Medication adherence and the risk of severe asthma exacerbations: a systematic review

Introduction

Asthma is a chronic inflammatory airway disease with a high prevalence, around 10% in children and 5% in adults in Western countries [1, 2]. Asthma is a major cause of disability and health resource utilisation, and reduces quality of life [3]. This is partly caused by asthma exacerbations, which have a huge impact on patients and their families. To minimise asthma exacerbations, treatment should be adjusted stepwise, driven by the patient’s asthma control level [4]. Asthma treatment includes daily use of a controller drug and use of short-acting bronchodilators when needed for quick symptom relief [5]. Adherence to treatment is essential to optimise the benefits of therapy. Poor adherence has been associated with outcomes like mortality [6], asthma symptoms [7], direct and indirect costs of care [3] and quality of life [8]. In asthma, adherence to treatment tends to be poor, with rates of <50% in children [9] and 30–70% in adults [4, 10, 11], depending on country, age, sex and ethnicity [12]. These low adherence rates have been attributed to safety concerns about inhaled corticosteroids (ICS) (“steroid phobia”) by both the patients and the caregivers [13]. Indeed, use of ICS has been associated with growth impairment in children and other systemic adverse effects, such as an increased risk of pneumonia [14]. In addition, most ICS need to be administered twice daily, which increases the risk of poor adherence compared with once-daily administration [5]. It has been suggested that poor adherence to ICS increases the risk of exacerbations. However, the literature on this topic is conflicting. With this systematic review, we aim to provide a critical appraisal of the literature, examining the association between adherence to asthma controller therapy and the risk of severe asthma exacerbations in children and adults.

Results

Overview of the included studies

The search strategy identified 2319 articles. Upon title and abstract review, 2268 of these were excluded, mainly for the following reasons: lack of information on medication adherence, no severe asthma exacerbation or no evaluation of the relationship between adherence and exacerbation (n>1000); articles on intervention, management or implementation strategies (n=370); audit/guidelines (n=122); focus on other diseases (n=139); animal studies (n=106); cell biology/immunology (n=62); not original articles (n=40). A total of 23 articles were finally included in the review (fig. 1).

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 1

Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) [20] flow diagram describing the steps for including studies in the review. ICS: inhaled corticosteroids; LABA: long-acting β₂-agonist.

Details of the 23 studies are shown in table 1. Briefly, all 23 studies were published between 1993 and 2013. The sample sizes ranged from 24 to 97 743 individuals [9, 33]. Two studies were multicentre, seven were single-centre and 14 were based on healthcare (pharmacy/insurance/claim) databases. Most studies were from the USA. The included studies mainly used a cohort design (n=19), with the remainder utilising a cross-sectional (n=1) [9], case–control (n=1) [30] or randomised design (n=2) [24, 40]. Studies analysed adherence rates over a follow-up period ranging from 13 weeks [9] to 4 years [40]. 10 studies included only children, seven studies included both adults and children/adolescents, and six studies included only adults.

View this table:

• View inline
• View popup

TABLE 1

Characteristics of the 23 studies using objective measures to assess adherence

Measures of medication adherence

The assessment of medication adherence varied across studies. Most studies (n=11) used the medication possession rate (MPR) as a measure, which is the number of days of medication supplied divided by the number of days between the first and the last refill [43]. Most studies chose a fixed time-frame for the refill interval rather than using the last refill as the end-point for the refill interval, and did not exclude the last refill. MPR calculated across multiple refills is also called the continuous measure of availability. In addition to the MPR, two studies also calculated a controller-to-total ratio [21, 27]. Williams et al. [28] used a unique method to calculate adherence, by calculating a moving 6-month average ICS adherence for each day of follow-up. Furthermore, six studies used number of (requested) refills in a certain period to assess adherence. Of the remaining studies, three used canister weighing/counting and three used electronic monitoring via a specific device.

Measures of severe asthma exacerbations

Of the 23 studies using objective measures to assess adherence, five defined exacerbation as emergency department visit and/or hospitalisation and/or oral corticosteroids (OCS), and 16 defined exacerbation as two out of three of these criteria. Five studies defined the need for OCS as a separate outcome. Only one study had need for OCS as the only outcome.

Asthma diagnosis

In four studies, the asthma diagnosis was based on physician diagnosis. In the remaining studies, diagnoses were based on disease codes (International Classification of Diseases, 9th revision (ICD-9)) (n=9), disease codes and prescriptions (n=2), asthma controller therapy prescriptions (n=3), guidelines (n=4), or study-specific asthma definitions (n=1).

