Real‐World Evidence Application in Translational Medicine: Making Use of Prescription Claims to Inform Drug–Drug Interactions of a New Psoriasis Treatment

Abstract

Patients with psoriasis often take multiple medications due to comorbidities, raising concerns about drug–drug interactions (DDIs) during the development of new medicines. DDI risk assessments of a new small molecule showed risks of CYP3A4 autoinduction and being a sensitive CYP3A4 substrate. We conducted a real‐world evidence (RWE) claims analysis to assess the frequency of prescription claims for up to 12 months from the date of the initial psoriasis diagnosis for drugs that may interact with CYP3A4 substrates. We used 2013 to 2018 patient data from the US Merative MarketScan Research Database. Among patients diagnosed with psoriasis, less than 1% had a claim for a moderate/strong inducer, but up to 15% had a claim for moderate/strong inhibitor. Most prescriptions for CYP3A4 inhibitors or inducers included antibiotics and anticonvulsants. While CYP3A4 inducers were rarely used, those treated received more than >90 days treatment. Then, these RWE data were used to inform the early translational medicine strategy for the new investigational drug by strategically integrating DDI evaluations into a first‐in‐human healthy volunteer trial prior to studies in patients with psoriasis. The resulting DDI substudy showed that the investigational small molecule did not induce midazolam clearance but was sensitive to CYP3A inhibition, leading to the decision to exclude concomitant use of strong CYP3A4 inducers or inhibitors from clinical trials.

Introduction

Psoriasis is a systemic immune‐mediated inflammatory disease and a serious public health concern as it affects at least 100 million patients globally, with a rising prevalence and substantial disease burden. ¹ , ² , ³ , ⁴ Anti‐inflammatory topical or systemic medications and phototherapy are used to treat psoriasis, depending on the disease severity and comorbidities at diagnosis. ³ , ⁵ Moderate‐to‐severe psoriasis is treated with biologic medications, mainly monoclonal injectable antibodies that target highly specific immune pathways. ² In addition, a significant effort to develop effective small molecule, orally bioavailable therapies is taking place. ⁶

Patients with psoriasis often take many concomitant medications, as the disease is associated with serious comorbidities such as cardiovascular diseases, mental disorders, lymphoma, psoriatic arthritis, skin cancer, and inflammatory bowel disease. ⁴ As a result, drug–drug interactions (DDIs) complicate therapy options and are a major concern in the development of new medicines in psoriasis. ⁶ , ⁷ , ⁸ This concern arises because DDIs may impact the pharmacokinetics (PK) and/or pharmacodynamics (PD) of the investigational and concomitant drug and alter the effectiveness and safety of concomitantly received medications. DDIs are linked with higher morbidity, increased hospitalizations, longer hospital stays, and poor outcomes. ⁹ , ¹⁰ Therefore, knowledge of comorbidities and concomitant medications of patients with psoriasis is important to address potential complex DDIs of new therapies. ⁸ For example, dermatologists, one of the specialist practitioners who may treat psoriasis, often have to think about the impact of CYP3A modulators as they may encounter DDIs between small molecules in conditions such as atopic dermatitis and onychomycosis. The sensitive or moderately sensitive CYP3A substrate (oral) cyclosporine is a treatment option for atopic dermatitis that affects up to a quarter of children and about 2.5% of adults. ¹¹ Oral azole antifungals that are moderate or strong CYP3A inhibitors are treatment options for onychomycosis, a condition that is common in 5.5% of the worldwide population. ¹² DDI management will continue to be necessary during development of new small molecule treatments for psoriasis and potentially as these treatments reach patients.

The investigational psoriasis medication of interest is a small molecule delivered orally. The DDI risk assessments projected it to be a CYP3A4 autoinducer and a sensitive CYP3A4 substrate. To our knowledge, there are no previous studies assessing the relative frequency of prescription claims for drugs that may interact with CYP3A4 inhibitors or inducers in the US population of adult patients with psoriasis. We thus conducted this study to assess, among patients with psoriasis, the frequency of prescription claims for drugs that may interact with CYP3A4 substrates. We used these data first to inform the DDI evaluations of an investigational drug for psoriasis and then to establish the study design of a first‐in‐human healthy volunteer trial.

