点击显示 收起
1 From the Information Center (KT, KU, YS, YT, and KE) and the Nutritional Education Program (SW), National Institute of Health and Nutrition, Tokyo, Japan
2 Supported by a grant from the Japanese Ministry of Health, Labor, and Welfare. 3 Address reprint requests to K Taku, Information Center, National Institute of Health and Nutrition, 1-23-1 Toyama, Shinjuku-ku, Tokyo 162-8636, Japan. E-mail: takuk{at}nih.go.jp.
ABSTRACT
Background: Clinical trials have reported the cholesterol-lowering effects of soy protein intake, but the components responsible are not known.
Objective: This meta-analysis was primarily conducted to evaluate the precise effects of soy isoflavones on lipid profiles. The effects of soy protein that contains enriched and depleted isoflavones were also examined.
Design: PUBMED was searched for English-language reports of randomized controlled trials published from 1990 to 2006 that described the effects of soy protein intake in humans. Eleven studies were selected for the meta-analysis.
Results: Soy isoflavones significantly decreased serum total cholesterol by 0.10 mmol/L (3.9 mg/dL or 1.77%; P = 0.02) and LDL cholesterol by 0.13 mmol/L (5.0 mg/dL or 3.58%; P < 0.0001); no significant changes in HDL cholesterol and triacylglycerol were found. Isoflavone-depleted soy protein significantly decreased LDL cholesterol by 0.10 mmol/L (3.9 mg/dL or 2.77%; P = 0.03). Soy protein that contained enriched isoflavones significantly decreased LDL cholesterol by 0.18 mmol/L (7.0 mg/dL or 4.98%; P < 0.0001) and significantly increased HDL cholesterol by 0.04 mmol/L (1.6 mg/dL or 3.00%; P = 0.05). The reductions in LDL cholesterol were larger in the hypercholesterolemic subcategory than in the normocholesterolemic subcategory, but no significant linear correlations were observed between reductions and the starting values. No significant linear correlations were found between reductions in LDL cholesterol and soy protein ingestion or isoflavone intakes.
Conclusions: Soy isoflavones significantly reduced serum total and LDL cholesterol but did not change HDL cholesterol and triacylglycerol. Soy protein that contained enriched or depleted isoflavones also significantly improved lipid profiles. Reductions in LDL cholesterol were larger in hypercholesterolemic than in normocholesterolemic subjects.
Key Words: Soy isoflavones • soy protein • lipid • total cholesterol • LDL cholesterol • HDL cholesterol • triacylglycerol • meta-analysis
INTRODUCTION
A meta-analysis published in 1995 reported that ingestion of 47 g soy protein/d reduced serum total cholesterol by 9.3%, LDL cholesterol by 12.9%, and triacylglycerol by 10.5% and increased HDL cholesterol by 2.4% (1). Two more recent meta-analyses reexamined the effects of soy protein that contained isoflavones on the lipid profile as compared with the effects of casein or other animal proteins (2, 3), and they found substantially weaker effects. Anderson et al (1) and Zhan and Ho (3) concluded that the effects were related to the initial serum lipid concentrations.
The studies did not address possible mechanisms of the effects of soy protein intake. Whether the changes were attributable to the soy protein per se, other soy-derived factors (eg, constitutive isoflavones), or both remained unclear. Zhan and Ho (3) reported that studies with isoflavone intakes >80 mg/d found better effects on lipid profiles, whereas Weggemans and Trautwein (2) concluded that consumption of soy-associated isoflavones was not related to changes in LDL or HDL cholesterol. Two of the authors of the current report (Kyoko Taku was formerly named Xing-Gang Zhuo) previously performed a meta-analysis (4), in which they found that the consumption of soy protein with a high isoflavone content reduced serum LDL-cholesterol concentrations more than did consumption of the same soy protein amount with low isoflavone content. The decrease in hypercholesterolemic subjects was larger than that in normocholesterolemic subjects.
Clinical trial studies used various intakes of soy protein with various amounts of isoflavones and different protocols. In the present meta-analysis, we primarily attempted to evaluate the precise effects of soy isoflavones, at the same time that we controlled for the ingested amounts of soy protein, on serum total, LDL, and HDL cholesterol and triacylglycerol. The effects of soy protein that contained enriched and depleted isoflavones compared with those of animal protein without isoflavones on lipid profiles were examined within the current dataset.
