A descriptive cohort study of withdrawal from inhaled corticosteroids in COPD patients.

Inhaled corticosteroid (ICS) therapy is widely prescribed without a history of exacerbations and consensus guidelines suggest withdrawal of ICS in these patients would reduce the risk of side effects and promote cost-effective prescribing. The study describes the prescribing behaviour in the United Kingdom (UK) in relation to ICS withdrawal and identifies clinical outcomes following withdrawal using primary and secondary care electronic health records between January 2012 and December 2017. Patients with a history ≥12 months' exposure who withdrew ICS for ≥6 months were identified into two cohorts; those prescribed a long-acting bronchodilator maintenance therapy and those that were not prescribed any maintenance therapy. The duration of withdrawal, predictors of restarting ICS, and clinical outcomes were compared between both patient cohorts. Among 76,808 patients that had ≥1 prescription of ICS in the study period, 11,093 patients (14%) withdrew ICS therapy at least once during the study period. The median time without ICS was 9 months (IQR 7-14), with the majority (71%) receiving subsequent ICS prescriptions after withdrawal. Patients receiving maintenance therapy with a COPD review at withdrawal were 28% less likely to restart ICS (HR: 0.72, 95% CI 0.61, 0.85). Overall, 69% and 89% of patients that withdrew ICS had no recorded exacerbation event or COPD hospitalisation, respectively, during the withdrawal. This study provides evidence that most patients withdrawing from ICS do not experience COPD exacerbations and withdrawal success can be achieved by carefully planning routine COPD reviews whilst optimising the use of available maintenance therapies.

Introduction

International clinical guidelines on the management of Chronic Obstructive Pulmonary Disease (COPD), Global Initiative for Chronic Obstructive Lung Disease (GOLD) 2020[1-3] recommend long-acting inhaled bronchodilators, including β2-agonists (LABA) and anti-muscarinic agents (LAMA) as maintenance therapies. These agents can be prescribed as a monotherapy dose, fixed dose of dual LAMA/LABA therapy, or in combination with inhaled corticosteroids (ICS) for the symptomatic management of COPD and the prevention of COPD-related exacerbations.

The GOLD treatment strategy recommends that ICS (prescribed as a combination inhaler with a LABA or as part of triple-therapy regimen with LAMA and LABA) are reserved for COPD patients with frequent or severe exacerbations, and research shows that certain features might identify patients more or less likely to respond well to treatment with ICS, including asthma-like phenotype or biomarkers such as eosinophilia[4-8]. However, considerable dissociation has been reported between guideline recommendations and clinicians’ practices[9-11] despite significant efforts to promote and disseminate the GOLD strategy[12]. ICS are often inappropriately prescribed for patients with mild or moderate COPD without a history of exacerbations resulting in up to 70% of patients in current practice receiving ICS-containing regimens[13,14]. There is a lack of evidence of the benefit of ICS in patients with preserved lung function and no history of exacerbations[15,16]. Safety concerns, particularly regarding increased incidence of pneumonia, osteoporosis, bruising, adrenal suppression, and diabetes have been reported in ICS users[17].

It has been proposed that ICS should be withdrawn in patients who may have been prescribed this therapy inappropriately, to reduce side effects and promote cost-effective prescribing[18-20]. The recently published European Respiratory Society (ERS) guidelines on ICS withdrawal suggest patients withdrawing ICS should receive ongoing maintenance therapy with a bronchodilator[21]. While guidelines and clinical opinion have been shared around how to withdraw ICS[22-25], there is limited evidence on the extent to which withdrawal recommendations have been adopted in primary care[26]. The aim of this study is to describe the trends in ICS withdrawal and the outcomes for patients receiving long-acting bronchodilator maintenance therapy (LAMA or LABA monotherapy or combination dual therapy), compared to patients not receiving any maintenance bronchodilators. Secondly, this study will aim to identify patient characteristics associated with successful withdrawal of ICS.

