificity, calibration, and assay performance.

To address these knowledge gaps and reconcile the conflicting evidence, we conducted a dual-component study integrating a systematic meta-analysis with a primary clinical investigation. The specific aims of this study were: (1) To perform a meta-analysis of existing observational studies to provide a pooled quantitative estimate of the difference in circulating irisin levels between individuals with prediabetes and healthy controls, and to explore potential sources of heterogeneity, particularly focusing on assay methodology; (2) To conduct an exploratory case-control study in a clinical cohort to directly compare irisin levels measured by two distinct, commercially available ELISA kits (ELK and Elabscience) in well-characterized individuals with prediabetes and normal glucose tolerance (NGT) controls, while meticulously accounting for key metabolic confounders; and (3) To evaluate the agreement between the two ELISA kits and to explore the potential correlation between irisin and glucagon levels, thereby providing empirical evidence on how kit choice might influence study outcomes and contribute to the observed literature inconsistency.

MATERIALS AND METHODS

Case-control study

Study design and participants: This was a single-center, exploratory cross-control study, conducted from December 2024 to March 2025. Initially, 50 individuals were screened, of which 35 eligible participants were enrolled. Inclusion criteria were: (1) Age ≥ 18 years; and (2) No history of using glucose-lowering medications. Exclusion criteria included: (1) Diseases affecting glucose metabolism, such as thyroid dysfunction, Cushing’s syndrome, or pancreatitis; (2) Use of medications interfering with glucose levels, such as glucocorticoids, thyroid hormones, or β-blockers; (3) Presence of severe hepatic or renal insufficiency, history of malignancy, or significant gastrointestinal diseases; and (4) Experiencing major illnesses (e.g., new-onset myocardial infarction or cerebral infarction within the past 6 months) or major stressful life events recently. All participants provided written informed consent. The study protocol was approved by the Ethics Committee of Tianjin Medical University General Hospital (Approval No. IRB2024-YX-139-01).

Given its exploratory nature aimed at investigating circulating irisin levels and their complex relationship with metabolic parameters (e.g., glucagon) in a relatively younger cohort with prediabetes, a formal prospective sample size calculation was not performed. The total sample size of 35 participants (prediabetes: n = 9; NGT controls: n = 26) was determined primarily based on feasibility considerations, including the availability of eligible participants meeting the stringent inclusion and exclusion criteria within the study period and the logistical resources available. We acknowledge that this limited sample size, particularly in the prediabetes subgroup, may constrain the statistical power to detect small-to-moderate effect sizes and increase the risk of type II errors. Therefore, the findings should be interpreted as preliminary and hypothesis-generating.

Data collection and oral glucose tolerance test: On the examination day, a trained healthcare professional measured the height, body weight, waist circumference, hip circumference, and blood pressure of all non-hospitalized volunteers. Body mass index (BMI) was calculated as weight in kilograms divided by the square of height in meters, with the result rounded to two decimal places. General information, including sex, age, history of smoking and alcohol consumption, family history of diabetes, and fatty liver status, was collected via a structured questionnaire.

All participants underwent a standard 75 g oral glucose tolerance test (OGTT). Before the test, participants were instructed to maintain a normal diet for at least 3 days, with a daily carbohydrate intake of no less than 150 g. After 22:00 on the day before the test, they underwent an overnight fast of at least 8 hours. Fasting venous blood samples were collected prior to glucose ingestion for the measurement of blood glucose, insulin, C-peptide, glucagon, and irisin. Subsequently, volunteers consumed 300 mL of a solution containing 75 g of anhydrous glucose within 5 minutes. The timing of the test commenced upon the first sip. Venous blood samples were then collected at 0.5-, 1-, 2-, and 3-hour after glucose ingestion to measure the levels of blood glucose, insulin, C-peptide, and glucagon. During the entire test period, participants were instructed to avoid strenuous exercise, consumption of strong tea, coffee, or other stimulants, smoking, and alcohol intake. The test was immediately suspended if any participant experienced discomfort such as vomiting. Due to the multiple blood draws, the OGTT was not repeated for any volunteer.

