Diets high in water-soluble fibers such as pectins, gums, mucilages and some hemicelluloses lower blood total cholesterol (TC)⁸ by 2-26% and LDL cholesterol (LDL-C) by 3 -29% in hyper- and normocholesterolemic adults. High intakes of some soluble fibers (e.g., 100 g/d dried beans or oat bran) lower TC and LDL-C levels significantly, even when subjects consume diets providing fat at 37% of total energy. High density lipoprotein cholesterol (HDL-C) concentrations are reported to decrease by 3-20%, increase by 3-35% or remain unchanged by high intakes of soluble fiber (Glore et al. 1994).

Psyllium, which is derived from the seed husk of Plantago ovata, is a gel-forming mucilage (Sandhu et al. 1981) that lowers blood lipid concentrations. As early as 1965, Garvin et al. (1965) showed that psyllium administered as a hydrophilic mucilloid (i.e., the commercial bulk-forming laxative, Metamucil; Procter and Gamble, Cincinnati, OH) reduced serum TC by 9% in five subjects who incorporated 9.6 g psyllium/d into their usual diets for 5 wk. More recent studies (Anderson et al. 1988, Bell et al. 1989, Everson et al. 1992, Neal and Balm 1990) have shown reductions in TC of 5-17% and in LDL-C of 8-20% in adults with mild to moderate hypercholesterolemia given 3.4-10.2 g of psyllium hydrophilic mucilloid two to three times daily for 6-12 wk. The HDL-C increased, decreased or remained unchanged in these studies.

The observation that consumption of psyllium as a bulk-forming laxative sometimes reduced HDL-C as well as TC and LDL-C concentrations led to investigations of the blood lipid effects of psyllium consumed in food form. Ready-to-eat cereals were an appropriate vehicle for psyllium administration, because they are widely consumed by the general public (Haines et al. 1996). Several studies of the effects of psyllium-enriched cereals on blood lipids have been undertaken, mainly among hypercholesterolemic men and women who consumed a low fat diet throughout the intervention. In these studies, TC and LDL-C levels were reduced to varying extents, whereas HDL-C levels were either unchanged or reduced slightly with psyllium cereal consumption.

We were interested in determining the effect of psyllium cereals on blood lipids and the size and consistency of the treatment effect, if any, using meta-analysis. Meta-analysis is a statistical method for evaluating a body of separate but similar experiments (Gibaldi 1993). A meta-analysis is usually undertaken to increase the statistical power for primary outcome measures, clarify results when uncertainties in outcome measures exist, evaluate study validity, improve estimates of the size of an effect or generate hypotheses for future research (Irwig et al. 1994, Sacks et al. 1987). This article describes a meta-analysis of studies of psyllium cereals and was undertaken to measure the consistency and size of the effect of psyllium cereals on blood TC, LDL-C and HDL-C levels in adults.

METHODS

Other food companies known to be involved in psyllium research were contacted as to unpublished human studies involving psyllium cereals. Unpublished studies were considered for inclusion to minimize the possibility of publication bias, because published trials are more likely to be positive than unpublished trials (Ellenberg 1988). Including unpublished trials that met the selection criteria increased the likelihood of obtaining a reliable estimate of the effects of psyllium cereals on blood lipids. Only the Kellogg Company had unpublished studies of psyllium cereals and blood lipids in humans.

Five studies used designs with parallel arms for treatment and control conditions. Seven studies used crossover designs with either two or three treatment periods. Because the length of the washout period between treatment periods was not uniform among the crossover studies, data from the first period only of all crossover studies were used in the meta-analysis to eliminate questions about carryover effects between treatment periods. One ramification of limiting the analyses of crossover studies to data from the first period is that the results of the meta-analysis are more conservative than those reported in the research reports for the individual studies.

In all studies, blood samples were collected for lipid analysis in serum or plasma. The raw data were inspected for missing values, outliers and other anomalies. Some blood lipid values included in the original research studies were excluded from the meta-analysis. The blood lipid values for two subjects in one study (Summerbell et al. 1994) were dropped, because one subject had a serum TC concentration that significantly exceeded the screening criterion and one had abnormal triglyceride levels throughout the study, which yielded questionable LDL-C levels. The data for some subjects in one study (Wolever et al. 1994b) were excluded because they had participated in another study sponsored by the research group and received an intervening treatment that may have influenced blood lipid concentrations. Data from another study (Rippe, J. M., et al., Study C) were excluded, because the actual intake of psyllium cereal in one group was lower than that specified by the study protocol, resulting in a soluble fiber intake similar to that of the control group; values for the remaining subjects were pooled for the meta-analysis. In the studies of Jenkins et al. (1997), data were included for those subjects in Study 1 who had completed the study at the time the meta-analysis was conducted; data for the remaining subjects in Study 1 and for all subjects in Study 2 were not available for inclusion in the meta-analysis.

