Supplements · Metabolic & Cardiometabolic
phytosterols decreases LDL cholesterol
In plain terms: Do plant sterols lower "bad" cholesterol?
Strong support Supplements 💰 Industry COI noted
RefutedContestedStrong support
consensus score 0.92
Yes — about 7-12% at the standard dose, but they need fat to absorb and some people don't respond.
📅 Last reviewed: 2026-07-15 ⓘ
Evidence ladder
How far up the ladder this claim has climbed. A high consensus on a low rung means "consistent so far," not "proven in people."
Top evidence so far: All trials, pooled (Meta-analysis)
MechanismIn-vitroAnimalObservationalRCTMeta-analysis
How the studies fall
11 support 0 contradict 0 tested null 2 mixed · 13 sources, 10 independent groups
What the evidence shows
Plant sterols/stanols lower LDL ~7-12% at ~2 g/d (high-grade, dose-responsive), but need a fat phase to be bioavailable and have a real non-responder fraction — a poor fit for a lean sourdough unless a sterol-bearing fat phase is built in.
The evidence (18)
| Source | Grade | Stance | Quality | Finding |
|---|---|---|---|---|
| Soto-Mendez, Rangel-Huerta, Ruiz-Lopez, Martinez de Victoria, Anguita-Ruiz, Gil 2019 · Adv Nutr | meta-analysis | supports | high | Systematic review with its OWN meta-analyses of RCTs (pools primary trial data - not an umbrella review), delivery vehicle = fortified DAIRY, which matters here because the claim's own scope line says phytosterols need a fat phase. SUBSET DENOMINATOR: 'Forty-one studies were finally selected for this synthesis; the selected studies tested dairy products fortified with the following nutrients and bioactive components: phytosterols (n = 31), FAs (n = 8), and vitamin D (n = 2)' - so this claim's estimate rests on the 31-trial PHYTOSTEROL subset, not the 41, and that subset's participant total is not reported in the available text. On the claim's object: 'the consumption of phytosterol-fortified dairy, led to an overall LDL cholesterol reduction of -0.36 (-0.41, -0.31) mmol/L, P < 0.001'. On dose, the authors' wording is weaker than the 'dose-dependent' previously recorded here - 'this decrease was mainly related to the dosage' - with no dose-response coefficient, dose bands or heterogeneity statistic in the available text. QUALITY high on what the paper establishes about itself: 'Risk of bias assessment according to Cochrane guidelines was performed to determine the quality of the trials', MEDLINE and SCOPUS searched to 30 April 2018, tight pooled interval; the RoB RESULT, publication-bias testing and I-squared are cannot-tell from available text and settle nothing either way. Authors' own caveat, which is about the vehicle question rather than the LDL estimate: 'more studies with homogeneous designs are needed to determine the advantages of using dairy products as fortification vehicles'. Off this claim's object and therefore not in the stance: omega-3-fortified dairy lowered LDL -0.18 (-0.27, -0.09) and triacylglycerols -0.18 (-0.32, -0.05) mmol/L, and the vitamin-D and non-omega-3 FA pools could not be run at all ('not possible because of the few available publications'). INDEPENDENCE FLAG, not acted on: these 31 phytosterol-dairy RCTs are a subset of the primary literature already pooled by ras-2014 and re-pooled by the umbrella reviews on this claim - one more pass over a shared corpus, not an independent observation. Abstract-only read (NCBI efetch; Europe PMC full text returned nothing and PMC6518140 is not in the BioC OA subset); venue is a journal supplement issue (Adv Nutr 10(suppl_2)) and funding/COI is not seen. |
| Zurbau A et al 2026 · study_type: meta-analysis | meta-analysis | supports | high | Umbrella review, 14 SRMAs of RCTs: LDL-C MD -0.26 to -0.36 mmol/L; inverse dose-response, ≥0.2 mmol/L reduction at doses as low as 0.5-1 g/day; GRADE certainty high for LDL-C. |
| Costabile A et al 2026 · study_type: RCT | RCT | mixed | low | 9-wk RCT, probiotic + plant sterols/stanols combo in coeliac adults: 'significant reductions in TC, LDL-C, non-HDL-C, ApoB' vs placebo at multiple timepoints; no isolated PS dose or Δ% given. |
| Fong V, Patel SB 2021 · Curr Opin Lipidol | mechanism | mixed | moderate | REVIEW, ZERO WEIGHT, AND OFF THE CLAIM'S SUBJECT. Curr Opin Lipidol invited review by Fong V and Patel SB (Cincinnati), PubMed PublicationType 'Review': 'In this review, we summarize the genetics and mechanisms of sitosterolemia and sterol trafficking'. It reports no measurement of its own — every number in it belongs to a cited primary (Brinton 207,926 samples; Lee 329 children; Tada 487 FH subjects; Reeskamp 3,031 FH patients, lipid data on only n=89; Helgadottir non-HDL-C N up to 943,891; Nomura 47 family members + >386,000 biobank participants). Its exposure is GENOTYPE at the STSL locus, not phytosterol intake. Where it touches this claim's object