Nutrition 27 (2011) 10081016

Contents lists available at ScienceDirect

Nutrition
journal homepage: www.nutritionjrnl.com

Applied nutritional investigation

Meta-analysis of the effect of b-glucan intake on blood cholesterol and glucose

levels
Uma Tiwari Ph.D., Enda Cummins Ph.D. *
Institute of Food and Health, School of Agriculture, Food Science and Veterinary Medicine, University College Dublin, Beleld, Dublin, Ireland

a r t i c l e i n f o a b s t r a c t

Article history: Objective: A meta-analysis was performed on epidemiologic studies to assess the relation between
Received 26 July 2010 b-glucan consumption from oats and from barley on blood cholesterol level, triglyceride/tri-
Accepted 6 November 2010 acylglycerol (TGL/TAG) level, and blood glucose level (BGL) in humans. In addition, the effect of
b-glucan on total cholesterol (TC) and BGL was translated into an empirical doseresponse model.
Keywords: Methods: Thirty research articles that evaluated the effect of different exposure levels of b-glucan
b-Glucan on blood cholesterol and BGL were analyzed, yielding 126 clinical studies.
Blood cholesterol
Results: There was a signicant inverse relation in TC (
0.60 mmol/L, 95% condence interval [CI]
Blood glucose
Meta-analysis
0.85 to
0.34), low-density lipoprotein (
0.66 mmol/L, 95% CI
0.96 to
0.36), and TGL/TAG
Doseresponse (
0.04 mmol/L, 95% CI
0.15 to 0.07) after consumption of b-glucan. In contrast, an increase in
high-density lipoprotein cholesterol was noted (0.03 mmol/L, 95% CI
0.06 to 0.13) with the
random-effect model. The analysis showed a signicant change in BGL (
2.58 mmol/L, 95% CI

3.22 to
1.84) with high heterogeneity between (I2 97%) and across (s2 5.88) the studies. The
xed-effect model showed a signicant change in TC, low-density lipoprotein, and BGL, whereas it
showed no signicant changes in high-density lipoprotein and TGL/TAG. The doseresponse model
showed that a 3-g/d dose of oat or barley b-glucan was sufcient to decrease TC.
Conclusion: Consumption of 3 g/d of oat or barley b-glucan is sufcient to decrease blood
cholesterol, whereas the effect on BGL is still inconclusive, with high heterogeneity, and requires
further clinical research studies with longer intervention periods.
2011 Elsevier Inc. All rights reserved.

Introduction claim on whole grain oat or barley (and derivatives) b-glucan to

Consumption of soluble ber from cereals and cereal prod-
ucts has been credited with the ability to promote benecial
human health effects. Cereals such as oats and barley are rich in
soluble bers, e.g., b-glucan. b-Glucan is a non-starch poly-
saccharide composed of linear chains of glucose with b-(1/3)
and b-(1/4) linkages [1]. Oat and barley b-glucan are well
recognized for their many health claims and current research is
focused on increasing soluble ber consumption through dietary
intervention to address growing consumer awareness. The Food
and Drug Administration (FDA) authorizes the use of a health
This project was funded by the Irish Department of Agriculture and Food under
the Food Institutional Research Measure as part of the National Development
Plan.
* Corresponding author. Tel.: 353-1-716-7476; fax: 353-1-716-7415.
E-mail address: enda.cummins@ucd.ie (E. Cummins).
decrease the risk of coronary heart disease [2,3].
Epidemiologic studies have reported that the intake of
soluble ber (b-glucan) lowers lipid absorption and in turn
decreases the risk of cardiovascular diseases [4]. Consequently,
due to its physiologic effects, many researchers have incorpo-
rated barley or oat b-glucan into various food products, including
breakfast cereals, beverages, bread, and infant foods, to improve
the nutritional and health benets [58]. Many intervention
studies have reported a varying degree of effect from b-glucan
intake on blood glucose and cholesterol levels, possibly due to
variations in study design and sample sizes, thus making it
difcult to compare studies and draw any overall conclusions.
A meta-analysis is a tool that can aid in resolving these issues.
Meta-analysis is a powerful technique that can synthesize
results from different studies by integrating the study, design,
sample size, and intervention period and producing a broader
generalizability of the overall impact of a treatment [9]. Thus, the
purpose of this study was to conduct a meta-analysis to assess

0899-9007/$ - see front matter 2011 Elsevier Inc. All rights reserved.
doi:10.1016/j.nut.2010.11.006
 U. Tiwari, E. Cummins / Nutrition 27 (2011) 10081016 1009

the relation between b-glucan consumption from oats and from estimates that is due to heterogeneity rather than sampling error [11]. To
barley on cholesterol and blood glucose levels (BGLs). The anal- examine the heterogeneity, Q (equation 3) and I2 (equation 4) statistics were
performed.
ysis also focused on a continuous doseresponse for b-glucan to
provide a quantitative evaluation in a standardized format, !2
Pk b
permitting a numerical analysis across studies. i1 q i wi
X
k 2
Q b
q i wi
(3)
Pk
i1 i1 wi
Materials and methods
q i is an estimate of bq from the ith study and wi 1=varbq i .
where b
Study design
Q
df
I 2 %  100 (4)
The electronic database of Medline (from 1990 through December 2009) was Q
searched using the search terms blood cholesterol, b-glucan or cholesterol, blood
where df (k
1) is the degree of freedom.
glucose, and b-glucan or glycemic and b-glucan limited to humans (male and/or
The I2 statistics can also describe the percentage of the total variability across
female) and clinical studies (excluding animal and in vitro studies). Only those
studies due to true heterogeneity based on I2 values of low (25%), moderate
that were published as full-length articles and in English (language restriction)
(50%), and high (75%) heterogeneity [11]. The s2 statistic (equation 5),
were considered without any geographic restriction. The reference lists of the
which indicates variation between studies, was also evaluated as described by
retrieved articles were also examined. The Web-of-Science database was also
DerSimonian and Kacker [12] (equation 5). The higher I2 index results in a higher
examined using similar search terms and the search was further narrowed to
s2, indicating that I2 is directly related to s2.
cereals such as oats and barley.
Q
df
Inclusion and exclusion criteria s2 Pk Pk Pk if Q > df and s2 0 if Q  df (5)
i1 wi
i1 wi = i1 wi
2