Diagnostic criteria and laboratory measurements: According to the American Diabetes Association criteria[27], NGT was defined as fasting plasma glucose (FPG) < 5.6 mmol/L and 2-hour plasma glucose (2hPG) during a 75 g OGTT < 7.8 mmol/L. Impaired glucose regulation was defined as FPG ≥ 5.6 mmol/L and < 7.0 mmol/L, and/or 2hPG ≥ 7.8 mmol/L and < 11.1 mmol/L.

Blood glucose levels were measured using the hexokinase method (Beckman 5800, United States). Insulin and C-peptide levels were determined by electrochemiluminescence immunoassay (Roche E602, Germany). Glucagon levels were assessed using a homogeneous chemiluminescence immunoassay (HomoG 100, China). Circulating irisin concentration was measured using commercially available ELISA kits. To account for potential inter-assay variability, we employed ELISA kits from two different manufacturers: ELK Biotechnology (ELK10769) and Elabscience (E-EL-H5735). The intra-assay and inter-assay coefficients of variation for both kits were reported to be < 8% and < 10%, respectively. All assays were performed in strict accordance with the manufacturers' protocols.

Meta-analysis

Search strategy: We searched the Cochrane Library, EMBASE, and PubMed from inception through November 2025. To uphold the comprehensiveness and precision of the literature search process, a hybrid approach combining medical subject headings and free vocabulary terms was utilized. The full details of the search methodology are documented in Supplementary Table 1. Moreover, to identify additional potentially eligible studies, the research team implemented a manual backward reference screening of the bibliographic lists from all incorporated studies.

To ensure the scientific rigor of literature screening and the reliability of subsequent pooled analyses, this study pre-established strict inclusion criteria, requiring eligible studies to simultaneously meet the following requirements: (1) They must be peer-reviewed observational studies (cohort, case-control, or cross-sectional designs, excluding randomized controlled trials, systematic reviews, case reports, and other non-observational studies); (2) The study subjects must consist of clearly diagnosed prediabetic individuals and control groups of healthy or verified non-diabetic populations; (3) Peripheral circulating irisin concentrations must be quantified by ELISA using plasma or serum samples; (4) The mean ± SD of irisin concentrations must be directly reported or indirectly calculable from available indicators [e.g., standard error, median and interquartile range (IQR)]; and (5) The publications must be in English to avoid translation bias in data extraction.

To minimize the confounding impact of irrelevant information on the analysis outcomes, the present study excluded two categories of literature: (1) Non-original research and duplicate publications, encompassing review articles, commentaries, conference abstracts lacking complete study data, preclinical studies (e.g., animal or cell experiments without human data), and duplicate reports from the same research team based on identical samples; and (2) Literature from which core outcome indicator data (including circulating irisin levels with mean and standard deviation, as well as diabetes subtype diagnosis) could not be directly extracted or indirectly estimated via formulas or graphs.

Quality assessment and data extraction: To guarantee the reliability of literature screening and data extraction, this study adopted a dual independent reviewer model: Two researchers completed literature retrieval, data extraction, and methodological quality appraisal separately before cross-verifying the results, with any discrepancies resolved by a third senior investigator as an arbitrator to reach a consensus. The core data extracted from eligible studies included first author, publication year, research region, study design, diagnostic criteria, case and control characteristics, separate sample sizes of the two groups, manufacturer of the ELISA kit for irisin detection, specific quantification values of peripheral circulating irisin concentrations (e.g., mean ± SD), and so on. All the included studies were cross-sectional studies, and thus the Agency for Healthcare Research and Quality scale was adopted for quality assessment, with studies scoring ≥ 7 classified as high-quality and 5-6 as moderate-quality.

Statistical analysis

For the clinical studies, statistical analysis was performed using IBM SPSS Statistics software (Version 27.0). Data conforming to a normal distribution are presented as mean ± SD and were compared between groups using independent samples t-tests. Non-normally distributed data are presented as median (IQR) and were compared using the Mann-Whitney U test. Categorical variables are presented as n (%) and were compared using χ² test or Fisher’s exact test, as appropriate. To control for potential confounding due to baseline imbalances (e.g., gender, BMI, fatty liver status), multiple linear regression analyses were performed with irisin levels (from each kit) as the dependent variable, and group (NGT/PreDM) along with the confounders as independent variables. Spearman’s rank correlation coefficient was used to assess correlations between variables. Agreement between the two ELISA kits was evaluated using Bland-Altman analysis and the Wilcoxon signed-rank test. A two-tailed P value < 0.05 was considered statistically significant.

