Comparing potential drug-drug interactions in veterinary medications using two electronic databases

Background: One of the most common global health issues in humans and animals is drug-drug interactions (DDIs). This issue increases the risks associated with healthcare in both human and veterinary medicine, as animals live long lives and receive many medicines to treat their illnesses. Currently, many electronic databases are being used as tools for potential DDI prediction, for example, Micromedex and Drugs.com. The purpose of this study was to examine the different abilities for the identication of potential DDIs in veterinary medicines by Micromedex and Drugs.com. Results: A list of 140 drugs, mainly used for the treatment of disease in animal hospitals, was complied, but the Micromedex and Drugs.com databases could recognise only 96 of these drugs. After inputting the recognised drug list into the databases, Micromedex showed 429 pairs of potential DDIs, whilst Drugs.com showed 842 pairs of potential DDIs. The analysis comparing results between the two databases showed 139 pairs (12.28%) with the same severity and 993 pairs (87.72%) with different severities. Major mechanisms of contraindicated and major potential DDIs were cytochrome P450 induction-inhibition and QT interval prolongation. Conclusion: Although Micromedex had a lower sensitivity to identify potential DDIs than Drugs.com, Micromedex provided more informative documentation. Veterinary pharmacists should evaluate potential DDIs from several databases and communicate with both the veterinarian and animal owner to ensure an appropriate drug prescription.

diseases and cancers. These diseases require multiple drug use and might result in DDIs in sick animals.
This study aimed to investigate the differences in performance of DDI databases for identifying potential DDIs with complicated disease treatments used in animals.

Results
From the 96 drugs used in this study, 1132 unduplicated pairs were found by the selected databases as potential DDIs. Micromedex identi ed 429 pairs of potential DDIs and Drugs.com identi ed 842 pairs of potential DDIs. Table 1   After comparing all of the potential DDI results analysed by the two databases, 139 pairs (12.28%) showed the same severity, whilst 993 pairs (87.72%) of results showed a difference in severity. From all of the results, contraindications and major DDIs identi ed by Micromedex and major DDIs reported by Drugs.com were selected for determination of the type of mechanism of each potential DDI report, as shown in Table 2. Among the 86 pairs of signi cant potential DDIs, 15 pairs were at the contraindication degree reported by Micromedex and classi ed at the major degree by Drugs.com. The remaining 71 pairs were reported at the major degree by both databases.   In terms of mechanism, 51% (44/86) were pharmacokinetics (PK)-based, 34% (29/86) were pharmacodynamics (PD)-based and 15% (13/86) were PK-PD-based. The majority of PK-based DDIs involved cytochrome P450 (CYP) induction and inhibition, whilst PD-based DDIs caused QT prolongation and potassium retention. We also found some con ict between the results of the two databases, in which one database reported potential DDIs as major but another one reported them as minor or not DDIs. For the dissimilar results as shown in Table 3, 32 pairs were identi ed by Micromedex as major DDIs but only minor or not DDIs by Drugs.com. Conversely, 53 pairs were speci ed as major DDIs by Drugs.com whilst Micromedex identi ed these as not DDIs.

