Research, Society and Development, v. 15, n. 9, e3115951680, 2026 (CC BY 4.0) | ISSN 2525-3409 | DOI: http://dx.doi.org/10.33448/rsd-v15i9.51680 Berberine and insulin resistance in non-diabetic adults: An updated systematic review and meta-analysis of randomized controlled trials Berberina e resistência à insulina em adultos não diabéticos: Revisão sistemática atualizada e metanálise de ensaios clínicos randomizados Berberina y resistencia a la insulina en adultos no diabéticos: Revisión sistemática actualizada y metaanálisis de ensayos clínicos aleatorizados Received: 09/05/2026 | Accepted: 09/08/2026 | Published: 09/09/2026 Vinícius Arantes Campos1 ORCID: https://orcid.org/0009-0008-1054-0285 Clínicas Expert, Brazil E-mail: drviniciusarantes@gmail.com Allan Gomes Doriguetto2 ORCID: https://orcid.org/0009-0007-0605-3591 Hospital IGESP, Brazil E-mail: allandoriguetto@gmail.com Abstract Berberine is a plant alkaloid with insulin-sensitizing properties. Recent meta-analyses have addressed glycemic endpoints and components of the metabolic syndrome, but none has specifically synthesized its effect on insulin resistance measured by the homeostatic model assessment of insulin resistance in non-diabetic populations. The objective was to estimate that effect in non-diabetic adults. We searched PubMed for randomized controlled trials comparing berberine, in any salt, formulation or dose, alone or as an add-on, with a non-berberine control or an active comparator, in non-diabetic adults with insulin resistance or metabolic risk. The primary outcome was the mean difference in the homeostatic model assessment of insulin resistance. A random-effects model was fitted by restricted maximum likelihood, with subgroup and leave-one-out sensitivity analyses and risk-of-bias appraisal using the RoB 2 tool. Of 56 records, nine trials met eligibility and six, comprising 485 participants, provided extractable data. Berberine reduced insulin resistance versus control by a mean difference of −0.95 (95% confidence interval −1.29 to −0.61), with substantial heterogeneity. The effect was larger and homogeneous in polycystic ovary syndrome and smaller in the metabolic and prediabetic subgroup. The prediction interval crossed the null. Five of six pooled trials carried some concerns and the single low-risk trial, which was the largest, showed no effect. In conclusion, berberine was associated with a reduction in insulin resistance, most consistently in polycystic ovary syndrome, but the pooled effect was driven by smaller trials at higher risk of bias. Larger and methodologically rigorous trials are needed before any clinical recommendation. Keywords: Berberine; Insulin resistance; Prediabetic state; Polycystic ovary syndrome; Meta-analysis as topic. Resumo A berberina é um alcaloide vegetal com propriedades sensibilizadoras à insulina. Metanálises recentes abordaram desfechos glicêmicos e componentes da síndrome metabólica, mas nenhuma sintetizou seu efeito sobre a resistência à insulina medida pelo modelo de avaliação da homeostase em populações não diabéticas. O objetivo foi estimar esse efeito em adultos não diabéticos. Buscou-se no PubMed ensaios clínicos randomizados que comparassem berberina, em qualquer formulação ou dose, com controle sem berberina ou comparador ativo, em adultos não diabéticos com resistência à insulina ou risco metabólico. O desfecho primário foi a diferença de médias no modelo de avaliação da homeostase. Ajustou-se modelo de efeitos aleatórios por máxima verossimilhança restrita, com análises de subgrupo, exclusão sequencial e risco de viés pela ferramenta RoB 2. De 56 registros, nove ensaios foram elegíveis e seis, com 485 participantes, forneceram dados extraíveis. A berberina reduziu a resistência à insulina, com diferença de médias de −0,95 (intervalo de confiança de 95%: −1,29 a −0,61) e heterogeneidade substancial. O efeito foi maior e homogêneo na síndrome dos ovários policísticos e menor no subgrupo metabólico e pré-diabético. O intervalo de predição cruzou a nulidade. Cinco dos seis ensaios suscitaram algumas preocupações e o único de baixo risco, o maior 1 2 Clínicas Expert, Itajaí, Santa Catarina, Brazil. Hospital IGESP, São Paulo, São Paulo, Brazil. 