For the meta-analysis, STATA 14.0 software (Stata Corporation, United States) was used. The pooled effect size was expressed as the standardized mean difference (SMD) and its 95% confidence interval (95%CI). Heterogeneity between studies was quantified using the I² statistic. A random-effects model was applied if significant heterogeneity was present (I² > 50%); otherwise, a fixed-effects model was used. Sensitivity analysis was performed by omitting one study at a time to assess the relative influence of individual studies on the pooled estimate. Publication bias was examined using two distinct methods: Begg’s adjusted rank correlation test and Egger’s regression asymmetry test. To explore potential sources of heterogeneity, pre-specified subgroup analyses were conducted based on ELISA kit manufacturer and specimen type (serum vs plasma).

RESULTS

Clinical and laboratory characteristics of study participants

A total of 35 subjects were recruited, including 9 participants with PreDM and 26 NGT controls. The basic clinical characteristics are shown in Table 1. Significant imbalances were observed between the two groups at baseline. The PreDM group had a significantly higher proportion of males (55.6% vs 11.5%, P = 0.015) and a higher prevalence of fatty liver (55.6% vs 11.5%, P = 0.015) compared to the NGT group. The PreDM group also exhibited significantly higher BMI, waist circumference, waist-to-hip ratio, and fasting and postprandial glucose levels (all P < 0.05). There were no statistically significant differences in age, family history of diabetes, smoking history, or alcohol consumption between the two groups (all P > 0.05).

Table 1 Baseline characteristics of the study participants.

Characteristic	NGT (n = 26)	PreDM (n = 9)	P value
Demographics
Age (year)	32.00 (30.00, 32.00)	33.00 (28.00, 39.00)	0.810
Male	3 (11.5)	5 (55.6)	0.015
Anthropometrics
Body mass index (kg/m2)	22.11 (19.00, 23.00)	25.01 (23.00, 28.00)	0.001
Waist circumference (cm)	76.48 ± 7.63	94.03 ± 9.41	0.000
Hip circumference (cm)	97.00 (94.00, 98.30)	105.00 (99.00, 106.30)	0.003
Waist-to-hip ratio	0.79 ± 0.06	0.90 ± 0.06	0.000
Clinical status
Fatty liver	3 (11.5)	5 (55.6)	0.015
Family history of DM	7 (26.9)	4 (44.4)	0.416
Smoking status	0 (0)	1 (11.1)	0.257
Alcohol drinking status	4 (15.4)	4 (44.4)	0.162
Biochemical parameters
Irisin (ng/mL) ELK	5.79 (4.99, 6.07)	5.89 (5.06, 6.57)	0.985
Irisin (pg/mL) Elabscience	22.99 ± 8.90	30.59 ± 10.95	0.045
Glu0 (mmol/L)	4.95 ± 0.38	5.62 ± 0.53	0.000
Glu0.5 (mmol/L)	7.86 ± 0.31	10.24 ± 0.88	0.000
Glu1 (mmol/L)	6.83 ± 1.79	10.38 ± 1.97	0.000
Glu2 (mmol/L)	5.48 (4.81, 6.07)	7.84 (6.95, 8.17)	0.000
Glu3 (mmol/L)	4.35 (3.75, 4.81)	4.30 (3.82, 6.23)	0.697
Ins0 (μIU/mL)	8.67 (6.00, 11.00)	15.61 (11.00, 24.00)	0.001
Ins0.5 (μIU/mL)	90.96 (58.00, 142.00)	117.22 (55.00, 183.00)	0.540
Ins1 (μIU/mL)	92.64 (51.00, 153.64)	154.21 (92.00, 217.21)	0.056
Ins2 (μIU/mL)	49.37 (32.00, 81.37)	101.87 (58.00, 204.87)	0.001
Ins3 (μIU/mL)	25.95 (12.00, 41.95)	37.83 (22.00, 54.83)	0.271
CP0 (ng/mL)	1.80 (1.45, 2.00)	2.90 (2.32, 3.33)	0.000
CP0.5 (ng/mL)	8.30 (6.01, 10.43)	9.36 (5.89, 11.97)	0.838
CP1 (ng/mL)	9.29 (6.01, 15.83)	12.72 (9.28, 14.34)	0.079
CP2 (ng/mL)	7.92 ± 2.42	12.04 ± 3.39	0.000
CP3 (ng/mL)	5.24 ± 2.13	6.26 ± 2.06	0.221
GCG0 (pmol/L)	14.79 ± 6.42	20.41 ± 8.77	0.048
GCG0.5 (pmol/L)	5.83 ± 3.19	10.85 ± 3.31	0.000
GCG1 (pmol/L)	3.67 (2.00, 5.00)	6.68 (4.00, 7.00)	0.005
GCG2 (pmol/L)	3.74 (2.00, 5.00)	5.51 (3.00, 8.00)	0.061
GCG3 (pmol/L)	5.46 (2.00, 8.00)	11.06 (6.00, 24.00)	0.016