The inclusion criteria were based on cohort studies, doseresponse studies, Publication bias was examined using the regression test of Egger et al. [13]
pre- and post-treatment studies, and reported study designs (randomized, and by visual inspection of funnel plots. The test for publication bias were
crossover, parallel, etc.), with or without subjects with health conditions. Clinical carried out by least-squares regression using the PROC GLM procedure (SAS 9.1,
trials involving healthy male or female, cholesterolemic or mildly cholester- SAS Institute, Cary, NC, USA) for the standardized estimate of treatment differ-
p p
olemic, diabetic or non-diabetic subjects and different age groups were included. ence (b
q wi ) and precision (1= wi ), where bq is the treatment difference and wi is
Types of blinding (open, single, or double), no intervention difference, and the estimated inverse variance of b q.
b-glucan dose (oats or barley food products) also met the inclusion criteria.
Studies that reported descriptive data with continuous outcome and mean 
standard deviation were selected for inclusion for the blood cholesterol meta- Continuous doseresponse modeling
analysis, whereas area under the curve (AUC) with mean  standard deviation
or standard error were selected for inclusion for the BGL meta-analysis. Eleven The doseresponse relation was studied between b-glucan and blood
studies for total cholesterol (TC) and low-density lipoprotein (LDL) cholesterol, cholesterol; the relation between b-glucan and blood glucose using the PROC
10 studies for triglyceride/triacylglycerol (TGL/TAG) and high-density lipoprotein NLIN program of SAS 9.1. The doseresponse curve was modeled using equation 6
(HDL) cholesterol, and 18 studies for BGL met the inclusion criteria. Studies that as described by Demonty et al. [14].
analyzed the b-glucan intake as a continuous variable were considered for this   
meta-analysis. Published research studies that involved in vitro studies were dose
Net changepred a 1
exp (6)
excluded. b=ln2

where net changepred is the predicted net change for blood cholesterol and
Data extraction and statistical analysis of studies glucose level and a and b are the parameters. Parameter a is the b-glucan dose at
which maximum net change can be achieved and parameter b is one-half the
Data extraction was performed to obtain the mean AUC and the standard previous dose effect needed to achieve an additional effect with an incremental
error or standard deviation. There are various methods for calculating AUC such dose step [14]. A Gauss-Newton algorithm was used to determine the parameters
as polynomial interpolation of third and fourth degrees, trapezoidal rule, Simp- of the mathematical equations adjusted by non-linear regression.
sons integration, and cubic interpolator splines [10]. According to Le Floch et al.
[10], the total AUC is a descriptive factor related to the blood glucose value,
whereas incremental and positive AUC more accurately describe glycemic
response to foods. In some studies, where the relevant data were available, the Results
trapezoid rule was used to calculate the AUC for a given blood glucose or insulin
concentration. The mean change in the AUC parameter was normalized to Quality of the studies
calculate the glycemic change per minute for the test and the control samples.
The net change for BGL, blood cholesterol (TC, LDL, and HDL), and TGL/TAG
Thirty research articles were chosen; some studies used two
were calculated as the standardized mean difference (SMD). The SMD is the mean
difference between the treated group and the control group divided by the or more b-glucan dose levels for oat and barley b-glucan as listed
pooled standard deviation (Hedgess g, equation 1) and corresponding with 95% in Tables 1 and 2. Therefore, the meta-analysis included 126
condence intervals (CIs) used for this continuous outcome: studies based on b-glucan dose levels. Of these 126 studies,
m
m 20 studies were selected for the TC and LDL meta-analysis,
SMD b
q T 2 C (1)
sb whereas 18 studies for HDL, 19 studies for TGL/TAG, and 49
s studies for blood glucose met the inclusion criteria. Of 126
sb2T bs2C studies, 82 studies were related to oat b-glucan and the other 44
sb2 (2)
2 studies were related to barley b-glucan. Some studies reported
where mT and mC are the population means in the treated and control groups and s b more than one trial with different consumption levels of
is the pooled standard deviation of the two groups. The pooled standard devia- b-glucan and these were considered as a separate study in this
tions sb are calculated as shown in equation 2, where s b and s
2 2 2
b represent the meta-analysis. Study design characteristics and participant
standard deviations of the treated and control groups, respectively.
number in addition to the level of oat or barley b-glucan
Overall pooled study estimates were assessed with the xed-effects and
random-effects estimators for blood cholesterol and glucose level. The xed- consumed are presented in Table 1 [5,7,8,1522] and Table 2 [5,6,
effect model assumes that all studies in the meta-analysis share a common 8,18,2240] and were included in the meta-analysis. The total
true size effect, whereas the random-effects model assumes that the studies were numbers of participants were 1154 for TC and LDL cholesterol
drawn from populations that differ from each other in ways that could affect the from 20 different studies, 1070 and 1000 participants from
treatment effect. Cochrans Q statistics were used to test for heterogeneity among
the trials [9]. Cochranes Q statistic may have low or high power depending on the
18 studies for HDL cholesterol, 1116 participant from 19 studies
number of studies included in the meta-analysis. The I2 statistics quantify for TGL/TAG, and 1250 participants for 49 different studies for
inconsistency across studies and describe the percentage of variability in point BGL. All studies included male and/or female subjects with
 1010
Table 1
Detailed study characteristics included in the meta-analysisdcholesterol*

