Effects of carotenoid supplementation on liver enzymes in adults: a GRADE-assessed systematic review and dose–response meta-analysis of randomized controlled trials

Abstract

Background and aims

Carotenoids are known for their beneficial effects in improving chronic diseases through their antioxidant properties. However, there has been no meta-analysis on the effects of carotenoids on liver enzymes and the evidence is inconsistent. So, this study aimed to evaluate the effects of carotenoid supplementation on liver enzyme levels in adults.

Methods

Through November 2024, the PubMed, Scopus and Web of Science electronic databases were searched for eligible trials evaluating the carotenoid supplementation effects on the Alanine aminotransferase (ALT), Aspartate aminotransferase (AST), Alkaline phosphatase (ALP) and Gamma-glutamyl transferase (GGT). The 95% confidence intervals (CIs) and weighted mean differences (WMDs) were calculated using the random effects model. As part of standard methods, dose–response, meta-regressions, sensitivity, and publication bias analyses were performed. Evidence certainty was assessed using GRADE (Grading of Recommendations for Assessment, Development, and Evaluation).

Results

Of the 15 studies (20 arms; n = 757), 7 included healthy participants, while 8 involved non-healthy individuals, including 4 studies on prediabetes or diabetes. Pooled estimates indicated non-significant effects of carotenoid supplementation on all enzymes including ALT (WMD: -2.25 IU/L, 95%CI: -4.84,0.34, P = 0.089), AST (WMD: -0.46 IU/L, 95%CI: -1.25,0.34, P = 0.259), ALP (WMD: -0.34 IU/L, 95%CI: -6.89,6.21, P = 0.918) and GGT (WMD: -0.43 IU/L, 95%CI: -3.06,2.21, P = 0.751) levels non-significantly. Significant reductions in ALT levels occurred in < 12 weeks (P = 0.028), BMI ≥ 25 (P = 0.045), and among non-healthy participants (P = 0.015). AST levels were significantly reduced in non-healthy participants (P = 0.003) with ages > 50 (P = 0.003) as well as GGT levels (P = 0.011) in non-healthy participants.

Conclusion

Carotenoid supplementation might be beneficial in reducing liver enzymes, especially in non-healthy participants and in those with a BMI ≥ 25 kg/m². However, more trials with standard methods are required.

PROSPERO registeration code

CRD42024612956.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12906-025-05201-5.

Introduction

Liver disease is a primary health issue that affects millions of people worldwide. The condition is characterized by progressive degeneration of liver tissue, resulting in impaired liver function and the potential development of severe and life-threatening complications [1]. Serum levels of Alanine aminotransferase (ALT), Aspartate aminotransferase (AST), Alkaline phosphatase (ALP), and Gamma-glutamyl transferase (GGT) enzymes are commonly utilized as biomarkers for liver diseases [2–4]. Increased levels of these enzymes are also related to many chronic diseases like diabetes, cardiovascular disease (CVD) and metabolic syndrome (MetS) [5–7]. Any form of liver cell damage can lead to an increase in ALT levels [8]. High serum levels of GGT and ALP have been linked to an increased risk of CVDs [6, 9, 10]. Additionally, previous studies have indicated a positive correlation between abdominal obesity and elevated ALT and GGT levels [11]. Population studies have reported that increased serum liver enzymes are present in up to 24.5% of individuals [12, 13]. According to a number of findings, medication therapy and lifestyle modification can control liver enzyme levels; Considering that medication therapy can cause other challenges such as side effects, a healthy lifestyle including a healthy diet [14] that is rich in vegetables and fruits and physical activity is a more effective therapy to improve serum liver enzymes levels [15–17].

Carotenoids defines as a diverse group of compounds synthesized by photosynthetic organisms such as algae, plants, and bacteria, as well as by certain non-photosynthetic fungi, archaea, protists, and bacteria. They belong to the terpenoid family, which also encompasses sterols, hopanoids, and ubiquinones [18]. In Western diets, there are over 50 different carotenoids; However, zeaxanthin, alpha-carotene, beta-carotene, Lycopene, lutein, and β-Cryptoxanthin account for more than 70% of the carotenoids present in plasma and tissues [19]. These compounds are particularly plentiful in orange-yellow vegetables and fruits, as well as in dark green leafy vegetables [20].

Carotenoids significantly contribute to human health by functioning as biological antioxidants. They help safeguard cells and tissues from the harmful impacts of free radicals and singlet oxygen [21], thereby boosting immune system performance [22]. Additionally, they may offer protection against cataracts [23], coronary heart disease, certain cancers [24, 25], diabetes [26], hypertension [27], and stroke [21].

Previous meta-analyses had reported the beneficial carotenoid effects on cognitive outcomes [28], inflammation [29], oxidative stress [30] and blood pressure [31]. However, the impact of carotenoids on liver enzymes is still unclear and there is a lot of inconsistency between randomized controlled trials (RCTs). Some studies suggest that carotenoids may improve liver enzyme levels [32, 33], while others indicate an increase in these enzymes [34, 35]. Given the potential effects of carotenoids on improving liver enzyme levels and the fact that no meta-analysis has yet examined their effects, the current dose–response meta-analysis aimed to evaluate the effects of carotenoid supplementation on liver enzyme levels in adults.