Data conforming to a normal distribution are presented as mean ± SD. Non-normally distributed data are presented as median (IQR). Categorical variables are presented as n (%). NGT: Normal glucose tolerance; PreDM: Prediabetes mellitus; DM: Diabetes mellitus; Glu0: Fasting blood glucose; Glu0.5: Oral glucose tolerance test 0.5-hour post-load blood glucose; Glu1: Oral glucose tolerance test 1-hour post-load blood glucose; Glu2: Oral glucose tolerance test 2-hour post-load blood glucose; Glu3: Oral glucose tolerance test 3-hour post-load blood glucose; Ins0: Fasting insulin; Ins0.5: Oral glucose tolerance test 0.5-hour post-load blood insulin; Ins1: Oral glucose tolerance test 1-hour post-load blood insulin; Ins2: Oral glucose tolerance test 2-hour post-load blood insulin; Ins3: Oral glucose tolerance test 3-hour post-load blood insulin; CP0: Fasting C-peptide; CP0.5: Oral glucose tolerance test 0.5-hour post-load blood C-peptide; CP1: Oral glucose tolerance test 1-hour post-load blood C-peptide; CP2: Oral glucose tolerance test 2-hour post-load blood C-peptide; CP3: Oral glucose tolerance test 3-hour post-load blood C-peptide; GCG0: Fasting glucagon; CGC0.5: Oral glucose tolerance test 0.5-hour post-load blood glucagon; GCG1: Oral glucose tolerance test 1-hour post-load blood glucagon; GCG2: Oral glucose tolerance test 2-hour post-load blood glucagon; GCG3: Oral glucose tolerance test 3-hour post-load blood glucagon.

Intergroup comparison of irisin levels

Simple comparisons: The irisin level measured by Elabscience was significantly higher in the PreDM group than in the NGT group (30.59 ± 10.95 vs 22.99 ± 8.90, P = 0.045). In contrast, the irisin level measured by ELK showed no significant difference between the two groups [median (IQR): 5.79 (4.99, 6.07) vs 5.89 (5.06, 6.57), P = 0.985].

Adjusted comparisons: To account for the significant baseline imbalances, multiple linear regression analyses were performed. After adjusting for gender, BMI, fatty liver status, waist circumference, hip circumference, and waist-to-hip ratio, the association between prediabetes status and irisin levels was reassessed. For ELK, the regression model was statistically significant (F = 3.534, P = 0.010, adjusted R² = 0.309). Within this model, BMI was positively associated with irisin level (β = 0.513, P = 0.015), while the presence of fatty liver was negatively associated (β = -0.483, P = 0.016). However, the group variable (NGT/PreDM) itself was not retained in the final model, likely due to model instability and overfitting given the small sample size and severe baseline imbalance. This further supports that the clinical data should be interpreted as strictly exploratory. For Elabscience, the overall regression model was not significant (F = 1.184, P = 0.343, adjusted R² = 0.031), indicating that after adjustment, prediabetes status and the included covariates did not significantly explain the variance in irisin levels measured by this kit.

Agreement between the two assay kits: (1) Wilcoxon signed-rank test: A highly significant difference was found between the irisin concentrations measured by ELK and Elabscience (P < 0.001). The median value from Elabscience was substantially higher than that from ELK; and (2) Bland-Altman analysis: The Bland-Altman plot (Figure 1) revealed poor agreement between the two kits. The mean difference (bias) was 5837.25 pg/mL. The 95% limits of agreement were very wide (limits of agreement: 4197.81-7476.69 pg/mL). Furthermore, a clear proportional bias was observed, with the difference between the two measurements increasing as the average irisin level increased.

