Potential drug-drug interactions of antiretrovirals and antimicrobials detected by three databases

Standard treatment for HIV infection involves a combination of antiretrovirals. Additionally, opportunistic infections in HIV infected patients require further antimicrobial medications that might cause drug-drug interactions (DDIs). The objective of this study was to to compare the recognition of DDIs between antiretrovirals and antimicrobials by three proprietary databases and evaluate their concordance. 114 items of antiretrovirals and antimicrobials from the National List of Essential Medicines of Thailand 2018 were used in the study. However, 21 items were not recognised by Micromedex, Drugs.com, and Liverpool HIV interactions. Only 93 items were available for the detection of potential DDIs by the three databases. Potential DDIs detected from the three databases included 292 pairs. Liverpool showed the highest number of DDIs with 285 pairs compared with 259 pairs by drugs.com and 133 pairs by Micromedex. Regarding the severity classifications, Liverpool reported 10% Contraindicated; Micromedex reported 14% contraindicated and 59% major; Drugs.com reported 21% major. The Fleiss’ kappa agreements were fair to poor among the three databases, higher agreement was observed for DDIs classified as severe. This study highlights the need to harmonize the evaluation and interpretation of DDI risk in order to produce standardized information to support prescribers.

www.nature.com/scientificreports/ addition of antimicrobials for opportunistic infections may increase the risk of DDIs 13,14 . There is a challenge in the selection of drug regimens to maximise the efficacy and minimise the toxicity of ART and antimicrobial usage in HIV patients.
Recently, electronic DDIs databases were introduced to determine potential DDIs in the prescription of HIV infected patients 15,16 . The most popular database for healthcare providers seems to be Micromedex, provided by IBM Corp., USA. This database is subject to annual subscription fees, either individually or institutionally. In contrast, Drugs.com is a free online database, favoured by patients to identify potential DDIs, as well as some pharmacists who do not subscribe to the paid database. However, several reports have mentioned that the potential DDIs detected by Micromedex and Drugs.com have low to moderate agreement 17,18 . This has resulted in differences in the identification and management of DDIs. Interestingly, there is a specialised database commonly used to determine DDIs of HIV medicines in HIV community. Liverpool HIV interaction database focused mainly on HIV drugs was selected into this study in order to determine the ability and agreement of the three electronic databases in detecting potential DDIs of antiretrovirals and antimicrobials in the national list of essential medicines of Thailand (NLEM, 2018). Furthermore, the ability and agreement of the databases was considered in order to find the most suitable information for health care providers and HIV patients to optimise drug regimens.

Materials and methods
Drug selection. All antiretrovirals and antimicrobials were selected from the NLEM of Thailand 2018 19 .
Databases. The Micromedex database used in this study was provided by IBM Corp., USA. This database was accessed under the copyright license of Chulalongkorn University (2020). Drugs.com is a free online database powered by four independent leading medical-information suppliers: American Society of Health-System Pharmacists, Cerner Multum, Micromedex, and Lexicomp. In total, 93 items of antiretrovirals and antimicrobials were inputted into the database on 15 October 2020, in order to determine potential DDIs among antiretrovirals and antimicrobials. Micromedex reported information on onset, severity, documentation, probable mechanism, summary, literature, clinical management, and references. Drugs.com reported potential DDIs with information on severity, description, management, and references. The Liverpool HIV interactions reported severity, quality of evidence, summary and description with references. Severity classification of potential DDIs determined by the three databases was shown in Table 1.

