Diabetes Care 1

Should Viscous Fiber Supplements Elena Jovanovski,1,2 Rana Khayyat,1,2 Andreea Zurbau,1,2 Allison Komishon,1 Be Considered in Nourah Mazhar,1,2 John L. Sievenpiper,1,2,3,4,5 Control? Results From a Sonia Blanco Mejia,1,2,3 Hoang Vi Thanh Ho,1 Dandan Li,1,2 Systematic Review and Alexandra L. Jenkins,1 Lea Duvnjak,6 and Vladimir Vuksan1,2,4,5 Meta-analysis of Randomized Controlled Trials https://doi.org/10.2337/dc18-1126

OBJECTIVE META-ANALYSIS Evidence from randomized controlled trials (RCTs) suggests that viscous dietary fiber may offer beneficial effects on glycemic control and, thus, an improved risk profile. Our purpose was to conduct a systematic re- view and meta-analysis of RCTs to synthesize the therapeutic effect of viscous 1Clinical Nutrition and Risk Factor Modification ’ fiber supplementation on glycemic control in . Centre, St. Michael s Hospital, Toronto, Canada 2Department of Nutritional Sciences, Faculty of RESEARCH DESIGN AND METHODS Medicine, UniversityofToronto,Toronto, Canada 3Toronto 3D Knowledge Synthesis and Clini- MEDLINE,Embase,andCochraneCentralRegister ofControlled Trialsweresearched cal Trials Unit, St. Michael’s Hospital, Toronto, through 15 June 2018. We included RCTs ‡3 weeks in duration that assessed the Canada 4 effects of viscous fiber on markers of glycemic control in type 2 diabetes. Two Division of Endocrinology and Metabolism, St. Michael’s Hospital, Toronto, Canada independent reviewers extracted data. Data were pooled using the generic inverse 5Li Ka Shing Knowledge Institute, St. Michael’s variance method and expressed as mean differences (MD) with 95% CIs. Hetero- Hospital, Toronto, Canada geneity was assessed (Cochran Q statistic) and quantified (I2 statistic). The Grading 6Vuk Vrhovac University Clinic for Diabetes, of Recommendations Assessment, Development, and Evaluation (GRADE) ap- Endocrinology and Metabolic Diseases, Merkur University Hospital, University of Zagreb School proach was used to evaluate the overall certainty of the evidence. of Medicine, Zagreb, Croatia RESULTS Corresponding author: Vladimir Vuksan, v.vuksan@ utoronto.ca fi n fi We identi ed 28 eligible trial comparisons ( = 1,394). Viscous ber at a median dose Received 23 May 2018 and accepted 19 Septem- of ∼13.1 g/day significantly reduced HbA1c (MD 20.58% [95% CI 20.88, 20.28]; P = ber 2018 0.0002), fasting blood glucose (MD 20.82 mmol/L [95% CI 21.32, 20.31]; P = 0.001), This article contains Supplementary Data online and HOMA-insulin resistance (IR) (MD 21.89 [95% CI 23.45, 20.33]; P = 0.02) at http://care.diabetesjournals.org/lookup/suppl/ compared with control and in addition to standard of care. The certainty of doi:10.2337/dc18-1126/-/DC1. E.J. and R.K. are co-first authors. evidence was graded moderate for HbA1c, fasting glucose, fasting insulin, and HOMA-IR and low for fructosamine. © 2019 by the American Diabetes Association. Readers may use this article as long as the work CONCLUSIONS is properly cited, the use is educational and not fi fi for pro t, and the work is not altered. More infor- Viscous ber supplements improve conventional markers of glycemic control mation is available at http://www.diabetesjournals beyond usual care and should be considered in the management of type 2 diabetes. .org/content/license. Diabetes Care Publish Ahead of Print, published online January 7, 2019 2 Viscous Fiber and Diabetes Diabetes Care

Despite advancements in preventive randomized controlled trials (RCTs), the Data Extraction and Quality medicine and pharmacotherapy, diabe- effect of viscous dietary fiber supplemen- Assessment tes remains an overwhelming health tation on glycemic parameters in indi- Using a standardized proforma, two in- problem. Diet and lifestyle are among viduals with type 2 diabetes receiving dependent reviewers (R.K. and N.M.) the main pillars in the management of usual care. assessed articles and extracted relevant type 2 diabetes, with fiber consistently data from each report, including fiber considered a significant component of RESEARCH DESIGN AND METHODS type, study design (crossover or parallel), dietary interventions steering glycemic Protocol and Registration participant characteristics, comparator, control (1). However, a 2014 position This study followed the guideline of the dose, duration, background diet, compli- statement from the American Diabetes Cochrane Handbook for Systematic Re- ance measures, statistical analysis, coun- Association (ADA) deemphasized the im- views of Interventions (14) and results try of conducted research, and funding pact of fiber in diets, reporting its po- were reported in accordance with the sources. Disagreement between reviewers tential for glucoregulation as marginal, Preferred Reporting Items for Sys- was resolved by consensus or when nec- with modest improvements of 0.2–0.3% tematic Reviews and Meta-Analyses essary by a third reviewer. If b-glucan fi in HbA1c requiring “unrealistic” quanti- (PRISMA) (15). The protocol is available was not reported, viscous ber from oat ties of .50 g/day (2,3). at ClinicalTrials.gov (NCT02629263). b-glucan was conservatively estimated at In contrast, there is a plethora of 5% (19). The mean and SD were extracted clinical evidence on using soluble viscous Data Sources and Searches for HbA1c, fasting glucose, fasting insulin, dietary fiber supplements in the regula- MEDLINE, Embase, and the Cochrane fructosamine, and HOMA-IR at change tion of hyperglycemia and reduction of Central Register of Controlled Trials from baseline for both control and inter- conventional cardiovascular disease risk were searched using the strategy pre- vention groups. When SD values were not factors (4–7). This has been reflected in sented in Supplementary Table 1 to reported, they were calculated from avail- the 2018 ADA Standards of Medical Care identify RCTs that investigated the able data (95% CIs or SEM) using standard in Diabetes, which recommends an in- effect of viscous fiber on glycemic out- formulae (14). Authors were contacted crease in viscous fiber intake from sources comes in individuals with type 2 diabetes. for additional information where neces- such as oats, legumes, and citrus (8). The search was performed through 15 June sary (20,21). Nonetheless, the commonly shared view 2018. A manual search of the references The Cochrane risk-of-bias tool was used is that it is difficult to achieve a high of included trials supplemented the elec- to assess the study risk of bias (14). Do- dietary fiber intake within the context tronic search. No language restrictions mains of bias assessment include sequence of a conventional Western diet without were applied. generation, allocation concealment, blind- the use of fortified foods or addition of ing, incomplete outcome data, and selec- fiber isolates (9,10). In response, many Study Selection tive outcome reporting. The study was isolated fiber supplements have been Included RCTs were those conducted considered low risk of bias when proper developed and extensively studied over in individuals with type 2 diabetes methods were taken to reduce bias, high the past three decades with the intention with $3 weeks duration (16) that in- risk of bias when improper study methods of offering convenience of use and facil- vestigated the effect of viscous fiber likely affected the true outcome, and un- itating clinical study of the potency of a supplementation (b-glucan, guar gum, clear risk of bias when insufficient infor- concentrated source, with a favorable konjac, psyllium, pectin, xanthan gum, mation was provided to permit judgment record on glycemic benefits (11). locust bean gum, alginate, ) com- of bias level. Although the mechanisms of action paredwithanappropriatecontrol(i.e., have yet to be elucidated, it is hypoth- fiber-free supplement or one containing Data Synthesis and Analysis esized that fiber isolates, such as guar insoluble fiber, background diet, pla- Review Manager (RevMan) version 5.3 gum, b-glucan, or psyllium, have the cebo) on at least one of these glycemic (The Nordic Cochrane Centre, The ability to increase viscosity in the hu- measures: HbA1c, fasting glucose, fast- Cochrane Collaboration, Copenhagen, man gut and reduce the rate of nutri- ing insulin, HOMA-insulin resistance Denmark) was used for primary data ent absorption, and thus demonstrate (IR), and fructosamine. For multiarm analyses and Stata version 14 (StataCorp, greater potential to flatten the post- trials, we included the groups that al- College Station, TX) for subgroup, dose- prandial glycemic and insulinemic re- lowed us to isolate the effect of vis- response, and publication bias analyses. sponses compared with nonviscous fibers cous fiber supplements from control The difference between change-from- (4,12,13). It is less certain, however, treatments. Trials that precluded the baseline values for intervention and whether and to what extent the post- isolation of the effect of the viscous control arms was derived from each trial prandial effects are reliably reflected in fiber because it was incorporated into for the end points of HbA1c, fasting glu- long-term improvements, such as reduc- a fiber mixture or included as part of cose, fasting insulin, fructosamine, and tion in HbA1c. Therefore, the magnitude a complex dietary pattern, or because HOMA-IR. When change from baseline of benefit from viscous fiber intake in of lack of comparison with a calorie- was not reported, the mean and SD for diabetes management remains ambigu- matched control, were excluded. Glyce- baseline and end values were used to cal- ous and warrants comprehensive and mic outcomes criteria were determined culate change from baseline for both con- robust assessment. Hence, the objective in accordance with the ADA and Diabe- trol and intervention groups. When of this study was to evaluate, through a tes Canada Clinical Practice Guidelines HOMA-IR was not reported, it was calcu- systematic review and meta-analysis of (17,18). lated using the equation (HOMA-IR = fasting care.diabetesjournals.org Jovanovski and Associates 3