Figure 1 Bland-Altman plot comparing irisin concentrations measured by the two enzyme-linked immunosorbent assay kits.

Correlations between irisin and metabolic parameters

Spearman correlation analysis was performed between irisin levels (Elabscience) and other metabolic parameters. Irisin level showed a significant positive correlation with fatty liver status (r = 0.350, P = 0.039) and with 0.5-hour postprandial glucose (r = 0.421, P = 0.012). No significant correlations were found with BMI, waist circumference, fasting glucagon, insulin, or C-peptide levels at various time points (all P > 0.05, Table 2).

Table 2 Correlation of irisin with other metabolic parameters.

	r value	P value
Irisin (ng/mL) ELK	-0.183	0.294
Body mass index	0.185	0.286
Gender	-0.04	0.818
Age	-0.026	0.884
Waist circumference	0.098	0.577
Hip circumference	-0.054	0.757
Waist-to-hip ratio	0.203	0.241
Fatty liver	0.350	0.039
Family history of DM	0.152	0.382
Glu0	0.168	0.336
Glu0.5	0.421	0.012
Glu1	0.307	0.073
Glu2	0.154	0.377
Glu3	0.207	0.232
GCG0	0.045	0.796
GCG0.5	0.066	0.707
GCG1	0.009	0.961
GCG2	-0.044	0.802
GCG3	-0.076	0.664
Ins0	-0.032	0.857
Ins0.5	0.018	0.919
Ins1	-0.063	0.718
Ins2	0.028	0.874
Ins3	0.254	0.141
CP0	0.046	0.793
CP0.5	-0.007	0.969
CP1	-0.049	0.779
CP2	0.003	0.987
CP3	0.184	0.289
Smoking history	-0.136	0.437
Alcohol drinking history	0.101	0.564

DM: Diabetes mellitus; Glu0: Fasting blood glucose; Glu0.5: Oral glucose tolerance test 0.5-hour post-load blood glucose; Glu1: Oral glucose tolerance test 1-hour post-load blood glucose; Glu2: Oral glucose tolerance test 2-hour post-load blood glucose; Glu3: Oral glucose tolerance test 3-hour post-load blood glucose; Ins0: Fasting insulin; Ins0.5: Oral glucose tolerance test 0.5-hour post-load blood insulin; Ins1: Oral glucose tolerance test 1-hour post-load blood insulin; Ins2: Oral glucose tolerance test 2-hour post-load blood insulin; Ins3: Oral glucose tolerance test 3-hour post-load blood insulin; CP0: Fasting C-peptide; CP0.5: Oral glucose tolerance test 0.5-hour post-load blood C-peptide; CP1: Oral glucose tolerance test 1-hour post-load blood C-peptide; CP2: Oral glucose tolerance test 2-hour post-load blood C-peptide; CP3: Oral glucose tolerance test 3-hour post-load blood C-peptide; GCG0: Fasting glucagon; CGC0.5: Oral glucose tolerance test 0.5-hour post-load blood glucagon; GCG1: Oral glucose tolerance test 1-hour post-load blood glucagon; GCG2: Oral glucose tolerance test 2-hour post-load blood glucagon; GCG3: Oral glucose tolerance test 3-hour post-load blood glucagon.

Meta-analysis of circulating irisin in prediabetes

A systematic search identified 51 articles related to the association between circulating irisin and diabetes. Study selection flow diagram is presented in Figure 2.

Figure 2 Literature screening flow chart.

A total of 8 studies comparing irisin levels between individuals with prediabetes and healthy controls were included in the meta-analysis. One study contributed three independent prediabetes subgroups, resulting in a total of 10 datasets for pooled analysis. Due to the significant heterogeneity observed among the included studies, the random-effects model was applied.

The pooled analysis demonstrated that individuals with prediabetes had significantly lower irisin levels compared to healthy controls (SMD = -0.524, 95%CI: -0.897 to -0.152, P = 0.006). There was substantial heterogeneity across the studies (I² = 86.7%, P < 0.001, Figure 3A).

Figure 3 Forest plot. A: Prediabetes mellitus for the association between circulating irisin levels and diabetes risk; B: Leave-one-out sensitivity analysis. SMD: Standardized mean difference; CI: Confidence interval.

