Pharmacokinetic Outcomes of the Interactions of Antiretroviral Agents with Food and Supplements: A Systematic Review and Meta-Analysis

Because pharmacokinetic changes in antiretroviral drugs (ARV), due to their concurrent administration with food or nutritional products, have become a clinical challenge, it is necessary to monitor the therapeutic efficacy of ARV in people living with the human immunodeficiency virus (PLWH). A systematic review and meta-analysis were conducted to clarify the pharmacokinetic outcomes of the interaction between supplements such as food, dietary supplements, and nutrients, and ARV. Twenty-four articles in both healthy subjects and PLWH were included in the qualitative analysis, of which five studies were included in the meta-analysis. Food–drug coadministration significantly increased the time to reach maximum concentration (tmax) (p < 0.00001) of ARV including abacavir, amprenavir, darunavir, emtricitabine, lamivudine, zidovudine, ritonavir, and tenofovir alafenamide. In addition, the increased maximum plasma concentration (Cmax) of ARV, such as darunavir, under fed conditions was observed. Area under the curve and terminal half-life were not significantly affected. Evaluating the pharmacokinetic aspects, it is vital to clinically investigate ARV and particular supplement interaction in PLWH. Educating patients about any potential interactions would be one of the effective recommendations during this HIV epidemic.


Introduction
The human immunodeficiency virus (HIV) infection is one of the major health challenges around the world. Approximately 33 million people have died from HIV/acquired immunodeficiency syndrome (AIDS)-related illness since the start of the HIV epidemic worldwide [1]. Globally, over 37 million people were living with HIV (PLWH) in 2020, of which more than 27 million were receiving HIV treatment or antiretroviral (ARV) therapy [2]. The commonly adopted ARV therapy includes the combined use of three or more ARVs from at least two different classes, such as nucleoside reverse transcriptase inhibitor (NRTI), non-nucleoside reverse transcriptase inhibitor (NNRTI), and protease inhibitor (PI). In adults and adolescents, tenofovir/lamivudine or tenofovir/emtricitabine is the preferred backbone, which is used in combination with a third drug, dolutegravir for post-exposure prophylaxis [3]. Due to the widespread use of effective ARV therapy and HIV care, about 66% of the overall infected population achieve viral suppression [2].
Dietary supplements including vitamins and minerals are widely used to boost the body's defense system in many patients with nutrient deficiencies. European Food Safety

Quality Assessment
The risk of bias of the included studies was assessed using the different validated tools including the risk of bias assessment tools for primary and secondary medical studies [18], ROBINS-I for non-randomized studies [19], Cochrane Collaboration's tool for randomized trials [20], a modified Newcastle-Ottawa Quality Assessment Scale for cross-sectional studies [21], a modified tool for case report described by Murad et al. [22], and Cochrane risk of bias for crossover studies [23]. The quality assessment of included studies is mentioned in Supplementary Material Table S2.

Statistical analysis
Review Manager (RevMan version 5.4.1: The Nordic Cochrane Center, Copenhagen, Denmark) was used for conducting meta-analysis. Random effect models with the inverse variance method were used to calculate the weights of the studies. Publication bias was assessed by visually inspecting the funnel plot. The heterogeneity among studies was assessed by I 2 statistics. The I 2 of <25% was assumed as negligible heterogeneity, whereas that of >75% was high heterogeneity [20].

Results
A total of 3038 articles were retrieved. After duplicates and irrelevant studies were removed by reviewing titles and abstracts, 28 articles were selected for potential inclusion. The number of records with reasons for exclusion is summarized in Figure 1. Records that possibly met the inclusion criteria were assessed for eligibility for systematic review. Finally, twenty-four articles were included in the systematic review. Thirteen articles of them were good quality and eleven articles were fair (Supplementary Material Table S2). In general, the heterogeneity was high, and the 95% confidence intervals of the pooled data were close to the most weighted and included studies whose intervals were narrow. The summary description of the included studies is characterized in Table 1. The absorption rate (t max and C max ) of abacavir, lamivudine, and zidovudine were reduced. -Since these changes were not considered clinically important, the drug formulation may be taken with or without food. There were no significant changes in plasma concentration (C 0h and C 3h ) of nelfinavir and its active metabolite M8.

