Real‐World Claims Data on Prescribed Drugs and the Associated Drug–Drug Interaction Risk in Sickle Cell Disease in the United States

ABSTRACT

Sickle cell disease (SCD) is a debilitating hereditary hematologic disorder characterized by severe physical complications, including painful vaso‐occlusive crises, chronic pain, multi‐organ damage, and reduced life expectancy. Diverse medications are prescribed to treat SCD or manage its associated complications. However, no comprehensive resource for concomitant medication and drug–drug interaction (DDI) risk exists for people living with SCD. Here, we utilized real‐world claims data from the Merative Commercial MarketScan Research Database to identify the 100 most prescribed drugs for adolescents and adults living with SCD in the United States with commercial health insurance. DDI risks were assessed by leveraging the CERTARA Drug Interaction Database. In total, 43/100 drugs on the list were associated with DDI risks. Inhibitors were more frequently represented than inducers, suggesting that interaction risk in this population may be driven primarily by inhibition‐mediated mechanisms. CYP3A4 accounted for the largest proportion of identified interactions. The high prescription rates of the CYP3A4 substrates oxycodone (40%) and fentanyl (18.5%), associated with potentially serious clinical consequences due to DDIs, highlight the importance of considering CYP3A4 interaction risk during development of novel therapies for SCD. Overall, this analysis provides a population‐level insight into prescription patterns and subsequent DDI risks for people living with SCD. We expect these findings to (1) create clinical DDI risk awareness for this population and (2) inform strategic planning of DDI assessments during development of novel therapies indicated for SCD.

Study Highlights

• What is the current knowledge on the topic?

  • ○
    People living with SCD are exposed to polypharmacy, increasing the potential for DDIs. There is currently limited information describing real‐world prescribing patterns and associated DDI risks specific to the SCD population.
• What question did this study address?

  • ○
    This study used real‐world US claims data from commercially insured adolescents and adults with SCD to identify the most frequently prescribed drugs and characterize the potential DDI risks.
• What does this study add to our knowledge?

  • ○
    This study highlights the potential for clinically relevant DDI risk in patients with SCD and provides a data‐driven basis to inform strategic planning of DDI assessments during the development of novel therapies for this population.
• How might this change clinical pharmacology or translational science?

  • ○
    This study illustrates how real‐world data can be used to prioritize clinically relevant DDIs and enable tailoring of drug development strategies toward the most meaningful risks in the intended patient population.

Real‐world prescription data from 6230 adolescents and adults living with sickle cell disease (SCD) were combined with a drug–drug interaction (DDI) database to characterize the population‐level DDI risk. The findings support a DDI risk assessment for development of novel SCD therapies and during clinical management of SCD.

1. Introduction

Sickle cell disease (SCD) is a severely debilitating inherited hematologic disorder affecting approximately 100,000 people in the US and 8 million worldwide [1]. People living with SCD experience acute and chronic complications, including painful vaso‐occlusive crises (VOCs), acute chest syndrome, hemolytic anemia, and progressive multi‐organ damage caused by tissue ischemia, microinfarctions, and inflammation. Over time, these complications often lead to mental distress, including depression and anxiety, with both psychological and physiological morbidities significantly reducing life expectancy by 20–30 years [2]. As a result, pharmacologic management becomes increasingly complex, and polypharmacy is common in this population, increasing the potential for clinically relevant drug–drug interactions (DDIs). DDIs are therefore an important consideration during both clinical development of new therapies for SCD and in routine clinical practice. Understanding and managing these interactions for people living with SCD is crucial for ensuring optimal patient care and outcomes of treatment, as DDIs can lead to potentially dangerous adverse effects or treatment failure.

