Introduction

DM is a metabolic disorder characterized by persistent hyperglycemia due to a deficiency in insulin production or its action or in both mechanisms. This chronic ailment has been classified as a troubling disease for the twenty-first century. According to the International Diabetes Federation Global Diabetes Atlas (10th Edition), DM reached about 537 million adults in 2021 was responsible for around 6.7 million deaths, and cost USD 966 billion in health spending in the last 15 years (International Diabetes Federation 2021).

Uncontrolled DM has deleterious consequences and several complications, such as circulatory problems (increased risk of heart disease, stroke, and peripheral blood vessels), kidney diseases, arterial hypertension, and a higher mortality rate when compared with non-diabetic and healthy patients; which can impair affected patients’ quality of life and also reduction or loss of productive capacity (Almutlaq et al. 2021; Bjornstad et al. 2022; de Souza Stork et al. 2022; Shrestha and Ghimire 2012).

It is known that OS, which is a state of imbalance between the production of ROs and endogenous antioxidant capacity; triggers the complications of DM (Reis et al. 2021). The main molecular mechanisms associated with OS in DM are related to glucose and lipid metabolism (Giacco and Brownlee 2010).

Under conditions of hyperglycemia, the excessive production of ROs during glycolysis reactions leads to damage in both DNA and DNA repair enzymes, causing accumulation of glyceraldehyde-3-phosphate content, which is responsible for the activation of other pro-oxidant pathways and the auto-oxidation of glucose. Thus, a boost in the production of hydrogen peroxide, a precursor of oxidizing substances, is observed. Furthermore, the auto-oxidation of glucose leads to the formation of glyoxal, an advanced glycation end-product precursor, which promotes cellular OS (Darenskaya et al. 2021).

Even though the complications of DM are diverse and clinically relevant, the pharmacological treatment options for either T1DM or T2DM are narrowed to a few common approaches. The treatment of T1DM involves the administration of insulin, but there is a chance of developing so-called insulin resistance, manifesting in most cases itself at the cellular level through post-receptor defects in insulin signaling. In another hand, T2DM is treated with oral antihyperglycemic drugs, which are associated with several side effects. Therefore, more patients seek lifestyle modifications combined with natural and safer options like nutraceuticals for the management of DM (Buzzetti et al. 2020; Shi et al. 2019).

Some nutraceuticals, like super fruits and spices exhibiting high anti-radical and anti-inflammatory activity, are sources of copious phytochemicals that can regulate alpha-glucosidase and lipase activities, improve pancreatic function and insulin release and reduce blood glucose levels, boosting the effects of hypoglycemic or antihyperglycemic agents (Alhujaily et al. 2022; Dehdashtian et al. 2020). Thus, they act as antioxidants, mitigating the imbalance of ROs and blocking the synthesis of prostaglandins, pro-inflammatory cytokines, and transcription factors, mainly the NF-κB factor. Consequently, such foodstuffs may be highly effective in the management of DM and its worsening (Alhujaily et al. 2022; Derosa et al. 2022; Zhang et al. 2015).

Rosemary (Rosmarinus officinalis L., Lamiaceae) is an aromatic household spice that has been used worldwide for culinary, food preservative, and medicinal purposes owing to its powerful antioxidant properties (Sánchez-Camargo and Herrero 2017). There is increasing scientific evidence supporting its efficacy in preclinical models in which the OS is involved, such as inflammation (Gonçalves et al. 2022), neurodegeneration (Capatina et al. 2020), cancer (Pérez-Sánchez et al. 2019) and DM (Bao et al. 2020; Hassani et al. 2016; Naimi et al. 2017). Its wide range of bioactivities can be due to the synergism of antioxidant phytochemicals such as diterpenes, triterpenes, phenolic acids, flavonoids, and volatile compounds (Ulbricht et al. 2010; Yashin et al. 2017).

Despite plenty of original papers and narrative reviews reporting the healing properties of R. officinalis, its efficacy in DM and DM complications has not been systematically reviewed and statistically compared so far. Therefore, healthcare professionals across the globe still lack high-level evidence that supports the use of R. officinalis in complementary and integrative medicine as a coadjutant in the management of DM.

Motivated by this scientific and clinical gap, this investigation aimed to provide a systematic review and meta-analysis concerning the effects of R. officinalis in the biochemical outcomes of animals that underwent pre-clinical models of DM. Insights on the overall methodological quality and bias of the included primary studies are also presented. Our findings are consistent with using R. officinalis in traditional medicine for managing DM. Besides, this scientific contribution is a step forward in evidence-based practices related to the rationale use of R. officinalis in primary healthcare and poses this nutraceutical as the target of further clinical investigations associated with DM and its complications.