Methods

Study design and Inclusion criteria

This retrospective cohort study included patients belonging to UK general practices contributing data to the Clinical Practice Research Datalink (CPRD), the world’s largest longitudinal primary care database that is utilised for pharmacoepidemiologic research[27,28]. The data encompasses approximately 60 million patients; GPs voluntarily contribute the data which is automatically transferred from the routine patient care and it is representative of the UK population[27].

Included patients in this study were those with a confirmed COPD record for adults ≥35 years of age at diagnosis, where the diagnosis of a COPD was defined using any record of a diagnostic Read Code for COPD in primary care records or COPD ICD-10 code in hospital records. The accuracy of COPD diagnosis in CPRD has been previously validated[29]. Patients were included if they were withdrawing a long-term ICS prescription between January 2012 to December 2017. To identify a group of patients with long-term ICS prescription, we required a minimum historical period of 12 months’ persistent exposure to ICS therapy (alone or in any combination) after allowing for the therapy period (calculated using amount prescribed or an imputed 28 days where data were missing). Additionally, patients were required to be currently registered in the observational period (January 2012 to December 2017) in general practices that contain research quality data linked with patient and practice-level secondary care hospital data in the Hospital Episode Statistics (HES) database. The final sample size for the study was based on the above basic eligibility criteria in addition to ICS-withdrawal requirements.

Definition of ICS withdrawal and data capture

Since gaps in prescription records of a few months are common within CPRD, a minimum ICS-free period was required to identify true withdrawal rather than a short period during which patients did not receive prescriptions. For the purposes of analysis, we therefore defined “withdrawal” as a period of at least 6 months with no record of ICS prescription. Patients’ withdrawal period was then considered from the index date—defined as the last date of ICS therapy based on the duration of medication prescribed, until the earlier date of subsequent ICS prescription (alone or in any combination), study end, acceptable research quality data (patient remains registered in GP practice without any data gaps), transferring out of the registered practice, or death. For patients with multiple periods of ICS withdrawal, we considered the first occurring period. Data capture consisted of a 2-year period prior to ICS withdrawal and minimum 6 months period after the withdrawal date to identify clinical outcomes.

Outcomes of interest

Outcomes of interest in this study included time without ICS; the number of exacerbations (defined as antibiotic and oral corticosteroid (OCS) prescriptions taken together for 5–14 days or lower respiratory tract infection Read Code, or an acute exacerbation Read Code where that did not coincide with a spirometry test)[30] during the patient’s ICS-withdrawal period; the number of COPD hospitalisations (ICD-10 code J44.0, J44.1 or J22 in any position of hospital episode); the number of interactions with a general practitioner (GP); the number of short-acting bronchodilator prescriptions (i.e. rescue therapy); and number of pneumonia episodes (Read Code and ICD-10 codes J12-J16 and J18). Measurements of these outcomes were taken from distinct patient groups (patients receiving bronchodilation maintenance therapy compared to patients not receiving bronchodilation therapy).

Patient characteristics and covariates

Baseline characteristics included age at the start of withdrawal, sex, BMI, smoking status, vaccination history (influenza or pneumococcal vaccination), co-morbidities (heart disease or asthma), prior history of COPD exacerbations, recent eosinophil count and percent predicted forced expiratory volume in 1 s (FEV1). In addition, the coding of routine COPD reviews (6 or 12-month intervals) with spirometry procedure on the day was also recorded. All clinical and demographic covariates were collected at baseline using the primary care electronic health data in CPRD.

Two distinct patient groups were defined—those receiving one or more long-acting bronchodilator maintenance therapies during the ICS-withdrawal period, and those with no long-acting bronchodilator treatment during the period of ICS withdrawal. Long-acting bronchodilator treatment in the maintenance group consisted of LAMA or LABA monotherapy, LAMA/LABA combination fixed dose, or LAMA + LABA combination free dose which includes separate inhalers prescribed for intended regular daily usage, not for as-needed use[31].

