1. Introduction
Scientists, clinicians, and regulatory agencies have long recognized that adverse drug events (ADEs) are between the third and sixth most common cause of death worldwide [
1,
2,
3,
4]. The association of ADEs with substantial morbidity and mortality has resulted in mandatory phase IV clinical trials and black box warnings and the withdrawal of drugs from the market [
5]. Drug safety monitoring through pharmacovigilance studies must remain in effect after the approval and marketing of medications [
5,
6]. Drug safety monitoring is especially critical for side effects that are rare (e.g., drug-induced torsade de pointes), occur in certain sub-groups of the population (e.g., genetic predisposition due to drug metabolism or drug transporters), or recognized long after drug approval (e.g., thalidomide and congenital malformations) [
7,
8,
9].
On one hand, (i) published case reports, (ii) Medwatch (since 1993) linked to the FDA Adverse Drug Event Reporting System (FAERS), (iii) the emergence of information technology such as electronic health records (EHRs) since the 2009 Health Information Technology for Economic and Clinical Health (HITECH) Act, (iv) the Sentinel initiative, (v) data partnership networks, and (vi) consortia, such as PedsNet and the Open Health Data Science Informatics (OHDSI) network, have increased the capability to capture information about drug safety [
5]. On the other hand, some consider these resources antiquated and believe that applied statistical signal detection methodologies have limited usefulness [
1]. A major limitation of all of these listed approaches is that they require that events occur before they can be captured and analyzed, at which point safety-related warnings are added and negative reports are generated. In other words, some patients must experience side effects and potentially mortality before any safety-related changes occur. For new market entities, some safety label changes occur 10 years after drug approval [
10,
11]. Therefore, continued surveillance through a drug’s lifecycle after approval is critical to trigger safety-related drug label changes [
10,
11].
One such monitoring solution for drug safety is a medication-based risk score, which can be used to identify at-risk individuals who would benefit from medication management interventions, as well as reduce inappropriate polypharmacy, adverse health outcomes, and avoidable healthcare utilization [
2,
12,
13]. Various medication risk scoring systems, like indices, have been developed to quantify the complexity of an individual’s medication regimen, risk of medication-related falls, sedative load, anticholinergic burden, or drug regimen appropriateness [
12,
14,
15]. In the last 25 years, the authors have contributed to the development and evolution of clinical decision support systems (CDSSs) [
16], where each CDSS included a drug regimen risk score that could be used to foresee patient outcomes, such as ADEs, falls, hospitalization, emergency department (ED) visits, and death [
17,
18,
19,
20]. The authors have shown that medication-based risk scores influence pharmacists’ interventions and affect patients’ outcomes [
21,
22].
On 31 December 2019, the World Health Organization issued an epidemiological alert in response to an unidentified pneumonia in Wuhan, China [
23,
24]. Rapidly, SARS-CoV-2 spread globally, causing the coronavirus disease 2019 (COVID-19) pandemic. The scientific community sought potential treatments using repurposed drugs such as hydroxychloroquine, chloroquine (alone or in combination with azithromycin), lopinavir/ritonavir, ivermectin, ebselen, remdesivir, molnupivir, favipiravir, bebtelovimab, sotrovimab, and crizanlizumab, among others [
25]. Considering the known toxicity of some of these agents and their potential for drug–drug interactions, a new “pre-emptive pharmacovigilance” strategy was developed [
26]. This strategy was used to design polypharmacy biosimulation studies where repurposed drugs were virtually added to the real drug regimens of individuals and changes in medication risk scores were monitored to identify potential ADEs and negative outcomes, including an increased risk of hospitalization, ED visits, and medical expenditures [
27,
28].
A similar strategy is applied to our study reported herein, where a new chemical entity, iclepertin (BI-425809), was virtually added to the drug regimens of millions of individuals. As iclepertin is intended to be used in patients with schizophrenia, a subset analysis was performed in this population. Risk assessments of drug-induced torsade de pointes and CYP450 drug–drug interaction burden were the focus of our study.
2. Results
Data were obtained for
n = 4,435,330 individuals. Based on the criteria listed above, the following numbers of individuals were retained in each group: commercially insured = 1,937,389; Medicaid = 1,983,976; and Medicare = 483,698; of those,
n = 4,405,063 were retained for analyses. Baseline characteristics for the total population and across the Commercial, Medicaid, and Medicare individuals are presented in
Table 1. There was a larger proportion of women in the Medicaid group. Further, Medicare beneficiaries were significantly older when compared to the Commercial and Medicaid groups. Medicare beneficiaries also had the highest Charlson Comorbidity Index (CCI), followed by the Medicaid and Commercial populations. Medicaid individuals took more drugs on average than Medicare beneficiaries and both groups were taking more medications than Commercial individuals. The 60–69 years of age group in the Commercial and Medicaid populations were taking more drugs than their younger counterparts (
Table 2). The top 50 prescribed medications per coverage group are reported in
Supplementary Table S1.