it is quoting guideline bodies, not evidence: 'plant sterols are endorsed by the American Heart Association (AHA), European Atherosclerosis Society (EAS) and European Society of Cardiology (ESC) for LDL-C lowering due to the ability to compete with cholesterol for absorption'. Its actual argument is on a DIFFERENT object — safety: 'all 3 of the studies highlighted above provided evidence that even mildly elevated levels of plant sterols are associated with risk of atherosclerosis, raising questions about the safety of phytosterol supplementation', concluding 'Until then caution should be used when considering supplementing diets with plant base sterols.' That safety signal is recorded in reread candidates for a claim of its own; it is not a mixed vote on LDL-lowering. Read from PMC full text (PMC8650851, 65,166 bytes), not the abstract. |
| Zhang Y, Zhang Q, Wang X, Jia Y, Niu Q, Ding S, Li W 2025 · Front Pharmacol | meta-analysis | supports | low | Systematic review that runs its OWN meta-analysis of 14 primary RCTs (PROSPERO CRD420251002645) in adults with hyperlipidemia - evidence, not an umbrella review. Exposure is phytosterol-ENRICHED FOODS at 1.6-4 g/day (spread, margarine, yogurt, milk and soy milk, biscuits, black tea, palm oil) for 3 weeks to 6 months; 'Studies on phytosterols combined with other drugs/nutrients intervention' and pharmaceutical phytosterol supplements were excluded by design, so this is squarely the claim's food-matrix exposure. SUBSET DENOMINATOR: the headline is 'This study included 14 randomized controlled trials with a total of 1,088 participants', but this claim's object rests on the 13-trial LDL-C pool - '13 studies demonstrated that phytosterols could significantly reduce LDL-C levels (MD = -0.52, 95% CI -0.66 to -0.38, P < 0.00001), with high heterogeneity (P < 0.00001, I 2 = 77%)' - which drops Wang 2015 (n = 100, LDL-C not among its reported outcomes); the paper never states that subset's own participant total, which is 988 by subtraction. UNIT NOT STATED: no unit accompanies the pooled MD anywhere in the paper or its GRADE table; from the included trials it can only be mmol/L, which would make -0.52 substantially larger than the -0.30 to -0.36 mmol/L pooled by the other phytosterol meta-analyses on this claim - a magnitude to distrust rather than quote. QUALITY low, largely on the authors' own authority: 'According to the GRADE assessment, the certainty of evidence was rated as low for TG, LDL-C, HDL-C, and TC outcomes', because 'These downgrades primarily stem from serious limitations in risk of bias (due to inadequate randomization, allocation concealment, and blinding) and imprecision (attributable to small sample sizes and wide confidence intervals)'; 'Allocation concealment was adequately described in two studies' of 14. What points the other way and is real: leave-one-out stability - 'The pooled effect sizes for TC, LDL-C, HDL-C, and CRP remained statistically significant ( P < 0.05) and unchanged after excluding any single study' - and 'The funnel plot of TC and LDL-C was symmetrical'; but an Egger test was run for TG (P = 0.0154, positive) and HDL-C (P = 0.6054) and is NOT reported for LDL-C, so publication bias on this claim's own outcome is cannot-tell from the paper. Two internal contradictions hold the quality down: the HDL-C section reads 'phytosterols could significantly reduce HDL-C levels' for a value of +0.08 (an increase), and the Strengths section asserts 'the evidence level was high' against the authors' own low and very-low GRADE ratings. CORRECTION to the previous extract, which recorded this paper as showing phytosterols improve 'inflammatory markers' - it showed the opposite: 'No significant change was observed for C-reactive protein (CRP) levels (MD = -0.00, 95% CI -0.01 to 0.00, P = 0.32)' and 'Yet, it fails to demonstrate anti-inflammatory activity as measured by CRP levels.' Off this claim's object and therefore not in the stance: TC -0.65 (95% CI -0.83 to -0.47), HDL-C +0.08 (0.05 to 0.10), TG -0.24 (-0.47 to -0.01) at I 2 = 85% with the authors' own withdrawal of it in the Discussion ('current evidence is insufficient to conclusively support the TG-lowering effect of phytosterols'), and the null CRP pool of 5 trials. INDEPENDENCE FLAGS, not acted on: the 13 LDL-C primaries are 2009-2024 trials (Theuwissen 2009, Athyros 2011, Eady 2011, Buyuktuncer 2013, Hallikainen 2013 and others) largely inside ras-2014's 124-RCT pool and re-counted again by the umbrella reviews on this claim; and group_id 'li-china' does not describe this paper - it is a Shanxi nursing group (first author Zhang Yihua, corresponding Zhang Qian, Shanxi Bethune Hospital / Shanxi University of Chinese Medicine). No commercial COI declared; funded by a Shanxi Bethune Hospital nursing research fund. Read from PMC full text (PMC12263549), not the abstract. |