RWE Methods

Study Design

In this longitudinal, observational, retrospective cohort study, we used patient data from the years 2013 to 2018 from the US Merative MarketScan Research Database. Institutional review board approval to conduct this study was not necessary because the database records were deidentified and fully compliant with US patient confidentiality requirements.

Patients with a psoriasis diagnosis were identified, and their concomitant medications were recorded for up to 12 months from the date of the initial psoriasis diagnosis, that is, the index date.

The primary objectives were to quantify the number of patients with psoriasis who had at least one claim for a strong or moderate CYP3A4 inhibitor or inducer and assess their therapy duration. The secondary objective was to calculate the frequencies of those prescription claims in the subgroup of patients with at least one claim of advanced psoriasis treatment.

Outcome Measures

The exposures of interest were concomitant administration of strong or moderate CYP3A4 inhibitors or inducers (CYP3A4 modulators). Concomitant medications were recorded as moderate or strong CYP3A4 modulators according to the classifications in appropriate submission‐based and publicly available databases. ¹³ , ¹⁴ , ¹⁵

Two cohorts of patients with psoriasis were considered: the overall population and a subpopulation consisting of patients receiving an advanced psoriasis medication, that is, any of the following medications during the 1 year of post index follow‐up: etanercept, adalimumab, infliximab, golimumab, certolizumab pegol, ustekinumab, brodalumab, secukinumab, ixekizumab, tildrakizumab, risankizumab, guselkumab, or apremilast.

Study Population

Eligible patients were adults (age ≥18 years at index date) with at least one diagnosis of plaque psoriasis (ICD‐9‐CM: 696.1 or ICD‐10: L40.0) recorded at one inpatient visit or two outpatient (minimum 30 days apart) visits between May 1, 2013 and April 30, 2017. Patients should also have had continuous enrollment with medical and pharmacy benefits for 12 months on and after the index date (a 29‐day gap in enrollment was allowed). Patients were excluded if, during the first post index year, they had a diagnosis of another autoimmune disease, including psoriatic arthritis, rheumatoid arthritis, Crohn's disease, ulcerative colitis, juvenile idiopathic arthritis, ankylosing spondylitis, hidradenitis suppurativa, or uveitis (Table S1).

Study Data and Sources

The Merative MarketScan Research Database contains a nationally representative data sample of the US population with employer‐provided health insurance, covered dependents, and retirees with supplemental Medicare coverage. ¹⁶ , ¹⁷ , ¹⁸ The database captures patient‐level data on demographics, real‐world treatment patterns, and healthcare resource utilization across inpatient, outpatient, prescription drug (i.e., retail, mail order, and specialty pharmacies), and carve‐out care. ¹⁶ , ¹⁷ The diagnoses in this database are captured by the International Classification of Disease Clinical Modification (ICD‐CM), 9th (ICD‐9) and 10th revisions (ICD‐10). ¹⁹ The pharmacy records of patients’ prescription drugs were coded using the National Drug Code (NDC) published by the FDA ²⁰ and Healthcare Common Procedure Coding System (HCPCS) codes. ²¹

Statistical Analysis

Descriptive statistics were conducted. Demographic information included age, sex, insurance type (commercial or Medicare), and area of residence by broad geographical region (Midwest, Northwest, South, West). We assessed the number and percentage of patients with claims for strong or moderate CYP3A4 inhibitors and strong or moderate CYP3A4 inducers during the follow‐up period, and we stratified by age. The mean Charlson comorbidity index ²² was calculated during the 1 year post index for the main cohort of patients and for the subgroup of patients receiving advanced therapy. The length of concomitant treatments with CYP3A4 modulators was determined by the days of each medication supplied.

The evaluation of the study outcome measures required complete records of the patients’ medical history. Only data from patients with complete claims, enrollment, and demographic information were used in this study. All analyses were performed using the Instant Health Data platform (Panalgo, Boston, MA).

RWE Results

Study Population

A total of 219,166 patients had a psoriasis diagnosis. Of these, 96,744 patients composed the full cohort of patients eligible for inclusion in our analysis, and among those, we identified 18,814 patients receiving advanced therapy (Figure 1, Table 1).

Figure 1.

