Treatment Persistence and Variations in Prescribing Oral, Injectable, and Inhaled Corticosteroids: A Population-Based Drug Utilisation Study

1 Introduction

Corticosteroids (CS) are used to treat multiple conditions such as allergies, eczema, asthma, chronic obstructive pulmonary disease (COPD), inflammatory diseases, and adrenal insufficiency, and have also been used widely as an effective treatment against severe Covid-19 infection [1]. It has been estimated that between 1% and 3% of the general population use systemic (oral or injectable) CS [2-4], often prescribed as long-term therapy [2]. In the UK, it has been estimated that 1.4% of patients over the age of 55 years received oral CS continuously for at least 3 months [5]. Furthermore, national drug utilisation data from the USA (1999–2008) and the UK (1989–2008) showed a rising incidence of long-term CS prescriptions despite increased use of disease-modifying drugs [2, 6].

CS are available through multiple administration routes, dosages, and treatment durations. Although they have been used widely, systemic CS may lead to several adverse events, including immunosuppression, osteoporosis, fractures, adrenal suppression, diabetes, psychiatric disturbances, dermatologic and cardiovascular events [7], as well as increased risk of hyperglycemia, especially for patients with diabetes [8]. It is important to study CS drug utilization and highlight variations in use to identify high users who are potentially at greatest risk of adverse events. Furthermore, understanding variation in prescribing and treatment adherence is crucial in this context, as inconsistent or prolonged CS use may further exacerbate the risk of adverse events.

Medication adherence is a concept that describes the degree to which a patient receives medications as advised. The EMERGE guidelines highlight three key phases: (a) treatment initiation; (b) implementation of the prescribed dosing; and (c) persistence with treatment [9, 10]. Nonadherence may occur at any stage, for instance, when patients do not initiate or delay treatment, take additional doses or omit doses, or if treatment is discontinued. Two of the methods that can measure treatment persistence are: (a) the refill-gap method that measures persistence based on gaps between refills of prescriptions, and (b) the proportion of patients covered (PPC) method, which estimates the proportion of patients covered by treatment at a specific time after treatment initiation [11].

This study aimed to report variation in prescribing, treatment persistence and combined treatment patterns among patients who were prescribed oral, injectable and inhaled CS in primary care in England.

2 Methods

3 Results

We identified 1 942 571 patients of whom 1 375 099 were prescribed oral CS, 140078 injectable and 912 454 inhaled CS across 1471 GP practices in England. Across the total CPRD population (15 681 632 patients), 8.8% of patients were prescribed oral CS, 0.9% of injectable CS, and 5.8% inhaled CS. Prednisolone (91.5%), methylprednisolone (75.5%), and beclometasone (84.2%) were the most frequently prescribed oral, injectable, and inhaled CS, respectively. More than half of the patients (54.6%) were female, 67.6% were older than 35 years, and 24.1% were enrolled with GP practices in the most deprived residential quintiles of England. The median follow-up was 6.8 years (interquartile range, IQR 2.8–12.5). Among patients receiving oral or inhaled CS, 67% were enrolled with GP practices in the 3rd, 4th, or 5th most deprived residential quintiles of England, with the corresponding number of patients receiving injectable CS being 57%. Regarding regional variation, nearly 33% of patients receiving either oral or inhaled steroids and 47% of patients in injectable CS were enrolled with GP practices in the South of England. Table 1 shows the baseline characteristics of the study cohort.

The median treatment duration among oral CS users was 28 days (IQR 28–36) and 30 days (IQR 30–133) among inhaled CS users. Only 2% received oral CS for longer than a year, with 80% of patients being on treatment for 44 days or fewer, while a small subset remained on treatment for significantly longer, with the maximum duration reaching 7542 days, which suggests a highly skewed distribution. Moreover, oral users had a median of 5 (IQR 2–14) treatment episodes, with the corresponding numbers for inhaled CS being 3 (IQR 1–6). Fifty-three percent (74250) of injectable CS users received only one prescription during their follow-up, with 20.6% (28818) receiving more than one prescription within 6 months, 11.8% (16456) within 6–12 months, 22.3% (31225) within 1–5 years, and 15.3% (21461) more than 5 years since the first prescription was issued. In terms of the number of prescriptions, 39.6% (55437) patients received between two and five, and 6.5% (9056) patients had more than five prescriptions of injectable CS.