Analysis of a linear model was performed using PROC GLM of the SAS software program (SAS/STAT versions 6.09 and 6.11, SAS Institute, Cary, NC) to estimate differences between baseline and end-of-study concentrations for TC, HDL-C and LDL-C, controlling for sex and age, two factors known or suspected to be associated with blood lipid concentrations. Least squares means provided the estimates of the size of the effect of psyllium cereal on blood lipid concentrations. A P value of 0.05 or less was considered significant.

Table 1. Study characteristics of published trials of effects of psyllium-enriched cereals on blood cholesterol in hypercholesterolemic adults [View Table]

Factors in the full model included treatment, study, age category, sex, the interaction of treatment with each of the others, and the three-way interaction of treatment, sex and age categories. Any or all nonsignificant interactions were eliminated in a reduced model. The variable "study" was included as a factor in the analysis because the studies differed in psyllium dosage, methodology and subject population. For the age analysis, four age categories grouped by 10-y intervals were created: 39, 40-49, 50-59 and 60 y.

A separate analysis was conducted using only female subjects, who were divided into two age groups (<50 y, 50 y) to provide a rough estimate of the effect of menopausal status on the blood lipid variables. The <50 y category served as a proxy for premenopausal status, and the 50 y category served as a proxy for postmenopausal status. The effect of hormone replacement therapy on blood lipid variables could not be determined because these data were not available for the female subjects.

To test the appropriateness of using only the first period of all crossover studies in the primary meta-analysis, a secondary meta-analysis was conducted using only the crossover studies. A crossover linear model with "study" as an additional term in the model was performed. Results showed an effect (P = 0.0001) due to psyllium for TC and LDL-C that was consistent with the findings of the primary meta-analysis in which only the first arm of the seven crossover studies was used. However, in the secondary meta-analysis, HDL-C decreased by 0.02 mmol/L (P = 0.06) during the control period, a change that was not significant at the a priori P value of 0.05 or less. Because the HDL-C difference was small and not likely to be of biological importance, and because there was no significant effect of period in this analysis, we concluded that it was appropriate to use only the first arm of the seven crossover studies in the primary meta-analysis. Only the results from the primary meta-analysis are described in this article.

RESULTS

Table 2. Study characteristics of unpublished trials of effects of psyllium-enriched cereals on blood cholesterol in hypercholesterolemic adults [View Table]

Table 3. Distribution of subjects by treatment, sex and age category in studies of the effects of psyllium-enriched or control cereals on blood cholesterol in hypercholesterolemic adults [View Table]

Most subjects consumed either a low fat diet (total fat 30% of energy) or the NCEP Step 1 diet during the experimental period. The NCEP Step 1 diet specifies a total fat intake 30% of energy, a saturated fat intake of 8-10% of energy and a cholesterol intake <300 mg/d. The NCEP Step 2 diet is typically prescribed if patients are already following a Step 1 diet at the time their hypercholesterolemia is detected or if the Step 1 diet fails to achieve the goals of diet therapy. The Step 2 diet specifies a total fat intake 30% of energy, a saturated fat intake <7% of energy and less than 200 mg of cholesterol per day (Expert Panel 1993). In general, subjects consumed the background diet, with or without psyllium cereal, for a period of 14-56 d (mean 42 d). The amount of soluble fiber in the psyllium cereals ranged from 3.0 to 12.0 g/d. One study used cornflakes as the control cereal, one used an oat loop cereal, and the remaining 10 studies used a wheat bran cereal as the control cereal.

Data from 404 hypercholesterolemic subjects, 68 women and 336 men, were included in the meta-analysis. Subjects ranged in age from 27 to 72 y (Table 3). Female subjects were enrolled in five of the smaller studies (Anderson et al. 1992; Jenkins, D.J.A., et al., Study A; Jenkins et al. 1997; Summerbell et al. 1994; Wolever et al. 1994b) and were generally underrepresented in these studies. Most female subjects were 50-59 y and older; male subjects were variably distributed across all age categories.

