Background

Dementia is characterized by a progressive decline in cognitive performance and executive and social functioning [1]. Based on its clinical characteristics and etiologies, dementia is divided primarily into vascular dementia (VaD) and neurodegenerative dementia, including Alzheimer’s disease (AD) and Lewy-body dementia (LBD) [2]. Mild cognitive impairment (MCI) represents the prodromal stage of dementia; over time, the condition ultimately progresses to dementia [3]. Other than four cholinesterase inhibitors, memantine, and etiological strategies for delaying the course of dementia, there are currently few effective disease-modifying dementia treatments available [4]. Naturally, these are not sufficient to treat dementia after the manifestation of clinical symptoms. Therefore, the goal of dementia therapy should be changed from passive symptomatic treatment to proactive disease prevention. Several pathological studies have suggested that irreversible neuronal apoptosis or neuronal loss induced by oxidative stress plays an important role in the different types and degrees of dementia [5]. Therefore, early detection of antioxidant levels and reduction of oxidative stress can provide opportunities to avoid the deterioration in cognitive function and achieve satisfactory preventive efficacy.

Carotenoids are lipid-soluble conjugated polyene pigments that are primarily present in red and orange fruits and vegetables, such as carrots, tomatoes, watermelons, and guava [6]. Based on their molecular structures (presence or absence of oxygen atoms), carotenoids can be classified into two major categories: carotenes (C40H56) and hydroxy-substituted carotenes, xanthophylls (C40H56O2) [7]. In humans, the former mainly includes lycopene, α-, and β-carotenes, while the latter mainly includes lutein, zeaxanthin, canthaxanthin, and β-cryptoxanthin [8, 9]. Lipophilic structures of carotenoids can allow them to cross the blood-brain barrier (BBB) and accumulate in the brain [10]. The existence of polyene chains in carotenoids determines their chemical properties to scavenge free radicals [11]. Hence, carotenoids exhibit excellent antioxidative and neuroprotective properties and have garnered significant attention in preventing and treating neurological diseases. Several animal studies have shown that oral carotenoid treatments attenuated oxidative stress and cognitive impairment in AD and VaD models [12,13,14,15,16,17]. Such carotenoid intervention tended to be protective against neurodegenerative processes including a cognitive decline in epidemiologic trials [18,19,20]. It should be pointed out that blood carotenoids in humans are mainly got from diet intake or antioxidant supplements [21]. So, after carotenoid ingestion, can the pathological or physiological status of dementia/MCI and healthy populations affect the blood carotenoid levels?

A multitude of trial studies have reported that dementia or cognitive decline was significantly linked to different plasma carotenoid levels, but not consistently [22, 23]. In addition, a recent study meta-analysis explored the association between plasma/serum carotenoids and AD and found that only plasma/serum lutein and zeaxanthin levels were associated with a reduced risk of AD [24]. However, this study did not detect the relevance between other carotenoids and AD risk on one hand and did not extensively examine the associations of blood carotenoids with the risk of MCI and other dementia subtypes, such as VaD and LBD. In contrast, another study found that higher plasma trans-β‐carotene and α-carotene tended to be associated with a higher risk of AD [25]. Thus, given the importance of disease prevention and the incompleteness of previous research, we aimed to perform a more comprehensive and systematic review and meta-analysis to summarize data from previous studies and clarify the association between blood carotenoid levels and multiple dementia subtypes and MCI.

Materials and Methods

Study design

This systematic review and meta-analysis adhered to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses [26]. The review protocol was registered prior to conducting the study and is published online in the PROSPERO database of systematic reviews (https://www.crd.york.ac.uk/prospero/registration number # CRD42020176174).

Patient and public involvement statement

This meta-analysis was based on data obtained from several databases. Patients and the public were not directly involved in this study.

