Introduction

Approved drugs

Currently in Europe there are six drugs legislated for use in the treatment of dementia. These are three cholinesterase inhibitors (donepezil, galantamine, rivastigmine), an NMDA antagonist (memantine), and two new anti-amyloid therapeutics used for an early stage, mild disorders (lecanemab, donanemab). In clinical trials, donepezil and memantine are the most commonly used positive controls (Dekker et al., Reference Dekker, Bouvy, O’Rourke, Thompson, Makady, Jonsson and Gispen-de Wied2019; Public Health - European Commission, n.d.). Donepezil is a reversible inhibitor of acetylcholinesterase (AChE), an enzyme that normally terminates nerve impulses by hydrolysing acetylcholine (ACh) to choline and acetate. By inhibiting AChE, donepezil prolongs the lifespan of acetylcholine in the synaptic cleft, resulting in increased stimulation of postsynaptic cholinergic receptors. It modifies the pathology of AD by increasing acetylcholine levels in the brain while simultaneously reducing AChE activity. Additionally, donepezil blocks the peripheral anionic site of AChE, preventing Aβ-induced damage to cholinergic neurons and inhibiting the nonclassical activity of AChE, which involves Aβ fibrillation. It also exhibits agonist activity at the sigma-1 receptor (σ1R) (Zhang & Gordon, Reference Zhang and Gordon2018; Brewster et al., Reference Brewster, Dell’Acqua, Thach and Sessler2019).

The action mechanism of memantine is primarily based on antagonism of the NMDA (N-methyl-D-aspartate) receptor, as excessive stimulation of the glutamatergic system (with chronic NMDA receptor activity) is thought to contribute to the pathogenesis of Alzheimer’s disease (AD) by inducing excitotoxicity and subsequent neuronal degeneration. Memantine acts as a noncompetitive NMDA receptor antagonist, allowing it to reduce chronic NMDA receptor activity, which leads to cell death. Furthermore, memantine also affects other ion channel receptors belonging to the serotonergic and cholinergic systems, which has a supportive effect on memory and learning. In the serotonin system, it is a noncompetitive 5-HT3 receptor antagonist, which facilitates learning and cognitive functions and may also have antiemetic effects. In the cholinergic system, it is a noncompetitive antagonist of the α7N (nicotinic) receptor, and its action leads to facilitation of LTP (long-term potentiation) and upregulation of nicotinic receptors. In animal models, memantine treatment also resulted in a reduction in Aβ levels and amyloid plaque burden in the brain, as well as a decrease in neuroinflammatory biomarkers (Kishi et al., Reference Kishi, Matsunaga, Oya, Nomura, Ikuta and Iwata2017; Balázs et al., Reference Balázs, Bereczki and Kovács2021). The mechanisms of action of donepezil and memantine are depicted in Figure 1.

Figure 1.

Mechanism of action of donepezil and memantine.

Figure 1. Long description

The diagram compares the molecular mechanisms of Alzheimer’s treatments Donepezil and Memantine. On the left, Donepezil enhances the cholinergic system through reversible AChE inhibition, increasing acetylcholine in the synaptic cleft, stimulating postsynaptic receptors, blocking the peripheral anionic site, exhibiting sigma-1 receptor agonist activity, and reducing amyloid beta levels and inflammation. On the right, Memantine regulates glutamatergic excitotoxicity through noncompetitive NMDA receptor antagonism, reducing excitotoxicity, exhibiting 5-HT3 and alpha7 nicotinic receptor antagonism, facilitating long-term potentiation, and reducing amyloid beta levels and inflammation.

Although donepezil and memantine are the first-line drugs for dementia, they are known for many adverse effects. The most common for memantine are agitation, insomnia, confusion, depression, headache, hypertension, dizziness, fall, accidental injury, urinary incontinence, diarrhoea and influenza-like symptoms (McShane et al., Reference McShane, Westby, Roberts, Minakaran, Schneider, Farrimond, Maayan, Ware and Debarros2019). Adverse effects of donepezil include diarrhoea, nausea, vomiting, dizziness, and anorexia (Chen et al., Reference Chen, Lai and Tao2024). Additionally, in the human studies the efficacy of donepezil was comparable to herbal interventions (Mazza et al., Reference Mazza, Capuano, Bria and Mazza2006; Akhondzadeh et al., Reference Akhondzadeh, Sabet, Harirchian, Togha, Cheraghmakani, Razeghi, Hejazi, Yousefi, Alimardani, Jamshidi, Zare and Moradi2010; Sadhu et al., Reference Sadhu, Upadhyay, Agrawal, Ilango, Karmakar, Singh and Dubey2014; Zhang et al., Reference Zhang, Lin, Zhang, Cui, Gu, Guo, Wu, Li and Song2015; Lin et al., Reference Lin, Huang, Huang and Zheng2020; Wang et al., Reference Wang, Liu, Zhang, Yang, Wang, Wei, Liu, Pei and Li2020).