1 Research, Society and Development, v. 15, n. 9, e3115951680, 2026 (CC BY 4.0) | ISSN 2525-3409 | DOI: http://dx.doi.org/10.33448/rsd-v15i9.51680 deles, não mostrou efeito. Conclui-se que a berberina associou-se à redução da resistência à insulina, sobretudo na síndrome dos ovários policísticos, mas o efeito agrupado foi determinado por ensaios menores e de maior risco de viés. São necessários ensaios maiores e rigorosos antes de qualquer recomendação clínica. Palavras-chave: Berberina; Resistência à insulina; Estado pré-diabético; Síndrome do ovário policístico; Metanálise como assunto. Resumen La berberina es un alcaloide vegetal con propiedades sensibilizadoras a la insulina. Los metaanálisis recientes han abordado desenlaces glucémicos y componentes del síndrome metabólico, pero ninguno ha sintetizado su efecto sobre la resistencia a la insulina medida por el modelo de evaluación de la homeostasis en poblaciones no diabéticas. El objetivo fue estimar ese efecto en adultos no diabéticos. Se buscó en PubMed ensayos clínicos aleatorizados que compararan berberina, en cualquier formulación o dosis, con un control sin berberina o un comparador activo, en adultos no diabéticos con resistencia a la insulina o riesgo metabólico. El desenlace primario fue la diferencia de medias en dicho modelo. Se ajustó un modelo de efectos aleatorios por máxima verosimilitud restringida, con análisis de subgrupos, exclusión secuencial y riesgo de sesgo mediante RoB 2. De 56 registros, nueve ensayos fueron elegibles y seis, con 485 participantes, aportaron datos extraíbles. La berberina redujo la resistencia a la insulina, con una diferencia de medias de −0,95 (intervalo de confianza del 95%: −1,29 a −0,61) y heterogeneidad sustancial. El efecto fue mayor y homogéneo en el síndrome de ovario poliquístico y menor en el subgrupo metabólico y prediabético. El intervalo de predicción cruzó la nulidad. Cinco de los seis ensayos presentaron algunas preocupaciones y el único de bajo riesgo, el mayor, no mostró efecto. Se concluye que la berberina se asoció a una reducción de la resistencia a la insulina, sobre todo en el síndrome de ovario poliquístico, pero el efecto agrupado estuvo determinado por ensayos menores y de mayor riesgo de sesgo. Palabras clave: Berberina; Resistencia a la insulina; Estado prediabético; Síndrome del ovario poliquístico; Metaanálisis como asunto. 1. Introduction Insulin resistance is a central mechanism linking obesity, the metabolic syndrome, prediabetes, polycystic ovary syndrome and non-alcoholic fatty liver disease, and it precedes the onset of type 2 diabetes (Tabák et al., 2012). Interventions that improve insulin sensitivity before diabetes develops are therefore of clinical interest for prevention. Berberine, an isoquinoline alkaloid extracted from plants such as Coptis chinensis and Berberis species, activates AMP-activated protein kinase and has demonstrated glucose-lowering and lipid-lowering effects (Lee et al., 2006; Xu et al., 2021). It has been proposed as a natural insulin sensitizer, and early randomized trials in type 2 diabetes reported reductions in glycemia comparable to those of metformin (Yin et al., 2008; Zhang et al., 2008). Several meta-analyses have evaluated berberine for glycemic control and for components of the metabolic syndrome. The most recent and comprehensive of these (Liu et al., 2025) pooled triglycerides, fasting plasma glucose, waist circumference, lipids, body-mass index and the two-hour oral glucose tolerance test, but did not synthesize the homeostatic model assessment of insulin resistance (Matthews et al., 1985). No meta-analysis has specifically quantified the effect of berberine on insulin resistance restricted to non-diabetic adults, a prevention-oriented question distinct from the glycemiccontrol question studied in type 2 diabetes. The objective was to estimate that effect in non-diabetic adults and to characterize its consistency across non-diabetic populations. 2. Methodology 2.1 Design and eligibility criteria This is a bibliographic study (Snyder, 2019), specifically classified as a systematic literature review with quantitative synthesis. Following the classification of research approaches proposed by Gil (2026) and by Risemberg et al. (2026), the study adopts a quantitative approach: the corpus was assembled through the systematic selection of randomized controlled 2 Research, Society and Development, v. 15, n. 9, e3115951680, 2026 (CC BY 4.0) | ISSN 2525-3409 | DOI: http://dx.doi.org/10.33448/rsd-v15i9.51680 trials, and the synthesis of their results was carried out by statistical meta-analysis, without a qualitative component in the treatment of the data. This review followed the PRISMA 2020 guideline (Page et al., 2021). Eligible studies were randomized controlled trials, parallel or crossover, enrolling non-diabetic adults aged 18 years or over with insulin resistance or metabolic risk, namely metabolic syndrome, prediabetes or impaired fasting glucose, polycystic ovary syndrome diagnosed by contemporary criteria (Teede et al., 2024), or non-alcoholic fatty liver disease. Trials had to compare berberine, in any salt or formulation, including dihydroberberine and phospholipid or phytosome preparations, at any dose, alone or as an add-on, against a nonberberine control such as placebo, no treatment, usual care or lifestyle, or against an active comparator such as metformin, where berberine was the only component differing between the compared arms, and had to report the homeostatic model assessment of insulin resistance. Studies enrolling populations with established type 2 or type 1 diabetes as the target condition, non-randomized designs, berberine given only within a fixed multi-component formulation, and studies without an extractable insulinresistance outcome were excluded. 