Individuals in the Medicaid population had a higher number of ED visits than the two other groups (
Table 1). Further, hospitalizations were more common in the Medicaid and Medicare groups, while hospital length of stay (LOS) showed a similar distribution among the three groups. Falls were more frequent in Medicare beneficiaries. The MRS was higher in Medicaid individuals, followed by Medicare beneficiaries, and more Medicaid individuals were in the Severe category when compared to the two other groups (
Figure 1). Medicaid beneficiaries also showed a higher risk for CYP450 drug interactions and an increased risk for drug-induced long QT syndrome (LQTS;
Table 1).
Table 3 reports our findings following the addition of iclepertin to the drug regimen of individuals from the different coverage groups. As no information was available for iclepertin in the FAERS, the combined effect of this parameter on the MRS could not be estimated. Clinical data for the anticholinergic properties and sedative load characteristics of iclepertin were also not available. A change in anticholinergic burden or sedative load is not expected based on iclepertin drug disposition characteristics (weak CYP3A4 substrate) and the low likelihood of iclepertin to cause a change in the exposure of other drugs. Note, the addition of iclepertin was associated with a slight increase (0.85 to 0.88 points; considered not significant) in the MRS on average. Nevertheless, an increase in the MRS was observed in about 50% of the population. Some individuals had an increase of up to 11 points; the various degrees of increase in the MRS led to changes in the distribution of individuals in each of the MRS categories (
Figure 1). For all three groups, fewer individuals were in the “Minimal” category following the addition of iclepertin, and more individuals were distributed towards higher risk categories (
Table 3,
Figure 1).
More discrete analyses were performed for both the CYP450 drug interaction burden and LQTS score as these two factors were significantly modified by CYP3A4 inhibitors and strong CYP3A4 affinity substrates: as iclepertin is expected to exhibit a weak affinity towards CYP3A4, it would behave like a victim drug.
Table 4 reports on the detailed analyses performed. The presence of CYP3A4 inhibitors and strong affinity substrates was associated with higher MRS at baseline in all populations tested. The list of clinically relevant perpetrator interacting drugs including CYP3A4 inhibitors, CYP3A4 inducers, and stronger affinity CYP3A4 competitive substrates are listed in
Supplementary Table S2.
The addition of iclepertin increased the MRS by 0.8 points in the Medicare population, 1.6 points in the Medicaid population, and 1.9 points in the Commercial population. As mentioned previously, this increase was mostly explained by the CYP450 drug interaction burden parameter.
Similarly, the presence of concomitant CYP3A4 inhibitors and CYP3A4 strong affinity substrates was associated with a higher LQTS risk score at baseline. The addition of iclepertin did not significantly modify the LQTS score, except for in some specific individuals (
Table 5). Overall,
n = 139 individuals had an increase in their LQTS score of 2 points and above, 123 had an increase in the LQTS score of 3 points and above, and 113 had an increase in their LQTS score of 5 points and above. An increase of 5 points and above was mostly observed in the Medicaid population.
Iclepertin is a glycine transporter 1 (GlyT1) inhibitor intended to be used for the treatment of cognitive impairment associated with schizophrenia. A sub-analysis was performed in individuals with listed ICD-10 codes associated with schizophrenia (
Table 6).
n = 123,722 individuals met the required conditions (ICD-10 codes of F-20 to F-29). In all sub-groups (Commercial, Medicaid, or Medicare), schizophrenia was associated with a much higher MRS (6.8 points on average). Changes in MRS produced by the addition of iclepertin in the schizophrenia population were of the same magnitude as those observed in the control population. Further, individuals in the schizophrenia population had higher CYP450 drug interaction burden score and higher drug-induced LQTS score than the control population. Changes in these parameters produced by the addition of iclepertin were also of the same magnitude as changes observed in the control population.
A pharmacoeconomic evaluation of the effect of adding iclepertin to the actual drug regimens of individuals on medical expenditures, ED visits, and hospitalizations based on the computed increase in MRS is presented in
Table 7. Overall, the increase in MRS (0.85 to 0.88) was associated with a postulated modest increase in average medical expenditure over the entire cohort (USD 91 on average). The model also predicted a 0.01% to 0.03% increase in ED visits and a 0.004% to 0.006% increase in hospitalizations.
