The effects of soy protein on chronic kidney disease: a meta-analysis of randomized controlled trials

Abstract

Background/Objectives:

There is a growing body of evidence to indicate that soy protein consumption may have a beneficial effect on kidney function. We performed a meta-analysis to evaluate the effects of soy protein consumption compared with animal protein consumption in patients with pre-dialysis chronic kidney disease (CKD).

Subjects/Methods:

We conducted a structured electronic search of the databases PubMed, EMBASE, Cochrane Library and Chinese Biological Medicine for randomized controlled trials published up to March 2014. The outcome measures were serum creatinine (SCR), triglyceride (TG), total cholesterol (TC), calcium (Ca) and phosphorus concentrations. Weighted or standard mean differences were calculated for net changes using random-effects models.

Results:

The meta-analysis consisted of nine trials, comprising 197 subjects. Soy protein intake significantly reduced SCR and serum phosphorus concentrations. The mean difference was −6.231 μmol/l (95% confidence interval (CI): −11.109, −1.352 μmol/l) for SCR (P=0.012) and −0.804 (95% CI: −1.143, −0.464 μmol/l) for serum phosphorus (P=0.00). It also significantly reduced serum TG, with a pooled estimated change of −0.223 mmol/l (95% CI: −0.396, −0.051 mmol/l; P=0.011) after the exclusion of one trial indicated by sensitivity analyses. No statistically significant effects were observed for TC (−0.135 mmol/l (95% CI: −0.289, 0.019 mmol/l)) or Ca (0.023 mmol/l (95% CI: −0.016, 0.062 mmol/l)).

Conclusions:

The meta-analysis suggested a protective effect of soy protein consumption on SCR and serum phosphorus concentrations in pre-dialysis CKD patients. It may also have a significant effect on lowering serum TG concentrations. However, nonsignificant effects on TC and Ca were observed. Evidence was limited because of the relatively small number of available trials and subjects.

Introduction

Chronic kidney disease (CKD) is a worldwide health problem with increasing incidence and prevalence, high costs and poor outcomes.¹ Therefore, the question of how to ameliorate the progression of CKD has remained an important and urgent problem. Early in 1836, Richard Bright was the first scholar to recommend controlling dietary protein consumption to protect kidney function.² Later, in 1882, Brenner et al.³ first proposed the hypothesis that low dietary protein could effectively ameliorate renal function deterioration. The hypothesis suggested that excessive protein intake would lead to hypertransfusion, hyperfiltration and hypertension of the glomeruli, thereby accelerating kidney function deterioration. Thus, dietary intervention has been an indispensable part of treatments for CKD over the years, and the findings of many studies and clinical practices have led to a dramatic improvement in the efficacy of low protein intake on pre-dialysis CKD patients.^{4, 5, 6, 7}

In the 1990s, instead of reducing protein intake, some interest was directed toward manipulating the quality of dietary protein, specifically by replacing animal protein with soy protein.⁸ Anderson et al.^{9, 10} proposed the soy protein hypothesis, which stated that substituting animal protein for soy protein resulted in reduced hyperfiltration and glomerular hypertension, with resultant protection from diabetic nephropathy (DN). Soy protein was shown to improve kidney function in animal models of polycystic kidney disease and in the rat remnant kidney model.^{11, 12, 13, 14, 15, 16} It was also shown to improve the lipid profile and reduce urinary albumin excretion in patients with nephrotic syndrome.^{17, 18, 19} Thus, there is a growing body of evidence indicating that soy protein consumption may have beneficial effects on kidney function. Furthermore, a number of studies have been carried out to explore the effect of soy protein on CKD patients. Nevertheless, the limited benefits and small sample size of these individual research projects have been recently challenged, and the findings of the relationship between soy protein intake and kidney function and the lipid profiles of CKD patients have not been conclusive. Therefore, we conducted a systematic meta-analysis of randomized controlled trials (RCTs) to quantitatively evaluate the overall effects of soy protein consumption compared with animal protein consumption on serum creatinine (SCR), triglyceride (TG), total cholesterol (TC), calcium (Ca) and serum phosphorus concentrations in pre-dialysis CKD patients.