insulin (microU/L) 3 fasting glucose to assess the overall certainty of the RESULTS (mmol/L)/22.5) (22). A previously pub- evidence. Quality can be graded as Search Results lished formula was used to derive SD for very low, low, moderate, or high. Evi- Figure 1 shows the flow of the literature. calculated values of HOMA-IR (23). If dence obtained from RCTs receives Our initial search yielded 2,716 pub- change-from-baseline values were not an initial grade of high. Scores can be lications, of which 66 articles were available, end-of-treatment values were downgraded based on study limitations reviewed in full and 27 (28 trial com- used. For multiarm trials, a weighted av- (assessed by the Cochrane risk-of-bias parisons) were included in the final anal- erage was used to create a single pairwise tool), inconsistency (substantial unex- ysis (n = 1,394) (20,21,25–49). Twenty 2 . , comparison and to reduce the unit-of- plained heterogeneity, I 50%, P trial comparisons reported on HbA1c analysis error. A correlation coefficient of 0.10), indirectness (presence of factors (n = 1,148) (21,25,27,29–35,37–42,44, 0.50 was assumed for SD of crossover that limit the generalizability of the find- 47–49), 28 on fasting glucose (n = trials. Sensitivity analysis was conducted ings), imprecision (CI for the effect esti- 1,394) (20,21,25–48), 9 on fasting insulin with the use of different correlation co- mates that are wide or cross a minimally (n = 228) (21,25,26,29,35,38,41,46,48), efficient values (0.25 and 0.75) to test for important difference for benefitorharm), and 2 on fructosamine (n = 23) (42,46), the robustness of the effect size. Pooled and publication bias (significant evidence and 11 trial comparisons reported on analyses were conducted using the generic of small-study effects). HOMA-IR directly or reported enough inverse variance method with random- effects models. When data from ,5 trials were available, fixed-effects models were used. Data were expressed as mean All Reports Identified in Literature Search: 2,716 difference (MD) with 95% CI and signifi- cance was considered as P , 0.05. Inter- MEDLINE (1946 to 15 June 2018): 938 Embase (1947 to 15 June 2018): 733 study heterogeneity was assessed using The Cochrane Library (to 15 June 2018): 1,042 the Cochran Q statistic and quantified using Manual Search: 3 the I2 statistic, with I2 $50% indicating substantial heterogeneity and P , 0.10 significance (14). If .10 studies were in- Reports Excluded Based on Title and/or Abstract: 2,650 cluded for an outcome, sources of het- Duplicate Reports: 1,191 erogeneity were explored with a priori Abstracts: 4 subgroup analyses for baseline values of Acute Studies: 11 Animal or In Vitro Studies: 25 HbA1c, fasting glucose, and HOMA-IR, as well as for dose, design, duration, fiber Editorials/Letters: 16 Guidelines/Recommendations: 7 type, and food matrix (i.e., powder, capsule, Intervention: 866 food source) with P , 0.05 significance. A Invalid End Point: 152 post hoc analysis was also conducted for Not a Randomized Controlled Trial: 2 baseline BMI. To determine whether any Observational Studies: 81 single study exerted particular influence Population: 113 on the overall results, an additional sen- Reports, Reviews, & Meta-Analyses: 173 Summaries & Supplementary Materials: 9 sitivity analysis was performed by re- moving each study individually from the meta-analysis and recalculating the effect size of the remaining studies. Dose- Reports Reviewed in Full: 66 response analysis was performed using meta-regressions to generate linear and nonlinear dose estimates using the Reports Excluded: 39 MKSPLINE procedure, with P , 0.05 Acute Studies: 6 significance. Visual inspection of funnel Insufficient Data to Determine Viscous Fiber: 5 Intervention: 7 plots was used to assess publication bias Invalid End Point: 1 and formally tested using Egger and Not a Randomized Controlled Trial: 14 Begg tests, where P , 0.05 was consid- Population: 4 ered evidence for small-study effects. If Not Published: 1 Inadequate Control: 1 funnel plot asymmetry was suspected, the Duval and Tweedie “trim and fill” method was performed to impute missing Reports Included: 27 (28 comparisons) (n = 1,394) n study data and correct for asymmetry. HbA1c ( = 1,148) Fasting Glucose (n = 1,394) Fasting Insulin (n = 228) Grading of the Evidence Fructosamine (n = 23) The Grading of Recommendations As- HOMA-IR (n = 652) sessment, Development, and Evalua- tion (GRADE) approach (24) was used Figure 1—Flow of the literature 4 Viscous Fiber and Diabetes Diabetes Care

information for calculation (n = 652) Continuous meta-regression analyses did Effect on HOMA-IR (21,25,26,29,35,37–39,41,46,48). not reveal an effect of dose, duration, Figure 3C shows the effect of viscous baseline fasting glucose, or baseline fiber supplementation on HOMA-IR in Trial Characteristics BMI. Categorical meta-regression analy- individuals with diabetes. A median dose Table 1 shows the characteristics of in- ses, however, revealed a greater reduc- of 10.5 g/day for a median duration of fi cluded studies. The majority of trials tion in trials with higher baseline HbA1c 6 weeks signi cantly reduced HOMA-IR were conducted in an outpatient setting, values (between-subgroup difference (MD 21.89 [95% CI 23.45, 20.33]; with 15 (54%) in Europe, 5 (18%) in Asia, 20.70% [21.36, 20.03]; P = 0.04), P = 0.02) compared with control, with 2 (7%) in North America, 4 (14%) in with residual I2 = 84%. substantial evidence of interstudy het- the Middle East, and 2 (7%) in South erogeneity (I2 = 94%, P , 0.00001). America. Of the included trials, 15 (54%) Effect on Fasting Glucose Removal of Abutair et al. (25) during were crossover design and 13 (46%) Figure 3A shows the effect of viscous sensitivity analyses reduced overall het- 2 parallel design. The median age of par- fiber supplementation on fasting glucose erogeneity (I = 63%, P = 0.004) and ticipants was 60 years (range 48–67), in individuals with diabetes. Compared modified the effect on HOMA-IR (MD with a median BMI of 27 kg/m2 (range with control, a median dose of 13.1 g/day 21.07 [21.88, 20.26]; P = 0.01). The 26–32). The median dose of viscous fiber for a median duration of 8 weeks resulted use of different levels of correla- supplementation for all included trials in a significant reduction in fasting glu- tion coefficients (0.25 and 0.75) for was 13.1 g/day (range 2.55–21.0) and cose (MD 20.82 mmol/L [95% CI 21.32, the crossover studies did not influence median duration was 8 weeks (range 20.31]; P = 0.001) with evidence of the HOMA-IR effect or heterogeneity in 3–52). The Cochrane risk-of-bias tool substantial heterogeneity (I2 =92%,P , the overall pooled results. (Supplementary Fig. 1) showed that 25 0.00001). Systematic removal of in- Supplementary Table 2 and Sup- trials (90%) had unclear risk of bias and dividual studies did not alter results or plementary Fig. 4 show the findings three trials (10%) had low risk of bias for explain heterogeneity. The use of differ- of a priori and post hoc subgroup anal- sequence generation. All trials (100%) had ent levels of correlation coefficients yses for HOMA-IR. Continuous meta- unclear risk of bias for allocation conceal- (0.25 and 0.75) for crossover studies regression analysis revealed that the ment. Seventeen trials (61%) had low risk did not influence the fasting glucose effect of viscous fiber on HOMA-IR is of bias, 8 trials (29%) unclear risk, and effect or heterogeneity in the overall modified by baseline values (MD 3 trials (10%) high risk of bias for blinding. pooled results. 20.42 [20.67, 20.16]; P , 0.01), with 2 Sixteen trials (57%) had low risk of bias, Continuous and categorical a priori residual I = 72%. Categorical meta- 8 trials (29%) unclear risk, and 4 (14%) high and post hoc subgroup analyses did regression analysis was consistent with risk of bias for incomplete outcome data. not reveal any significant subgroup these findings, revealing a greater re- The majority of trials (96%) had low effects and failed to explain hetero- duction in trials with higher baseline risk but one (4%) had high risk of bias geneity (Supplementary Table 2 and HOMA-IR values (between-subgroup for selective outcome reporting. Fund- Supplementary Fig. 3). difference 23.49 [25.85, 21.14]; ing sources included agency for 6 trials P , 0.01), with a residual I2 = 67%. (21%), industry for 5 (18%), agency- Effect on Fasting Insulin Additional subgroup analyses were not industry for 5 (18%), and were not re- Figure 3B shows the effect of viscous fi- significant. ported for 12 (43%). ber supplementation on fasting insulin in individuals with diabetes. With a median Effect on Fructosamine