Studies Included in the Qualitative Analysis
Two-hundred and seventy-nine healthy subjects, and five-hundred and six PLWH who participated in twenty-four studies were included in the qualitative analysis ( Table 1). The participants include various ethnic and racial groups such as people of European, African, and Asian descent. There were 3 randomized controlled trials, 13 cross-over studies, 2 longitudinal studies, 4 pharmacokinetic studies, 1case report, and 1cross-sectional study. Among the included studies, three studies reported the effects of grapefruit juice or Seville orange juice on ARV pharmacokinetics [24,26,30]. While nine studies investigated the effects of food or nutrient supplement on ARV pharmacokinetics [12,14,25,27,32,38,[42][43][44]. The rest of the studies explored the pharmacokinetic interactions between ARV and vitamins or minerals [15,29,34,36,37,40], and dietary supplements [28,31,33,35,39,41]. Most studies showed non-significant changes in ARV pharmacokinetic parameters when coadministered with meals. However, there was a significant decrease in the extent of absorption of indinavir with liquid meals reported by Carver et al. [12]; protein, carbohydrate, fat, and viscosity meal treatments reduced indinavir AUC 0 -∞ by 68%, 45%, 34%, and 30%, respectively, while these meals decreased C max by 74%, 59%, 46%, and 36%, respectively. Regarding the meal compositions, fat meals resulted in the highest variability on the plasma indinavir levels. The consumption of garlic supplement decreased AUC and C max of saquinavir by 51% and 54%, respectively [28], while other dietary supplements such as quercetin, Ginkgo extract, Ginseng extract, β-carotene did not significantly affect saquinavir, lopinavir, ritonavir, and nelfinavir pharmacokinetics [31,33,35,41]. Grapefruit juice administration increased the oral bioavailability of saquinavir [24] while its concurrent intake did not lead to any change in indinavir concentrations [26]. A pharmacokinetic study by Slain et al. [29] indicated that after a week of 1000 mg vitamin C supplementation, the steady-state indinavir plasma concentration was reduced by 20%. On the contrary, some studies have reported that the use of multivitamins or minerals did not lead to any pharmacokinetic variation in ARV such as dolutegravir and nelfinavir [34,36,37,40]. In a case report of an HIV-infected man receiving raltegravir, concomitant calcium administration resulted in virologic failure showing detectable plasma HIV-1 RNA levels [15]. Coadministration of protease inhibitors such as nelfinavir with β-carotene led to a significant increase not only in mean CD4%, but also in the CD4:CD8 ratio in PLWH in one study [41]. In addition, the study also found an increase in CD4 counts; although, it was not clinically significant.