Study ID Products b-Glucan dose (g/d) n Age (y) Sex Subjects Study design Intervention References
period (wk)
1 oat meal or oat bran oat meal : 1.2 or 2.4 or oat meal: 20 or 21 or 20, 5055 M/F low-density lipoprotein controlled, single-blinded, 6 [15]
3.6; oat bran: 2 or 4 or 6 oat bran: 23 or 20 or21 cholesterol doseresponse
2 barley 3 or 6 27 M/F mild hypercholesterolemia Latin square design 5 [16]
3 oat cereal 3 150 30w70 M/F healthy randomized 6 [17]
4 oat ready 4 22 35w72 MF healthy parallel 3 [18]
mealsoup
5 oat muesli 5 40 18w65 M/F mild hypercholesterolemia randomized, double-blinded, 4 [7]
multiple crossover
6 oat 5 25 18w70 M/F normal healthy double-blinded, parallel 3 [19]
7 barley/oat beverage 5 or 10 19/16 18w70 MF healthy randomized, dose-controlled 5 [5]
8 oat cereal 5.5 17 50-75 M healthy randomized 12 [20]
9 oat 6 35 22w65 M/F hypercholesterolemia randomized 6 [21]

U. Tiwari, E. Cummins / Nutrition 27 (2011) 10081016

10 barley 7 19 33w50 M hypercholesterolemia randomized, double-blinded 12 [8]
11 barley 8 21 30w59 M mild hypocholesterolemia crossover 4 [22]

F, female; M, male
* Cholesterol includes total cholesterol, triglycerides or triacylglycerol, low-density lipoprotein, and high-density lipoprotein.

Table 2
Characteristics of studies included in the meta-analysis for blood glucose level

Study ID Products b-Glucan n Age Sex Subjects Study Intervention References

dose (g/d) (y) design period
4 oat ready mealsoup 4 22 35w72 M F healthy parallel 3 wk [18]
7 barley beverage/oat beverage 5 or 10 19 or 16 18w70 M F healthy randomized, dose-controlled 5 wk [5]
11 barley food 8 21 30w59 M mild hypocholesterolemia crossover design 4 wk [22]
12 oat gum 2 or 4 or 7 9 29w34 M F healthy doseresponse 3h [23]
13 barley bread; oat or barley porridge 8 or 14, 2 or 8 9 24w46 M F healthy random order 2w3 h [24]
14 barley bread 2 or 4 or 7 8 20w27 M F healthy random order 2h [25]
15 barley/oat tempe 2 12 25w75 M F healthy randomized block design 2h [26]
16 oat/barley bran and extracts 3 or 6 for oats and 20 35w57 M F healthy Latin square design 2w3 h [27]
5 for barley
17 oat/barley our and akes 3 or 12 10 28w58 F healthy, overweight Latin square design 3h [28]
18 oat bran: crisp and our 3 or 9 12 59w73 MF type 2 diabetes randomized, controlled, 3h [29]
repeated measure
19 oat muesli 3 or 4 19 22w53 MF healthy 2-series test 2h [30]
20 prototype b-glucan breakfast/bars 6 or 7 or 4 16 59w63 MF type 2 diabetes randomized, crossover 3h [31]
and oat bran
21 barley crackers/cookies 4 10 25w26 M F healthy randomized 2  2 2 h [6]
22 oat bran akes 4 12 22w35 M F healthy randomized, crossover, double-blinded 1 h [32]
23 oat bran mufns 4 or 8 10 32w43 M F healthy randomized block 3 h [33]
24 barley bread 6 or 8 or 12 10 23w27 M healthy random order 3 h [34]
25 oat concentrates drink 6 11 29w39 M F healthy randomized, crossover, double-blinded 2 h [35]
26 barley food/drink 6 18 21w33 M healthy randomized crossover 6 h [36]
27 oat bran mufns 8 or 12 11 20w50 M F healthy randomized block 3 h [37]
28 barley breakfast w12 15 23w29 M F healthy random order 3 h [38]
29 OBC (bread products) 9 8 45w47 M NIDDM crossover study 8 h [39]
30 barley pasta 12 5 30w60 M F healthy non-random order 2 h [40]

F, female; M, male; NIDDM, noninsulin-dependent diabetes mellitus; OBC, oat bran concentrate
Table 3
Results of the meta-analysis of study group to evaluate effect of b-glucan on blood cholesterol and glucose response

Study groups TC (mmol/L) LDL (mmol/L) HDL (mmol/L) TGL/TAG (mmol/L) BGL (mmol/L)
Pooled studies
No. of studies 20 20 18 19 49
Participants 1154 1154 1000 1116 1250
Fixed effect
Net change (95% CI)
0.59 (
0.71 to
0.48)
0.65 (
0.77 to
0.53) 0.02 (
0.10 to 0.14)
0.07 (
0.19 to 0.04)
2.48 (
2.59 to
2.37)
Heterogeneity
Q statistics 82.8 117.8 4.8 9.6 1843.1
P <0.0001 <0.0001 0.9983 0.9445 <0.0001
I2 (%) 77 84 0 0 97
Random effect
Net change (95% CI)
0.6 (
0.85 to
0.34)
0.66 (
0.96 to
0.36) 0.03 (
0.06 to 0.13)
0.04 (
0.15 to 0.07)
2.53 (
3.22 to
1.84)
Heterogeneity
s2 0.25 0.38 0.02 0.00 5.88
Excluded studies 3 4 N/A N/A 4
Fixed effect

