Abstract

Introduction Pulmonary arterial hypertension (PAH) and chronic thromboembolic pulmonary hypertension (CTEPH) are life-threatening conditions that can progress to death without treatment. Although strong medication adherence (MA) is known to enhance outcomes in chronic illnesses, its association with PAH and CTEPH was sporadically explored. This study aims to examine the MA of patients with PAH or CTEPH, identify factors associated with low adherence and explore the resulting outcomes.

Methods A systematic review was conducted by searching multiple databases (Medline, Embase, Cochrane Central, ClinicalTrials.gov, Scopus, Web of Science and Google Scholar) from 6 March 1998 to 6 July 2023. We included studies reporting MA as primary or secondary end-points. Study selection, data extraction and methodological quality assessment were performed in duplicate.

Results 20 studies involving 22 675 patients met the inclusion criteria. Heterogeneity was observed, particularly in the methods employed. MA means ranged from 0.62 to 0.96, with the proportion of patients exhibiting high MA varying from 40% (95% CI 35–45%) to 94% (95% CI 88–97%). Factors associated with low adherence included increased treatment frequency, time since diagnosis and co-payment. High MA seems to be associated with reduced hospitalisation rates, inpatient stays, outpatient visits and healthcare costs.

Conclusions This systematic review underscores the heterogeneity of MA across studies. Nevertheless, the findings suggest that high MA could improve patients’ clinical outcomes and alleviate the economic burden. Identifying factors consistently associated with poor MA could strengthen educational efforts for these patients, ultimately contributing to improved outcomes.

Introduction

Pulmonary hypertension (PH) encompasses a diverse range of pathophysiological disorders [1], with pulmonary arterial hypertension (PAH) and chronic thromboembolic pulmonary hypertension (CTEPH) standing out as rare, life-threatening conditions leading to right ventricular (RV) failure and death in the absence of treatment. The management of PAH and CTEPH involves a multifaceted approach, incorporating specific and nonspecific therapies, as well as adherence to lifestyle and dietary guidelines [1, 2].

Over the past two decades, significant strides in pharmaceutical advancements and optimised treatment strategies have markedly enhanced the survival and quality of life of patients with PAH and CTEPH. An increased comprehension of the pathogenic factors associated with PAH and CTEPH has paved the way for the development of targeted therapies, focusing on specific pathways, such as the prostacyclin, endothelin, nitric oxide and guanylate cyclase pathways, as well as calcium-channel blockers (CCBs) [3–5]. Current guidelines recommend using a combination of these treatments for PAH based on the risk status at the time of diagnosis and during follow-up [1, 6]. For CTEPH, these pharmaceutical agents should be integrated into a customised multimodal management plan for each patient [7]. Alongside these specific drugs, general measures and nonspecific medications play a crucial role in achieving optimal management [2]. Notably, maintaining long-term curative anticoagulation is critical in treating patients with CTEPH. In addition, avoiding fluid retention through diuretics is a key objective in managing patients with all forms of PH and oxygen therapy is acknowledged for its role in reducing pulmonary vascular resistance and enhancing exercise tolerance in patients with PAH [2].

Guidelines underscore the pivotal role of optimal medication adherence (MA) in effectively managing PAH and CTEPH, given the association between high adherence to medication in chronic diseases and improved health outcomes [8]. MA, defined as the extent to which a patient embraces their treatment and overall care [9–11], can be assessed through various direct and indirect tools [10, 12]. Regular monitoring of MA by a member of the multidisciplinary team is recommended to promptly identify adherence issues or any treatment changes initiated by the patient or nonspecialist physicians [1]. Multiple factors, including those related to the patient, their disease, their treatment, their relationship with the healthcare team and their economic status as defined by the World Health Organization, can influence MA [9, 13, 14].

This systematic review aims to 1) determine the extent of MA in patients with PAH and CTEPH, encompassing both PH-specific and nonspecific treatments, 2) provide an insight into the methodologies employed to assess MA, 3) investigate the factors linked to low or high levels of MA, and 4) explore the clinical outcomes associated with low MA.

