Four databases, Pubmed, Embase, Web of Science, and Cochrane Library (Central Database), were searched for relevant literature from 2000–01 to 2023–03. Because according to the “results by year” function of Pubmed, the skin health of orally consumed fruits or fruit extracts has been extensively studied only in the last 23 years. A computerized search of the relevant literature was performed for the following terms: (1) skin aging or skin condition or skin hydration; (2) fruit or fruit extracts. These search terms were adapted accordingly for each database. Supplementary Table S1 provides a comprehensive description of the search method. In this paper, only English RCTs were selected.

Studies were selected when they met all of the following inclusion criteria:

Three reviewers (H.L., L.W.and J.W.) independently screened titles and abstracts. Full texts were imported into Endnote (Version X9, Clarivate Analytics) for further review when they were deemed potentially eligible by the initial screening process. The full texts were then independently evaluated against the aforementioned inclusion criteria by reviewers (H.L. and L.W. or J.W.) to determine their eligibility for inclusion in the study. Throughout the process, whenever any question arose, we consulted with other researchers (J.F. and L.J.) or referenced relevant literature and books for clarification.

The characteristics of each included study are documented in Supplementary Table S2, including details of the study such as author(s), country and year of publication, characteristics of the study population (age, sex and health status), study duration, intervention content of each group, daily dose, the form of applications, testing environment [°C room temperature (R.T.) and % relative humidity (R.H.)], the start and end time of the study, the measurement instruments used, and the skin sites assessed. The safety of the intervention was assessed by comparing adverse events (AEs), treatment-related adverse effects (TEAs), and treatment-related discontinuations (TAW) in the intervention and placebo groups.

The other collected data, such as the baseline and endpoint measure (average) and its variability SD (standard deviation) or SE (standard error), were input into an Excel spreadsheet. Five outcomes were evaluated: skin hydration, minimum erythema dose (MED), overall skin elasticity (R2), transcutaneous water loss (TEWL), and wrinkle depth. When multiple measurement sites were provided in the study, facial and arm test areas commonly used by females were preferred as the measurement sites for assessing moisturizing effects. When data from different racial populations were provided, subgroup analysis would be performed, and studies from different races would be considered as independent studies. The study data with the longest observation period, particularly those with multiple time points of measurement results, were selected for analysis.

During the data collection process, whenever any issues arose, we consulted with other researchers to discuss and resolve them.

Three reviewers (H.L. and L.W.or J.F.) evaluated the risk of bias (ROB) in the randomized controlled trials (RCTs) using the Cochrane Collaboration’s assessment tool (18), which includes seven criteria: selection biases (SB), performance biases (PB), detection biases (DB), attrition biases (AB), reporting biases (RB), and other biases in the trial design or methodology. Examples of other biases include changes in life habits during the study, unstable conditions in test rooms that may affect the measurement results, and potential conflicts of interest. Each study was assessed for each of these criteria and rated as “low risk,” “high risk,” or “unclear.” ROB was considered low when the study adequately reported methods without potential biases according to the Cochrane Collaboration’s guidelines (18). Conversely, high ROB was assigned when the method used in the study could not remove underlying biases. If there was insufficient information to make a determination, or if the information provided was inadequate, the trial was labeled with an“unclear risk of bias.”

A total of 4,794 references (Figure 1) were retrieved, including 795 from PubMed, 962 from Embase, 2,251 from Web of Science, and 786 from Cochrane Library. After screening, 7 randomized controlled trials were included in the Systematic review and Meta-analysis.

Considering that Nobile 2022 involved participants from both China and Italy, studies conducted with participation from different ethnic groups are considered as separate studies.

