The Interplay Between Molecular Architecture, Pharmacology, and Suspected Adverse Drug Reactions Associated with Non-Steroidal Androgen Antagonists in The United Kingdom
School of Pharmacy, University of Birmingham, Edgbaston, B15 2TT, United Kingdom
ProCURE, Catalan Institute of Oncology (ICO) & Oncobell, Bellvitge Institute for Biomedical Research (IDIBELL), Barcelona, Catalonia, Spain
Center for Cancer Drug Discovery, Division of Cancer Therapeutics, The Institute of Cancer Research, London SM2 5NG, UK
*Corresponding author: a.m.jones.2@bham.ac.ukAbstract
Aims
To correlate potential links between the suspected adverse drug reaction (ADR) profile of licensed non-steroidal androgen receptor antagonists (NSARA) with their unique chemical properties and known off-target polypharmacology.
Methods
Physiochemical and polypharmacology data was curated from the Electronic Medicines Compendium, FDA New Drug Applications documents, and ChEMBL databases. System organ class (SOC, MedDRA) suspected ADRs and fatalities were curated from the United Kingdom Medicines and Healthcare products Regulatory Authority (MHRA) Yellow card spontaneous reporting scheme for their respective prescribing period; apalutamide (Jan 2019-), bicalutamide (Aug 2018-), enzalutamide (Aug 2018-), flutamide (Aug 2018-) and darolutamide (March 2019-) until Oct 2023. The number of daily doses (dd) was extracted from OpenPrescribing and NHS Digital secondary care medicines data. Data was standardised before comparison to suspected ADRs and fatality reports per 100,000 dd.
Results
A total of n = 2,480 suspected ADRs were associated with 42,903,000 dd of NSARAs in the United Kingdom. The highest number of ADRs were associated with enzalutamide (n = 1,091) and bicalutamide (n = 749). Enzalutamide was found to have the most off-target pharmacological interactions of the NSARAs studied (n = 5) including potent inhibition of γ-aminobutyric acid, GABA receptor (IC50 = 2.6 µM vs Cmax = 7.7 µM) associated with nervous system disorders (n = 72, accounting for 73% of all NSARA ADRs in this SOC). Apalutamide, the only other GABA inhibitor (IC50 = 3 µM vs Cmax = 2.9 µM) had the highest relative rate of suspected nervous system ADRs at 1.08 per 100,000 dd.
Apalutamide was also a modest inhibitor of the human Ether-à-go-go-Related Gene (hERG) ion channel (IC50 = 6 µM vs Cmax = 2.9 µM) and had the highest rate of suspected cardiac arrhythmia ADRs, 30-fold over, enzalutamide, a significantly weaker hERG inhibitor (15.7 µM vs Cmax = 7.7 µM).
Darolutamide was the only NSARA to show effects at 5-HT (serotonin) receptor at < 10 µM but did not translate to psychiatric disorders due to low clinical BBB penetration but a an association with hepatobiliary and cardiac disorders was identified based on this inhibitory axis. Suspected skin and subcutaneous SOC ADRs was associated with all NSARAs (except flutamide) but did not reach statistical significance (P = .25). A rationale for epidermis reactions relating to apalutamide containing a masked arylamine was explored but molecular matched pair (MMP) analysis with enzalutamide suggests it may not be a chemical cause. Statistical significance (P < .05) was identified in reported fatalities associated with NSARAs, flutamide had n = 24 or 897.5 fatalities per 100,000 dd which was likely due to both the indication and the small number of dd (n = 3,000) during the time period of the study.
Conclusions
An investigation of suspected ADRs, standardised to the number of dd for the novel NSARA drug class identified SOCs of potential interest. The highest number of reports related to enzalutamide and bicalutamide. Suspected skin and subcutaneous ADRs approached statistical significance and was interrogated for chemical and pharmacological connections for the first time with the aid of MMP analysis. A potential correlation to nervous system disorders and cardiac arrhythmia for the GABA and hERG inhibitors, enzalutamide and apalutamide, respectively was identified. Darolutamide’s interaction with 5-HT may influence ADRs associated with cardiac and hepatobiliary SOCs. Statistically significant number of suspected fatalities with flutamide was identified.
Article notes
Competing Interest Statement
The authors have declared no competing interest.