There are unmet clinical needs to find more effective, better tolerated or complementary therapeutic options, especially in cases of mild to moderate disease severity. This need is driven by many aspects that require improvement in dementia treatment regimens. These include the numerous side effects of currently used medications, the need to improve the tolerability and safety of drugs, and the limited effectiveness of conventional treatments.

Interestingly, in addition to the pharmacotherapy described above, research is increasingly using plant-derived therapeutics. Herbal preparations are administered in many forms, such as infusions, drops, extracts, or encapsulated powders. In studies different preparations of varying concentrations, obtained using different methods are used. Currently, there are no registered herbal medicines on the pharmaceutical market indicated for the treatment of dementia.

The primary goal of this publication is to comprehensively analyse the results of clinical trials and other human studies regarding the efficacy and safety of natural remedies and complex herbal formulas used in patients with dementia. This review aimed to systematically evaluate the effectiveness and safety of herbal interventions and to compare their clinical effects with standard pharmacological treatments and placebo, including the consistency of outcomes across different types of dementia and the areas of greatest clinical benefit.

Materials and methods

Search strategy

Databases Medline (PubMed), Web of Science, Cochrane Library, and Embase were screened for English-language publications concerning phytopharmaceutical interventions in patients suffering from different types of dementia until September 18th, 2025. We used the search terms ‘therapeutic AND plants AND alzheimer’, ‘therapeutic AND plants AND dementia’. Additionally, the search criteria were expanded to include the following terms: ‘herbal and Alzheimer’, ‘phytotherapy and Alzheimer’, ‘botanical and Alzheimer’, ‘medicinal plants and Alzheimer’, and ‘traditional medicine and Alzheimer’. It resulted in a total of 3 255 publications. Databases were screened by two independent researchers.

The exclusion criteria were non-research articles (reviews), in-vitro studies, animal studies, computational studies, study protocols, case reports, administration other than oral, studies on healthy subjects, intervention including phytotherapeutic and other treatments (standard drugs, cognitive training, physical activity), study populations with major comorbidities.

We included studies on different types of dementia – Alzheimer’s disease, vascular dementia, mild cognitive impairment, and others. We included studies on the intervention with one active substance as well as a mixture or multi-component preparations, intervention with standardised substances (for ex. Ginkgo biloba extract EGb 761) or non-standardised (for ex. TCM preparations). Randomised controlled trials (RCTs), double-blind, placebo-controlled studies, and, where necessary, other clinical trials (e.g. open-label, uncontrolled) were included. The search scheme is presented in Figure 2.

Figure 2.

Search scheme.

Figure 2. Long description

Flowchart illustrating the stages of identifying studies via databases and registers. The process begins with records identified from databases such as PubMed, Web of Science, Embase, and registers, totaling 3059 records. Duplicate records, amounting to 790, are removed before screening. No records are marked as ineligible by automation tools, and three records are removed for other reasons. The remaining 2462 records are screened, with 2358 records excluded. Reports sought for retrieval total 82, with 11 reports not retrieved. Reports assessed for eligibility number 71, with exclusions due to language, multiple publications, administration methods, inconsistent study population, inconsistent studied parameters, and other reasons. Finally, 48 studies are included in the review.

The sought data consisted of the type of study, aim, used cognitive tests or scales, results, characteristics of the study group (dementia type, number of patients, sex and age of participants), study duration. In case of missing data on the age range we used mean (SD) given in the publication.

For the risk of bias assessment, two researchers worked independently. We analysed each of the research projects, the presence of a control group/control measurements, the use of validated tests.