Materials and methods

Search strategy

We searched the databases PubMed (http://www.ncbi.nlm.nih.gov/pubmed/), EMBASE (http://www.embase.com/), Cochrane Library (http://www.thecochranelibrary.com/) and Chinese Biological Medicine (CBM) (http://www.sinomed.ac.cn/) for published RCTs designed to assess the effects of soy protein intake on pre-dialysis CKD patients. We undertook our first comprehensive search in April 2013 and a second search for more recent data in March 2014. In our structured searches, we used the following keyword or MeSH word (medical subject heading) formats: (soy OR soybean OR soya OR soy protein) AND (chronic renal insufficiencies OR chronic kidney insufficiencies OR chronic kidney diseases OR chronic renal diseases OR diabetic nephropathies OR diabetic kidney diseases OR diabetic glomerulosclerosis) AND sensitive RCT filters (the sensitive search strategies used to ensure optimal collection of RCTs in electronic searches). We also limited the results to adult subjects. References listed in the searched articles were used for additional screening for relevant data.

Study selection

The inclusion criteria for selecting studies for this meta-analysis from the database search results were as follows:

• 1) The study was an RCT trial with either a crossover or a parallel design.

• 2) The target population was patients with CKD before dialysis.

• 3) The study used a concurrent control group; the treatment and control groups differed in the use of soy protein, or soybeans or soy products. The eligible treatments and controls were: (a) soybeans or soy products in the treatment diets that were replaced with usual foods or animal products in the control diets, and (b) isolated soy protein (ISP) compared with milk or animal protein.

• 4) The soy protein supplied by ISP or soybeans or soy products comprised more than 30% of the total protein in the treatment diets.

• 5) The initial or end point values for SCR or lipid measurements (or the differences between the values) and the s.d. or s.e. or 95% confidence intervals (CIs) of each group were available.

• 6) The study involved no other factors which may have had a potentially active effect on kidney function.

Data extraction and quality assessment

Two of the current authors independently evaluated the studies to consider whether the inclusion criteria had been met after primary screening. A piloted data extraction form was then used to collect the following data: the first author’s name; year of publication; gender and kidney function of participants; type of CKD (DN or non-DN); number of participants; type of intervention; diet protein and percentage of soy protein; duration of follow-up; study design; and trial qualities. Baseline data and final concentrations (or net changes) of SCR, TG, TC, Ca and phosphorus were also extracted. The main outcome was net changes in SCR. The secondary outcomes were net changes in lipid profiles (TG and TC) and blood electrolytes (serum Ca and phosphorus).

The criteria by which we assessed the quality of each study were: (1) allocation concealment, (2) randomization, (3) blindness (participants and outcome assessor), (4) compliance and (5) withdrawals. If allocation concealment, randomization and blindness were all coded ‘yes’, and if a compliance evaluation had been conducted, and the number of withdrawals plus reasons had been cited, then the study was considered at low risk of bias. If a study did not meet ⩾3 criteria, then it was considered at high risk of bias. Other studies were considered at moderate risk of bias.

Statistical analyses

Cochrane collaboration meta-analysis review methodology was used to conduct the meta-analysis. The net changes in continuous variables with formal distribution were described by weighted median differences or standard median differences. We used the results of each item obtained at the end of each intervention. The Q-statistic was used to evaluate heterogeneity between studies and I² was used to quantify the degree of inconsistency between studies.²⁰ We calculated the pooled effect size based on a random-effects model (DerSimonian and Laird method).²¹ The statistical software package Stata (version 12.0; StataCorp LP, College Station, TX, USA) was used for analyses.

We conducted subgroup analyses to explore the potential effect of study design (crossover or parallel), trial duration and type of CKD (DN or non-DN) on outcomes. Sensitivity analyses were also performed to assess the influence of the quality of trials on the overall effect sizes by eliminating trials rated with a high risk of bias. Where significant heterogeneity existed, sensitivity analyses were also conducted to find the source of the heterogeneity.