Effect on HbA1c dose of 15.0 g/day and median duration Figure 3D shows the effect of viscous Figure 2 shows the effect of viscous fiber of 8 weeks, no significant effect on fast- fiber supplementation on fructosamine supplementation on HbA1c in individ- ing insulin was observed (MD 217.56 in individuals with type 2 diabetes. Only uals with diabetes. A median dose of pmol/L [95% CI 237.54, 2.42]; P = 0.08) two trials reported on this outcome 10.9 g/day for a median duration of 8 compared with control, with substan- measure, with a median dose of 13.2 weeks resulted in a significant reduction tial interstudy heterogeneity (I2 = 90%, g/day and median duration of 7.5 weeks. in HbA1c (MD 20.58% [95% CI 20.88, P , 0.00001). Removal of Abutair et al. Compared with control, no significant 20.28]; P = 0.0002) compared with (25) during our sensitivity analyses re- effect was observed for fructosamine control, with evidence of substantial in- duced overall heterogeneity (I2 = 43%, (MD 20.12 mmol/L [95% CI 20.39, terstudy heterogeneity (I2 = 91%, P , P = 0.09) and modified the effect size 0.14]; P = 0.37) and no evidence of 0.00001). Systematic removal of individ- (MD 29.18 pmol/L [218.97, 0.60], interstudy heterogeneity was present ual studies did not alter the results or P = 0.07). The use of different levels (I2 = 0%, P = 0.60). Systematic removal explain heterogeneity. The use of differ- of correlation coefficients (0.25 and of individual studies and the use of ent levels of correlation coefficients 0.75) for crossover studies did not in- different levels of correlation coeffici- (0.25 and 0.75) for crossover studies did fluence the fasting insulin effect or het- ents (0.25 and 0.75) for crossover stud- not influence the HbA1c effect or hete- erogeneity in the overall pooled results. ies did not influence the fructosamine rogeneity in the overall pooled results. Continuous and categorical a priori effect or heterogeneity. Supplementary Table 2 and Sup- and post hoc subgroup analyses were Continuous and categorical a priori plementary Fig. 2 show the results of not performed for fasting insulin as ,10 and post hoc subgroup analyses were continuous and categorical a priori and trial comparisons were available for not performed as ,10 trial comparisons post hoc subgroup analyses for HbA1c. analyses. were available for analyses. care.diabetesjournals.org

Table 1—Characteristics of included trial comparisons Baseline BMI,* HbA1c,* % Baseline Design, Duration, Viscous Dose,*† Food Trial Participants Age,* years kg/m2 (mmol/mol) FG,* mmol/L blinding weeks fiber Control g/day matrix Diet Funding Country Abutair et al., 36 47.5‡ 31.7 C: 8.5 (69) C:8.7 P, n/r 8 Psyllium No psyllium 10.5 Powder Usual n/r Palestine 2016 (18M:18F) T: 8.5 (69) T: 9.0 Abutair et al., 36 47.5 n/r n/r 8.9 P, n/r 8 Psyllium No psyllium 10.5 Powder Usual n/r Palestine 2018 (18M:18F) Aro et al., 9 53 n/r n/r C:11.5 Cr, DB 12 Guar gum Wheat flour 21 Granules Usual A-I Finland 1981 (5M:4F) T:10.6 Baker, 1988 30 C: 64.3 n/r C: 11.9 (107) C:11.7 P, SB 9 Guar gum n/r 15 Tablets Usual n/r U.K. T: 59.1 T: 12.0 (108) T:12.1 Chen et al., 22 64 25.5 n/r 9.1 Cr, DB 4 Konjac Cornstarch 3.6 Capsules NCEP A-I Taiwan 2003 (10M:12F) Chuang et al., 13 54.9 25.8 7.9 (63) 10.6 Cr, DB 8 Guar gum n/r 15 Powder Liquid A Taiwan 1992 meals Cugnet-Anceau 53 C: 61.8 C: 29.0 C: 7.5 (58) C: 8.4 P, DB 8 b-Glucan Soup 3.5 Soup Usual A Sweden et al., 2010 T: 61.9 T: 30.5 T: 7.3 (56) T: 8.8 Dall’Alba et al., 44 C: 63.6 C: 29.3 C: 7.0 (53) C: 7.9 P, OL 6 Guar gum n/r 10 Powder Usual A-I Brazil 2013 (27M:17F) T: 60.5 T: 30.2 T: 6.9 (52) T: 7.4 Feinglos et al., 33 C: 56.5 n/r C: 7.6 (60) C: 11.8 P, DB 12 Psyllium Fiber-free 10.2 n/r Restricted n/r U.S. 2013 TA: 61.8 TA: 7.4 (57) TA: 11.2 placebo TB: 64.8 TB: 7.9 (63) TB: 10.3 Fuessl et al., 18 61.3 30.1 C: 9.3 (78) 9.1 Cr, DB 4 Guar gum Wheat 15 Granules Usual A U.K. 1987 (12M:6F) T: 9.7 (83) Holman et al., 29 54.2 26.5 n/r n/r Cr, n/r 8 Guar gum n/r 15 Tablets n/r I U.K. 1987 (24M:5F) Laajam et al., 39 51.5 31.2 11.4 (101) 12.5 Cr, DB 4 Guar gum Beef gelatin 15 Granules Usual n/r Saudi Arabia 1990 (11M:28F) Lalor et al., 19 56.5 31.5 n/r 10.9 Cr, DB 12 Guar gum n/r 15 Granules Usual A-I U.K. 1990 (8M:11F) Li et al. (A), 238 59 27 C: 8.1 (65) C: 9.5 P, n/r 4 b-Glucan No b-glucan 2.65 Oat Low fat, I China 2016 (116M:85F) TA: 8.4 (68) TA: 9.9 high fiber TB: 8.3 (67) TB: 9.7 Li et al. (B), 228 59 27 C: 8.1 (65) C: 9.5 P, n/r 52 b-Glucan No b-glucan 5.3 Oat Low fat, I China 2016 TA: 8.4 (68) TA: 9.9 high fiber TB: 8.3 (67) TB: 9.7 oaosiadAscae 5 Associates and Jovanovski Liatis et al., 41 C: 66.5 C: 27.0 C: 6.9 (52) C: 7.7 P, DB 3 b-Glucan No b-glucan 3 Bread Usual n/r Greece 2009 T: 60.2 T: 29.6 T: 7.3 (56) T: 8.8 Ma et al., 186 57.5‡ C: 26.6 C: 9.7 (83) C: 9.5 P, SB 4 b-Glucan No b-glucan 2.55 Oat SDI I China 2013 TA: 26.6 TA: 9.9 (85) TA: 10.1 TB: 26.7 TB: 9.8 (84) TB: 10.1 Continued on p. 6 icu ie n Diabetes and Fiber Viscous 6