Table 4. Baseline cholesterol concentrations by study and treatment groups and for all studies combined in hypercholesterolemic adults consuming psyllium-enriched or control cereals1 [View Table]

Table 5. Analysis of the difference between baseline and end-of-study cholesterol concentrations in studies of the effects of psyllium-enriched or control cereals on blood cholesterol in hypercholesterolemic adults1 [View Table]

Table 6. Estimates of the size of the effect of psyllium-enriched and control cereals on blood cholesterol in hypercholesterolemic adults1 [View Table]

Table 7. Analysis of data for hypercholesterolemic women categorized by age as a proxy for menopausal status in studies of the effects of psyllium-enriched and control cereals on blood cholesterol1,2 [View Table]

Table 8. Change in distribution of subjects across LDL cholesterol (LDL-C) categories from baseline to end of study in studies of the effects of psyllium-enriched and control cereals on blood cholesterol in hypercholesterolemic adults [View Table]

DISCUSSION

We believe these results represent a conservative estimate of the consistency and size of the effect of psyllium cereal consumption on blood lipid concentrations for two reasons. First, data from the first period only of eligible crossover studies were used in the meta-analysis. This decision resulted in instances in which the significant results achieved in an individual crossover study did not carry over to the meta-analysis, thereby reducing the power of specific studies and yielding more conservative results. Secondly, a variable "study" was included in the meta-analysis to minimize differences in study population, psyllium dose and methodology. There was no significant interaction between the variables "study" and "treatment" for any of the lipid values.

Both published and unpublished studies were included in the meta-analysis. It is generally recognized that restricting the meta-analysis to published studies introduces a bias, because published studies are more likely to report positive research outcomes than unpublished ones (Dickersin and Berlin 1992). Even so, including unpublished studies in the meta-analysis has been criticized on the grounds that they may be unreliable or less rigorous in their design or implementation, factors that possibly contributed to their being rejected for publication (Sacks et al. 1987). Several methods have been proposed to address this problem, including keeping the quality reviewer unaware of such particulars as the researchers, study location and journal of publication. These methods are not foolproof, however. Quality reviewers who are not given information about the study particulars, for example, can identify study researchers through their familiarity with the literature (Haselkorn et al. 1994). In this meta-analysis, the issue of publication bias was addressed by having one author (BHO) systematically apply the same criteria to published and unpublished studies to determine their eligibility for inclusion. One of the major design criteria stipulated that eligible studies must be randomized, controlled experiments using either a crossover design or a design with parallel arms for treatment and control conditions. This decision ensured that studies selected for the meta-analysis were conducted in a consistent, rigorous and valid manner.

The results of the meta-analysis indicate significant reductions in TC and LDL-C, with no change in HDL-C, among adults with mild to moderate hypercholesterolemia who consumed a psyllium-enriched cereal as part of a low fat diet. Soluble fibers other than psyllium, such as legumes, oats and oat bran, pectin, apples and some vegetable fibers, also reduce TC and LDL-C significantly without altering HDL-C, although the effect has not been consistent; in some instances, HDL-C fell 20% following soluble fiber supplementation (Glore et al. 1994).

Incorporating a psyllium-containing bulk laxative into a low fat diet has yielded reductions in TC of about 5% and in LDL-C of about 5-8%, with no change in HDL-C, among adults with mild to moderate hypercholesterolemia (Bell et al. 1989, Neal and Balm 1990). Adding a psyllium-containing bulk laxative to mixed diets providing 30-40% of total energy as fat, a level typical of the North American diet and higher than the fat intakes reported in the studies analyzed in the meta-analysis, has yielded variable results. Reductions in TC and LDL-C concentrations ranged from about 6% (Everson et al. 1992) to 15-20% (Abraham and Mehta 1988, Anderson et al. 1988). These data suggest that adding psyllium in the form of a bulk laxative to typical mixed diets high in fat may result in substantial reductions in TC and LDL-C. These health gains, however, are roughly equivalent to those which can be expected by adhering to a low fat diet that includes a psyllium-enriched cereal or a psyllium-containing bulk laxative.

An important finding of the meta-analysis is that HDL-C levels were not affected by incorporating psyllium-enriched cereals into a low fat diet. Levels of HDL-C are inversely related to the development of coronary heart disease (CHD). Because low HDL-C levels (<0.9 mmol/L) are now recognized as a major risk factor for CHD, interventions that protect or elevate HDL-C are desirable (Expert Panel 1993). The results of the meta-analysis provide support for diet therapies that focus on low fat eating patterns and include psyllium cereals.