Discussion
The two databases used in our study could recognise animal medicines less frequently than human medicines. This may be because the databases do not include complete information about animal medicines in their DDI databases. However, Drugs.com included a list of veterinary products that covered many animal species and provided useful information, for example, dosage, administration, precautions and adverse reactions. Therefore, veterinary pharmacists should use this drug database for searching for drug information and as a source for reference documents. The potential DDI results from the drug lists were different for the two electronic databases. In this study, the results from Drugs.com exhibited a higher number of potential DDIs than Micromedex by nearly 2.0 times. This result was correlated with Lauren et al., who found that Drugs.com had higher sensitivity than Micromedex for screening DDIs in oral cancer treatment [13]. Suriyapakorn et al. compared the capability of the databases to identify potential DDIs with metabolic syndrome medications and also found that Drugs.com provided more sensitivity than the other database [15]. The reason for the high sensitivity of Drugs.com in identifying DDIs may be caused by using databases from many providers to analyse data that is contained in Micromedex. We compared the result from Micromedex between the drug list used for metabolic syndrome in human and animal diseases in identity contraindicated and major potential DDIs. The drug list for animal disease treatment identi ed more pairs of contraindicated and major potential DDIs; the reason might be that the drugs used in animals included many drugs related to antiarrhythmic agents, antimicrobials and antihypertensive drugs, which often show a high incidence of potential DDI when prescribed with other drugs [16][17][18]. Veterinary pharmacists should realise when prescribing these drug groups to avoid the severe adverse reactions.
The combination of drugs prescribed to treat canine atopic dermatitis were identi ed as a potential DDIs; for instance, a co-prescription of ketoconazole with cyclosporine has been suggested, which could reduce the therapeutic cost and is convenient to use. This combination appears to provide greater clinical bene ts for the treatment than disadvantages. Nevertheless, an excessive number of alerts of potential DDI lacking supporting information could cause wearying to healthcare providers as well as animal owners. Many healthcare providers dislike using drug interaction databases [19] for several reasons, including alert fatigue [20][21], work ow disturbance [22] and believing that there is no clinical signi cance related to most DDI alerts [23]. Ideally, an applicable drug interaction database should have both high sensitivity in identifying signi cant interactions and high speci city in excluding insigni cant interactions. As a result, healthcare providers should be sure to use more than one DDI reference for reaching the best nal answer when identifying potential DDIs [24]. Apart from using several DDI databases, healthcare providers should share their decision-making with animal owners regarding any signi cant potential DDI pairs, to preclude animals from adverse events and minimise liability.
In multiple-drug prescriptions, drug dosages usually have a relationship with drug interactions. For example, giving high doses of some drugs may cause interactions, but if they are used at lower doses, the possibility of a DDI may decrease. Ideally, the DDI database should be able to overlook an interaction if the given drugs are at doses that will not likely develop into a DDI [25][26], which the two databases in this study were not able to do. Therefore, the input of dosage should be added to databases, so healthcare providers can select the dose of drugs under consideration. Interestingly, Micromedex and Drugs.com provide detailed information differently. Micromedex adds more information on allergy interaction, alcohol interaction, lab interaction, tobacco interaction, pregnancy interaction and lactation interaction; Drugs.com provides more information only on food interactions and presents results into two categories: consumer and professional. The diversity of information from these two databases gives many bene ts to increase the con dence of healthcare providers when many health conditions are discussed with their clients.
For determination of the mechanisms of potential DDIs in contraindicated and major severity levels identi ed by the two databases, PK-based was the main mechanism of DDI, followed by PD-based and PK-PD-based. The PK-based DDI causes a change of drug concentration in plasma or at the targeted organ by altering the absorption, distribution, metabolism and elimination. In this study, CYP enzyme inhibition was the main result of PK-based DDI, which leads to the accumulation of co-administered drugs, for example, cipro oxacin inhibits CYP1A2 activity and decreases the clearance of theophylline, resulting in fatal adverse drug reaction via drug toxicity. PD-based DDI is caused by one drug interfering with another drug at the target site. The main result of PD-based DDI in our study was QT prolongation, which may lead to irregular heart rhythm and sudden death, for example, co-administration of erythromycin with sotalol may result in an additive QT prolongation. As a result, healthcare providers should truly understand these DDI mechanisms to prescribe multiple drugs properly.
This study has several important limitations. The most important one is that only two drug databases were used for the evaluation, so future studies should include additional databases of both subscription and open-access type, such as Lexicomp and Epocrates Free, respectively. Changing of the drug list was also one of our concerns; each year new drugs were developed whilst old drugs disappeared. The drug list used in this study was gathered in the rst quarter of 2020, so it might have changed at any time. The updating frequency of databases for their potential DDI reports might affect the results of the analysis. The potential DDI result produced by the updated version of Micromedex and Drugs.com at different time points might give different outcomes from our study, performed in the rst quarter of 2020. Additionally, the two databases have no data on several drugs used in animal hospitals, resulting in incomplete potential DDI analysis. Hence, some differences may occur once all drugs are added into the databases.
Finally, the results of all probable mechanisms of action in our study referred to humans, which might differ from animals due to dissimilar physiology and drug-metabolising enzyme systems [27]. Therefore, more study of potential DDIs in animals is recommended to improve medical care and decrease the possibility of DDIs as a result of multidrug therapy in animals.

Conclusions
Drug interaction databases showed highly variable performance in assessing the DDIs of veterinary drugs. Open-access resources, such as Drugs.com, could detect more potential DDIs. However, Micromedex, a subscription database, provided more supportive information and special features. The judgement of healthcare providers should be used, with the consent of animal owners, to determine appropriate treatments for animals and avoid potential DDIs by using several databases for the data evaluation.

Materials And Methods
Drug selection A list of 578 drugs was taken from those used in one animal hospital in Thailand and the VetList database [28], on 9 January 2020. From the total, 140 drugs were selected as frequently prescribed for the treatment of diseases in animal hospitals. Remarkably, Micromedex and Drugs.com could not recognise 44 of these drugs, so those remaining were used for this analysis, as shown in Fig. 2 and Table  4. The unrecognised 44 items were aditroprim, afoxolaner, avoparcin, baquiloprim, carprofen, clomocycline, dano oxacin, demethylchlortetracycline, deracoxib, di oxacin, enro oxacin, eprinomectin, pronil, rocoxib, glutathione, iba oxacin, imidacloprid, imidapril, ivermectin, levosimendan, limecycline, marbo oxacin, methacycline, methoprene, milbemycin oxime, moxidectin, oclacitinib, orbi oxacin, ormetoprim, pimobendan, pirlimycin, prado oxacin, rolitetracycline, samylin, spiramycin, sulfadimethoxine, sulfadoxine, sulfamethazine, teicoplanin, tepoxalin, tilmicosin, tulathromycin, tylosin and virginiamycin. Table 4 List of drugs used for the potential DDIs analysis from controlled studies; good -the interaction is strongly suspected but lacks well-controlled studies; fair -availability of documentation is poor but the interaction is suspected to exist on the basis of pharmacological considerations, or a pharmacologically related drug provides good documentation. All results of the potential DDIs in this study were obtained from searches in the two databases and gathered in January 2020.

Availability of data and materials
All data are presented in the manuscript.

Competing interests
Each author declares no competing interests.

Funding
No funding was obtained for this study.
Authors' contributions TB collected, analysed, and interpreted the data and drafted the manuscript. PK contributed to study design, data analysis, and revising the manuscript.
AK contributed to study design, data analysis, and revising the manuscript.
All authors have read and approved the nal version of the manuscript.

Figure 1
Result of the potential DDIs in each documentation level of Micromedex