Adherence and exacerbations

Due to differences in design, exposure, outcome, cut-off values, assessments and definitions, studies differed substantially in effect magnitudes and this precluded a formal meta-analysis with pooling of results. Therefore, we report the findings separately for each type of objective adherence measure. The order of reporting within each category is based on quality score, indicated with the letter q. We reported results separately for children and adults. Details of the 23 studies included in this review are summarised in table 1. The overall conclusions remained similar when papers with low quality (q<6) were excluded.

An overview of the range of risk estimates, for those studies that reported risk estimates for the association between adherence and asthma exacerbation, is shown in figure 2 (for children) and figure 3 (for adults).

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 2

Overview of paediatric studies publishing odds ratios (OR), sorted by quality score, number of participants and year published. Hosp: hospitalisation for asthma; ED: emergency department visit for asthma; CTR: controller-to-total medication ratio; Comb: combined; MPR: medication possession rate; OCS: oral corticosteroids.

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 3

Overview of adult studies publishing odds ratios (OR) or relative rate (RR), sorted by quality score, number of participants and year published. Comb: combined; ED: emergency department visit for asthma; hosp: hospitalisation for asthma; MPR: medication possession rate; perc: percentile; OCS: oral corticosteroids; CMA: continuous measure of availability; Adh: adherence.

Objective adherence measurements

To assess overall treatment adherence, 17 studies used refill data; of these, 11 used MPR measurements [21, 22, 25–29, 31–33, 35] and six used number of prescription refills [23, 24, 30, 34, 36, 37]. The remaining studies used either electronic monitoring devices (n=3) [9, 38, 39] or weighted canisters (n=3) [40–42].

The overall treatment adherence was low in paediatric and adult studies. In children, adherence measured as average MPR was only 20–33.9% for ICS [26, 38]. The number of prescription fillings over the course of 1 year ranged from 4.7 to 5.5 times for fluticasone [24, 37]. In adults, MPR for ICS ranged between 15% [26, 31] and 54% [32].

Williams et al. [28] demonstrated, with a moving 6-month average ICS adherence for each day of follow-up, that adherence to ICS medications began to increase just before the first asthma exacerbation, and continued to increase after the event.

Discussion

To our knowledge, this is the first systematic review on the association between treatment adherence and severe asthma exacerbations. In the identified articles, we observed that low adherence was common. Despite heterogeneity among studies in terms of definitions of adherence and asthma exacerbations, the majority of the high-quality studies consistently reported an association between low adherence and higher risk of severe asthma exacerbations, both in adults and children.

We identified important differences and limitations in the included studies. Notably, the majority of studies were classified as moderate to low quality, largely because of flaws in the study methods, e.g. small sample sizes, which may compromise power, and cross-sectional designs that are prone to bias. When excluding studies of poor or moderate quality (q<6) or excluding studies that used a design other than cohort design, the overall conclusion remained the same, namely that good adherence was associated with fewer severe asthma exacerbations.

Another limitation was that the included studies varied widely with regard to the definition of adherence. There is no standardised method to measure adherence, and each measure has its own strengths and limitations [11, 44, 45]. Objective measures are considered to be more reliable and accurate than indirect measures [11]. Electronic device monitoring is usually considered the gold standard because of a detailed assessment of adherence patterns. Although it has the benefit of being able to identify and exclude “dumping” (deliberate emptying of inhaler before study visits to conceal nonadherence) from analyses, unfortunately such monitoring is expensive and often prone to device failure [11, 35]. The most cost-effective method to assess adherence is by self-report. However, the reliability of this method is questionable [46], as it is well known that patients over-report their adherence [9, 11, 47], even in a clinical trial setting [40]. Hence, self-report was not included in this study.

The majority of studies calculated the MPR, which is a common way to measure medication adherence from claims or pharmacy data that has been found to be useful and reasonably accurate [48, 49]. MPR data have the advantage of being easily accessible and inexpensive [45]. The drawback of these databases is that details about devices and day-to-day patterns of adherence are not always recorded [11, 45].

Another method used to measure adherence was the number of prescriptions, but this is highly influenced by the level of asthma control. Indeed, patients appear to self-titrate their medication, showing greater adherence to therapy when they have a worse level of control [27]. Additionally, the number of prescriptions is also influenced by the quality and reliability of physicians prescribing. The best method to tease this out is by using a ratio of long-term controller to total asthma medications, in which the denominator and numerator increase with increasing severity, but the numerator is more specifically reflective of adherence to therapy [21].