Methods

Search strategy

The present study was conducted and reported based on the PRISMA guidelines (Preferred Reporting Items of Systematic Reviews and Meta-Analyses) [36]. In addition, the protocol has been registered by "PROSPERO" (no. CRD42024612956). Web of Science, SCOPUS, and PubMed databases were searched for relevant papers on our topic with no limitation on the date or language of the studies. Related RCTs published till November 2024 were considered in the search. Detailed information about the search strategy used for these databases is provided in Table S1.

Study selection

To determine eligibility, two independent investigators (SSH and MA) evaluated inclusion criteria. Disagreements resolved through discussion. A third author was consulted to resolve conflicts between the authors (MB). In accordance with PICO (Population, Intervention, Comparator, Outcome) (Table 1), studies were selected on the basis of the following criteria:

Table 1.

PICOS criteria for inclusion and exclusion of studies

Parameter	Criteria
Participant	Adults
Intervention	Carotenoid supplementation
Comparator	Matched control group
Outcomes	ALT, AST, ALP, GGT
Study design	Randomized controlled trial

Abbreviations: ALT Alanine aminotransferase, AST Aspartate aminotransferase, ALP Alkaline phosphatase, and GGT Gamma-glutamyl transferase

The design of parallel or crossover RCTs;

The article could provide information on the effects of carotenoid supplementation on liver function; (At baseline and at the end of follow-up liver enzyme levels with standard deviations [SDs], standard errors [SEs], or 95% confidence intervals [CIs] were available for the control and intervention groups).

Using only the carotenoid supplement to differentiate the control group from the intervention group;

A minimum of 2 weeks of intervention;

Adult participants (age ≥ 18);

Studies with supplement interventions (e.g., capsules, tablets, or fortified foods), not diets (e.g., high tomato diet as intervention).

The exclusion of studies was caused by several factors:

A non-RCT study design;

Inability to extract the carotenoids' net effects (e.g., multi-nutrient interventions);

Carotenoid intervention duration was < two weeks;

Lack of baseline or follow-up liver parameters in studies;

Dietary intervention studies, not supplementation studies.

Data extraction

Data extraction was preformed independently by two authors (SSH and MVB). As part of the data extraction process, the following information was collected from each article: first author’s name, publication year, sample size, study design, study location, health status of participants, type, duration, dose of interventions and placebos, mean age and gender of participants. Each dose of supplementation was considered in the meta-analysis for studies reporting data at multiple doses.

Quality assessment

The Cochrane Risk of Bias 2.0 (RoB 2) was used to determine the risk of bias in the studies included in this analysis. It was classified into some concerns, low and high bias risk based on the bias risk of the study [37]. In order to assess and summarize the quality of all evidence across studies, GRADE (Grading of Recommendations Assessment, Development, and Evaluation) was used. GRADE is an approach for evaluating recommendations and evidence based on methodologically rigorous and transparent criteria. There are four main components of the GRADE approach: assessing evidence quality, evaluating the benefit-harm balance, considering preferences and values, and analyzing recommendations in detail and making explicit judgments regarding their strength. Through systematic consideration of these factors, the GRADE approach ensures that recommendations are based on the best available evidence and are transparent [38]. GRADEpro GDT (McMaster University) was used to standardize assessments and generate summary tables (e.g., Table 3).

Table 3.

GRADE profile of carotenoid supplementation on liver enzymes in adults

Certainty assessment							№ of patients		Effect	Certainty	Importance
№ of studies	Study design	Risk of bias	Inconsistency	Indirectness	Imprecision	Other considerations	[intervention]	[comparison]	Absolute (95% CI)
ALT levels
15	Randomized trials	not serious	very seriousa	not serious	seriousb	none	442	403	MD 2.25 IU/L lower (4.84 lower to 0.34 higher)	⨁◯◯◯ Very lowa,b	NOT IMPORTANT
AST levels
14	Randomized trials	not serious	not serious	not serious	seriousc	none	386	367	MD 0.46 IU/L lower (1.25 lower to 0.34 higher)	⨁⨁⨁◯ Moderatec	NOT IMPORTANT
ALP levels
9	Randomized trials	not serious	not seriousd	not serious	seriouse	none	233	212	MD 0.34 IU/L lower (6.89 lower to 6.21 higher)	⨁⨁⨁◯ Moderated,e	NOT IMPORTANT
GGT levels
7	Randomized trials	not serious	not seriousf	not serious	seriousg	none	203	198	MD 0.43 IU/L lower (3.06 lower to 2.21 higher)	⨁⨁⨁◯ Moderatef,g	NOT IMPORTANT

Abbreviations: ALT: Alanine aminotransferase; AST: Aspartate aminotransferase; ALP: Alkaline phosphatase; GGT: Gamma-glutamyl transferase; CI: confidence interval; MD: mean difference