To explore potential sources of heterogeneity, pre-specified subgroup analyses were performed based on the specimen type (plasma vs serum) and the ELISA kit manufacturer. The results are presented in Table 3.

Table 3 Subgroup analyses of irisin levels.

Subgroup	Number of studies	SMD (95%CI)	P value	I2 (%)
By specimen type
Plasma	5	-0.769 (-1.304 to -0.234)	0.005	87.0
Serum	5	-0.282 (-0.826 to 0.263)	0.311	87.4
By ELISA kit manufacturer
Adipogen	3	-0.565 (-0.791 to -0.338)	< 0.001	0
CusabioTM	1	-0.533 (-0.915 to -0.151)	0.006
Bio Vision, Milpitas	1	-1.307 (-1.774 to -0.840)	< 0.001
Phoenix	1	0.517 (-0.047 to 1.082)	0.072
SinoGeneClon Biontec	1	-0.124 (-0.525 to 0.276)	0.542
USCN Life Science	1	-0.071 (-0.530 to 0.389)	0.763
Not specified (NO)	2	-1.116 (-3.388 to 1.155)	0.335	97.3
Overall	10	-0.524 (-0.897 to -0.152)	0.006	86.7

SMD: Standardized mean difference; 95%CI: 95% confidence interval; ELISA: Enzyme-linked immunosorbent assay.

The subgroup analysis based on specimen type showed a significant reduction in irisin levels in the prediabetes group when plasma was used (SMD = -0.769, P = 0.005), whereas no significant difference was observed in studies utilizing serum (SMD = -0.282, P = 0.311). This discrepancy underscores the potential influence of pre-analytical factors.

The subgroup analysis by ELISA kit manufacturer provided critical insights. Notably, studies employing kits from Adipogen, Cusabio^TM, and Bio Vision consistently reported significant reductions in irisin levels among prediabetic individuals. In contrast, the study using a Phoenix kit reported a non-significant increasing trend, while those using SinoGeneClon Biontec and USCN Life Science kits found no significant difference. The marked heterogeneity within the subgroup of studies that did not specify the kit brand (I² = 97.3%) further highlights the confounding effect of this variable.

These observations suggest that assay related factors may contribute to inconsistent findings. However, several manufacturer-based subgroups contained only one study, limiting their ability to fully explain heterogeneity. Thus, the role of immunoassay standardization in explaining literature inconsistencies should be interpreted cautiously.

Sensitivity analysis using the leave-one-out method indicated that the overall pooled estimate was robust, as the omission of any single study did not substantially alter the direction or significance of the pooled SMD, with all 95%CIs overlapping that of the combined estimate (Figure 3B).

Both Begg’s test (z = -0.09, P = 0.929) and Egger’s test (bias coefficient = -6.00, P = 0.247) showed no statistical evidence of significant publication bias (Figure 4).

Figure 4 Begg’s plot and Egger’s plot for prediabetes mellitus. SMD: Standardized mean difference.

DISCUSSION

Main findings and comparison with meta-analysis

This integrated study, comprising a systematic meta-analysis and an exploratory clinical investigation, yields insights into the association between circulating irisin and prediabetes. The meta-analysis of eight studies demonstrated a statistically significant reduction in irisin levels among individuals with prediabetes compared to healthy controls. This pooled finding suggests that diminished irisin may be a feature associated with early dysglycemia. In contrast, our initial clinical observation using one of the two ELISA kits indicated higher irisin levels in the prediabetes group. This apparent discrepancy underscores the complexity and inconsistency prevalent in the existing literature regarding irisin in prediabetes.

Our study also observed a similar discrepancy: Although the mean age was comparable (around 30-35 years) between the prediabetes and control groups, the control group had a significantly higher proportion of females. Given that age and sex are established factors influencing irisin levels[28-30], we postulate that differences in these baseline characteristics may be a significant source of the inconsistency across studies. Specifically, older age and a higher female proportion are likely associated with lower baseline irisin levels, which may partly explain the overall decreasing trend observed in the meta-analysis. Our relatively younger cohort in the cross-sectional study might have contributed to the observed overall higher irisin levels, while the predominance of females in the control group could have lowered its levels. These demographic variations add another layer to the considerable heterogeneity quantified in our meta-analysis (I² = 86.7%). More directly pertinent to our clinical cohort, and as detailed in Table 1, we observed significant baseline imbalances in several key metabolic parameters—specifically, a higher proportion of males, greater adiposity (higher BMI, waist circumference), and a higher prevalence of fatty liver in the prediabetes group. It is these particular imbalances that necessitated the use of multivariate analyses to isolate the potential independent association between prediabetes status and irisin levels.