Flowchart of patient selection. Notes:^*Psoriatic arthritis, rheumatoid arthritis, Crohn's disease, ulcerative colitis, juvenile idiopathic arthritis, ankylosing spondylitis, hidradenitis suppurativa, or uveitis 12 months from the date of the initial psoriasis diagnosis. †Twelve months from the date of the initial psoriasis diagnosis

Table 1.

Patient Characteristics

	Full Cohort	Advanced Therapy Cohort *
Characteristic	(n = 96,744)	(n = 18,814)
Sex, male, n (%)	47,883 (49.5)	10,706 (56.9)
Age
Mean (SD)	50.0 (15.0)	46.4 (12.9)
Median	51	47
Range	18‐100	18‐93
Age years group, n (%)
18‐24	5900 (6.1)	1132 (6.0)
25‐34	10,732 (11.1)	2576 (13.7)
35‐44	17,133 (17.7)	43,336 (23.1)
45‐54	23,240 (24.0)	5229 (27.8)
55‐64	26,029 (26.9)	4518 (24.0)
65‐74	8562 (8.9)	770 (4.1)
75+	5148 (5.3)	253 (1.3)
Insurance type, n (%)
Commercial	83,535 (86.4)	17,822 (94.7)
Medicare	13,209 (13.7)	992 (5.3)
Region, n (%)
Midwest	21,362 (22.1)	3974 (21.1)
Northeast	22,414 (23.2)	3275 (17.4)
South	37,687 (39.0)	8762 (46.6)
West	13,900 (14.4)	2419 (12.9)
Missing	1381 (1.4)	384 (2.0)
Charlson Comorbidity Index, mean (SD)	0.49 (1.11)	0.34 (0.87)

SD, standard deviation.

^a Advanced therapy includes etanercept, adalimumab, infliximab, golimumab, certolizumab pegol, ustekinumab, brodalumab, secukinumab, ixekizumab, tildrakizumab, risankizumab, guselkumab, and apremilast.

The mean age of the full cohort was 50 years (standard deviation [SD], 15). Men and women were equally represented in the full cohort, but there were more men in the advanced therapy cohort (Table 1). The mean Charlson comorbidity index was 0.49 (SD, 1.11) in the full cohort and 0.34 (SD, 0.87) in the advanced therapy cohort (Table 1).

Strong or Moderate CYP3A4 Μodulators

Moderate inhibitors were the most frequently used in the full cohort (15.6%) and in the advanced therapy cohort (14.3%), followed by strong inhibitors (3.2% and 2.8%, respectively), strong inducers (0.6% and 0.7%, respectively), and moderate inducers (0.5% and 0.3%, respectively) (Table 2). The same pattern was recorded by the age subgroup analysis (Table S2).

Table 2.

Type of Strong or Moderate CYP3A4 Modulator Prescribed to Patients with Psoriasis During the 12 Months Post Index, ^a Merative MarketScan Data 2013‐2017

Strong or Moderate, CYP3A4 Modulator Type, n (%)	Full Cohort (n = 96,744)	Advanced Therapy Cohort b (n = 18,814)
Strong inducer	615 (0.6)	124 (0.7)
Moderate inducer	445 (0.5)	53 (0.3)
Strong inhibitor	3114 (3.2)	527 (2.8)
Moderate inhibitor	15,046 (15.6)	2695 (14.3)

^a Twelve months from the date of the initial psoriasis diagnosis.

^b Advanced therapy includes etanercept, adalimumab, infliximab, golimumab, certolizumab pegol, ustekinumab, brodalumab, secukinumab, ixekizumab, tildrakizumab, risankizumab, guselkumab, and apremilast.

The most common CYP3A4 inducer prescribed in both cohorts was the anticonvulsant carbamazepine (0.25% in the full cohort, 0.24% in the advanced therapy cohort; Table S3). The most common CYP3A4 inhibitors prescribed in both cohorts were the antibiotic ciprofloxacin (8.0% in the full cohort and 6.7% in the advanced therapy cohort) and the antifungal fluconazole (6.0% and 5.5%), followed by the Ca‐channel blocker diltiazem (1.4% and 1.2%), clarithromycin (1.3% and 1.2%), and cyclosporine (0.6% and 1.4%) (Table S3).