MethodsReview protocol

This systematic review was conducted according to Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines [15]. The study protocol was registered in the International Prospective Register of Systematic Reviews (PROSPERO) on 30 January 2023 (registration number: CRD42023382341), with modifications detailed in the supplementary files.

Eligibility criteria

This systematic review encompassed all the studies that reported MA for specific and/or nonspecific treatments of PAH and CTEPH. The following methods were employed for MA assessment:

medication possession ratio (MPR): this measure gauges the percentage of time a patient has access to medication [16, 17];

proportion of days covered (PDC): this measure assesses the proportion of days during which a person has access to medication over a given period of interest [18, 19]; and

questionnaires: including the Morisky Medication Adherence Scale-8 (MMAS-8) [20], the Girerd Score [13, 21] and the Medication Adherence Report Scale (MARS-5) [22, 23], as previously described.

Given potential variations in defining appropriate or high MA between studies, we considered the thresholds defined by each study. We included randomised controlled trials (RCTs), cluster RCTs, controlled (nonrandomised) clinical trials or cluster trials, as well as prospective and retrospective comparative cohort studies, case–control or nested case–control studies and grey literature (abstracts and theses). Eligible studies had MA as either a primary or secondary objective and a follow-up time of at least 6 months. Excluded from the review were reviews, letters, commentaries/editorials, case series, case reports and preclinical/animal studies. Included studies were restricted to articles published in English and French. Additionally, studies involving patients under 18 years of age or those with psychiatric disorders were excluded.

Information sources and search strategy

On 6 July 2023, we conducted a comprehensive search starting from 6 March 1998 (the date of marketing authorisation for epoprostenol, the first specific treatment for PH). This extensive search covered seven databases, as follows: Medline via PubMed, Embase, Cochrane Central, ClinicalTrials.gov, Scopus, Web of Science and Google Scholar. The search involved a combination of medical subject headings and text words associated with each database. The search strategy was approved by a university librarian and two independent clinical pharmacists. Detailed search strategies are provided in the Appendix. In addition to the primary database search, we manually screened the selected studies and monitored newly published articles in PubMed through an iterative alert system to identify potential additional studies.

Study selection

The results of the literature search were imported into Rayyan software, an online platform designed to facilitate peer collaboration in the study selection process according to recommendations (supplementary data) [24]. Cohen's kappa coefficient was used to gauge the degree of agreement between the two reviewers. A coefficient between 0.60 and 0.70 was considered satisfactory, while a coefficient surpassing 0.70 was deemed very good. This approach aimed to enhance the reliability and consistency of the study selection process.

Data collection process

For each study included in the analysis, the following information was systematically collected:

general characteristics (authors, year of publication, country of origin, publication type);

study design and sample size;

population characteristics: clinical disease, age, sex and PH type;

specific treatments: phosphodiesterases type 5 inhibitors (PDE5i), endothelin receptor antagonists (ERA), stimulator of soluble guanylate cyclase (sGCS), prostacyclin analogues (PAN), prostacyclin agonist (PAG) and CCBs;

nonspecific treatments: oxygen therapy, diuretics and anticoagulants;

methods of MA assessment: MPR, PDC and questionnaires;

MA results: mean±standard deviation (sd) (MPR, PDC), number and proportion of patients with a high level of MA (MPR, PDC and questionnaires);

factors associated with MA: polypharmacy, age, sex, comorbidities, number of PH-specific and nonspecific treatments, and liver and renal functions; and

all reported patient-important outcomes: adverse event (AEs), hospitalisation/rehospitalisation occurrences, economic cost and quality of life.

Quality assessment and risk of bias

Two reviewers conducted an independent evaluation of the risk of bias in each study using the Quality Assessment Tool for Quantitative Studies (QATQS) developed by the Effective Public Health Practice Project [25]. This tool is recommended for assessing the risk of bias in studies, whether randomised and controlled or not [26]. In the event of disagreement, it was pre-determined that a third reviewer would be consulted to reach a consensus. The QATQS evaluates five sources of bias: selection bias (D1), study design (D2), data collection method (D3), withdrawal and drop-outs (D4), and statistical analyses (D5). The risk of bias for each component is rated as strong, moderate or weak. Subsequently, each study receives a global rating of low, with some concerns or high risk of bias. A low risk of bias is assigned if there are at least two strong components without a weak component, a study with some concerns is defined if there is only one weak component and a high risk of bias if there are two weak components [27]. Cohen's kappa coefficient was employed to assess the level of agreement between the two reviewers.