U. Tiwari, E. Cummins / Nutrition 27 (2011) 10081016

Net change (95% CI)
0.38 (
0.51 to
0.26)
0.37 (
0.49 to
0.24) N/A N/A
2.02 (
2.13 to
1.90)
Heterogeneity N/A N/A
Q statistics 8.6 7.4 N/A N/A 1001.8
P 0.9380 0.9470 N/A N/A <0.0001
I2 (%) 0 0 N/A N/A 96
Random effect
Net change (95% CI)
0.35 (
0.47 to
0.22)
0.32 (
0.44 to
0.20) N/A N/A
2.05 (
2.60 to
1.50)
P <0.0001 <0.0001 N/A N/A <0.0001
Heterogeneity N/A N/A
s2
0.03
0.04 N/A N/A 3.41
Subgroups
Oat intake studies 14 14 14 14 26
Fixed effect
Net change (95% CI)
0.53 (
0.66 to
0.40)
0.56 (
0.69 to
0.42) 0.02 (
0.12 to 0.15)
0.06 (
0.20 to 0.07)
2.56 (2.71 to
2.41)
Heterogeneity
Q statistics 40.6 41.8 4.7 6 782.1
P <0.0001 <0.0001 0.9809 0.9461 <0.0001
I2 (%) 68 69 0 0 97
Random effect
Net change (95% CI)
0.55 (
0.80 to
0.30)
0.58 (
0.84 to
0.33) 0.03 (
0.09 to 0.16)
0.24 (
0.39 to
0.08) 0.43 (
3.46 to
1.78)
P <0.0001 <0.0001 <0.0001 <0.0001 <0.0001
Heterogeneity
s2 0.14 0.15 0 0 4.63
Barley intake studies 6 6 4 5 23
Fixed effect
Net change (95% CI)
0.80 (1.04 to
0.55)
0.98 (
1.22 to
0.74) 0.03 (
0.27 to 0.34)
0.18 (
0.29 to
0.07)
2.39 (
2.56 to
2.23)
Heterogeneity
Q statistics 38.7 66.9 0.07 3.4 1058.9
P <0.0001 <0.0001 0.9951 0.4860 <0.0001
I2 (%) 51 93 0 0 98
Random effect
Net change (95% CI)
0.72 (
1.11 to
0.33)
0.87 (
0.27 to
0.47) 0.03 (
0.18 to 0.24)
0.18 (
0.29 to
0.07)
2.43 (
3.57 to
1.28)
P <0.0001 <0.0001 0.0167 <0.0001 <0.0001
Heterogeneity
s2 0.63 1.16 0 0 7.7

BGL, blood glucose level; CI, condence interval; HDL, high-density lipoprotein; LDL, low-density lipoprotein; N/A, not applicable; TAG, triacylglycerols; TC, total cholesterol; TG, triglyceride

1011
1012 U. Tiwari, E. Cummins / Nutrition 27 (2011) 10081016

A
Study 1a
Study 1b
Study 1c
Study 2a
Study 3
Study 1d
Study 1e
Study 4
Study 5
Study 6
Study 7a
Study 7b
Study 8
Study 9
Study 2b
Study 1f
Study 10
Study 11
Study 7c
Study 7d

Fixed effect
Random effect

-2.5 -2.0 -1.5 -1.0 -0.5 0.0 0.5 1.0

B Study 1a
Study 1b
Study 1c
Study 2a
Study 3
Study 1d
Study 1e
Study 4
Study 5
Study 6
Study 7a
Study 7b
Study 8
Study 9
Study 2b
Study 1d
Study 11
Study 7c
Study 7d

Fixed effect
Random effect

-1.0 -0.5 0.0 0.5 1.0

Net change (mmol/l)

Fig. 1. Net change and 95% condence interval plot for effect of b-glucan on (A) total cholesterol and (B) triglyceride/triacylglycerol.

b-glucan intake from oat- or barley-based products, with
consumption levels ranging from 2 to 14 g of b-glucan per day.
Net change in blood cholesterol and glucose level
The meta-analysis assessed net change in blood cholesterol
level and BGLs with the consumption of oat or barley b-glucan, as
presented in Table 3. The mean net change among individual
studies for TC was
2.08 to
0.03 mmol/L. The overall pooled
estimate for the xed- and random-effect models showed net
changes of
0.59 mmol/L (95% CI
0.71 to
0.48, P < 0.0001)
and
0.60 mmol/L (95% CI
0.85 to
0.34, P < 0.0001; Fig. 1). The
analysis indicated a substantial heterogeneity, with I2 77%
across the studies with the xed-effect model, whereas a small
variance was observed between studies, with s2 0.25 for the
random-effect model. The high heterogeneity may be due to
signicant variations in the study design.
The mean net change among individual studies for LDL
cholesterol was
2.46 to
0.04 mmol/L, with overall pooled
estimates of
0.65 mmol/L (95% CI
0.77 to
0.53, P < 0.0001)
for the xed-effect model and
0.66 mmol/L (95% CI
0.96 to

0.36, P < 0.0001) for the random-effect model (Fig. 2). Similar
 U. Tiwari, E. Cummins / Nutrition 27 (2011) 10081016 1013

A
Study 1a
Study 1b
Study 1c
Study 2a
Study 3
Study 1d
Study 1e
Study 4
Study 5
Study 6
Study 7a
Study 7b
Study 8
Study 9
Study 2b
Study 1f
Study 10
Study 11
Study 7c
Study 7d