Most patients who were prescribed CYP3A4 inducers had ≥3 such claims: 66.3% (688/1038) in the full cohort and 56.3% (99/176) in the advanced therapy cohort (Table 3). Most patients who were prescribed CYP3A4 inhibitors had only one such claim: 56.4% (9787/17,351) in the full cohort and 56.7% (1761/3105) in the advanced therapy cohort (Table 3).

Table 3.

Duration of Therapy and Number of Concomitant Strong or Moderate CYP3A4 Modulators Prescribed to Patients with Psoriasis During the 12 Months Post Index, ^a Merative MarketScan Data 2013‐2017 ^b

	Full Cohort (n = 96,744)	Advanced Therapy Subgroup c (n = 18,814)
CYP3A4 strong or moderate inducers	N = 1038	N = 176
Number of medications per patient, mean (SD)	4.97 (4.4)	3.77 (3.4)
Number of inducers per patient, n (%)
1	243 (23.4)	57 (32.4)
2	107 (10.3)	20 (11.4)
≥3	688 (66.3)	99 (56.3)
Duration of treatment
Mean days (SD)	175.5 (121.6)	138.9 (114.2)
Median days	180	99
Inducer‐treated days, n (%)
Acute	225 (21.7)	53 (30.11)
Chronic d	647 (62.3)	89 (50.6)
Subchronic e	166 (16.0)	34 (19.3)

CYP3A4 strong or moderate inhibitors	N = 17,351	N = 3105
Number of medications per patient, mean (SD)	2.35 (2.52)	2.32 (2.40)
Number of inhibitors per patient, n (%)
1	9787 (56.4)	1761 (56.7)
2	2992 (17.2)	515 (16.6)
≥3	4572 (26.4)	829 (26.7)
Duration of treatment
Mean days (SD)	48.5 (90.2)	47.8 (86.8)
Median days	10	10
Inhibitor‐treated days, n (%)
Acute	13,478 (77.7)	2,382 (76.7)
Chronic d	2460 (14.2)	442 (14.2)
Subchronic e	1413 (8.1)	281 (9.1)

SD, standard deviation.

^a Twelve months from the date of the initial psoriasis diagnosis.

^b Table 3 only includes patients with complete prescription information, specifically the days supplied of the medication.

^c Advanced therapy includes etanercept, adalimumab, infliximab, golimumab, certolizumab pegol, ustekinumab, brodalumab, secukinumab, ixekizumab, tildrakizumab, risankizumab, guselkumab, and apremilast.

^d Chronic: >90 days.

^e Subchronic: 31‒90 days.

Among those receiving CYP3A4 strong or moderate inducers, these were prescribed for chronic use in 62.3% of the patients, whereas CYP3A4 strong or moderate inhibitors were prescribed for acute use in 77.7% (Table 3). In the full cohort, the duration of treatment was on average 175.5 (SD, 121.6) days (median: 180 days) for inducers and 48.5 (SD, 90.2) days (median: 10 days) for inhibitors (Table 3).

RWE Application to Translational Medicine

As per standard drug development process (Figure 2a), prior to testing an investigational drug in patients, in vitro studies and proposed or measured clinical exposures inform stepwise quantitative DDI risk assessment using basic, mechanistic static, or dynamic physiologically based pharmacokinetic models. ²³ , ²⁴ Post risk assessment, inclusion and exclusion criteria are developed for specific patient populations. These criteria are often based on broad categories, such as “Exclude strong inhibitors and inducers of CYP3A4.”

Figure 2.

Actionable drug development solution using RWE. (a) The standard drug development process, in which prior to testing an investigational drug in patients, in vitro studies and proposed or measured clinical exposures inform stepwise quantitative DDI risk assessment models. Based on such investigations, it was determined that our molecule was a CYP3A4 substrate and a CYP3A4 inducer. This raised the question of whether concomitant medication could confound its exposure–response profile. Within the standard process (a), that question could typically be answered at any time, including post approval observational studies, when the molecule of interest would be administered to patients receiving concomitant medications not allowed within the standard clinical development process. However, early answers could be obtained by the alternative procedure of conducting an early RWE study investigating the frequency of CYP3A4 modulators among the concomitant medications prescribed to patients with psoriasis. (b) The incorporation of an RWE study into the early stages of the standard process. In the present case, our RWE study showed that DDIs were indeed a risk in the psoriasis patient population. A DDI substudy was subsequently conducted and showed that the investigational small molecule was a sensitive CYP3A4 substrate, and exposures were expected to change >5‐fold in the presence of strong inhibitor (and by inference, a strong inducer). This information was fed back to the standard clinical development process (a), finally leading to the decision to exclude concomitant use of strong CYP3A4 inducers or inhibitors from clinical trials. DDI, drug–drug interaction; PD, pharmacodynamics; PK, pharmacokinetics; RWE, real‐world evidence.

