Background

Although information pertaining to drug-drug interactions (DDIs) constitutes a fundamental aspect of drug information, the provision of such information has proven, over the past decades, to be extremely lacking. DDIs information has traditionally been provided through tens of sources including DDIs dedicated textbooks, drug information databases and programs or brochures of drug manufacturers. These sources invariably provide DDIs data as part of monographs of single drugs or classes/groups of drug substances thus making the healthcare professionals responsible for the mundane task of identifying interactions (IAs) in cases where more than one drug is prescribed to their patients (this problem is accentuated with elderly subjects who would be prescribed many drugs on long-term basis).

The following paragraphs will discuss some of the shortcomings that would be experienced by healthcare professionals relying on the above said traditional sources to form an informed decision on DDIs:

  1. Incomplete and erroneous data, e.g.
    • the results of cross-referencing the American DDIs database (including USP-DI, AHFS, Facts and Comparisons and P. Hansten) against its European counterpart (including Martindale, Ivan Stockley, Nebenwirkungen und Wechsel Wirkungen and Dei Rote Liste) showed that more than 1,500 critical or clinically significant DDIs reported in the European databases are not acknowledged by any major drug information source in the United States or Canada;
    • Jankel and Martin (Jankel CA, Martin BC. Evaluation of six computerized drug interaction screening programs. Am J Hosp Pharm. 1992 Jun; 49(6):1430-1435) also evaluated six widely used commercially available drug IAs screening programs; they tested nine known drug interactions against each program or database and discovered that only two of the programs detected all nine interactions;
    • in a survey of 50 community pharmacies in the Washington, D.C. area, Hazlet et al concluded that the performance of most DDIs-detecting programs was suboptimal.
  2. Deficient information regarding induced drug effect.

  3. Deficient information or non-existent information on the effect of liver enzymes and transporters on DDIs. It could be readily verified that more than 75% of drugs are cleared from the body by liver enzymes. Hence, alteration of the activity of such enzymes may affect the disposition of drugs in the body (in their 2015 edition of Top 100 Drugs Interactions, P. Hansten and J. Horn list only thirteen enzymes despite the fact that close to sixty enzymes are currently documented by the FDA).

  4. Lack of standardized terminology for drug names. This is evident by the presence of about 4.9 names (on average) for any drug substance. Failure to standardize such names has prompted the WHO to publish the International Non-proprietary Names (INN) Glossary for drugs more than three decades ago. Notwithstanding, the vast majority of these names are still not acknowledged by most of the computerized DDIs applications.

  5. The inability of all existing databases to account for the impact of dosage forms or routes of administration on potential interactions. A starking example could be found in DrugBank regarding the formation of non-absorbable caused by Iron Dextran and other drugs (25 flaws).

  6. Editorial misjudgements when prioritizing or reporting IAs information. This alarming deficiency is clearly demonstrated in reporting IAs of Moclobemide by DrugBank (see Annex II) in two consecutive years. One IA was reported to have caused death in 2014 and was non-existent in 2015. The same applies to the anti-progesterone drug Mifepristone where 382 reports are currently provided by its online service versus 10 reports in the 2015. Out of the 382 reports, 126 pertain to the QTc-interval prolongation agents. Interestingly, 23 reports state that the hypoglycaemic effects of this drug are increased upon its concomitant administration with other hypoglycaemic agents!!

    Similarly, 197 interaction cases were reported for Mifepristone with other drugs in the year 2015 by Lexi-Comp online service. Out of these, 152 interactions were attributed to additive QTc-IP. These are presently reduced to 95 cases with only 14 interactions caused by the same effect.

  7. It is established that about 75% of drugs are eliminated from, or transported within, the body by liver enzymes and transporters. Hence, factors affecting the activity of these enzymes or transporters, by ways of inhibition or induction, may affect their kinetic and, consequently, their therapeutic characteristics. Hitherto, reporting of the implication of these enzymes by most databases is too generalized, incomplete or erroneous. The following paragraph on Omisertinib (a protein kinase inhibitor) has been exactly extracted from Lexi-Comp online database and is meant to substantiate the aforementioned assumption.

    Metabolism/Transport Effects

    Substrate of BCRP, CYP3A4 (minor), P-glycoprotein; Note: Assignment of Major/Minor substrate status based on clinically relevant drug interaction potential; Inhibits BCRP

    CYP1A2 Substrates: Osimertinib may decrease the serum conc. of CYP1A2 Substrates. Risk C: Monitor therapy

    CYP3A4 Substrates: Osimertinib may increase the serum conc. of CYP3A4 Substrates. Osimertinib may decrease the serum conc. of CYP3A4 Substrates. Risk C: Monitor therapy

    In addition to the generalized nature of this expression, the obvious flaws within such brief report may be attributed to deficient coding since the text appears to be code-generated. It could be argued that the discrepancies reported in Annex I represent an odd example that may not be generalized and should not be extrapolated to other databases. However, by ways of refuting this argument, a more exhaustive inter edition comparison has been undertaken between two different editions of the BNF, namely, versions 66 and 67. A sample of the results of this thorough evaluation is presented in an abbreviated table in Annex V.