Table 8 reports the impact of CYP3A4 inhibitors and CYP3A4 strong affinity substrates on the increases in medical expenditures, ED visits, and hospitalizations.
The effect of perpetrator CYP3A4 inhibitors and CYP3A4 strong affinity substrates was observed both at baseline and after the addition of iclepertin.
Table 9 reports on the postulated impact of iclepertin on medical expenditures, ED visits, and hospitalizations in individuals with schizophrenia. Changes observed were of the same magnitude as those observed in the non-schizophrenia population. However, the benefits associated with the use of iclepertin in individuals with schizophrenia were not considered in our pharmacoeconomic evaluation.
3. Discussion
In this study, we determined the safety element profile of a new chemical entity, iclepertin, by using real-world claims data and virtually adding iclepertin to the actual drug regimens of over 4 million individuals. Based on some pharmacokinetic (CYP3A4 weak affinity substrate with partial metabolic clearance of 80%) and pharmacodynamic (IC50 for block of IKr) properties, we estimated how the virtual intake of this drug would expose subjects to significant multidrug interactions and potential side effects. Relevant side effect frequency for iclepertin is absent from the FAERS, and therefore, our approach did not estimate any benefits associated with iclepertin or consider other potential side effects. However, we demonstrated that the safety profile of iclepertin was similar in the targeted population of individuals with schizophrenia compared to the non-schizophrenia population. This approach represents a new, pre-emptive, polypharmacy biosimulation strategy that adds to the pharmacovigilance armamentarium; this science is proactive rather than reactive. More importantly, information is obtained without exposing any individuals to drugs and potential side effects, including death.
A change in MRS was observed following the addition of iclepertin in about 50% of the population (
n = 2,138,247). The percentages of individuals with a change in MRS and the mean increases in MRS were similar between the three tested groups (47.3–50.6% change in MRS and percent increase of 0.85–0.88 units, respectively). The MRS was classified into Minimal, Low, Intermediate, High, and Severe categories that have been associated with health outcomes [
18,
19,
22,
29]. It was previously demonstrated that Intermediate and High/Severe categories are associated with increased risk of poor health outcomes, including ADEs, fall, death, and medical expenditures, compared to the Minimal category [
18,
19]. When looking at individuals with a change in their MRS category from the baseline after the addition of iclepertin, the simulation predicted that 3.33% of the population would have a change in MRS leading to a higher risk category. The number of individuals classified in the High/Severe MRS category increased by 0.33% in the overall population. By sub-group, an increase of
n = 4070 (0.21%), 9201 (0.46%), and 1256 (0.26%) individuals was estimated in the High/Severe MRS category for the Commercial, Medicaid, and Medicare groups, respectively.
In this study, changes observed in MRS secondary to the virtual introduction of iclepertin to the drug regimen of commercially insured and Medicare or Medicaid beneficiaries were 3 to 10 times less than those observed in similar populations from our previous studies [
28]. Previously, the virtual addition of drugs such as hydroxychloroquine or chloroquine (alone or with azithromycin) or lopinavir/ritonavir to the drug regimen of commercially insured and Medicare beneficiaries during the COVID-19 outbreak led to more significant increases in the MRS [
28]. In the later study, changes in MRS were mostly due to an increase in CYP450 drug interaction and drug-induced LQTS risk indices. In another study, simulations performed in a group of participants in the Program for All-inclusive Care of the Elderly (PACE) with the same COVID-19 repurposed drugs led to similar results [
27].
Based on available data, iclepertin was considered a weak affinity substrate for CYP3A4. Therefore, iclepertin systemic exposure can be affected if administered with CYP3A4 inhibitors, CYP3A4 inducers, or CYP3A4 higher affinity substrates (i.e., other drugs becoming competitive substrate inhibitors). This study looked at relevant CYP3A4 interactions, as most drugs are metabolized by CYP450 enzymes, driving the effect on drug interaction burden [
30]. Across the three populations, atorvastatin was the most common clinically relevant concomitant CYP3A4 interacting drug, followed by omeprazole, amlodipine, simvastatin, buspirone, doxycycline, topiramate, buprenorphine, fluconazole, and risperidone. Following the addition of iclepertin, the CYP450 interaction burden score was estimated to increase by approximately one unit for the overall population (
Table 1 and
Table 3). The simulation indicates that a similar magnitude of CYP450 interaction burden score change is expected among the various groups tested.