Beyond methodological differences, variations in population baseline characteristics may also contribute to inconsistent findings. Our clinical cohort, with a mean age of 30-35 years, is considerably younger than participants in many published studies reporting lower irisin levels in prediabetes, which often involved middle-aged or older populations. As irisin levels are positively associated with muscle mass and basal metabolic rate—parameters that typically decline with age—the relatively young individuals with prediabetes in our study might have preserved a higher baseline irisin secretory capacity. This may partly explain why, in unadjusted comparisons using the Elabscience kit, we observed a trend opposite to the meta-analysis conclusion (i.e., higher levels in the prediabetes group). This disparity in age composition, coupled with the higher proportion of females (who may have lower irisin levels) in our control group, adds complexity to cross-study comparisons and contributes to the substantial heterogeneity quantified in our meta-analysis. Therefore, to account for these confounders (gender, BMI, fatty liver status, etc.), we performed multiple linear regression analyses. After adjustment for covariates such as gender, BMI, and fatty liver status, the association between prediabetes status and irisin levels was attenuated and lost statistical significance for the kit that initially showed a difference. This highlights the critical influence of metabolic confounders, particularly adiposity-related measures, on circulating irisin levels. The variance in irisin levels within our cohort appeared to be more closely associated with these confounding factors than with glucose tolerance status per se. Therefore, the initial unadjusted between-group difference should be interpreted with caution, and the adjusted analyses do not provide strong evidence for an independent association between prediabetes and irisin in this specific cohort.

The “meta-analysis-primary study” dual-model design employed in this research offers distinct methodological advantages. First, the systematic meta-analysis quantitatively synthesized the overall heterogeneity in the field and identified assay methodology as a likely key contributor. This finding directly informed the core design of our clinical investigation: The parallel measurement of irisin using two distinct commercial ELISA kits within the same cohort. This approach allowed us to empirically test the methodological variability hypothesis suggested by the meta-analysis under controlled conditions, tracing observed literature discrepancies to specific detection tools. Second, the clinical component enabled meticulous adjustment for confounders (e.g., sex, adiposity, fatty liver) that are difficult to fully account for in meta-analyses. Therefore, this study not only reports on the association of irisin with prediabetes but also provides a comprehensive analytical framework “from detecting heterogeneity to empirically investigating its sources”, which is challenging to achieve with a single study design. This represents a primary contribution of our work towards clarifying the controversies surrounding irisin.

Reconciling discrepancies: The pivotal role of methodological variability

A pivotal finding from our clinical study is the profound lack of agreement between the two commercially available ELISA kits used. Quantitative comparison via Bland-Altman analysis revealed substantial mean bias and wide limits of agreement, indicating poor interchangeability. Furthermore, the observed proportional bias suggests that the discrepancy between kits is not constant but varies with the concentration level. This empirical demonstration of significant inter-assay variability provides a quantifiable and plausible explanation for the heterogeneous and often contradictory results reported in the literature. It underscores that differences in assay methodology, including antibody specificity and calibration, are a major source of the current controversy surrounding irisin levels in metabolic disorders. Consequently, comparisons of absolute irisin values across studies employing different kits are problematic, and conclusions drawn from single-assay studies require careful consideration of methodological context.

However, the interpretation of this finding must be tempered by the substantial heterogeneity observed across studies (I² = 86.7%). Subgroup analyses suggested that this heterogeneity might be partly associated with methodological differences, including specimen type and ELISA kit manufacturer.

Specimen type: The subgroup analysis based on specimen type showed a significant reduction in irisin levels in the prediabetes group when plasma was used (SMD = -0.769, P = 0.005), whereas no significant difference was observed in studies utilizing serum (SMD = -0.282, P = 0.311). This discrepancy suggests the potential influence of pre-analytical factors. Differences in anticoagulants used in plasma collection, clotting processes in serum preparation, and the stability or binding profiles of irisin in different matrices may affect measured concentrations, limit the comparability of results across studies.