| Liu B, Chen K, Chen X, Wang J, Shu G, Ping Z, Zhang S 2024 · Phytomedicine | meta-analysis | supports | moderate | UMBRELLA REVIEW THAT POOLS NOTHING — ZERO WEIGHT (evidence_role: review). The paper's own method is appraisal of other people's syntheses: 'We conducted an umbrella review, which incorporated verification spanning a number of meta-analyses and systematic reviews to clarify the link that existed between PS consuming and health outcomes', and 'Methodological quality and the validity of evidence received designation in the included meta-analyses leveraging the Assessment of Multiple Systematic Reviews (AMSTAR-2) and the Grading of Recommendations, Assessment, Development, and Evaluation (GRADE).' MEDLINE indexes it 'Systematic Review' + 'Journal Article', NOT 'Meta-Analysis', and the abstract contains no pooled estimate, effect size, confidence interval or heterogeneity statistic — the RESULTS section carries no number at all. It therefore re-describes the primary literature the vault already counts through ras-2014, s27539156, s31089744, s32823643 and s36180973, and counting it alongside them double-counts the same trials. DENOMINATOR, which the previous one-line extract did not name: 'The databases PubMed, Embase, Web of Science and The Cochrane Library were searched for appropriate research and ultimately included 23 articles involving 79 results' — and those 23 articles / 79 results span ALL outcomes (hypercholesterolemia, diabetes, obesity, hypertension, apolipoproteins, ASCVD), so this claim's object rests on an unnamed SUBSET of the 79 whose own review count and participant total are never reported. ON THE CLAIM'S OBJECT the direction is right, hence stance retained at supports, but it is stated qualitatively with no magnitude: 'Its most essential function is to decrease cholesterol absorption, leading to dramatically reductions in total cholesterol and low density lipoprotein cholesterol.' QUALITY moderate rather than high on what the paper establishes about ITSELF: the previous 'high' rested on the phrase 'with quality assessment', but while the paper says it RAN AMSTAR-2 and GRADE it never reports the RESULT of either in the available text — whether the 23 included reviews rated high, low or critically low is exactly the value an umbrella review is supposed to deliver, and it is absent; no PROSPERO registration, search date range, interval or heterogeneity statistic is reported either. Per the rubric, that absence settles nothing about the underlying literature, but it does mean 'high' is unearned here. GRADE FLAGGED, NOT CHANGED: the note grades this meta-analysis (weight 8) and no meta-analysis was performed — for cultivar to rule on, as with sibling s37739728 on this claim. Off this claim's object and therefore not in the stance, recorded in reread candidates instead: 'The consumption of PS makes a contribution to the alleviation of metabolic conditions such as hypercholesterolemia, diabetes, obesity, and hypertension' and 'utilizing PS products can have a favorable impact on managing apolipoprotein levels along with decreasing the probability of obtaining atherosclerotic cardiovascular disease' — both stated with no estimate or denominator. INDEPENDENCE FLAG, not acted on: this is one of THREE umbrella reviews on this claim (with s37905579 and s42275989) that were each counting at full meta-analysis weight over the same shared primary literature; the other two need the same judgement made on their own text. Authors are seven, all College of Public Health, Zhengzhou University (Liu B, Chen K, Chen X, Wang J, Shu G, Ping Z, Zhang S), the corresponding author additionally at the Food Laboratory of Zhongyuan; 'No conflicts of interest are declared by the authors', and the paper's stated conclusion is developmental — 'highlighting the promising potential for the development of PS into functional foods'. Abstract-only read (NCBI efetch; Europe PMC reports isOpenAccess 'N', inPMC 'N', pmcid null, so no full text was obtainable). |
| Li L et al 2026 · study_type: RCT | RCT | mixed | moderate | RCT, 2 g/day PS alone x60d: TC ↓6.2% (P<0.001) but LDL-C reduction not significant at 60d in PS-only arm; PS+phospholipid combo arm showed LDL-C ↓7.7% only at 30d (P=0.017), lost by 60d. |