There is currently no comprehensive resource describing SCD medication usage, let alone emphasis on the DDI risks posed. An overview of concomitant medication use can guide prioritization of the early preclinical DDI evaluation and help determine the need and timing of clinical DDI trial during clinical development. This is also emphasized in regulatory DDI guidance where it is recommended to consider the likely concomitant medications in the intended patient population when planning a drug development strategy [3, 4, 5]. Utilizing real‐world data (RWD) is a novel and growing approach that captures prescription patterns (e.g., claims data in specific patient populations), thereby highlighting real‐life risks of DDIs rather than hypothetical [6, 7, 8]. Recent publications within this area have for example highlighted its application for covid‐19 treatment, multiple sclerosis and hormone receptor‐positive breast cancer [9, 10, 11]. Here, we utilized real‐world claims data to identify the 100 most frequently prescribed drugs among adolescents and adults with SCD in the United States with commercial health insurance and to assess their potential DDI risks and relevance for drug development.

2. Methods

2.1. Drug Data Source

This study used the Merative Commercial MarketScan Research Database, an administrative claims database containing longitudinal medical and pharmacy claims for commercially insured individuals in the United States. Drug prescriptions were derived primarily from outpatient pharmacy claims, which capture reimbursed prescription dispensing. Although the database environment includes linked inpatient claims, detailed and comprehensive information on inpatient medication administration is generally limited compared with outpatient pharmacy claims, as medications administered during hospitalizations are typically included within bundled payments rather than recorded as individual drug‐level claims. Any analysis, interpretation, or conclusion based on these data is solely that of the authors and not Merative.

Patients included in the current study were aged ≥ 12 years at the start of the observation period (01 January 2018–31 December 2022) and had both a SCD diagnosis code and a VOC diagnosis code recorded as separate entries within the database period (for detailed code list, refer to Supporting text). To be included in our analyses, patients also needed to have minimum 6 months of observation time, continuous health insurance coverage and available drug prescription information within the study period.

2.2. Top 100 Drug Selection

Data on drug prescriptions for the eligible patients were collected and matched with drug information using the drug concept ID to identify the total number of prescriptions for each drug. The number of patients who had been prescribed each drug was counted and divided by the total number of patients to estimate the proportion of patients who had been prescribed a specific drug. Drugs that appeared multiple times were only counted once for each patient. Listed drugs were ranked based on total patient counts associated with each drug (Table S1). Vaccines, inorganic salts, vitamins, and dietary supplements were excluded to maximize clinical applicability, as these classes are unlikely to present a DDI risk. Excluded substances with patient proportion rates are provided in Table S2. Diagnosis codes for SCD and VOC events are provided in Tables S3 and S4. Analysis was performed using Rv4.3.0.

2.3. Drug–Drug Interaction Analysis

The Drug Interaction Database (DIDB) curated by CERTARA was accessed in December 2024 to evaluate potential DDIs of the top 100 drugs by utilizing the drug interaction characteristics provided by DIDB. Additional clinically relevant DDI information for known individual drug metabolites was searched, and no effects were found except for prodrugs. For these prodrugs, DDI effects are reported for their active metabolite form. Importantly, the DDI analysis is focused solely on metabolism‐ and transporter‐mediated PK interactions and excludes other interactions such as pharmacodynamic (PD) drug interactions. Finally, we note the generated dataset does not provide details on administration route, which may be relevant when evaluating DDI risk in specific contexts.

3. Results

3.1. Study Population

We identified 26,319 people with both an SCD diagnosis and a VOC diagnosis code (Tables S3 and S4) recorded in the Commercial MarketScan Research Database, resulting in a prevalence of 12.42 per 100,000 people. Of these, 7418 had at least 6 months of observation time during the period of 01 January 2018–31 December 2022. After excluding patients under 12 years of age, the final study population was 6230 patients (56.5% female). In total, 5993 were defined as having had a drug prescription during the observation period. The mean age at inclusion was 33.5 years (SD: 13.9, Min, Max: [12, 65]). Included patients had a mean follow up duration of 2.54 years (SD: 1.52).