Statistical analysis

Comparative analysis of the studies was conducted based on p-value and effect size (Cohen’s d). Effect sizes were calculated based on the numerical values provided in the publication. In the absence of SD change values that were necessary to assess Cohen’s d a sensitivity analysis was performed to determine this value, using three r values; r = 0.2, r = 0.5, r = 0.8. The MMSE and ADAS-cog tests were selected for comparative analysis as the most universal parameter and the ones present in most of the included publications. Calculations were conducted in Numbers (Apple Inc., version 14.4) and Jamovi (version 2.7.13). The eligibility of the studies for the synthesis was assessed by tabulating the interventions and comparing them against the characteristics of the remaining studies. The data we pulled were: phytotherapeutic used as an intervention, dose administered, control type (standard drug/placebo) if applicable, dementia type, used test, number of participants in both groups, p-value of the result, mean and standard deviation of the result.

The certainty of the evidence was not formally assessed due to the heterogeneity of study designs and results. However, key factors such as methodological quality and consistency of findings were considered in the narrative synthesis.

Risk of bias

Risk of bias assessment was performed independently for each included study. Particular attention was paid to elements such as blinded study design, randomisation, and the presence and nature of the control group (placebo-controlled or nonplacebo-controlled).

Statistical synthesis and analysis (methods for calculating effect size)

For all paired comparisons (Intervention vs. Control/Placebo or baseline and endpoint in case of no control group) for which sufficient numerical data (mean, standard deviation, and count) were available, the standardised mean difference (SMD), or Cohen’s effect size (d), was calculated.

The formula for pooled standard deviation (SD_pooled) was used in this calculation to obtain the standardised denominator for the mean difference:

$d = {M_A - M_C \over SD_{{\rm pooled}}}$

where:

$SD_{{\rm pooled}} = \sqrt{{(n_A - 1)SD_A^2 + (n_C - 1)SD \over n_A + n_C - 2}}$

A – active group, C – control group

Cohen’s d values were interpreted according to conventional thresholds (e.g. d = 0.20 small, d = 0.50 medium, d = 0.80 large). If SD change was needed to be calculated, we used r = 0.2, r = 0.5, and r = 0.8 to show the sensitivity of obtained results. Additionally, we calculated Cochran’s Q and I ² for partial heterogeneity analysis.

Discussion and conclusions

The main purpose of this study was to assess the effects of phytopharmaceutics on cognition and neuropsychiatric symptoms in dementia patients. Most of the studied plants, herbs and formulas had a great positive clinical impact on treating mild to moderate dementia, especially Ginkgo biloba, Crocus sativus, and Salvia officinalis, which supports the hypothesis of their therapeutic value (Akhondzadeh et al., Reference Akhondzadeh, Noroozian, Mohammadi, Ohadinia, Jamshidi and Khani2003b, Reference Akhondzadeh, Sabet, Harirchian, Togha, Cheraghmakani, Razeghi, Hejazi, Yousefi, Alimardani, Jamshidi, Zare and Moradi2010; Ihl & Bachinskaya, Reference Ihl, Tribanek and Bachinskaya2010; Ihl et al., Reference Ihl, Tribanek, Bachinskaya and Group2012; Tsolaki et al. Reference Tsolaki, Karathanasi, Lazarou, Dovas, Verykouki, Karacostas, Georgiadis, Tsolaki, Adam, Kompatsiaris and Sinakos2016).