2.2 Search and study selection PubMed was searched from inception to 2024, combining terms for berberine, insulin resistance and metabolic risk with a randomized-trial filter, restricted to human studies. Titles and abstracts were screened with the assistance of a large language model; potentially eligible reports underwent independent full-text assessment. Reference lists of relevant reviews, including a meta-analysis on berberine and metabolic-syndrome components (Liu et al., 2025), were checked for additional trials. 2.3 Data extraction and risk of bias Two reviewers independently extracted, for the berberine and control arms, the sample size and the homeostatic model assessment of insulin resistance at baseline and at the end of the intervention, or the change from baseline where reported, together with the population, the berberine formulation and dose, the duration of the intervention and the comparator. Risk of bias was assessed with the Cochrane RoB 2 tool across five domains with an overall judgement (Sterne et al., 2019). Discrepancies were resolved by consensus. 2.4 Statistical analysis The primary outcome was the mean difference in the homeostatic model assessment of insulin resistance, calculated as berberine minus control, so that negative values favour berberine. Endpoint values were pooled together with change-frombaseline values, which estimate the same intervention effect under randomization. Change scores were used where a baseline imbalance was present (Rondanelli et al., 2023) or where only change was reported (Ming et al., 2021; Yan et al., 2015). Where only a 95% confidence interval was reported, the standard error was derived from its width; where only baseline and final values were available for a change analysis, the change-score standard deviation was imputed assuming a pre-post correlation of 0.5. A random-effects model was fitted by restricted maximum likelihood using the metafor package (Viechtbauer, 2010). Heterogeneity was quantified with the inconsistency statistic and with tau squared, a 95% prediction interval was computed as recommended for random-effects syntheses (IntHout et al., 2016), and pre-specified subgroup analyses, comparing polycystic ovary syndrome with metabolic and prediabetic populations, and leave-one-out sensitivity 3 Research, Society and Development, v. 15, n. 9, e3115951680, 2026 (CC BY 4.0) | ISSN 2525-3409 | DOI: http://dx.doi.org/10.33448/rsd-v15i9.51680 analyses were conducted. Small-study effects were not formally tested, since graphical and regression-based methods are unreliable with fewer than ten studies (Egger et al., 1997). 2.5 Role of artificial intelligence The literature search and the initial title and abstract screening were assisted by a large language model. All subsequent steps, namely full-text eligibility, data extraction and risk-of-bias assessment, were conducted independently by the two authors, who reconciled any differences by consensus and are responsible for the data reported. 2.6 Ethical considerations This study is a systematic review based exclusively on aggregate data from previously published studies, without access to individual participant data and without direct involvement of human subjects. It therefore did not require appraisal by a Research Ethics Committee. 3. Results 3.1 Study selection The search returned 56 records. At title and abstract screening, 22 records were excluded and 34 reports underwent full-text assessment, of which 25 were excluded: no extractable insulin-resistance outcome (12), berberine not isolable from a co-intervention (8), review or protocol (2), ineligible design (2), and diabetic population (1). Nine randomized controlled trials were included in the systematic review. Six provided extractable data and were meta-analysed; three were included in the narrative synthesis only, namely Pérez-Rubio et al. (2013), in which insulin sensitivity was reported as the Matsuda index, Pu et al. (2021), available as an abstract only and without extractable per-arm data, and Nikseresht et al. (2024), in which the outcome was reported only graphically. The selection process is summarized in Figure 1. Figure 1 — Flow diagram of study selection according to PRISMA 2020. Source: Prepared by the authors (2026). The diagram summarizes the number of records identified, screened, assessed in full text and included, with the reasons for exclusion at the full-text stage. 