Fixed effect
Random effect

- 3.0 -2.5 -2.0 -1.5 -1.0 -0.5 0.0 0.5 1.0

B Study 1a
Study 1b
Study 1c
Study 2a
Study 3
Study 1d
Study 1e
Study 4
Study 5
Study 6
Study 7a
Study 7b
Study 8
Study 9
Study 2b
Study 11
Study 7c
Study 7d

Fixed effect
Random effect

-1.5 -1.0 -0.5 0.0 0.5 1.0

Net change (mmol/l)

Fig. 2. Net change and 95% condence interval plot for effect of b-glucan on (A) low-density lipoprotein cholesterol and (B) high-density lipoprotein cholesterol.

to TC, the analysis of LDL cholesterol indicated a considerable
heterogeneity, with I2 84% across studies for the xed-effect
model, with a small variance between studies, with s2 0.38
for the random-effect model. By excluding three studies from TC
and four studies from LDL cholesterol, the I2 and s2 statistics
decreased to zero, indicating no heterogeneity in TC and LDL
cholesterol. This indicates that most of the heterogeneity was
due to variations in study design and can be decreased by
excluding outlier studies.
In the case of HDL cholesterol, the mean net change among
individual studies was
0.60 to 0.18 mmol/L, with overall pooled
estimates of 0.02 mmol/L (95% CI
0.10 to 0.14, P 0.99)
for the xed-effect model and 0.03 mmol/L (95% CI
0.06 to 0.13,
P < 0.0001) for the random-effect model (Fig. 2). HDL cholesterol
showed no heterogeneity for the xed-effect or for the random-
effect model. For TGL/TAG, the mean net change among indi-
vidual studies was
0.69 to 0.19 mmol/L, and overall pooled
estimates were
0.07 mmol/L (95% CI
0.19 to 0.04, P 0.94) for
the xed-effect model and
0.04 mmol/L (95% CI
0.15 to 0.07,
P < 0.0001) for the random-effect model, indicating no sign of
heterogeneity among studies (Fig. 1).
The mean net change for BGL showed a wide variation within
individual studies compared with blood cholesterol levels
ranging from
10.33 to 0.21 mmol/L, with an overall pooled
estimate for the xed-effect model at
2.48 mmol/L (95% CI

2.59 to
2.37, P < 0.0001), with high heterogeneity (I2 97%)
1014 U. Tiwari, E. Cummins / Nutrition 27 (2011) 10081016

Study 12a
Study 13a
Study 13b
Study 14a
Study 15a
Study 15b
Study 16a
Study 17a
Study 17b
Study 18a
Study 19a
Study 12b
Study 14b
Study 20a
Study 21a
Study 21b
Study 4
Study 22
Study 19b
Study 23a
Study 7a
Study 7b
Study 16b
Study 16c
Study 20b
Study 16d
Study 24a
Study 25
Study 26a
Study 26b
Study 12c
Study 14c
Study 20c
Study 11
Study 13c
Study 13d
Study 27a
Study 28
Study 23b
Study 29
Study 18b
Study 7c
Study 7d
Study 30
Study 17c
Study 17d
Study 24b
Study 27b
Study 13e

Fixed effect
Random effect

-11.0 -8.0 -5.0 -2.0 1.0

Net change (mmol/l)
Fig. 3. Net change and 95% condence interval plot for effect of b-glucan on blood glucose level.

across studies (Table 3). Overall pooled estimate for the random-
effect model was
2.53 mmol/L (95% CI
3.22 to
1.84, P <
0.0001) with a high s2 value equal to 5.88, which indicates a high
level of heterogeneity (Fig. 3). Hence, the xed- and random-
effect models showed considerable heterogeneity for all 49
studies. However, by excluding the most prominent outliers (four
studies) for the xed-effect (I2 96%) and random-effect (s2
3.41) models, the heterogeneity remained the same. This indi-
cates that the studies may have large variation in study design.
Similarly, Higgins et al. [11] reported that statistical heteroge-
neity is inevitable due to clinical and methodologic diversity,
which always occurs in a meta-analysis.
For consuming oat or barley b-glucan individually,
the random-effect model showed that oat b-glucan signicantly
(P < 0.0001) decreased blood cholesterol levels. Similar obser-
vations were noted for barley b-glucan, although there was no
signicant difference seen for HDL cholesterol. The heteroge-
neity level in LDL cholesterol was notably high for oat and barley
b-glucan (oat b-glucan, I2 69%, s2 0.15; barley b-glucan, I2
93%, s2 1.16). The BGL effects on oat and barley b-glucan
consumed separately were
2.62 and
2.43 mmol/L for oat and
barley intakes, respectively. The heterogeneity was considerably
high, with I2 97% and s2 4.63 for oat intake and I2 98% and
s2 7.70 for barley intake, for BGL (Table 3).
Eggers regression test suggest no signicant publication
bias (P 0.18, P 0.25, P 0.73, P 0.64, and P 0.55 for
TC, LDL, HDL, TGL/TAG, and BGL, respectively). Eggers
regression test was used to measure funnel plot asymmetry
and indicated no substantial evidence of a publication bias.
SAS 9.1 was used for this analysis. In addition, when Eggers
regression plot (not shown) was observed, it showed that HDL
cholesterol increases with b-glucan intake, as evident from
a positive intercept of 0.11, thus indicating no publication bias.
This method is based on the fact that when there is no
publication bias the intercept is expected to be zero; hence,
deviations from this point (positive or negative) will indicate
different degrees of publication bias [9]. In contrast, b-glucan
intake was inversely proportional to TC, LDL, and TGL/TAG,
which showed negative intercept values of
0.20,
0.40, and