The review of every ROB item presented as percentages across the entire selected research as per the judgment of the researcher was displayed in Figure 2, and the judgment for every trial was described in Supplementary Figure S1. All studies have fully reported randomization, including the use of computer-generated randomization schedules and random number tables. Therefore, the selection bias of all studies is considered to be at low risk. Five studies (63%) reported a clear process of blinding for both participants and outcome assessors, while 36% of the studies had an unclear risk of bias. Two studies (25%) were identified as having low risk of detection bias, while the remaining studies had an unclear risk of bias. Attrition bias was low risk in 7 research works (88%), and 1 research (12%) was unclear. The risk of reporting bias was high in one study (12%), while concrete data were provided and trials were pre-registered in a publicly available trial registry in seven studies (88%). Four studies (50%) were identified as having a high risk of “other biases,” with commonly observed factors including potential conflicts of interest, maintenance of participants’ living habits, and insufficient details regarding physical environmental conditions.

Supplementary Table S2 summarizes the main characteristics of each study. Seven studies were included in a Meta-analysis to evaluate the effect of oral intake of fruits or fruit extracts on improving skin aging. Among the seven trials included in the Meta-analysis, two involved Asian participants (Japanese and Korean). One study consisted of participants from China and Italy, and three trials were conducted among U.S. participants. The primary participants were women between the ages of 30 and 70 with visible signs of skin aging, such as wrinkles and dryness, at baseline levels.

Skin hydration levels were assessed in five studies (n = 341), including a study conducted in both China and Italy, which were considered as two independent studies. Three studies focused on Asian populations (24–26), and three studies included non-Asian populations (27, 28). Four studies examined the effects of different fruit extract formulations, including capsules and tablets (24–27), while one study involved direct ingestion of avocado (28). These studies lasted 8–12 weeks and evaluated skin hydration levels.

Figure 3 displays the forest plot of a Meta-analysis of six studies, which estimates the difference in skin hydration between intervention and placebo groups.

The results of the Meta-analysis of six studies showed no significant improvement in skin hydration in the Asian group after ingestion of fruit or fruit extract (SMD = 1.51; 95% CI: -0.2-3.22, p = 0.08); and no significant difference in the non-Asian group compared to the placebo group after ingestion of fruit or fruit extract (SMD = 0.83; 95% CI: −0.36- 2.01, p = 0.41). However, the ability of ingesting fruits or fruit extracts to promote skin hydration was supported by statistical evidence based on the overall effect size summary analysis (SMD = 1.16; 95% CI: 0.22–2.10, p = 0.02).Overall, both the Asian group (I² = 96%) and the non-Asian group () exhibited a high degree of heterogeneity. Sensitivity analysis using the one-by-one exclusion method found that the heterogeneity significantly decreased(I² = 23%，I² = 36%) when the study by Nobile 2022 was excluded, with no change in the pooled effect result for skin hydration (SMD = 0.35; 95%CI: 0.07–0.62,p = 0.01) (Supplemental Figure S2); the heterogeneity mainly originated from Nobile 2022, which may be due to the skin measurements differed in location, the study measured skin moisture on the back, which is less exposed to the external environment, while other studies measured skin hydration on the face and forearm, which are more exposed.

Skin elasticity was monitored in three studies (n = 225) lasting from 8 to 12 weeks. One study was carried out in both China and Italy and was considered as two independent studies. Two studies were included in the Asian population group (24, 25), whereas the other two studies were included in the non-Asian population group (24, 27). The four studies focused on different fruit extract formulations (including capsules and tablets).

Figure 4 displays the forest plot of a Meta-analysis of four studies, which estimates the difference in skin elasticity between intervention and placebo groups.

The Meta-analysis of four studies (n = 225) revealed no significant effect of consuming fruits and fruit extracts on increasing skin elasticity compared to placebo (SMD = 1.43; 95% CI: −0.07-2.93, p = 0.06) (Figure 4). No significant correlation was observed in both the Asian and non-Asian groups.In the Meta-analysis of these two subgroups, the overall study exhibited high heterogeneity (I² of 92 and 98%, respectively), which may be attributed to different sources of fruit extracts, including red-orange complex extract, Citrus sinensis extracts (PTE), and pomegranate extracts, different extraction methods such as standardized extracts and 50% ethanol extracts, as well as the differences in body parts of the tested subjects (back, eye contour, and arm).