Our model predicted a change in drug-induced LQTS score in 0.0032% of the population (
n = 139). Overall, changes in the LQTS score were minimal. LQTS scores were classified into Low-, Moderate-, and High-risk groups for QT prolongation. Based on our model, the risk of experiencing torsade de pointes is associated with the High-risk group [
26]. Following the addition of iclepertin, no individual who had an increase in their LQTS score became at high risk of having QT prolongation and torsade de pointes;
n = 137 remained as low risk and two individuals moved from low to moderate risk. Notably, women are at increased risk of torsade de pointes [
31,
32], and the current simulation did not find a higher proportion of women in the increased LQTS score group. An increased risk for women was previously demonstrated in simulations studies of known QT prolongation drugs [
27,
28]. This suggests that QT prolongation risk of iclepertin is likely negligible.
Across all populations, the average MRS increased from 6.3 ± 6.6 to 7.2 + 6.6 following the virtual addition of iclepertin to individuals’ drug regimens. Further, 1,382,567 individuals had an MRS increase of 2 units, 72,506 individuals had an increase of 5 units, 375 individuals had an increase of 10 units, and 5 individuals had an increase of 11 units. A previously published model, trained using claims data, was used for the current simulation [
18]. When examining medical expenditures, ED visits, and hospitalizations based on an increase in MRS following the virtual addition of iclepertin, the updated model predicted a USD 91 increase per individual (USD 3172 to USD 3263) for the total population. The Commercial population was predicted to have a USD 131 increase in cost, while the Medicaid and Medicare populations had a USD 62 and USD 137 predicted increase in cost, respectively. The pharmacoeconomic estimation performed is biased at this stage as it does not consider beneficial effects and potential savings associated with drug efficacy.
Iclepertin was developed as a potent and selective glycine 1 transporter (GlyT1) inhibitor to improve symptoms of cognitive impairment associated with schizophrenia [
33]. Sub-analyses were conducted in individuals with a schizophrenia diagnosis based on ICD-10 codes to compare changes in MRS in this population versus the non-schizophrenia population. Our results demonstrated that changes in MRS, in CYP450 interaction burden, and drug-induced long QT syndrome indices did not differ between subjects with schizophrenia vs. without known schizophrenia. As we have observed that individuals with schizophrenia in the three populations tested had about a 2-fold increase in their MRS compared to the non-schizophrenia population, these findings are important.
Previously, Michaud et al. [
18] explored the association of the MRS and health outcomes in a large Medicare population. Study results showed that a 1-unit change in the MRS was associated with an 8.5% increase in total medical expenditures (Part A and B); therefore, a 2-unit change in MRS would be associated with an increase of 17.7%, and a 10-unit change in MRS with 126% increase in medical costs [
18]. Based on the current model developed for the total population (three groups), a 1-unit change in MRS was associated with an increase of 4% in medical costs, which would represent an increase in medical expenditures of 8.1% and 47.3% for individuals with a predicted increase of 2 and 10 units of the MRS, respectively. In the current simulation, the MRS was derived using the most recent drug claims in 2019 (with retroactive drug claim overlap), while the published model used the maximum MRS generated over the year [
18]. Therefore, the calculated MRS and, consequently, the impact on medical expenditures may be underestimated for some individuals.
The previously developed model by Michaud et al. [
18] found that per 1-unit increase of the MRS per year, the number of expected ED visits increased by 7%, and the number of hospital admissions was predicted to increase by 3%. The current model trained for iclepertin drug simulation estimated a 4.2% increase in ED visits and a 5.8% increase in hospital admission per 1-unit increase in MRS, comparatively. The previously published model estimated a 5.8% increase in odds ratio of having an ADE per increase in MRS unit. Michaud et al. [
18] also reported that a 2-unit increase in the MRS (as estimated for 65% of individuals in this simulation) could translate into a 12% increase in odds ratio for experiencing at least one ADE. Therefore, an increase of up to 86% in the odds ratio of having an ADE is estimated for an increase of 11 units of the MRS, which is the highest increase in the MRS observed in this simulation.
As mentioned, relevant side effect frequency for iclepertin is absent from the FAERS, and therefore our approach did not estimate benefits associated with iclepertin or consider other potential side effects. This is a significant limitation as only two out of five factors included in the MRS could be assessed. Further, using real-world data presents several advantages, including the acquisition and analysis of data on many individuals in a short time. However, real-world data often requires significant clean-up, is static, and is not always uniform between all individuals, which can be a limitation when needing specific inclusion criteria to perform analyses.
Finally, the presence of CYP3A4 inhibitors or CYP3A4 strong affinity substrates was the most important parameter driving changes in medical expenditures, ED visits, and hospitalizations, both at baseline and following the addition of iclepertin; this observation remained true in the schizophrenia population.