Studies Included in the Meta-Analysis
A total of 122 participants from 5 studies were included in the meta-analysis. Studies by Moore et al. [14], Yuen et al. [25], Falcoz et al. [27], Sekar et al. [32], and Yamada et al. [38] reported the pharmacokinetic outcomes of the interaction between food and ARV were eligible for meta-analysis. In general, the heterogeneity was high, and the 95% confidence intervals of the pooled data were close to the most weighted and included studies whose intervals were narrow. Figure 2 shows the meta-analysis of the pharmacokinetic effects of food on the AUC inf of ARV. There were no significant changes in AUC inf of ARV between fasted and fed conditions (mean difference -24.64, 95% CI -141.34to 92.07, p = 0.68), with a substantial heterogeneity of 75%. difference 0.67, 95% CI 0.44-0.91, p < 0.00001), showing the heterogeneity of 63%. Similarly, we found a significant increase in tmax in fed state compared to fasted condition in PI receiving groups (mean difference 1.29, 95% CI 0.81-1.77, p < 0.00001) ( Figure 5). For both NRTI and PI groups, none of these show any clinically significant changes in t1/2 ( Figure  6). We observed no statistical heterogeneity on t1/2 of both NRTI and PI regimens.    The overall effect of food on C max of NRTI group shows no clinical significance between fasted and fed conditions (mean difference −0.56, 95% CI -2.31 to 1.20, p = 0.53) in fixed effect models, while C max of NRTI group shows a significant decrease during fed state (mean difference -204.70, 95% CI -337.14 to -72.26, p = 0.002) in random effect models with the heterogeneity value of 84%. Interestingly, there was a significant increase in the C max of PI group during fed state (mean difference 845.51 (110.10-1580.92, p = 0.02) as shown in Figure 3. Figure 4 indicates the overall effect of food on C 24 , which was not significant (mean difference −0.13, 95% CI −3.08 to 2.82, p = 0.93).
The results of the subgroup meta-analysis of pharmacokinetic changes in tmax of ARV depending upon food or supplement administration are shown in Figure 5. The overall effect in NRTI group indicates a significantly higher tmax value under fed condition (mean difference 0.67, 95% CI 0.44-0.91, p < 0.00001), showing the heterogeneity of 63%. Similarly, we found a significant increase in tmax in fed state compared to fasted condition in PI receiving groups (mean difference 1.29, 95% CI 0.81-1.77, p < 0.00001) ( Figure 5). For both NRTI and PI groups, none of these show any clinically significant changes in t1/2 ( Figure  6). We observed no statistical heterogeneity on t1/2 of both NRTI and PI regimens.    nificant (mean difference −0.13, 95% CI −3.08 to 2.82, p = 0.93).
The results of the subgroup meta-analysis of pharmacokinetic changes in tmax of ARV depending upon food or supplement administration are shown in Figure 5. The overall effect in NRTI group indicates a significantly higher tmax value under fed condition (mean difference 0.67, 95% CI 0.44-0.91, p < 0.00001), showing the heterogeneity of 63%. Similarly, we found a significant increase in tmax in fed state compared to fasted condition in PI receiving groups (mean difference 1.29, 95% CI 0.81-1.77, p < 0.00001) ( Figure 5). For both NRTI and PI groups, none of these show any clinically significant changes in t1/2 ( Figure  6). We observed no statistical heterogeneity on t1/2 of both NRTI and PI regimens.    The results of the subgroup meta-analysis of pharmacokinetic changes in t max of ARV depending upon food or supplement administration are shown in Figure 5. The overall effect in NRTI group indicates a significantly higher t max value under fed condition (mean difference 0.67, 95% CI 0.44-0.91, p < 0.00001), showing the heterogeneity of 63%. Similarly, we found a significant increase in t max in fed state compared to fasted condition in PI receiving groups (mean difference 1.29, 95% CI 0.81-1.77, p < 0.00001) ( Figure 5). For both NRTI and PI groups, none of these show any clinically significant changes in t 1/2 ( Figure 6). We observed no statistical heterogeneity on t 1/2 of both NRTI and PI regimens.

Publication Bias
Publication bias was assessed by visual inspection of the funnel plot, which is shown in Figure 7. It was found that there were seven missing studies for tmax. The distribution of the standard error was close to the peak of the funnel plot, which demonstrated that there may be systematic difference or publication bias between smaller and larger studies. Further statistical test for the funnel plot asymmetry was not performed since the forest plot that had the largest number of included studies included only five studies [45].

Publication Bias
Publication bias was assessed by visual inspection of the funnel plot, which is shown in Figure 7. It was found that there were seven missing studies for tmax. The distribution of the standard error was close to the peak of the funnel plot, which demonstrated that there may be systematic difference or publication bias between smaller and larger studies. Further statistical test for the funnel plot asymmetry was not performed since the forest plot that had the largest number of included studies included only five studies [45].

Publication Bias
Publication bias was assessed by visual inspection of the funnel plot, which is shown in Figure 7. It was found that there were seven missing studies for t max . The distribution of the standard error was close to the peak of the funnel plot, which demonstrated that there may be systematic difference or publication bias between smaller and larger studies. Further statistical test for the funnel plot asymmetry was not performed since the forest plot that had the largest number of included studies included only five studies [45].

