Interactions of Antiretroviral Drugs with Food, Beverages, Dietary Supplements, and Alcohol: A Systematic Review and Meta-analyses

Multiple factors may affect combined antiretroviral therapy (cART). We investigated the impact of food, beverages, dietary supplements, and alcohol on the pharmacokinetic and pharmacodynamic parameters of 33 antiretroviral drugs. Systematic review in adherence to PRISMA guidelines was performed, with 109 reports of 120 studies included. For each drug, meta-analyses or qualitative analyses were conducted. We have found clinically significant interactions with food for more than half of antiretroviral agents. The following drugs should be taken with or immediately after the meal: tenofovir disoproxil, etravirine, rilpivirine, dolutegravir, elvitegravir, atazanavir, darunavir, lopinavir, nelfinavir, ritonavir, saquinavir. Didanosine, zalcitabine, zidovudine, efavirenz, amprenavir, fosamprenavir, and indinavir should be taken on an empty stomach for maximum patient benefit. Antiretroviral agents not mentioned above can be administered regardless of food. There is insufficient evidence available to make recommendations about consuming juice or alcohol with antiretroviral drugs. Resolving drug-food interactions may contribute to maximized cART effectiveness and safety. Supplementary Information The online version contains supplementary material available at 10.1007/s10461-022-03880-6.


Introduction
Introducing combined antiretroviral therapy (cART) to the management of human immunodeficiency virus (HIV) infection not only did significantly reduce morbidity and mortality rates but also improved the quality of patients' life [1]. However, multiple factors may alter the effectiveness and safety of antiretroviral treatment, e.g., potential cumulative toxicity, suboptimal patients' adherence, drug-induced resistance, autoinduction, and inter-individual or inter-ethnical variability in drug response [2].
cART does not eradicate HIV, hence lifelong antiretroviral therapy is necessary. Newer classes of antiretroviral agents are better tolerated by patients with temporary gastrointestinal discomfort and fatigue as the most commonly reported side effects [3].
Nevertheless, chronic HIV treatment and aging may both contribute to the greater risk of metabolic disorders (e.g. dyslipidemia, insulin resistance, diabetes mellitus), cardiovascular diseases, hepatotoxicity, and renal impairment in people living with HIV (PLWH) [3][4][5]. Both temporary and prolonged adverse effects may entail therapy discontinuation and patients' poor adherence [6].
The causes of poor adherence to cART are diverse: starting from the individual (such as forgetting, depression, alcohol misuse), through treatment-related (e.g. the complexity of dosing regimen, side effects), ending with health-service barriers (such as poor patient-physician relationship, distance to the clinic) [6]. Prolonged suboptimal adherence to the cART may result in disease progression, a higher plasma viral load (that implies the increased risk of HIV transmission), and the development of drug-resistant HIV strains [6,7].
There are to main types of HIV drug resistance: pretreatment or acquired. Pretreatment HIV drug resistance occurs approximately in 10 percent of PLWH receiving initial treatment and its prevalence is the highest in low-and middle-income countries [8]. Acquired HIV drug resistance develops during the virus replication in the systemic presence of antiretroviral medications. According to the World Health Organisation (WHO) global report from 2019, the levels of resistance to commonly used nucleoside reversetranscriptase inhibitors (NRTIs) and non-nucleoside reverse transcriptase inhibitors (NNRTIs) may range from 21 to 91% and from 50 to 97%, respectively [8]. Medications like dolutegravir, lopinavir, or darunavir are characterized by higher genetic barriers to drug resistance [8,9].
Even in non-cART-resistant patients with high adherence levels, therapy can be ineffective, e.g. due to the autoinduction process. Several antiretroviral drugs, e.g. efavirenz, nevirapine, or nelfinavir may induce their own metabolism, leading to subtherapeutic plasma concentrations and unsatisfying treatment outcomes [10,11].
Individual and ethnic global population variability settles the antiretroviral drug response as well. There are both qualitative and quantitative differences in genetic variants of drug-metabolizing enzymes (especially cytochrome P450) between racial and ethnic groups [12]. Genetic polymorphisms determine levels of expression, activity, and stability of enzymes, and thus impact the drug metabolism rate.
Considering all the above-mentioned factors, cART optimization is critical for achieving effective virologic response and a satisfactory safety profile. In our previous publications, it was scientifically proved that the dosing regimen and drugfood interactions can either positively or negatively influence the treatment with several groups of drugs [13][14][15]. So far, two systematic reviews have been published addressing the topic of antiretroviral drug-food interactions. The first, by de Souza et al. covered studies to 2012, and only 11 studies were included [16]. Since then, many new antiretroviral drugs have been registered and novel formulations of already registered drugs have been developed. The second systematic review, by Siritientong et al., is up-to-date, however, in this review, very general keywords were used, making the number of included studies relatively small [17]. In addition, the effect of food was investigated either for all antiretroviral drugs simultaneously or within pharmacological groups, and not for individual drugs. Hence, there is a need for an up-to-date, detailed, comprehensive systematic review that considers the effect of food on particular antiretroviral drugs.
In this systematic review, we assessed the potential impact of food, beverages, dietary supplements, and alcohol on the pharmacokinetic and pharmacodynamic parameters of antiretroviral drugs. We decided to examine all antiretroviral drugs that were on the global market at the time the review was prepared or have been registered in the past. We included 33 drugs in our analysis. The main aim of this study was to identify the clinically significant drug-food interactions and to propose practical guidelines on how to take antiretroviral agents in relation to food.