Statistical analysis

Baseline characteristics and follow-up patient outcomes were described for patients receiving or not receiving inhaled maintenance therapy during their ICS-withdrawal period. Cross-tabulated summary statistics were used to describe these patient characteristics and included mean ± standard deviation (SD) for continuous data and number (percentage) for categorical data. Missing data for covariates were quantified without imputation due to the relative percentage of missing data.

Kaplan–Meier plots to describe the probability of ICS-free time (months) and median ICS-free time (months) were calculated in patients that were prescribed a maintenance therapy compared with those without maintenance therapies. The frequency and percentage of patients lost to follow-up were also quantified by corresponding reasons for the loss to follow-up. A univariable Cox-proportional hazards model was initially performed to identify significant predictors of restarting ICS therapy at the level of α = 0.05. These significant predictors were used to develop a two-sided multivariable predictive model using backwards variable selection at the level of α = 0.2 to produce a parsimonious complete-case model adjusting for confounders. Hazard function (hazard ratio) for patients on maintenance therapy and for patients without maintenance therapy after withdrawal are assumed to be proportional, with a constant hazard ratio over time. Potential interactions were also tested using the final multivariable model. For these analyses, those who received their first maintenance therapy prescription after the first 6 months of withdrawal were considered not on maintenance therapy at the index to avoid immortal time bias.

In a sensitivity analysis, we used a time-dependent variable, which changed from no maintenance therapy to the use of maintenance therapy in patients who received their first maintenance therapy after the first 6 months. In an additional sensitivity analysis, the median duration of ICS-free time was repeated based on a minimum 9 months of ICS-free time prior to the index date to understand whether a longer baseline ICS-free time would influence the number of exacerbations in patients receiving maintenance treatment or not receiving maintenance treatment.

The analyses were performed in accordance with relevant regulations and guidelines. This study was reviewed and approved by the Independent Scientific Advisory Committee for Medicines and Healthcare products Regulatory Agency (MHRA) database research (ISAC number 17195RA) and by an internal scientific committee of the study sponsor. As this was a non-interventional study using anonymised data, no patient consent was necessary. Primary and secondary database access and data extraction were restricted to the corresponding author who was also responsible for the creation of the final study population and statistical analysis. Raw data extracts were assessed for outliers and cleaned to provide a clinically and statistically complete dataset. Data linkage between primary and secondary care data was performed by merging and appending based on a common patient identification number for both databases. Covariates were stratified by clinical thresholds and aggregated for n < 10 patients. All cleaning methods and analyses were conducted using SAS (SAS Institute Inc., Cary, NC, USA) and STATA/IC 15.0.

Table 1

Characteristics of patients withdrawing ICS.