Funding Statement
This study did not receive any funding
Introduction
In the United Kingdom, 52,300 men per year are diagnosed with prostate cancer, [1] making it the most common cancer with peak diagnosis between 70-74 years old. [1] Hormone therapy is recommended as the first-line treatment strategy by the National Institute for Health and Care Excellence (NICE). [2-3] Androgen receptor antagonists (ARA) inhibit the exertion of androgen-like testosterone and dihydrotestosterone (DHT) pharmacological effects by competitively inhibiting testosterone binding to ARs [4]. Steroidal ARAs treat a range of conditions including sexual deviation and hot flushes with gonadorelin analogue therapy. [5]
Non-steroidal ARAs (NSARAs) are specifically used for the treatment of prostate cancer. NSARAs target prostate cancer cells by inhibiting androgen production, therefore, decreasing prostate cancer cell growth. [6] NSARAs can be either first generation (bicalutamide, flutamide) which exclusively target AR translocation to the nucleus or second generation (enzalutamide, apalutamide, darolutamide) which have further improvements on this mechanism of action (MoA). [7-8] Monitoring adverse drug reactions (ADRs) reported to the MHRA Yellow Card Scheme in the United Kingdom [9] is important as 1 in 16 patients admitted to hospital are suspected to be experiencing an ADR [10]. The ability to determine correlation of NSARAs ADRs to molecular structure and/or off-target pharmacological factors is therefore timely. [11-18]
Methods
The licensed first generation NSARAs are bicalutamide [19] and flutamide [20]; and second generation, enzalutamide [21], apalutamide [22], darolutamide [23] (Figure 1) were selected based on inclusion and exclusion criteria of this study (Table 1).
Prescribing Data
Drug pharmacology
NSARA’s molecular and physiochemical properties were extracted from the Chemical database of bioactive molecules with drug-like properties, European Molecular Biology Laboratory (ChEMBL), [26] FDA New Drug Application (NDA) documents, [27-30] and the Electronics Medicines Compendium (EMC). [31] Dosing regimens were extracted from the British National Formulary (BNF). [32]
The following parameters were calculated: pIC50 was calculated from the negative log of the average AR IC50 data. Lipophilic ligand efficiency (LLE) was calculated using the following equation: pIC50 – log10P. [33]. The predicted blood-brain barrier (BBB) penetration requirements were interrogated as follows: molecular weight (MW) <450 Da, <6 hydrogen bond donors (HBD), <2 hydrogen bond acceptor (HBA), a neutral or basic drug (indicated by pKa), topological polar surface area (tPSA) <90Å, log10D (at pH 7.4) between 1 to 3 and a low p-glycoprotein affinity. The Cmax was converted from ng/mL to nM based on the unique molecular weight of each NSARA.
Adverse drug reaction reporting
Suspected adverse drug reactions (ADRs) of all United Kingdom licensed NSARAs were extracted from the MHRA Yellow Card interactive drug analysis profile (iDAPs). [9] Suspected ADRs reported for apalutamide (since Jan 2019), bicalutamide (since Aug 2018), enzalutamide (since Aug 2018), flutamide (since Aug 2018), and darolutamide (since Mar 2019) with data extracted corresponding to the longitudinal timeline of each NSARAs prescribing in the United Kingdom.
Standardised NSARA ADRs were determined by calculating the ADR incidence rate per 100,000 daily doses (dd) for each drug:
- Total tablets quantity dispensed divided by the respective standard of NSARA daily number of tablets = total dd.
- Scale factor: Total dd dispensed / 100,000.
- ADRs multiplied by the scale factor gave the ADRs per 100,000 dd of the NSARAs.
Target Affinity
ChEMBL was used to extract the target proteins for each NSARA (accessed 29th November 2023). [26] Where > 1 IC50 for a human protein target was available, the mean IC50 value was calcualated.
Ethical Approval
Ethical approval was not required by the School of Pharmacy sub-ethics committee due to the use of fully anonymised patient data.
Statistical Analysis
Excel for Microsoft 365 was used to perform chi-squared (χ2) tests on the standardised ADR’s/100,000 dd data. A P < .05 was set for statistical significance-.