Although Ginkgo biloba showed results of a large effect size, the outcomes were not consistent throughout all of the studies. Some authors obtained significant improvement of EGb 761 interventions in the field of SKT, ADL, and NPI (Kanowski et al., Reference Kanowski, Herrmann, Stephan, Wierich and Hörr1996; Le Bars et al., Reference Le Bars, Katz, Berman, Itil, Freedman and Schatzberg1997; Kanowski & Hoerr, Reference Kanowski and Hoerr2003; Mazza et al., Reference Mazza, Capuano, Bria and Mazza2006; Scripnikov et al., Reference Scripnikov, Khomenko, Napryeyenko and Group2007; Napryeyenko et al., Reference Napryeyenko, Sonnik and Tartakovsky2009; Ihl & Bachinskaya, Reference Ihl, Tribanek and Bachinskaya2010; Ihl et al., Reference Ihl, Tribanek, Bachinskaya and Group2012; Herrschaft et al., Reference Herrschaft, Nacu, Likhachev, Sholomov, Hoerr and Schlaefke2012; Chowdhury et al., Reference Chowdhury, Roy, Reddy, Gupta, Nigam and Hoerr2024), whereas some did not achieve any difference between Ginkgo biloba and placebo (Por & Evans, Reference Por and Evans1998; van Dongen et al., Reference van Dongen, van Rossum, Kessels, Sielhorst and Knipschild2000; Schneider et al., Reference Schneider, DeKosky, Farlow, Tariot, Hoerr and Kieser2005; McCarney et al., Reference McCarney, Fisher, Iliffe, van Haselen, Griffin, van der Meulen and Warner2008). These inconsistencies might be influenced by an abnormal deterioration rate in the placebo group, like in the case of studies by Schneider et al. (Schneider et al., Reference Schneider, DeKosky, Farlow, Tariot, Hoerr and Kieser2005). The other reason might be the differences in test sensitivity. In most cases, the effects measured by the SKT score were significantly better, whereas the same intervention did not show any differences on other scales. This happened in the study by Le Bars et al. on Ginkgo biloba efficacy, a significant improvement on the SKT scale was noted, yet the measurements on ADAS were negligible (Le Bars et al., Reference Le Bars, Katz, Berman, Itil, Freedman and Schatzberg1997). Similar outcomes were revealed by Mazza et al. where the SKT improved significantly, while MMSE did not show any changes (Mazza et al., Reference Mazza, Capuano, Bria and Mazza2006). The differences in the results suggest that the SKT scale may have been more sensitive to changes compared to the ADAS-Cog or MMSE scales in the studied populations, which creates a bias in the interpretation of the results. However, there was a consistent effectiveness of EGb761 on the neuropsychiatric symptoms. NPI scores improved in the studies of Ihl et al., Herrschaft et al. and Scripnikov et al. (Scripnikov et al., Reference Scripnikov, Khomenko, Napryeyenko and Group2007; Ihl and Bachinskaya, Reference Ihl, Tribanek and Bachinskaya2010; Herrschaft et al., Reference Herrschaft, Nacu, Likhachev, Sholomov, Hoerr and Schlaefke2012).

Crocus sativus extract has been repeatedly demonstrated to be comparable to conventional medications. Studies have shown that saffron (at a dose of 30 mg/day) was as effective as donepezil (10 mg/day) in the treatment of mild to moderate AD. In these clinical trials, no statistically significant differences were observed in ADAS-cog and CDR-SB scores between the saffron and donepezil groups (Akhondzadeh et al., Reference Akhondzadeh, Sabet, Harirchian, Togha, Cheraghmakani, Razeghi, Hejazi, Yousefi, Alimardani, Jamshidi, Zare and Moradi2010). Despite indications of potential cognitive benefits, the currently available studies are short-term and include relatively small samples, which limits the ability to draw conclusions about long-term efficacy. Furthermore, saffron demonstrated similar efficacy to memantine in reducing cognitive decline in patients with moderate to severe AD, with no statistically significant differences on the SCIRS, FAST, and MMSE scales (Farokhnia et al., Reference Farokhnia, Shafiee Sabet, Iranpour, Gougol, Yekehtaz, Alimardani, Farsad, Kamalipour and Akhondzadeh2014a).

Similar results were observed for complex TCM formulas. The Jiannao Yizhi formula proved equally effective as donepezil in terms of cognitive function, memory, executive function, and language skills. In this case, scores on the ADAS-cog and MMSE significantly improved in both groups, but the differences at the end of the study were not significant (Guo et al., Reference Guo, Zhou, Wang, Huang, Lee, Su and Lu2013). Similar efficacy in slowing cognitive decline, comparable to donepezil, was also observed in a study evaluating a complex preparation containing, among others, Bacopa monnieri extract (Sadhu et al., Reference Sadhu, Upadhyay, Agrawal, Ilango, Karmakar, Singh and Dubey2014).

