Curcumin and Cognitive Function: Why a 2025 Update Reversed the Earlier Null Result
A 2021 meta-analysis found curcumin did nothing for overall cognition. A 2025 update, using different trials, found a significant benefit, but only above a dose threshold, past 24 weeks, and mainly in older and Asian participants. Here is what changed and what did not.
Research context: This article discusses peer-reviewed systematic reviews and meta-analyses of randomized controlled trials. It is intended for informational and educational purposes only and does not constitute medical advice. Consult a qualified healthcare professional before making changes to diet, lifestyle, or supplementation.
Two meta-analyses, four years apart, reached opposite headline conclusions
Curcumin, the primary polyphenol in turmeric, has one of the more confusing evidence trails in the cognitive-supplement space, because the two most recent attempts to pool its randomized trial evidence arrived at different headline answers. A 2021 systematic review and meta-analysis found no significant effect of curcumin on overall cognitive function. A 2025 update, published in Frontiers in Nutrition, found a significant benefit on global cognition versus placebo. Neither paper is wrong. They pooled an overlapping but different set of trials, and the 2025 update's own subgroup analysis explains most of the gap: the benefit depends heavily on dose, duration, and who is taking it (Wang W, Zhao R, Liu B, Li K, "The effect of curcumin supplementation on cognitive function: an updated systematic review and meta-analysis," Frontiers in Nutrition 2025;12:1549509, DOI: 10.3389/fnut.2025.1549509, PMID: 40308636).
What curcumin is supposed to do
Curcumin's proposed route to cognitive benefit runs through anti-inflammatory and antioxidant signaling, primarily inhibition of NF-κB-driven neuroinflammation, along with preclinical evidence of upregulated brain-derived neurotrophic factor (BDNF), the same neurotrophic pathway discussed in our BDNF and neuroplasticity review. The practical problem has always been bioavailability: unmodified curcumin is poorly absorbed and rapidly metabolized, which is why many of the trials below used enhanced-bioavailability formulations rather than plain turmeric extract. Curcumin is also one of ten plant compounds ranked head-to-head on cognition in a separate January 2026 network meta-analysis, where it placed second on executive function in healthy older adults behind a bacopa compound, a different trial population and a different statistical approach from the two meta-analyses discussed here. Readers can see the full ranking in our plant compounds network meta-analysis review.
The 2021 verdict: null on overall cognition, positive on one domain, a safety flag
The earlier meta-analysis, led by I-Chen Tsai, pooled eight randomized controlled trials covering 389 participants and broke the results down by individual cognitive domain rather than reporting a single pooled score. On overall cognition, six of those studies showed no significant difference between curcumin and placebo (Hedges' g = 0.340, 95% CI -0.353 to 1.033, p = 0.337). The one domain that did separate from placebo was working memory, assessed across three studies (Hedges' g = 0.396, 95% CI 0.078 to 0.714, p = 0.015, I² = 0%), a result that held up under sensitivity testing. Processing speed showed a borderline, non-significant trend favoring curcumin (Hedges' g = 0.303, 95% CI -0.013 to 0.619, p = 0.060). The same review also reported a safety finding that gets less attention than the cognitive numbers: curcumin carried significantly higher odds of gastrointestinal adverse events than placebo (odds ratio 3.019, 95% CI 1.118 to 8.150, p = 0.029), though overall adverse event rates and trial withdrawals due to side effects did not differ significantly (Tsai IC, Hsu CW, Chang CH, Tseng PT, Chang KV, "The Effect of Curcumin Differs on Individual Cognitive Domains across Different Patient Populations: A Systematic Review and Meta-Analysis," Pharmaceuticals 2021;14(12):1235, DOI: 10.3390/ph14121235, PMID: 34959636).
The 2025 update: a different trial set, and a significant pooled result
The 2025 update pooled nine randomized controlled trials, comprising 12 independent comparisons across 501 participants, a larger and partly non-overlapping set of trials than the 2021 review. This time, global cognition showed a significant benefit of curcumin over placebo (standardized mean difference 0.82, 95% CI 0.19 to 1.45, p = 0.010), though heterogeneity across the pooled trials was high (I² = 88.7%), meaning individual trials varied substantially in how large an effect they found. The authors explicitly note that their result contradicts the null overall-cognition finding from the 2021 review and an earlier 2019 review, and attribute the shift to the newer trials added to the pool rather than to a reanalysis of the same data.
Why the headline number is less useful than the subgroup breakdown
The pooled 0.82 SMD is the number most likely to get quoted, and it is also the number most dependent on who, how much, and how long. A linear dose-response meta-regression across the 12 comparisons was not statistically significant (p = 0.636), but a curvilinear spline analysis found an optimal dose near 0.8 g/day, with the benefit rising with dose up to that point (p = 0.047) and a non-significant decline above it (p = 0.194), based on only 12 data points. Duration mattered more clearly: trials running 24 weeks or longer showed a significant benefit (SMD 1.15, 95% CI 0.13 to 2.18, p = 0.027), while trials shorter than 24 weeks did not (SMD 0.22, p = 0.321). Age and population mattered too: participants aged 60 and over showed a significant benefit (SMD 1.12, 95% CI 0.03 to 2.21, p = 0.044) while those under 60 did not (SMD 0.31, p = 0.085), and trials conducted in Asian countries showed a significant benefit (SMD 0.96, 95% CI 0.08 to 1.83, p = 0.032) while the Western trials, run in the United States and Australia, did not (SMD 0.69, p = 0.199). Enhanced-bioavailability curcumin formulations showed a larger, significant effect (SMD 1.57, 95% CI 0.29 to 2.84, p = 0.016) compared with standard formulations, consistent with the absorption problem that has dogged curcumin research for years.