The in vitro findings that our molecule was potentially a CYP3A4 substrate and CYP3A4 inducer (and thus an autoinducer) prompted the question of whether concomitant medication could confound its exposure–response profile. To investigate the answer to this question as early as possible, we incorporated a real‐world evidence (RWE) analysis into our DDI research, as outlined in Figure 2. This deviation from the standard process gave us the opportunity to incorporate into the drug development process, specific, actionable, population‐tailored information for investigators using a combination of curated lists of relevant concomitant medications within a category (Figure 2b). Our lists of CYP3A4 inhibitors and inducers were curated from key trusted sources. ¹³ , ¹⁴ , ¹⁵

Ideally, key DDI questions are answered in early clinical studies, after which continued development can be reevaluated. For this molecule, a midazolam substudy revealed no meaningful CYP3A4 induction activity, thus resolving a key question for future development. However, the molecule was shown to be a CYP3A4 sensitive substrate with a >5‐fold change in exposure when coadministered with a strong CYP3A4 inhibitor.

Discussion

The present RWE analysis of a large US claims database revealed that up to 15% of patients with psoriasis had claims for medications with potential DDIs. This finding underscores the risk of DDIs in this population and highlights the need for improved recognition and management of such interactions to minimize the risk of therapeutic failure and adverse drug effects in both clinical practice and drug development. The CYP3A4 modulator was usually an inhibitor as inducers were only prescribed in a very low percentage of patients. These RWE data were used to inform the early translational medicine strategy for the new investigational drug. This was done by strategically integrating DDI evaluations into a first‐in‐human healthy volunteer trial prior to studies in patients with psoriasis. The resulting DDI substudy showed that the investigational small molecule was not an inducer of midazolam clearance at a dose of 300 mg but was sensitive to strong inhibition of CYP3A by itraconazole, with a >5‐fold increase in exposure. Therefore, the initial decision to exclude concomitant use of strong CYP3A4 inducers or inhibitors from the first‐in‐human clinical trial was appropriate, and the decision to continue restrictions on use of strong CYP3A4 inducers or inhibitors could then be justified based on clinical data.

The present findings show that DDIs might pose a risk to a portion of patients receiving psoriasis treatments that are CYP3A substrates and are sensitive to CYP3A‐mediated DDI. This necessitates strategies to reduce therapeutic failure and adverse effects, simplify treatment outcomes, and improve compliance. ⁶ By integrating RWE into early DDI assessments, the present study proposes a methodological framework for the development of a complete clinical pharmacology strategy. Such a strategy could be used to better understand the characteristics of target patient populations, thereby presenting decision‐makers with an added layer of information about whether to include design elements like DDI substudies in Phase 1 clinical trial protocols and inclusion/exclusion criteria. Factors such as duration of treatment (acute or chronic), route of administration (including oral, topical, and others), ability to temporarily pause, should be considered when determining specific concomitant medications to include or exclude in trials. It can be wise to invest early in research, including combining real‐world data and clinical evidence, that could reduce uncertainty around real or perceived factors that affect eligibility criteria and thus may slow enrollment in Phase 2 to 3 clinical studies. While efforts exist to ensure that drug labeling, which becomes available post approval, is improved in terms of accessibility and utility, ²⁵ RWE can help improve the utility of the DDI information provided to investigators during drug development. RWE can help shift the focus from broad exclusions, whose implementation depends on investigator experience, to population‐specific exclusions that (1) minimize guesswork, (2) are likely to be commonly encountered in the daily practice of treating a specific condition, and (3) may increase the efficiency of enrollment.