Results
Physicochemical properties and pharmacokinetics
Properties of the NSARAs (clog10 P, pIC50, LLE) are shown in Table 2. Enzalutamide was the most lipophilic drug (clog10P = 3.99) and darolutamide was the most potent AR antagonist (pIC50 = 7.59). All AR antagonists had an LLE < 5. NSARAs that comply with the BBB criteria: bicalutamide, flutamide and darolutamide successfully passed the MW requirements; with apalutamide being the heaviest drug (477 Da). All the NSARA’s failed the guidelines for number of HBAs by having >2 HBA’s, however, passed the HBD criteria (<6). Apalutamide, enzalutamide and darolutamide are weak bases and bicalutamide and flutamide are neutral. Apalutamide, enzalutamide and flutamide had a tPSA < 90 Å. Bicalutamide and darolutamide met the criteria for log10D7.4 (2.71 and 2.44, respectively). Enzalutamide and flutamide were not P-glycoprotein substrates. Overall, flutamide passed the most BBB requirements out of all AR antagonists (achieving 5/7 BBB requirements), whilst apalutamide met the least (3/7 conditions).
In clinical settings, darolutamide did not significantly alter cerebral blood flow (CBF), consistent with its low clinical BBB penetration and subsequent low risk of CNS-related adverse events. [34] However, a significant reduction in CBF was observed with enzalutamide suggestive of increased BBB penetration. [34] This was further supported by cerebrospinal fluid (CSF) for enzalutamide in rats showing a 3-6% penetration. [35] CSF measurement in dogs for apalutamide show a 2-5% penetration, [36] A single case report for CSF leak with flutamide was identified [37] but no clinical data was identified for bicalutamide.
Flutamide required the most daily tablets; with the regime consisting of two tablets, three times a day due to the short half-life (10 h). Apalutamide had the largest volume of distribution (276 L/kg) with bicalutamide and flutamide having undetermined volumes of distribution (Vd).
Target affinity
Table 3 shows the strength of inhibition assessed by using the mean IC50 values. Apalutamide and darolutamide hadthe most potent inhibition of the AR receptors, with IC50 = 63 nM and 26 nM, respectively. Bicalutamide and flutamide had the weakest AR antagonist inhibition (1.2 and 1.3 µM, respectively). Bicalutamide had additional interactions with progesterone receptors at clinically achievable levels (IC50 = 1.8 µM vs Cmax = 1.7 µM). Enzalutamide was found to have the most off-target pharmacological interactions of the NSARAs studied (n = 5) including potent inhibition of γ-aminobutyric acid, GABA receptor (IC50 = 2.6 µM vs Cmax = 7.7 µM). Apalutamide, was the only other GABA inhibitor (IC50 = 3 µM vs Cmax = 2.9 µM) and a modest inhibitor of the human Ether-à-go-go-Related Gene (hERG) ion channel (IC50 = 6 µM vs Cmax = 2.9 µM). Enzalutamide was a significantly weaker hERG inhibitor (15.7 µM vs Cmax = 7.7 µM). Furthermore, darolutamide was the only NSARA to show effects at 5-HT (serotonin) receptor at < 10 µM.
Total ADR’s and fatalities
ADR profiles of the NSARAs were standardised based on the number of dd dispensed in the time period of the study (Table 4). Overall, flutamide had the largest number of ADR’s/100,000 dd (15.8); followed by darolutamide (15.0) with enzalutamide the least (3.1). Flutamide had the largest number of fatalities/100,000 dd (n = 24 (897.5 per 100,000 dd)). Bicalutamide had the next highest at 0.4 deaths per 100,000 dd. Darolutamide was the only NSARA to have no reported deaths. For the selected SOC based on relative proportions of suspected ADRs and > 100,000 dd during the time period are shown in Figure 2.
Blood and lymphatic system disorder ADR’s and fatalities
Darolutamide and apalutamide had the highest reports of 0.44 and 0.24 ADRs per 100,000 dd, respectively. Patients taking darolutamide experienced suspected neutropenia (0.44/100,000 dd) and apalutamide experiencing suspected thrombocytopenia (0.12/100,000 dd).
Cardiac disorders ADR’s and fatalities
Darolutamide had the highest suspected cardiac disorder profile (0.89/100,000 dd), apalutamide had the second highest ADRs/100,000 dd (0.78), along with being the only AR antagonist to have a suspected fatality in this category. The main cardiac event possessed by apalutamide patients were cardiac arrhythmias and heart failure, 0.3 and 0.36 per 100,000 dd respectively. Whilst enzalutamide exhibited the least cardiac events (n =12, 0.03 per 100,000 dd).