Search strategy

A systematic and comprehensive search was performed using the Web of Science, PubMed, Embase, and Cochrane Library electronic databases, from their inception until February 23, 2023. Relevant publications were initially searched using multiple keywords as follows: “carotenoids”, “lycopene”, “carotene”, “lutein”, “zeaxanthin”, “cryptoxanthin”, “plasma”, “serum”, “dementia”, “vascular dementia”, “Lewy bodies dementia”, “Alzheimer’s disease”, “mild cognitive impairment”, and “MCI”, following controlled vocabulary terms “AND” and “OR”. The relevance of the retrieved papers was evaluated by reviewing their titles and abstracts. The search was limited to peer-reviewed articles published in English. Duplicates of studies were identified and excluded. We also manually screened the reference lists of the relevant articles and reviewed the relevant studies.

Inclusion and exclusion criteria

Two investigators independently evaluated the studies identified for analysis. Any disagreements involving inclusion or exclusion were resolved by a third investigator (QHJ) through consensus and adjudication. In this meta-analysis, eligible studies were selected if they met the following inclusion criteria: (1) case-control, cross-sectional and case-cohort studies with dementia/MCI and control groups; (2) diagnosis of MCI and dementia based on NINDS-AIREN or NINCDS-ADRDA or other diagnostic criteria; (3) articles published in English; and (4) reported plasma or serum carotenoid concentrations. Studies were excluded if they met any of the following exclusion criteria: (1) reviews, letters, book chapters, short communications, abstracts, editorials, or case reports; (2) studies without normal cognition controls; (3) studies without available detailed data; (4) animal or in vitro experiments; (5) duplicated publications; (6) studies of patients without dementia or cognitive examinations and (7) dead patients.

Data extraction

Relevant key information, including first author names, publication year, country, diagnostic criteria, analytical methods, sample sources, dementia type, participant characteristics, and blood carotenoid concentrations, were independently extracted from each selected study by two investigators and re-examined for accuracy by the third investigator. In included studies that reported median and range data, these values were converted into means and standard deviations (SD) using the equations of Hozo et al. [27].

Quality assessment

The overall study quality was independently assessed and scored by two investigators using the Newcastle-Ottawa Scale (NOS) [28], which was based on the following three quality parameters: (1) selection (case definition, representativeness, control selection, control definition; 4 points), (2) comparability (comparability of cases and controls based on the design or analysis; 2 points), and (3) outcome (ascertainment of exposure, the same method of ascertainment for cases and controls, and nonresponse rate; 3 points). The total points varied from 0 to 9; studies with ≥ 7 points were generally considered high-quality; studies that ranged between 5 and 6 points were considered moderate quality, and studies with points < 5 were considered low-quality. The details of these criteria and points of individual studies are listed in Supplemental Table 1.

Data synthesis and statistical analysis

All statistical analyses were performed using STATA 15.0 software (Stata Corp, College Station, TX, USA). Standardized mean differences (SMDs) and 95% confidence intervals (CIs) were considered measures of continuous outcomes in all studies. Statistical heterogeneity across the studies was assessed using the I2 statistic. I2 values < 50% indicated low heterogeneity, and in such cases, a fixed-effect model was selected; if I2 values were > 50%, which indicated high heterogeneity, a random-effect model was employed [29]. Subgroup analyses for disease type and sample source were conducted to ascertain the potential sources of heterogeneity. If enough studies were available (≥ 10 studies), meta-regression was performed on mean age, country, percentage of men, NOS score, and diagnostic criteria as numerical covariates to further investigate the association between blood carotenoid levels and disease with dementia and MCI [30]. Sensitivity analyses were performed by excluding one study at a time to assess the robustness of the pooled SMD [31]. Funnel plots and Egger’s test were used to analyze possible publication bias when the number of studies permitted (≥ 10 studies) [32]. For all analyses, a two-tailed P value < 0.05 was considered statistically significant.

Results

Search results

A flow chart of the study search and selection process is presented in Fig. 1. Altogether, 2140 articles were obtained from the electronic databases in the initial search, and 1698 duplicate files were deleted. After careful screening of titles and abstracts, 403 unrelated articles, such as reviews, abstracts, or animal studies, were removed. Consequently, 23 studies were eligible for inclusion in this systematic review and meta-analysis.