Biopharmaceutical Characteristics of Antiretroviral Drugs
Even within the same pharmacological group, antiretroviral drugs may differ considerably in their chemical structure and physicochemical properties. In Table 1 we present biopharmaceutical characteristics of investigated antiretroviral drugs, namely the Biopharmaceutical Classification System (BCS) class, predicted logarithm of the partition coefficient (log P), solubility in water, and formulations available on the worldwide market (as of the date 10.10.2021). The data was obtained from http:// www. drugb ank. ca and Micromedex database).

Biopharmaceutical Classification System (BCS)
The Biopharmaceutical Classification System (BCS) divides drugs into 4 classes, based on their solubility in water and intestinal permeability: • class 1 (high solubility and high permeability)-such compounds are well absorbed and less vulnerable to factors that may affect bioavailability; food usually has no significant impact on absorption, • class 2 (low solubility and high permeability)-the bioavailability of those compounds is limited by their solvation rate; food usually has a positive impact on absorption (e.g. high-fat meal promotes dissolution of lipophilic drugs), • class 3 (high solubility and low permeability)-the drug dissolves fast but the absorption is limited by the permeation rate; food may negatively affect absorption by altering the process of drug dissolution, • class 4 (low solubility and low permeability)-those compounds have poor bioavailability; it is hard to predict the impact of food [18].

Logarithm of the Partition Coefficient (log P)
Partition coefficient (P) is the ratio of concentrations of an un-ionized compound in a mixture of two immiscible solvents (one is lipophilic, e.g. octanol, and the second hydrophilic, e.g. water). Log P value describes drug lipophilicity-the higher log P, the more lipophilic drug is: • log P < 1-hydrophilic drug, absorption can be lower in the presence of food rich in fat, • log P between 1 and 3-a drug with moderate lipophilicity, • log P > 3-lipophilic drug, food rich in fat may have a positive impact on drug absorption [19].
The abovementioned biopharmaceutical characteristics explains the rationale for considering the interactions with food for individual antiretroviral drugs, rather than within the pharmacological group.

Methods
This systematic review was conducted in adherence to the Preferred Reporting Items for Systematic Reviews and Metaanalyses (PRISMA) statement. The protocol was not prepared and not registered.

Information Sources and Search Strategy
In July 2022, the search in two databases was performed: Medline (via PubMed) and Embase, covering reports from the date of database inception to the date of the search. Additional records that can be classified as grey literature were identified via a Google Scholar search. Further reports were found by checking the product characteristics of antiretroviral drugs registered on the global market, as well as the reference lists of previously identified scientific publications.
During the searching process, the following keywords and phrases were applied: antiretroviral drugs names in combination with "food", "food-drug interaction", "meal", "diet", "breakfast", "dietary supplement", "alcohol", and "juice". When possible, MeSH terms and Emtree terms were used. In Medline (via Pubmed) and Embase, the keyword search was restricted to titles and abstracts, while in Google Scholar to titles only. The detailed searching strategy is provided in Supplementary Material S1.

