Prevalence of potentially harmful multidrug interactions on medication lists of elderly ambulatory patients

Adverse drug reactions have been estimated by the FDA to be the fourth leading cause of death in the US resulting in costs of over $500 billion [11, 12]. Drug-drug interactions increase the likelihood of ADEs either because both drugs have the same target or effect, or one drug inhibits the metabolism of the other. It follows that the probability of patients experiencing ADEs would be further amplified if three or more drugs interacted with each other. We found that 1.3 % of the 6,545 medication lists of a cohort of elderly ambulatory patients contained MDIs. The prevalence of MDIs in medication lists was considerably higher in those that included one of a subset of medications; MDIs were identified in more than 20 % of the medication lists containing amiodarone or methotrexate and more than 10 % of lists containing bupropion, tramadol, trazodone, cyclobenzaprine, fluoxetine, ondansetron, or quetiapine (Table 2).

Medications for psychiatric conditions were the medications most frequently involved in MDIs, accounting for 35.1 % of all medications involved, a finding consistent with several prior studies of two-drug interacting pairs [13‐20]. Given the frequency of psychotropic medications in MDIs, the most common potential serious ADEs were those associated with psychotropic medications, namely prolonged QT, seizures, and serotonin syndrome [21‐23].

Nearly all MDIs could amplify the risk of patients experiencing associated ADEs. There are two predominant mechanisms of amplification. First, there are pharmacodynamic additive effects of medications such as prolonging the QT interval, lowering the seizure threshold, or causing the serotonin syndrome [23‐27]. Second, there were pharmacokinetic additive effects in which multiple interactions could result in steep increases in the concentration of participating medications resulting from CYP inhibition. Sutherland et al. found that 4 % of elderly patients in the cohort they studied were taking multiple inhibitors of the same CYP enzyme [4]. Psychotropic medications, which were the most common class of medications represented in MDIs from our study, operate by both mechanisms to amplify the risk of serotonin syndrome, seizures, and torsades de pointes [28, 29]. Patient F (Supplementary Table 1) is likely to have significantly elevated levels of the psychotropic mirtazapine resulting from the inhibition of its metabolism by both bupropion and fluoxetine and the inhibition of fluoxetine’s metabolism by bupropion [22, 28, 30‐32]. Age-related changes in pharmacokinetics due to diminished hepatic and renal function in the elderly make pharmacokinetic interactions especially likely [33].

Prior studies of DDIs found that between 6 % and 20 % of patients with DDIs identified in their records experienced an associated ADE [17, 20, 34‐36]. While there are as yet no studies to ascertain what proportion of patients exposed to MDIs experience an ADE, it is likely that the number will exceed that observed in two-drug interactions given the amplifying effect of additional interacting drugs. For example, patients taking warfarin and aspirin are more likely to have a hemorrhagic episode if they are also taking amiodarone. Polypharmacy has been associated with increased bleeding risk in patients on warfarin, even after adjusting for confounding factors [37]. We observed that the average number of medications on the lists of patients with MDIs was considerably higher (8.6 ±3.4) than those without (3.1 ± 3.4).

Prospective studies of ambulatory patients who experience ADEs due to multidrug interactions will be challenging because ADEs experienced outside of the hospital may be unaccounted for if the patient is admitted to another healthcare facility or suffers a lethal reaction. We note that an increased rate of sudden death associated with psychotropic medications has been observed [38‐45]. The FDA adverse event reporting system (FAERS) has recently been mined to identify multidrug interactions and the resulting ADEs actually experienced by patients [46]. While this approach cannot determine accurately the prevalence of ADEs associated with MDIs, it can establish that MDIs are associated with documented serious ADEs. In the absence of prospective studies, an examination of medication lists provides an opportunity to estimate the potential for amplification of the risk of ADEs.

In 2019, approximately 30 million CMS Part D Medicare beneficiaries were prescribed one of the top 15 medications involved in the MDIs we identified. Based on the prevalence we observed for MDIs in patients prescribed these drugs, approximately 3.5 million patients could have been exposed to amplifying MDIs (Table 2) [10]. Psychotropic medications were prescribed to nearly 16 million CMS Part D Medicare beneficiaries of whom 1.6 million were potentially exposed to amplifying MDIs. The use of amiodarone is particularly worrisome because, in addition to its effects on the lung, liver, and thyroid, it inhibits several key CYP enzymes (1A2, 2C9, 2D6, 3A4) and prolongs the QT interval [47, 48]. The number of CMS Part D Medicare beneficiaries prescribed amiodarone in 2019 was 706,029 of which 192,746 patients were potentially exposed to serious MDIs. Assuming that, on average, 10 % of patients with drug interactions experience the associated ADEs, approximately 20,000 patients taking amiodarone could have experienced a serious ADE [17, 20, 34‐36].