Even if studies used a common methodology to assess adherence, a wide variety of cut-off values to define “adherent patients” was observed. This arbitrary selection of adherence cut-off points is of considerable concern. Traditionally, medication adherence is dichotomised using a cut-off value (e.g. 80%), which is derived from studies in other chronic diseases such as HIV and hypertension [50] but does not necessarily hold for asthma.

It was quite remarkable that most studies did not attempt to measure whether and how medication was actually taken. Inhaler competence, the skill to inhale correctly, is particularly relevant for asthma medication, as inhaling of drugs requires considerable skill and practice [51]. Even if medication is taken daily, deposition in the lungs will be low with incorrect inhalation technique [52]. Newer devices such as breath-actuated inhalers and smart nebulisers, which impose breathing patterns and record lung deposition, may partly overcome this problem, but still require a high level of cooperation. Smart nebulisers are expensive and the experience in children with asthma is limited [53]. Nevertheless, all of these methods can overestimate adherence and there remains a number of complex issues affecting optimal medication use in children with asthma, e.g. a lack of parental knowledge about asthma medications, parental beliefs and fears, and the child’s self-image [54].

We noted heterogeneity in outcome definitions, despite international guidelines on the definitions and assessment of severe asthma exacerbations [55]. This might partly be explained by the data being used, for example pharmacy data do not always have detailed information on disease codes and/or symptoms and diagnosis. As asthma is a difficult diagnosis, especially in children, risk of asthma misclassification is high if based on prescription data only.

In a pragmatic trial, patients are randomised within a real-life setting that combines the sound methodology of randomised controlled trials with daily practice, enlarging the external validity of study results. This external validity is crucial for studies investigating treatment adherence and associated factors [56]. All methods have their sets of benefits and limitations [57]. It has been suggested that the best design would probably be a combination of observational studies, pragmatic trials and randomised controlled trials, as all have advantages and drawbacks and they are not designed to answer the same questions [58].

Some of the studies were at risk of bias. There are limitations arising from the observational nature of the included studies. One important bias is the “healthy adherer effect”, where healthier people are predisposed to follow medical advice, because they are more concerned about their health. Other biases in this type of research are recall bias and observer/investigator bias when using questionnaires [59]. Furthermore, misclassification of adherence caused by dumping or over-reporting might have contributed to an underestimation of the role of adherence in mediating the observed treatment effects.

Another limitation was that some studies did not adjust for potential confounders. “Asthma severity” is an important confounder that is difficult to control for, as details on asthma severity are often lacking in database studies. Other potential confounders are environmental factors, such as pollution or smoking habits of the parents.

Although the majority of papers of good quality indicated that higher levels of adherence were associated with a reduced risk of severe asthma exacerbations, this was not confirmed in all studies. Some studies even reported an inverse association between treatment adherence and risk of severe asthma exacerbations. This might be explained by the fact that the number of prescriptions was used as a proxy for adherence [35, 37], or because unusual cut-off values for adherence were used [34]. Furthermore, there are general potential explanations for this inverse association. Treatment need is higher in patients with poorly controlled asthma, who are by definition at risk of asthma exacerbation. This higher need for treatment will result in increased prescription of asthma controller therapies, with a consequently higher MPR [37, 46]. Another explanation is self-titrating of medications: patients show greater adherence to therapy with worse levels of control, thus positively influencing the MPR [27]. A third explanation could be the heterogeneity among asthma patients in treatment response: some patients reduce their prescribed controller medication without negative consequences [37], whereas other patients continue to have poor outcomes despite good adherence [60]. Additionally, there is no known dose of medication or duration of treatment. Therefore, the “low” adherence shown by some people may be adequate for them most of the time.

The main strengths of this review are the comprehensiveness of the searches and our standardised approach to study selection, data extraction and quality assessment. However, our review has some limitations that need to be considered. Only published data was included, and the relatively small number of relevant publications suitable for this review warrants caution given the possibility of publication bias. Publication bias is a type of error that may affect the results of a meta-analysis because studies with statistically significant positive findings are more likely to be published than studies with negative results [61]. However, given the heterogeneity, inclusion of additional studies would have been unlikely to undermine our overall conclusion.

Overall, this review highlights the importance of adherence to prevent severe asthma exacerbations. This is in agreement with recent predictive models, which observed that improvements in medication adherence could lead to significant improvements in asthma outcomes [62]. It is clear that better studies about adherence to treatments are needed, and that more homogeneous outcomes and study design would be useful in order to reach conclusions. For clinicians, it seems evident that efforts directed at improving, evaluating and measuring adherence should be a routine component of asthma care [63]. The recognition of nonadherence is an important first step towards optimal asthma control.