Eligibility Criteria
All articles describing or investigating the impact of food, beverages, dietary supplements, and alcohol on pharmacokinetic and/or pharmacodynamic parameters of orally taken antiretroviral drugs were considered for inclusion in this systematic review. The pharmacokinetic parameters of interest were primarily AUC-area under the plasma drug concentration-time curve that reflects the extent of exposure to a drug, C max -the maximum (or peak) serum drug concentration, and t max -the time to reach C max that both relate to the rate of drug absorption. Pharmacodynamic indicators of drug efficacy were mainly plasma HIV-RNA count, mean CD4 gain, and frequency of virological failure. To present possibly the most complete and reliable evidence, no restrictions were applied regarding study type, study year, the number of participants, or their characteristics (e.g. age, gender, race). Both studies involving healthy volunteers and HIV (+) patients were considered. We excluded review studies, in vitro studies, and studies performed on animals.

Selection Process
The selection process was carried out using the Rayyan software. The authors, namely AW and PP, independently screened titles and abstracts of each record and selected those eligible for inclusion in the systematic review. Any disagreements between authors were discussed among the remaining authors and resolved by consensus.

Data Collection Process
From included studies, the authors, namely AW and MS, independently extracted available data of study type, the number of participants and their characteristics (health state, gender, race), antiretroviral drug dose and formulation, quantitative food composition (caloric load, percentage or weight amount of fat, carbohydrates, and protein), qualitative meal composition, alcohol concentration (in studies of interactions with alcohol), pre-and postprandial values of pharmacokinetic parameters (AUC, C max , t max ), pharmacodynamic parameters, and possible mechanism of interaction between antiretroviral drug and food. Additionally, statistically and/or clinically significant percentage changes in pharmacokinetic parameters were collected (if given) or calculated (if not provided by study authors). The data collection process was supervised by PP, who resolved any discrepancies.

Study Risk of Bias Assessment
Quality assessment of each included study was performed independently by two authors, namely AW and PP. Depending on the study design, different tools were used, such as version 2 of the Cochrane risk-of-bias tool for parallel trials (RoB 2) [20], Cochrane risk-of-bias tool for crossover studies [21], National Institutes of Health (NIH) quality assessment tool for observational cohort, and cross-sectional studies [22], and NIH quality assessment tool for before-after (pre-post) studies [22]. Any discrepancies in the assessment were discussed between the authors (AW and PP) and a consensus was made.

Data Synthesis
Quantitative analyses were performed for each drug if 2 or more food-effect studies with specified and comparable study designs were available, e.g. randomized and non-randomized studies were not synthesized in the same metaanalysis, as well as parallel and cross-over studies. The effect measures were mean differences (fed vs. fasted) of the three main outcomes: AUC, C max , and t max . If values of pharmacokinetic parameters were presented as geometric means with confidence intervals or the coefficient of variation, they were converted to arithmetic means and standard deviations using the method designed by Higgins et al. [23].
When median values and range or interquartile range were reported, the approach proposed by Wan et al. [24] was used to estimate arithmetic mean and standard deviation. For AUC, the adopted unit was ug·h/mL, for C max -ug/mL, and for t max -h. Results reported in other units have been transformed accordingly. Meta-analyses were conducted in the Review Manager (RevMan) [Computer program] Version 5.4.1, The Cochrane Collaboration, 2020. As the heterogeneity of studies was predicted to be high, the random effects model with the inverse variance method was used for the calculation of study weights. The results of meta-analyses were visually displayed as forest plots. To identify and measure the heterogeneity of studies included in the meta-analyses, the I 2 statistics and Chi 2 tests were calculated. I 2 < 25% together with the P-value from the Chi 2 test < 0.1 indicated the low heterogeneity, 25% < I 2 < 75%-the moderate heterogeneity, whereas I 2 > 75% and P > 0.1-the high heterogeneity [25]. In cases of moderate or high heterogeneity, subgroup analyses were performed. Grouping variables were: the type of meal, drug formulation, the health state of the participants, or study risk of bias. The grouping variables differed between the meta-analyses, depending on the characteristics of the included studies. Due to the small number of studies available, we chose to conduct a subgroup analysis if at least two studies were included in each subgroup. Additionally, to assess the robustness of the synthesized results, sensitivity analyses were conducted by changing the analysis model. Since none of the meta-analyses included 10 or more studies, funnel plots were not generated.
For drugs, for which meta-analyses could not be performed due to the unknown/variable study designs or lack of pharmacokinetic data, the results of available studies were summarized and discussed.