Fig. 1
figure 1

Flowchart of study selection process

General characteristics

The total number of participants was 6610 (50.1–85.9 years), including 1422 with dementia, 435 with MCI, and 4753 controls. Of these, 19 and 4 studies were used for meta-analyses investigating the association between blood carotenoid levels and dementia and MCI, respectively. All the trials were performed between 1998 and 2020, and 13 were published after 2010. These studies included data on lycopene (n = 16), α-carotene (n = 13), β-carotene (n = 17), lutein (n = 11), zeaxanthin (n = 11), and β-cryptoxanthin (n = 9) levels. Fifteen studies were conducted in Europe and two in the USA, with only one in South America, one in Oceania, and one in Asia. The blood carotenoid levels in included studies were measured using high-performance liquid chromatography. The sample size varied from 10 to 1628. The proportion of men patients varied from 0 to 71%. The quality scores of the included studies ranged from 6 to 9, with a mean score of 7.8. The majority of the studies were of high quality, and three studies were of moderate quality. The detailed characteristics of each included study are provided in Table 1.

Table 1 General characteristics of the included studies

Lycopene

Sixteen studies [33,34,35,36,37,38,39,40,41,42,43,44,45,46,47,48] investigated blood lycopene concentrations in 1078 patients with dementia compared with 2433 controls. Dementia was positively associated with reduced blood lycopene levels in a random-effect model (SMD: -0.521; 95%CI: -0.741, -0.301) with a considerable degree of heterogeneity across the studies (I2 = 86.1%) (Fig. 2). Subgroup analysis accounting for dementia type revealed that this association was found for AD and VaD but not for indefinite dementia (ID). Additionally, subgroup analysis stratified by sample source showed that lycopene levels were reduced in both the plasma and serum of patients with dementia (Supplemental Table 2). Meta-regression analysis revealed that the mean age of patients, proportion of men, country, NOS score, and diagnostic criteria did not contribute to the heterogeneity of the included studies (Supplemental Table 3). The sensitivity analysis illustrated that removing any of the studies did not significantly affect the outcome (Supplemental Fig. 1A). Moreover, no publication bias was observed according to Egger’s test (P = 0.686) (Supplemental Table 4).

Fig. 2
figure 2

Forest plot of the relation between blood lycopene (µmol/L) and dementia and MCI, 1999–2020. The data are expressed as SMDs with 95%CIs. The overall effect is represented by a hollow diamond. The horizontal lines represent 95%CI. The sizes of the shaded squares are proportional to study weight

Two studies evaluated the blood lycopene levels in 70 patients with MCI compared to 101 controls [37, 49]. One study reported remarkably lower plasma levels of lycopene in patients with MCI than that in controls (P < 0.05) [49]. In the other study, no significant difference in serum lycopene levels was reported between the two groups [37]. Our investigation found no noticeable association between lower blood lycopene levels and MCI (SMD: 0.059; 95%CI: -0.252, 0.370), applying a fixed-effect model with low heterogeneity (I2 = 0.0%) (Supplemental Fig. 3).

α-Carotene

Thirteen studies [33,34,35,36,37,38,39, 42, 43, 46,47,48, 50] compared blood α-carotene concentrations in 973 patients with dementia and 2244 controls. Meta-analysis results revealed that patients with dementia had remarkably lower blood α-carotene levels compared with the controls (SMD: -0.489; 95%CI: -0.697, -0.281). Because of the substantial heterogeneity across the studies (I2 = 81.6%), a random-effect model was employed (Fig. 3). Subgroup analysis indicated that reduced α-carotene levels were observed in patients with AD and VaD and both plasma and serum samples (Supplemental Table 2). Furthermore, meta-regression analysis showed that none of the aforementioned covariates had a marked impact on heterogeneity (Supplemental Table 3). Sensitivity analysis showed that the inverse association was not completely reversed by omitting any one study, which in turn suggested that the stability of the overall result was robust (Supplemental Fig. 1B). Moreover, a visual funnel plot revealed approximate symmetry (Supplemental Fig. 2A). Egger’s test showed no evidence of publication bias for blood α-carotene levels (P = 0.956) (Supplemental Table 4).