Most EHRs alert providers of potential drug-drug interactions by posting a warning of the potential ADE associated with the pair and the severity of the interaction. A patient prescribed warfarin, amiodarone and aspirin would trigger EHR warnings of two potential interactions, [amiodarone-warfarin] and [aspirin-warfarin], on separate lines along with the severity of each of the interactions. The amplifying effect of amiodarone on the warfarin-aspirin interaction would not be factored into the severity because there is no reported interaction between aspirin and amiodarone. Considering the amplifying effect of amiodarone on the warfarin-aspirin interaction, the combined severity of the drug interactions should probably be categorized as contraindicated. The Multum warnings of monitor closely for the [amiodarone-warfarin] pair and generally avoid for the [aspirin-warfarin] pair would underestimate the potential risk for hemorrhage (Patient D, Supplementary Table 1).

An EHR warning system that recognizes MDIs could be incorporated into the drug-drug interaction decision support but would likely result in alert fatigue that often results in providers overriding the alerts. Providers’ ignoring the alerts has been shown to increase the number of ADEs, at great cost to patients and the healthcare system [49, 50]. One solution would be to post the warning only when high-risk drugs such as psychotropics or amiodarone are involved. Until acceptable interventions are implemented, providers will need to look for multiple overlapping drug-drug interactions which share a common drug (and thus constitute an MDI).

Once potential MDIs are recognized, providers could substitute one of the interacting drugs with a more benign, non-interacting one [51]. Amiodarone could be replaced with propafenone which would eliminate the amplified risk of hemorrhage resulting from the warfarin-aspirin interaction. The FDA and the American Geriatrics Society 2019 updated the AGS Beers Criteria® for Potentially Inappropriate Medication Use in Older Adults recommend that amiodarone should not be a first-line choice [52, 53]. Ondansetron, which prolongs the QT, could be avoided in patients already on medications that prolong the QT if prescribed anti-emetics that are not associated with changes in the QT interval [54]. Tramadol, which acts as both an opiate and inhibitor of neurotransmitter uptake, exposes patients to multiple serious ADEs. Non-opiate analgesics could be substituted for tramadol and, if there was no indication for the psychotropic effect, no additional medication would be necessary. Substitution of methotrexate, an immunosuppressant prescribed for patients with various autoimmune diseases, would be difficult to replace given its unique role in treating patients with autoimmune diseases. Other strategies would have to be employed such as meticulous attention to those medications which interact with methotrexate to increase the risk of renal or hepatic injury.

Deprescribing rather than substitution may be a more effective strategy, especially for patients on psychotropics [55]. It is not uncommon for patients to be prescribed three or more psychotropic medications, likely reflecting the increasing prevalence of psychotropic polypharmacy [56]. Given the amplifying effect of psychotropic drugs, it is essential to assess the necessity and suitability of their use in elderly patients [57‐61].Trazadone, one of the most prescribed psychotropic medications participating in multidrug interactions (Table 2) and associated with serotonin syndrome and seizures, is one of the least efficacious medications in treating depression and is associated with a higher suicide rate than other psychotropic medications [62‐66]. Despite the warnings, nearly 3 million Medicare beneficiaries received prescriptions for trazodone in 2019 of which over 400,000 would be expected to be exposed to an MDI including that drug (Table 2). Cyclobenzaprine, despite the characterization by the AGS Beers Criteria® for Potentially Inappropriate Medication Use in Older Adults as “questionable,” is contraindicated in patients already taking drugs that prolong the QT, cause serotonin syndrome, or cause seizures [53]. Nevertheless, 1.8 million Medicare beneficiaries were prescribed cyclobenzaprine in 2017, potentially exposing recipients to increased risk of life-threatening ADEs. Although deprescribing these medications might be the best strategy, to do so would require primary care physicians to coordinate care with the specialists who originally prescribed the other medications which is often a significant challenge [67].

In 2015 Roughead warned that polypharmacy would inevitably result in multidrug interactions which would lead to an increased risk of ADEs [7]. Since that call to arms, however, there have been no studies prior to ours that identify MDIs. This may because doing so requires a more complex computational approach. To identify two-drug DDIs, drug pairs on the patients’ medication lists can be directly compared and matched to known two-drug interactions found in an enterprise EHR such as Cerner’s. Identification of MDIs, however, requires either network analysis or complex structured queries of the institutional database (Supplemental Fig. 1). Both approaches, however, are well known to clinical informaticians and could easily be applied in clinical studies. Perhaps our results will inspire other to identify and characterize MDIs in their institutions’ patient populations to expand the growing list of MDIs.

There are several limitations to this study. First, we did not take into consideration the doses of the medications that were involved in the MDIs. It is possible that patients were prescribed lower doses of a particular medication than recommended because a second or third drug could influence its metabolism, thus mitigating any potentially amplifying interaction. Second, we cannot say if the medications on the list were intended for daily use or only intermittent use, as needed. Aspirin, which appeared on many lists of patients with MDIs, could have been intended to be taken daily for cardioprotective effects or only as necessary for intermittent pain. Daily aspirin would pose greater risk of an ADE, such as gastric hemorrhage, than if taken only intermittently. Third, identification of three and four MDIs is based on the drug-drug interaction tables at our institution. The identification and rating of various interactions might differ across institutions given that there is significant variability in the drug-drug interaction databases [68‐73].