The Yishen Huazhuo decoction demonstrated superiority over donepezil in improving ADAS-cog scores. The change in ADAS-cog scores between baseline and post-baseline was significantly different in favour of the herbal formula, and achieved a medium effect size (d=−0.51) (Zhang et al., Reference Zhang, Lin, Zhang, Cui, Gu, Guo, Wu, Li and Song2015). A comparative study suggested that Pushen capsule – a formula consisting mainly of Polygonum multiflorum, Pollen Typhae, Salvia miltiorrhiza, Ligusticum chuanxiong, and paeoniflorin – may have superior beneficial effects over Ginkgo biloba extract in vascular MCI. Pushen capsule improved cognitive function, especially in the field of memory (Li et al., Reference Li, Cao, Deng, Zhu and Yan2019). Wang et al. in a meta-analysis from 2021, showed a promising potential in numerous Chinese herb formulas. One of them was Bushen Jianpi Huatan Pills (formula containing Cistanche deserticola, Rhizoma Alismatis, Polygonum multiflorum thumb), which scored significantly higher compared to placebo in the MMSE scale. It also showed an optimistic effect of the Guipi Decoction (mixture consisting of 14 herbs), combined with donepezil and nimodipine (versus both drugs alone) on the MMSE. The last one, which achieved a significant effect was the Dream sweet oral liquid (17 herbs) with citicoline, compared to the citicoline alone (Wang et al., Reference Wang, Diwu, Liu, Zhou, Sayed, Wang and Gou2021).

In addition to stabilising cognitive function, a key therapeutic goal in dementia treatment is managing behavioural and psychological symptoms of dementia (BPSD), such as apathy, anxiety, agitation, and depression. These symptoms significantly lower patients’ quality of life and increase caregiver distress. Ginkgo biloba extract (EGb 761®) is well documented in this context, demonstrating significant improvements in the NPI score and reducing caregiver stress (Scripnikov et al., Reference Scripnikov, Khomenko, Napryeyenko and Group2007; Ihl et al., Reference Ihl, Tribanek, Bachinskaya and Group2012; Kandiah et al., Reference Kandiah, Ong, Yuda, Ng, Mamun, Merchant, Chen, Dominguez, Marasigan, Ampil, Nguyen, Yusoff, Chan, Yong, Krairit, Suthisisang, Senanarong, Ji, Thukral and Ihl2019). Other phytopharmaceuticals also play a significant role in mood and behaviour modulation. In clinical trials, Melissa officinalis and Salvia officinalis demonstrated a positive effect not only on improving cognitive function (ADAS-cog) but also on reducing agitation (Akhondzadeh et al., Reference Akhondzadeh, Noroozian, Mohammadi, Ohadinia, Jamshidi and Khani2003a, Reference Akhondzadeh, Noroozian, Mohammadi, Ohadinia, Jamshidi and Khani2003b). These findings are consistent with other authors’ reports. Noguchi-Shinohara et al. described Melissa’s properties in the treatment of AD, through anti-Aβ action, cholinergic pathway modulation, and both anti-inflammatory and anti-oxidative characteristics. Participants received the intervention via aromatherapy (Noguchi-Shinohara et al., Reference Noguchi-Shinohara, Ono, Hamaguchi, Nagai, Kobayashi, Komatsu, Samuraki-Yokohama, Iwasa, Yokoyama, Nakamura and Yamada2020). Kennedy et al. used encapsulated dried leaves of Melissa and achieved a great improvement in patient’s calmness and cognition (Kennedy et al., Reference Kennedy, Wake, Savelev, Tildesley, Perry, Wesnes and Scholey2003). Burns et al. also utilised aromatherapy as a form of Melissa’s delivery, and reached similar effects to donepezil concerning, agitation and NPI scores (Burns et al., Reference Burns, Perry, Holmes, Francis, Morris, Howes, Chazot, Lees and Ballard2011). Regarding sage, Lopresti described its wide-range anti-oxidant and anti-inflammatory properties, which were confirmed in numerous in vitro and animal studies (Lopresti, Reference Lopresti2017). Kennedy et al. studied Salvia’s impact on mood in AD patients and found it effective in reducing anxiety, improving calmness, alertness, and task performance (Kennedy et al., Reference Kennedy, Pace, Haskell, Okello, Milne and Scholey2006). Other combination formulations, such as Yokukansan, have shown significant improvement in behavioural and psychological symptoms (BPSD) in dementia with Lewy bodies (DLB), including delusions, hallucinations, and aggression/agitation, without compromising cognitive function (Nagata et al., Reference Nagata, Yokoyama, Yamazaki, Takano, Maeda, Takahashi and Terayama2012). In a study of Cosmos caudatus in elderly individuals with MCI, significant improvements in tension and total mood disturbance were observed (You et al., Reference You, Shahar, Rajab, Haron, Yahya, Mohamad, Din and Maskat2021b).