Fig. 3
figure 3

Forest plot of the relation between blood α-carotene (µmol/L) and dementia and MCI, 1999–2020. The data are expressed as SMDs with 95%CIs. The overall effect is represented by a hollow diamond. The horizontal lines represent 95%CI. The sizes of the shaded squares are proportional to study weight

Two studies [37, 51] examining blood α-carotene levels in 224 patients with MCI and 1684 controls reported that patients with MCI exhibited a slight decrease in lycopene levels compared to controls (P < 0.05), which was in line with the pooled findings (SMD: -1.108; 95%CI: -1.937, -0.280). However, significant heterogeneity was observed in the pooled analysis (I2 = 90.7%) (Supplemental Fig. 3).

β-Carotene

Seventeen articles [33,34,35,36,37,38,39,40,41,42,43, 46,47,48, 50, 52, 53] including 1280 patients with dementia and 2703 controls provided relevant data regarding the link between blood β-carotene levels and dementia. There was an association between low blood β-carotene levels and dementia in a random-effect model (SMD: -0. 476; 95%CI: -0.784, -0.168) with substantial heterogeneity among these trials (I2 = 93.9%) (Fig. 4). Subgroup analysis found no significant difference in β-carotene concentrations between patients with dementia and controls when the dementia type was not restricted to AD or the sample source was serum (Supplemental Table 2). The meta-regression analysis showed that none of the above-mentioned covariates owned an impact on heterogeneity. In addition, none of the available moderators could explain the source of the heterogeneity (Supplemental Table 3). Based on sensitivity analysis, the final SMD did not change appreciably after each study was excluded, indicating a stable meta-analysis result (Supplemental Fig. 1C). Egger’s test detected no publication bias in any of the included studies (P = 0.284) (Supplemental Table 4).

Fig. 4
figure 4

Forest plot of the relation between blood β-carotene (µmol/L) and dementia and MCI, 1998–2020. The data are expressed as SMDs with 95%CIs. The overall effect is represented by a hollow diamond. The horizontal lines represent 95%CI. The sizes of the shaded squares are proportional to study weight

Four studies [37, 49, 51, 53] investigating the link between blood β-carotene levels and MCI were included, including 435 patients with MCI and 1916 controls. Of these studies, two studies [37, 49] reported that blood β-carotene levels did not differ between patients with MCI and controls; however, two studies [51, 53] reported markedly lower plasma β-carotene levels in patients with MCI than in controls. Pooled analysis showed a negative correlation between blood β-carotene levels and MCI (SMD: -0.872; 95%CI: -1.226, -0.518; I2 = 84.4%) (Supplemental Fig. 3).

Lutein

Eleven studies [16, 34, 35, 37,38,39, 42, 43, 45, 48, 54] reported blood lutein levels in 851 patients with dementia and 2109 controls. Meta-analysis results showed that patients with dementia had lower blood lutein levels compared with the controls (SMD: -0.516; 95%CI: -0.753, -0.279), and the heterogeneity was significant (I2 = 84.2%) (Fig. 5). In the subgroup analysis by dementia type, lower blood lutein levels were observed in patients with AD and VaD but not in patients with ID. Concerning the sample source, a significant relationship was noted in studies assessing lutein levels in both plasma and serum samples (Supplemental Table 3). In the meta-regression analysis, the percentages of male and NOS score led to 40.63% (P = 0.036) and 20.84% (P = 0.119) heterogeneity, respectively (Supplemental Table 3). Moreover, the sensitivity analysis showed that the statistical significance of the meta-analysis results was not significantly altered after omitting any of these studies (Supplemental Fig. 1D). The funnel plot appeared approximately symmetric, and no publication bias was found according to Egger’s test (P = 0.959) (Supplemental Fig. 2B and Table 4).