| Lobo IMB et al 2023 · Clin Nutr ESPEN | meta-analysis | supports | moderate | SYSTEMATIC REVIEW THAT POOLS NOTHING - ZERO WEIGHT (evidence_role: review). MEDLINE indexes it '<PublicationType UI='D000078182'>Systematic Review</PublicationType>' and NOT as Meta-Analysis, and the methods confirm no synthesis of its own: 'A systematic review was carried out in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analysis, with a PROSPERO protocol registered under number CRD42021249539', after which 'The studies were separated into groups according to the association of the intervention with changes in lifestyle and the characteristics extracted from the studies were summarized and displayed in tables.' There is no pooled estimate, no effect size, no confidence interval and no heterogeneity statistic anywhere in the paper's abstract - every number in it belongs to one of the primary RCTs it describes, so it contributes no observation the vault does not already hold and cannot count alongside the meta-analyses that pool the same literature (ras-2014, s27539156, s31089744). DENOMINATOR, which the previous one-line extract did not name: 'The searches resulted in 977 articles. 22 articles were selected, whose full text was read, and according to the eligibility criteria 10 were incorporated into the review.' - so the whole basis is 10 RCTs, and their combined participant total is not reported in the available text. WHAT IT ACTUALLY CONCLUDES, weaker than the previous extract's 'phytosterols in capsule/tablet form lower LDL': 'Most studies have revealed a positive association between phytosterol supplementation and cholesterol reduction, despite the short duration of interventions', and 'The analyzed studies showed that phytosterol supplements can be useful to modulate the lipid profile, helping to reduce the plasma concentration of LDL cholesterol. However, more research with the aforementioned supplementation in such pharmaceutical formats should be encouraged.' The direction is right, hence stance left at supports, but the role zeroes its weight. SCOPE is genuinely on-claim and on the claim's own open question - the delivery format: 'The use of phytosterols in capsules or tablets can bring advantages in the context of diet therapy, but such format is still less studied when compared to fortified foods', which speaks to this claim's scope line that phytosterols need a fat phase; that finding is recorded in reread candidates rather than as counting weight. QUALITY moderate is retained on what the paper establishes about itself: PRISMA-conformant, PROSPERO-registered, 'The process was conducted by two independent reviewers', five databases plus hand search with no time or language restriction, and 'The authors declare that there are no conflicts of interest related to this study'; against that, a risk-of-bias result, GRADE certainty and publication-bias testing are cannot-tell from the available text and settle nothing either way, and the authors flag 'the short duration of interventions' themselves. GRADE FLAGGED, NOT CHANGED: the note grades this meta-analysis (weight 8) and no meta-analysis was performed - for cultivar to rule on. INDEPENDENCE FLAG, not acted on: single group (all five authors University of the State of Bahia / Federal University of Bahia, group lobo-brazil), but this claim now carries six synthesis-layer rows over one shared primary literature. Abstract-only read (NCBI efetch; Europe PMC returns no PMCID and isOpenAccess=N). |
| Han S, Jiao J, Xu J, Zimmermann D, Actis-Goretta L, Guan L, Zhao Y, Qin L 2016 · Sci Rep | meta-analysis | supports | moderate | Fixed-effect meta-analysis of RCTs adding plant stanols/sterols (1.8-6 g/d, median 2.5 g/d, mostly enriched margarine) to ONGOING statin therapy; both arms received the statin, so the phytosterol was the randomised variable and the design bears on this claim. On the claim's object: LDL-C net change -0.30 mmol/L (95% CI -0.35 to -0.25). SUBSET DENOMINATOR: the headline 'We included 15 RCTs involving a total of 500 participants' is the TOTAL-cholesterol pool; the LDL-C estimate comes from 14 of those trials (Table 3, 'LDL-cholesterol Total 14') because one trial 'lacked LDL-cholesterol index', and the LDL subset's own participant total is not reported separately. Consistency is genuinely good: I-squared = 0% at every endpoint, 'the net change for LDL-cholesterol ranged from -0.55 mmol/L to -0.13 mmol/L' with 8 of 14 trials individually significant, leave-one-out stable (-0.29 to -0.31), meta-regression null, and 'there was no publication bias for total cholesterol, LDL-cholesterol and HDL-cholesterol from the Begg funnel plot' (LDL Egger P = 0.88). QUALITY moderate, not high, on what the paper establishes about ITSELF: no risk-of-bias instrument - 'We assessed the quality of individual studies by reporting the key components of the study designs instead of providing aggregate scores' - plus the authors' own limitations, 'the sample size of individual trials was relatively small, thereby restricting the capacity of randomization to minimize the potential influences of confounding factors', 'allocation concealment, quality of randomization, and details of withdrawals were not always reported', and 'the power of this analysis is limited because of the relatively low number of studies' (trials of 8-141 participants, median 6 weeks); and an industry sponsor on a product-category question - 'This study was supported by Nestec Limited', with four of eight authors employees of Nestle Research Centres. Main analysis used stanol data where a trial tested both; the sterol-substituted sensitivity analysis gave the same -0.30 mmol/L. Off this claim's object and therefore not in the stance: no effect on HDL-C (0.00, 95% CI -0.01 to 0.02) or triglycerides (-0.04, 95% CI -0.09 to 0.01). |