Although the largest evidence base concerns Ginkgo biloba and Crocus sativus, this review indicates the presence of other, less-researched substances that have demonstrated exceptionally large clinical effect sizes in smaller trials, requiring further clinical research. Zataria multiflora achieved a very large effect size (d = 1.6) on the MMSE scale in patients with mild to moderate AD compared to placebo. The benefits of Zataria on brain health were already proven in animal studies. Arab et al. used a rat model of AD and revealed that it reversed the disorders caused by lipopolysaccharide in the Morris maze and dark chamber tests. It also lowered the oxidative stress through increasing the levels of superoxide dismutase, catalase, and IL-6 (Arab et al., Reference Arab, Hosseini, Marefati, Beheshti, Anaeigoudari, Sadeghnia and Boskabady2022). Eskandari-Roozbahani et al. however, did not reveal any changes in the oxidative levels in rat’s brain tissue. Nevertheless, they reported improvement in the Morris maze test, reduction of AChE and increased levels of Brain-Derived Neurotrophic Factor (BDNF) (Eskandari-Roozbahani et al., Reference Eskandari-Roozbahani, Shomali and Taherianfard2019).

Another plant to achieve a large effect size (d = 0.81) on the MMSE scale in patients with MCI was Cosmos caudatus, which is attributed to its high concentration of polyphenols, such as quercetin and quercitrin. You et al. performed an interesting study on Task-Based Dorsolateral Prefrontal Cortex Activation through an MRI imaging. They administered 500 mg C. caudatus capsules, twice a day for 12 weeks to older adults with MCI. They revealed that this intervention significantly improved the digit span test scores, working memory test scores, and Stroop test score, which implicates a better cognitive control and information processing (You et al., Reference You, Shahar, Mohamad, Rajab, Haron, Che Din and Abdul Hamid2021a).

Many of the beneficial effects of phytopharmaceuticals go beyond the anticholinergic mechanism of action, which is the one modulated by conventional drugs. Oxidative stress is a key factor in the pathogenesis of AD. Studies clearly indicate that antioxidant properties are important. Cosmos caudatus, for example, significantly reduced serum malondialdehyde (MDA) concentrations and increased glutathione (GSH) concentrations (You et al., Reference You, Shahar, Rajab, Haron, Yahya, Mohamad, Din and Maskat2021b). Similarly, Bacopa monnieri prevented oxidative stress by restoring redox balance and increasing the activity of antioxidant enzymes such as SOD, catalase, and glutathione peroxidase (Aguiar & Borowski, Reference Aguiar and Borowski2013). Action against β-amyloid peptide (Aβ) and tau protein aggregation is another key mechanism. In the case of Crocus sativus, crocin and crocetin have been shown to protect against Aβ-induced toxicity. Crocin inhibits tau protein aggregation and the production of Aβ oligomers (Hatziagapiou et al., Reference Hatziagapiou, Kakouri, Lambrou, Bethanis and Tarantilis2019). EGb 761 has been shown to protect neurons against Aβ-induced cell death and inhibit Aβ aggregation (Singh et al., Reference Singh, Srivastav, Castellani, Plascencia-Villa and Perry2019; Lorca et al., Reference Lorca, Mulet, Arévalo-Caro, Sanchez, Perez, Perrino, Bach-Faig, Aguilar-Martínez, Vilella, Gallart-Palau and Serra2023). Some substances act as transient cognitive enhancers. In the case of Panax ginseng, the observed improvement in the ADAS-cog and MMSE scales disappeared after treatment discontinuation, suggesting that its effect is rather the result of neurotransmitter modulation (dopamine and serotonin) than modification of the disease pathology (Lee et al., Reference Lee, Chu, Sim, Heo and Kim2008).