Fig. 5
figure 5

Forest plot of the relation between blood lutein (µmol/L) and dementia and MCI, 2002–2017. The data are expressed as SMDs with 95%CIs. The overall effect is represented by a hollow diamond. The horizontal lines represent 95%CI. The sizes of the shaded squares are proportional to study weight

A single publication [37] that investigated plasma lutein levels in 25 patients with MCI and 56 controls reported significantly low plasma lutein concentrations in patients with MCI (P < 0.01) (Supplemental Fig. 3).

Zeaxanthin

Eleven studies [16, 34, 35, 37,38,39, 42, 43, 45, 48, 54] provided data on blood zeaxanthin levels in 851 patients with dementia compared with 2109 controls. A random-effect model was used because significant heterogeneity (I2 = 92.5%, P = 0.000) was observed across these studies. The results showed that blood zeaxanthin levels were substantially lower in patients with dementia compared with controls (SMD: -0.571; 95%CI: -0.910, -0.232) (Fig. 6). In the subgroup analysis, when stratified by disease type, a negative association was observed in both patients with AD, but not in patients with VaD and ID. When the data were stratified by sample source, a significant difference was observed in the plasma but not the serum subgroup (Supplemental Table 2). The meta-regression results indicated that these covariates had no significant impact on the heterogeneity among the studies (Supplemental Table 3). Sensitivity analysis indicated that the elimination of any study did not affect the combined SMD (Supplemental Fig. 1E). Egger’s test for the association between blood zeaxanthin levels and dementia showed no publication bias (P = 0.201) (Supplemental Table 4).

Fig. 6
figure 6

Forest plot of the relation between blood zeaxanthin (µmol/L) and dementia and MCI, 2002–2017. The data are expressed as SMDs with 95%CIs. The overall effect is represented by a hollow diamond. The horizontal lines represent 95%CI. The sizes of the shaded squares are proportional to study weight

Two studies including 224 patients with MCI and 1684 controls, reported on plasma zeaxanthin levels [37, 51]. A single study [37] showed that plasma zeaxanthin levels in patients with MCI were notably lower than in controls (P < 0.01); however, one study [51] showed that patients with MCI had the same plasma zeaxanthin levels as controls. The pooled data suggested no significant between-group difference in plasma zeaxanthin levels (SMD: -0.317; 95%CI: -1.009, 0.376; I2 = 86.8%) (Supplemental Fig. 3).

β-Cryptoxanthin

Nine papers [34, 35, 37,38,39, 42, 43, 46, 48] retrieved blood β-cryptoxanthin levels, including 780 patients with dementia and 2047 controls. The pooled findings showed that patients with dementia had lower blood β-cryptoxanthin levels than controls in a random-effect model (SMD: -0.617; 95%CI: -0.953, -0.281). Statistically, significant heterogeneity was noted among studies on blood β-cryptoxanthin levels (I2 = 91.7%, P = 0.000) (Fig. 7). The subgroup analyses showed that blood carotenoid levels were lower in patients with AD than in the controls, and the pooled SMD was significant both in the subgroups of serum and plasma (Supplemental Table 2). The meta-regression analyses showed that these covariates have no significant contribution to the between-study heterogeneity (Supplemental Table 3). In sensitivity analysis, leaving out any individual study did not cause large fluctuations in the combined SMD results (Supplemental Fig. 1F). Egger’s test for the association between blood β-cryptoxanthin levels and dementia detected no publication bias among these studies (P = 0. 492) (Supplemental Table 4).