SUBJECTS AND METHODS
Study identification and selection
The PUBMED database (National Library of Medicine, Bethesda, MD) was searched on May 29, 2006, for English-language reports of randomized controlled trials published from 1990 to 2006 that described effects of soy protein intake and isoflavones on serum lipid concentrations in humans. The search was performed with the key words "(soy protein OR soy OR soybean OR soya) AND (isoflavones OR isoflavone) AND (cholesterol OR lipid)," with the constraints noted previously. We also looked for relevant articles in the reference lists of 3 meta-analyses (1-3). Studies were selected for analysis if they met the following criteria: 1) adult subjects ingested soy protein for 1–3 mo; 2) the study was a randomized controlled trial with either a parallel or a crossover design; 3) the study had comparable groups with enriched and depleted isoflavones for the same ingested amount of soy protein; 4) the study provided the intake amount of soy isoflavones; and 5) the starting and endpoint lipid concentrations were available. On the basis of these criteria, 15 studies (5-19) were identified. Two studies (5, 8) secondarily analyzed a subset of subjects in the primary study (13), 1 study (6) reported only 1 of 4 lipid concentrations for the same study subjects in another study (12), and 1 study did not provide SD values (7). These 4 studies (5-8) were excluded, and 11 studies (9-19) were selected for the meta-analysis.
Data extraction
The number of participants in each group and the means and SDs of serum lipid concentrations were extracted from the 11 studies and tabulated for analysis. The isoflavone-depleted soy protein used in each study contained trace amounts of isoflavones because of incomplete extraction. Although 3 studies had 3 (9) or 2 (17, 19) soy protein groups containing different amounts of isoflavones, we selected the group with the highest concentrations of isoflavones for analysis. One study had 2 groups of soy protein that contained enriched isoflavones (10): the group with normal phytate was selected for analysis. One study had 2 groups of animal protein with and without isoflavones (13): we selected animal protein without isoflavones for analysis. One study also reported midstudy (ie, 4 wk) measurements, but we used measurements from the full study period (ie, 12 wk) for analysis (11). One study measured lipid concentrations during 4 phases of the menstrual cycle, but we used mean values for this meta-analysis (17). Because previous studies suggested that the effects of soy protein are related to the initial plasma lipid concentrations (1, 3, 9, 13), we divided our data into 2 subcategories (hypercholesterolemic and normocholesterolemic) with the use of a baseline LDL-cholesterol cutoff value of 4.14 mmol/L (160 mg/dL). Two studies also divided their subject populations into 2 groups according to baseline LDL-cholesterol concentrations (9, 13), but Lichtenstein et al (13) did not provide baseline values for each subgroup, and we used their data on the group as a whole for analysis.
Meta-analysis and statistical analysis
We performed the meta-analysis with the use of REVMAN for WINDOWS software (version 4.2.8; updated on July 8, 2005; Cochrane Collaboration, Oxford, United Kingdom). The estimate of the principal effect on lipid profiles of soy isoflavones after control for amounts of soy protein ingestion was defined as the mean difference in endpoint serum lipid concentrations between comparison groups (ie, value for the subjects ingesting soy protein that contained enriched isoflavones minus that for the subjects ingesting isoflavone-depleted soy protein). The estimates of principal effect of isoflavone-depleted soy protein and soy protein that contained enriched isoflavones compared with that of animal protein without isoflavones were also defined as the mean differences between comparison groups. We use a fixed-effect model or a random-effects model to calculate weighted mean differences; 95% CIs for each comparison; a combined overall effect, with P values for the total population and the hypercholesterolemic and normocholesterolemic subcategories; and the P value for testing heterogeneity. Each comparison was assigned a weight consisting of the reciprocal of its variance based on n and SD. We present the results based on a random-effects model when the test for heterogeneity of the total population was significant (20). Otherwise, the results of a fixed-effect model are presented. We examined potential publication bias on the basis of a funnel plot, which plotted the SEM of the studies against their corresponding effect sizes. Pearson correlation coefficients were calculated with the use of SPSS for WINDOWS software (version 14.0J; SPSS Inc, Chicago, IL).
