This study adhered to the PRISMA guidelines for reporting systematic reviews and was prospectively registered with PROSPERO (https://www.crd.york.ac.uk/PROSPERO/#myprospero) (ID: CRD42023464661).

PubMed, Embase, Cochrane Library, Web of Science, CNKI, Wanfang, VIP, and SinoMed databases were systematically searched from their inception until 03 September 2023. In addition, a manual search of the references of the included studies was performed. The search strategy was designed based on the following criteria: (1) study population: animal models of PCa and (2) intervention: curcumin. Table 1 presents the search strategy used for the PubMed database.

(1) Studies with incomplete or unanalyzable data; (2) cell tests, reviews, abstracts, letters, plans, or in vitro studies; (3) curcumin was used in the control group; and (4) curcumin was not the main component of the intervention in the experimental group.

Two researchers (Wang and Zhang) independently used the literature management software EndnoteX9 to screen literature. First, duplicate literature is excluded by searching for duplicate literature. Titles and abstracts were read according to the inclusion and exclusion criteria. If the PICOS met the inclusion criteria, the study was included. Documents were excluded if the exclusion criteria were met. Finally, we read the full text and judge whether the literature meets the inclusion criteria, and if it does, we include it. After screening was completed, the two participants were compared, and if the results were the same, the included study was finally determined. Otherwise, a third researcher (Jing Chen) was consulted or the decision was discussed.

Two researchers (Wang and Zhang) designed the data extraction tables based on the information required for the research. After reaching a mutual agreement, the necessary information was filtered. Two studies independently extracted data from articles or graphs. The extracted information included the first author, year, sample size, age, intervention, study duration, dose, and outcome indicators. If the data could not be extracted, the author was contacted via email for missing or additional data. Finally, the extracted data were entered into EXCEL. If the extracted outcome index data were continuous, they were uniformly expressed as means and SD. If the outcome index data were binary, they were expressed as the number of occurrence cases ^® and number of samples (n). If not, they were converted. If the information extracted by the two people was inconsistent, it was first discussed upon. On still being unsolved, it was either further discussed upon or a third researcher was consulted to decide (Chen).

Two researchers (Wang and Zhang) used the SYRCLE Animal Studies Bias Risk Tool (Hooijmans et al., 2014a) to conduct bias risk assessment. SYRCLE’s Risk of Bias tool is an adapted version of the Cochrane risk-of-bias tool used in animal intervention studies. This was an objective assessment of possible biases or confusion in the tools used for the design, conduct, and measurement of the animal experiments. Its suitability for different types and domains of animal experiments covered six areas of bias: selection, implementation, measurement, loss to follow-up, reporting, and others. In each area, the risk of bias was judged to be low (+), unclear (?), or high risk (−). Two researchers independently assessed the risk of bias in the included studies, and if there was a disagreement, a third researcher (Chen) was consulted for resolution.

The meta-analysis was conducted using Revman5.4 and Stata 16.0 software. Relative risk (RR) was employed for dichotomous data, whereas standardized mean difference (SMD) was used for continuous data. The effect estimates were calculated using 95% confidence intervals (CIs). Initially, a heterogeneity test was conducted to address anticipated heterogeneity, with the I2 statistic (Higgins and Thompson, 2002) being employed for this purpose. In light of the absence of uniformity across animal studies, a random-effects model was chosen, and various statistical techniques, such as meta-regression, subgroup analyses, and sensitivity analyses, were employed to investigate the origins of heterogeneity. Funnel plots were generated to examine publication bias, adhering to the Cochrane Handbook’s recommendation to include at least 10 relevant literature sources. Assessment of publication bias involved a visual assessment of funnel plot asymmetry.

A comprehensive search was conducted on 3,332 papers, eliminating 2,005 duplicates. Subsequently, 1,267 papers were excluded based on an evaluation of their titles and abstracts. Upon further examination of the full text, an additional 43 papers were excluded for various reasons, such as the absence of endpoint indicators suitable for inclusion, lack of a control group, incomplete data, inadequate trial plan or abstract, intervention not involving curcumin, and absence of PCa. Ultimately, 17 animal trials were deemed eligible for inclusion. Figure 1 shows the results of the literature review.