The Lorenz curve in Figure 1 demonstrates that nearly 20% of oral and inhaled users accounted for 80% of CS use. A comparison between KM and PPC plots for oral and inhaled CS in Figure 2 shows distinct patterns across the two formulations. Almost 80% of oral CS users received continuous treatment during the first month of their follow-up, with a sharp drop to 20% (Hazard Ratio, HR: 0.205; 95% Confidence Interval, CI: 0.200–0.206) after 2 months. The proportion of persistent users for longer than 100 days was less than 3% (HR: 0.026; 95% CI: 0.026–0.026). Similar patterns were observed for inhaled formulations, with approximately 50% (HR: 0.475; 95% CI: 0.471–0.477) of persistent users 2 months following treatment initiation. The PPC method showed that almost 30% (HR: 0.301; 95% CI: 0.300–0.304) of oral and 60% (HR: 0.600; 0.600–0.603) of inhaled users were covered by treatment 6 months following treatment initiation. In sensitivity analyses, the KM plots showed distinct patterns for oral and inhaled CS when different grace periods of 30, 60, and 90 days were applied (Figure S3). Similarly, in Figure S4, the PPC method showed that most patients were covered by treatment 6 months following treatment initiation, with the application of different grace periods shifting the observed patterns proportionately.

Examination of the dose impact on CS prevalence (Figure 3) showed that a higher proportion of women older than 35 years and men older than 45 years were using high daily doses of oral CS (> 7.5 mg), while patients on low doses had the lowest prevalence of use. In contrast, 74.6% of patients on inhaled CS were using low daily doses (< 500 μg), followed by medium (13.9%) and high doses (11.5%), with some overlap among older patients. Post hoc analysis regarding asthma and COPD CS users showed that most asthma patients were prescribed low doses, whilst COPD patients older than 45 years were prescribed high or medium doses (Figure S5). Stratification of the KM plots according to comorbidities at baseline (Figure 4) showed that 50% (HR: 0.50; CI: 0.51–0.55) of all patients with the examined comorbidities were most likely to remain on continuous treatment after 28 days. Post hoc analyses indicated that the results were sensitive to changes in treatment duration, suggesting that longer treatment episodes may increase the probability of persistence (Figure S6). Among inhaled users, patients with COPD remained on continuous treatment for longer periods compared to patients with asthma.

Finally, examination of combined treatment patterns, including patients who were prescribed oral and inhaled CS, showed that most patients were prescribed solely oral formulations, with 6.9% (133524) of the studied population receiving concurrently oral and inhaled CS. Sensitivity analyses using June 30th as an audit date between 2005 and 2021 showed slightly different prescription rates over the study period with no impact on the overall inferences (Table S1).

4 Discussion

We found that 20% of users of oral and inhaled CS contributed 80% of the total prescribing in general practice. In addition, KM plots showed that 50% of patients who were prescribed inhaled formulations were treatment persistent for 2 months until their first treatment gap. Almost 20% of oral users were treatment persistent for a month. The PPC method showed that less than 20% of oral and 50% of inhaled CS users were covered by treatment during the first year of follow-up. We also observed that the prevalence of high-dose oral CS use increased significantly with age, peaking in the 66–75 age group for both men and women. Medium dose also rose with age but to a lesser extent, while low-dose use remained consistently low across all age groups, with the increase in prevalence for high and medium doses being more pronounced in women compared to men. Finally, patients with inflammatory bowel disease, rheumatoid arthritis, systemic lupus erythematosus, and respiratory conditions were most likely to remain on continuous oral CS treatment, while asthma patients on inhaled CS had shorter treatment periods compared to COPD patients.

Multiple methods allow the examination of treatment persistence using EHRs [20]. For instance, treatment persistence can be assessed using KM plots and the PPC method, although only the PPC method is often applied, whilst skewness in data dosage can be visually illustrated using Lorenz curves [20]. A previous study that examined skewness of CS exposure and focused on oral formulations for treating asthma found that 10% of patients was accounting for more than 50% of dispensed CS [21]. Systematic reviews of the methods used to assess adherence using EHR databases indicate a lack of clear consensus regarding the methodology, concluding that, ideally, both treatment duration and implementation should be considered. This is because in the PPC method, patients with longer treatment gaps that might indicate low implementation are excluded [22, 23]. This means that patients can cycle in and out of the numerator as many times as they are prescribed CS. In contrast, using the KM plots, patients with longer treatment gaps are censored. This allows patients with low implementation to be included in the analysis, which may influence the observed patterns in the PPC curves. Therefore, the extent to which the observed patterns that include patients with low implementation are representative depends on the representation of those patients at the population level. Finally, in our study, changes in grace periods showed some variation, and this is important to consider, especially when the observation period is shorter than a year. The results from the PPC method were less sensitive to changes in grace periods compared to the KM plots.

In this study, we also observed that the use of high and medium oral CS doses increases with age. A previous study that examined the use of oral CS in the UK found that the highest use was by patients between 70 and 79 years, and patients with arthropathies were most likely to receive uninterrupted CS treatment [4]. Our findings generally align, but some key factors that might be responsible for the differences in treatment continuation according to comorbidities include the study's observation period, the population and sample size, differences in person-years of follow-up, and the definition of treatment episodes, or grouping of disease categories [2, 24]. In the sensitivity analysis for patients prescribed both oral and inhaled CS over time, crude rates appear to be declining, likely due to the increased availability and use of biologics for inflammatory conditions. However, our findings showed that 20% of users were associated with the highest prescribing, which emphasizes the importance of identifying these patients and supporting interventions to minimize potential risks in the future.