Discussion
This systematic review and meta-analysis mainly presented no significant pharmacokinetic changes in AUC, C24, and t1/2 of ARV when co-administered with food or dietary supplements. However, it can be highlighted that there was an increase in tmax of NRTI and PI under fed conditions. Of interest, the higher Cmax in fed state was observed in participants receiving the PI regimen in this study. Among the articles included in this systematic review, most of the ARV's pharmacokinetic parameters were not clinically altered by the concurrent administration of food and other supplements such as quercetin, Ginkgo extract, vitamin supplement, Ginseng, or β-carotene.
The different classes of ARV have varied pharmacokinetic metabolisms. PIs are extensively metabolized by the cytochrome P (CYP) 450 enzymes in the liver and small intestine [46]. Thus, ingestion of some foods or juices that have an inhibitory effect on CYP450 may increase concentrations of PIs. Kupferschmidt et al. [24] reported an increase in the bioavailability of saquinavir bioavailability after an intake of grapefruit juice, which is widely known as a CYP3A4 inhibitor [47]. In addition, a previous study reported an increase in AUC of atazanavir, lopinavir/ritonavir, nelfinavir, and saquinavir when ingested with food [48]. Under the fed condition, there was a long tmax of amprenavir from 1 to 4 h compared with the fasting state [27]. The tmax of darunavir under fasted conditions was approximately 1.5 h, whereas it was increased to 3-4 h after administration with food including standard breakfast, croissant with coffee, and high-fat breakfast. Similarly, the median tmax of 2 h of ritonavir without food extended to 4 or 5 h when food was administered concurrently [32]. Regarding the changes in plasma concentration of PI such as darunavir, Sekar et al. [32] reported the increased Cmax by about 30% under fed condition compared with the fasted state. Another class of ARV, NNRTIs such as efavirenz and nevirapine are also metabolized by several liver CYP isoenzymes [49]. The administration of efavirenz with high fat or high caloric meal was associated with the increased mean AUC by 28% and mean Cmax by 29%, respectively, compared to fasted conditions [50]. Most NRTIs are degraded by liver enzymes from the purine or pyrimidine nucleoside salvage pathway, depending upon the NRTI analogs [51]. Since they are not extensively metabolized by CYP450 [51], they have less interaction with food or other drugs. The delayed absorption of abacavir, lamivudine, and zidovudine was noted by Yuen et al. [25] an hour later in median tmax under fed conditions compared to fasted conditions. Similarly, Moore et al. [14] observed the slower tmax of lamivudine and zidovudine by 30 min and 45 min, respectively, after the administration of combined tablets with a high-fat meal. The mean tmax of emtricitabine and tenofovir alafenamide under standard breakfast was 2 h and 1 h, respectively, which was longer than those under fasted conditions [38,52]. The