Explanations

a. Serious inconsistency since I2 = 97.18%

b. Serious imprecision since combined results from the random-effects model showed a non-significant reduction in ALT levels following carotenoid supplementation (p = 0.089). Downgraded

c. Serious imprecision since combined results from the random-effects model showed a non-significant reduction in AST levels following carotenoid supplementation (p = 0.259). Downgraded

d. Serious inconsistency since I2 = 57.06%. However, the value of was I2 < 50% in the subgroup of trials with Age (years) ≥ 50, and the significance, direction, and magnitude of the effect remained unchanged (WMD: −2.27, 95% Cl: −9.04, 4.48; n = 4, I2 = 0.0%). Not downgraded

e. Serious imprecision since combined results from the random-effects model showed a non-significant reduction in ALP levels following carotenoid supplementation (p = 0.918). Downgraded

f. Serious inconsistency since I2 = 72.23%. However, the value of was I2 < 50% in the subgroup of trials with Dose (mg/d) ≥ 12, and the significance, direction, and magnitude of the effect remained unchanged (WMD: −0.95, 95% Cl: −5.22, 3.31; n = 4, I2 = 0.0%). Not downgraded

g. Serious imprecision since combined results from the random-effects model showed a non-significant reduction in GGT levels following carotenoid supplementation (p = 0.75). Downgraded

Quantitative data synthesis and statistical analysis

An evaluation of carotenoid effects on liver enzymes was conducted. Weighted mean differences (WMDs) and 95% confidence intervals (CIs) were used to calculate effect sizes. For carotenoid and control groups, we calculated mean differences (MD) and standard deviations (SD) of ALT, AST, ALP and GGT levels before and after intervention: Trial end value—trial baseline value. In addition, MD was calculated as follows: (final outcome value in the carotenoid population compared to the baseline in this group)—(final outcome value for the comparison population, compared to the baseline in this group). In the absence of SDs, the SD was calculated as follows: SD = square root [(SD pre-treatment)² + (SD post-treatment)²—(2 R × SD pre-treatment × SD post-treatment)], and a correlation coefficient (R) = 0.8 [39]. When studies report standard error of the mean (SEM) instead of SD, this formula converts SEM to SD: SDs = SEs × square root (n), where n relates to the number of participants in the groups. We calculated ranges, medians, and CIs based on the method used by Hozo et al. [40]. For the calculation of the overall effect size, random effects were used. To measure heterogeneity, Cochran's Q-test and I² test were conducted. (p < 0.1 and I² value ≥ 50% indicate significant heterogeneity, respectively) [41]. A predefined subgroup analysis was performed to identify probable heterogeneity sources. Participants' body mass index (BMI) and health status, treatment dose and duration were included in this analysis. Meta-regression was performed to determine whether moderating variables, like dose and duration, influence effect size. A non-linear dose–response analysis on the basis of Crippa et al.'s method was conducted to evaluate liver enzyme levels in response to carotenoid dosages. [42]. A leave-in-one-out sensitivity analysis was conducted to determine if each study had a significant effect on the overall effect size. We also used Begg's rank correlations and Egger's weighted regression tests to detect publication bias. We used the trim and fill method and the fail-safe N method to adjust for publication bias [43]. In this meta-analysis, Comprehensive Meta-Analysis (CMA) V3 software was used [44] except for the dose–response analysis, which was performed with Stata, version 17 (Stata Corp, Texas, USA). To be statistically significant, a probability value (p-value) below 0.05 was required.

Results

Flow and characteristics of included studies

According to Fig. 1, there were 1215 initial identifications. Upon removing duplicate articles (n = 153), 1062 articles remained. 1040 articles were excluded after reviewing their titles and abstracts because they weren't RCTs or directly related to our inclusion criteria. As a result, we selected 22 articles that may be relevant for further full-text evaluation. We excluded 8 more articles for the following reasons: not reported SD (n = 3) [45–47], dietary intervention instead of supplements (n = 2) [48, 49], participants younger than 18 years of age (n = 2) [50, 51] and lack of baseline details and randomization (n = 1) [52]. Additionally, we found 1 study by performing a manual search of the literature [33]. Following the inclusion criteria, 15 qualified RCTs with 20 treatment arms were selected for analysis [32–35, 53–63]. Table 1 presents the PICOS criteria for studies included in this review.

Fig. 1.

Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) 2020 flow diagram of database searches, registers, and other sources

Characteristics of included studies

A total of 757 participants were randomly assigned to either the carotenoid treatment group (n = 422) or the placebo group (n = 335) in 15 qualified studies, comprising 20 treatment arms (Table 2) [32–35, 53–63]. Trials were conducted with a range of participants from 20 [34, 54] to 100 [55]. The included studies were published between 2003 and 2023 and were conducted in Japan (n = 6 studies) [32, 54, 58, 59, 61, 62], Iran (n = 5) [33, 35, 55, 56, 63], United States (n = 2) [53, 57], United Kingdom [34] and Australia (n = 1 each in the latter 2 countries) [60]. Among the participants, the mean age ranged from 31.3 [35] to 62.6 [58] years. One trial involved only men [63], one exclusively conducted with women [59] and the remaining trials included both genders.