4 Research, Society and Development, v. 15, n. 9, e3115951680, 2026 (CC BY 4.0) | ISSN 2525-3409 | DOI: http://dx.doi.org/10.33448/rsd-v15i9.51680 3.2 Study characteristics The six pooled trials, comprising 485 participants, were conducted in China, India and Italy between 2012 and 2023, in polycystic ovary syndrome (two trials), prediabetes or impaired fasting glucose (two), non-alcoholic fatty liver disease (one) and newly diagnosed non-diabetic hyperglycemia (one). Berberine doses ranged from 1.0 to 1.5 g per day, or a bioavailabilityenhanced phospholipid preparation, over 8 to 16 weeks. Five trials used a placebo control and one a lifestyle control. Their characteristics are presented in Table 1. Table 1 — Characteristics of the six pooled randomized controlled trials. Country Population n (B/C) Berberine and dose Weeks Comparator Wei et al. (2012) Study China Polycystic ovary syndrome 31/28 Berberine hydrochloride 500 mg three times daily 12 Placebo Some concerns Risk of bias An et al. (2014) China Polycystic ovary syndrome (in vitro fertilization) 44/43 Berberine 500 mg three times daily 12 Placebo Some concerns Yan et al. (2015) China Non-alcoholic fatty liver disease 55/53 Berberine 0.5 g three times daily 16 Lifestyle Some concerns Ming et al. (2021) China Non-diabetic hyperglycemia 49/99 Berberine 0.5 g twice daily 16 Placebo Low Panigrahi & Mohanty (2023) India Prediabetes 17/17 Berberine 500 mg three times daily 12 Placebo Some concerns Rondanelli et al. (2023) Italy Impaired fasting glucose, overweight 24/25 Berberine phospholipid 550 mg twice daily 8 Placebo Some concerns Source: Prepared by the authors (2026). B, berberine arm; C, control arm. The reader should note that only one trial was judged at low risk of bias and that it was also the largest of the series. 3.3 Effect of berberine on insulin resistance Berberine significantly reduced the homeostatic model assessment of insulin resistance compared with control, with a pooled mean difference of −0.95 (95% confidence interval −1.29 to −0.61; p < 0.001; six trials, 485 participants). Individual estimates ranged from −1.30 to +0.16: Wei et al. (2012) −1.21; An et al. (2014) −1.30; Yan et al. (2015) −0.80; Panigrahi and Mohanty (2023) −0.99; Rondanelli et al. (2023) −0.91; and Ming et al. (2021) +0.16. Heterogeneity was substantial, with an inconsistency of 72% and tau squared of 0.10, and the 95% prediction interval, from −2.12 to +0.21, crossed the null, indicating that a future trial could plausibly show no effect. 3.4 Subgroup and sensitivity analyses The effect was concentrated in trials of polycystic ovary syndrome, with a mean difference of −1.27 (95% confidence interval −1.52 to −1.01; two trials; inconsistency 0%), and was smaller and heterogeneous in the metabolic and prediabetic subgroup, with a mean difference of −0.69 (95% confidence interval −1.22 to −0.16; four trials; inconsistency 63%). The subgroup difference was not statistically significant (p = 0.093). In leave-one-out analysis the pooled estimate remained between −0.84 and −1.08 and statistically significant throughout. Omitting the null trial (Ming et al., 2021) tightened the estimate to −1.08 and reduced the inconsistency to 24%, identifying it as the principal source of heterogeneity. 3.5 Risk of bias Among the six pooled trials, five were judged to have some concerns and one, Ming et al. (2021), which was the largest with 148 participants, was judged at low risk. The domain-level judgements are shown in Table 2. The three narrative- 5 Research, Society and Development, v. 15, n. 9, e3115951680, 2026 (CC BY 4.0) | ISSN 2525-3409 | DOI: http://dx.doi.org/10.33448/rsd-v15i9.51680 only studies were all judged at high risk. Notably, the single low-risk trial was the one that found no effect on insulin resistance. Table 2 — Risk of bias of the pooled trials according to the Cochrane RoB 2 tool. Study D1 D2 D3 D4 D5 Overall Wei et al. (2012) Low Some concerns Low Low Low Some concerns An et al. (2014) Low Low Some concerns Low Low Some concerns Yan et al. (2015) Some concerns Some concerns Some concerns Low Low Some concerns Low Low Low Low Low Low Some concerns Low Low Low Low Some concerns Low Low Some concerns Low Low Some concerns Ming et al. (2021) Panigrahi & Mohanty (2023) Rondanelli et al. (2023) Source: Prepared by the authors (2026). D1, randomization process; D2, deviations from intended interventions; D3, missing outcome data; D4, measurement of the outcome; D5, selection of the reported result. The reader should observe that the only overall judgement of low risk corresponds to the trial reporting no effect. 3.6 Studies not pooled Pérez-Rubio et al. (2013), in metabolic syndrome with 24 participants, reported improved insulin sensitivity by the Matsuda index but did not report the homeostatic model assessment. Pu et al. (2021), in olanzapine-associated metabolic disturbance with 132 participants, reported a significant reduction with berberine in the abstract, without extractable per-arm data. Nikseresht et al. (2024), in prediabetic men in a six-arm design combining high-intensity interval training with or without berberine and 54 participants, reported that berberine alone reduced insulin resistance versus untreated control but added no benefit over exercise; values were presented only graphically. All three were directionally consistent with the pooled result but could not contribute numeric data. A currency check performed at revision confirmed that recent randomized trials of berberine in non-diabetic populations, such as Koperska et al. (2025) in metabolic dysfunction-associated fatty liver disease, do not report the homeostatic model assessment of insulin resistance and therefore did not alter the pooled estimate. 