Discussion
This systematic review and meta-analysis mainly presented no significant pharmacokinetic changes in AUC, C 24 , and t 1/2 of ARV when co-administered with food or dietary supplements. However, it can be highlighted that there was an increase in t max of NRTI and PI under fed conditions. Of interest, the higher C max in fed state was observed in participants receiving the PI regimen in this study. Among the articles included in this systematic review, most of the ARV's pharmacokinetic parameters were not clinically altered by the concurrent administration of food and other supplements such as quercetin, Ginkgo extract, vitamin supplement, Ginseng, or β-carotene.
The different classes of ARV have varied pharmacokinetic metabolisms. PIs are extensively metabolized by the cytochrome P (CYP) 450 enzymes in the liver and small intestine [46]. Thus, ingestion of some foods or juices that have an inhibitory effect on CYP450 may increase concentrations of PIs. Kupferschmidt et al. [24] reported an increase in the bioavailability of saquinavir bioavailability after an intake of grapefruit juice, which is widely known as a CYP3A4 inhibitor [47]. In addition, a previous study reported an increase in AUC of atazanavir, lopinavir/ritonavir, nelfinavir, and saquinavir when ingested with food [48]. Under the fed condition, there was a long t max of amprenavir from 1 to 4 h compared with the fasting state [27]. The t max of darunavir under fasted conditions was approximately 1.5 h, whereas it was increased to 3-4 h after administration with food including standard breakfast, croissant with coffee, and high-fat breakfast. Similarly, the median t max of 2 h of ritonavir without food extended to 4 or 5 h when food was administered concurrently [32]. Regarding the changes in plasma concentration of PI such as darunavir, Sekar et al. [32] reported the increased C max by about 30% under fed condition compared with the fasted state. Another class of ARV, NNRTIs such as efavirenz and nevirapine are also metabolized by several liver CYP isoenzymes [49]. The administration of efavirenz with high fat or high caloric meal was associated with the increased mean AUC by 28% and mean C max by 29%, respectively, compared to fasted conditions [50]. Most NRTIs are degraded by liver enzymes from the purine or pyrimidine nucleoside salvage pathway, depending upon the NRTI analogs [51]. Since they are not extensively metabolized by CYP450 [51], they have less interaction with food or other drugs. The delayed absorption of abacavir, lamivudine, and zidovudine was noted by Yuen et al. [25] an hour later in median t max under fed conditions compared to fasted conditions. Similarly, Moore et al. [14] observed the slower t max of lamivudine and zidovudine by 30 min and 45 min, respectively, after the administration of combined tablets with a high-fat meal. The mean t max of emtricitabine and tenofovir alafenamide under standard breakfast was 2 h and 1 h, respectively, which was longer than those under fasted conditions [38,52]. The concurrent administration of integrase inhibitors (IIs) and polyvalent cations such as magnesium and calcium should be monitored since these cations may bind with ARV, which leads to decreased plasma levels. Although Buchanan et al. [37] reported no clinical significance, dolutegravir, according to its medication package insert, should be taken 2 h before or 6 h after administering a cation containing antacids or laxatives, sucralfate, oral iron supplements, oral calcium supplements, or buffered medications [36,53].
The impacts of vitamin and mineral supplements on ARVs were mostly investigated in IIs [15,34,36,37] and some PIs [29,40]. Co-administration of 1200 mg calcium carbonate and 324 mg ferrous fumarate with dolutegravir during fasted conditions decreased in AUC 0-∞ , C max , and C 24 of dolutegravir by respective 37% to 39% and 54% to 57% due to chelation with polyvalent cations [36]. However, coadministration of these divalent cations and dolutegravir under fed conditions counteracted such the interaction [36]. Moreover, Roberts et al. [15] reported subtherapeutic raltegravir levels when co-administered with 1000 mg calcium carbonate. Nonetheless, information on the time interval between raltegravir and calcium intake or whether they were administered under fed or fasted conditions were not reported. Given that calcium supplement is necessary for the prevention of osteoporosis, especially in PLWH [54], future studies investigating the magnitude of such interaction between raltegravir and calcium supplement under fasted and fed conditions should be conducted. If the fed condition can counteract the interaction, calcium supplements may be co-administered with raltegravir. Concerning the impacts of vitamin and mineral supplements on PIs, while the divalent calcium ion did not significantly alter nelfinavir and its active metabolite M8 concentrations [40], co-administration with vitamin C significantly decreased steady-state indinavir concentrations [15]. This could be explained by the inductive property of vitamin C on CYP isoenzymes observed in animal studies [55][56][57][58][59]. However, another study reported no significant effect of vitamin C on CYP3A4, the primary enzyme used to metabolize indinavir activity [60]. Based on the contradictory effects of vitamin C on CYP3A4, further studies are needed to confirm the results.
Concerning the impacts of various dietary supplements, different classes of ARVs were investigated, mostly the PIs. Piscitelli et al. [28] reported that garlic supplement significantly decreased the extent of saquinavir absorption, as indicated by the decrease in AUC 0-8 , C trough , and C max . No definite underlying mechanism could be drawn from this study, but the authors proposed that it could be due to the induction of CYP450, or P-glycoprotein (P-gp) produced by long-term use of garlic. Nonetheless, evidence has shown that the effects of garlic on CYP3A4 are controversial as some in vitro studies reported that garlic had an inhibitory effect on CYP3A4 [61]. While other studies showed no significant effect of garlic on CYP3A4 [62,63]. Given that garlic supplements might have a negative effect on saquinavir exposure, garlic supplementation should be avoided in patients receiving treatment with saquinavir.