Table 2.

Demographic characteristics of the included studies

Author (Year)	Design	Country	Participant’s status	Sample size (intervention/control)	Age (years) (intervention/control)	BMI (intervention/control)	Gender	Duration (weeks)	Intervention/control (type and dosage)	Registered	Funding
Spiller et al. (2003)	Rn/Db/Pa	United States	Healthy	19/16	58/55	NR	Both	8	6 mg Astaxanthin/500 mg high-oleic safflower oil	Yes	Investigator-Initiated
Saito et al. (2011)	Rn/Db/Pa	Japan	Healthy	10/10	38.2/38.8	NR	Both	4	12 mg Astaxanthin/P	Yes	Investigator-Initiated
Nakagawa et al. (2011)	Rn/Db/Pa	Japan	Healthy	10/10 10/10	56.3/56.6 56.1/56.6	27.4/27.7 27.6/27.7	Both	12	6 mg Astaxanthin/maize oil 12 mg Astaxanthin/maize oil	Yes	Investigator-Initiated
Mohamadpour et al. (2013)	Rn/Db/Pa	Iran	Healthy	22/20	31.3	24.9	Both	4	20 mg crocin/P	Yes	Investigator-Initiated
Mousavi et al. (2015)	Rn/Db/Pa	Iran	Schizophrenia	20/21	48.1/48.1	NR	Male	12	30 mg crocin/P	Yes	Investigator-Initiated
Coombes et al. (2016)	Rn/Db/Pa	Australia	Undergone renal transplantation	33/28	49.1/50.9	26.2/27.7	Both	52	12 mg Astaxanthin/P	Yes	Industry
Mikami et al. (2017)	Rn/Sb/Pa	Japan	Normal-weight and obese	19/20 20/20	57.5/53 55.2/53	26.6/25.2 27.2/25.2	Both	8	1 mg Fucoxanthin-enriched akamoku-oil/akamoku oil 2 mg Fucoxanthin-enriched akamoku-oil/akamoku oil	Yes	Investigator-Initiated
Chen et al. (2017)	Rn/Db/Pa	Japan	Healthy	14/15	51/52	21.3/21.9	Female	13	12 mg Astaxanthin/P	Yes	Industry
Sepahi et al. (2018)	Rn/Db/Pa	Iran	Diabetic Maculopathy	20/20 20/20	54.31/57.17 56.09/57.17	NR	Both	13	5 mg crocin/P 15 mg crocin/P	Yes	Investigator-Initiated
Kakutani et al. (2018)	Rn/Db/Pa	Japan	Healthy	41/39	48.9/50.8	27.19/27.09	Both	12	9 mg Xanthophyll capsules (Paprika-oil)/P vegetable oil capsules	Yes	NR
Chernyshova et al. (2019)	Rn/Db/Co	United Kingdom	Healthy	10/10	33.4/33.4	23.2/23.2	Both	4	7 mg Lycopene/P	Yes	Investigator-Initiated
Chan et al. (2019)	Rn/Db/Pa	Japan	Type 2 diabetes	18/18 18/18	58.7/57.7 62.6/57.7	26.3/25.9 25.2/25.9	Both	8	6 mg Astaxanthin/Cellulose starch 12 mg Astaxanthin/Cellulose starch	Yes	Investigator-Initiated
Haidari et al. (2020)	Rn/Db/Pa	Iran	NAFLD patients with grade 2–3 hepatic steatosis	23/23 23/23	38.1/35.6 38.4/36.4	33.2/33.1 32.8/33.9	Both	12	6 mg β-Cryptoxanthin and hypocaloric normal protein diet/P and hypocaloric normal protein diet 6 mg β-Cryptoxanthin and hypocaloric high protein diet/P and hypocaloric high protein diet	Yes	Investigator-Initiated
Sepahi et al. (2022)	Rn/Tb/Pa	Iran	Uncontrolled Type 2 diabetes	50/50	57.58/56.92	NR	Both	13	30 mg crocin/P	Yes	Investigator-Initiated
Ciaraldi et al. (2023)	Rn/Db/Pa	United States	Prediabetes and Dyslipidemia	22/12	51/53	31.81/31.89	Both	24	12 mg Astaxanthin/P	Yes	Industry

Abbreviations: Rn Randomized, Db Double-blinded, Pa Parallel, NR Not reported, P Placebo, Co Crossover, Sb Single-blinded, P Placebo, NAFLD Non-alcoholic fatty liver disease, Tb Triple-blinded

The supplementation duration ranged from 4 [34, 35, 54] to 52 [60] weeks. It was a crossover design in one study [34]; however, the rest (n = 14) were conducted in parallel. Among the 15 studies, one was a single-blinded RCT [62], one was a triple-blinded RCT [55], and the others were double-blinded RCTs. In 7 studies, participants were healthy [32, 34, 35, 54, 57, 59, 61]; two studies involved patients with type 2 diabetes [55, 58]; in one study, participants had undergone renal transplantation [60]; one study was conducted on NAFLD patients [33]; in one study normal or obese participants were involved [62]; another study was conducted on schizophrenia patients [63]; a study involved patients with prediabetes and dyslipidemia [53]; and in one study participants had diabetic maculopathy [56].