Fig. 7
figure 7

Forest plot of the relation between blood β-cryptoxanthin (µmol/L) and dementia and MCI, 2002–2020. The data are expressed as SMDs with 95%CIs. The overall effect is represented by a hollow diamond. The horizontal lines represent 95%CI. The sizes of the shaded squares are proportional to study weight

Two publications reported plasma β-cryptoxanthin levels in 224 patients with MCI and 1684 controls [37, 51]. One study [37] measured significantly higher plasma levels of β-cryptoxanthin in patients with MCI than in controls (P < 0.001), while another study [51] reported appreciably lower plasmaβ-cryptoxanthin levels in patients with MCI than in controls (P < 0.05). The pooled data revealed that no significant differences were observed in plasma β-cryptoxanthin levels between patients with MCI and controls (SMD: -0.839; 95%CI: -2.553, 0.874; I2 = 97.9%) (Supplemental Fig. 3).

Discussions

Main findings

Given the inconsistency of the prior literature, the differences in blood carotenoid concentrations between patients with dementia and cognitively intact controls remain controversial. Meanwhile, information on carotenoid concentrations in the blood of patients with MCI is limited. Therefore, to address these critical issues, we performed the first extensive systematic review and meta-analysis to estimate their relationships and to provide a comprehensive critique of the evidence to establish whether relationships exist.

Emerging evidence suggests that oxidative damage is closely associated with the pathogenesis of dementia and MCI. Disadvantageous events in the brain, such as Aβ deposits, cerebral ischemia, or Lewy body formation, continuously damage mitochondrial function, produce reactive oxygen species (ROS), and induce oxidative stress [55]. ROS containing unpaired electrons with immense chemical reactivity actively reacts with intercellular proteins and lipids that are important for neuronal function, eventually propelling membrane lipid peroxidation and protein structural changes [56]. These alterations cause substantial neuronal dysfunction or loss, in turn forming neuropathological lesions that eventually contribute to dementia onset [57]. Consequently, a reduction in oxidative stress damage is a vital protective mechanism against dementia. Carotenoids with numerous conjugated olefinic bonds in their molecules can easily provide ROS with electrons and yield an inactive p-π conjugated system, thus quenching ROS and terminating harmful oxidative chain reactions [58]. Owning to the structural lipophilicity, several carotenoids allow them to traverse the blood-brain barrier into the brain, wherein they enable the embedding of neuronal membranes, further participating in scavenging ROS and avoiding the oxidation of lipid precursor membrane components [59]. Additionally, carotenoids are involved in the activation of antioxidant signaling pathways and upregulation of downstream antioxidant enzymes, thereby forming antioxidant defenses to combat dementia [60, 61]. Therefore, sustaining higher blood carotenoid levels appears to be key in protecting neurons against oxidative stress, which is also beneficial for preventing and improving dementia and MCI.

In our study, by incorporating the data of all included studies, we attained an answer for our research question: in comparison with cognitively normal controls, patients with dementia had a systematic deficiency in levels of blood carotenoids (P < 0.05). Several studies enabled us to provide consistent evidence of the correlation between lower blood carotenoid levels and dementia. A possible internal pathophysiological cause may be that patients with dementia have to consume more antioxidants, including carotenoids than normal cognitive controls to counteract excessive ROS production, which results in abnormal blood carotenoid changes [62]. From external cause, lower blood carotenoid levels in patients with dementia might result from poor dietary intake of fruits and vegetables. Unfortunately, it is very difficult to identify this issue as exists some degree of potential confounders. Several studies have suggested that patients with dementia have poor diets lacking in fruit and vegetables [63, 64]. However, in this study, most of the cited studies established a baseline prior to clinical studies to largely exclude the influence of carotenoid dietary intake on the comparison between patients with dementia and controls using dietary questionnaires or vegetables/fruits or carotenoid supplementation restriction. Meanwhile, we have to admit that these questionnaire investigations suffer from confounding bias on the long-term dietary habits and the accuracy of recent food/supplement compositions among participants. No matter whether the blood carotenoid deficits were caused by excessive anti-oxidative consumption or possible insufficient intakes and unbalanced diets, patients with dementia should be described as having special carotenoid-rich dietary requirements.