Characteristica	All patients (N = 11,093)		Patients receiving maintenance therapy (N = 3849)		No maintenance therapy prescribed (N = 7244)
Female	5857	52.8%	1896	49.3%	3961	54.7%
Mean ± SD age, years	70.0	15.6	71.5	11.0	69.1	13.3
35–44	351	3.2%	44	1.1%	307	4.2%
45–54	1101	9.9%	256	6.7%	845	11.7%
55–64	2019	18.2%	666	17.3%	1353	18.7%
65–74	3221	29%	1239	32.2%	1982	27.4%
75–80	1905	17.2%	762	19.8%	1143	15.8%
>80	2496	22.5%	882	22.9%	1614	22.3%
Mean ± SD body mass index (BMI) kg/m2	28.1	6.6	28.0	6.6	28.2	6.6
Missingb	2914	26.3%	431	11.2%	2483	34.3%
Smoking status
Current smoker	2920	26.3%	1115	29%	1805	24.9%
Non-smoker/never	1025	9.2%	236	6.1%	789	10.9%
Ex-smoker	5349	48.2%	2330	60.5%	3019	41.7%
Data not entered	1799	16.2%	168	4.4%	1631	22.5%
Mean ± SD years since COPD diagnosis	5.1	6.3	6.8	6.2	4.3	6.2
Missingb	484	4.4%	180	4.7%	304	4.2%
Mean ± SD latest stable blood eosinophil count (absolute count × 109 cells/L)	0.23	0.15	0.23	0.15	0.23	0.15
Missingb	2718	24.5%	888	23.1%	1830	25.3%
Mean ± SD latest FEV1 % of predicted	65%	21.2%	61%	20.4%	67%	21.4%
GOLD 1— ≥80	1561	14.1%	517	13.4%	1044	14.4%
GOLD 2—50–79	3250	29.3%	1538	40.0%	1712	23.6%
GOLD 3—30–49	1358	12.2%	724	18.8%	634	8.8%
GOLD 4—<30	587	5.3%	152	3.9%	135	1.9%
Missingb	4637	41.8%	918	23.9%	3719	51.3%
MRC Dyspnoea Score
1–2	3328	30.0%	1334	34.7%	1994	27.5%
≥3	3382	30.5%	1827	47.5%	1555	21.5%
Missingb	4383	39.5%	688	17.9%	3695	51.0%
CAT Score
<10	149	1.3%	80	2.1%	69	1.0%
≥10	394	3.6%	216	5.6%	178	2.5%
Missingb	10,550	95.1%	3553	92.3%	6997	96.6%
Exacerbations (mean ± SD in year prior)	1.0	1.8	1.4	2.1	0.8	1.5
Hospitalised exacerbation	964	8.7%	424	11.0%	540	7.5%
Non-hospitalised exacerbation	4311	38.9%	1859	48.3%	2452	33.8%
GOLD Group (2020)
A	2720	23.6%	1032	26.8%	1688	23.3%
B	2055	17.8%	1072	27.9%	983	13.6%
C	609	5.3%	303	7.9%	306	4.2%
D	1329	11.5%	755	20%	574	7.9%
Undetermined	4380	38.0%	687	17.8%	3693	51.0%
Asthma history
>2 years before COPD diagnosis	4665	42.1%	1320	34.3%	3345	46.2%
≤2 years before COPD diagnosis or withdrawal date	3394	30.6%	1251	32.5%	2413	33.3%
No history	3034	27.4%	1278	33.2%	1486	20.5%
Pneumonia (year prior)	2136	19.3%	892	23.2%	1244	17.2%

aN,% unless otherwise stated.

bImputation methods were not conducted due to the high amount of missing data for stable blood eosinophil count, FEV1% GOLD groups, MRC dyspnoea scores, and CAT scores.

Table 2

Characteristics of withdrawal periods in patients withdrawal ICS.

Characteristica	All patients (N = 11,093)		Patients receiving maintenance therapy (N = 3849)b		No maintenance therapy prescribed (N = 7244)
Maintenance therapy prescribedb
LAMA + LABAc	~	~	265	6.9%	~	~
Mean months withdrawn ICS (95% CI)	~	~	20.2	18.4, 22	~	~
LAMA/LABAd	~	~	276	7.2%	~	~
Mean months withdrawn ICS (95% CI)	~	~	18.3	16.9, 19.8	~	~
LAMA only	~	~	2,965	77%	~	~
Mean months withdrawn ICS (95% CI)	~	~	13.6	13.2, 14.1	~	~
LABA only	~	~	345	9%	~	~
Mean months withdrawn ICS (95% CI)	~	~	14.7	13.3, 16	~	~
Withdrawal periods that end due to ICS prescription	7846	70.7%	2521	65.5%	5325	73.5%
Last ICS therapy is the same as the first	6515	83.0%	1894	75.1%	4621	86.8%
ICS therapy is different	1331	17.0%	627	24.9%	704	13.2%
Withdrawal periods coinciding with a COPD annual/6-monthly review	1929	17.4%	1132	29.4%	797	11.0%
At start of withdrawal	1394	12.6%	877	22.8%	517	7.1%
At end of withdrawal	690	6.2%	364	9.5%	326	4.5%

aN,% unless otherwise stated.

bIncluded patients observed to have clear periods of using different maintenance therapy regimens.

cFree-dose combination.

dFixed-dose combination.