Material and methods

Study design and guiding question

This systematic review was carried out according to the PRISMA guideline (Page et al. 2021). This study was registered in PROSPERO with protocol number CRD42021250556 (https://www.crd.york.ac.uk/prospero/display_record.php?ID=CRD42021250556). Our guiding question was: Does R. officinalis improve the biochemical outcomes of animals in experimental models of diabetes mellitus?

Sources of information and search strategies

The electronic databases PubMed (MEDLINE), Scopus, ScienceDirect, Web of Science (Science Citation Index), and VHL were used for the search of relevant publications issued until December 31st, 2022. The authors also searched citation reference lists of relevant included studies and any previously published reviews and gray literature to include other studies in Google Scholar, through the D.O.I of the intended study. Thus, all the searches were performed by two independent researchers (V.M.O and L.R.S) on January 13th, 2023.

None of the collection language was restricted during the identification of the studies. The authors of the unavailable articles were contacted twice by e-mail, through which they were requested access to the articles and other information (raw data).

The descriptors were delimited and defined from the MeSH and DeCS tools. The research strategies used in this review were carried out using the PECOS acronym, i.e.: population (P); exposure (E); comparator (C); outcomes (O); and study design (S). The research terms were combined using the Boolean logical terms “AND” and “OR”. Research filters (document type = “article”) for the identification of preclinical studies in the Web of Science and VHL were applied to increase filtering and research efficiency. The research strategies used for each database are detailed in Table S1 of the Supplementary Material.

Eligibility criteria

The eligible and selected articles from the databases were read in full and evaluated according to the inclusion and exclusion criteria predefined from the PROSPERO protocol prioritized in terms of i) type of exposure/intervention, ii) type of population, iii) study design and, iv) type of publication.

Inclusion criteria:

  1. (i)

    treatment with preparations of R. officinalis at any dosage, administered at any time, by any administration route, and any dose frequency;

  2. (ii)

    all animal models of DM (deficiency in insulin production and failure of β cells to ensure construct validity), all sexes, all ages, and all species of animals/strains;

  3. (iii)

    experimental studies with separate control groups; cross-design, randomized, and non-randomized study designs;

  4. (iv)

    original articles and short communications (published or ahead of print) were considered.

Exclusion criteria:

  1. (i)

    combined treatment of R. officinalis L., or isolated pure compounds or R. officinalis combined with standard oral hypoglycemic agents;

  2. (ii)

    studies in humans, animals with any other comorbidity

  3. (iii)

    in vitro and ex vivo study designs, in silico, studies before and after without a control group;

  4. (iv)

    case reports, review articles, editorials, letters to the editor, and articles presented at scientific events, news, comments, dissertations, and thesis were excluded.

After searching the databases, the articles were transferred to the Rayyan QCRI application (Ouzzani et al. 2016) to organize, remove duplicates, and apply the eligibility criteria of each article and thus associate them with eligible or not for inclusion.

Throughout the selection of the articles, each reviewer (VMO, LRS) was blinded about the other’s decisions, and the discrepancies were discussed until a consensus or, if it was not possible, a third author (ROC) was consulted. It is noteworthy that the final decision of the articles to be included is made by consensus among all the researchers involved.

The Kappa coefficient (Landis and Koch 1977) was used to assess the level of agreement between the two authors (VMO, LRS). In this pursuit, we considered a 95% CI and used the Stata 11.0 software package.

Data extraction

The primary and secondary data of the included studies were extracted for a standardized and pre-formatted electronic spreadsheet. Data were collected considering the animal models (species/strains of animals used; sex of animals; number of animals per group; age and weight of animals; type of feeding used); study design (number of experimental groups and duration of follow-up; method of allocation to the treatment group; type of diabetes whether T1DM or T2DM; and blind or non-blind evaluation method); intervention characteristics (taxonomic identification of the herb; voucher number; method of preparation of R. officinalis extract; methods and parameters of quality control; chemical composition; dose; dose frequency; time and route of administration; method of induction of DM and time to begin exposure; parameters to be considered diabetic; whether nephrectomy was performed or not); identification of the study (authors; year of publication; country where the research was conducted); and publishing language.

The main results evaluated were: FPG (mg/dL); HbA1c (%); SIL (mU/L); lipid profile [Tg (mg/dL), HDL-c (mg/dL), LDL-c (mg/dL), and TC (mg/dL); ALP (U/L), ALT (U/L), AST (U/L), and Alb (g/dL) for liver function; BUN (mg/dL), AUR (mg/dL), and CRE (mg/dL) for renal function.