Table 1—Continued Baseline BMI,* HbA1c,* % Baseline Design, Duration, Viscous Dose,*† Food Trial Participants Age,* years kg/m2 (mmol/mol) FG,* mmol/L blinding weeks fiber Control g/day matrix Diet Funding Country McGeoch et al., 27 60.9 31.5 6.8 (51) n/r Cr, NB 8 b-Glucan No b-glucan 4 Oat-based Usual A U.K. 2013 (18M:9F) products Niemi et al., 18 63 27 C: 11.4 (101) C: 12.5 Cr, DB 12 Guar gum Micro 15 n/r Usual A-I Finland 1988 T: 12.1 (109) T: 11.7 crystalline cellulose Peterson et al., 16 60 27.3 C: 11.3 (100) C: 9.7 Cr, n/r 6 Guar gum No guar gum 16.6 Granules Usual A U.K. 1987 (10M:6F) T: 11.2 (99) T: 9.5 Rodr´ıguez- 123 C: 56.5 C: 28.6 n/r C: 7.8 P, DB 6 Psyllium Micro 15 Powder Low fat n/r Mexico Moran´ et al., (55M:68F) T: 57.0 T: 29.1 T: 10.7 crystalline 1998 cellulose Sels et al., 12 62 25.8 11.3 (100) 9.8 Cr, n/r 12 Guar gum High-fiber 11.2 Bread Usual n/r The 1993 (6M:6F) bread Netherlands Uusitupa et al., 9 47.8 n/r n/r 10.6 Cr, DB 4 Guar gum Wheat flour 15 Granules Usual n/r Finland 1990 (4M:5F) Uusitupa et al., 39 C: 60.9 n/r C: 9.4 (79) C: 12.8 P, DB 12 Guar gum Wheat flour 15 Granules Usual A Finland 1984 (13M:26F) T: 60.1 T: 8.9 (74) T: 12.2 Uusitupa et al., 1984 17 62 n/r n/r 9.7 Cr, DB 18 Guar gum Wheat flour 21 Granules Usual n/r Finland Vuksan et al., 11 M: 62 n/r M: 7.4 (57) C: 9.3 Cr, DB 3 Konjac Wheat bran 15.1 Biscuit NCEP I Canada 1999 (5M:6F) F: 59 F: 8.3 (67) T: 9.6 Wolffenbutt 12 62 25.8 11.3 (100) 9.8 Cr, n/r 12 Guar gum Bread 11.2 Powder Usual n/r The et al., 1992 (6M:6F) Netherlands Ziai et al., 36 C: 53.6 C: 27.5 C: 9.1 (76) C: 9.9 P, DB 8 Psyllium Micro 10.2 Powder Usual n/r Iran 2005 T: 51.9 T: 26.6 T: 10.5 (91) T: 11.6 crystalline cellulose A, agency; A-I, agency-industry; C, control; Cr, crossover; DB, double blind, F, female; FG, fasting glucose; I, industry; M, male; NB, not blinded; NCEP, National Education Program; n/r, not reported; OL, open label; P, parallel; SB, single blind; SDI, Structured Dietary Intervention; T, treatment; TA, treatment A; TB, treatment B. *Mean values presented. †Dose of viscous fiber supplement. ‡ Range reported, mean of upper and lower ranges displayed. ibtsCare Diabetes care.diabetesjournals.org Jovanovski and Associates 7

Trials, year Viscous Fiber (n)Control (n)Weight MD [95% CI] in HbA1c (%) Abutair et al, 2016 18 18 6.20% -1.00 [-1.28, -0.72] Baker, 1988 15 15 4.80% 0.00 [-0.71, 0.71] Chuang et al, 1992 29 24 5.50% -0.30 [-0.82, 0.22] Cugnet-Anceau et al, 2010 29 24 6.20% -0.17 [-0.44, 0.10] Dall'Alba et al, 2013 23 21 6.00% -0.20 [-0.54, 0.14] Feinglos et al, 2013 25 8 6.40% -0.59 [-0.80, -0.37] Fuessl et al, 1987 18 18 4.50% -0.79 [-1.57, -0.01] Holman et al, 1987 29 29 5.00% 0.20 [-0.44, 0.84] Laajam et al, 1990 39 39 4.80% -1.60 [-2.31, -0.89] Li et al (A), 2016 159 79 6.20% -0.13 [-0.42, 0.17] Li et al (B), 2016 152 76 6.20% -0.56 [-0.85, -0.26] Lias et al, 2009 23 18 6.40% -0.15 [-0.32, 0.02] Ma et al, 2013 127 59 6.20% -2.00 [-2.28, -1.71] McGeoch et al, 2013 27 27 6.30% 0.10 [-0.14, 0.34] Niemi et al, 1988 18 18 3.90% -1.10 [-2.08, -0.12] Peterson et al, 1987 16 16 2.40% -0.50 [-2.03, 1.03] Sels et al, 1993 9 9 2.20% 0.20 [-1.47, 1.87] Uusitupa et al, 1989 20 19 4.90% -0.44 [-1.11, 0.23] Wolffenbuel et al, 1992 12 12 2.60% 0.20 [-1.25, 1.65] Ziai et al, 2005 21 15 3.30% -3.00 [-4.17, -1.83]

Total 809 544 100.00% -0.58 [-0.88, -0.28] Heterogeneity: τ² = 0.36; χ² = 212.00, df = 19 (P < 0.00001); I² = 91% Test for overall effect: Z = 3.77 (P = 0.0002) Favors Viscous Fiber Favors Control

Figure 2—The effect of viscous fiber supplementation in individuals with type 2 diabetes on primary outcome HbA1c. Diamond represents the pooled effect estimate for overall analysis. Data are represented as MD with 95% CI, using the generic inverse variance random-effects model. Interstudy heterogeneity quantified by I2 with significance P , 0.10. (A high-quality color representation of this figure is available in the online issue.)

Dose-Response Analyses in an adjusted MD of 22.67 (95% absolute reduction of 0.58% in HbA1c, There was no significant evidence of a CI 24.17, 21.18), P , 0.01, suggesting 0.82 mmol/L in fasting blood glucose, dose-response effect (Supplementary evidence of small-study effects. Publi- and1.89inHOMA-IRfollowingamedian Figs. 5 and 6). Visual inspection of cation bias was not assessed for fast- dose of ;13.1 g/day for a median du- data suggests doses .10 g/day may ing insulin and fructosamine as there ration of ;8 weeks. No significant ef- be more effective in HOMA-IR improve- were ,10 trial comparisons available. fects were revealed for fasting insulin ment, but the difference in slopes for and fructosamine. Subgroup analyses ,10 vs. .10 g/day was not significant Grading of the Evidence revealed those with higher baseline fi (P = 0.06). Because of insuf cient data, Supplementary Table 3 shows the sum- HbA1c and HOMA-IR values appear to dose-response analyses could not be mary of the GRADE assessment for each show greater reductions. There did not conducted for fructosamine and only outcome. The effect estimates for HbA1c, appear to be any subgroup effects of linear analysis was performed for fast- fasting glucose, fasting insulin, and HOMA- dose, design, duration, baseline BMI, ing insulin. IR were graded as moderate quality fiber type, or the form of intervention. based on downgrades of serious incon- Results from our analyses suggest that Publication Bias sistency for HbA1c, fasting glucose, and viscous fiber may be clinically meaningful Supplementary Fig. 7 shows the funnel HOMA-IR and serious imprecision for in the management of type 2 diabetes, plots for HbA1c, fasting glucose, and fasting insulin. Evidence for fructosamine with reductions in HbA1c exceeding the HOMA-IR. Visual inspection of funnel was graded low quality owing to down- U.S. Food and Drug Administration plots suggests no asymmetry in HbA1c grades for very serious imprecision. threshold of $0.3% established for and fasting glucose and mild asymmetry new antihyperglycemic drug develop- in HOMA-IR. Formal testing with the ment (50). Our findings build on those Egger and Begg tests was not significant CONCLUSIONS of an earlier systematic review and for evidence of small-study effects. Trim This systematic review and meta-analysis meta-analysis by Silva et al. (51), who and fill analyses was conducted for quantified the effect of viscous fiber reported a decrease of 0.52% in HbA1c HOMA-IR, identifying four additional supplementation on indices of glycemic and 0.55 mmol/L in fasting blood glucose studies imputed to adjust for funnel control in 28 RCT comparisons involv- in type 2 diabetes patients following a plot asymmetry (Supplementary Fig. ing individuals with type 2 diabe- high intake of various types of dietary 8). Inclusion of imputed studies resulted tes. Pooled analyses demonstrate an fiber, including soluble and insoluble 8 Viscous Fiber and Diabetes Diabetes Care