ResultsResults of the literature search

From the initial search across seven databases, we identified 717 references. After removing 73 duplicates, we screened 644 studies based on titles and abstracts. At this stage, 589 irrelevant studies were excluded, leaving us with 55 publications for further review. Following a thorough examination of the full texts, 20 studies met the eligibility criteria. Two studies were excluded due to a lack of information on the calculation method of MA. Additionally, we checked for potential supplementary publications. No additional study was identified through the citing analysis, while alerts from the screened databases identified two other studies. Consequently, a total of 20 studies were included in the present analysis (figure 1). The level of agreement between reviewers was considered good, as evidenced by a Cohen's kappa coefficient of 0.67 (95% CI 0.55–0.79).

FIGURE 1

Flow diagram of the selection process.

General study characteristicsStudy design

The characteristics of the selected studies are presented in table 1. These studies were published between 2012 and 2023. The majority, 13 out of 20 (65%) were conducted in the USA [19, 28–39], six (30%) in Europe [17, 40–44] and one (5%) in Asia [45]. 13 studies specifically focused on MA as the primary study end-point [19, 28, 30–32, 38–45]. In terms of study design, most studies (15 out of 20, 75%) were retrospective. Prospective studies included surveys with questionnaires (n=4) and one randomised study (n=1). 16 studies were original journal articles and four studies were obtained from grey literature, with three abstracts [38, 39, 43] and one thesis [44].

TABLE 1

General characteristics of studies included in the systematic review, description of medication adherence (MA) assessment method and results

Study population

The analysis included 22 675 patients, with a median age of 58 years (interquartile range: 53–66 years). Among these patients, 66% were female (n=14 962). The type of PH was specified in six studies [17, 39–42, 44], all of which analysed baseline differences between patients with PAH and CTEPH, and one study assessed and compared MA based on the type of PH [41]. Four studies provided details on the time since diagnosis, with a median ranging from 1.7 to 11.1 years. PAH therapies were detailed in 15 studies, representing 21 124 patients (PDE5i: 13 938, 66.0%; ERAs: 3847, 18.2%; sGCS: 1882, 8.9%; PCA: 295, 1.4%). Among patients treated with PDE5i and ERAs, sildenafil was used in 75% of cases (n=6155) and bosentan in 58% (n=1271), respectively. In studies describing treatment characteristics, 4825 patients (83%) received monotherapy, while 981 patients (17%) received combination therapy (dual oral combination therapy or multiple therapy including PCA) [29, 33, 37, 41, 42, 45]. Five studies reported nonspecific PH treatments [17, 29, 34, 35], the most reported being diuretics (4346, 65.5%), anticoagulants (2303, 34.7%) and oxygen therapy (1560, 23.5%).

Methods of MA assessment

Table 1 provides a comprehensive overview of the various methods employed to measure MA. Two primary types of MA evaluation were identified: mean±sd of MA, reported in 14 studies, and the proportion of high MA, assessed by 14 studies. PDC, MPR and auto questionnaires were used in 10 (71%), eight (57%) and six (43%) studies, respectively. Moreover, four studies (25%) used two different methods to assess MA. Specifically, two studies used both PDC and MPR [37, 41], while two studies used MPR and questionnaires [17, 44]). None of the studies used more than two methods.

Risk of bias within studies

The evaluation of the risk of bias is presented in figure 2. The majority of the studies were categorised with a moderate (with some concerns) global risk of bias (9/20) [19, 28, 33–35, 37–40]. Six studies demonstrated a low global risk of bias [17, 29, 31, 32, 41, 45], while five were identified with a high global risk of bias [30, 36, 42–44]. More than half of the studies exhibited notable issues, particularly in the domains of selection bias and study design, with concerns arising from factors such as retrospective study design and nonrepresentative PH patient populations. Furthermore, information on withdrawals and dropouts was frequently lacking. The level of agreement between the two reviewers was considered good, as reflected by a Cohen's coefficient of 0.63 (95% CI 0.28–0.98).

