A Systematic Review With Targeted Meta-Analyses of Curcumin and Berberine In Vitro Cytotoxicity Models (2014-2026).

Introduction and background
Traditional Asian medical systems, particularly siddha, ayurveda, and traditional Chinese medicine (TCM), have long used botanical, mineral, and herbo-mineral preparations for chronic disease management, including conditions now recognized as malignant diseases. These systems remain scientifically relevant because natural products have historically contributed disproportionately to anticancer drug discovery, and modern integrative oncology increasingly examines traditional therapies as potential sources of bioactive compounds and mechanistic leads [1]. Despite the rapid growth of the preclinical evidence base, interpretation remains difficult because studies often differ substantially in experimental design, including model systems, exposure durations, assay chemistries, and reporting completeness [2,3]. This heterogeneity limits cross-study comparability and frequently undermines broad quantitative synthesis.
Two compounds commonly studied across these traditions are curcumin and berberine, both frequently evaluated in in vitro cytotoxicity models and therefore serving as pragmatic candidates for targeted pooling under tightly matched experimental conditions [4-9]. Several additional challenges constrain translation. For curcumin, poor systemic bioavailability and chemical instability are major barriers, motivating delivery and formulation strategies [10,11]. Berberine similarly exhibits low oral bioavailability and complex pharmacokinetics that complicate extrapolation from in vitro potency to clinically achievable exposures [12]. Moreover, complex siddha and ayurvedic herbo-mineral and polyherbal formulations introduce additional variability through differences in sourcing, processing, and standardization; recent reports underscore the need for rigorous chemical characterization and reproducible reporting in such studies [13,14]. The aim of this review is twofold: (1) to systematically identify and narratively synthesize preclinical evidence (2014-2026) regarding anticancer activity from siddha, ayurveda, TCM, and related traditional systems; and (2) to conduct targeted meta-analyses restricted to narrowly comparable experimental strata defined by cell line, exposure duration, and MTT-like viability assays to generate defensible quantitative estimates of in vitro cytotoxic potency for curcumin and berberine [2-9].

Review
Methodology
This systematic review was prepared in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) 2020 statement and the PRISMA extension for search reporting (PRISMA-S) [2,3]. Because preclinical systematic reviews are not routinely eligible for clinical trial registries, we adopted a transparency approach analogous to the Clinical Trials Registry-India (CTRI). We preregistered the protocol and analysis plan in an open platform (Open Science Framework; Center for Open Science, Charlottesville, VA, USA); shared extracted data tables (half-maximal inhibitory concentration (IC50) values, assay conditions, and model metadata) and analysis scripts; and reported protocol deviations explicitly. Study identification, screening, eligibility assessment, and inclusion are summarized in the PRISMA flow diagram shown in Figure 1.
Eligibility Criteria
We included 15 preclinical studies published between January 2014 and January 2026 that evaluated the anticancer activity of siddha, ayurveda, TCM, or other traditional systems using either in vitro cancer cell line models with viability or proliferation endpoints and/or in vivo rodent tumor models with tumor burden or survival endpoints. For quantitative synthesis, studies were additionally required to match a prespecified pooling stratum (cell line, 48-hour exposure, and MTT-like viability assay) and to report an extractable IC50 value for the compound of interest.
Information Sources and Search Strategy
Searches were conducted on PubMed and PubMed Central and supplemented with structured web searches to locate full texts, using combinations of controlled vocabulary and the following keywords: system of medicine (siddha OR ayurveda OR “traditional chinese medicine” OR TCM), compound or formulation (curcumin, berberine, rasathailam, vaalai rasa mezhugu, rasagandhi mezhugu, V2S2), and model and endpoint (MCF-7, HepG2, IC50, MTT, cytotoxicity, apoptosis). Reference lists of eligible articles and relevant reviews were screened for additional studies [2,3].