The great advantage of phytotherapeutic interventions is their good tolerability and high safety profile. Studies have shown that Melissa officinalis was well-tolerated and did not cause safety issues (Noguchi-Shinohara et al., Reference Noguchi-Shinohara, Ono, Hamaguchi, Nagai, Kobayashi, Komatsu, Samuraki-Yokohama, Iwasa, Yokoyama, Nakamura and Yamada2020). Furthermore, phytotherapeutic agents often outperform conventional medications in terms of safety. In clinical trials, Crocus sativus demonstrated similar efficacy and a better safety profile compared to donepezil or memantine (Akhondzadeh et al., Reference Akhondzadeh, Sabet, Harirchian, Togha, Cheraghmakani, Razeghi, Hejazi, Yousefi, Alimardani, Jamshidi, Zare and Moradi2010; Farokhnia et al., Reference Farokhnia, Shafiee Sabet, Iranpour, Gougol, Yekehtaz, Alimardani, Farsad, Kamalipour and Akhondzadeh2014b). Ginkgo biloba extract (EGb 761) was also well tolerated, and its use was not associated with (serious) adverse events (van Dongen et al., Reference van Dongen, van Rossum, Kessels, Sielhorst and Knipschild2000; Mazza et al., Reference Mazza, Capuano, Bria and Mazza2006). Furthermore, substances such as Melissa and Salvia have a direct positive effect on neuropsychiatric symptoms (BPSD), including a reduction in agitation and irritability, with agitation being less frequently observed in the groups treated with these herbs than in the placebo groups (Akhondzadeh et al., Reference Akhondzadeh, Noroozian, Mohammadi, Ohadinia, Jamshidi and Khani2003a, Reference Akhondzadeh, Noroozian, Mohammadi, Ohadinia, Jamshidi and Khani2003b).

Many of the clinical trials analysed, especially those involving new or less frequently studied substances, are characterised by significant limitations in sample size. Examples include studies with very small sample sizes, such as the Cistanche tubulosa study and Ginkgo biloba study, which included only 18 patients each (Le Bars et al., Reference Le Bars, Katz, Berman, Itil, Freedman and Schatzberg1997; Guo et al., Reference Guo, Zhou, Wang, Huang, Lee, Su and Lu2013), the Melissa officinalis study (23 participants) (Noguchi-Shinohara et al., Reference Noguchi-Shinohara, Ono, Hamaguchi, Nagai, Kobayashi, Komatsu, Samuraki-Yokohama, Iwasa, Yokoyama, Nakamura and Yamada2020), Choto-san (20 participants) (Yamaguchi et al., Reference Yamaguchi, Matsubara and Kobayashi2004), Yokukansan (12 participants) (Nagata et al., Reference Nagata, Yokoyama, Yamazaki, Takano, Maeda, Takahashi and Terayama2012), ferulic acid and Angelica archangelica (20 participants) (Kimura et al., Reference Kimura, Hayashida, Murata and Takamatsu2011), or the grape powder formulation study (10 participants) (Lee et al., Reference Lee, Torosyan and Silverman2017). This small sample size means that conclusions drawn from even large effect sizes (Cohen’s d), such as those observed in small studies with Vitis vinifera (d = 1.47) (Lee et al., Reference Lee, Torosyan and Silverman2017) or Zataria multiflora (d = 1.6) (Ansari et al., Reference Ansari, Sargazi-Moghadam, Tajadini, Kamali, Kamali, Dehesh, Shojaei and Sharififar2025), are subject to a high risk of bias. Small sample sizes make it difficult to assess whether the obtained large effect sizes (Cohen’s d) are reliable and whether a given phytopharmaceutical could actually be considered a disease-modifying drug (DMT).

The lack of standardisation of plant extracts is one of the most serious challenges in phytotherapy, which is directly reflected in the sources analysed. The use of different extraction methods and preparation forms (standardised extracts, powders, decoctions, concentrates) significantly contributes to the inconsistency of reported results and hinders their direct comparison. The analysed studies use dramatically different formulations, which directly translates into variability in results. For example, one study on Melissa officinalis used an extract standardised for rosmarinic acid content, achieving a medium effect size (d = −0.55) (Noguchi-Shinohara et al., Reference Noguchi-Shinohara, Ono, Hamaguchi, Nagai, Kobayashi, Komatsu, Samuraki-Yokohama, Iwasa, Yokoyama, Nakamura and Yamada2020). However, an older study, which described the preparation only as ‘extract’, reported an extremely high effect size (d = −7.72) (Akhondzadeh et al., Reference Akhondzadeh, Noroozian, Mohammadi, Ohadinia, Jamshidi and Khani2003a). This discrepancy may be due to differences in the concentration of active ingredients obtained using different extraction methods. Studies on Panax ginseng and Cosmos caudatus used powder, while studies on Morus nigra used concentrate (Lee et al., Reference Lee, Torosyan and Silverman2017; You et al., Reference You, Shahar, Rajab, Haron, Yahya, Mohamad, Din and Maskat2021b; Güçer Öz et al., Reference Güçer Öz, Naharcı, Çelebi, Rakıcıoğlu and Göktaş2024). These formulations typically have lower compositional consistency than solvent-based extracts, which may explain why in some cases (e.g. ginseng) no difference was observed compared to placebo. The high heterogeneity index values calculated in our study (I ² from 88.3% to 98.2%) provide mathematical evidence that the differences in results are not accidental. They result largely from the fact that different chemicals were actually tested in different studies under the same plant name (e.g. saffron or ginkgo) in terms of their phytochemical profile.