FIGURE 2

Risk of bias across studies.

Results of MAMean MA

14 studies provided an insight into the mean MA of patients with PH (table 1). Of the 13 studies assessing MA to PH-specific treatments, three did not specify the class of treatment [17, 33, 42]. The remaining two studies determined MA to PH treatments without specifying their type (specific and/or nonspecific) [39, 44]. The reported means of MA ranged from 0.62 to 0.96. Means of MPR and PDC ranged from 0.66 to 0.96 and from 0.62 to 0.98, respectively (figure 3).

FIGURE 3

Forest plot of mean values of treatment medication adherence among pulmonary hypertension patients according to the method used. MPR: medication possession ratio; MRAW: mean raw; PDC: proportion of days covered.

Proportion of high MA

14 studies reported the proportion of patients with high MA (table 1). Among these studies, three detailed the proportion of patients with high MA without providing information on the type of PH treatments [39, 43, 44]. The proportion of patients with high MA varied based on the method employed and ranged from 40% (95% CI 35–45%) to 94% (95% CI 88–97%). Proportion values of MPR, PDC and Questionnaires ranged from 51 to 80%, from 40 to 94% and from 47 to 94%, respectively (figure 4).

FIGURE 4

Forest plot of proportions of pulmonary hypertension patients with high medication adherence level according to the method used. MPR: medication possession ratio; PDC: proportion of days covered.

Factors associated with MA

This systematic review identified ten studies investigating potential factors associated with MA in PH [19, 28, 31, 32, 38–42, 45]. The most consistently associated factor was the type of therapy (monotherapy versus combination therapy) [19, 32, 42, 45]. Conflicting results emerged regarding the impact of combination therapy versus monotherapy on the proportion of patients with high MA, with Tsai et al. [45] reporting an odds ratio (OR) of 2.66 (0.97–7.29), p=0.06, and Grady et al. [42] reporting an OR of 0.51 (0.30–0.97), p=0.01. Notably, patients with PH seemed less likely to maintain high MA when prescribed PDE5i, particularly sildenafil (Waxman et al. [19]: tadalafil versus sildenafil; OR 2.59 (1.60–4.22)). Similarly, Frantz et al. [32] reported a higher proportion of patients treated by ERA exhibiting high MA (75.6%) compared to those treated with PDE5i (61.5%). Grady et al. [42] found that an increased number of daily doses was associated with a lower level of MA (OR 0.70 (0.52–0.96), p=0.03). Several other co-medications were explored as potential risk factors for MA. Kjellström et al. [41] demonstrated that good adherence was associated with a lower number of concomitant chronic treatments (OR 1.17 (1.03–1.33)), especially for CTPEH patients (OR 1.43 (1.07–1.91)). In contrast, Grady et al. [42] reported that concomitant medicines were a positive factor for adherence (OR 1.11 (1.03–1.19), p=0.007) [42]. Studies examining the delay since PH diagnosis found that an increased delay was responsible for lower adherence, with significant differences (Kjellström et al. [41]: OR (≤3 years versus >3 years) 1.93 (1.30–2.86); Ivarsson et al. [40]: 4.1±3.6 years (high adherence) versus 7.5±6.7 years (low adherence), p<0.001) except for one study that found a positive effect of length of time on therapy to achieve moderate MA (MMAS-8≥6) (Grady et al. [42]: OR 0.86 (0.75–0.98), p=0.023).