Four studies (n = 280) evaluated wrinkle depth over a period of 8 to 12 weeks. One of the trials included both Chinese and Italian participants, and studies with different ethnic groups were considered as independent studies. Therefore, three studies were included in the Asian group (24–26), while two studies were included in the non-Asian group (24, 29). Among them, the focus of four studies (24–26) was on different formulations of fruit extracts, including capsules and tablets, while one study (29) directly consumed mangoes.

Figure 5 displays the forest plot of a Meta-analysis of four studies, which estimates the difference in wrinkle depth between intervention and placebo groups.

Among the five studies, a total of 280 participants were included, and no significant improvement in wrinkle depth was observed after consuming fruit or fruit extracts (SMD = -0.73; 95% CI: −1.65-0.20, p = 0.12) (Figure 5). No significant correlation was also observed for subgroup analysis: Asian group (SMD = −1.04; 95% CI: −2.13-0.05, p = 0.06) or non-Asian group (SMD = −0.21; 95% CI: −2.61-2.20, p = 0.87). However, both analyses showed significant heterogeneity between studies (I² = 92 and 95%). Sensitivity analysis using the one-by-one exclusion method found that the results of the study changed when Fam 2020 was removed, and a significant improvement in skin wrinkle depth was observed after ingesting fruit extracts (SMD = -1.13; 95% CI: −1.97--0.28, p = 0.87) (I² = 89%) (Supplemental Figure S3). This may be because the Fam 2020 study used a low-dose placebo group and a high-dose experimental group, while the rest of the studies used a placebo control group. Nevertheless, high heterogeneity still existed, possibly due to the complex composition of the supplement products, different extraction methods for the extracts, and differences in dosages between studies.

MED was measured in two studies (n = 89), one with fruit extract intake (27) and the other with fruit intake (28). Both studies included participants from non-Asian groups and reported changes in MED after 8–12 weeks of fruit (27) or fruit extract (28) consumption.

Figure 6 displays the forest plot of a Meta-analysis of four studies, which estimates the difference in MED between intervention and placebo groups.

Among 89 participants in two studies, oral consumption of fruit or fruit extracts showed no statistical difference in MED compared to placebo (SMD = 0.27; 95% CI: −0.28-0.83; p = 0.34) (Figure 6). Meta-analysis showed moderate heterogeneity (I2 = 42%), largely due to a limited number of randomized controlled trials and the fact that the supplements consumed included both fruit extracts and whole fruits.

TEWL was measured in two studies (n = 186), one of which was conducted simultaneously in China and Italy and was considered as two independent studies. A total of 2 studies were included in the Asian group (24, 25) and 1 study was included in the non-Asian group (24). All studies investigated the effects of fruit extracts for a duration of 8 to 12 weeks.

Figure 7 displays the forest plot of a Meta-analysis of three studies, which estimates the difference in Transepidermal water loss (TEWL) between intervention and placebo groups.

A Meta-analysis of these three studies showed a significant difference between the fruit extract group and the placebo group (SMD = −6.54; 95% CI: −12.15- -0.92; p = 0.02) (Figure 7), indicating that the intake of fruit extracts has the ability to reduce TEWL. There was no significant difference in the change in TEWL for the Asian group (SMD = −8.11; 95% CI: −24.37-8.15; p = 0.33); the non-Asian group showed a significant improvement in TEWL (SMD = −4.09; 95% CI: −5.14- -3.04; p < 0.00001).However, there was high heterogeneity between the two studies, possibly due to differences in the tested areas (back, eye area), different types of fruit extracts (qingken orange extract, red orange (Citrus sinensis (L.) Osbeck) extract), and variations in extract preparation methods (standardized extract, 50% ethanol extract).