Sulforaphane ameliorates lipid profile in rodents: an updated systematic review and meta-analysis

Sulforaphane (SFN), a naturally-occurring isothiocyanate enriched in cabbage and broccoli, has been provided as food supplements to improve weight management and reduce lipid levels. However, its effects on serum lipid profiles are contradictory. In this review, a meta-analysis and systematic review of SFN on lipid reduction and weight control is assessed with mice and rats fed on high-fat diet. The effects of SFN supplementation were evaluated by weighted mean difference (WMD) in body weight (BW), liver weight (LW) and also by its effect on serum lipids. A random-effects model was applied to estimate the overall summary effect. SFN reduced BW (WMD: − 2.76 g, 95% CI: − 4.19, − 1.34) and LW (WMD: − 0.93 g, 95% CI: − 1.63, − 0.23) significantly in our ten trials. Its effects on serum total cholesterol (TC) (WMD: − 15.62 mg/dL, 95% CI: − 24.07, − 7.18), low-density lipoprotein cholesterol (LDL-C) (WMD: − 8.35 mg/dL, 95% CI: − 15.47, − 1.24) and triglyceride (TG) (WMD: − 40.85 mg/dL, 95% CI: − 67.46, − 14.24) were significant except for high-density lipoprotein cholesterol (HDL-C) component (WMD: 1.05 mg/dL, 95% CI: − 3.44, 5.54). However, species, disease model, duration, SFN dosage as well as route of administration did not explain the heterogeneity among studies. In summary, these findings provide new insights concerning preclinical strategies for treating diseases including obesity, diabetes, hypertension, non-alcoholic fatty liver disease as well as cardiovascular disease with SFN supplements.

Statistical analysis. Treatment effects were considered as weighted mean difference (WMD) and the corresponding standard error (SE) in BW, LW and concentrations of serum lipids (TC, LDL-C, HDL-C and TG). To estimate the overall effect, we used a random-effect model, previously described by DerSimonian and Laird, which considers both within and between-study heterogeneity 27 . Heterogeneity among the studies was estimated using the I 2 statistic, with values of 0-25%, 25.1-75%, and 75.1-100% representing a low, moderate, and high degree of heterogeneity, respectively. When standard deviations or SEs were not shown in studies, they were calculated using 95% CI. In addition, when studies have reported median and interquartile range, they were converted to mean and SE using available formulas 28 . Statistical analyses were done using Stata, version 13 (Stata Corp., College Station, TX, USA). P-values less than 0.05 were considered statistically significant.

Selection of articles.
A total of 654 studies were involved by our database search. 279 duplicate articles were removed. After reading the title and abstract of papers, 20 articles were selected to analyze the full text with removing 355 studies. We considered inclusion and exclusion criteria and then excluded 10 further studies owing to prescribing broccoli supplement or broccoli sprout extract instead of SFN (n = 3), acute SFN action (n = 2), RCTs (n = 2), rabbit model (n = 1), alcohol-induced liver steatosis model (n = 1) and lacking of clear data (n = 1) (Fig. 1). Ultimately, this meta-analysis was conducted on ten trials of rodents [29][30][31][32][33][34][35][36][37][38] (Table 1), including 5 batches of C57BL/6 mice, 4 batches of Wistar rats and 1 batch of Sprague Dawley rats. In terms of gender, all the trials selected male animals aged 4-10 weeks. The methods to build disease models included feeding mice or rats with high-fat diet, high-fructose diet or highly palatable diet or injecting of streptozotocin (STZ) into rodents. Specifically, four trials induced obesity by feeding with high-fat diet, one trial by feeding with high-fructose diet, one trial by feeding with high-fat high-sucrose diet, and one trial by feeding with highly palatable diet, two trials evoked diabetes by feeding with high-fat diet and then injecting of STZ and one trail by injection of STZ. The animals were treated with SFN using multiple routes, including by oral administration, oral gavage, subcutaneous injection, and intraperitoneal injection. The intervention duration was 3 to 16 weeks. The dosage of SFN ranged from 0.5 mg/kg to 30 mg/kg.