Taxonomic evaluation

Taxonomic and nomenclature accuracy was evaluated by comparing the taxonomic information reported with existing patterns in an open botanical database accessible at https://wfoplantlist.org/plant-list. A classification was then elaborated according to methods proposed elsewhere (Rivera et al. 2014) to evaluate the taxonomy and identification of plant species used.

For studies that included complete information on plant species, identification and proof of the specimen presented, it was qualified as “LEVEL 1”. For studies that did not present information on identification and proof of the specimen, a classification of “LEVEL 2” was then assigned; finally, studies with incomplete information or that did not present any information about the species, collection, identification, and proof of specimens, received a classification of “LEVEL 3” (Rivera et al. 2014).

Data synthesis

The qualitative data of the included studies were described in narrative synthesis and summarized in tables to establish patterns and variations. Statistically, we used signs to indicate significant increment (↑), decrement (↓), or equality ( ↔) effect size measured between treatment and control groups.

Quantitative data were grouped into a statistical meta-analysis using RevMan 5.3 (Copenhagen: The Nordic Cochrane Centre). The meta-analysis was carried out exclusively for two or more studies depicting data on a specific outcome of interest. Herein, given the great diversity of the experimental models, only the most effective dose and the greatest frequency of exposure were considered. As the same outcome could be reported in different measurement scales, we performed mathematic conversions to the same scale. Data of FPG (mg/dL), HbA1c (%), SIL (mU/L), Tg (mg/dL), HDL (mg/dL), LDL (mg/dL), TC (mg/dL), ALP (U/L), ALT (U/L), AST (U/L), Alb (g/dL), BUN (mg/dL), AUR (mg/dL), and CRE (mg/dL) underwent metanalysis. MD and SD were used to assess continuous variables with a 95% CI. The p-value < 0.05 was considered statically significant. For studies reporting, the standard errors of means, and the corresponding SD were calculated by multiplying by the square root of the respective sample size.

The inverse method of weighted variance was used to attribute the relative contribution of each study included to the effect of the grouped standard deviation of R. officinalis and its 95% CI. The random effects model was used for pooling effect estimates because the effect sizes from animal studies probably differ more due to the distinction in design features (Peter et al. 2021).

Heterogeneity assessment

We used the I2 statistic to evaluate the severity of heterogeneity. The I2 ≥ 75% was considered indicative of substantial heterogeneity (Borenstein et al. 2010). Subgroup analysis was carried out to identify potential factors that influence heterogeneity in the FPG and SIL.

Assessment of the risk of bias and quality of the included articles

Two reviewers (VMO and LRS) independently evaluated the risk of bias using SYRCLE’s tool (Hooijmans et al. 2014b). By using this tool, it was possible to evaluate selection bias, performance bias, detection bias, friction bias, reporting bias, and other types of biases.

The ten domains evaluated in this tool are random sequence generation; basic characteristics; allocation concealment; random housing; blinding of participants and staff; evaluation of random results; incomplete results data; reports of selective results; and other sources of bias. The risk of bias was assessed using RevMan 5.3 (Copenhagen: The Nordic Cochrane Centre). Then, two independent researchers (VMO and LRS) evaluated all the studies included through the CAMARADES checklist (Bahor et al. 2021) for study quality. Disagreements were resolved with discussion.

Results

Search results

Figure 1 depicts the PRISMA flowchart with the selection of studies performed in this systematic review. A total of 4909 studies were identified in the electronic databases search. After reviewing the title and abstract, 322 papers were excluded for being duplicates (coming from the two different databases). During the title, keywords, and abstract screening 4876 articles were further discarded. The full text of the 33 potentially relevant articles was assessed and 15 papers were excluded due to the following reasons: not an R. officinalis preparation (n = 6), not a DM model (n = 4), dissertation (n = 1), wrong experimental design (n = 1), in vitro experiments (n = 2), unseparated control group (n = 1).

Fig. 1
figure 1

PRISMA flow diagram of studies selected, included and excluded in the systematic review on the effect of R. officinalis in animal models of diabetes mellitus

Full size image

After carrying out a thorough review of the literature, including both published articles and gray literature, a total of five articles were selected for full-text reading. In addition, two articles were identified as potentially relevant (Erenmemisoglu et al. 1997; Malek et al. 2020) and we contacted the authors requiring full access, but only one was obtained (Malek et al. 2020).