Study Selection and Data Extraction
Two independent reviewers screened titles and abstracts, followed by full texts against eligibility criteria. Conflicts (include/exclude/unclear) were resolved by discussion and consensus. From each included study, we extracted the model (cell line or animal strain), exposure duration, assay type, IC50 (and variance if reported), and mechanistic endpoints (e.g., caspase activation, B-cell lymphoma 2 (Bcl-2) family changes, epithelial-mesenchymal transition (EMT) markers, microRNA (miRNA) changes). When IC50 values were only available in figures, values were extracted when numerically stated in figure labels or captions. Study selection is summarized in Figure 1. Targeted quantitative synthesis was feasible for two tightly matched strata: curcumin in Michigan Cancer Foundation-7 (MCF-7) breast cancer cells at 48 hours with MTT-like assays (three studies) and berberine in HepG2 hepatocellular carcinoma cells at 48 hours with MTT-like assays (two studies) [4-8]. 
Risk of Bias (RoB) and Reporting Quality
For in vivo studies, reporting quality was assessed using the Animal Research: Reporting of In Vivo Experiments (ARRIVE) 2.0 guidance, and RoB was considered using the Systematic Review Centre for Laboratory Animal Experimentation (SYRCLE) risk-of-bias tool where applicable (Table 1) [15,16]. For in vitro studies, key reproducibility variables (cell line authentication, mycoplasma testing, solvent controls, and replicate structure) were recorded (Table 2), recognizing that validated RoB tools for in vitro cytotoxicity are not standardized. The summarized data obtained from the 15 included studies are shown in Table 3. 
Statistical Analysis
The IC50 values were analyzed on the natural logarithmic scale to reflect their approximate log-normal distribution. Random-effects meta-analyses were performed using restricted maximum likelihood (REML) estimation of the between-study variance (tau-squared (τ²)). Results are reported as pooled geometric mean IC50 values with 95% CI and 95% prediction intervals, back-transformed to micromolar (µM). Heterogeneity was summarized using the I-squared statistic (I²). When studies did not report uncertainty for IC50, a conservative within-study coefficient of variation of 20% was assumed (sensitivity range: 10% to 30%). Statistical analyses were performed using R, version 4.3.1 (R Foundation for Statistical Computing, Vienna, AUT). Meta-analyses were conducted using the Metafor package, version 4.2-0 (Wolfgang Viechtbauer, Maastricht, NLD).
Results
Targeted Meta-Analysis of Curcumin MCF-7 48-Hour MTT-Like Assays
Three studies [4-6] reported IC50 values for curcumin in MCF-7 cells at 48 hours using MTT-like viability assays. The IC50 values ranged from 11.21 to 53.18 µM. The pooled geometric mean IC50 was 22.85 µM (95% CI: 11.04-47.27). Between-study heterogeneity was high (I² = 93.54%), with a wide 95% prediction interval (5.63-92.67 µM), indicating that experimental context and/or curcumin preparation likely modify observed potency. The curcumin-targeted meta-analysis is shown in Table 4. A forest plot of curcumin IC50 values in MCF-7 cells at 48 hours is shown in Figure 2. Representative mechanistic evidence includes berberine-mediated NF-kappaB signaling [18]. 
Targeted Meta-Analysis of Berberine in HepG2 48-Hour MTT-Like Assays
Two studies [7,8] reported IC50 values for berberine in HepG2 cells at 48 hours using MTT-like assays and reported results in molar concentration or convertible units. The IC50 values were 25.01 µM and 14.87 µg/mL (converted to 39.99 µM assuming berberine chloride, MW 371.8 g/mol). The pooled geometric mean IC50 was 31.63 µM (95% CI: 22.84-43.79), with moderate heterogeneity (I² = 63.70%). The berberine-targeted meta-analysis is shown in Table 5. The forest plot of berberine IC50 values in HepG2 cells at 48 hours is shown in Figure 3.