RESULTS
Characteristics of the studies
The characteristics of treatment groups selected for analysis in the 11 randomized controlled trials are summarized in Table 1. Four studies used a parallel design (9-11, 14), and the others used a crossover design (12, 13, 15-19). Seven studies focused on postmenopausal women (10-14, 18, 19), 1 focused on men (15), 1 focused on premenopausal women (17), and 2 studies included both sexes (9, 16). Two studies used casein as the animal protein in the control group (9, 16), 3 used milk (11, 15, 18), 1 used dairy and eggs (12), and 1 used dairy and meat (13). The study durations ranged from 4 wk (12) to 93 d (19). In each study, subjects in different groups ingested the same amount of soy or animal protein, ranging from 25 g/d (9, 18) to 133 g/d (16), and the measured isoflavone intakes ranged from 1.64 mg/d (15) to 317.9 mg/d (16). Three studies had moderately hypercholesterolemic subjects (9, 11, 13), 2 studies had hypercholesterolemic subjects only (12, 14), and 6 studies had normocholesterolemic subjects only (10, 15-19), according to the initial lipid concentrations. The reported mean starting serum total cholesterol concentrations ranged from 3.87 mmol/L (17) to 7.47 mmol/L (14) (150–289 mg/dL).
View this table:
TABLE 1. Characteristics of treatment groups selected for analysis in 11 randomized controlled trials1
Six studies reported serum lipid concentrations (10, 12-15, 18), and the others reported values for plasma (9, 11, 16, 17, 19). Normally, serum cholesterol concentrations are 3% higher than corresponding plasma concentrations (21), but, because we were interested in mean differences in each study, we analyzed plasma and serum concentrations without correction for this difference; we report all results as serum concentrations. The mean age of subjects ranged from 26.3 y (17) to 62.7 y (13), and the mean body mass index (in kg/m2) ranged from 22.8 to 27.9 (17). Isoflavone amounts were reported in aglycone units in all but 1 study (14), and we considered all amounts as being in aglycone form to calculate the mean intake amount. Ten studies used alcohol (ethanol) extraction to prepare isoflavone-depleted soy protein (9-12, 14-19), and 1 did not report the extraction method used (13).
In addition, in most of the comparisons, subjects consumed similar diets, with similar amounts of fat (total and saturated), cholesterol, and fiber. No significantly different baseline body weights, body mass indexes, or lipid concentrations were reported between comparison groups. Most of the studies were designed to maintain body weight, and no significant weight changes were reported (data not shown).
Effects of isoflavones after control for soy protein
Pooled estimates of treatment effects on the changes in serum lipid concentrations are summarized in Table 2. To more precisely evaluate the effects on lipid profiles of isoflavones after control for the ingestion of soy protein amount, 12 comparisons (both groups ingested the same amount of soy protein that contained either enriched or depleted isoflavones) in 11 studies were selected for meta-analysis.
View this table:
TABLE 2. Pooled estimates of treatment effect on the lipid profile1
For the total population, total cholesterol decreased by 0.10 mmol/L [3.9 mg/dL (95% CI: –0.17, –0.02 mmol/L or –6.6, –0.8 mg/dL); P = 0.02] in the group consuming soy protein that contained enriched isoflavones (ISP+ group) compared with the group consuming the corresponding isoflavone-depleted soy protein (ISP– group), or by 1.77% of the starting value in the ISP– group. The test for heterogeneity was not significant (P = 0.54), which suggested that combining these studies for a meta-analysis was valid. Of 12 comparisons (5 in hypercholesterolemic and 7 in normocholesterolemic subcategories), 9 (75%) reported reductions (Table 3, Figure 1). Only 2 comparisons in the hypercholesterolemic subcategory reported significant reductions (P < 0.05 and P = 0.005, respectively) (9, 11). No significant reductions in total cholesterol were found for the hypercholesterolemic (P = 0.13) or normocholesterolemic (P = 0.06) subcategories (Table 2).
View this table:
TABLE 3. Blood total cholesterol concentrations in groups ingesting soy protein that contains enriched isoflavones (ISP+) and isoflavone-depleted soy protein (ISP–) and estimated treatment effect1
FIGURE 1.. Weighted mean difference (WMD) with the fixed-effect model (fixed) in endpoint total and LDL cholesterol between the group ingesting soy protein that contained enriched isoflavones (ISP+) and the group ingesting isoflavone-depleted soy protein (ISP–). The horizontal lines denote the 95% CI.
For the total population, LDL cholesterol decreased by 0.13 mmol/L [5.0 mg/dL (95% CI: –0.20, –0.07 mmol/L or –7.7, –2.7 mg/dL); P < 0.0001] in the ISP+ group compared with the corresponding ISP– group, or by 3.58%. The test for heterogeneity was not significant (P = 0.83), which suggests that combining these studies for a meta-analysis was valid. All comparisons (5 in hypercholesterolemic and 7 in normocholesterolemic subcategories) reported the reductions, and 2 comparisons in the normocholesterolemic subcategory observed that the upper limit of the 95% CI was <0 (17, 19) (Table 4, Figure 1
View this table:
TABLE 4. Blood LDL-cholesterol concentrations in groups ingesting soy protein that contains enriched isoflavones (ISP+) and isoflavone-depleted soy protein (ISP–) and estimated treatment effect1
FIGURE 2.. Linear correlation between isoflavone intake difference and LDL-cholesterol reduction for soy protein that contained enriched isoflavones compared with isoflavone-depleted soy protein.