The study comprised 23 articles (Al-badry 2017; Al-Hader et al. 1994; Alnahdi 2012; Ayaz 2012; Bakirel et al. 2008; Belmouhoub et al. 2016, 2017, 2020; Benkhedir et al. 2022; El-Boshy et al. 2015; Emam 2012; Khalil et al. 2012; Koga et al. 2006; Malek et al. 2020; Mohamed 2021; Nazem et al. 2015; Ramadan et al. 2013; Rasoulian et al. 2019; Sebai et al. 2015; Selmi et al. 2017; Shen et al. 2020; Silva et al. 2011; Wahba 2022), of which 18 were eligible for the meta-analysis (Al-Hader et al. 1994; Alnahdi 2012; Ayaz 2012; Bakirel et al. 2008; Benkhedir et al. 2022; El-Boshy et al. 2015; Emam 2012; Khalil et al. 2012; Malek et al. 2020; Mohamed 2021; Nazem et al. 2015; Ramadan et al. 2013; Rasoulian et al. 2019; Sebai et al. 2017; Shen et al. 2020; Silva et al. 2011; Wahba 2022). Throughout the eligibility evaluation, the degree of agreement between the two researchers (Kappa coefficient) was of 0.992, which is considered almost perfect (Landis and Koch 1977).

Description of the included studies

General Features

Table 1 summarizes the experimental design of the studies herein included, Moreover, Table 2 presents the qualitative synthesis of their main findings. The majority of included studies used rats (69.6%; n = 16), while 21.7% (n = 5) used mice and 8.7% (n = 2) used rabbits. A total of 21 studies (91.3%) used only male animals, 1 (4.3%) study used only females and 1 (4.3%) used both sexes. About 56.5% of the studies (n = 13) did not specify the animal’s age, being that the age of the animals ranged from 13 days to 3 months.

Table 1 General features of the studies included in the systematic review on the efficacy of Rosmarinus officinalis in animal models of diabetes mellitus
Full size table
Table 2 Summary of the primary and secondary outcomes of the studies assessing the effects of R. officinalis in animal models of diabetes mellitus
Full size table

About 47.8% (n = 11) of studies used R. officinalis aqueous extract, 34.8% (n = 8) used hydroalcoholic extract, 8.7% (n = 2) used fortified power diet and 8.7% used the volatile oil. A total of 10 studies (43.5%) presented data of qualitative and quantitative analysis of the phytochemicals contained in the extracts.

The doses of R. officinalis administered ranged from 25 to 800 mg/kg (200.7 ± 156.8 mg/kg). Only two studies (8.7%) did not specify the dose (mg/kg) used, and the other 2 studies (8.7%) blended leaves powder in animals’ diets with doses ranging from 0.05% to 10%.

Oral administration was the most frequent among studies (82.6%; n = 19), while 3 studies (13%) used intraperitoneal administration and 1 study (4.3%) used intramuscular administration. The once-a-day regimen of administration was performed in all the included studies (100%).

The histopathology and anatomopathological analysis were assessed in 8 studies (34.8%), which related to nonharmful effects in the evaluated organs. R. officinalis reduced histopathological changes induced by DM on the kidney and liver of the treated animals, also displayed amelioration in the degeneration of Langerhans islets, attenuated hepatic DM-induced alteration in hepatocytes, preserved renal architecture, decreased plasma inflammatory cytokine and enhanced the reproductive performance by reversing semen abnormalities.

Table S2 of the Supplementary Material presents the characteristics of DM induction protocols in the included studies. All of the studies carried out T1DM experimental models, with FPG levels at baseline greater than 200 mg/dL; since the researchers chemically induced the illness by using alloxan or streptozotocin (STZ) (King 2012).

About 39.1% of the studies (n = 9) used alloxan and 60.9% (n = 14) used STZ. The STZ doses used range from 40 to 110 mg/kg (60.8 ± 28.3 mg/kg). In turn, the doses of alloxan used ranged from 120 to 220 mg/kg (146.1 ± 32.0 mg/kg). The time taken to confirm that the diabetes is stabilized after the exposure to induction material was between 2 and 7 days. Only 5 studies (21.7%) reported pancreatectomy, which was used to prevent pancreatic beta regeneration.

The animal’s diet was assessed in 13 studies (56.5%), in which 2 studies (8.7%) reported the amount of nutrients contained in the food. Concerning the exposure time, 5 studies (21.7%) carried out acute follow-up (up to 24 h), while 13 (56.5%) studies performed subacute follow-up (from 1 up to 21 days) and 7 studies (30.4%) performed subchronic follow-up (>22 days). Noteworthy, in some studies more than one follow-up period was investigated depending on the outcome assessed.