A Trials, year Viscous Fiber (n) Control (n) Weight MD [95% CI] in Fasng Glucose (mmol/L) Abutair et al, 2016 18 18 4.30% -2.08 [-2.50, -1.66] Abutair et al, 2018 18 18 4.20% 2.26 [1.72, 2.80] Aro et al, 1981 9 9 2.50% -1.20 [-3.35, 0.95] Baker, 1988 15 15 2.40% 0.90 [-1.26, 3.06] Chen et al, 2003 22 22 3.90% -2.30 [-3.19, -1.41] Chuang et al, 1992 13 13 3.70% -0.94 [-2.02, 0.14] Cugnet-Anceau et al, 2010 29 24 4.10% -0.69 [-1.41, 0.03] Dall'Alba et al, 2013 23 21 3.80% 0.50 [-0.51, 1.51] Feinglos et al, 2013 25 8 4.00% -1.97 [-2.77, -1.17] Fuessl et al, 1987 18 18 3.80% -1.06 [-2.05, -0.07] Holman et al, 1987 29 29 4.20% 0.00 [-0.53, 0.53] Laajam et al, 1990 39 39 3.80% -1.80 [-2.74, -0.86] Lalor et al, 1990 19 19 3.00% -1.90 [-3.55, -0.25] Li et al (A), 2016 159 79 4.30% -0.19 [-0.68, 0.29] Li et al (B), 2016 152 76 4.30% -0.09 [-0.54, 0.35] Lias et al, 2009 23 18 4.20% -0.65 [-1.17, -0.13] Ma et al, 2013 127 59 4.30% -1.56 [-1.91, -1.21] McGeoch et al, 2013 27 27 4.20% 0.30 [-0.20, 0.80] Niemi et al, 1988 18 18 3.50% 0.50 [-0.72, 1.72] Peterson et al, 1987 16 16 3.10% -0.20 [-1.79, 1.39] Rodríguez-Morán et al, 1998 60 63 4.30% -2.17 [-2.54, -1.80] Sels et al, 1993 12 12 2.60% 0.20 [-1.81, 2.21] Uusitupa et al, 1984 17 17 3.00% -0.80 [-2.42, 0.82] Uusitupa et al, 1989 20 19 3.50% -1.30 [-2.54, -0.06] Uusitupa et al, 1990 9 9 2.50% -1.60 [-3.71, 0.51] Vuksan et al, 1999 11 11 3.00% -0.71 [-2.38, 0.96] Wolffenbuel et al, 1992 12 12 2.60% 0.20 [-1.81, 2.21] Ziai et al, 2005 21 15 2.80% -4.98 [-6.79, -3.17]

Total 961 704 100.00% -0.82 [-1.32, -0.31] Heterogeneity: τ² = 1.49; χ² = 332.99, df = 27 (P < 0.00001); I² = 92% Test for overall effect: Z = 3.18 (P = 0.001) Favors Viscous Fiber Favors Control B

Trials, year Viscous Fiber (n) Control (n)Weight MD [95% CI] in Fasng Insulin (pmol/L) Abutair et al, 2016 18 18 13.20% -67.80 [-80.99, -54.61] Aro et al, 1981 9 9 10.10% -0.60 [-33.78, 32.58] Chuang et al, 1992 13 13 10.50% -28.80 [-59.33, 1.73] Laajam et al, 1990 39 39 9.60% 10.92 [-25.33, 47.17] Lias et al, 2009 23 18 11.60% -42.00 [-66.07, -17.93] McGeoch et al, 2013 27 27 13.60% -2.40 [-11.46, 6.66] Peterson et al, 1987 16 16 12.60% -6.00 [-23.71, 11.71] Vuksan et al, 1999 11 11 5.50% 2.00 [-64.66, 68.66] Ziai et al, 2005 21 15 13.30% -5.40 [-17.35, 6.55]

Total 177 166 100.00% -17.56 [-37.54, 2.42] Heterogeneity: τ² = 743.10; χ² = 80.27, df = 8 (P < 0.00001); I² = 90% Test for overall effect: Z = 1.72 (P = 0.08) Favors Viscous Fiber Favors Control

Figure 3—The effect of viscous fiber supplementation in individuals with type 2 diabetes on secondary outcomes: fasting glucose (A), fasting insulin (B), HOMA-IR (C), and fructosamine (D). Diamond represents the pooled effect estimate for overall analysis. Data are represented as MD with 95% CI, using the generic inverse variance random-effects and fixed-effects models. Interstudy heterogeneity is quantified by I2 with significance P , 0.10. (A high-quality color representation of this figure is available in the online issue.) care.diabetesjournals.org Jovanovski and Associates 9

C Trials, year Viscous Fiber (n) Control (n) Weight MD [95% CI] in HOMA-IR Abutair et al, 2016 18 18 11.40% -6.30 [-7.33, -5.27] Aro et al, 1981 9 9 6.60% -1.07 [-5.13, 2.99] Chuang et al, 1992 13 13 7.50% -2.77 [-6.24, 0.70] Laajam et al, 1990 39 39 5.40% -1.57 [-6.54, 3.40] Li et al (A), 2016 159 79 10.30% -1.33 [-3.16, 0.49] Lias et al, 2009 23 18 10.00% -3.41 [-5.42, -1.40] Ma et al, 2013 127 59 12.00% -0.11 [-0.23, 0.01] McGeoch et al, 2013 27 27 11.60% 0.10 [-0.78, 0.98] Peterson et al, 1987 16 16 10.20% -0.53 [-2.41, 1.35] Vuksan et al, 1999 11 11 3.60% -0.59 [-7.45, 6.27] Ziai et al, 2005 21 15 11.10% -2.20 [-3.53, -0.87]

Total 463 304 100.00% -1.89 [-3.45, -0.33] Heterogeneity: τ² = 5.25; χ² = 159.33, df = 10 (P < 0.00001); I² = 94% Test for overall effect: Z = 2.37 (P = 0.02) Favors Viscous Fiber Favors Control D Trials, year Viscous Fiber (n) Control (n) Weight MD [95% CI] in Fructosamine (mmol/L) Sels et al, 1993 12 12 48.70% -0.05 [-0.43, 0.33] Vuksan et al, 1999 11 11 51.30% -0.19 [-0.56, 0.18]

Total 23 23 100.00% -0.12 [-0.39, 0.14] Heterogeneity: χ² = 0.27, df = 1 (P = 0.60); I² = 0% Test for overall effect: Z = 0.90 (P = 0.37) Favors Viscous Fiber Favors Control