General disorders and administration site conditions ADR’s and fatalities
General disorders and administration site conditions had the highest number of ADRs across all AR antagonists. Enzalutamide had a fatality rate of 0.8 per 100,000 dd. Apalutamide also experienced a high prevalence of general disorders side effects (2.04/100,000 dd) and fatalities (0.06/100,000 dd). Darolutamide followed with 1.78 ADRs/100,000 dd.
Nervous system ADR’s and fatalities
Similarly, to general disorder ADR’s, apalutamide (1.08/100,000 dd) and darolutamide (0.44/100,000 dd) had the highest nervous system ADRs. Apalutamide’s main nervous system ADR was neurological disorders (0.48/100,000 dd).
Psychiatric disorders ADR’s and fatalities
Darolutamide had 0.44 ADRs per 100,000 dd, which was the highest number of suspected events for the NSARA series. Patients on enzalutamide experienced 0.1 suspected psychiatric reactions /100,000 dd.
Hepatobiliary disorders ADR’s and fatalities
The highest suspected ADRs of patients with hepatobiliary disorders were experience with flutamide (37.4 per 100,000 dd) but may be an artefact of the very low number of dd in this time period.
Skin and subcutaneous tissue disorder ADR’s and fatalities
Similar to general disorders, skin and subcutaneous disorders had the largest number of suspected ADRs/100,000 dd. Darolutamide (2.67/100,000 dd) and apalutamide (2.4/100,000 dd) had the vast number of ADRs, with all darolutamide patients experiencing epidermal ADR’s and 2.16 ADR’s/100,000 dd for apalutamide.
Age range of NSARA ADR reports
Apalutamide consisted of 72 patients who experienced a suspected ADR between the ages of 50-89 years old. Bicalutamide had 2 patients between 0-19 years old, with the remaining 35 patients between 60-99 years old. Enzalutamide consisted of 140 patients with side effects between 50-99 years old. Flutamide’s single report from the selected SOCs was from a male patient that was between 80-89 years old; and darolutamide having 14 patients between the ages of 60-89.
Molecular matched pair (MMP) analysis was possible between apalutamide versus enzalutamide (Figure 1) due to a single point variation present - gem-dimethyl versus spirocyclobutane, respectively on the thiohydantoin scaffold. This led to a logarithmic difference in activity at the AR (63 vs 806 nM), and a 2.5-fold difference in hERG activity (6 vs 15.7 µM), comparable inhibition at the GABA receptor, and weak inhibition at glucocorticoid and progesterone receptors, respectively.
Discussion
Total ADR’s
The highest number of total ADRs reported was with enzalutamide (n = 1091), when standardised to the number of dd, flutamide (n = 423, 15.82 ADRs per 100,000 dd) and darolutamide (n = 35, 15.55 ADRs per 100,000 dd) emerged as have a higher suspected incidence rate. Flutamide’s data was excluded when comparing individual SOC ADRs per 100,000 dd as there was limited prescribing during the time frame of this study (3,000 dd) vs all other NSARAs 42,900,000 dd.
Total fatalities
Enzalutamide emerged as having the highest number of fatalities (n = 50), when standardised to the number of dd, flutamide (n = 23, 897.5 fatalities per 100,000 dd) emerged as having a higher fatality rate. Darolutamide had zero reported fatalities during the time period of this study.
Blood and lymphatic systems disorders ADR’s and fatalities
Darolutamide had a prevalence of 0.4 ADR’s/100,000 dd with one report of neutropenia. Apalutamide followed with 0.24 ADR’s/100,000 dd, with reports of thrombocytopenia. Interactions between apalutamide and CYP2C8 and CYP2C9 can cause drug-drug interactions (DDI),. Enzalutamide patients experiencing both (immune thrombocytopenia and neutropenia) again possibly due to the DDI’s with CYP2C8 and CYP2C9. Patients taking flutamide experienced no blood and lymphatic disorder ADR’s. None of the NSARAs studied were suspected of inducing a blood and lymphatic system fatality.