ELISA kit manufacturer: The subgroup analysis by ELISA kit manufacturer provided notable differences. Notably, studies employing kits from Adipogen, Cusabio^TM, and Bio Vision generally reported significant reductions in irisin levels among prediabetic individuals. In contrast, the study using a Phoenix kit reported a non-significant increasing trend, while those using SinoGeneClon Biontec and USCN Life Science kits found no significant difference. The marked heterogeneity within the subgroup of studies that did not specify the kit brand (I² = 97.3%) also indicates the confounding effect of this variable. These observations suggest that the lack of standardization across commercial immunoassays possibly due to variations in antibody specificity, epitope recognition, calibration standards, and dynamic ranges may be an important source of the inconsistent findings in the literature regarding irisin levels in prediabetes.

This assay-dependency likely explains why some studies in obese or metabolic syndrome populations reported higher irisin in disease states[31-33], while others reported higher levels in diabetics on medication, and yet others, including our meta-analysis, report decreases[20,23,24,34].

Acknowledging these methodological challenges, our present case-control study was deliberately designed to use plasma samples. This choice aligns with the subgroup of studies in our meta-analysis that showed a more consistent signal of irisin reduction. Furthermore, to directly address the critical issue of assay variability highlighted by our meta-analysis, we employed two distinct ELISA kits for parallel measurement of irisin in the same set of samples. This intra-study comparison allows for a direct assessment of the concordance (or lack thereof) between different commercial assays within a single, well-defined cohort, thereby providing unique insights into the methodological discordance observed across different studies.

The persistent heterogeneity, even within certain subgroups, underscores the urgent need for the development and adoption of a gold-standard reference method (e.g., mass spectrometry) or internationally harmonized immunoassays for irisin measurement. Future large-scale studies should rigorously report and, if possible, standardize pre-analytical procedures and the specific assay employed. Until such standardization is achieved, conclusions regarding absolute irisin levels in metabolic conditions must be drawn with caution, with explicit reference to the methodological context.

Correlation with metabolic parameters

Our exploratory correlation analysis revealed a significant positive association between irisin levels and both fatty liver status and early postprandial glucose. The correlation with fatty liver aligns with emerging evidence of the liver as a site of irisin action and potential secretion, linking this myokine to hepatic steatosis—a common condition in prediabetes. The association with 0.5-hour postprandial glucose suggests that irisin may be responsive to acute glycemic changes. This correlation aligns with a subset of literature suggesting irisin is associated with measures of glycemia[31,35], warranting further investigation into its dynamic secretion profile during metabolic challenges. No significant correlation was found with fasting glucagon levels in this study, indicating that any potential interplay between irisin and glucagon in the fasting state, if it exists, is not readily apparent in this cohort and under these conditions.

Limitations and strengths

Several limitations of this meta-analysis should be acknowledged. First, the number of studies included, particularly within each manufacturer-specific subgroup, was relatively small, limiting the statistical power and generalizability of subgroup findings. Second, we could not account for all potential confounding factors from the original studies, such as differences in participant age, sex distribution, BMI, and PA levels, which may also contribute to heterogeneity. Finally, publication bias, though not statistically detected in our analysis, cannot be entirely ruled out.

The primary strength of this research is the integrated approach of meta-analysis and primary data collection, which allowed us to first quantify the overall heterogeneity in the field and then empirically demonstrate key sources of it. We systematically compared two common assays and correlated irisin with glucagon, a previously unexplored relationship in prediabetes.

CONCLUSION

In conclusion, this study suggests that circulating irisin levels in prediabetes may be influenced by assay methodology and metabolic confounding. Variability between commercial ELISA kits likely contributes to the inconsistent literature, but definitive confirmation requires larger, standardized studies. These findings highlight the importance of assay standardization and transparent methodological reporting in future irisin research. To definitively elucidate the physiological role of irisin in dysglycemia, future investigations should be larger longitudinal studies with adequate power. They must employ standardized methodologies, incorporate comprehensive metabolic phenotyping, and prioritize matched designs (e.g., for age, sex, and BMI) based on a priori sample size calculations to minimize confounding bias.

ACKNOWLEDGEMENTS

The authors gratefully thank all the participants and research staff for their invaluable contributions to this study.