Data quality

For the quality assessment of studies, Fig. 2 summarizes the findings regarding the risk of bias in studies. Only one study was classified as “some concerns” quality [56], with the remaining studies classified as high quality. Within the framework of the GRADE assessment (Table 3), the evidence regarding the impact of carotenoid supplementation on liver enzymes was classified with varying degrees of certainty. ALT was evaluated as having very low certainty, indicating substantial uncertainty in the effect estimates due to notable inconsistency and imprecision in the available data. AST was assigned a moderate grade, reflecting a reasonable level of reliability, albeit with some uncertainty attributed to imprecision. Similarly, ALP and GGT were also graded as moderate, signifying that while the evidence is generally consistent, minor concerns remain regarding precision and inconsistency.

Fig. 2.

Risk of bias tool: traffic light chart and summary table by domain

Meta-analysis

Findings from the meta-analysis of carotenoid supplementation and ALT levels

ALT levels were measured in 15 studies with 20 treatment arms, including a total of 825 participants (Carotenoid: 422, control: 403). ALT was reduced non-significantly after carotenoid treatment (WMD: −2.25 IU/L, 95% CI: −4.84, 0.34, P = 0.089). Moreover, there was significant heterogeneity among the studies (I² = 97.18%, P < 0.001) (Fig. 3 panel A). As a result of subgroup analysis, the health status of participants could be the source of heterogeneity (P-_between = 0.010). Significant effects were observed in duration < 12 weeks (P = 0.028), BMI ≥ 25 (P = 0.045) and in the non-healthy participants (P = 0.015) (Table 4). According to the meta-regression, both carotenoid dosage and intervention duration had no significant effect on ALT levels (Table s2) (Fig. s2 panel A, E). No non-linear dose–response relationship between carotenoid dosage and ALT levels was found (P_{dose-response} = 0.205, P_{non-linearity} = 0.238) (Fig. 4 Panel A).

Fig. 3.

A-D. Forest plots for the effect of carotenoid supplementation on: ALT (A), AST (B), ALP (C) and GGT (D). CI; confidence interval, ALT; Alanine aminotransferase, AST; Aspartate aminotransferase, ALP; Alkaline phosphatase, and GGT; Gamma-glutamyl transferase

Table 4.

The results of subgroup analysis of the included randomized controlled trials in the meta-analysis of carotenoid supplementation on liver enzymes in adults

Variable		Number of effect sizes	(I2)1	P value	Pheterogenity2	WMD (95%CI)3	P-between4
Alanine aminotransferase (IU/L)
Overall		20	97.18	0.089	< 0.001	−2.251 (−4.84, 0.34)
Dose (mg/d)							0.599
	< 12	10	96.79	0.120	< 0.001	−3.037 (−6.86, 0.79)
	≥ 12	10	97.33	0.502	< 0.001	−1.486 (−5.82, 2.85)
Duration (weeks)							0.104
	< 12	8	92.66	0.028	< 0.001	−3.395 (−6.42, −0.37)
	≥ 12	12	66.42	0.332	0.001	−0.652 (−1.98, 0.67)
BMI (Kg/m^2)							0.229
	< 25	3	39.94	0.749	0.189	−0.643 (−4.59, 3.3)
	≥ 25	11	98.23	0.045	< 0.001	−4.08 (−8.06, −0.1)
Age (years)							0.775
	< 50	8	77.26	0.083	< 0.001	−2.373 (−5.06, 0.312)
	≥ 50	12	97.16	0.322	< 0.001	−1.738 (−5.18, 1.7)
Health status							0.010
	H	10	0	0.205	0.517	0.346 (−0.19, 0.88)
	NH	10	97.57	0.015	< 0.001	−4.447 (−8.04, −0.85)
Aspartate aminotransferase (IU/L)
Overall		18	46.27	0.259	0.017	−0.457 (−1.25, 0.34)
Dose (mg/d)							0.860
	< 12	9	0	0.948	0.729	0.015 (−0.44, 0.47)
	≥ 12	9	64.61	0.869	0.004	−0.14 (−1.81, 1.53)
Duration (weeks)							0.124
	< 12	6	52.78	0.342	0.060	1.039 (−1.1, 3.18)
	≥ 12	12	44.99	0.072	0.045	−0.762 (−1,59, 0.07)
BMI (Kg/m^2)							0.890
	< 25	3	84.94	0.836	0.001	0.581 (−4.93, 6.09)
	≥ 25	9	0	0.431	0.851	0.191 (−0.283, 0.664)
Age (years)							0.229
	< 50	8	58.83	0.794	0.017	0.239 (−1.56, 2.04)
	≥ 50	10	0	0.003	0.680	−0.924 (−1.52, −0.32)
Health status							0.162
	H	10	50.79	0.866	0.032	0.117 (−1.24, 1.47)
	NH	8	0	0.003	0.565	−0.948 (−1.57, −0.32)
Alkaline phosphatase (IU/L)
Overall		11	57.06	0.918	0.010	−0.342 (−6.89, 6.21)
Dose (mg/d)							0.823
	< 12	5	74.62	0.793	0.003	−1.264 (−10.7, 8.17)
	≥ 12	6	10.87	0.961	0.346	0.223 (−8.77, 9.21)
Duration (weeks)							0.648
	< 12	3	0	0.6	0.375	−2.005 (−9.51, 5.50)
	≥ 12	8	58.75	0.887	0.018	0.599 (−7.69, 8.88)
Age (years)							0.629
	< 50	7	64.9	0.901	0.009	0.608 (−8.94, 10.16)
	≥ 50	4	0	0.509	0.960	−2.279 (−9.04, 4.48)
Health status							0.474
	H	6	41.83	0.551	0.126	2.179 (−4.98, 9.34)
	NH	5	38.92	0.653	0.162	−2.212 (−11.85, 7.43)
Gamma-glutamyl transferase (IU/L)
Overall		11	72.23	0.751	< 0.001	−0.426 (−3.06, 2.21)
Dose (mg/d)							0.830
	< 12	7	82.57	0.824	< 0.001	−0.368 (−3.61, 2.87)
	≥ 12	4	0	0.661	0.939	−0.955 (−5.22, 3.31)
Duration (weeks)							0.072
	< 12	3	0	0.122	0.37	3.299 (−0.89, 7.48)
	≥ 12	8	78.89	0.346	< 0.001	−1.502 (−4.63, 1.62)
Age (years)							0.115
	< 50	6	84.6	0.332	< 0.001	−1.689 (−5.1, 1.73)
	≥ 50	5	0	0.211	0.517	2.35 (−1.33, 6.03)
Health status							0.006
	H	8	62.91	0.276	0.009	1.385 (−1.11, 3.88)
	NH	3	24.16	0.011	0.268	−5.655 (−10.04, −1.27)