Eligible Studies
During an extensive databases search, 7814 records were identified in total: 3438 in Medline (via Pubmed) and 4376 in Embase. 4084 duplicate records were removed using the automation tool (Rayyan), and the other 1478 were deleted manually. Titles and abstracts of remained 2252 papers were meticulously screened. 2158 studies did not address the research question or met the exclusion criteria. Of 94 studies that were assessed for eligibility, 7 were excluded for reasons as follows: drug other than antiretroviral being assessed [26][27][28], a study design that does not allow to resolve the impact of food on an antiretroviral drug [29,30], a study performed on hepatitis B patients [31], the impact of food being assessed for intravenously given antiretroviral drug [32].
Additional 349 records were identified during a search in other information sources such as Google Scholar (322), product characteristics (16), reference lists (4), and conference reports (8). 41 records were sought for retrieval, and 15 were not retrieved since we did not find either an abstract or full text available. Of 26 reports assessed for eligibility, 4 were found to be duplicates after reading the full text [33][34][35][36], and 23 remaining reports were included.
Ultimately, in our systematic review, we included 109 reports of 120 studies. The flowchart of the search strategy is presented in Fig. 1.

Study Characteristics
The majority of included studies were open-label, cross-over clinical trials, as this study design is recommended by FDA for assessing the food effect [37]. The list of studies included in the systematic review is presented in Table 2. Detailed study characteristics are available in Supplementary Material S2, in which studies were organized by pharmacological groups and pooled for each antiretroviral drug. Hence, studies in which the effect of food on the pharmacokinetic/ pharmacodynamic parameters of several antiretroviral drugs was assessed, appeared more than once, with results collected for each drug. Additionally, studies investigating the impact of alcohol and juices were presented separately.

Risk of Bias Assessment
The risk of bias assessments carried out by AW and PP were generally consistent. For 12 studies, the assessment differed in one domain, however, the final assessments for all studies were the same. 32 studies (26%) were judged as having a high risk of bias, and the remaining were of moderate quality (at least one domain with some concerns). A detailed risk of bias assessment is presented in Supplementary Material S3.

Quantitative Syntheses
For 16 antiretroviral drugs, quantitative syntheses were performed. Studies excluded from meta-analyses (with reasons) are listed in Supplementary material S4. Overall, 50 metaanalyses were conducted. In Table 3, the results of metaanalyses for individual antiretroviral drugs are presented. Forest plots of each meta-analysis are available in Supplementary material S5.

Subgroup Analyses
In 29 (58%) of meta-analyses, moderate or high heterogeneity was revealed, and for those, subgroup analyses were conducted (when possible). In Table 4, we present the results of only those subgroup analyses where the grouping variables potentially explain the heterogeneity of the studies included in meta-analyses. Forest plots of all performed subgroup analyses are available in Supplementary material S5.

Sensitivity Analyses
After changing the statistical model from the random effects model to the fixed effects model, no significant qualitative differences in overall effect were found for 45 of 50 meta-analyses, however, the fixed effects model generally produced higher values of overall effect and narrower confidence intervals of a mean difference than the random effects model. For the remaining 5 meta-analyses, differences between random and fixed effects models are listed in Table 5. Forest plots of these meta-analyses are provided in Supplementary material S5.

Qualitative Syntheses
For 17 antiretroviral drugs, conducting meta-analyses was not possible due to the insufficient number of studies eligible for quantitative synthesis. For those drugs, we summarized the available evidence in Table 6, whereas a detailed description of studies is provided in Supplementary material S2.

Interactions with Dietary Supplements
We found limited evidence for interactions between several INSTIs and mineral supplements. Dolutegravir and bictegravir AUC and C max both decreased by 33-42% in the presence of calcium carbonate. Ferrous fumarate analogically caused a decrease of AUC and C max by 55-58% for dolutegravir, and by 63-70% for bictegravir [90,135]. Interestingly, when dolutegravir and bictegravir were given with both supplements and a moderate-fat meal, no significant changes in pharmacokinetic parameters occurred. Another solution proposed to avoid malabsorption was maintaining the 2 h break between dolutegravir or bictegravir and calcium or iron supplements intake [90,135].
HIV-positive patients often use garlic supplements due to their immunomodulating and cholesterol-lowering effects. Garlic products may either induce or inhibit CYP enzymes and thus the potential risk of pharmacokinetic interaction with several protease inhibitors exists. Garlic supplements were found to interfere with saquinavir absorption by decreasing AUC and C max by 51% and 54%, respectively [118]. However, individual concentration-time profiles revealed that in 3 patients AUC was slightly increased during garlic supplement consumption. Interestingly, after  Regarding darunavir, we found case studies of 2 patients with substantial garlic consumption, in whom subtherapeutic C through darunavir concentrations were revealed. After garlic eviction, darunavir concentrations normalized within 1 month [47].
After ritonavir co-intake with a garlic-containing dietary supplement, no significant changes in AUC and C max occurred. However, these results cannot be extrapolated to steady-state conditions [58]. Moreover, regular garlic consumption during the treatment with ritonavir may exacerbate gastrointestinal adverse effects [58].