We used Egger’s linear regression analysis to evaluate the publication bias.²²

Results

Identification of relevant studies

Detailed processes of the study selection are shown in Figure 1. After searching the databases, there were 34 reports in all. We excluded 23 reports which were clearly not eligible by prime screening according to titles and abstracts. One additional article was discarded because the full text was not available. After reading the full texts of 10 studies carefully, we excluded one study because of its data mistakes. Finally, nine studies were selected for the meta-analysis.^{9,23, 24, 25, 26, 27, 28, 29, 30}

Figure 1

Flow diagram of identification of relevant articles. Pubmed (http://www.ncbi.nlm.nih.gov/pubmed/), EMBASE (http://www.embase.com/), Cochrane Library (http://www.thecochranelibrary.com/), CBM (http://www.sinomed.ac.cn/).

Study features

Table 1 shows the features of the nine studies that met the inclusion criteria. In one study patients in the treatment group received ISP powder, whereas in the other studies patients received soybeans or soy products as sources of soy protein. In six studies, patients were given 0.6–0.8 g/kg protein, of which 50–60% of the total protein was soy protein. In the other three studies, patients received 1 g/kg protein, of which 35% of the total protein was soy protein. Six studies involved CKD subjects with DN and the other three studies involved non-DN CKD subjects. In two trials, the patients were all male, while the other trials involved both males and females. Five trials lasted for 6–8 weeks, three trials for 6 months to 1 year, and one trial for 4 years. Five studies were designed as randomized crossover trials of short duration (6–8 weeks) and four studies were randomized parallel trials of longer duration (6 months to 4 years), suggesting that study duration was closely related to the study design.

Table 1 Features of the nine randomized controlled trials selected for analysis

Study quality

The study quality of the nine RCTs differed. Eight studies in which treatment group patients received soybeans or soy products were not blinded, and only one study that used ISP was blinded. Allocation concealment was adequate in eight studies and not carried out in one study. One study masked both participants and outcome assessors, seven studies masked only outcome assessors and one study was unclear. Compliance was assessed in eight studies, but was unclear in one study. Seven studies reported withdrawals and reasons for withdrawal, and two studies failed to report withdrawals. One study was judged to be at low risk of bias, seven studies at moderate risk of bias and one study at high risk of bias (Table 1).

Overall effects of soy consumption on SCR, TG, TC, Ca and serum phosphorus concentrations

Table 2 shows the overall effects of soy consumption. Eight studies reported data on SCR concentrations. Compared with animal protein, the consumption of soy protein significantly reduced SCR in CKD patients. The pooled estimated change in SCR was −6.231 μmol/l (95% CI: −11.109, −1.352 μmol/l; P=0.012) using a random-effects model (Figure 2). Five studies reported data on serum phosphorus concentrations. As different methods were used for phosphorus assays, the net change in phosphorus was described by standard median difference. We found that soy protein significantly reduced levels of serum phosphorus in CKD patients, with a pooled estimated change of −0.804 (95% CI: −1.143, −0.464; P=0.00) using a random-effects model (Figure 3).

Table 2 Pooled estimates of treatment effect of soy protein on SCR, TG, TC, Ca and serum phosphorus concentrations

Figure 2

Pooled estimated effect of soy protein on SCR. Horizontal lines represent 95% CIs, one of which extends beyond the limits of the scales. The sizes of the data markers indicate the weight of each study. The meta-analysis used the weighted mean difference (WMD) in the random-effects model.

Figure 3

Pooled estimated effect of soy protein on serum phosphorus. Horizontal lines represent 95% CIs. The sizes of the data markers indicate the weight of each study. The meta-analysis used the standard mean difference (SMD) in the random-effects model.

We noted that there were no significant effects on TG, TC and Ca concentrations compared soy protein with animal protein. Seven studies reported data on TG and TC concentrations and the pooled estimated effect was, respectively, −0.111 mmol/l (95% CI: −0.345, 0.122 mmol/l; P=0.351) and −0.135 mmol/l (95% CI: −0.289, 0.019 mmol/l; P=0.086) using a random-effects model. Four studies reported data on Ca concentrations and the pooled estimated effect was 0.023 mmol/l (95% CI: −0.016, 0.062 mmol/l; P=0.256) using a random-effects model.