3.2. Top 100 Most Common Prescribed Drugs

From our analysis, we identified the 100 top drugs prescribed to patients with SCD (Table S1). The most common drug classes were antibiotics and analgesics (18% and 14% of the top 100 drugs, respectively). Prescribed treatment options for SCD were hydroxyurea (31% of patients), voxelotor (3.8% of patients), and crizanlizumab (2.9% of patients). Folic acid (vitamin B9) (39%) and glutamine (3.8%) were also prescribed but not included in Table S1 as vitamins and dietary supplements were part of the exclusion criteria for further analysis (see Table S2). Meanwhile, novel CRISPR‐Cas9 gene‐editing therapies were not included as they were approved in 2023, after the database observation period (01 January 2018–31 December 2022). The most prescribed drug in the studied population was the analgesic acetaminophen (49% of patients). The second most prescribed drug was the antiemetic ondansetron (44%), followed by four analgesic drugs (oxycodone, ketorolac, morphine, and ibuprofen). In total, 14 analgesic drugs were commonly prescribed: acetaminophen (49%), oxycodone (40%), ketorolac (40%), morphine (35%), ibuprofen (35%), hydromorphone (34%), hydrocodone (29%), fentanyl (19%), codeine (12%), tramadol (12%), naproxen (11%), meloxicam (7%), diclofenac (7%), and aspirin (4%). Other commonly prescribed drugs identified were the antibiotic amoxicillin (35%) and the antihistamine diphenhydramine (22%). Notably, the study period spanned the height of the covid era, and corticosteroids such as dexamethasone (17%) and prednisone (16%) were subsequently prescribed to a high degree despite recent studies recommending to avoid systemic corticosteroids to people living with SCD due to an increased risk of vaso‐occlusive episodes [12].

3.3. Drug–Drug Interaction Risk of Prescribed Drugs

We next assessed potential DDI risks associated with the identified drugs. Notably, no such risk was identified for the two most prescribed drugs on the list: acetaminophen and ondansetron. However, 43 drugs on the list were associated with a DDI risk. Of the 18 different antibiotics on the list, four are associated with DDI risks: azithromycin (weak CYP3A4 inhibitor and P‐glycoprotein [P‐gp] inhibitor), trimethoprim (weak CYP2C8 inhibitor), sulfamethoxazole (weak CYP2C8 inhibitor), and ciprofloxacin (moderate CYP1A2 inhibitor). Of the three antifungals on the list, two are associated with a DDI risk: fluconazole is a strong CYP2C19 inhibitor and moderate CYP3A4 and CYP2C9 inhibitor, while ketoconazole is a strong CYP3A4 inhibitor and weak CYP2C8 and CYP2C9 inhibitor. Ketoconazole is also a moderately sensitive CYP3A4 substrate. Interpretation of ketoconazole‐related DDI risk should be made with caution, as current use is predominantly topical, resulting in minimal systemic exposure.

Of the seven opioids on the list, four showed a DDI risk: oxycodone, fentanyl, tramadol, and codeine. Oxycodone and fentanyl are defined as moderately sensitive CYP3A substrates, and codeine and tramadol are classified as sensitive and moderately sensitive CYP2D6 substrates, respectively.

The specific identified inhibitor, inducer, and substrate drugs were extracted per CYP enzyme and presented in Figure 1. Several drugs associated with CYP‐mediated interactions were prescribed to more than 10% of the study population. These included fluconazole, midazolam, oxycodone, fentanyl, dexamethasone, and codeine. Fluconazole was identified as a strong and moderate inhibitor of CYP2C19 and CYP3A4/CYP2C9, respectively. Midazolam, oxycodone, fentanyl, and dexamethasone are all classified as CYP3A4 substrates, and codeine is a sensitive CYP2D6 substrate. Together, these frequently prescribed medications span multiple CYP enzymes and include both substrate and inhibitor mechanisms. The overall CYP interactions of the top 100 prescribed drugs are then further depicted in Figure 2. The most common interaction was with CYP3A4. Specifically, 14 drugs were substrate of CYP3A4, and 12 drugs were either inhibitor or inducer. The second most involved interacting enzyme was CYP2D6, with 5 substrates, where 3 are sensitive (codeine, dextromethorphan, and metoprolol), and 5 drugs were classified either as inhibitor or inducer. No strong CYP inducers were found across all enzymes, only one moderate and four weak inducers (Figure 2B). In contrast, three strong inhibitors were identified overall, affecting CYP3A4, CYP1A2, and CYP2C19. A total of nine sensitive substrates were observed, spanning all evaluated enzymes except CYP2C8 and CYP2B6. Notably, no drugs were identified that interacted with CYP2B6 as an inhibitor, inducer, or substrate.