0.04 , respectively, which also indicated the absence of
publication bias. Similarly, the consumption of b-glucan indi-
cated benecial effects by decreasing TC, LDL, and TGL/TAG.
However, visual inspection of the funnel plot (not shown)
indicated publication bias among studies when plotted against
the standardized estimate on inverse standard error (preci-
sion). The differential levels of publication bias observed in
this study indicate how bias affects the estimation of the
heterogeneity parameter between studies (I2) and across
studies (s2) as reported by Jackson [41].
Effect of b-glucan dose on TC and BGL
The net change in TC and BGL against the b-glucan dose
(grams per day) is shown in Figure 4. The mean b-glucan doses in
the study were 1.2 to 10.0 g/d for TC and 2.0 to 14.0 g/d for BGL.
A b-glucan dose of 1 g/d produced changes in TC and BGL of

0.079 mmol/L (range
0.236 to
0.006 mmol/L) and
0.084
mmol/L (range
0.61 to 0.011 mmol/L), respectively. However,
no statistically signicant doseresponse relation was observed
for net change in LDL, HDL, or TGL/TAG.
The predicted doseresponse curve for the relation between
b-glucan dose (grams per day) and net change for TC and BGL can
be obtained from equations 7 and 8 by tting model parameters
(a and b) to equation 6 describing the observed net change for TC
 U. Tiwari, E. Cummins / Nutrition 27 (2011) 10081016 1015

cholesterol by 0.03 mmol/L (P < 0.0001) based on the random-

A 0.2 TC-Net Change BGL -Net Change
Predicted 5th Percentile 95th Percentile effect model. The observed decrease in cholesterol may be due
0.0
to an increase in bile acid excretion induced by b-glucan [19]. The
meta-analysis showed higher I2 and s2 values for TC and LDL
cholesterol, indicating high heterogeneity and variation among
-0.2
Net change (mmol/l)

studies and across studies. However, by excluding three outlier

studies from TC and 4 studies from LDL cholesterol, the analysis
-0.4
showed an absence of heterogeneity among studies. Intake of oat
or barley b-glucan separately showed signicant (P < 0.0001)
-0.6
decreases in TC, LDL cholesterol, and TGL/TAG for oat or barley b-
glucan. HDL cholesterol increased signicantly (P < 0.0001) with
-0.8
oat b-glucan, whereas no signicant change was observed with
barley b-glucan intake when investigated separately. These
-1.0
results are in agreement with those of Li et al. [4] who showed an
0.0 2.0 4.0 6.0 8.0 10.0 12.0 14.0
B 0.5 association between barley b-glucan consumption and a decrease
in cholesterol concentrations.
0.0 The meta-analysis showed signicant changes in BGL of

2.53 mmol/L (random-effect model) with high heterogeneity
-0.5 (I2 97% and s2 5.88) after consuming oat and barley
Net change (mmol/l)

b-glucan. A separate meta-analysis for oat and barley b-glucan

-1.0
-1.5
-2.0
-2.5
0 2 4 6 8 10
-glucan dose (g/day)
Fig. 4. Net change in (A) TC and (B) BGL with corresponding b-glucan dose. BGL,
blood glucose level; TC, total cholesterol.
and BGL. The best-t model parameters were: a
0.30 (stan-
dard error 0.04, CI
0.39 to
0.20) and b
0.37 (standard error
1.0, CI
2.48 to 1.74) for TC and a
0.45 (standard error 0.10, CI

0.66 to
0.24) and b
2.64 (standard error 1.27, CI
3.82 to
1.29) for BGL.
 
dose
Net changepredTC
0:30  1
exp
0:37=ln2
 
dose
Net changepredBGL
0:45  1
exp
2:64=ln2
Based on equations 7 and 8, the predicted net change
(millimoles per liter) for TC (Fig. 4a) and BGL (Fig. 4b) and the CI
were obtained with the corresponding b-glucan dose (grams per
day). The predicted doseresponse models for TC and BGL were
found to be signicant (P < 0.0001). The doseresponse indi-
cated average net changes from predicted values of
0.28
mmol/L per day for TC and 0.32 mmol/L per day for BGL.
Discussion
This meta-analysis of 126 studies demonstrates the inuence
of oat and barley b-glucan consumption on TC, LDL, HDL, TGL/TAG,
and BGL in humans. This study also shows a continuous dose
response relation of b-glucan dose (grams per day) on TC and BGL.
Signicant changes were observed by the meta-analysis for blood
cholesterol and glucose with b-glucan consumption. The results
of this meta-analysis suggest that oat and barley b-glucan
consumption helps lower TC by
0.60 mmol/L, LDL cholesterol by

0.66 mmol/L, TGL/TAG by
0.04 mmol/L and increase HDL
showed high heterogeneity among studies, indicating more
variation among BGL studies. However, by excluding four
studies from the 49 studies, the analysis showed no change in
heterogeneity in BGL. Several research studies have related
b-glucan consumption to maintenance of a normal BGL, yet
the effect of b-glucan in maintaining a normal BGL was not
conclusive based on the present meta-analysis due to signi-
12 14 cant variations among studies. A recent report by the Euro-
pean Food Safety Authority (EFSA) [42] concluded that the
proposed claim lacks evidence to show relation between the
consumption of b-glucans and long-term maintenance of
normal blood glucose level.
The doseresponse model demonstrated a net change in TC
and BGL with a corresponding increase in b-glucan dose (Fig. 4).
The overall predicted net change for TC was
0.28 mmol/L and
that for BGL was
0.32 mmol/L, with no signicant change in TC
observed when increasing the b-glucan dose above 3 g/d. A 3-g/
d dose of b-glucan produced a net change for TC of
0.30 mmol/L
(CI
0.17 to
0.42). This is in agreement with FDA recommen-