Interactions with Juices
Ingredients of grapefruit and Seville orange juices act as inhibitors of intestinal CYP enzymes (especially CYP3A4). Among antiretroviral drugs, protease inhibitors present the highest potential to interact with juices, due to being extensively metabolized by CYP enzymes.
Amprenavir acts both as a substrate and inhibitor of CYP3A4. Juice ingredients can reduce amprenavir first-pass metabolism by inhibiting intestinal CYP enzymes. Nevertheless, co-administration with 200 mL of grapefruit juice only slightly decreased amprenavir C max (by 22%) and delayed t max (by 0.4 h), without significantly affecting AUC [54]. These results indicate that the gut metabolism of amprenavir is low.
For indinavir, the mean AUC and C max values remained unchanged after co-intake with grapefruit juice [117,130]. However, in one study, individual changes in indinavir AUC ranged from a 25% decrease to even a 25% increase [117]. Such a high variability may negatively impact the treatment, by either causing its ineffectiveness or increasing the risk of adverse drug reactions.
After indinavir administration with Seville orange juice, no significant changes in AUC and C max occurred [117]. However, t max was slightly longer (by 0.6 h), since a high amount of carbohydrates (especially pectins) in juice may delay gastric emptying [117].
In a study of saquinavir, AUC and C max significantly increased (by 50% and 93%, respectively) after co-intake with grapefruit juice [77]. By inhibiting intestinal CYP enzymes, juice ingredients reduced the extensive first-pass metabolism of saquinavir. However, that effect is variable and should not be considered as therapeutic guideline to improve low saquinavir bioavailability [77].
In patients with gastric hypoacidity, delavirdine absorption can be even 50% lower [131]. Co-intake with acidic beverages, such as orange juice, may improve delavirdine bioavailability. Regular orange juice ingredients do not inhibit CYP enzymes but may lower gastric acidity. In a  Heterogeneity can be explained either by the type of meal or health state For low-fat and light meals and in healthy participants-a lower mean difference in C max between fasted and fed states single study, the elevations of delavirdine AUC and C max were observed (by 57% and 53%, respectively) after coadministration with regular orange juice. Improved AUC and higher C max , however, were only visible in patients with gastric pH ≥ 3 [131].

Interactions with Alcohol
The pooled prevalence of alcohol use disorders among people living with HIV/AIDS is almost 30% [147]. Excessive alcohol consumption not only impairs a patient's immunity but also may cause liver damage that contributes to the disease progression [147]. There are concerns that alcohol may alter the effectiveness and safety of antiretroviral therapy due to the shared metabolism pathways with antiretroviral drugs. On the other hand, when the recommendation to avoid co-intake of alcohol and antiretroviral drugs is made, patients often miss doses of their medications if they plan on consuming alcohol. Such non-adherence can lead to the development of resistance to therapy. We found only several studies addressing the issue of interactions between antiretroviral drugs and alcohol. Cointake of maraviroc with ethanol did not produce any significant changes in maraviroc pharmacokinetics. Contrastingly, ethanol AUC increased slightly but significantly (by 12%), indicating that maraviroc can potentially enhance alcohol concentration and toxicity [59].
In a study of abacavir, a single dose did not significantly affect ethanol pharmacokinetics, however, abacavir AUC and t 1/2 significantly increased (by 41% and 26%, respectively) [92]. Both abacavir and ethanol are metabolized by alcohol dehydrogenase and hence may compete in the metabolism phase [92].
After the co-intake of efavirenz and ritonavir with alcohol, no significant changes occurred in the pharmacokinetic parameters of these drugs [91]. However, in the presence of both efavirenz and ritonavir, ethanol blood concentration slightly decreased (AUC by 14% and C max by 12%) [91].

Findings and Interpretations
Results of our review indicate that food may have a diverse and significant impact on the pharmacokinetic and pharmacodynamic parameters of antiretroviral drugs. The overall high heterogeneity of investigated studies suggests that the basis of antiretroviral drug-food interactions is multifactorial. Based on quantitative and qualitative syntheses, we pointed out several factors that may influence the Table 5 Significant qualitative differences in the results of meta-analyses after changing statistical model from random to fixed effects model  High-and medium-fat meals: ↑ by 510-520% in a fed state Low-and very low-fat meals: ↑ by 220-310% in a fed state magnitude and clinical relevance of antiretroviral drug-food interactions.