| Hallikainen MA, Sarkkinen ES, Uusitupa MI 2000 · J Nutr | RCT | supports | moderate | Single-blind dose-response CROSSOVER, n=22 hypercholesterolemic men and women, each consuming all five doses for 4 wk: 'In a single-blind design each of 22 men or women consumed five different doses of plant stanol [target (actual) intake 0 (0), 0.8 (0.8), 1.6 (1.6), 2.4 (2.3), 3.2 (3.0) g/d] added as plant stanol esters to margarine for 4 wk. The order of dose periods was randomly determined.' DENOMINATOR: all 22 subjects contribute to every dose estimate - no per-arm subset, but also no independent groups. On the claim's own object: 'The respective decreases for LDL cholesterol were 1.7% (P = 0. 892), 5.6% (P < 0.05), 9.7% (P < 0.001) and 10.4% (P < 0.001)' at 0.8 / 1.6 / 2.4 / 3.2 g - so the LOWEST dose was NULL on LDL-C and significance starts at 1.6 g/d. Total cholesterol fell 2.8% (P = 0.384, also null), 6.8%, 10.3%, 11.3%. CONDITIONS THE AUTHORS ATTACH, and they are most of the paper: 'Although the decreases were numerically greater with 2.4 and 3.2 g doses than with the 1.6 g dose, these differences were not significant (P = 0.054-0.516)', and they conclude 'increasing the dose from 2.4 to 3.2 g does not provide clinically important additional effect' - a plateau their n=22 is underpowered to establish, and one that cuts against ras-2014's pooled 'no plateau below 3 g/d' on this same claim. Also: ApoB down 8.7% (P < 0.001) already at 0.8 g; 'Apolipoprotein E genotype did not affect the lipid responses'; serum plant stanols rose dose-dependently (P < 0.001). QUALITY moderate, not high, on what the paper establishes about itself: single-blind rather than double-blind, n=22, no confidence intervals reported anywhere in the abstract, and a null lowest dose on the claim's object. Abstract-only read (PubMed efetch; not in PMC OA) - washout between periods, allocation concealment and funding are cannot-tell from available text. |
| Ras RT, Geleijnse JM, Trautwein EA 2014 · Br J Nutr | meta-analysis | supports | moderate | 124 RCTs / 201 arms pooled by dose band: -5.7% LDL (95% CI -7.1,-4.4) at mean 0.6 g/d from 24 arms, rising to -12.4% (95% CI -13.6,-11.2) at mean 3.3 g/d from 27 arms; P<0.001 for dose. NO plateau found below 3 g/d - authors state 'at least up to approximately 3 g/d of PS, there is a proportional dose-response effect' and only EXPECT tapering 'at doses slightly higher than 3 g/d'; the 5 arms above 4 g/d were never pooled. Quality moderate: authors state 'heterogeneity tests and publication bias tests were not carried out', publication bias was assumed not tested, 'the quality of studies was not assessed', pool inherited from Demonty 2009 + Musa-Veloso 2011 searches, and 2 of 3 authors are employed by Unilever, which markets plant-sterol foods. |
| O'Neill FH et al] # corrected 2026-08-17 from PubMed; was [Plat 2004 · Nutr Metab Cardiovasc Dis | RCT | supports | moderate | At 1.6 g/d, plant sterol and stanol esters lowered LDL cholesterol by 4-8% and total cholesterol by 3-7% with 'no significant differences during the study' between them, and none between 1.6 and 2.6 g/d stanol; those ranges span two products, two timepoints (1 and 2 months) and two populations (normal subjects and treated familial hypercholesterolaemia patients), with no point estimate, interval or sample size in the available text. The authors' own qualifier is part of the finding: 'the cholesterol-lowering effect of plant sterol esters was attenuated between 1 and 2 months', accompanied by increased serum plant sterols and decreased 7alpha-hydroxy-cholestenone, while stanols 'maintained their cholesterol-lowering efficacy'. |