dation of 3 g of b-glucan per day to lower the risk of cardiovas-
(7)
cular diseases. However, the predicted net change for BGL
showed uncertain changes with b-glucan dosage, which may be
 due to greater variation in studies.
(8) Although, the doseresponse curve showed a decrease in BGL
with increasing b-glucan level, the EFSA has claimed that there is
insufcient evidence to show the long-term maintenance of
normal BGL after b-glucan consumption [42].
The meta-analysis conducted in this study was based on
published research clinical studies. Observed heterogeneity
among study results may in part be related to different accuracy
levels in the assessment of b-glucan intake across studies due to
different food matrices. In addition, statistical heterogeneity is
inevitable due to the wide variation in study design and in clinical
studies as indicated by Higgins et al. [11]. However, Castro et al.
[43] adopted multivariate techniques to evaluate the results and
to decrease heterogeneity. This study highlights how meta-
analysis can be used as a tool to integrate experimental results
to yield a concerted evaluation of the impact of a treatment on
biological response levels across sets of empirical studies by
evaluating the effects of potentially moderating variables.
This study concludes that consumption of 3 g/d of oat or
barley b-glucan is sufcient to decrease blood cholesterol levels.
Further analysis of the b-glucan effect on BGL requires longer
intervention periods as recommended by the EFSA.
1016 U. Tiwari, E. Cummins / Nutrition 27 (2011) 10081016

References [24] Liljeberg HGM, Granfeldt EY, Bjo rck IME. Products based on a high bre
barley genotype, but not on common barley or oats, lower postprandial
glucose and insulin responses in healthy humans. J Nutr 1996;126:
[1] Tiwari U, Cummins E. Factors inuencing b-glucan levels and molecular
45866.
weight in cereal-based products. Cereal Chem 2009;86:290301.
[25] Cavallero A, Empilli S, Brighenti F, Stanca AM. High (1/3, 1/4)-b-glucan
[2] Food and Drug Administration (FDA). Final rule for food labeling: health
barley fractions in bread making and their effects on human glycemic
claims. Oats and coronary heart disease. Fed Reg 1997;62:35843681.
response. J Cereal Sci 2002;36:5966.
[3] Food and Drug Administration (FDA). Health claims: soluble bre from
[26] Alminger M, Eklund-Jonsson C. Whole-grain cereal products based on
certain foods and coronary heart disease (CHD). Fed. Reg 2005;70:7615062.
a high-bre barley or oat genotype lower post-prandial glucose and insulin
[4] Li J, Kaneko T, Qin L-Q, Wang J, Wang Y. Effects of barley intake on glucose
responses in healthy humans. Eur J Nutr 2008;47:294300.
tolerance, lipid metabolism, and bowel function in women. Nutrition
[27] Hallfrisch J, Scholeld DJ, Behall KM. Physiological responses of men and
2003;19:9269.

rklund M, van Rees A, Mensink RP, Onning women to barley and oat extracts (Nu-trim X). II. Comparison of glucose
[5] Bio G. Changes in serum lipids
and insulin responses. Cereal Chem 2003;80:803.
and postprandial glucose and insulin concentrations after consumption of
[28] Behall KM, Scholeld DJ, Hallfrisch J. Comparison of hormone and glucose
beverages with b-glucans from oats or barley: a randomised dose-
responses of overweight women to barley and oats. J Am Coll Nutr
controlled trial. Eur J Clin Nutr 2005;59:127281.
2005;24:1828.
[6] Casiraghi MC, Garsetti M, Testolin G, Brighenti F. Post-prandial responses to
[29] Tapola N, Karvonen H, Niskanen L, Mikola M, Sarkkinen E. Glycemic
cereal products enriched with barley b-glucan. J Am Coll Nutr
responses of oat bran products in type 2 diabetic patients. Nutr Metab
2006;25:31320.
Cardiovasc Dis 2005;15:25561.
[7] Theuwissen E, Mensink RP. Simultaneous intake of b-glucan and plant
[30] Granfeldt Y, Nyberg L, Bjo rck I. Muesli with 4 g oat b-glucans lowers
stanol esters affects lipid metabolism in slightly hypercholesterolemic
glucose and insulin responses after a bread meal in healthy subjects. Eur J
Subjects. J Nutr 2007;137:5838.
Clin Nutr 2008;62:6007.
[8] Shimizu C, Kihara M, Aoe S, Araki S, Ito K, Hayashi K, et al. Effect of high b-
[31] Jenkins AL, Jenkins DJA, Zdravkovic U, Wu rsch P, Vuksan V. Depression of
glucan barley on serum cholesterol concentrations and visceral fat area in
the glycemic index by high levels of b-glucan bre in two functional foods
Japanese menda randomized, double-blinded, placebo-controlled trial.
tested in type 2 diabetes. Eur J Clin Nutr 2002;56:6228.
Plant Foods Hum Nutr 2008;63:215.
[32] Hlebowicz J, Darwiche G, Bjo rgell O, Alme
r L- O. Effect of muesli with 4 g
[9] Whitehead A. Meta-analysis of controlled clinical trials. Chichester, UK:
oat b-glucan on postprandial blood glucose, gastric emptying and satiety in
John Wiley & Sons; 2002. p. 601, 204.
healthy subjects: a randomized crossover trial. J Am Coll Nutr
[10] Le Floch JP, Escuyer P, Baudin E, Baudon D, Perlemuter L. Blood glucose area
2008;27:4705.
under the curve. Methodological aspects. Diabetes Care 1990;13:1725.
[33] Tosh SM, Brummer Y, Wolever TMS, Wood PJ. Glycemic response to oat
[11] Higgins JPT, Thompson SG, Deeks JJ, Altman DG. Measuring inconsistency
bran mufns treated to vary molecular weight of b-glucan. Cereal Chem
in meta-analyses. BMJ 2003;327:55760.
2008;85:2117.
[12] DerSimonian R, Kacker R. Random-effects model for meta-analysis of
[34] Ostman E, Rossi E, Larsson H, Brighenti F, Bjorck I. Glucose and insulin
clinical trials: an update. Contemp Clin Trials 2007;28:10514.
responses in healthy men to barley bread with different levels of (1/3;
[13] Egger M, Smith GD, Schneider M, Minder C. Bias in meta-analysis detected
1/4)-b-glucans; predictions using uidity measurements of in vitro
by a simple, graphical test. BMJ 1997;13:62934.
enzyme digests. J Cereal Sci 2006;43:2305.
[14] Demonty I, Ras RT, van der Knaap HCM, Duchateau GSMJE, Meijer L, Zock PL,
[35] Panahi S, Ezatagha A, Temelli F, Vasanthan T, Vuksan V. b-Glucan from two
et al. Continuous dose-response relationship of the LDL-cholesterol
sources of oat concentrates affect postprandial glycaemic in relation to the
lowering effect of phytosterol intake. J Nutr 2009;139:27184.
level of viscosity. J Am Coll Nutr 2007;26:63944.
[15] Davidson M, Dugan LD, Burns JH, Bova J, Story K, Drennan KB. The hypo-
[36] Poppitt SD, van Drunen JDE, McGill A-T, Mulvey TB, Leahy FE. Supple-
cholesterolemic effects of b-glucan in oatmeal and oat bran: a dose-
mentation of a high-carbohydrate breakfast with barley b-glucan improves
controlled study. JAMA 1991;265:18339.
postprandial glycaemic response for meals but not beverages. Asia Pac J
[16] Behall KM, Scholeld DJ, Hallfrishch J. Diets containing barley signicantly
Clin Nutr 2007;16:1624.
reduce lipids in mildly hypercholesterolemic men and women. Am J Clin
[37] Lan-Pidhainy X, Brummer Y, Tosh SM, Wolever TM, Wood PJ. Reducing
Nutr 2004;80:118593.
beta-glucan solubility in oat bran mufns by freeze-thaw treatment
[17] Karmally W, Montez MG, Palmas W, Martinez W, Branstetter A,
attenuates its hypoglycaemic effect. Cereal Chem 2007;84:5127.
Ramkrishnan R, et al. Cholesterol-lowering benets of oat-containing
[38] Nilsson AC, Ostman EM, Holst JJ, Bjo rck IME. Including indigestible
cereal in Hispanic Americans. J Am Diet Assoc 2005;105:96770.