Figure 3dContinued. sources, from 12 RCTs. Although the two remains of important clinical interest resulted in flattening of the glycemic reviews cannot be directly compared (53). Although residual confounding might response and reduced insulin require- given the variation in study inclusion preclude a potential case for causality, ments, ultimately leading to a reduction criteria, it is interesting that there was the recently hypothesized link of modu- in HbA1c, as observed in the present a similar reduction in HbA1c, suggest- lating gut microbiota by extended cereal analysis. More recently, Chandalia ing that the benefits observed by Silva fiberintakedeservesfurtherinvestigation et al. (56) demonstrated that an effective et al. (51) may be mostly attributed to (54). The protective role of cereal fiber level of intake can be achieved by con- viscous fiber. Despite this similarity, against type 2 diabetes should therefore suming fiber-rich foods. In that study, grouping viscous soluble and insoluble not be overlooked and fits within a they found that doubling fiber intake fiber does not provide a reliable estimate broader public health case for increasing above the level recommended by of metabolic benefit as it is analogous overall dietary fiber intake. ADA at the time achieved a powerful to grouping therapeutic entities with dif- The variable efficacy of fiber on gly- reductionin24-hbloodglucoseand ferent physiochemical and, hence, phys- cemic control was first highlighted in a hyperinsulinemia (56). The authors con- iological characteristics. While insoluble seminal study in which Jenkins et al. (55) cluded that these effects seem to be fiber seems to have specific application, compared the response to an oral glu- predominantly a result of increasing likely in the area of colonic health and cose tolerance test supplemented with the viscous fiber foods from the typically stool bulking, viscous fibers appear to five fiber types varying in level of viscos- recommended 8 g/day in the control arm have pleiotropic effects metabolically, as ity. A positive correlation was found to 25 g/day in the test arm, a difference they improve glycemic control, lipid lev- between the viscosity of the fiber type that approximates the median effective els, , and possibly weight, and reduction in peak postprandial blood dose of 13.1 g/day seen in the current in addition to potential prebiotic activity glucose and insulin concentrations (55). study. Further supporting the case for (52). Conversely, the strong and con- This benefit was abolished when fiber viscosity, a study comparing several sistent paradoxical association of non- was hydrolyzed to its nonviscous form types of viscous fiber supplements re- viscous cereal fiber and whole-grain (55). Thus, through its effect in chang- vealed that the most viscous fiber, konjac consumption to reduced type 2 diabetes ing the rate of nutrient delivery and glucomannan, resulted in the greatest incidence from prospective cohort data endocrine response, viscous fiber intake reduction of postprandial blood glucose 10 Viscous Fiber and Diabetes Diabetes Care

(57). These acute benefits were trans- outcome and the study included trials Balancing the strengths and limita- lated to long-term metabolic improve- spanning multiple countries, thereby al- tions, we graded the overall certainty ments in individuals with type 2 diabetes lowing for generalizability of findings of available evidence as moderate for and metabolic syndrome (46,58), as well and reducing potential confounders as- HbA1c, fasting glucose, fasting insulin, as lipid lowering in healthy individuals, sociated with a single geographic loca- and HOMA-IR and low for fructosamine. compared with lower viscosity fibers in- tion. Finally, the overall quality and cluding insoluble fiber or wheat bran strength of evidence was assessed using Conclusion control (52). Similarly, a 6-month inter- the GRADE approach. This study illustrated that viscous fiber vention study in participants with met- Limitations of this analysis should also supplementation improved conventional abolic syndrome showed no glycemic be recognized when interpreting the markers of glycemic control beyond benefits on the American Heart Associ- findings. First, we downgraded the cer- usual care in individuals with type 2 di- ation Step II diet, but addition of 7 g/day tainty of the evidence for serious incon- abetes. Future dietary guidelines should of psyllium fiber lead to a sustained sistency in the estimates across trials for be revisited in light of these findings, improvement in glycemic and insuline- some of the assessed outcomes. This was although taking into consideration the mic response, including a significant due to evidence of heterogeneity that limitations raised by GRADE. Additional reduction of 0.6% in HbA1c (59). could not be explained by sensitivity and high-quality RCTs are required to further Despite a relatively small quantity of subgroup analyses. Further, the certainty explore the effect by fiber type and to viscous fiber being required to obtain of the evidence was downgraded for optimize the incorporation of highly vis- clinically meaningful benefits in diabetes, serious and very serious imprecision. cous supplements into the daily diet. the main challenge remains how to in- Although the 95% CI of the pooled effect corporate it into foods while preserving estimate for some of our outcomes did sensory characteristics. From a palat- not overlap with our minimally important Acknowledgments. The authors thank Teruko ability standpoint, semimoist foods in- difference for harm (i.e., did not contain Kishibe, Information Specialist, Scotiabank Health ’ cluding crisp breads, crackers, muffins, evidence for harm), the upper bound of Science Library at St. Michael s Hospital, for herhelpinthedevelopmentofsearchterms and biscuits are suggested to be the the 95% CI included 0. The number of used. most suitable vehicles for optimal fiber participants for the fructosamine out- Funding and Duality of Interest. R.K. and N.M. delivery (60). come was also less than the optimal have received funding from the King Abdullah Even though the prevalent notion is to information size criterion, which resulted Scholarship Program, Saudi Arabia. J.L.S. was favor food rather than supplement use in an additional downgrade for impreci- funded by a PSI Graham Farquharson Knowledge fi fi Translation Fellowship; Canadian Diabetes As- as a primary source of ber, the use of the sion owing to insuf cient power. Due to sociation Clinician Scientist Award; Canadian latter would allow one to achieve the goal the small number of observations for Institutes of Health Research Institute of Nutri- of an individualized eating plan even in some outcomes, meta-regression analy- tion, Metabolism and Diabetes/Canadian Nutri- the absence of major dietary restrictions, ses could not be conducted, limiting our tion Society New Investigator Partnership Prize; such as in the context of lack of willing- exploration of these outcomes. Addition- and Banting & Best Diabetes Centre Sun Life Financial New Investigator Award. J.L.S. has also ness or ability of an individual to change ally, of the included trials, only 11 had a received research support from Diabetes Can- (8). Notably, with the exception of one duration of 12 weeks or longer and 13 ada, the American Society for Nutrition, Calorie study where treatment was through diet trials were less than 8 weeks in length. Control Council, INC International Nut and Dried only, studies included in our analysis Given the conventional estimate that Fruit Council Foundation, National Dried Fruit fi fl Trade Association, the Tate and Lyle Nutritional utilized ber supplements as the in- HbA1c re ects blood glucose levels in Research Fund at the University of Toronto, and tervention, with antihyperglycemic oral the preceding three months, inclusion the Glycemic Control and Cardiovascular Disease medication and insulin having not of shorter trials potentially underesti- in Type 2 Diabetes Fund at the University of been altered throughout the study pe- mated the effect size, as they may not Toronto (a fund established by the Alberta Pulse riod. This suggests that the effect of have been of sufficient duration. The Growers). He has received speaker fees and/or fi honoraria from Diabetes Canada, the Canadian viscous ber, primarily through supple- certainty of evidence was not down- Nutrition Society, Dr. Pepper Snapple Group, ments, seen within the pooled analysis graded for indirectness as subgroup Dairy Farmers of Canada, Sprim Brasil, White- is beyond that of standard pharmaceu- analyses revealed no effect modulation Wave Foods, Rippe Lifestyle, mdBriefcase, Al- tical therapy. by duration. It was also not downgraded berta Milk, FoodMinds LLC, Memac Ogilvy & The current study has several for risk of bias as plausible selection Mather LLC, PepsiCo, The Ginger Network LLC, International Sweeteners Association, Nestle´ strengths. Importantly, to our knowledge bias was unlikely to seriously alter the Nutrition Institute, Pulse Canada, Canadian So- this is one of the largest and most results. Finally, there was evidence of ciety for Endocrinology and Metabolism (CSEM), comprehensive systematic reviews and publication bias. Although visual inspec- and GI Foundation. He has ad hoc consulting meta-analyses of RCTs on dietary fiber, in tion of the funnel plots suggested arrangements with Winston & Strawn LLP, Per- particular investigating the isolated ef- asymmetry for HOMA-IR and trim and kins Coie LLP, and Tate & Lyle. He is a member of the European Fruit Juice Association Scientific fect of viscous fiber sources in diabetes. fill analyses suggested small-study ef- Expert Panel. He is on the clinical practice guide- No prior study had differentiated dietary fects, we elected not to downgrade lines expert committees of Diabetes Canada, the fiber on the basis of viscosity but rather for publication bias, as Egger and Begg European Association for the Study of Diabetes had done so on the basis of the gravi- tests were not significant and the ad- (EASD), and the Canadian Cardiovascular Society, as well as an expert writing panel of the American metrics. Furthermore, the majority of justed pooled effect estimate after trim Society for Nutrition. He serves as an unpaid trials in this study reported glycemic and fill analyses did not change direc- scientific advisor for the Food, Nutrition, and control end points as the primary tion or significance. Safety Program and the Technical Committee on care.diabetesjournals.org Jovanovski and Associates 11