    Another sticking example has been noted regarding the interactive profile of Diltiazem (a calcium channel blocker) and Moclobemide (a reversible MAO A inhibitor) as provided be the online service of DrugBank in two consecutive years (2014 and 2016). Details of this evaluation are provided in Annex I and I respectively. For Diltiazem, 53 and 152 interactions have been reported in 2014 and 2016 respectively. It may be assumed that such increase in the interaction report is expected. However, 30 interaction reports have been dropped from the current online version. These are marked with RED in Annex I. In addition, the description of some cases is reported in both years is considerably different. Similar inconsistencies exist in the case of Moclobemide.

    It is conceivable to notice growth in the content of some interaction reports as a result of increase in the size in of relevant information in the clinical literature. However, it would be difficult to account for the trimming of significant information from the previous edition (e.g. Imidazole Antifungals vs. Anticoagulants). This is related to the fact that once an interaction has been reported, it must have been based on sufficient evidence or sound clinical experience/judgement. Hence, its elimination from consequent edition becomes questionable.

    It would be more alarming to realise that reporting consistency in Stockley’s DDIs textbook. The table provided hereunder summarizes the discrepancy in the attribution of clinical significance rating, or advice to the clinical practioners, for 2,080 IAs report. It may be appreciated if the size of such reports has increased due to marketing new drugs. However, it is beyond comprehension to notice an alteration of the clinical significance rating in such an inconsistent manner.

    Summary of the discrepancy in the attribution of clinical significance rating, or advice to the clinical practioners, for 2,080 IAs report.
    No. of IAs Reports Old New
    Fatalities 166 94
    Contra-Indicated 144 516
    Life-Threatening 86 29
    Avoid Usage 620 425

  8. Design errors that fail to correctly capture the information stored in the database. This may be related to inadequate code and/or data structures or erroneous or rigid coding logic. Such deficiency is manifested in reporting dynamic interactions for classes or groups of drugs with other individual drugs. This has led to significant reporting redundancy in several databases such as Lexi-Comp, Dynamed and DrugBank. As demonstrated above, the 126 interactions reports involving Mifepristone and QTc-interval prolongating agents could be expressed by one record in a data which correctly employ extended search logic.

    With the advent of the IT era, most drug information databases tend to offer computerized systems which provide computer-generated drug interaction texts. The validity of such texts greatly depends on the efficiency of data and code structures that ensures correct classification of drugs. This is particularly relevant when and effect is attributed to classes or groups of drugs. Hence, the misplacement of a drug within a group would result in significant flaws and misleading information such as reporting 107 cases for Citric Acid and 115 cases for Adetic Acid identifying them as being anti-coagulants (refer to snapshots in Annex VII). Interestingly, 115 interactions for the anti-coagulant Anisindione, that were reported in 2014, have been dropped from the current online service of DrugBank. Additional details on such deficiencies are also provided in Annex VII.

The above said shortcomings directly impact healthcare professionals working in clinical setups that do not have the bandwidth or experience to research all traditional sources and databases / programs; the end result is the inability of said healthcare professionals to access all the relevant information required to make a completely informed decision in real-time. Within this context, DDI-Explorer has emerged with the promise of resolving all these concerns or shortcomings.

Discrepancies involving prodrugs and active metabolites

Reporting IAs involving represents a common deficiency in many databases wherein DDIs information pertaining to a prodrug or an active metabolite is independently reported by all sources. Prodrugs consist of the same drug substance that has been slightly modified for the purpose of enhancing its availability to systemic circulation. Typically, prodrugs break down or dissociate in the systemic circulation to produce its parent drug. This implies that both forms of such chemical entities should have the same interaction profile. In spite of the aforesaid, a drug or its prodrug is often differently reported by numerous DDIs databases (in this context, it suffices to cite the following cases: 1) Amprevavir / Fosamprenavir, 2) Amphetamine / Lisdesamphetamine, 3) Fosaprepitant to Aprepitant, 4) Phenytoin / Fosphenytoin). The pyrimidine analogue, Flourouracil, is reported by the DrugBank to be metabolized by12 different enzymes. Notwithstanding, two of its prodrugs, namely, Tegafur and Capecitabine, are reported to be metabolized by three and one enzymes respectively.

Annexes II &III provide examples of reporting discrepancies for drug prodrug of Phenytoin/ Fosphynetoin by DynaMed and Sirolimus/Temsirolimus by Lexi-Comp respectively.

In addition, the latest edition of the British National Formulary (BNF), published in 2015, only two prodrugs are recognized despite the fact that the number of such entities exceeds 45 cases.

Annex II provides a clear example of the magnitude of this deficiency. This also applies to active metabolites of many drug substances (typical cases include: 1) Primidone / Phenobarbitone, 2) Valacyclovit / Acyclovir, 3) Heroin / Morphine, 4) Enalapril / Enalaprilat and 5) Droxidopa / Adrenaline) where different interaction profiles are provided to a drug substance and its active metabolite.

Print