Physicochemical Properties of a Drug
Physicochemical drug properties rather than belonging to the pharmacological group may explain the impact of food. The majority of NRTIs are hydrophilic drugs, with log P values < 1 (Table 1), which makes them highly soluble in water. Hydrophilic compounds dissolute slower in the presence of fat, and gastric emptying is prolonged after a meal as well. It may explain significantly higher postprandial values of t max reported for emtricitabine, lamivudine, stavudine, zalcitabine, and zidovudine. On contrary, tenofovir disoproxil and alafenamide are lipophilic drugs (with log P values between 1 and 3), which may explain better absorption in the presence of meals rich in fat. Our review indicates that food has a positive or neutral impact on the bioavailability of NNRTIs. Drugs from this group belong mostly to the 2nd BCS class (Table 1). They exert low solubility in water and are highly permeable through the intestinal membrane. Except for delavirdine and nevirapine, all NNRTIs are lipophilic, with log P > 3 (Table 1). Fat from food induces bile acids and pancreatic juice secretion and thus promotes the solubilization of lipophilic NNRTIs. High-fat meals delay gastric emptying so it takes longer for NNRTIs to dissolve and their absorption is better.
Food has shown a positive or neutral impact on INSTIs absorption. Most of INSTIs are members of the 2nd BCS class (Table 1) with low solubility in water and high intestinal permeability. Only raltegravir belongs to the 3rd BCS class, hence it is highly soluble and hardly permeable. INSTIs vary in terms of lipophilicity-from the most lipophilic elvitegravir (with log P > 3) to the most hydrophilic raltegravir (with log P < 0).
The influence of food differed among drugs that belong to the PIs group. PIs are heterogeneous in terms of BCS classification (class 1st, 2nd, and 4th), lipophilicity (log P from 1.8 to 6.29), and solubility in water (Table 1). For lipophilic drugs, such as atazanavir, lopinavir, nelfinavir, ritonavir, and saquinavir, our review has confirmed more efficient dissolution and absorption in the presence of food rich in fat.

Type of Meal
For several drugs, the impact of food on pharmacokinetics differed depending on the type of meal. Didanosine co-administration with a high-fat meal or a standard meal produced more distinct changes than with low-fat and light meals such as yogurt or apple [53,72,75]. It may be explained by delayed gastric emptying in the presence of food rich in fat. Furthermore, gastric juice secretion is increased after a meal, resulting in lower pH values. In an acidic environment, didanosine is unstable due to hydrolysis [53].
For rilpivirine, the bioavailability was improved in the presence of a high-fat meal, whereas the impact of a standard or moderate-fat meal differed from neutral to beneficial [48,51,81,93]. However, no positive effects were observed after the co-administration of rilpivirine and the protein-rich cocktail. It can be explained by the slower tablet disintegration and drug liberation in the presence of nutritional drinks [48].
Also, raltegravir and nelfinavir absorption depended on the type of meal and improved with the increasing content of fat in food [39,43,76,107,148]. The opposite pattern was observed for atazanavir-light meals improved bioavailability, whereas high-fat meals exhibited a neutral effect [99,128]. The impact of a meal type was revealed for indinavir as well. Due to its basic chemical character, indinavir is mostly absorbed in the upper small intestine. Meals rich in fat delay gastric emptying, neutralize acidic pH, and may cause indinavir precipitation. Consequently, less drug can reach the absorption site, and the overall bioavailability is lower [144]. A negative food impact was not observed for low-fat and low-calorie meals [38,144].

Drug Formulation
The formulation of an antiretroviral drug could be another factor determining the onset and magnitude of interaction with food. For didanosine, in studies of enteric-coated capsules instead of tablets, less distinct postprandial changes in pharmacokinetic parameters occurred [53,61]. Moreover, for enteric-coated capsules taken with or without food, similar antiretroviral activity was reported after 28 days [61]. The enteric-coated formulation may be preferable over tablets, as it protects didanosine from hydrolysis in the presence of gastric acid and improves its AUC as well [53].
Lamivudine, tenofovir disoproxil, and doravirine tablets (single or combined with other antiretroviral agents) were absorbed more slowly after the meal, whereas no significant postprandial changes in t max were observed for coated and uncoated granules [42,143]. Moreover, our findings suggest that ingesting lamivudine granules in the fasting state may slightly improve its bioavailability, while the impact of food on the AUC of lamivudine tablets remains neutral.
For tenofovir alafenamide and rilpivirine, the food effect was lower when both drugs were given with other antiretroviral agents as a combined tablet.
In the case of darunavir, a positive food influence on bioavailability was observed only for tablets. No significant deviations were reported when the oral suspension was administered with a standard meal, except for the delay of t max [66]. For lopinavir tablets, contrastingly, the food effect was diminished, possibly due to the presence of hydrophilic excipients in a formulation [74].
For ritonavir, the bioavailability of an oral solution and capsules remained unchanged in the presence of meals rich in fat, whereas tablets and oral powder had slightly lower values of AUC and C max when taken with a high-fat meal [67,97,109,124].