Taxonomic evaluation

The full taxonomic analysis of the included studies is presented in Fig. S1 of the Supplementary Material. The majority of the studies (56.5%, n = 13) were classified as LEVEL 1, with higher quality and complete information about the specimens; 8 studies (34.8%) were classified as LEVEL 2 because they did not use an appropriate style to formally differentiate the nomenclature from other text, and/or provide data and evidence of specimens used and/or provided data on how these specimens were identified. Finally, 2 studies (8.7%) were classified as LEVEL 3, because did not fulfill 3 or more specified items.

The effects of R. officinalis in the biochemical parameters of animals with DM

Lowering the FPG, but not the HbA1c levels

As can be seen in Fig. 2A, seventeen preclinical studies (n = 256) were pooled and indicated that R. officinalis derivatives significantly reduced FPG compared with vehicle control groups, representing −120.84 of MD (95% CI; −157.09, −84.59), I2 = 99% (indicating a heterogeneity between the groups). The great majority of the studies (88.2%; n = 15) favored R. officinalis. In turn, Fig. 2B showcases that R. officinalis did not significantly reduce HbA1c as compared to the control groups (2 studies, n = 28; MD = −1.77 [95% CI; −6.16, + 2.61]; I2 = 70%).

Fig. 2
figure 2

Forest plot of preclinical studies of diabetes mellitus comparing R. officinalis and vehicle control; measuring FPG (A), HbA1c (B), and SIL (C)

Full size image

Boosting the insulin release

Figure 2C depicts the metanalysis for SIL. The data from nine studies were pooled and provided evidence that R. officinalis significantly increases the release of insulin (n = 138; MD =  + 3.73 [95% CI; + 3.17, + 4.29]) in comparison to the vehicle control groups. The I2 was 100%, which showcases heterogenicity between the groups. Noteworthy, all studies consistently favored R. officinalis.

Decreasing hepatic damage

As reported in Fig. 3A, B, respectively, data of three studies (n = 38) were pooled and indicated that treating the animals with R. officinalis significantly reduced both ALT (MD = −36.42; [95% CI; −55.69, −17.14]; I2 = 99%) and AST (MD = −24.05; [95% CI; −37.84, −10.27]; I2 = 85%) enzyme levels compared to vehicle control groups. On the other hand, R. officinalis did not significantly decrease the ALP enzyme levels (MD = −13.28; [95% CI; −33.08, +6.52]; I2 = 88%) as shown in Fig. 4C. Moreover, the Alb levels (Fig. 4D) were not significantly affected following the treatment with R. officinalis (3 studies, n = 36; MD =  + 2.00 [95% CI; −0.86, + 4.86]; I2 = 100%).

Fig. 3
figure 3

Forest plot of preclinical studies of diabetes mellitus comparing R. officinalis and vehicle control; measuring ALT (A), AST (B), ALP (C) and Alb (D)

Full size image
Fig. 4
figure 4

Forest plot of preclinical studies of diabetes mellitus comparing R. officinalis and vehicle control; measuring BUN (A), AUR (B), and CRE (C)

Full size image

Slightly recovering the kidney function

Figures 4A–C show the metanalysis for the several biochemical markers herein used to assess the kidney function of the diabetic animals treated with R. officinalis or vehicle. The data of 4 studies (n = 48) were pooled and indicate that R. officinalis significantly decreased both BUN (MD = −24.84 [95% CI; −34.78, −14.90]; I2 = 93%) and CRE (MD = −0.40 [95% CI; –0.74, −0.06]; I2 = 98%) levels compared to vehicle control groups. However, the data of these same studies were pooled and indicate that R. officinalis did not significantly reduce AUR (MD = −0.43 [95% CI; −2.05, + 1.19]; I2 = 100%) level in comparison to the vehicle control group.

Improving the lipid profile

Figure 5A indicates that there was a significant lowering in LDL-c level of diabetic animals receiving R. officinalis compared to the vehicle control group (3 studies, n = 38; MD = −11.74 [95% CI; −21.27, −2.21]; I2 = 92%). Nor the HDL (3 studies, n = 38; MD =  + 3.66 [95% CI; −1.89, + 9.20]; I2 = 93%), neither the TC (4 studies, n = 52; MD = −27.18 [95% CI; −60.21, + 5.86]; I2 = 99%) or Tg (4 studies, n = 52; MD = −18.52 [95% CI; −37.96, + 0.93]; I2 = 97%) serum levels were significantly affected following the treatments with R. officinalis in comparison to vehicle control groups (Fig. 5B–D, respectively).

Fig. 5
figure 5

Forest plot of preclinical studies of diabetes mellitus comparing R. officinalis and vehicle control; measuring LDL-c (A), HDL-c (B), TC (C), and Tg (D)

Full size image

Subgroup analysis

The subgroup analyses for FPG enabled us to compare the results for the exposure time and type of extracts. As shown in Fig. S2, there was a subgroup effect statistically significant (p < 0.00001; I2 = 79.3%), i.e., the exposure time and/or the extract type impact the antihyperglycemic effect of R. officinalis.