Seventeen animal experiments involving the use of mouse transplantation tumor models were included in the analysis. The experimental group consisted of 132 mice, whereas the control group consisted of 131 mice, resulting in a total of 263 mice used in the study. Among the included studies, 13 originated in China (Limin et al., 2007; Qi, 2009; ZHAO et al., 2010; LI et al., 2013; Zhang, 2013; Huang et al., 2015; QI et al., 2015; Yang et al., 2015; Ye, 2015; Roy et al., 2016; Zhao et al., 2018; MAO et al., 2019; Cheng et al., 2020). Additionally, three studies were conducted in the United States (Dorai et al., 2001; Khor et al., 2006; Barve et al., 2008), while one study was conducted in Spain (Fernández-Martínez et al., 2009). Curcumin, obtained from Sigma (USA), was administered to the test group predominantly intraperitoneally, with only a few studies utilizing oral administration. The control group, in contrast, received a placebo intervention. Table 2 summarizes the basic characteristics of the included studies.

Three studies (QI et al., 2015; Roy et al., 2016; MAO et al., 2019) used the random number table method to address selection bias, whereas one study (Dorai et al., 2001) employed randomization to allocate participants into intervention groups. The remaining studies did not provide sufficient information on their approaches to selection bias. Regarding implementation bias, only one study (ZHAO et al., 2010) reported the use of randomization and blinding, whereas the remaining studies did not explicitly state their methods. Blinding was not described in any of the studies, indicating potential measurement bias. The risks of lost visits, reporting, and other biases were relatively low. Figures 2, 3 show the results of the risk of bias assessment.

We performed a meta-analysis of duration and dose. The results showed that the curcumin dose affected the tumor inhibition rate (p < 0.05) (Table 3). Subsequently, we performed a subgroup analysis of the doses. Subgroup analysis showed that the effect of 100 mg/kg curcumin (SMD: 0.83, 95% CI: 0.294.1.37) was better than 50 mg/kg curcumin (SMD: 0.66, 95% CI: 0.21.1.10), which proved that the meta regression results were correct (Table 4).

We performed a subgroup analysis of PCa cell types and doses. The results showed that: (1) ginger had different effects on different PCa cells. The effect of the PC - 3 (SMD: 11.30, 95% CI: 20.07, 2.53) was superior to that of the LNCaP (SMD: 4.41, 95% CI: 11.49, 2.66). The effect of the PC - 3 (SMD: 11.30, 95% CI: 20.07, 2.53) was superior to the LNCaP (SMD: 4.41, 95% CI: 11.49, 2.66). (2) Different doses of curcumin had different effects on outcome indexes. In terms of tumor volume, effect of ranking was 100 mg/kg (SMD: 11.30, 95% CI: 20.07, 2.53) > 50 mg/kg (SMD: 9.13, 95% CI: 14.44, 3.83) > 30 mg/kg (SMD: 5.00, 95% CI: 10.64, 0.63). The effect on tumor weight was 100 mg/kg (SMD: 7.80, 95% CI: 12.13, 3.46) > 50 mg/kg (SMD: 6.30, 95% CI: 9.01, 3.58). The tumor inhibition rate was 100 mg/kg (SMD: 0.83, 95% CI: 0.294, 1.37) to 50 mg/kg (SMD: 0.66, 95% CI: 0.21, 1.10). The results of the subgroup analysis showed a positive correlation between the curcumin dose and results; the higher the dose, the better the effect (Table 4; Figure 9). Contour-weighted funnel plots and cut-and-complement methods showed a low likelihood of publication bias (Figure 9, Table 5).

Sensitivity analysis was performed on the results of the meta-analysis. The following results were obtained: tumor volume (SMD: −3.04, 95% CI: 4.51,-1.57, p = 0.0001), tumor weight (SMD: 3.62, 95% CI: 5.14,-2.09, p = 0.0001), tumor inhibition rate (SMD: 0.90, 95% CI: 0.44, 1.36, p = 0.0001), PCNA mRNA (SMD: 3.45, 95% CI: 5.10, 1.81, p = 0.0001), and MMP2 mRNA (SMD: 3.56, 95% CI: 6.46, 0.67, p = 0.0001). The results of sensitivity analysis showed that the results of the meta-analysis were stable (Table 6).

This study included 17 meta-analyses that aimed to examine the evidence derived from animal studies on the therapeutic efficacy of curcumin in the treatment of PCa. The findings indicated that curcumin exhibited inhibitory effects on PCNA and MMP2 mRNA expression and the growth of malignant tumors in PCa cells of animal models when compared to control groups. The stability of the results was confirmed through a sensitivity analysis. Furthermore, the meta-regression analysis demonstrated that the dosage of curcumin had a significant impact on the rate of tumor inhibition in the animal model. Through subgroup analysis, a positive correlation between the dosage of curcumin and the outcomes was determined, indicating that higher doses of curcumin yielded more favorable effects.