Cardiac disorders ADR’s and fatalities
Darolutamide had the greatest prevalence of ADRs for cardiac disorders. The increase in cardiovascular events on darolutamide could be potentially ascribed to SERT inhibition, given its importance in the cardiovascular system.[38] Furthermore, darolutamide is a known CYP3A4 inhibitor related to potential DDIs. Prostate cancer patients are within the same age group to have cardiac co-morbidities. Cardiac medications are known CYP3A4 substrates (e.g. simvastatin/ atorvastatin [39]). Additionally, p-gp modulation causes disturbances in ion efflux which can be of concern to those taking ion channel inhibitors, such as verapamil and amiodarone. [40] P-glycoprotein inhibitors can be proton-pump inhibitors (e.g. omeprazole) and macrolides antibiotics (e.g. clarithromycin & erythromycin). The combination of these factors could have a detrimental impact on the patients’ health by increasing the risk for a cardiovascular event that is highly dependent on polypharmacy in this age group.
Apalutamide was a modest inhibitor of the human Ether-à-go-go-Related Gene (hERG) ion channel (IC50 = 6 µM vs Cmax = 2.9 µM) and had the highest rate of suspected cardiac arrhythmia ADRs, 30-fold over the only other NSARA with this suspicion, enzalutamide, a significantly weaker hERG inhibitor (15.7 µM vs Cmax = 7.7 µM). Apalutamide (0.36 ADR’s /100,000 dd) also interacts with CYP2C8, CYP2C9, CYP2C19 and P-gp. Inducing CYP2C19 can cause patients taking clopidogrel increase metabolism to its active metabolite, which has a half-life of approx. 30 minutes. This causes the therapeutic effect to diminish quickly after administration, making the patient at risk for an ischemic attack again dependent on polypharmacy confounders not available in these datasets.
Enzalutamide is also known to interact with CYP2C8, CYP2C9, CYP2C19, CYP3A4. Bicalutamide is a CYP3A4 inhibitor and p-gp inducer, cardiac events are potentially due to DDIs. Flutamide had no cardiac ADRs and no reported cytochrome P450 or p-gp interactions in contrast.
General disorders and administration site conditions ADR’s and fatalities
A higher prevalence of ADR’s was observed with enzalutamide, apalutamide and darolutamide for general disorders and administration site conditions. Apalutamide’s interaction with CYP3A4, CYP2C19 and p-gp can induce DDI with other drugs that utilise their metabolic pathways. Darolutamide and bicalutamide affects CYP3A4 and p-gp, whilst enzalutamide interacts with CYP3A4. Enzalutamide had 0.08 fatalities/100,000 dd, followed by apalutamide (0.06).
Nervous system ADR’s and fatalities
Effects of AR on CNS and cognitive function have been reported.[41] Apalutamide and darolutamide had 1.08 and 0.44 nervous system ADRs/100,000 dd, respectively. These higher reports are potentially due to their unique pharmacology. Inhibition of GABA receptor can lead to increased anxiety, stress, and fear responses and apalutamide, a GABA inhibitor (IC50 = 3 µM vs Cmax = 2.9 µM) had the highest relative rate of suspected nervous system ADRs at 1.08 per 100,000 dd. BIn clinical settings, darolutamide did not significantly alter cerebral blood flow (CBF), consistent with its low clinical and predicted BBB penetration and subsequent low risk of CNS-related adverse events. [34] In contrast, CSF measurement in dogs for apalutamide show a 2-5% penetration which may inform patient stratification. [36]
Psychiatric disorders ADR’s and fatalities
Darolutamide and apalutamide led with the highest number of psychiatric disorders per 100,000 dd (0.44 and 0.33, respectively). Darolutamide was identified as the sole NSARA to show effects at 5-HT (serotonin) receptor at < 10 µM. 5-HT has a well-known role in depression, obsessive-compulsive disorder, and social phobia and there have been reports of increased depression in patients taking AR antagonists.[42] However, low BBB penetration of darolutamide made this off-target interaction an unlikely contributor to suspected psychiatric disorders reported. Similar to the nervous system side effect profile bicalutamide and enzalutamide followed with lower numbers of suspected reports, 0.15 and 0.1 ADRs per 100,000 dd.