Based on this analysis it’s possible to assert that either the volatile oil (2 studies, n = 24; MD = −18.20 [95% CI; −188.89, + 152.50]; I2 = 98%) or the acute exposure (2 studies, n = 26; MD = −19.45 [95% CI; −195.76, + 156.86]; I2 = 97%) do not yield significant outcomes in decreasing the FPG.

Concerning the SIL, it was possible to compare the results for the type of extract and animal strain whether rabbits or rats. The results of the subgroup analyses are presented in Fig. S3. The test for subgroup differences indicated that there was a significant subgroup effect statistically (p < 0.00001; I2 = 96.6%) so that both extract type and animal strain significantly impact the efficacy of R. officinalis in increasing the release of insulin when compared to vehicle control.

The treatment effect favored R. officinalis over vehicle control for all animal strains and extract types; which suggests that the subgroup effect is quantitative. Both FPG and SIL heterogeneity between results in studies within sub-groups requires further exploration.

Risk of bias and methodological quality

Figure 6A displays the summary of the results concerning the bias risk assessment according to the SYRCLE tool for controlled preclinical trials herein included. The results reveal that all studies did not perform allocation concealment, blinding of outcome assessment, and blinding of evaluation of random results. In addition, an important risk of bias was verified by checking whether there was a pancreatectomy to assess outcome bias or indicating that there was no pancreatic beta regeneration not attributed to R. officinalis exposure.

Fig. 6
figure 6

The risk of bias assessment results using SYRCLE’s tool criteria (A) and methodological quality of the studies included (B) based on CAMARADES checklist

Full size image

Complementarily, Fig. 6B reports the methodological quality evaluated by the CAMARADES checklist for the pre-clinical studies included. The quality score of the majority of studies in this analysis was in mean, 5.3 ± 1.0, indicating an average methodological quality. All 23 studies reported statements of compliance with regulatory requirements and the use of animals with hypertension or diabetes.

However, none of these studies described the blinded assessment of outcome and sample size calculation, and 15 studies did not report how the animals used in the study were euthanized (avoidance of anesthetics with marked intrinsic properties).

Discussion

The efficacy and safety of R. officinalis

To the best of our knowledge, this is the first systematic review and meta-analysis shedding light on the antihyperglycemic, hepatoprotective, nephroprotective, and antihyperlipidemic effects of R. officinalis extracts in preclinical experimental models of T1DM. Our meta-analysis confirmed the beneficial effect of this nutraceutical in reducing FPG besides kidney and hepatic injury markers; in decreasing the LDL-c and increasing SIL in animals.

The reduction in the FPG of the diabetic animals treated with the different R. officinalis extracts is outstanding and, on average, greater than that observed with Momordica charantia L. (n = 815; SMD = −6.86 [95% CI; −7.95, −5.77]; I2 = 90%), which is a nutraceutical widely consumed in Asia and Africa as adjuvant in the management of DM (Peter et al. 2021). Moreover, R. officinalis displays a more appreciable antihyperglycemic effect when compared to the Brazilian superfruit Eugenia uniflora (n = 36; MD = −49.97 [95% CI; −57.18, −42.75]; I2 = 67%) (Máximo et al. 2022) and to the medicinal plant Moringa oleifera (n = 699; MD =  −3.92, [95% CI; − 4.65, − 3.19; I2 = 90.15%) (Watanabe et al. 2021).

It’s important to consider that in all studies herein included, the treatment with R. officinalis was not able to reduce the FPG to normal levels, i.e., < 100 mg/dL (Elsayed et al. 2023). Therefore, using R. officinalis does not ensure controlling the hyperglycemia by itself but may be helpful to enhance the hypoglycemic effects achieved with insulin administration. In turn, decreasing the dose of insulin or its administration frequency can increase patient compliance and adherence to the treatment, and reduce the adverse effects often associated with insulin administration errors.

The chance of herb-drug interactions following the long-term intake of R. officinalis with either insulin or oral hypoglycemic drug administration is to be taken into account because this may exacerbate the hypoglycemic effect. On one hand, it would be beneficial when envisioning the deprescribing of oral hypoglycemic drugs and decreasing their major side effects such as vomiting, flatulence, and diarrhea, which impair the patient’s adherence to the treatment (Ghadge and Kuvalekar 2017). On the other hand, it could provoke hypoglycemia (FPG < 70 mg/dL) which brings several undesired symptoms to patients (e.g., shakiness, sweating, fatigue, difficulty concentrating, etc.) (Nakhleh and Shehadeh 2021).