Abbreviations: H Healthy and NH Non-Healthy, WMD Weighted mean difference

¹Inconsistency, percentage of variation across studies due to heterogeneity

²Obtained from the Q-test

³Obtained from the random-effect model

⁴Heterogeneity between groups

Fig. 4.

Nonlinear dose–response effects of carotenoid dose (mg/d) on ALT (A), AST (B), ALP (C) and GGT (D) in adults. The 95%CI is demonstrated in the dashed lines. Abbreviations: CI, confidence interval; ALT; Alanine aminotransferase, AST; Aspartate aminotransferase, ALP; Alkaline phosphatase, and GGT; Gamma-glutamyl transferase

Findings from the meta-analysis of carotenoid supplementation and AST levels

There were 14 studies involving 18 treatment arms and 753 participants (Carotenoid: 386, control: 367) that measured AST levels. AST levels were not significantly affected by carotenoid supplementation (WMD: −0.46 IU/L, 95% CI: −1.25, 0.34, P = 0.259) with a significant level of heterogeneity within the studies (I² = 46.27%, P = 0.017) (Fig. 3 panel B). As shown in Table 4, studies conducted on non-healthy participants (P = 0.003) with ages ≥ 50 (P = 0.003) resulted in significant AST level reductions. Analysis of meta-regression in a random effect model revealed that (Table s2), AST levels were not dependent on the dose and duration of carotenoid supplementation (Fig. s2 panel B, F). Non-linear dose–response analysis between AST and dose of carotenoid wasn’t significant (P_{dose-response} = 0.986, P_{non-linearity} = 0.926) (Fig. 4 Panel B).

Findings from the meta-analysis of carotenoid supplementation and ALP levels

A total of 445 participants (Carotenoid = 233, control = 212) were included from 9 studies involving 11 treatment arms that measured levels of ALP. Carotenoid supplementation did not result in a significant reduction in ALP (WMD: −0.34 IU/L, 95% CI: −6.89, 6.21, p = 0.918) while there was significant heterogeneity among the studies (I² = 57.06%, p = 0.01) (Fig. 3 panel C). Subgroup analysis revealed greater effects in doses < 12 mg/d, duration < 12 weeks, ages ≥ 50 years and in the non-healthy participants. However, these effects weren’t significant (Table 4). Neither the dose nor the duration were significantly associated with ALP levels according to meta-regression (Table s2) (Fig. s2 panel C, G). No non-linear dose–response association was observed between doses of carotenoid and ALP levels (P_{dose-response} = 0.998, P_{non-linearity} = 0.990) (Fig. 4 panel C).

Findings from the meta-analysis of carotenoid supplementation and GGT levels

GGT levels were measured in 7 studies including 401 participants (Carotenoid = 203, control = 198). GGT levels were reduced following carotenoid supplementation (WMD: −0.43 IU/L, 95% CI: −3.06, 2.21, p = 0.751) non-significantly and there was significant heterogeneity among the studies (I² = 72.23%, p < 0.001) (Fig. 3 panel D). The health situation of participants could be the source of heterogeneity (P_-between = 0.006) (Table 4). Also, significant reductions were observed in the non-healthy participants (P = 0.011) (Table 4). Meta-regression for both duration and dose wasn’t significant (Table s2) (Fig. s2 panels D, H). No non-linear dose–response association was observed between GGT levels and dose of carotenoid supplementation (P_{dose-response} = 0.952, P_{non-linearity} = 0.862) (Fig. 4 panel D).