| Escolà-Gil JC et al] # corrected 2026-08-17 from PubMed; was [Patel 2014 · Curr Atheroscler Rep | mechanism | mixed | moderate | REVIEW, ZERO WEIGHT, AND OFF THE CLAIM'S OBJECT. Narrative review of sitosterolemia (PubMed PublicationType 'Review'); its denominator is the published case literature — 'Around 100 cases have been reported in the literature' — with no data of its own. Subject mismatch: the exposure is a monogenic transporter defect, 'mutations affecting ABCG5 or ABCG8', producing 'a 30-fold to 100-fold increase in plasma concentrations of plant sterols' — not dietary or supplemental phytosterol intake, and orders of magnitude above what enriched foods produce. Object mismatch: no LDL-cholesterol outcome is reported at all; the endpoints are 'xanthomas, premature atherosclerosis, hemolytic anemia, and macrothrombocytopenia', plus mouse-model findings the review explicitly declines to extend to patients ('It remains unclear whether all disease traits are present in sitosterolemic patients'). It also reports 'The drug ezetimibe appears to be effective in reducing plasma plant sterol levels, promotes xanthoma regression', which belongs to claim-ezetimibe-treats-sitosterolemia, not here. Stance retained as recorded: the role zeroes the row and no stance value honestly describes a paper that measured nothing on the object. Read abstract-only (PubMed efetch); the full text was not opened, but review status is a metadata fact, not an inference from the abstract's silence. |
| Xia W, Xiang S, Gaman MA, Jamilian P, Prabahar K, Du G, Gao D 2022 · Phytother Res 2022 Dec;36(12):4398-4408 | meta-analysis | supports | moderate | Random-effects (DerSimonian-Laird) meta-analysis pooling its OWN primary RCT data - evidence, not an umbrella review - but in a POPULATION narrower than this claim: postmenopausal women only ('the purpose of this study was to assess the effect of PS supplementation on the lipid profile in postmenopausal women based on a systematic review of the literature and a meta-analysis of randomized controlled trials'). On the claim's object: 'PS supplementation resulted in a significant decrease in TC (weighted mean difference [WMD]: -16.73 mg/dl) and LDL-C (WMD: -10.06 mg/dl) levels.' DENOMINATOR UNKNOWN: neither the number of pooled RCTs nor the pooled participant total appears anywhere in the available text, and no confidence interval accompanies any WMD there - so the -10.06 mg/dl figure is stored with no interval and no n behind it. CORRECTION to the previous extract's 'strongest at >=2 g/day': that holds for TOTAL cholesterol, not for this claim's object. 'In the stratified analysis, there was a notable reduction in TC and LDL-C levels when the PS dose was >=2 g/day (TC: -22.22 mg/dl and LDL-C: -10.14 mg/dl)' - i.e. TC moves -16.73 -> -22.22 while LDL-C moves -10.06 -> -10.14, which is no move at all. The stratum that does shift LDL-C is adiposity: 'participants with a body mass index >=25 kg/m2' gave '(TC: -20.22 mg/dl and LDL-C: -14.85 mg/dl)'. Both strata are subsets whose own study counts and participant totals are not reported. QUALITY moderate rather than high on what the paper establishes about ITSELF: it reports a four-database search ('PubMed/Medline, Scopus, Embase, and Web of Science were searched to identify suitable papers published until January 18, 2022') and a random-effects model, but the available text carries no risk-of-bias result, no heterogeneity statistic, no publication-bias test result and no interval - whether the full text reports them is cannot-tell from available text and settles nothing either way, so 'high' is unearned here; and the authors' own closing hedge is part of the finding: 'Further high-quality studies are needed to firmly establish the clinical efficacy of PS usage in postmenopausal females.' Off this claim's object and therefore not in the stance: 'No effect of PS supplementation on TG (WMD: -1.14 mg/dl) or HDL-C (WMD: -0.29 mg/dl) concentrations was detected.' Exposure pools sterols and stanols with no separate estimate, which the claim's own scope line already covers. INDEPENDENCE FLAGS, not acted on: the vault stores 'authors: [Xia]', which hides the co-author network - Gaman M-A, Jamilian P and Prabahar K, whose other meta-analyses the paper self-cites (Xiong 2020 berberine; Zhu 2021 tibolone); and the postmenopausal primaries in its reference list (Gylling 2006, Takeshita 2007, Stelmach-Mardas 2014, Granado-Lorencio 2014, Hsu 2017) are old trials likely already inside ras-2014's 124-RCT pool and re-counted by the three umbrella reviews on this claim. Abstract-only read (NCBI efetch; PMC BioC returned 'No result can be found' - not in the OA subset). |
| Zhang YF et al 2024 · Medicine | meta-analysis | supports | moderate | Random-effects (DerSimonian-Laird) meta-analysis pooling its OWN primary RCT data - evidence, not an umbrella review. SUBSET DENOMINATOR: the headline 'Twenty-eight RCTs with a total of 1777 participants (895 cases and 882 controls)' is the TOTAL-CHOLESTEROL pool; this claim's object rests on a smaller one - 'From the 27 studies analyzed, which included 1759 participants (886 cases and 873 controls), data on LDL-c measures was obtained.' On the claim's object: 'This analysis showed a marked reduction in LDL-c levels upon PS supplementation (WMD: -0.38 mmol/L, 95% CI: [-0.50, -0.27], P < .01), with significant heterogeneity observed (I 2 = 98%, P < .01)'; leave-one-out stable. The ApoB half of the old one-line