FIGURE 1.

Drug‐level overview of CYP‐mediated DDI risks among the 100 most frequently prescribed medications in people living with SCD. Panels display enzyme‐specific interaction profiles: (A) CYP3A4, (B) CYP2D6, (C) CYP1A2, (D) CYP2C19, (E) CYP2C8, and (F) CYP2C9. For each enzyme, drugs are categorized according to their role as inhibitors, inducers, or substrates following the definitions outlined in the ICH Guideline M12 on drug interaction studies, with percentages indicating the proportion of patients prescribed each drug [5].

FIGURE 2.

Distribution of CYP‐mediated interactions among the most frequently prescribed drugs in people living with SCD. The x‐axis indicates the number of drugs out of the top 100 drugs, and bars are grouped by interaction potency. The identified interactions are color‐coded based on the potency of the interaction. Perpetrators (inhibitors [A] and inducers [B]): green; weak interaction, that is, inhibitor: AUCR ϵ [1.25–2[and inducer AUCR ϵ]0.50–0.80], orange; moderate interaction that is, inhibitor: AUCR ϵ [2–5[and inducer: Moderate interaction, that is, AUCR ϵ]0.20–0.50], and red; Strong; that is, inhibitor: AUCR ≥ 5 and inducers AUCR ≤ 0.20. Object interactions (substrates [C]) dark blue; Sensitive substrate, that is, AUCR ≥ 5 and light blue; Moderately sensitive substrate, that is, AUCR ϵ [2–5].

Of the top 100 prescribed drugs evaluated, a subset was identified as interacting with drug transporters, including P‐gp inhibitors (azithromycin, atorvastatin, ketoconazole), a BCRP inhibitor (pantoprazole), an OATP1B1/1B3 substrate (atorvastatin), OAT1/OAT3 substrates (ciprofloxacin, famotidine, furosemide), an additional OAT3 substrate (oseltamivir), and an OCT2/MATE1/MATE2‐K substrate (metformin).

4. Discussion

Here, we identify the 100 most prescribed drugs to people living with SCD using a commercial insurance claims database and assess their DDI risk by leveraging data available from DIDB. These findings provide a population‐level reference of commonly prescribed medications in SCD and highlight potential interaction liabilities relevant to drug development and clinical use.

4.1. DDI Mechanisms Among Commonly Prescribed Drugs in SCD

Our evaluation revealed a clear predominance of CYP3A4‐mediated interactions. CYP3A4 accounted for the largest proportion of identified substrates, as well as inhibitors and inducers, consistent with its known central role in drug metabolism. Notably, no strong CYP inducers were identified among the top prescribed drugs, and only a limited number of moderate or weak inducers were observed. Sensitive substrates were, except for CYP2C8 and CYP2B6, identified across all evaluated enzymes. Interactions involving these substrates are particularly relevant for drugs with a narrow therapeutic window (tizanidine, fentanyl, clindamycin, and levothyroxine), where even modest changes in exposure may have clinical implications. In contrast to the predominance of CYP3A4‐mediated interactions, transporter‐mediated interactions were less frequently represented among the top prescribed drugs. Only four drugs were classified as transporter inhibitors, all acting at the level of gastrointestinal efflux transporters, including P‐gp (azithromycin, atorvastatin, and ketoconazole) and BCRP (pantoprazole). Notably, no inhibitors of hepatic or renal transporters were identified, and the remaining transporter‐associated drugs were classified as substrates of OATP1B1/1B3, OAT1/3, OCT2, and MATE1/2‐K.