Discussion
In this updated meta-analysis, ten articles were utilized to assess SFN supplementation effects on body weight and lipid profile in preclinical animal models. Our analysis clearly demonstrates that SFN supplementation significantly decreased BW, LW, TC as well as LDL-C levels, apart from HDL-C. This is the first meta-analytic study that  www.nature.com/scientificreports/ summarizes the function of SFN mono-treatment on lipid profile in rodents with metabolic syndrome. Clearly, SFN has a positive effect in reducing BW and LW and promotes physiologically lipid profile in animal models. Our results reveal that different disease models (with or without STZ) do not influence the previous results of meta-analysis on lipid-related parameters and weight in rodents after supplementation with SFN. Duration, dosage of SFN or route of administration has no effect on this result.
Our research focuses on the effects of SFN as a mono-therapeutic drug on animal lipid profile. Majority clinical trials used of broccoli and broccoli sprouts instead of SFN mono-treatment. However, clinical trials with intake of broccoli (enriched with SFN) can provide some auxiliary support for SFN mono-therapeutic research. Adriana Conzatti and coworkers' study revealed that broccoli sprouts could improve lipid profile and blood gulcose 13 . Armah et al. 17 , also found that plasma LDL-C was significantly downregulated with intake of high glucoraphanin broccoli. However, according to the result of Sudini et al. 39 , intake of broccoli sprouts lasting for half a week did not ameliorate inflammation and oxidative stress markers, in spite of causing a remarkable increase in serum SFN levels. Overall, all related clinical trials we mentioned here, using food like broccoli instead of SFN mono-treatment as therapeutic intervention, required longer time courses to obtained reliable physiologically effects.
In vivo and in vitro studies had shown that SFN can improve lipid-related metabolic indicators and ameliorate cardiovascular disorders 40,41 . Recently, SFN was reported to attenuate HFD-induced obesity through inhibiting lipogenesis via AMP-activated protein kinase (AMPK) pathway 20,29 . SFN played a positive role in cardiomyopathy  www.nature.com/scientificreports/ that was specifically related to the Nrf2-mediated antioxidant pathways (nuclear factor, erythroid 2 like 2) and the AMPK-upregulated lipid metabolism 29,42,43 . Clinical and animal studies shows that the preventive effect of SFN on CVD could also be through Nrf2 activation 44 . Apart from mediating metabolic syndromes, SFN was also found to reduce glycated hemoglobin and fasting blood glucose in type 2 diabetes patients 14,42,45 . Fu et al. 46 showed that SFN supplements attenuated reactive oxygen species stimulated by glucose, and thereby decreased insulin secretion. SFN could ameliorate obesity and insulin resistance in parallel experiments 31,[47][48][49] . Clearly, the take home message is that SFN could attenuate certain risk factors of metabolic syndrome through weight management and reduction of lipid abnormalities 12,50 . This is the first systematic review and meta-analysis of SFN monotreatment vs. whole broccoli on lipid distribution in rodents with metabolic syndrome. Our subgroup analysis focused on model method, age, SFN dosage, intervention duration and route of administration. Furthermore, the meta-analysis involved 10 studies from various countries and animal models.
However, some limitations of this study must be kept in mind. First, it is not clear whether gender responses differently to the effect of SFN, because all researches use male animals rather than female. Female and male animals have different sex hormones, which may affect serum lipid concentration 51 . It's necessary to conduct research using both male and female to evaluate the effect of SFN on the lipid profile. Moreover, our meta-analysis is based on animal experiments rather than RCTs. Results obtained through animal models are not necessarily applicable to humans. In addition, only ten studies meet our requirements. The number of studies with clearly delineated data about metabolic parameters is too small to engage in further subgroup analysis. Most articles were not exclusively performed on rodents fed with HFD and the trail using other food supplements with SFN was not utilized here owning to exclusion criterion.
Overall, our analysis supports the conclusion that SFN supplements decrease the level of BW, LW and lipid profile such as TC, TG, LDL-C in rodents. However, this needs to be validated by relevant clinical trials. In addition, it will be necessary to design and perform a more comprehensive panel of indicators in patients with conditions including dyslipidemia, obesity, CVD, NAFLD and related metabolic disorders.