Sensitivity analysis

The overall effects of carotenoid supplementation on all ALT, AST, ALP and GGT variables did not depend on any single study according to the sensitivity analysis, (Fig. s1 panel A-D).

Publication bias

Following the "trim and fill" method, 0, 1, 1 and 0 probable missing studies were imputed for ALT, AST, ALP and GGT, respectively (Fig. s3 panel A-D). Begg's rank correlation and Egger's linear regression tests were not significant for all outcomes except ALP which Egger’s test was significant (P = 0.047) (Table s3).

Discussion

Summary of main results

The meta-analysis assessed the impact of carotenoid supplementation on liver enzyme levels, incorporating 15 studies for ALT, 14 for AST, 9 for ALP, and 7 for GGT. Despite observing reductions in these enzyme levels, the changes were statistically non-significant overall. Subgroup analyses revealed significant reductions in non-healthy participants, individuals with BMI ≥ 25 kg/m², and shorter intervention durations (< 12 weeks). These results suggest that the efficacy of carotenoids may depend on participant health status, baseline characteristics, and intervention specifics. Heterogeneity was notable across studies, and meta-regressions confirmed that neither dosage nor duration consistently affected enzyme levels.

Comparison with other studies

The results of this meta-analysis are broadly in agreement with previous findings. studies on carotenoids, such as lycopene and astaxanthin. For instance, Zamani et al. highlighted that while lycopene consumption significantly reduced malondialdehyde (MDA), it did not consistently improve inflammatory markers such as TNF-α, CRP or IL-6 [64]​. Similarly, Arefpour et al., found that astaxanthin supplementation was associated with variable effects on liver enzymes, showing a potential increase in ALT in certain contexts [65]​. These variations may be due to differences in health status, baseline values, and carotenoid formulations used in the studies.

According to previous studies, such as those by Wu et al. [66], observed limited effects of carotenoids on liver enzymes, suggest that their impact might be greater in individuals with pre-existing metabolic conditions like non-alcoholic fatty liver disease (NAFLD). Similarly, Lim et al. (2018) noted that short-term interventions might show a wide range of results due to differences in absorption rates and metabolic responses [67]. As a result of these data, the current meta-analysis supports the study's subgroup results, where significant reductions were noted in non-healthy participants and interventions lasting 12 weeks. Age-related responsiveness was also observed by Tan et al. (2009), supporting our findings in participants aged ≥ 50 [68]. Furthermore, a study conducted by Ben-Dor et al. reported modest but non-significant reductions in AST, which is consistent with this analysis and suggests that antioxidant benefits may differ from individual to individual [69]. Despite the non-significant effect of carotenoid supplementation on ALP, the results were consistent with those of Lietz et al. (68), who proposed that carotenoid benefits may be more evident when combined with other micronutrients [70]. This synergistic interaction might explain the lack of strong effects when carotenoids are used in isolation. Furthermore, a previous study by Wu et al. emphasized that baseline liver function significantly influences responses to supplementation, echoing the heterogeneity seen in the current analysis [71].

In this meta-analysis, carotenoids significantly reduced GGT levels in non-healthy individuals, in line with research by Lee et al., which indicated that carotenoids could reduce the markers of oxidative stress in populations with compromised health [72]. It is important to note, however, that heterogeneity in outcomes has also been observed in a study conducted by Furubayashi et al., who speculated that differences in carotenoid formulations (natural vs. synthetic) could be responsible for the inconsistent findings [73].

The results of this meta-analysis reflect broader inconsistencies in the literature, possibly due to differences in study design, such as dosages, supplementation duration, and participant demographics. In spite of some studies reporting promising results, such as Elvira-Torales et al., for specific carotenoid types like lycopene, the overall evidence suggests that these benefits would be more likely to be achieved by targeting specific subgroups rather than general populations [47, 74].

Mechanisms

General antioxidant mechanisms of carotenoids

Carotenoids, like lycopene and astaxanthin, exhibit potent antioxidant and anti-inflammatory properties that contribute to their potential effects on liver enzyme levels [64]. In humans, these compounds work by neutralizing reactive oxygen species (ROS) and reducing oxidative stress, a key factor in liver injury and dysfunction [75]. In recent studies, lycopene supplementation significantly lowers malondialdehyde (MDA) levels, a marker of lipid peroxidation, indicating improved oxidative defense mechanisms ​[64, 76]. The antioxidant properties of carotenoids improve the endogenous antioxidant enzymes activity like superoxide dismutase (SOD) and glutathione peroxidase, which play crucial roles in the detoxification of harmful oxidative molecules [77].