Still, in the wake of this food or herb-drug interaction scenario, it has been witnessed in the literature that the rosmarinic acid, the major phytochemical comprising R. officinalis acts in vitro as a weak or moderate inhibitor (even mixed or competitive mechanisms) of some human cytochrome P450 monooxygenases (CYP2C19 and CYP2E1) and, mainly, uridine diphosphate glucuronosyltransferases (UGT1A1/1A6/2B7) isoforms (Kim et al. 2019). This last group of enzymes is needful for the hepatic metabolism of some antidiabetic drugs such as canagliflozin and troglitazone (Zhou et al. 2016), besides antitumoral (etoposide and SN-38), antilipidemic (ezetimibe) and non-steroidal anti-inflammatory (acetaminophen, diclofenac, and naproxen) drugs (Kim et al. 2019).

These concerns are of utmost importance because diabetic patients can be up to two-fold more susceptible to using oral anti-diabetic agents associated with medicinal plants (Chelghoum et al. 2021). Therefore, the use of R. officinalis in association with other pharmaceutical drugs in the management of DM must be closely accompanied by a multidisciplinary healthcare professional team.

The overall mechanisms of action behind the antihyperglycemic effect of R. officinalis are not yet fully understood, but it has been suggested that the increment in insulin secretion somehow plays a key role in lowering the FPG (Bao et al. 2020). This is because an increased insulin level can help to the lowering of glucose levels by promoting glucose uptake and storage in the cells, then, inhibiting glucose production in the liver, and stimulating glycogen synthesis while inhibiting glycogenolysis (Rahman et al. 2021).

For clarification sake of this hypothesis, herein we plotted the MD obtained for SIL versus the respective MD of FPG arising from our meta-analyses (Figs. 2A and 3, respectively). There was a very low correlation between the increment of SIL and the decrement of FPG (R2 = 0.027, data not shown), thus further investigations with a greater number of studies are still required to ratify or refute this.

Non-alcoholic fatty liver disease and chronic kidney disease, which are metabolic conditions putatively interdependent, are often reported as DM complications (Byrne and Targher 2020). In a study included in the meta-analysis regarding the biochemical markers of hepatic damage related to DM (Selmi et al. 2017), the administration of R. officinalis lowered the ALT and AST serum levels to normal values, i.e., < 40 U/L (Kim et al. 2004).

Even though all the treatment effects favored R. officinalis over vehicle control in the meta-analysis of BUN, the decrement observed for this parameter associated with kidney injury did not reach normal levels, i.e., 5 ≤ BUN ≤ 20 mg/dL (Hosten 1990). Nonetheless, in the group treatment with R. officinalis the CRE blood levels have been lowered to the normal ranges, which are 0.7–1.3 mg/dL for adult males and 0.6–1.2 mg/dL for adult females (Hosten 1990). Therefore, R. officinalis may promote hepatoprotection more markedly than nephroprotective effects in DM.

The hepatoprotective effects of R. officinalis have been confirmed by several in vitro and in vivo investigations irrespective of the method used to induce liver damage (Amin and Hamza 2005; Guimarães et al. 2023; Hegazy et al. 2018; Ielciu et al. 2021; Mohamed et al. 2022; Rašković et al. 2014). The lipid alterations that occur in DM, such as the increment of the LDL-c, are related to cardiovascular diseases, one of the main causes of mortality and morbidity in people with DM (Cole and Florez 2020).

Among the hypotheses related to these alterations, we have the high formation of advanced glycation end products and the increment in reactive oxygen species that are favored by hyperglycemia (Méndez et al. 2010). This process can lead to the accumulation of oxidized lipids, which can contribute to the development of lipid metabolism disorders and the progression of related diseases. Furthermore, a study published by Jiang et al. 2020, suggested that oxidative stress-induced lipid peroxidation can contribute to the development of insulin resistance, which is a key feature of DM.

Thence, besides the other benefits aforementioned, R. officinalis may reduce cardiovascular risk to a certain extent by decreasing the serum LDL-c level and avoiding the fatty deposits assembled in the arteries.

The role of OS in the pathogenesis of DM is extensively revised elsewhere (Darenskaya et al. 2021). Since R. officinalis is an outstanding source of antioxidants (Andrade et al. 2018; Celiktas et al. 2007; Gonçalves et al. 2011; Sánchez-Camargo and Herrero 2017; Takayama et al. 2022), several studies herein included assessed in vivo antioxidant activity of the tested extracts (Bakirel et al. 2008; El-Boshy et al. 2015; Emam 2012; Khalil et al. 2012; Mohamed 2021; Nazem et al. 2015; Sebai et al. 2015; Selmi et al. 2017; Shen et al. 2020; Silva et al. 2011; Wahba 2022). However, the methods used by the authors in pursuit of this and the antioxidant outcomes varied widely. For this reason, it was not possible neither to perform metanalysis nor to correlate the in vivo biochemical markers of antioxidant activities with the preventive effect in DM.