There was evidence of inter-study heterogeneity for levels of TG (I²=49.7%), but not for SCR, TC, Ca and serum phosphorus levels (I²=11.8%, 0.0%, 0.0%, 0.0%, respectively).

Sensitivity and subgroup analysis

Table 3 shows the results of sensitivity and subgroup analyses for SCR, TG, TC, Ca and serum phosphorus concentrations.

Table 3 SCR, TG, TC, Ca, serum phosphorus by sensitivity analyses and subgroups analyses ranked by type of kidney disease and study design (duration)

The results of sensitivity analyses for SCR and serum phosphorus levels showed that the pooled effects remained the same after excluding the trial rated as having a high risk of bias (Mao, which referred to SCR and serum phosphorus data).²⁵

Subgroup analyses showed that the effects of soy protein on reducing SCR concentrations was greater in DN patients, with a net change of −8.202 μmol/l (95% CI: −14.646, −1.757 μmol/l), than in non-DN patients (5.5 μmol/l (95% CI: −15.281, 26.280 μmol/l)), but the subgroup difference was insignificant (P=0.31). There were no statistically significant differences in the pooled effects of the subgroups ranked by study design (crossover or parallel).

Considering the high heterogeneity for TG (I²=49.7%), we conducted a sensitivity analysis for TG and found that the pooled estimated effect was entirely different when we excluded the study by Soroka et al.²³ (−0.223 mmol/l (95% CI: −0.396, −0.051 mmol/l; P=0.011)), whereas the results were the same when other studies were excluded. These findings suggested the significant effect of soy protein consumption on reducing TG levels in CKD patients (Figure 4). We then performed subgroup analysis stratified by study design and types of CKD and noted that the effects of soy protein consumption on TG in DN patients and crossover trials (study duration <24 weeks) were greater than those in non-DN patients and parallel trials (study duration ⩾24 weeks). However, the subgroup differences were insignificant (P=0.37 and 0.65 for subgroup differences, respectively).

Figure 4

Pooled estimated effect of soy protein on TG after excluding the trial by Soroka et al.²³ Horizontal lines represent 95% CIs. The sizes of the data markers indicate the weight of each study. The meta-analysis used the weighted mean difference (WMD) in the random-effects model.

Also, we found the effects of soy protein on reducing TC concentrations were greater in parallel trials, with a net change of −0.435 μmol/l (95% CI: −0.775, 0.095 μmol/l), than in crossover trials (−0.058 μmol/l (95% CI: −0.230, 0.114 μmol/l)), but the subgroup difference was insignificant (P=0.052). No statistically significant differences were found in the pooled effects on TC in subgroups ranked by the type of CKD.

Differences in study design or CKD type did not appear to significantly influence pooled mean differences in Ca and serum phosphorus concentrations.

Publication bias

Egger’s regression test was used to evaluate the publication bias of the studies. The test suggested no significant asymmetry for the overall effect of SCR, TG, TC, Ca and serum phosphorus concentrations (P=0.397, 0.288, 0.126, 0.780, 0.956, respectively).

Discussion

First, this meta-analysis of nine RCTs indicated a significant efficacy of soy protein consumption in improving SCR and serum phosphorus concentrations in pre-dialysis CKD patients compared with animal protein consumption. Second, it indicated that soy protein may have a significant effect on lowering TG concentrations, but the result would be more conclusive following further investigations. Finally, the meta-analysis suggested that there was no significant efficacy of soy protein in improving TC and Ca concentrations.