Carbohydrates of the International Life Science individuals with diabetes: a quantitative assess- 22. Muniyappa R, Lee S, Chen H, Quon MJ. Institute North America. He is a member of the ment and meta-analysis of the evidence. J Am Current approaches for assessing insulin sensi- International Carbohydrate Quality Consortium, Coll Nutr 2004;23:5–17 tivity and resistance in vivo: advantages, limi- Executive Board Member of the Diabetes and 8. American Diabetes Association. 9. Cardiovas- tations, and appropriate usage. Am J Physiol Nutrition Study Group (DNSG) of the EASD, and cular disease and risk management: Standards Endocrinol Metab 2008;294:E15–E26 director of the Toronto 3D Knowledge Synthesis of Medical Care in Diabetesd2018. Diabetes 23. Pentz M, Shott M, Aprahamian F. Handling and Clinical Trials Foundation. His wife is an Care 2018;41(Suppl. 1):S86–S104 Experimental Data. Milton Keynes, U.K., Open employee of Unilever Canada. A.L.J. is part owner 9. Agricultural Research Service. Dietary fiber (g): University Press, 1988 and Vice President at Glycemic Index Laborato- usual intakes from food and water, 2003–2006, 24. Guyatt GH, Oxman AD, Vist GE, et al.; GRADE ries, Inc. (GI Labs), a contract research organi- compared to adequate intakes. In What We Eat Working Group. GRADE: an emerging consensus zation. L.D. has previously received honoraria for in America, NHANES 2003–2006. Beltsville, MD, on rating quality of evidence and strength of speaking and consultancy and grants for attend- U.S. Department of Agriculture, 2010 recommendations. BMJ 2008;336:924–926 ing scientific conferences from Abbott, Amgen, 10. Mobley AR, Jones JM, Rodriguez J, Slavin J, 25. Abutair AS, Naser IA, Hamed AT. Soluble AstraZeneca, Boehringer Ingelheim, Eli Lilly, Zelman KM. Identifying practical solutions to fibers from psyllium improve glycemic re- Merck, Merck Sharp & Dohme, Novo Nordisk, meet America’s fiber needs: proceedings from sponse and body weight among diabetes type Novartis, Sanofi, Servier, and Takeda. V.V. holds a the Food & Fiber Summit. Nutrients 2014;6: 2 patients (randomized control trial). Nutr J 2016; research grant from the Canadian Diabetes As- 2540–2551 15:86 sociation for study of dietary intervention in- 11. Wheeler ML, Dunbar SA, Jaacks LM, et al. 26. Aro A, Uusitupa M, Voutilainen E, Hersio K, cluding viscous soluble fiber and holds the Macronutrients, food groups, and eating pat- Korhonen T, Siitonen O. Improved diabetic con- Canadian (2,410,556) and American (7,326,404) terns in the management of diabetes: a system- trol and hypocholesterolaemic effect induced patents on the medical use of viscous fiber blend atic review of the literature, 2010. Diabetes Care by long-term dietary supplementation with guar for reducing blood glucose for treatment of 2012;35:434–445 gum in type 2 (insulin-independent) diabetes. diabetes, increasing insulin sensitivity, and re- 12. McRorie JW Jr. Evidence-based approach Diabetologia 1981;21:29–33 ducing systolic blood pressure and . to fiber supplements and clinically meaningful 27. Baker P. Placebo-controlled trial of guar in None of the sponsors had a role in any as- health benefits, part 1: what to look for and poorly controlled type II diabetes. Pract Diabetes pect of the current study, including design and how to recommend an effective fiber therapy. 1988;5:36–38 conduct of the study; collection, management, Nutr Today 2015;50:82–89 28. Chen HL, Sheu WH, Tai TS, Liaw YP, Chen YC. analysis, and interpretation of the data; and 13. McRorie JW Jr. Evidence-based approach to Konjac supplement alleviated hypercholesterol- preparation, review, and approval of the man- fiber supplements and clinically meaningful emia and hyperglycemia in type 2 diabetic uscript or decision to publish. health benefits, part 2: what to look for and subjectsda randomized double-blind trial. J Author Contributions. E.J., A.Z., A.K., H.V.T.H., how to recommend an effective fiber therapy. Am Coll Nutr 2003;22:36–42 and V.V. designed the research. R.K. and N.M. Nutr Today 2015;50:90–97 29. Chuang LM, Jou TS, Yang WS, et al. Thera- conducted the research. R.K., A.K., J.L.S., S.B.M., 14. Higgins JPT, Green S (Eds.). Cochrane Hand- peutic effect of guar gum in patients with non- and D.L. performed or assisted in performing the book for Systematic Reviews of Interventions, insulin-dependent diabetes mellitus. J Formos statistical analysis of the data, E.J., R.K., A.Z., A.K., Version 5.1.0 [Internet]. The Cochrane Collabo- Med Assoc 1992;91:15–19 A.L.J., and L.D. wrote the manuscript draft. E.J., ration, 2011. Available from https://handbook- 30. Cugnet-Anceau C, Nazare JA, Biorklund M, J.L.S., and V.V. had primary responsibility for the 5-1.cochrane.org/. Accessed 15 June 2018 et al. A controlled study of consumption of final content. All authors contributed to the 15. Moher D, Liberati A, Tetzlaff J, Altman DG; b-glucan-enriched soups for 2 months by critical revision of the manuscript for important PRISMA Group. Preferred reporting items for type 2 diabetic free-living subjects. Br J Nutr intellectual content and approved the final man- systematic reviews and meta-analyses: the 2010;103:422–428 uscript. PRISMA statement. PLoS Med 2009;6:e1000097 31. Dall’Alba V, Silva FM, Antonio JP, et al. Prior Presentation. Parts of this work were 16. U.S. Food and Drug Administration. Guid- Improvement of the metabolic syndrome profile presented in abstract and poster form at the ance for Industry: Evidence-Based Review Sys- by soluble fibre – guar gum – in patients with 2017 CSEM/Diabetes Canada Professional Con- tem for the Scientific Evaluation of Health type 2 diabetes: a randomised clinical trial. Br J ference and Annual Meetings, Edmonton, Claims. Silver Springs, MD, U.S. Department of Nutr 2013;110:1601–1610 Canada, 1–4 November 2017, and in abstract Health and Human Services, 2009. Available from 32. Feinglos MN, Gibb RD, Ramsey DL, Surwit RS, form and short oral presentation at the DNSG- https://www.fda.gov/Food/GuidanceRegulation/ McRorie JW. Psyllium improves glycemic control EASD 36th Symposium on Diabetes and Nutri- GuidanceDocumentsRegulatoryInformation/ in patients with type-2 diabetes mellitus. Bio- tion, Opatija, Croatia, 27–30 June 2018. LabelingNutrition/ucm073332.htm. Accessed active Carbohydr Diet Fibre. 2013;1:156–161 30 March 2018 33. Fuessl HS, Williams G, Adrian TE, Bloom SR. References 17. Berard LD, Blumer I, Houlden R, Miller D, Guar sprinkled on food: effect on glycaemic 1. McRae MP. Dietary fiber intake and type 2 Woo V; Canadian Diabetes Association Clinical control, plasma lipids and gut hormones in diabetes mellitus: an umbrella review of meta- Practice Guidelines Expert Committee. Monitor- non-insulin dependent diabetic patients. Diabet analyses. J Chiropr Med 2018;17:44–53 ing glycemic control. Can J Diabetes 2013;37 Med 1987;4:463–468 2. Evert AB, Boucher JL, Cypress M, et al. Nu- (Suppl. 1):S35–S39 34. HolmanRR, SteemsonJ,DarlingP,TurnerRC. trition therapy recommendations for the man- 18. American Diabetes Association. 2. Classifica- No glycemic benefit from guar administration agement of adults with diabetes. Diabetes Care tion and diagnosis of diabetes: Standards of in NIDDM. Diabetes Care 1987;10:68–71 2014;37(Suppl. 1):S120–S143 Medical Care in Diabetes––2018. Diabetes Care 35. Laajam MA, Jim LK, El-Bolbol S, et al. Guar 3. American Diabetes Association. 4. Lifestyle 2018;41(Suppl. 1):S13–S27 gum in non-insulin-dependent diabetes mellitus: management: Standards of Medical Care in 19. Anderson JW, Bridges SR. Dietary fiber con- a double-blind crossover, placebo-controlled Diabetesd2018. Diabetes Care 2018;41(Suppl. tent of selected foods. Am J Clin Nutr 1988;47: trial in Saudi Arabian patients. Ann Saudi Med 1):S38–S50 440–447 1990;10:525–530 4. Vuksan V, Rogovik AL, Jovanovski E, Jenkins 20. Rodr´ıguez-Moran´ M, Guerrero-Romero F, 36. Lalor BC, Bhatnagar D, Winocour PH, et al. AL. Fiber facts: benefits and recommendations Lazcano-Burciaga G. Lipid- and glucose-lowering Placebo-controlled trial of the effects of guar for individuals with type 2 diabetes. Curr Diab efficacy of Psyllium in type II diabetes. J gum and metformin on fasting blood glucose and Rep 2009;9:405–411 Diabetes Complications 1998;12:273–278 serum lipids in obese, type 2 diabetic patients. 5. Anderson JW, Baird P, Davis RH Jr, et al. Health 21. McGeoch SC, Johnstone AM, Lobley GE, Diabet Med 1990;7:242–245 benefits of dietary fiber. Nutr Rev 2009;67:188–205 et al. A randomized crossover study to assess 37. Li X, Cai X, Ma X, et al. Short- and long-term 6. Tabatabai A, Li S. Dietary fiber and type 2 the effect of an oat-rich diet on glycaemic con- effects of wholegrain oat intake on weight diabetes. Clin Excell Nurse Pract 2000;4:272–276 trol, plasma lipids and postprandial glycaemia, management and glucolipid metabolism in over- 7. Anderson JW, Randles KM, Kendall CW, Jenkins inflammation and oxidative stress in type 2 di- weight type-2 diabetics: a randomized control DJ. Carbohydrate and fiber recommendations for abetes. Diabet Med 2013;30:1314–1323 trial. Nutrients 2016;8:E549 12 Viscous Fiber and Diabetes Diabetes Care