Noteworthy, none of the preclinical studies included in this systematic review reported any adverse effects, mortality, or behavioral changes in the tested animals that were attributed to R. officinalis exposure. These findings hint that R. officinalis is safe in all of the tested doses (25–800 mg/kg for extracts and 0.5% to 10% for leaves powder). Indeed, the European Food Safety Authority has classified R. officinalis extracts as a Generally Recognized as Safe (GRAS) food (CFR182.10; 182.20)(Aguilar et al. 2008), not showing high levels of toxicity at therapeutic doses.

Study limitations

Some studies evaluated in this work reported having assessed FPG in diabetic animals but did not demonstrate the raw results in the tables or figures (Belmouhoub et al. 2016, 2017, 2020; Ramadan et al. 2013). Hence, as we did not receive any answer about these raw data from the authors, it was not possible to include them in our meta-analysis.

Regarding the experimental design of the included articles, a large number of studies included only male rats in their experimental protocols, and it was not possible to compare the effects between males and females.

We also point out the lack of standardization when reporting the dose of extract administered to the animals. One study did not report the weight of the animals or the concentration of extract administered (Al-badry 2017), so it was not possible to define the dose administered (mg/Kg).

Many articles did not inform how the sample size calculation, housing, allocation, and blinding were carried out, thus revealing a high risk of bias for these domains and impairing the internal validity of these investigations. This trend is not an exclusivity of our findings, since it has been very often reported in systematic reviews and meta-analyses dealing with preclinical trials in animals (Gupta 2019; Máximo et al. 2022; Mignini and Khan 2006; Peter et al. 2021; Rostamkhani et al. 2022; Shojaei-Zarghani et al. 2022).

Furthermore, a scanty number of studies assessing the HbA1c levels in the animals was included, and R. officinalis displayed no appreciable impact in this parameter, which represents an index of average glucose and is considered the gold standard for monitoring the patient’s glucose control over time (Sherwani et al. 2016). This inefficacy is likely related to the short-lasting of the studies assessing this biochemical marker., i.e., much lower than 90 days (the average time required for erythrocyte turnover). which might have been insufficient to adequately capture such effects.

The meta-analysis of animal studies has been considered a valuable statistical toolset for improving healthcare since it can be further used to generate new hypotheses and guide the design of clinical trials in humans (Hooijmans et al. 2014a; Sauvant et al. 2020).

The high heterogeneity displayed in our meta-analysis is likely attributed to the various methodological design aspects of the included articles, e.g., variability related to the animals used (strains, age, etc.), small sample sizes, different drugs administrated to induce DM and their doses, diverse intervention features (plant origin, harvesting time, solvents and procedures used to prepare the extract, route of administration, doses administrated, and treatment period), distinct outcome measurement methods, etc. This is also a recurrent trend for animal studies, that can restrict but not prevent the animal-to-human translation of findings arising from systematic reviews and meta-analyses (Bahadoran et al. 2020).

Finally, a few articles reported the accurate quantitative phytochemical analysis of the extracts administered in the animals (Rasoulian et al. 2019; Sebai et al. 2015; Shen et al. 2020), thereby revealing warning analytical standardization issues. Although displaying a great challenge, the standardization of herbal extracts plays a key role throughout the rationale development of phytopharmaceuticals once it is required for i) assessing the seasonal and batch-to-batch variability to assure consistent chemical profile; ii) identifying and controlling critical processing parameters such as extraction, drying, transportation and storage; iii) establishing the major active compounds involved in the biological activities and their putative mechanisms of action; and iv) enabling to achieve improved and reproducible treatment outcomes (Alamgir 2017).

Conclusions

The administration of R. officinalis derivatives dwindles the FPG besides hepatic damage in diabetic animals, thence this nutraceutical may help in the management of DM and some of its complications. The pre-clinical trials herein analyzed display relevant issues concerning the risk of bias and methodological quality. Therefore, although plausible, so far, it’s not possible to draw a solid conclusion regarding the efficacy of this nutraceutical as an alternative in the primary care of DM.

Patients and healthcare professionals would greatly benefit from further high methodological quality studies, especially controlled randomized clinical trials performed with accurately standardized extracts and assessing HbA1c as a primary outcome.