Population-specific effects in overweight/non-healthy individuals

The meta-analysis revealed significant reductions in ALT levels among participants with BMI ≥ 25 kg/m² (P = 0.045) and non-healthy individuals (P = 0.015). This aligns mechanistically with the heightened oxidative stress and chronic inflammation characteristic of these populations [78]. For instance, obesity and metabolic dysfunction are associated with elevated ROS production and impaired antioxidant defenses [30], creating a physiological context where carotenoids’ ROS scavenging capacity becomes clinically impactful. Similarly, non-healthy individuals (e.g., NAFLD or diabetic patients) exhibit upregulated NF-κB-driven inflammatory pathways, which carotenoids suppress via inhibition of redox-sensitive transcription factors [79]. These mechanisms collectively explain why enzyme reductions were more pronounced in subgroups with elevated baseline oxidative stress.

It is important to note that while elevated liver enzymes are nonspecific biomarkers, their reduction in high-risk populations (e.g., NAFLD patients) holds clinical relevance. For instance, a 5 IU/l decrease in ALT has been associated with a 10% reduction in liver-related mortality in observational studies [80]. Thus, the significant ALT reduction observed in NAFLD subgroups may translate to meaningful health benefits, particularly as adjunct therapy alongside lifestyle modifications [81].

Molecular signaling pathways

The subgroup-specific benefits are further supported by carotenoid modulation of metabolic and inflammatory pathways. For example, astaxanthin upregulates FGF21, a hepatoprotective hormone that improves mitochondrial function and lipid metabolism in obese individuals [68, 82]. This pathway likely contributed to the significant ALT reduction in participants with BMI ≥ 25 kg/m² (WMD: −4.08 IU/l). Similarly, lycopene’s inhibition of NF-κB and activation of Nrf2 in hepatic cells correlates with the marked AST reduction in non-healthy subgroups (WMD: −0.948 IU/l, P = 0.003), as these pathways regulate both inflammation and hepatocyte apoptosis [83].

Strengths and weaknesses of this study

The study has several strengths, including its comprehensive scope, which encompasses a wide range of populations and intervention parameters, providing an unbiased assessment of the impact of carotenoid supplementation on liver enzymes. By using subgroup analyses, the study was further enhanced by identifying populations, such as non-healthy individuals and those with BMI ≥ 25 kg/m², who might derive the most significant benefits, offering critical insights for targeted interventions. Furthermore, the robustness of the findings is supported by sensitivity analyses, which showed that our results were not unduly influenced by any specific study, and the minimal publication bias observed enhances the reliability of the conclusions. In spite of this, there are several notable limitations to be considered. the heterogeneity in carotenoid types, dosages, and participant characteristics across studies. While RCTs are designed to infer causality, the high degree of heterogeneity across studies, which is caused by differences in participant characteristics, carotenoid formulations, and intervention duration, limits definitive causal conclusions and makes the findings less generalizable. The inclusion of diverse populations (healthy individuals, diabetics, NAFLD patients) introduced heterogeneity, complicating interpretation for specific liver conditions. While this diversity enhances generalizability, future meta-analyses should prioritize cohorts with uniform liver diagnoses to isolate carotenoid effects on hepatic health. It is also difficult to draw direct comparisons between studies due to the inconsistent reporting of liver enzyme measurements and carotenoid dosages, and the absence of mechanistic insights beyond liver enzyme measurements limits the understanding of the pathways involved. Lastly, most outcomes were moderately certain according to the GRADE. Considering these limitations, future research should be directed toward improving these areas.

Conclusion

While carotenoids supplementation appears to have potential benefits under specific conditions, such as in < 12 weeks, BMI ≥ 25, and among non-healthy participants, this meta-analysis highlights a lack of significant overall effects on liver enzyme levels. Clinically, integrating carotenoid-rich diets (e.g., tomatoes, paprika, seafood) or standardized supplements into lifestyle interventions may complement existing strategies for managing metabolic liver diseases. However, future research should focus on long-term interventions, standardized dosages, high-quality, well-powered RCTs with standardized carotenoid interventions, clearly defined populations and mechanistic studies to better understand the therapeutic potential of carotenoids on liver enzyme levels.

Authors’ contributions

The authors' responsibilities were as follows: MB conceived the study. SSH and MA carried out the literature search. SSH and MVB carried out data extraction and independent reviewing. MA and EB conducted the quality assessment of the included studies. MB conducted data analysis and interpretation. SSH, MB and MVB wrote the manuscript. All authors performed critical revisions and approved the final manuscript.

Funding

The research protocol was approved and supported by Shiraz University of Medical Sciences (grant number: 32683).

Data availability

Data described in the manuscript will be made available upon reasonable request pending application and approval by contacting the Corresponding author.

Declarations

Ethics approval and consent to participate

All of the included studies provided a statement on ethics approval and consent with a registration number.Considering that in this study we did not conduct experiments directly on humans and only extracted the information available in other studies, therefore having a code of ethics was not necessary for this study. Also, according to the type of study, we registered our protocol in Prospero with this code (no. CRD42024612956), which is accessible.

Consent for publication

Not applicable.

Competing interests

The authors have no relevant interests to declare.