We further analyzed blood carotenoid levels in patients with MCI. We observed that individuals with MCI only possessed lower blood α-carotene, β-carotene, and lutein levels as compared to controls (P < 0.05). This evidence suggests that not all blood carotenoid levels reflect the condition of MCI. A major cause of these results might be the insufficient number of available publications. Another possibility is that the decrease in blood carotenoid levels is not typical in the early and mid-stages of dementia. Therefore, they could not identify sensitive biomarkers for assessing MCI. However, after pooling the overall carotenoid levels in patients with MCI, we found that the pooled SMD for MCI was − 0.650 (-1.045, -0.252), suggesting a tendency toward lower carotenoid levels (Supplemental Fig. 3). In this sense, our data indicate that earlier intervention with carotenoid supplementation in patients with MCI could result in better therapeutic effects. Consistent with this finding, a recent trial reported a significant increase in cognitive function measures following the use of avocado supplements for 6months in patients with MCI, which may be explained by the fact that avocado is rich in multiple carotenoids, and its consumption may increase blood carotenoid levels (ClinicalTrials.gov Identifier: NCT01620567). In 2020, another cognitive impairment study (CARES) demonstrated that a 12-month combined nutritional intervention ofω-3FAs, xanthophyll carotenoids, and vitamin E were able to restore cognitive performance and delay memory impairment in patients with MCI [65]. Notably, because of the relatively small number of studies, the results of the specified subgroup analysis should be interpreted cautiously, and more studies are needed for further assessment.

Furthermore, subgroup analyses were conducted to determine whether the results could be altered according to the different specified subgroups. Subgroup analysis of disease type showed that except for β-carotene, the concentrations of the other five carotenoids in the blood were significantly low in patients with AD and VaD (P < 0.05). Similarly, a previous meta-analytical study reported lower plasma levels of carotenes in individuals with AD as compared to controls [66]. Our observations not only corroborated their findings but also expanded their scope of applicability to the association between plasma/serum carotenoid levels and VaD. Moreover, we pooled their data and discovered that ID was possibly linked with lower blood carotenoid levels (SMD: -0.284; 95%CI: -0.464, -0.101) (Supplemental Fig. 4). Additionally, we assessed the impact of the sample source on the results. Generally, the concentrations of most compounds in plasma and serum samples should not differ according to pharmacokinetic theory. However, subgroup analysis of sample sources showed that lycopene, α-carotene, β-carotene, lutein, zeaxanthin, and β-cryptoxanthin levels were considerably lower in the plasma of patients with dementia. In contrast, no significant difference was observed in β-carotene and zeaxanthin levels in the serum samples. This result was attributed to the limited number of serum samples used for subgroup analyses in the included studies (only six studies), thereby leading to imprecision and inconsistency of the subgroup analyses.

Our meta-analysis has several limitations, which should be acknowledged when interpreting findings. Although this study found evidence of lower blood carotenoid levels in patients with dementia as compared to controls, the relationship between blood carotenoid levels and MCI is not entirely clear. Hence, more prospective studies should be conducted to provide greater evidence of the role of blood carotenoid levels at the MCI stage. Additionally, there are limited studies supplying data regarding blood carotenoid levels in patients with VaD and other dementias. Moreover, because of the lack of special forms of diet, body mass index, and Mini-Mental State Exam scores, we could not directly infer whether nutrient status and dementia severity affect blood carotenoid levels. Therefore, further in-depth studies are needed to identify potential associations. It would also be worth elucidating the cause of the systematic deficiency in blood carotenoids in individuals with dementia.

Conclusion

In conclusion, this meta-analysis found that patients with dementia had significantly lower blood carotenoid levels, despite high heterogeneity. However, we could not draw clear and stable evidence on blood carotenoid levels in patients with MCI, because of less severe disease progression and insufficient data. Therefore, further well-designed and large-scale studies are necessary to reduce the heterogeneity and consolidate our conclusions. As a final recommendation, given their convenience and safety, the intake of carotenoid-rich food or medicine supplements may be beneficial to reducing the risk of dementia in the future.