Documentation of DDIs. Micromedex classifies DDIs as excellent documentation when controlled studies
have established the existence of the interaction; good documentation is defined as a strong suggestion that the interaction exists, but well-controlled studies are lacking. With fair documentation, the information is poor, but pharmacologic considerations have led clinicians to suspect that the interaction exists, or the documentation is good for a pharmacologically similar drug. Drugs.com did not provide information of documentation in their database. The Liverpool documentation was reported as quality of evidence. High quality is data obtained from a randomised, controlled interaction trial with clinical or validated surrogate endpoints. Moderate is obtained from crossover or parallel, steady state pharmacokinetics (PK) study with area under curves (AUCs). Low and very low are other kinds of information, e.g., editorial comments, animal studies, case reports without AUCs.  Fig. 1. Interestingly, we found that contraindicated and major severity DDIs had a higher tendency for good and excellent documentation than moderate and minor DDIs.
Of the 292 pairs detected by the three databases, similar severity, especially contraindicated and major, was found for 26 pairs (Table 2). For example, coadministration of atazanavir and nevirapine was characterised as contraindicated by Micromedex, major by Drugs.com, and do not coadminister by Liverpool. CYP3A4 inhibition by atazanavir and CYP3 induction by nevirapine were determined as pharmacokinetic-based mechanisms (Table 2S). Among these serious potential DDI pairs, pharmacokinetic-based DDIs had a 65% share, with CYP induction or inhibition as the main mechanisms. Meanwhile, 35% were pharmacodynamic-based DDIs, with QT-interval prolongation as the lead mechanism (Fig. 2).
Surprisingly, there are a large number of DDIs with poor concordance in severity grading across the three databases, e.g., minor or none by Micromedex and Drugs.com versus do not administer or potential interactions by Liverpool database (76 DDI pairs). For example, coadministration of Atazanavir and azithromycin was reported as minor or none by Micromedex and Drugs.com, meanwhile Liverpool database determined as potential interaction. This discordance might be due to the quality of evidence of this potential DDI was very low, and coadministration has never been studied. Both compounds could develop QT prolongation and atazanavir shows inhibitory effect on efflux transporters that responsible for azithromycin elimination. Therefore, atazanavir could potentially increase azithromycin exposure, leading to the increased risk for QT prolongation. This information might lead to the discrepancy in potential DDI classification among databases ( Table 3).
The agreement among the severity reports of the three databases, as determined by Fleiss' kappa value, was 0.129 (0.127 to 0.132, p < 0.001), which was considered to be a slight agreement among the three databases, as detailed in Table 4. While the agreement between two databases, Micromedex and Drugs.com, for the determination of all potential DDIs was 0.160 (0.156 to 0.165, p < 0.001), indicated slight agreement. Higher severity DDIs showed tendency for better agreement among databases.