extract has its own third denominator, 26 studies / 1607 participants (810 cases and 797 controls), WMD -0.04 g/L (95% CI -0.06 to -0.02, P < .01), and belongs to a claim the vault does not yet hold. QUALITY moderate rather than high, on what the paper establishes about ITSELF, in the sense of the Huxley/Ding precedent - the authors report bias IN THEIR OWN LDL RESULT: 'the result of Begg test demonstrated a significant publication bias for LDL-c (z = -2.11, P = .0352), but Egger regression asymmetry tests did not yield statistically significant evidence of publication bias (t = 1.32, df = 25, P = .1983)', and they conclude 'These results suggest that publication bias is a possible cause of heterogeneity among study results.' Alongside that: I 2 = 98% on the LDL pool, the highest of any endpoint here, and blinding is weak in the underlying trials - 'Six of the 28 studies included in this analysis adhered to double-blinding protocols, while 15 were evaluated as single-blinded.' A Cochrane risk-of-bias assessment WAS run and reported ('Most parameters employed in risk assessment indicated either a low or ambiguous bias risk within the incorporated trials'), which is why this is moderate and not low. REPORTING DEFECTS FLAGGED, not acted on: the Apo-A1 result is described as 'a marked decline in Apo-A1 levels when supplemented with PS (WMD: -0.01 g/L, 95% CI: [-0.01,0], P = .64)' - P = .64 is a null, not a marked decline; and the LDL Egger P of .1983 is character-identical to the HDL-c Egger P quoted two sentences earlier. DOSE, and the paper contradicts itself across analyses: the categorical subgroup is null on this claim's object - 'The results indicated that the participants' dosage, duration, and health condition did not have a statistically significant impact on the effectiveness of PS supplementation in reducing LDL-c levels' - while the spline dose-response is significant (LDL-c chi 2 = 93.99, P < .001) and the Discussion claims a plateau at 3 g/d. Only the TC subgroup separated by dose (-0.43 vs -0.29 mmol/L, P for subgroup difference = .04). Off this claim's object and therefore NOT in the stance, recorded in reread candidates: TC -0.40 mmol/L (95% CI -0.46 to -0.34, 28 studies / 1777), HDL-c +0.04 mmol/L (95% CI 0.01 to 0.06, 28 studies / 1777), ApoB as above, and Apo-B/Apo-A1 ratio -0.04 (95% CI -0.05 to -0.04) from just 5 studies / 342 participants. The authors' own scope limit is part of the finding: 'the absence of investigations exploring the association between PS intake and clinical endpoints such as cardiovascular events and mortality hampers a comprehensive assessment of the preventive effects of PS on CVDs', and intervention durations 'ranged from 3 to 85-wk, with only 5 extending beyond 3 months'. INDEPENDENCE FLAG, not acted on: 28 trials published 2000-2023, the same primary literature already pooled by ras-2014 (124 RCTs) and re-passed by the three umbrella reviews on this claim; its reference list includes Fuentes 2008, which s32823643 also pools. Vault stores 'authors: [Zhang 2024]', which hides that first author Zhang Yi-Feng and corresponding author Zhang Yang are both Sun Yat-sen University (Shenzhen) - one group, two Zhangs. Funded by the National Natural Science Foundation of China (82000462 and 82170883); 'The authors have no conflicts of interest to disclose.' Read from PMC full text (PMC11495771, fulltext.xml, 167,643 bytes), not the abstract. |
| Nguyen TT et al] # corrected 2026-08-17 from PubMed; was [Hallikainen 1999 · Mayo Clin Proc | RCT | supports | high | 8-wk RCT, 318 mildly hypercholesterolemic US adults randomized after a run-in to four arms - EU 3 G, US 3 G, US 2 G (0.67 g stanol ester x3/d) or placebo spread. SUBSET DENOMINATOR: the headline effect is the US 3 G arm ALONE, not all 318 - 'In the US 3 G group, 3 g daily of stanol esters lowered TC and LDL-C levels by 6.4% and 10.1%, respectively'; per-arm n (~80) is not reported in the abstract, and no CI or p-value is given for the LDL-C change. Dose-dependent vs US 2 G, but the 2 g effect size is not reported; the EU 3 G arm's result is not reported either. TG and HDL-C unchanged; adverse events, serum vitamin A and 25-hydroxyvitamin D not different from placebo. Abstract-only read. |