Curcumin exhibits anticancer properties, as demonstrated in a meta-analysis conducted by de Oliveira et al. (de Oliveira et al., 2022). The analysis revealed that curcumin exerts therapeutic effects on various tumor types, including the inhibition of tumor growth (SMD: −3.03; 95% CI: −3.84, −2.21; p < 0.00001), reduction in tumor volume (SMD: −7.30; 95% CI: −11.39,-3.21; p < 0.00001), and decrease in tumor weight (SMD: −3.96; 95% CI: −6.22, - 1.70; p = 0.0006). Furthermore, studies have indicated that curcumin exhibits a greater affinity for PCa cells than for healthy prostate epithelial cells in individuals without the disease, suggesting its potential as a chemopreventive agent against human PCa (Srivastava et al., 2007). However, the existing body of meta-analytical studies on the efficacy of curcumin in PCa remains insufficient. This study, in contrast, revealed significant reductions in both prostate tumor volume and weight, as well as the inhibition of PCa cell growth, following curcumin administration. These effects were attributed to three primary pathways of action. First, curcumin can impede PCa LNCaP cell viability and induce apoptosis (Mukhopadhyay et al., 2001; Nakamura et al., 2002). Curcumin exhibited anti-proliferative properties in LNCaP cells, resulting in a dose-dependent reduction in DNA synthesis efficiency (Mukhopadhyay et al., 2002). Additionally, it downregulated the expression of G1/S-specific cyclin D1, which is frequently overexpressed in various tumor cells and is a crucial target for inhibiting tumor cell proliferation (Bharti et al., 2003; Shankar and Srivastava, 2007; Mahammedi et al., 2016; Yang et al., 2017; Choi et al., 2019; Termini et al., 2020; Ferlay et al.). Curcumin exerted inhibitory effects on the proliferation of PCa cells by downregulating cyclin D1 expression, thereby impeding the transition of tumor cells from the G1 to S phase. Disruption of cell cycle led to apoptosis. In addition, curcumin upregulated Bax, which promotes apoptosis, and downregulated Bcl-2, which inhibits apoptosis in cancer cells. Bcl-2 and Bax expression is closely associated with apoptosis in cancer cells.

In the present study, curcumin exhibited a noteworthy advantage in suppressing PCNA mRNA and MMP2 mRNA in comparison to the control group (p < 0.05). PCNA mRNA, a recently discovered probe for assessing the proliferative activity of PCa cells, reflects the extent of cell proliferation (Squires et al., 2003). PCNA mRNA serves as a robust positive marker of the proliferative activity of PCa cells, thereby emerging as a novel determinant for assessing the biological behavior of PCa. Strong positive expression of PCNA mRNA signifies the presence of PCa cells with heightened proliferative activity, elevated malignancy levels, and an unfavorable prognosis, thus establishing a novel criterion for evaluating the biological behavior of PCa (Park et al., 2005; Prusty and Das, 2005). MMP2 mRNA is a member of the matrix metalloproteinases (MMPs), a group of proteins that have been demonstrated to be associated with tumor metastasis (Dos Reis et al., 2009). Additionally, MMPs play a role in tumor growth regulation by maintaining the integrity of cellular pathways (Dorai et al., 2000a). The expression of MMP2 is minimal in benign prostate hyperplasia tissues, whereas in PCa tissues, its expression level correlates with the degree of proliferation. MMP2 is highly expressed in PCa tissues and plays a crucial role in the metastatic progression of prostate tumors (Dorai et al., 2000b). The findings from the cellular assay demonstrated that curcumin could impede the expression of PCNA and MMP2 in PCa cells, thereby highlighting its distinctive merits in PCa management (Hong et al., 2006; Teiten et al., 2011). However, given the limited number of incorporated studies, further rigorous investigations encompassing both the fundamental and clinical realms are imperative to substantiate the efficacy of curcumin in suppressing PCNA mRNA and MMP2 mRNA.

Currently, the use of curcumin for PCa treatment relies primarily on preliminary investigations and limited clinical studies, resulting in insufficient evidence. However, few published clinical studies have substantiated the efficacy and safety of curcumin in the treatment of PCa. For instance, Mahammedi conducted a study involving 30 patients with desmoplasma-resistant PCa, administering a combination of docetaxel/prednisone and curcumin. The findings of the study conducted by Choi et al. (2019) revealed that a significant reduction in PSA levels was observed in 59% of the participants, with 14% achieving complete normalization. Additionally, 40% of the patients reported experiencing symptom relief. Furthermore, the study demonstrated that the adverse effects were attributable to the administration of docetaxel rather than curcumin, as the latter exhibited no toxic effects. The trial, which was randomized and double-blinded, encompassed a sample of 97 individuals diagnosed with PCa and indicated a statistically significant decrease in the proportion of patients experiencing PSA progression over a 6-month treatment period in the curcumin group compared with that in the placebo group (10.3% vs 30.2%, p = 0.00259). Additionally, the incidence of adverse events was lower in the curcumin group (15.56%) than that in the placebo group (34.78%). Notably, the Clinical Trial Registry platform (https://beta.clinicaltrials.gov) currently hosts numerous ongoing clinical studies on curcumin for PCa, with some already completed or in progress (e.g., NCT02064673, NCT04731844, NCT03211104, NCT03769766, etc.). Increasing attention on curcumin treatment for PCa indicates its potential for broad clinical applications. Consequently, our future focus will be directed towards closely monitoring clinical studies on curcumin for PCa, aiming to gather additional clinical evidence.