Although duration data for oral formulations were imputed using a published algorithm [15], and duration data pertaining to inhaled CS were determined based on the published literature [17], some degree of exposure misclassification is possible. For example, standard metered-dose inhalers, such as beclomethasone, typically contain 200 doses. The maximum recommended dose for most CS inhalers is two inhalations twice daily, equating to four inhalations per day. Under conditions of full implementation, this regimen would result in an inhaler lasting approximately 50 days for most patients. Conversely, some patients may implement a lower dosing regimen, such as one inhalation twice daily, in which case an inhaler would last up to 100 days, which is the case for more than 30% of the patients, as the KM plot shows. Furthermore, in clinical practice, the majority of oral CS courses are prescribed for individuals experiencing exacerbations of asthma or COPD. Typical regimens for acute exacerbations include 6 to 8 tablets of 5 mg prednisolone daily for 5 to 7 days, resulting in a total course of 30 to 40 tablets. Similarly, another common regimen consists of 8 tablets of 5 mg prednisolone taken once daily for 5 days. A 30-day treatment duration was assumed to delineate episodes of inhaled CS according to the literature, with a 30-day grace period applied to capture late refills. Other published studies have utilised 60- or 90-day episodes [25], however, we believe that this matter should be examined in sensitivity analyses. Additionally, the methods that were applied to handle duplicated and overlapping prescriptions may also have contributed, to some degree, to exposure misclassification. A previously conducted cross-sectional study by Joseph et al. [26] utilised oral and injectable CS prescriptions from 78 patients and compared self-reported CS use to prescription-based CS exposure using the CPRD. The study found that 84.2% of patients receiving oral CS were correctly classified as being exposed (‘on treatment’) to CS and 87.5% were correctly classified as unexposed (‘off treatment’), whilst the concordance was low for patients receiving injectable CS, indicating that EHRs per se are not a reliable source to utilise exposure to injections [26]. This was the reason why our approach regarding injections was purely descriptive. The investigators also examined the impact that exposure has had on the relative risk of a hypothetical population and found that exposure misclassification may underestimate the risk for studied outcomes [26].

We acknowledge some potential limitations. For instance, data on medication supply were not available because there is no linkage between prescribed and dispensing data, nor were medication adherence data available. Furthermore, since adherence data were not available, we were unable to examine whether the medications were consumed as prescribed. This is a common limitation among studies that utilize EHRs. Moreover, the index date was assumed to represent treatment initiation, which might not necessarily indicate the exact date that patients initiated treatment; however, we accounted for this potential limitation by including patients who had at least 1 year of continuous registration with their general practice.

Another possible limitation is that patients with chronic conditions, severe diseases, or multimorbidity that need complex therapeutic schemes might have been exposed to CS for longer periods compared to patients who were prescribed CS occasionally (e.g., skin or lung infections). This is expected to influence the observed treatment persistence and/or combined treatment patterns, with patients treated for longer periods being those who have chronic medical conditions. A study demonstrated that users of oral CS with arthropathies remained on treatment for longer periods compared to patients with COPD [4]. Patterns of combined treatment that involved drugs other than CS have not been considered, nor were the reasons that might have led to early treatment discontinuation or switching to therapeutic alternatives (e.g., due to adverse events) investigated. Topical CS and other formulations used as short-term therapies including intranasal, ocular, ear drops, or rectal CS were not examined due to recognised difficulties in determining treatment duration [27]. Future research should consider whether the observed treatment persistence patterns insist for patients with multiple chronic conditions and those who are prescribed complex therapeutic schemes.

In conclusion, we found that a fifth of patients receiving CS accounted for over 80% of oral and inhaled CS prescribing in primary care. Those patients are more likely to be older, with comorbidities including inflammatory bowel diseases, respiratory disease, or rheumatoid arthritis and systemic lupus erythematosus. Identification of those patients will be important for targeting future interventions to promote patient safety in the use of CS in a cost-effective manner.

Author Contributions

D.M.A. conceptualised the study. E.D., D.M.A., and M.J.C. contributed to study design. E.D. developed the dataset and analysed the data. E.D. and M.J.C had full access to and verified the underlying study data. E.D. drafted the manuscript. All authors contributed to the interpretation of data. All authors reviewed results, revised the manuscript, approved its final version, and agreed to be accountable for all aspects of the work.

Ethics Statement

The study was approved by the CPRD Independent Scientific Advisory Committee (reference: 22_001811). No further ethical approval was needed.

Conflicts of Interest

A.J.A. is the National Clinical Director for Prescribing for NHS England. The other authors declare no conflicts of interest.