| Barkas, Nomikos, Liberopoulos, Panagiotakos 2020 · Nutrients | meta-analysis | supports | moderate | Systematic review that runs its OWN pairwise meta-analyses of primary RCTs in familial hypercholesterolemia (not an umbrella review), so it is evidence on this claim rather than a report about evidence. SUBSET DENOMINATORS - the headline 'A total of 17 trials were finally included, with a total of 376 participants across 8 comparison groups' spans EIGHT different dietary interventions; this claim's object rests on two much smaller pools, neither of which has its participant total reported by the paper. Plant STANOLS added to a cholesterol-lowering diet, 2 RCTs, both in CHILDREN with FH (Jakulj 2006 n=42, Gylling 1995 n=14; ~56 summed from Table 1): 'the increased intake of plant stanols reduced study participants' TC (MD: -0.62 mmol/L, 95% CI: -1.13 to -0.11, p = 0.02) and LDL-C (MD: -0.58 mmol/L, 95% CI: -1.08 to -0.09, p = 0.02)'. Plant STEROLS, 4 RCTs (Amundsen 2002 n=41 and De Jongh 2003 n=41 children, Neil 2001 n=62 and Fuentes 2008 n=30 adults; ~174 summed): LDL-C MD -0.45 mmol/L (95% CI -0.74 to -0.16, p < 0.01), TC -0.46 (95% CI -0.76 to -0.17, p < 0.01). No significant heterogeneity in either pool; included trials ran 3-13 weeks. The authors read their own effect as LARGER than in non-FH populations: 'the average LDL-C reduction was 0.45-0.58 mmol/L in our analyses, whereas the corresponding reduction was 0.34 mmol/L in another one including RCTs with dyslipidemic individuals'. QUALITY moderate on what the paper establishes about ITSELF: a risk-of-bias assessment WAS run and reported ('The included trials had either a low or unclear bias risk for most of the parameters'), but 'Only one trial reported adequately on the randomization sequence' of 17, and the authors state 'publication bias cannot be ruled out; there was no adequate data to assess selection, performance and detecting bias. However, a high-risk attrition bias was noticed.' plus 'only a few RCTs have investigated the impact of dietary interventions in patients with FH. Not only their samples were small, but also, they were short-term.' No external funding, no declared COI. NOTE THE PAPER'S OWN TYPO: section 3.4.5 is headed plant sterols but its sentence says 'plant stanols' - the -0.46/-0.45 figures are the STEROL pool (Table S5, Figure 5), as the Discussion confirms. Off this claim's object and therefore NOT in the stance: neither pool moved TG or HDL-C, apoB was null in the sterol pool (MD -0.06 g/L, 95% CI -0.14 to 0.03), and the one head-to-head trial found no stanol-vs-sterol difference (n=18). The review's own PRIMARY outcome was never testable: 'Cardiovascular incidence or mortality were not evaluated in any of the included trials' - the '~12.5% CVD risk reduction' in the Discussion is extrapolated through a third paper's regression, not measured here. INDEPENDENCE FLAG, not acted on: these six 1991-2008 FH trials are old primaries very likely already inside ras-2014's 124-RCT pool and re-counted again by the three umbrella reviews on this claim. Read from PMC full text (PMC7468930), not the abstract. |
| Wang L, Feng L, Prabahar K, Hernández-Wolters B, Wang Z 2024 · Phytother Res 2024 Feb;38(2):507-519 | meta-analysis | supports | moderate | Umbrella meta-analysis of interventional meta-analyses, searched to June 2023. DENOMINATOR IS SYNTHESES, NOT TRIALS: 'Seventeen meta-analyses with 23 study arms were included in the umbrella meta-analysis'; the number of underlying primary RCTs and participants is not reported in the abstract. 'Data pooled from the 23 eligible arms revealed that phytosterol supplementation reduces low-density lipoprotein cholesterol (LDL-C) (ES = -11.47 mg/dL; 95% CI: -12.76, -10.17, p < 0.001)', with TC -13.02 mg/dL (95% CI -15.68, -10.37) and TG -3.77 mg/dL (95% CI -6.04, -1.51); HDL-C unchanged. SUBSET, denominator not reported: 'Subgroup analyses showed that phytosterol administration with dosage ≥2 g/day and duration over 8 weeks and in hypercholesterolemic subjects was more likely to decrease LDL-C, TC, and TG' — no per-subgroup arm count, effect size or interaction test is given, and the threshold reading is in tension with ras-2014's proportional dose-response with no plateau below 3 g/d. Quality moderate: the published abstract reports HDL-C as '95% CI: -0.13, -0.51, p = 258' — an interval whose upper bound is below its lower bound and excludes its own point estimate (0.18), with a p-value that is not a probability; abstract-only read (not open access, Wiley PDF route blocked), so whether the 17 pooled meta-analyses were de-duplicated for shared primary trials, and whether publication bias or AMSTAR-2/GRADE appraisal was performed, cannot be told. |
Disagree, or know a study we missed?
We grade by evidence, not opinions. The way to weigh in is to point us to a study we haven't cited (check the evidence table above first), or to flag a problem with one we have. Every submission is reviewed; if it holds up, the grade updates and shows in Science Changes Its Mind.
Opens a short form. You'll sign in with Google so submissions are tied to a real account — we don't display your identity, and we only accept a link we can verify (PubMed, DOI, ClinicalTrials.gov).
Educational only, not medical advice. Grades and scores reflect published evidence weighted by study design and quality; see the methodology.