What the evidence does not establish
Several limits apply to the 2025 update specifically. Two of the nine included trials carried a high risk of bias; excluding them left the global cognition result significant but noticeably weaker (SMD 0.63, p = 0.046). Several of the subgroup splits, including the dose-response curve, rest on as few as 12 independent comparisons, which is a thin base for estimating an "optimal dose." The review was not prospectively registered and had no published protocol, unlike the plant-compounds network meta-analysis discussed above, which was PROSPERO-registered before data collection began. And the 2025 paper's own discussion describes the stronger-response group as including "European" subjects, while its methods and subgroup table state that none of the included trials were conducted in Europe and that the Western group consisted of trials from the United States and Australia, an inconsistency worth flagging rather than resolving on the reader's behalf. Neither meta-analysis reports results in people already diagnosed with dementia or mild cognitive impairment as a distinct curcumin-cognition outcome, so this evidence speaks to general and age-related cognitive performance, not to a clinical population.
Key takeaways
A 2021 meta-analysis of eight trials found curcumin produced no significant benefit to overall cognition, a significant benefit to working memory specifically, and a higher rate of gastrointestinal side effects than placebo. A 2025 update, pooling nine partly different trials, found a significant benefit to global cognition overall, but subgroup analysis showed that benefit was concentrated in trials lasting 24 weeks or longer, in participants aged 60 and over, in trials conducted in Asian countries, and in enhanced-bioavailability formulations, with a proposed optimal dose near 0.8 g/day based on a thin evidentiary base. High heterogeneity across trials, a high risk of bias in two of the nine studies, and the absence of a registered protocol all temper how far the 2025 pooled result should be read as a settled finding.
FAQ
Does curcumin improve cognition or not?
The honest answer is that the two most recent pooled analyses disagree on the headline number, and the disagreement is explained mostly by which trials were included. The 2021 review found no significant overall effect; the 2025 update, using a different and partly non-overlapping trial set, found a significant effect. The 2025 update's own subgroup data suggests the benefit shows up mainly in longer trials (24 weeks or more), older participants (60 and over), and enhanced-bioavailability formulations, not uniformly across all curcumin use.
What dose of curcumin did the 2025 analysis identify as most effective?
A curvilinear analysis of 12 trial comparisons pointed to an optimal dose near 0.8 g/day, with effects increasing with dose up to that point and a non-significant decline above it. This is a dose-response estimate from a small number of comparisons, not a clinical dosing recommendation, and a simple linear model across the same data did not reach statistical significance.
Is curcumin safe to take for cognitive purposes?
The 2021 meta-analysis found curcumin carried significantly higher odds of gastrointestinal adverse events than placebo, though overall adverse event rates and discontinuations due to side effects did not differ significantly from placebo. Neither meta-analysis discussed here reports a serious safety signal beyond gastrointestinal tolerability, but neither pools data in people with existing gastrointestinal, liver, or gallbladder conditions as a distinct subgroup.
Why did a newer meta-analysis reach a different conclusion than an older one?
The 2025 update pooled nine trials and 501 participants, the 2021 review pooled eight trials and 389 participants, and the two sets of included studies were only partly the same. The 2025 authors attribute the shift to the newer trials entering the literature rather than to a reanalysis of the 2021 data, and they note that heterogeneity across the pooled trials was high (I² = 88.7%), meaning individual trials varied widely in the size of effect they found. That volatility is itself informative: a single updated meta-analysis is a snapshot of the current trial pool, not a final verdict.
Primary sources: Wang W, Zhao R, Liu B, Li K. "The effect of curcumin supplementation on cognitive function: an updated systematic review and meta-analysis." Frontiers in Nutrition 2025;12:1549509. DOI: 10.3389/fnut.2025.1549509. PMID: 40308636. Tsai IC, Hsu CW, Chang CH, Tseng PT, Chang KV. "The Effect of Curcumin Differs on Individual Cognitive Domains across Different Patient Populations: A Systematic Review and Meta-Analysis." Pharmaceuticals 2021;14(12):1235. DOI: 10.3390/ph14121235. PMID: 34959636. This article summarizes published, peer-reviewed research and is not a substitute for individualized medical advice.
Disclaimer
This article is for educational and research-context purposes only. It does not constitute medical advice and is not a substitute for consultation with a qualified healthcare practitioner. The meta-analyses discussed reflect published, peer-reviewed findings pooled across trials of varying quality and design, not personalized clinical guidance. Curcumin can interact with blood-thinning medications and may affect people with existing gastrointestinal, liver, or gallbladder conditions. Always consult a doctor before beginning, modifying, or stopping any supplement, particularly if you have a diagnosed cognitive or gastrointestinal condition, take prescription medications, are pregnant or breastfeeding.