Several clinical studies have demonstrated that curcumin dosage plays a significant role in its effectiveness (de Waure et al., 2023). High doses of curcumin, such as 3.6 g/d for 6 months or 8 g/d for 28 days, have been shown to effectively reduce tumor volume and enhance survival rates (Kuriakose et al., 2016; Santosa et al., 2022). These findings are consistent with those of the present study. Furthermore, studies have indicated that the physicochemical properties of curcumin influence its efficacy. Notably, curcumin exhibits poor water solubility and low systemic bioavailability of approximately 0.47% when administered orally (Mirzaei et al., 2017). Encapsulation of curcumin within polymeric nanocarriers has been suggested as a potential solution to improve its physicochemical properties (Klippstein et al., 2015). Polymer nanocapsules, which are characterized by vesicular nanostructures encompassing an oily core encased by a polymer wall, have been extensively studied to enhance curcumin stability, augment its apparent water solubility, improve its bioavailability, and enhance its in vivo anti-inflammatory (Asadirad et al., 2022), neuroprotective (Hoppe et al., 2013), and antitumor effects (Zanotto-Filho et al., 2013). This will significantly enhance the advancement of curcumin-based pharmaceuticals and clinical utilization of curcumin. The findings of this study indicate that curcumin has a more pronounced effect on PC-3 cells than on LNCaP cells. Currently, there is a lack of evidence elucidating the underlying reasons, which is potentially attributable to the limited number of LNCaP cells included in this study (only two studies). Given the limited number of included studies, further research is warranted to validate the effects of curcumin dosage, treatment duration, administration method, PCa cell type, and other relevant factors on its efficacy.

Strengths of this study: This study is the inaugural systematic review of animal studies investigating the effects of curcumin on PCa. The evidence derived from these studies holds immense potential for evaluating the effectiveness and possible clinical applications of curcumin in cancer treatment. Consequently, this evidence can serve as a foundation for the design and implementation of clinical trials as well as the advancement of novel therapeutic interventions.

Limitations of this study: The inability to conduct a meta-analysis of the clinical studies was attributed to the limited number of available trials. This limitation was further compounded by the absence of standardized protocols in the included animal trials, resulting in substantial heterogeneity in terms of study design, disease staging, control group, dose, and regimen. Meta-analysis of animal trials may exhibit even greater heterogeneity than clinical studies (Hooijmans et al., 2014b). Consequently, numerous meta-analyses on animal tests have encountered challenges in addressing this issue (de Oliveira et al., 2022; Luís et al., 2023; Dan et al., 2024). Subgroup analyses have been commonly performed to address heterogeneity. However, owing to the limited number of articles within each subgroup, we conducted a subgroup analysis solely on dosage, precluding the examination of factors, such as the mode of administration, treatment duration, and treatment frequency. The findings from the subgroup analyses indicated that dosage exerted an impact on the results, albeit without a substantial reduction in heterogeneity. The meta-analysis conducted by e Oliveira TV et al. (de Oliveira et al., 2022) examined the impact of curcumin’s mode of administration, dose, and treatment duration on the outcomes reported in various articles. The findings revealed that these factors did not significantly influence the results and there was no substantial reduction in heterogeneity. These findings are consistent with the results of this study. Furthermore, a significant number of experimental studies lack comprehensive reporting of randomization methods, allocation concealment, and randomization of outcome assessments, thereby leading to an indeterminate risk of selection, implementation, and measurement biases. Consequently, there is a pressing need to enhance the methodological rigor of animal testing, given that the findings of such studies frequently inform clinical applications. Petersen et al. (Petersen et al., 2016). Discussed the limitation of interpreting preclinical studies, arguing that this increases the difficulty of evaluating the results of preclinical studies owing to differences in pharmacokinetics and pharmacodynamics between humans and animals. He recommends the creation of consensus guidelines for evaluating the conduct and reporting of preclinical research findings. Meta-analyses of animal experiments should be referenced in the future, which will help improve the quality of evidence in animal experiments.