Quercetin, an inhibitor of various CYP450 including CYP3A4 as well as P-gp, could theoretically be used as a booster of saquinavir levels [64,65]. However, DiCenzo et al. [31] did not find significant effects of quercetin on saquinavir concentrations. The nonsignificant effect of quercetin could partially be explained by the intersubject and intrasubject variability. Given that only ten subjects were included in the study by DiCenzo et al. [31], further studies are required to confirm such results.
Ginkgo biloba, another widely used dietary supplement, has been shown to induce CYP3A activity [66,67]. However, Robertson et al. [33] reported no significant effects of Ginkgo biloba extract on lopinavir and ritonavir pharmacokinetics. These could be explained by the inhibitory effect of co-administered ritonavir on CYP3A. Nevertheless, the impacts of Ginkgo biloba extract on un-boosted protease inhibitors were not investigated in that study, and hence coadministration of Ginkgo biloba extract with un-boosted PI is not theoretically recommended. Similar to the effect of Ginkgo biloba, Panax ginseng has shown an inductive effect on CYP3A, which in turn may reduce PI concentrations [68]. However, the two-week administration of Panax ginseng did not alter lopinavir and ritonavir pharmacokinetics [35]. The same reason concerning the inhibitory effect of concurrently administered ritonavir could be applied here. Therefore, a similar recommendation on the use of Panax ginseng and un-boosted PI to that of Ginkgo biloba extract is proposed.
Inconclusive effects of β-carotene on CYP3A4 have been reported from in vitro studies, as one study found an inhibitory effect [69], while another reported an inductive property [70]. A clinical study investigating the impacts of β-carotene supplementation at the dose of 25,000 IU twice daily on nelfinavir and its active metabolite M8, which is metabolized by respective CYP2C19 and CYP3A4, indicated no clinically significant effect [41]. This implies that PLWH receiving ARVs that are substrates of CYP2C19 or CYP3A4 may be able to use β-carotene as a dietary supplement. However, it should be noted that these findings were based only on 15 HIV-infected subjects. Further prospective studies investigating the impact of β-carotene on other CYP2C19 and CYP3A4 substrates may be warranted.
This study has some limitations. The included studies did not report the component of fruit juices, i.e., whether the juice contained pulp or additives. This information may be crucial since these compositions might not be pharmaceutically inert. Since most studies were conducted with different ARV drugs and different supplements, only limited studies were included for meta-analysis. Only food (meal) and ARV pharmacokinetic interaction in healthy subjects were analyzed due to the limited number of studies with supplements and ARV pharmacokinetic interaction. Additionally, caution should be taken when the stratified results are interpreted as the minimal number of studies for subgroup analysis should be more than 10 [71]. Some studies from the qualitative analysis could not be included in the meta-analysis because of the diverse types of supplements among studies. The number of participants in each study was relatively small, and there may also have been interpatient or intrapatient variability. Furthermore, most studies were conducted in different regions with different study designs. Moreover, our systematic review may not be generalizable to all nutrient or food supplements as we did not focus on some products, e.g., probiotics or prebiotics, which may have effects on ARV pharmacokinetics. Of note, the heterogeneity across studies may also affect the results due to the potential effects of some factors such as gender, comorbidities, food compositions, dosage regimen, and duration of treatment.
This study found new information on the potential impact of supplement use on ARV pharmacokinetics. The study highlighted the decreased absorption of NRTIs such as abacavir, emtricitabine, and PIs such as ritonavir and darunavir after co-administration with food. The increased t max and plasma concentration of some ARV such as darunavir due to ARV-food interaction can suggest nutrition monitoring in relation to HIV and ARV treatment. As a further matter, due to the diversity of supplements, further research that considers not only the variability of interactions but the likelihood of an individual patient to develop potential outcomes should be examined. Improving awareness of therapeutic monitoring followed by pharmacokinetic evaluation of specific interaction should be implemented because there is extensive use of supplements in the HIV population.

Conclusions
This study pointed out the delayed absorption of NRTIs such as abacavir, emtricitabine, lamivudine, zidovudine, tenofovir alafenamide, and PIs including amprenavir, darunavir, and ritonavir under fed conditions including low-fat and high-fat meals. The higher plasma concentration of PI such as darunavir and cobicistat was also observed in a fed state. Considering the effects of various products on drug-metabolizing enzymes, more evidence in pharmacokinetic fields is required. So far, it is still necessary to investigate the potential interactions of ARV and particular supplement or complementary products in PLWH. Further pharmacokinetic studies are highly recommended to explore the potential pharmacokinetic interaction between food and ARV that might affect therapeutic outcomes. Physicians and healthcare providers need to be aware of the potential interaction between prescribed ARV and any mineral supplements that may lead to virologic failure or delayed absorption due to chelation. Since even a small change in plasma drug level can affect therapeutic efficacy, PLWH on ARV should be educated or given information on supplement use.

Supplementary Materials:
The following are available online at https://www.mdpi.com/article/10 .3390/nu14030520/s1, Table S1: Search terms for study screening, Table S2: The quality assessment of included studies.