Hepatobiliary disorders
Bicalutamide and flutamide both weakly inhibited BSEP. Chronic inhibition of BSEP can result in cholestasis due to the restriction of the bile passage. [43]
Flutamide had the largest reports at 0.44 per 100,000 dd, this could be due to being the only NSARA that was cleared by hepatic metabolism (approximately 50%) rather than renal clearance. Darolutamide followed with 0.44 ADR’s /100,000 dd. Darolutamide was the only NSARA to show effects at 5-HT (serotonin) receptor at < 10 µM and an association of inhibiting 5-HT and dysregulation of crucial components of liver function including hepatic blood flow, innervation and wound healing was noted.[44]
Apalutamide and enzalutamide did not reveal significant reports of hepatic injuries, due to possessing similar drug chemical properties. They are cleared via the kidney with a renal excretion of 65% and 71%, respectively.
Skin and subcutaneous tissue disorder ADR’s and fatalities
NSARAs exhibited suspected skin and subcutaneous tissue disorder ADRs impacting the epidermis. Apalutamide, darolutamide and bicalutamide possessed the largest number of ADR’s/100,000 dd (2.40, 2.67, and 0.33, respectively). This may be due to the apalutamide containing an arylamine (Figure 1), which are molecular characteristics known to induce skin reactions.[45] Conversely, flutamide also contains an arylamine, but there were no reports of suspected skin or subcutaneous tissue ADRs. Bicalutamide (Figure 1) contains a sulfonyl-arylamine which is known to induce inflammation via IgE and IgG immune pathways.[46] Resulting in patients with a sulfonyl-arylamine allergy that are taking bicalutamide to experience epidermal side effects due to the hypersensitivity as a potential cause.
Limitations
Reports from the Yellow Card scheme are suspected reports, and therefore, no causal relationship must be proved before submitting a report. Therefore, reported ADRs in any registry may have no defined relationship to the pharmacology of the NSARAs. Under-reporting of ADRs is also a common issue with pharmacovigilance schemes. [47] Independent datasets were used for frequency of dispensed prescriptions and frequency of reported suspected ADRs; neither dataset enables the interpretation of comorbidity and co-prescription confounders, for example. This limits the assessment of causality and identification and adjustment for these confounding factors. Prostate cancer occurs in later stages of life where issues of medications from co-morbidities can arise.
Furthermore, the prescriber’s choice of a NSARA may be influenced by known interactions (and contraindications) with drugs co-prescribed for comorbidities, i.e., there may be a significant bias by indication (or contraindication) as these comorbidities may affect sensitivity for ADRs. For example, certain NSARAs are currently cautioned as follows enzalutamide: seizures, encephalopathy, hypersensitivity, ischemic heart disease, falls and embryo-foetal tox; darolutamide: seizures, ischemic heart disease and embryo-foetal tox; and apalutamide: seizures, ischemic heart disease, fractures, falls and embryo-foetal tox. The study cannot adjust for these confounders. There is also a NSARA dose versus risk of association with an ADR that cannot be described within these datasets.
Conclusions
The detection of suspected ADR’s of NSARAs and potential correlations to their respected polypharmacology and physiochemical properties was identified. The highest number of ADRs were associated with enzalutamide (n = 1,091) and bicalutamide (n = 749). Enzalutamide was found to have the most off-target pharmacological interactions of the NSARAs studied (n = 5) including potent inhibition of GABA receptor (IC50 = 2.6 µM vs Cmax = 7.7 µM) and had the most ADR reports for nervous system disorders (n = 72, accounting for 73% of all NSARA ADRs in this SOC). Apalutamide, the only other GABA inhibitor (IC50 = 3 µM vs Cmax = 2.9 µM) had the highest relative rate of suspected nervous system ADRs at 1.08 per 100,000 dd.
Apalutamide was a modest inhibitor of the hERG ion channel (IC50 = 6 µM vs Cmax = 2.9 µM) and had the highest rate of suspected cardiac arrhythmia ADRs, 30-fold over the only other NSARA with this suspicion, enzalutamide, a significantly weaker hERG inhibitor (15.7 µM vs Cmax = 7.7 µM). Darolutamide was the only NSARA to show effects at 5-HT (serotonin) receptor at < 10 µM but did not translate to a correlation to psychiatric disorders due to low BBB penetration but may be associated with both cardiac and hepatobiliary suspected ADRs reported.
Data Availability
All data produced in the present work are contained in the manuscript
Acknowledgments
We thank OpenPrescribing, NHS secondary care medicines data set, EMC, FDA, ChEMBL, and MHRA Yellow Card Scheme for open access public data that underpinned this research.