The NCEP recommends pursuing diet therapy over drug therapy among individuals with low to moderate CHD risk. Drug therapy should be reserved for high-risk individuals (Expert Panel 1993), a group that may constitute about 7% of U.S. adults (Gaziano et al. 1996). Compared with interventions that focus on diet and physical activity, drug therapy is relatively expensive. A cost analysis of therapy with three alternative agents  cholestyramine resin, colestipol and oat bran  found that the cost per year of life saved among adults with high blood TC (>6.85 mmol/L) was highest for cholestyramine resin packets ($117,400), intermediate for colestipol ($70,900) and lowest for oat bran ($17,800); the yearly cost per person was estimated at $1442 for cholestyramine resin, $879 for colestipol and $248 for oat bran. Most of the cost associated with oat bran use was due to the cost of medical supervision by physicians and dietitians and not due to the cost of oat bran itself (Kinosian and Eisenberg 1988). Treatment for hypercholesterolemia by a dietitian has been estimated at $163 compared with $1450 for drug therapy (McGehee et al. 1995).

The emerging view is that diet therapy alone is not sufficient for patients with overt CHD or for those with severe hyperlipidemia (Caggiula et al. 1996, Kannel 1996). Our findings suggest that some high-risk individuals may respond, even in short-term periods lasting 8 wk, to low fat dietary interventions that include psyllium cereals. Long-term consumption of psyllium cereals for 24-26 wk by adults with hypercholesterolemia has also resulted in reductions in TC and LDL-C of 3-6.7%, with no effect on HDL-C (Anderson, J. W., Davidson, M. H., Blonde, L., Brown, W. V., Howard, W. J., Ginsberg, H., Allgood, L. D. & Weingand, K. W., Veterans Administration Medical Center, Lexington, KY, unpublished data; Davidson, M. H., Maki, K. C., Kong, J. C., Dugan, L. D., Torri, S. A., Hall, H. A., Drennan, K. B., Anderson, S., Fulgoni, V., Saldanha, L. & Olson, B., Chicago Center for Clinical Research, Chicago, IL, unpublished data).

For the ~30% of U.S. adults with mild to moderate hypercholesterolemia and for whom nonpharmacologic interventions to lower cholesterol are appropriate and efficacious (Gaziano et al. 1996, Shaffer and Wexler 1995), our findings support incorporating a psyllium cereal into low fat eating patterns to lower TC and LDL-C. In this meta-analysis, the reduction in TC associated with eating psyllium cereal was attributed to the decrease in LDL-C. Similar results have been shown among studies of hypercholesterolemic adults consuming a low fat diet with added psyllium hydrophilic mucilloid. Because both approaches seem to produce equivalent benefits, the issue for consumers is partly one of taste and preference. Although bulk-forming laxatives containing psyllium are generally well-tolerated, some subjects have reported gastrointestinal discomfort and nausea on first consuming such products (Borgia et al. 1983, Neal and Balm 1990). Adding a psyllium-enriched cereal to a low fat diet may be a more palatable choice than bulk laxatives for some consumers. Having both options available provides alternatives to improve compliance with fiber intake recommendations.

From the standpoint of general eating patterns, adding cereal to the diet provides many benefits. The intake of selected vitamins and minerals increases (Morgan et al. 1986) and the overall intake of fat and cholesterol decreases when cereals are consumed as part of breakfast (Stanton and Keast 1989). In addition, total dietary fiber intake, especially the intake of cereal fiber, is inversely associated with risk of myocardial infarction in men (Rimm et al. 1996), and high dietary fiber intakes are associated with lower risk of CHD (He et al. 1995, Khaw and Barrett-Connor 1987, Kushi et al. 1985). Moreover, adding soluble fiber to diets already low in saturated fat and cholesterol helps lower blood cholesterol concentrations (Jenkins et al. 1993). Finally, with cereal consumption the emphasis remains on foods as sources of nutrients and fiber, which reinforces the basic principles of the 1995 Dietary Guidelines for Americans (Kennedy et al. 1996).

For more than a decade, the dietary recommendations of government agencies (Expert Panel 1993, National Institutes of Health Consensus Development Panel 1985, U.S. Department of Agriculture 1992, U.S. Surgeon General 1988) and nonprofit health organizations (Grundy et al. 1985) have called for a dietary pattern that is low in fat, saturated fat and cholesterol and high in dietary fiber, which can be achieved by eating plenty of grain products, fruits and vegetables. Even so, the average fiber intake of the U.S. population is low, with children and adults consuming less than half of the recommended 20 to 35 g/d (Albertson and Tobelmann 1995, Nicklas et al. 1995). Incorporating a psyllium-enriched cereal into the diet is a relatively simple and inexpensive means of increasing dietary fiber intake, reducing blood cholesterol concentrations and approaching an overall eating pattern that more closely resembles current dietary guidelines.