Patient's Age
Age-related differences in pharmacokinetics between children and adults may influence the food effect. For example, unlike in adults, in HIV-positive children, co-intake of didanosine with a standard meal did not produce significant changes in AUC but delayed the absorption [136].

Autoinhibition and Autoinduction
Autoinduction and autoinhibition can modulate the effect of food on antiretroviral drugs and vice versa, dietary interventions may potentially compensate for changes in drug concentrations caused by influence on its own metabolism.
For example, delavirdine can inhibit its own metabolism, so with increasing drug doses, plasma concentrations higher than proportional can be observed [149]. This effect may partially compensate for the potentially negative food impact on pharmacokinetic parameters.
Contrastingly, efavirenz induces CYP enzymes and thereby accelerates its own metabolism. The autoinduction effect varies based on treatment duration-the longer the treatment is, the higher enzyme induction could be. Autoinduction can cause suboptimal therapy effectiveness, so the need for administering an increased efavirenz dose to maintain the optimal clinical response. Moreover, autoinduction contributes to drug resistance development [10]. Autoinduction is the most prevalent among CYP2B6 extensive metabolizers. In CYP2B6 slow metabolizers, the presence of a CYP3A5 genotype allele can be associated with a greater impact of efavirenz autoinduction on plasma drug concentrations [150]. Efavirenz's co-intake with food may potentially mitigate aberrations caused by autoinduction. Increased etravirine bioavailability after co-intake with some types of food may compensate for the metabolic autoinduction effect as well [151]. However, close monitoring of a drug concentration is recommended during the intentional nutritional intervention.
Also ritonavir, as a potent CYP3A4 inhibitor, can induce its own metabolism, causing instability in plasma concentrations during the first 2 weeks of therapy [152]. Irregular drug intake with regard to food may additionally worsen these fluctuations.

Inter-individual and Inter-ethnical Variability
When discussing antiretroviral drug-food interactions, the aspect of inter-individual and inter-ethnical variability cannot be overlooked. For example, efavirenz is metabolized primarily by the CYP2B6 enzyme that is characterized by high genetic polymorphism, e.g. the presence of CYP2B6 516T allele leads to the 50-75% reduction of enzyme activity, whereas CYP2B6 785 G is associated with the increased activity [153]. A meta-analysis from 2019 confirmed that homozygous individuals with the T allele have substantially higher efavirenz plasma concentrations than those with the G allele [154]. CYP2B6 516G > T polymorphism is more frequent in African-Americans compared to Hispanic, Caucasian, and Asian populations (46%, 27%, 21%, and 17% respectively) [155,156]. Moreover, approximately 4-12% of African-Americans, are the rare carriers of the CYP2B6*18 variant that inhibits functional protein expression [157]. Therefore, in African-American poor metabolizers, reduced efavirenz clearance is relatively frequent. In consequence, serum drug concentrations are higher, and the risk of side effects increases. The genotype-based efavirenz dose adjustment strategy might be a resolution for the problem of CYP2B6 polymorphism.
Similar to efavirenz, inter-individual variability in nevirapine plasma levels can be observed. The CYP2B6 516 TT genotype is associated with increased nevirapine concentrations if compared to the 516TG and 516GG genotypes [158]. During the first month of the therapy, nevirapine pharmacokinetic parameters can be unstable due to the autoinduction of CYP3A4 and CYP2B6 enzymes [159]. Fluctuations in plasma drug levels can be partially compensated by administering nevirapine in a constant relationship to food.
In reference to etravirine metabolism, CYP2C19 enzyme genetic polymorphism may occur. In patients with the CYP2C19*2 gene variant, etravirine clearance can be reduced by 23%, leading to a higher drug concentration.
Interestingly, the impact of a low-fat meal on lopinavir absorption depended on the race. In Asian patients, no significant changes in AUC and C max were observed, whereas in Caucasian patients AUC increased by 72%, and C max by 38% [113]. Inter-racial diversity in lopinavir metabolism is obvious in this case. Single nucleotide polymorphism in the CYP3A4*22 gene is responsible for decreased clearance and higher plasma concentrations of lopinavir. This gene variant is more frequent in Caucasians and virtually absent within the East Asian population [12].