38. Liatis S, Tsapogas P, Chala E, et al. The levels in type 2 (non-insulin-dependent) diabetic 53. InterAct Consortium. Dietary fibre and in- consumption of bread enriched with betaglucan patients with high blood pressure. Ann Clin Res cidence of type 2 diabetes in eight European reduces LDL-cholesterol and improves insulin 1984;16(Suppl. 43):126–131 countries: the EPIC-InterAct Study and a meta- resistance in patients with type 2 diabetes. Di- 46. Vuksan V, Jenkins DJ, Spadafora P, et al. analysis of prospective studies. Diabetologia abetes Metab 2009;35:115–120 Konjac-mannan (glucomannan) improves glyce- 2015;58:1394–1408 39. Ma X, Gu J, Zhang Z, et al. Effects of Avena mia and other associated risk factors for coronary 54. Davison KM, Temple NJ. Cereal fiber, fruit nuda L. on metabolic control and cardiovascular heart disease in type 2 diabetes. A randomized fiber, and type 2 diabetes: explaining the par- disease risk among Chinese patients with di- controlled metabolic trial. Diabetes Care 1999; adox. J Diabetes Complications 2018;32:240–245 abetes and meeting metabolic syndrome cri- 22:913–919 55. Jenkins DJ, Wolever TM, Leeds AR, et al. teria: secondary analysis of a randomized clinical 47. Wolffenbuttel BH, Sels JP, Heesen BJ, Dietary fibres, fibre analogues, and glucose tol- trial. Eur J Clin Nutr 2013;67:1291–1297 Menheere PP, Nieuwenhuijzen Kruseman AC. erance: importance of viscosity. BMJ 1978;1: 40. Niemi MK, Keinanen-Kiukaanniemi¨ SM, Effects of dietary fiber and insulin treatment 1392–1394 Salmela PI. Long-term effects of guar gum and on serum levels of lipids and lipoprotein (a) in 56. Chandalia M, Garg A, Lutjohann D, von microcrystalline cellulose on glycaemic control patients with type II diabetes mellitus. Ned Bergmann K, Grundy SM, Brinkley LJ. Beneficial and serum lipids in type 2 diabetes. Eur J Clin Tijdschr Geneeskd 1992;136:739–742 [in Dutch] effects of high dietary fiber intake in patients Pharmacol 1988;34:427–429 48. Ziai SA, Larijani B, Akhoondzadeh S, et al. with type 2 diabetes mellitus. N Engl J Med 41. Peterson DB, Ellis PR, Baylis JM, et al. Low Psyllium decreased serum glucose and glycosy- 2000;342:1392–1398 dose guar in a novel food product: improved lated hemoglobin significantly in diabetic out- 57. Vuksan V, Sievenpiper JL, Xu Z, et al. metabolic control in non-insulin-dependent di- patients. J Ethnopharmacol 2005;102:202–207 Konjac-mannan and American ginsing: emerg- abetes. Diabet Med 1987;4:111–115 49. Abutair AS, Naser IA,Hamed AT.The effect of ing alternative therapies for type 2 diabetes 42. Sels JP, Postmes TJL, Nieman F, Ebert R, soluble fiber supplementation on metabolic syn- mellitus. J Am Coll Nutr 2001;20(5 Suppl.): Wolffenbuttel BHR, Kruseman ACN. Effects of drome profile among newly diagnosed type 2 370S–380S; discussion 381S–383S guar bread in type 1 and type 2 diabetes melli- diabetes patients. Clin Nutr Res 2018;7:31–39 58. Vuksan V, Sievenpiper JL, Owen R, et al. tus. Eur J Intern Med 1993;4:193–200 50. Food and Drug Administration.. Guidance Beneficial effects of viscous dietary fiber from 43. Uusitupa M, Sodervik¨ H, Silvasti M, for Industry. Diabetes Mellitus: Developing Konjac-mannan in subjects with the insulin re- Karttunen P. Effects of a gel forming dietary Drugs and Therapeutic Biologics for Treatment sistance syndrome: results of a controlled met- fiber, guar gum, on the absorption of glibencla- and Prevention. Rockville, MD, U.S. Department abolic trial. Diabetes Care 2000;23:9–14 mide and metabolic control and serum lipids in of Health and Human Services, 2008 59. Cicero AFG, Derosa G, Bove M, Imola F, patients with non-insulin-dependent (type 2) 51. Silva FM, Kramer CK, de Almeida JC, Borghi C, Gaddi AV. Psyllium improves dyslipid- diabetes. Int J Clin Pharmacol Ther Toxicol Steemburgo T, Gross JL, Azevedo MJ. Fiber intake aemia, hyperglycaemia and hypertension, while 1990;28:153–157 and glycemic control in patients with type 2 guar gum reduces body weight more rapidly in 44. Uusitupa M, Siitonen O, Savolainen K, Silvasti diabetes mellitus: a systematic review with patients affected by metabolic syndrome follow- M, Penttila¨ I, Parviainen M. Metabolic and nutritional meta-analysis of randomized controlled trials. ing an AHA Step 2 diet. Mediterr J Nutrition effects of long-term use of guar gum in the treat- Nutr Rev 2013;71:790–801 Metab 2010;3:47–54 ment of noninsulin-dependent diabetes of poor 52. Vuksan V, Jenkins AL, Rogovik AL, Fairgrieve 60. Ellis PR, Kamalanathan T, Dawoud FM, Strange metabolic control. Am J Clin Nutr 1989;49:345–351 CD, Jovanovski E, Leiter LA. Viscosity rather than RN, Coultate TP. Evaluation of guar biscuits for use 45. Uusitupa M, Tuomilehto J, Karttunen P, Wolf quantity of dietary fibre predicts cholesterol- in the management of diabetes: tests of physio- E. Long term effects of guar gum on metabolic lowering effect in healthy individuals. Br J Nutr logical effects and palatability in non-diabetic vol- control, serum cholesterol and blood pressure 2011;106:1349–1352 unteers. Eur J Clin Nutr 1988;42:425–435