Discussion
Antiretroviral combinations are a standard regimen for the treatment of HIV infected patients. Combination of antiretroviral drugs increase antiretroviral activity and reduces the risk of acquiring resistance 22,23 . Antiretroviral drugs have a high potential for drug-drug interactions which can occur with comedications used to treat for instance opportunistic infections 24,25 . In recent years, several online databases have been developed to determine potential DDIs. The present study selected three common databases, including Micromedex, Drugs. com, and Liverpool database, to determine the potential DDIs of antiretrovirals and antimicrobials available in the NLEM of Thailand 2018. Among 93 medicines included in this study, there were approximately 8,000 pairs of medicines. However, only 3-4% (292/8,000 pairs) could develop potential DDIs as determined by the three databases, compared with 18-20% (1,285/7,000 pairs) of potential DDIs in metabolic syndrome medications described in our previous report 18 . Even though the percentage of potential DDIs among antiretrovirals and antimicrobials was lower than that found for metabolic syndrome medications, the degree of severity seemed to be higher. Approximately 70% (97/133 pairs) of potential DDIs among antiretrovirals and antimicrobials were classified as contraindicated or major by Micromedex. Meanwhile, only 20% (155/724 pairs) of potential DDIs in metabolic syndrome medications were determined as contraindicated or major by Micromedex. Therefore, Among 292 pairs of potential DDIs, there were 26 pairs of major concerns. The three databases reported that these 26 pairs are contraindicated or major DDIs. Most of these serious DDIs have mechanisms related to CYP induction or inhibition and QT interval prolongation. Among the antiretroviral combinations in ART, protease inhibitors seemed to be the most troublesome class of medicines with moderate to strong CYP inhibition, and a less degree of CYP induction. For example, ritonavir is an enzyme inducer of CYP1A2 and CYP2C; meanwhile, it is an enzyme inhibitor of CYP3A4 and CYP2D6 26 . Induction and inhibition of CYPs could interfere with the metabolism of commonly available medicines because CYPs are the major drug-metabolising enzymes; approximately 50% of the top 200 prescribed drugs were biotransformed by CYPs 27 . Besides, non-nucleotide reverse transcriptase inhibitors such as Efavirenz is a moderate inducer of CYP3A4 whereas nevirapine is a moderate-weak inducer, which might reduce exposure of concomitant drugs if they are a substrate of CYP3A4. Avoidance of CYP overlapping antiretroviral combinations or therapeutic drug monitoring would be useful to prevent further complications with antiretroviral usage in HIV patients.
In the case of bacterial infections, the addition of antibacterials to ART seems to raise precaution for two groups of medicines, i.e. macrolides and quinolones. Erythromycin and clarithromycin have moderate to strong CYP inhibitory properties, and a class effect with QT-interval prolongation, which might lead to cardiac www.nature.com/scientificreports/ arrhythmia. Quinolones also showed similar properties to macrolides, especially CYP inhibition and QT-interval prolongation 28 . Therefore, coadministration of these antibacterials should be exempted from concomitant use with protease inhibitors and non-nucleotide reverse transcriptase inhibitors. Beta-lactam antibacterials seemed to be the better choice for coadministration with protease inhibitors or non-nucleotide reverse transcriptase inhibitors. Because most beta-lactams have good water solubility and minimal CYP biotransformation, they are mainly excreted via the kidney. Interestingly, the standard treatment of mycobacterial infections, e.g., tuberculosis, needs a combination of three to five antimycobacterials to eradicate the pathogen and prevent long-term resistance. Rifampicin is a potent CYP inducer; meanwhile, ethambutol is a CYP inhibitor; these two drugs are the backbone of the standard treatment for tuberculosis in Thailand 29 . It is based on the fact that enzyme inhibition could occur immediately after administration, whereas enzyme induction need more time for protein expression. Therefore, rifampicin is likely to have delay and stronger induction effect compared with rapidly inhibitory effect of ethambutol. In this case, therapeutic drug monitoring would be useful if adverse events or treatment failure were observed from any suspected drugs. Fungal infections in HIV patients can range from mild to severe. Azole derivatives seemed to be the most popular antifungals due to good safety profiles and convenient dosage forms, i.e. intravenous, oral, and topical preparations. These azole antifungals have excellent inhibitory properties to fungal CYP450s and to a lesser extent to human CYP450s. Itraconazole and posaconazole are more potent inhibitors of human CYP3A4 compared with fluconazole and voriconazole. Also, fluconazole and voriconazole are strong inhibitors of CYP2C9 and CYP2C19. Therefore, coadministration of azole antifungals and some protease inhibitors or non-nucleotide reverse transcriptase inhibitors would be careful with potential DDIs. Alternative antifungals against specific pathogens or therapeutic drug monitoring would be possible choices to manage the potential DDIs between antifungals and antiretrovirals.
Optimistically, the majority of antivirals against opportunistic infections are nucleoside analogues e.g., ganciclovir, acyclovir. These compounds are endowed with fairly high water solubility owing to the heterocyclic base and sugar in their structures. There is thus no need for CYP biotransformation and these drugs are predominantly eliminated through renal excretion. Therefore, potential DDIs, especially pharmacokinetic-based interactions,  Table 3. Drugs pairs with poor concordance in DDI severity grading across databases. www.nature.com/scientificreports/ are not likely to arise via CYP inhibition or induction. In addition, anti-hepatitis B and C drugs also show similar results compared with those from the aforementioned antivirals. Tenofovir and sofosbuvir are nucleotide analogues that contain base, sugar and phosphate moieties in their structures. Actually, high water solubility can be expected in tandem with minimal biotransformation via hepatic CYP enzymes, but these antiviral agents might be substrate of some transporters during excretion. Organic anion transporters are important for renal tubular secretion of guanine containing antivirals, e.g., ganciclovir, acyclovir. In addition, sofosbuvir and other NS5A inhibitors were reported as substrates of some efflux transporters, e.g., p-glycoprotein. These drugs can be victims of DDIs via inhibition or induction of drug transporters. Further information in antiviral DDIs via transmembrane transports are urgently required for database development. Management of potential DDIs among antiretrovirals and antimicrobials is a crucial step in HIV treatment. The three databases were able to detect serious DDIs with contraindicated and major severity. Alternative drugs with different pharmacokinetic or pharmacodynamic pathways seem to be appropriate options 30 . Other possible management involves close monitoring and dose adjustment to prevent adverse events 31 . Therefore, patient consultation and physician communication should be stressed to improve awareness of significant potential DDIs. These actions would be useful to prevent treatment failure and improve patient compliance with HIV treatment. One limitation of this study is the inclusion of antiretrovirals and antimicrobials, only drugs on the NLEM of Thailand 2018 were included, which did not cover all available drugs in each class. Additionally, the three databases might update information frequently, so during different search periods might provide different results. The difference between the drug interaction databases might be caused by the use of different references, unique rating priorities, and display formats 32 . Difference sources of references could be resulting in different severity classification and inconsistency among databases has been reported 33 . Not finding a comedication does not mean that there is no risk of DDIs. Combination of antibacterials or antivirals with similar metabolic pathways and toxicity meachnisms could lead to serious potential DDIs. Careful determination of potential DDIs by multiple databases and consideration for their agreement is necessary. Currently, there are no standard criteria for defining potential DDIs, resulting in complex interpretation of the results 34,35 . Therefore, a method to standardise DDIs information is required and the use of more than one database may be necessary.

Conclusion
A large number of potential DDIs were detected among antiretrovirals and antimicrobials used for opportunistic infection in the National List of Essential Medicines of Thailand 2018. Liverpool database showed the highest number of potential DDIs. DDIs with high severity as classified by contraindicated or major had better agreement value among the three databases. CYP induction or inhibition and QT prolongation were leading causes of serious potential DDIs. The observed poor agreement among databases highlights the need to harmonize the evaluation and interpretation of DDI risk in order to produce standardized information to support prescribers.