Table 3

Univariable and multivariable model of time without ICS based on patient characteristics (hazard ratios >1 present increased risk of restarting ICS or shorter-ICS free time).

	Univariable analysis		Multivariable analysisa (N = 6009)
Characteristic	HR (95% CI)	P value	HR (95% CI)	P value
Age
35–44	Reference		Reference
45–54	0.94 (0.82, 1.08)	0.38	0.96 (0.74, 1.25)	0.75
55–64	0.95 (0.84, 1.08)	0.46	0.98 (0.77, 1.26)	0.88
65–74	0.91 (0.80, 1.03)	0.14	0.97 (0.75, 1.24)	0.80
75–80	0.90 (0.79, 1.02)	0.10	0.97 (0.75, 1.25)	0.82
>80	0.83 (0.73, 0.94)	0.003	0.94 (0.73–1.21)	0.63
Sex
Female	Reference		Reference
Male	0.96 (0.92, 1.0)	0.064	0.95 (0.89, 1.01)	0.12
Body mass index (BMI) kg/m2
Normal	Reference		Reference
Underweight	1.13 (0.95, 1.35)	0.16	0.98 (0.79, 1.21)	0.84
Overweight	1.11 (0.96, 1.28)	0.17	1.02 (0.85, 1.22)	0.86
Obese	1.03 (0.89, 1.19)	0.72	0.91 (0.76, 1.09)	0.32
Severely obese	1.12 (0.97, 1.30)	0.12	0.99 (0.83, 1.19)	0.94
Smoking cessation	0.97 (0.92, 1.01)	0.13	–	–
Latest stable blood eosinophil count (absolute count in 109 cells/L)	1.22 (1.03, 1.44)	0.024	–	–
FEV1% predicted	0.99 (0.99, 1.0)	<0.001	0.99 (0.99, 1.0)	<0.001
Exacerbations (number in year prior)	1.02 (1.01, 1.04)	0.033	1.04 (1.02, 1.06)	<0.001
Maintenance therapy during withdrawal period
No	Reference		Reference
Yes	0.85 (0.81, 0.89)	<0.001	0.95 (0.88, 1.02)	0.16
Asthma history
No asthma history	Reference		Reference
>2 years before COPD diagnosis	1.44 (1.37, 1.52)	<0.001	1.28 (1.18, 1.38)	<0.001
≤2 years before COPD diagnosis or withdrawal date	1.22 (1.15, 1.30)	<0.001	1.12 (1.04, 1.22)	0.004
History of heart disease
No	Reference		Reference
Yes	0.91 (0.86, 0.96)	<0.001	0.95 (0.88, 1.03)	0.19
Vaccination history
No	Reference		–	–
Yes	1.10 (1.01, 1.19)	0.026	–	–
Start of withdrawal coincides with a review
No	Reference		Reference
Yes	0.67 (0.62, 0.72)	<0.001	0.86 (0.76, 0.97)	0.01

aAge and sex covariates were included in the final model regardless of their threshold for the backward variable selection process.

Table 4

Modification of the effect of maintenance therapy group on restarting ICS therapy by the COPD review at the date of withdrawal.

	Date of withdrawal coincides with annual/6-month review period
	No		Yes
Treatment prescribed	HRa (95% CI)	P value	HRa (95% CI)	P value
No maintenance	Reference	–	0.86 (0.76, 0.97)	0.01
Maintenance therapy	0.95 (0.88, 1.02)	0.16	0.72 (0.61, 0.85)	<0.001

aHazard ratios correspond to the interaction between subgroups with patients that did not receive a COPD review and maintenance therapy as a reference group.