The clinical efficacy of phytopharmaceuticals must always be considered in the context of their safety, especially in the geriatric population, which often has multiple comorbidities. For instance, Ginkgo biloba’s phytocompounds (gingkolides) may inhibit platelet-activating factor (PAF). Therefore, patients taking anticoagulants (e.g. warfarin) or antiplatelet drugs (e.g. aspirin) should be monitored for an increased risk of bleeding, which is particularly important in older adults with cardiovascular disease (Bent et al., Reference Bent, Goldberg, Padula and Avins2005; Koch, Reference Koch2005). The lack of long-term studies on the toxicity and interactions of these extracts after many months of use requires caution in recommending them as stand-alone therapies in advanced stages of AD. Furthermore, many of the substances analysed, such as Crocus sativus, Melissa officinalis, Salvia officinalis, and Bacopa monnieri, affect the cholinergic pathway by inhibiting acetylcholinesterase (AChE) activity or receptor activity. Using them as a complementary option alongside standard medications, such as donepezil, may lead to additive pharmacodynamic effects. Although phytopharmaceuticals are often better tolerated than conventional medications, the lack of long-term studies on their use makes it difficult to fully assess the risk of cumulative adverse effects in complex drug regimens.

Another significant limitation is the short duration of interventions. Although some studies, such as those involving Crocus sativus, lasted up to 12 months (Farokhnia et al., Reference Farokhnia, Shafiee Sabet, Iranpour, Gougol, Yekehtaz, Alimardani, Farsad, Kamalipour and Akhondzadeh2014a; Tsolaki et al., Reference Tsolaki, Karathanasi, Lazarou, Dovas, Verykouki, Karacostas, Georgiadis, Tsolaki, Adam, Kompatsiaris and Sinakos2016), a significant number of interventional studies lasted only 3 to 6 months, and the shortest only 4 weeks (Kimura et al., Reference Kimura, Hayashida, Murata and Takamatsu2011; Iwasaki et al., Reference Iwasaki, Kosaka, Mori, Okitsu, Furukawa, Manabe, Yoshita, Kanamori, Ito, Wada, Kitayama, Horiguchi, Yamaguchi, Takayama, Fukuhara, Ouma, Nakano, Hashimoto and Kinoshita2012; Nagata et al., Reference Nagata, Yokoyama, Yamazaki, Takano, Maeda, Takahashi and Terayama2012 Albertyn et al., Reference Albertyn, Guu, Chu, Creese, Young, Velayudhan, Bhattacharyya, Jafari, Kaur, Kandangwa, Carter and Aarsland2025). Such a short time frame may be insufficient for evaluating phytotherapeutics as disease-modifying treatments (DMTs). We strongly recommend long-term follow-up periods (e.g. 12–24 months) for future research designs.

Summary

In light of the evidence, phytopharmaceuticals have a promising role as a co-therapeutic option or alternative for patients with dementia who do not tolerate or have contraindications to standard medications. However, further research is necessary to translate these initial promising results into clinical practice. Longer-term, multicenter, randomised controlled trials (RCTs) with larger patient groups using consistent and sensitive outcome measures (such as the ADAS-Cog and NPI) are recommended. This is particularly important for compounds with large effects (e.g. Zataria multiflora and Cosmos caudatus), for which preliminary conclusions are limited by very small sample sizes.

Given that dementia represents a growing public health challenge, this area offers a critical, extensive, and high-priority scope for exploring novel therapeutic avenues.