Effects of isoflavone-depleted soy protein
Eight comparisons in 7 studies (Table 2) were subsequently selected from the current data set of 11 studies to examine effects on the lipid profile of the ISP– group compared with the group consuming animal protein without isoflavones. For the hypercholesterolemic subcategory, total cholesterol decreased by 0.20 mmol/L [7.7 mg/dL (95% CI: –0.40, 0 mmol/L or –15.5, 0 mg/dL); P = 0.05] in the ISP– group compared with the corresponding group consuming animal protein without isoflavones, or by 3.56% of the starting value in the latter group. The test for heterogeneity was not significant (P = 0.41), which indicated that it was valid to combine the comparisons in the hypercholesterolemic subcategory for a meta-analysis. No significant reductions in total cholesterol were observed for the total population (P = 0.13) or the normocholesterolemic subcategory (P = 0.63).
For the total population, LDL cholesterol decreased by 0.10 mmol/L [3.9 mg/dL (95% CI: –0.19, –0.01 mmol/L or –7.3, –0.4 mg/dL); P = 0.03] in the ISP– group compared with the group consuming the corresponding animal protein, or by 2.77%. The test for heterogeneity was not significant (P = 0.69), which suggested that combining these studies for a meta-analysis was valid. No linear correlation (Pearson's r = –0.321, P = 0.439) was observed between reductions (in percentage terms) and the starting LDL concentrations in the group consuming animal protein without isoflavones. No linear correlation (Pearson's r = 474, P = 0.236) was observed between the reductions and the ingested soy protein amounts. No significant reductions in LDL cholesterol were found for the hypercholesterolemic (P = 0.19) or normocholesterolemic (P = 0.09) subcategories.
For the normocholesterolemic subcategory, HDL cholesterol increased by 0.06 mmol/L [2.3 mg/dL (95% CI: 0.01, 0.11 mmol/L or 0.4, 4.3 mg/dL); P = 0.03] in the ISP– group compared with the group consuming the corresponding animal protein without isoflavones, or by 4.5%. The test for heterogeneity was not significant (P = 0.48), which indicated that it was valid to combine comparisons in the normocholesterolemic subcategory for a meta-analysis. No significant reductions in HDL cholesterol were observed for the total population (P = 0.07) or the hypercholesterolemic subcategory (P = 0.99). No significant reduction in triacylglycerol was found in the ISP– group compared with the group consuming the corresponding animal protein without isoflavones.
Effects of soy protein that contained enriched isoflavones
Eight comparisons in 7 studies (Table 2) were selected from the current dataset to examine effects on the lipid profile of the ISP+ group compared with the group consuming animal protein without isoflavones. For the hypercholesterolemic subcategory, total cholesterol decreased by 0.32 mmol/L [12.4 mg/dL (95% CI: –0.50, –0.14 mmol/L or –19.3, 5.4 mg/dL); P = 0.0006] in the ISP+ group compared with the group consuming the corresponding animal protein without isoflavones, or by 5.69% of the starting value in the latter group. The test for heterogeneity was not significant (P = 0.57), which indicated that it was valid to combine the comparisons in the hypercholesterolemic subcategory for a meta-analysis. No significant reductions in total cholesterol were observed for the total population (P = 0.09) or the normocholesterolemic subcategory (P = 1.00).