Continuous technological advancements have facilitated the vast generation of digital patient data, such as the laboratory, genomic, and clinical data giving rise to RWE research. ²⁶ RWE applications in drug discovery, development, regulatory authorization decisions, and post authorization lifecycle management have gained increasing interest in recent years and are encouraged by the US and European regulatory authorities. ²⁶ , ²⁷ , ²⁸ , ²⁹ , ³⁰ RWE can also be used in clinical pharmacology research to address specific drug development questions. ³¹ , ³² In that respect, RWE has been used to support label expansion, ³³ , ³⁴ generate control groups for rare disease studies, ³⁵ , ³⁶ identify biomarkers, ³⁷ and inform dosing recommendations for patients with organ impairment and for pediatric patients. ³¹ , ³⁸ , ³⁹ , ⁴⁰ To our knowledge, our work is the latest addition to only three previous efforts incorporating RWE into DDI assessments. ⁴¹ , ⁴² , ⁴³ Duke et al used large databases containing electronic patient records to identify DDIs that, based on in vitro pharmacology evidence, could lead to myopathy. ⁴¹ They found five drug pairs that increase the risk of myopathy compared to the risk with either drug alone: loratadine‐ropinirole (relative risk [RR], 3.21), promethazine‐tegaserod (RR, 3.00), loratadine‐duloxetine (RR, 1.94), loratadine‐alprazolam (RR, 1.86), and loratadine‐simvastatin (RR, 1.69). ⁴¹ Lorberbaum et al used adverse event reports and electronic patient records in combination with laboratory experiments to investigate QT‐interval DDIs. ⁴² They found that when ceftriaxone and lansoprazole are combined, they produce prolongation in QT intervals by blocking the human ether‐à‐go‐go‐related gene channel. ⁴² More recently, Yee et al used databases containing data on patients with COVID‐19 to complement their in vitro assessments of 25 small molecules used in COVID‐19 research to include transporter‐mediated DDIs. ⁴³ They found that 20 of those molecules had a risk of DDI, indicating that patients with COVID‐19 who receive various concomitant medications should be closely followed for known adverse drug reactions. ⁴³ Clinical pharmacology applications of RWE research could be increasingly impactful and broadened with collaboration with scientists and experts from a variety of fields, such as epidemiology, clinical research, medical affairs, who could leverage their expertise in the selection of patients and sources and in data extraction and cleaning. ³¹

Limitations

Patients with diagnosis of more than one autoimmune disease were excluded from this evaluation to align with the relevant patient population considered in the clinical trial with the small molecule drug of interest. This also helped minimize misclassification of moderate or severe psoriasis patients at risk for DDI who might be receiving advanced therapy for other indications. We also included in our study a subset of patients receiving advanced psoriasis treatment to ensure the inclusion of patients with only psoriasis, that is, without other underlying diseases. Furthermore, the inclusion of such a group allowed for more conservative estimates for our compound's DDIs, given that these patients are likely to have moderate to severe disease and are generally more compliant with their treatment. However, it would be interesting in future work to investigate the frequency and prevalence of plaque psoriasis among patients with additional autoimmune disease or other conditions and the subsequent impact on the concomitant medications. Our study conclusions are affected by the nature of the data recorded in the database used. ³¹ Specifically, these results may not be generalizable to populations beyond those identified in the commercial claims database. Additionally, the database does not provide information as to whether the patients were adherent to the prescriptions they received and does not include over‐the‐counter medications. Lastly, the list of concomitant CYP3A inhibitor and inducer medications was curated from multiple independent sources and was thus considered as comprehensive as possible. Portions of the list are presented herein, but we recommend that sponsor companies curate lists based on their own institutional knowledge. With new approvals and rare withdrawals, similar lists should be updated on a frequency that enables patient safety to be maintained.

Conclusions

The present RWE findings demonstrate that, considering DDI risks among certain patients treated with psoriasis medications, strategies are required to reduce therapeutic failure and maximize patient safety while preserving the ability to enroll patients in efficacy trials. This study further illustrates the application of RWE to inform clear decision‐making in the early development plan and evaluation of an investigational drug risk profile.

Conflicts of Interest

Casey Kar‐Chan Choong, Jessica Rehmel, and Amita Datta‐Mannan are employees and/or stockholders of Eli Lilly and Company.

Funding

This study was funded by Eli Lilly and Company.

Supporting information

Supporting Information

Data Availability Statement

The datasets generated and/or analyzed during the current study are not publicly available due to individual data privacy but may be available from the corresponding author on reasonable request.