Limitations of Studies Included in the Review
The majority of investigated food-effect studies were randomized, open-label, cross-over clinical trials, as recommended by Food and Drug Administration (FDA). Unfortunately, in almost all cases the randomization process was not described in detail and information was insufficient to judge whether the allocation sequence was truly random and concealed. Additionally, we identified several issues regarding the cross-over design, such as a lack of reporting data from each period of a trial separately, a lack of information on the number of participants allocated to study sequences, or too short washout period. For 15 studies, the design was not specified at all, and 8 were non-randomized clinical trials. To obtain possibly the most complete evidence, we decided to include all food-effect studies (from the 90 s and 00 s as well). In older studies, the methodology was usually described very basically, making it difficult to assess the risk of bias.
Regarding study participants, their number was relatively small, as almost half of the studies (47%) involved only 15 participants or less. However, according to FDA guidelines, the minimum number of participants in a food-effect study should be 12. The apparent limitation is that 73% of studies involved healthy volunteers. Moreover, in almost half of the studies the gender and/or race of participants were not specified, and in the majority of the remaining studies, African-American race was generally the least represented.
The results of such studies can be difficult to translate into clinical practice, given the already discussed inter-ethnical variability in response to antiretroviral drugs.
Our findings indicate that drug formulation and type of meal could be potential factors influencing the magnitude of food effects on certain antiretroviral drugs. However, for many of investigated drugs, not all available formulations were tested in the presence of food, and in several studies, it was not even mentioned which drug formulation is being investigated.
Regarding meal composition, in half of the studies, the quantitative or qualitative meal composition was not specified. Another frequent problem was that the same types of meals (e.g. high-fat, high-protein, low-fat, etc.) substantially differed in terms of their qualitative and quantitative composition, and in several studies the patient's typical diet instead of the standardized meal was tested.
Overall, given all the abovementioned limitations, studies included in this review were judged as having a moderate or high risk of bias (see Supplementary material S3 for details).

Limitations of the Review
The apparent limitation of our review is the unproportionate evidence of antiretroviral drug-food interactions. Quantitative syntheses were only possible for 16 of 33 antiretroviral drugs since for the remaining, the data on food impact was usually limited to one study, often mentioned only in product characteristics.
Due to the overall high heterogeneity of studies included in this systematic review, a substantial number were not eligible for inclusion in meta-analyses (see Supplementary material S4). The most frequent reasons for exclusion were: inappropriate or unknown study design or effect measure, and missing data of investigated outcomes. According to Cochrane guidelines, the minimal number of studies for subgroup analysis and testing for funnel plot asymmetry should be more than 10 [25]. In this review, the average number of studies in a single meta-analysis was 3-4, and the maximum was 6. Nevertheless, we have performed subgroup analyses 1 3 to find possible reasons for high heterogeneity, however, their results should be interpreted with caution. The reporting bias cannot be excluded, but we did not investigate it due to the limitations outlined above.

Final Recommendations
The summary of recommendations for the optimal intake of antiretroviral drugs with regard to meals is presented in Fig. 2.
For the drugs included in the qualitative syntheses, the strength of the recommendations is very low, whereas for drugs included in quantitative syntheses is low to moderate, depending on the drug.
The data on the interactions of antiretroviral drugs with dietary supplements, juice, and alcohol is scarce and limited to individual drugs, hence we were not able to outline even general recommendations.

Summary
In this comprehensive, substantive systematic review, we found evidence for clinically significant interactions with food for more than half of 33 investigated antiretroviral drugs. It is a clear indication that further education about drug-food interactions is necessary. Raising awareness about the proper intake of antiretroviral agents with food should be a priority to optimize HIV patients' cART. Figure 2 shows an overview of recommendations for the optimal intake of antiretroviral drugs with regard to meals.
Our review revealed existing gaps in the knowledge of interactions between antiretroviral drugs and dietary supplements, juice, and alcohol. In our opinion, further in-depth studies are urgently needed. New evidence could be used in the future as the cornerstone of the informed decisionmaking process regarding HIV therapy.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/.