4. Discussion 4.1 Main findings In this updated meta-analysis restricted to non-diabetic adults, berberine was associated with a clinically meaningful reduction in the homeostatic model assessment of insulin resistance, with a mean difference of −0.95, most consistently in polycystic ovary syndrome, where two trials agreed closely with an inconsistency of 0%. This is compatible with the proposed action of berberine as an insulin sensitizer mediated by AMP-activated protein kinase (Lee et al., 2006) and with the direction of the three studies that could not be pooled. 4.2 Interpretation and the weight of the evidence The result must be read against its internal structure. Five of the six pooled trials carried some concerns, and the large reductions came predominantly from smaller trials. The single trial at low risk of bias, which was also the largest with 148 participants (Ming et al., 2021), found no effect and was the dominant source of heterogeneity. Removing it left the effect intact, but its inclusion is precisely what a cautious reader should weigh. The prediction interval crossed the null, which means that the interval within which the effect of a future comparable trial is expected to fall includes no benefit at all (IntHout et al., 6 Research, Society and Development, v. 15, n. 9, e3115951680, 2026 (CC BY 4.0) | ISSN 2525-3409 | DOI: http://dx.doi.org/10.33448/rsd-v15i9.51680 2016). Taken together, the evidence supports a probable and population-dependent benefit rather than an established one, and does not yet warrant a clinical recommendation. 4.3 Relation to previous reviews Prior meta-analyses of berberine addressed type 2 diabetes, glycemic traits or metabolic-syndrome components (Xu et al., 2021; Yin et al., 2008; Zhang et al., 2008). The most recent and comprehensive (Liu et al., 2025), which searched multiple databases including Chinese-language sources, meta-analysed lipids, fasting glucose, anthropometry and the two-hour glucose tolerance test, but did not pool the homeostatic model assessment of insulin resistance. The present review complements that work by isolating the insulin-resistance question in a non-diabetic population, and by incorporating trials of polycystic ovary syndrome, non-alcoholic fatty liver disease and prediabetes that a scope restricted to metabolic syndrome would exclude. An overview of systematic reviews of berberine across health outcomes reached a comparable conclusion regarding the fragility of the underlying evidence base, noting that most syntheses rest on small trials of limited methodological quality (Shi et al., 2025). 4.4 Limitations Only six trials with 485 participants contributed to the primary analysis, several were small, and heterogeneity was substantial. The pooled populations, berberine formulations, comparators and durations differed, and endpoint and changefrom-baseline values were combined. Two variance inputs relied on assumptions, namely a conversion from confidence interval to standard error (Yan et al., 2015) and a pre-post correlation of 0.5 (Rondanelli et al., 2023), and one influential trial reported an unusually small standard deviation (Panigrahi & Mohanty, 2023); these are flagged for confirmation against source. The unequal arm sizes in Ming et al. (2021), of 49 versus 99, reflect a four-arm design, in which berberine alone and berberine combined with Bifidobacterium were each allocated 49 participants against a placebo arm of 99, and are therefore a feature of the allocation rather than a reporting error. The search was limited to PubMed, so non-English trials reporting the outcome, for example in Chinese databases, may have been missed, although cross-checking a multi-database review (Liu et al., 2025) yielded no additional poolable trial. Publication bias could not be formally assessed with six studies (Egger et al., 1997). 5. Conclusion In non-diabetic adults, berberine was associated with a reduction in the homeostatic model assessment of insulin resistance, most consistently in polycystic ovary syndrome. Because the effect was driven by smaller trials at higher risk of bias while the only low-risk trial did not confirm it, and because the prediction interval crossed the null, the finding should be regarded as promising but not definitive. Larger, adequately powered trials at low risk of bias reporting this outcome are needed to establish whether berberine meaningfully improves insulin resistance before the onset of diabetes.