rklund M, Holm J, Onning carbohydrates in the evening meal of healthy subjects improves glucose
[18] Bio G. Serum lipids and postprandial glucose and
tolerance, lowers inammatory markers, and increases satiety after
Insulin levels in hyperlipidemic subjects after consumption of an oat b-
a subsequent standardized breakfast. J Nutr 2008;138:7329.
glucancontaining ready meal. Ann Nutr Metab 2008;52:8390.
[39] Pick ME, Hawrysh ZJ, Gee MI, Toth E, Garg ML, Hardin RT. Oat bran
[19] Naumann E, van Rees AB, Onning G, Oste R, Wydra M, Mensink RP. b-
concentrate bread products improve long term control of diabetes: a pilot
Glucan incorporated into a fruit drink effectively lowers serum LDL-
study. J Am Diet Assoc 1996;96:125461.
cholesterol concentrations. Am J Clin Nutr 2006;83:6015.
[40] Yokoyama WH, Hudson CA, Knuckles BE, Chiu MCM, Sayre RN, Turnlund J.
[20] Davy BM, Davy KP, Ho RC, Beske SD, Davrath LR, Melby LC. High-bre oat
Schneeman, RO. Effect of barley b-glucan in durum wheat pasta on human
cereal compared with wheat cereal consumption favorably alters LDL-
glycaemic response. Cereal Chem 1997;74:2936.
cholesterol subclasses and particle numbers in middle-aged and older
[41] Jackson D. The implications of publication bias for meta-analysis other
men. Am J Clin Nutr 2002;76:3518.
parameter. Stat Med 2006;25:291121.
[21] Queenan KM, Stewart ML, Smith KN, Thomas W, Fulcher RG, Slavin JL.
[42] European Food Safety Authority (EFSA). Scientic opinion on the
Concentrated oat b-glucan, a fermentable bre, lowers serum cholesterol
substantiation of health claims related to beta-glucans and maintenance
in hypercholesterolemic adults in a randomised controlled trial. Nutr J
or achievement of normal blood glucose concentrations (ID 756,
2007;6:18.
802, 2935) pursuant to Article 13(1) of Regulation (EC) No 1924/2006.
[22] McIntosh GH, Whyte J, McArthur R, Nestel JP. Barley and wheat foods:
2010.
inuence on plasma cholesterol concentrations in hypercholesterolemic
[43] Castro IA, Monteiro VCB, Barroso LP, Bertolami MC. Effect of eicosa-
men. Am J Clin Nutr 1991;53:12059.
pentaenoic/docsahexaenoic fatty acids and soluble bers on blood lipids
[23] Wood PJ, Braaten JT, Scott FW, Riedel KD, Wolynetz MS, Collins MW. Effect of
of individuals classied into different levels of lipidemia. Nutrition
dose and modication of viscous properties of oat gum on plasma glucose
2007;23:12737.
and insulin following an oral glucose load. Br J Nutr 1994;72:73143.