Evidence Map Across Traditional Systems
Beyond single-compound exemplars, the broader preclinical literature (Table 6) includes siddha herbo-mineral formulations, ayurvedic polyherbal extracts, and TCM herbs and formulae. However, formulation complexity and variable standardization complicate cross-study comparability. Representative examples identified in this review are summarized in Table 7.
Discussion
This review applied a deliberately restrictive quantitative synthesis strategy to address a core obstacle in preclinical systematic reviews, i.e., extreme methodological heterogeneity [2,3]. When pooling was constrained to closely comparable experimental strata, both curcumin and berberine demonstrated reproducible in vitro cytotoxicity in widely used cancer models, with pooled IC50 values in the tens of micromolar range at 48 hours. For curcumin, the pooled geometric mean IC50 was 22.85 µM but with very high heterogeneity (I² = 93.54%) [4-6]. This level of heterogeneity suggests that curcumin preparation, assay execution, and experimental context likely influence observed potency. Mechanistic endpoints within the included studies implicate apoptosis-related pathways and regulation of anti-apoptotic signaling, including altered myeloid cell leukemia 1 (Mcl-1) expression and miRNA-associated effects [4,6]. For berberine, the pooled geometric mean IC50 was 31.63 µM with moderate heterogeneity (I² = 63.70%) [7,8]. Reported mechanisms commonly implicated EMT-linked signaling, migration, and invasion phenotypes, and TGF-β/Smad pathway involvement [7,8]. Additional studies outside the strict pooling stratum provide a supportive mechanistic context, including NF-κB-linked apoptosis signaling and AMPK-associated cell death signaling in HepG2 models [9,17].
Across traditional systems, evidence from siddha and ayurveda includes complex herbo-mineral and polyherbal formulations with reported cytotoxic and apoptotic signatures in vitro; however, reproducibility depends heavily on chemical characterization and standardization, particularly when mineral-containing preparations are used [13,14]. Ayurvedic polyherbal preparations have been evaluated in carcinoma cell models [19], and pharmaceutically viable polyherbal mixtures have been studied in NSCLC settings [20]. In vivo evidence exists in related traditional medicine investigations, such as DAL models, but integration with in vitro potency estimates requires careful interpretation due to pharmacokinetic and exposure constraints [17].
Bioavailability remains a major bottleneck for curcumin, and delivery strategies are central to any translational argument [10,11]. Berberine’s pharmacokinetics similarly complicate translation from in vitro potency to clinically achievable tissue concentrations [12]. These realities reinforce the importance of aligning preclinical experimental exposures with physiologically plausible ranges, implementing chemical fingerprinting (e.g., high-performance liquid chromatography (HPLC)), and reporting formulation details rigorously.
Limitations
This review has several limitations. First, the targeted meta-analyses were restricted to small numbers of studies due to strict matching criteria; results should therefore be interpreted as hypothesis-generating rather than definitive. Second, because many IC50 reports did not provide variance estimates, a within-study coefficient of variation was assumed for meta-analytic weighting. Although sensitivity analyses did not materially change pooled point estimates, confidence intervals depend on this assumption. Third, the evidence mapping across siddha and some polyherbal formulations included heterogeneous sources with variable reporting quality, underscoring the need for standardized reporting and chemical characterization rather than serving as strong confirmation of efficacy [13-16].

Conclusions
When quantitative synthesis is restricted to carefully matched experimental conditions, curcumin and berberine demonstrate consistent in vitro anticancer activity in widely used models (MCF-7 and HepG2), with pooled IC50 values in the tens of micromolar range at 48 hours. Nevertheless, methodological heterogeneity, incomplete reporting, and translational constraints, particularly bioavailability and standardization, remain major barriers to robust synthesis and clinical inference. Future preclinical investigations should adopt standardized reporting practices, comprehensive chemical characterization, and transparent data sharing to enable credible evidence synthesis and to support rational development within integrative oncology.