4.2. Implications for DDI Risk Assessment in Drug Development

The presented real‐world claims data provide important context for interpreting in vitro DDI findings during drug development and guide DDI screening strategy in drug discovery. Specifically, the clinical relevance of identified enzyme‐ and/or transporter‐based liabilities can be put into broader clinical context by assessing the actual SCD medication usage. As an illustrative example, a compound identified during lead selection as a potent CYP2C8 in vitro inhibitor would typically raise concerns regarding the potential for clinically relevant DDIs. However, the clinical significance of such a finding depends on the likelihood of co‐administration with CYP2C8 substrates in the intended patient population. As shown in Figure 1, only one drug among the 100 most frequently prescribed medications in the SCD population, montelukast, is a CYP2C8 substrate, and it is used by approximately 5.6% of patients. Moreover, montelukast is considered a moderately sensitive substrate with a broad therapeutic window and a well‐established safety profile with no restrictions on CYP2C8 inhibitors [13]. Taken together, these data suggest that the development of potent CYP2C8 inhibitors is unlikely to result in a high clinical DDI risk in the SCD population. Furthermore, understanding medication use in a target population can help evaluate how DDI‐related restrictions may affect patient recruitment in Phases 2 and 3 trials. Similar approaches have been suggested for other disease areas where a framework using real‐world data has been suggested to support a risk profile of drug candidates [8, 11, 14].

4.3. Potentially High‐Risk Drug Combinations in Clinical Practice

Figure 1 also provides the opportunity to highlight potential drug combinations of risk that may take place in clinical practice based on the prescription frequencies. Notably, oxycodone and fentanyl are moderate CYP3A4 substrates prescribed to 40% and 18.5% of people living with SCD in our analysis. The most prescribed CYP3A4 inhibitor identified was the antifungal drug fluconazole (11.5% of people living with SCD in our analysis). Two published case reports describe serious opioid‐related adverse events, including respiratory depression and death, following concomitant administration of fluconazole with oxycodone or fentanyl [15, 16]. While these reports represent low‐level evidence, they illustrate the potential clinical severity of CYP3A4‐mediated interactions involving commonly prescribed opioids in the SCD population. These examples support prioritization of CYP3A4 interaction risk during drug development. Figure 1 also highlights potential CYP1A2‐mediated interaction risks based on prescription frequencies. Ciprofloxacin, a moderate CYP1A2 inhibitor, was prescribed to approximately 9.7% of people living with SCD in our analysis, while the sensitive CYP1A2 substrates tizanidine and duloxetine were prescribed to 3.9% and 3.6% of patients, respectively. Two published reports describe serious adverse events following concomitant administration of ciprofloxacin with tizanidine or duloxetine, including hypotension and central nervous system toxicity [17, 18]. In addition, analysis of real‐world pharmacovigilance data has shown that ciprofloxacin–tizanidine co‐administration continues to occur in clinical practice despite the known CYP1A2 interaction risk, with reported adverse reactions primarily involving nervous system and cardiovascular toxicity, including rare fatal outcomes [19]. Together, these observations highlight that CYP1A2‐mediated interactions may have clinically significant consequences and warrant consideration during drug development DDI risk prioritization.