In the total population, LDL-cholesterol concentrations decreased by 0.18 mmol/L [7.0 mg/dL (95% CI: –0.27, –0.09 mmol/L or –10.4, –3.5 mg/dL); P < 0.0001] in the ISP+ group compared with the group consuming the corresponding animal protein without isoflavones, or by 4.98%. The test for heterogeneity was not significant (P = 0.51), which suggested that combining these studies for a meta-analysis was valid. Significant reductions in LDL cholesterol of 0.28 mmol/L or 7.75% [10.8 mg/dL (95% CI: –0.44, –0.12 mmol/L or –17.0, –4.6 mg/dL); P = 0.0006] and 0.14 mmol/L or 3.87% [5.4 mg/dL (95% CI: –0.24, –0.03 mmol/L or –9.3, –1.2 mg/dL); P = 0.01] were found for the hypercholesterolemic and normocholesterolemic subcategories, respectively. Tests for heterogeneity in neither subcategory were significant (P = 0.63 and P = 0.49, respectively), which indicated that it was valid to combine the comparisons in each subcategory. The LDL-cholesterol reduction was larger in the hypercholesterolemic subcategory (subjects with higher baseline LDL-cholesterol concentrations) than in the normocholesterolemic subcategory, but no linear correlation (Pearson's r = –0.443, P = 0.272) was observed between reductions (in percentage terms) and the starting LDL-cholesterol concentrations of the group consuming animal protein without isoflavones. No linear correlations between LDL-cholesterol reductions and soy protein ingestion (
FIGURE 3.. Linear correlation between isoflavone intake and LDL-cholesterol reduction for soy protein that contained enriched isoflavones compared with animal protein without isoflavones.
In the total population, HDL-cholesterol concentrations increased by 0.04 mmol/L [1.6 mg/dL (95% CI: 0, 0.08 mmol/L or 0, 3.1 mg/dL); P = 0.05] in the ISP+ group compared with the group consuming the corresponding animal protein without isoflavones, or by 3.00% (Table 2). The test for heterogeneity was not significant (P = 0.51), which suggested that combining these studies for a meta-analysis was valid. No significant reductions in HDL cholesterol were observed for the hypercholesterolemic (P = 0.47) or normocholesterolemic (P = 0.06) subcategories. No significant reduction in triacylglycerol was found in the ISP+ group compared with the group consuming the corresponding animal protein without isoflavones.
Publication bias
The funnel plots of the effects on lipid profiles in comparison groups did not indicate obvious publication bias (data not shown).
DISCUSSION
This meta-analysis of 11 randomized controlled trials showed that soy protein that contained enriched isoflavones significantly decreased serum total and LDL cholesterol compared with the same amounts of isoflavone-depleted soy protein. The results suggest that ingesting 102 mg soy-derived isoflavones/d (the mean difference between the 2 groups), independent of the amount of soy protein ingested, for 1–3 mo would lower total cholesterol by a mean of 0.10 mmol/L (1.77%) and LDL cholesterol by a mean of 0.13 mmol/L (3.58%). The LDL cholesterol–lowering effect was consistent with our previous meta-analysis (4). An intake of 102 mg soy isoflavones/d (daidzein and genistein) is equivalent to approximately twice the amount consumed habitually in Japan (
Although most of the 11 individual studies in this meta-analysis found a slight decrease in total and LDL cholesterol, the 95% CI often included zero. However, the combined overall effects in this meta-analysis were significantly different from zero (95% CI), which suggested that limited sample sizes often prevent detection of significant effects in individual studies.
A meta-analysis of 7 studies showed that isoflavone-depleted soy protein when compared with animal protein without isoflavones was associated with a significant decrease in serum total cholesterol (0.20 mmol/L or 3.56%) in the hypercholesterolemic subcategory and in LDL cholesterol (0.10 mmol/L or 2.77%) in the total population. It also showed a significant increase in serum HDL cholesterol (0.06 mmol/L or 4.50%) in the normocholesterolemic subcategory. These results suggest that ingesting an average of 49 g soy protein that contained extremely low isoflavones/d (
One possible explanation for the lack of linear correlations between reductions in LDL cholesterol and pretreatment values, soy protein ingestion, or isoflavone intakes may be the different amounts of constituent isoflavones or other components in the various soy proteins tested. Another explanation may be the individual differences in the capacity of intestinal flora to convert daidzein into its metabolite, equol. Equol is easily absorbed and possesses substantial estrogenic activity because of its affinity for both the estrogen and ß receptors (26).
In conclusion, soy isoflavones significantly lowered serum total and LDL cholesterol but did not change HDL cholesterol and triacylglycerol. Soy protein with or without isoflavones also significantly improved lipid profiles. Reductions in LDL cholesterol were larger in hypercholesterolemic subjects than in normocholesterolemic subjects, but they had no linear correlation with pretreatment values or isoflavone intakes. When provided concurrently with soy protein, soy isoflavones would have synergistic or additive effects on cholesterol lowering.
ACKNOWLEDGMENTS
KT contributed to the study search and selection, data extraction, meta-analysis, and preparation of the first draft of the manuscript; YS contributed to the collection of full-text articles; KE and YT confirmed the study selection and extracted data; and KU and SW contributed to the final version of the manuscript. None of the authors had any financial or personal conflict of interest.
REFERENCES