In this meta-analysis, we conducted a sensitivity analysis on TG for its high heterogeneity and found that the pooled estimated effects were closely related to those found in the study by Soroka et al.²³ After re-reading the full text, we found that both the treatment group and the controlled group in the study of Soroka et al.²³ received three eggs per week—showing a clear difference from other studies. After discussing with experts, we decided to exclude this study from the analysis of TG and show the favorable effects of soy protein on lowering TG, which concurred with the results of some other studies.^{31, 32, 33} Remarkably, no differences were indicated by sensitivity analyses on other variables after excluding the study by Soroka et al.²³ In addition, subgroup analysis of TG, stratified by type of CKD and study design, suggested differences between the subgroups, but they were insignificant. Very few studies may result in such differences, so further studies are necessary.

Although several studies have shown that soy protein can significantly reduce cholesterol concentrations,^{31, 32, 33, 34} we failed to observe a significant reduction in TC concentrations in this meta-analysis. Nevertheless, we found that the effect of soy protein on lowering TC concentrations was greater in parallel trials (⩾24 weeks) than in crossover trials. This suggested that reduction in TC concentrations may be related to study duration. Hence, long-term RCTs (⩾24 weeks) are needed which aim to explore the effect of soy protein on lipid profiles in CKD patients. Additionally, the relatively small number of available trials and subjects studied limited the efficacy of our meta-analysis to show any significant change in TC levels. Moreover, several studies have shown that soy protein significantly reduces TC only in individuals with initially above-normal concentrations.^{35, 36, 37} Most CKD subjects in our meta-analysis had desirable and optimal cholesterol concentrations, which may have led to insignificant change in TC in this meta-analysis.

In this meta-analysis, we found a significant reduction in serum phosphorus concentrations, which may have important clinical implications, since hyperphosphatemia is an often intractable problem in pre-dialysis CKD patients.²³ The decrease is likely due to a combination of the reduced phosphate intake and reduced intestinal absorption. Phosphorus from animal proteins is well absorbed by humans, whereas that from plant sources is less well absorbed because much of the phosphorus is in the form of phytic acid, which is not well absorbed. If a significant reduction in phosphate intake and absorption can be achieved by the simple and tolerable method of replacing animal protein with soy protein, this could make the control of hyperphosphatemia easier and more successful.²³

Several potential limitations in the study merit consideration. The number of trials selected in the meta-analysis was small, so the pooled estimated changes may have been affected by certain studies, especially those with the highest weighting. In addition, most studies reported only on SCR concentrations, whereas other important variables for evaluating kidney function, such as glomerular filtration rate and 24-h urinary protein levels, were not considered. Glomerular filtration rate is a more reliable variable than SCR for reflecting kidney function, and 24-h urinary protein levels are essential for evaluating kidney function, especially early kidney damage. Furthermore, only 4–5 studies reported serum Ca and phosphorus concentrations, which are important for evaluating the Ca–phosphorus disorder that occurs in the later stages of CKD. All of these factors limited the efficiency of the meta-analysis. In addition, most of the trials included were judged to be of moderate quality. This is largely due to inadequate blinding of subjects from using either soy products or soy beans as the treatment intervention. Finally, although a variety of potentially conflicting factors were controlled in the original studies, it was feasible that uncontrolled conflicting factors still affected the outcomes.

Our analysis, however, had some strengths. First, only RCTs that followed rigorous protocols were included, thereby minimizing bias. Second, most of our results showed relatively little or no heterogeneity between the studies. Third, no evidence of publication bias of the studies was found using Egger’s regression test. Fourth, since there was a wide variation in the number and formulation of soy products used, and the control diet regimen varied remarkably across studies—although most of our results showed relatively little or no heterogeneity between the studies—we used a random-effects model to calculate the pooled effect size. Use of a random-effects model is less likely to produce significant results for pooled effect sizes than use of a fixed-effects model.²¹

In conclusion, soy protein consumption showed a notable favorable effect on reducing SCR and serum phosphorus concentrations compared with animal protein, suggesting a protective effect of soy protein consumption in pre-dialysis CKD patients. Moreover, it may also have a significant effect on reducing TG concentrations. Further well-reported RCTs are needed, with comprehensive clinical outcomes and long-term study durations, which are specifically designed to assess the effects of soy protein consumption on pre-dialysis CKD patients.