4.4. Cohort Definition Considerations

Identifying individuals with SCD from administrative claims data is challenging, and validated case‐finding algorithms have been developed to improve diagnostic accuracy, particularly for epidemiologic surveillance research. These approaches commonly require multiple SCD‐coded healthcare encounters over a defined observation period to reduce the risk of misclassification [20]. The objective of the current study differed from these applications. Rather than estimating disease prevalence or characterizing the broader SCD population, the aim was to assess concomitant medication use and potential DDI liabilities relevant to SCD drug development. Recent SCD drug development programs, beyond those focused on the transfusion‐dependent population, incorporate a history of VOCs into eligibility criteria [21, 22]. Accordingly, a different cohort definition was applied. The selected cohort incorporated both an SCD diagnosis code and a documented VOC with the intent to identify individuals with clinically manifest disease and active healthcare contact. Requiring a documented VOC may preferentially identify individuals with greater disease‐related healthcare utilization and underrepresent patients with well‐controlled disease, including those receiving effective disease‐modifying therapy or chronic transfusion support, yet will also provide effective control of the false‐positive rate. Accordingly, the findings should be interpreted as representative of the selected claims‐based cohort included in this analysis.

4.5. Limitations

Our study has limitations. The used database solely contains information on people living in the US with commercial health insurance often provided by their employer. As a result, the prescribing patterns and concomitant medication use described here are most representative of commercially insured individuals with SCD and should not be interpreted as nationally representative of all people living with the disease in the United States. Disease burden associated with SCD may limit some people from sustaining a job, meaning the results described here may generalize best to people living with SCD in the US workforce. Indeed, SCD prevalence in this database (12.75 in 100,000) is lower than that of the general US population. For context, findings from a study analyzing Medicaid enrollment, claims, and encounter data from 52,524 enrollees in 2021 indicated that approximately 50% of people living with SCD in the United States were covered by the public insurance Medicaid [23]. Despite this limitation, the selected database was chosen as it can accurately capture prescription patterns linked to insurance reimbursements. Furthermore, this study should be interpreted in the context of limitations inherent to administrative claims databases. In particular, detailed and comprehensive information on inpatient medication administration is generally limited, as medications administered during hospitalizations are typically captured through bundled billing rather than as individual drug‐level claims. This limitation is common across claims data sources and is not specific to the database used in this study. Despite these constraints, claims databases have been widely and successfully used in SCD research to characterize real‐world treatment patterns, specific concomitant medication use, and healthcare utilization in the outpatient setting [24, 25, 26].

Accordingly, the presented analysis is consistent with established real‐world approaches in SCD and provides relevant insights into prescribing patterns from a drug development perspective.

5. Conclusions

This study provides a real‐world evaluation of the most frequently prescribed drugs in adolescents and adults with SCD in the United States who have commercial health insurance, and their associated potential for DDIs. By characterizing the DDI risk profiles of these commonly prescribed medications, our analysis offers insight into the real‐world co‐medication landscape relevant to this population. The presented findings are intended to support early DDI risk assessment and inform clinical pharmacology strategy, including the identification and mitigation of interaction risk for novel therapies intended for use in people living with SCD.

Author Contributions

Wrote the manuscript: Ifeyinwa Osunkwo, Laura Grenier, Peter Bruun‐Rasmussen, Rune Aabjerg Nørgaard and Carolina Säll. Designed the research: Peter Bruun‐Rasmussen, Rune Aabjerg Nørgaard and Carolina Säll. Performed the research: Laura Grenier and Peter Bruun‐Rasmussen. Analyzed the data: Ifeyinwa Osunkwo, Laura Grenier, Peter Bruun‐Rasmussen, Rune Aabjerg Nørgaard and Carolina Säll.

Funding

The authors have nothing to report.

Conflicts of Interest

Rune Aabjerg Nørgaard, Peter Bruun‐Rasmussen, and Carolina Säll are employees of Novo Nordisk and may hold shares in the company. Laura Grenier and Ifeyinwa Osunkwo were employed by Novo Nordisk at the time of the study.

Supporting information

Table S1: Top 100 drugs prescribed to people living with SCD.

Table S2: List of excluded substances.

Table S3: Diagnosis codes for SCD.

Table S4: Diagnosis codes for VOC events.