Five studies reported the effect of age on MA [19, 40–42, 45]. Most found no clear association between older age and higher MA, except for two studies (Grady et al. [42]: OR 1.07 (1.03–1.11), p<0.001; Ivarsson et al. [40]: 68±13 years (high adherence) versus 59±16 years (low adherence), p<0.001). Two studies explored comorbidities, defined as a health problem currently affecting the participant or as a high comorbidity index and showed nonsignificant results with a good MA [42, 45]. Finally, studies highlighted the negative impact of co-payment on MA, depending on the presence of financial assistance [19, 28, 31, 39]. Schikowski et al. [31] demonstrated that high co-payment for the combination of ERA and PDE5i was significantly associated with poor MA. Furthermore, a higher 30-day co-payment (>$250 versus <$50) was also associated with poorer MA (Waxman et al. [19]: OR 0.57 (0.39–0.83), p<0.05). Details are provided in table S1 (for t-test and χ2) and table S2 (for ORs). Only two studies explored the impact of ethnicity on MA, with no apparent or significant influence found.

Clinical and economic outcomes related to MA

The studies included in this review primarily focused on healthcare utilisation and costs associated with varying levels of MA. Among the six studies describing healthcare use and cost, four explored a potential connection with MA [28, 32, 38, 39]. A high level of MA of PH patients was consistently associated with a significant decrease in hospitalisation rate, fewer inpatient hospital days and reduced outpatient visits. However, Shah et al. [28] reported a nonsignificant slight increase in the hospitalisation rate [32, 38, 39]. Healthcare utilisation and more substantial healthcare costs were linked to a low degree of MA, especially for patients in the PDE5i group compared to the ERA group (p=0.0015) [32]. AEs related to PH treatments were only explored by Shah et al. [28], who found a significant decrease in AEs in the high MA group (AEs: 100% (low MA group) versus 44% (high MA group), p<10−2). Only one study investigated overall survival according to MA, with no significant effect [39].

Two studies explored the association between MA and quality of life. Ivarsson et al. [40] detected no significant association between MA and levels of the EuroQol five dimensions index and the Beliefs about Medicines Questionnaire–Specific scale, while Robbins et al. [39] showed that better MA was significantly associated with an increase in SF-12 score and a decrease in EmPHasis 10 score. Results are presented in table S3.

Discussion

In this systematic review, we identified 20 studies exploring the MA of patients with PH. These studies revealed heterogeneous results regarding mean MA values or the proportion of patients with high MA. We highlighted potential factors associated with low MA that could be linked to poorer outcomes.

MA according to diagnosis and management

Despite our intent to specify MA levels according to PH type (PAH or CTEPH) or PH treatments (specific or nonspecific), the data did not allow for a conclusive answer. Several studies described nonspecific PH treatments (such as diuretics, oxygen and anticoagulants), but none specifically explored them. Regarding the type of PH, only one study assessed MA by differentiating between PAH and CTEPH patients [41]. A lack of precision in the profiles of PH patients could affect the accuracy of determining their MA levels. It would be more appropriate to better characterise PH patients to assess MA by stratifying results according to PH characteristics, including type and treatments.

Different MA assessment methods and thresholds

We chose to specify the proportion of MA according to different determination methods, both direct and indirect, to better capture any difference in the level of MA [13, 46, 47]. There is no clear consensus on the best MA assessment method, so guidelines suggest using at least two different methods. Only four studies in our review used two methods to assess MA, with two using MPR and PDC and two using MPR and questionnaires [17, 37, 41, 44]. Interestingly, no study used a direct method. Future studies should ideally incorporate different methods (direct and/or indirect) to capture the degree of MA accurately. In this systematic review, MPR and PDC were the most used methods for determining MA. They presented significant differences. The main limitation of MPR lies in its inability to indicate whether patients obtain treatments before exhausting their initial supply, potentially leading to an overestimation. PDC is also a ratio, but unlike MPR, it represents the sum of the days with effective treatment, avoiding artificially increasing adherence levels [48, 49]. Heterogeneity between studies might also be explained by different thresholds used to define a high level of MA. Haynes et al. [10] recommend a cutoff for MA of 80% to reliably predict clinical and economic outcomes [50]. Our systematic review included studies with MA thresholds as high as 90% or 95%; unsurprisingly these were associated with poorer adherence [38, 41]. Similar interpretation issues were observed in studies using questionnaire-based MA measuring tools [17, 40, 42, 43]. All included studies using questionnaires established the maximum value as a high MA threshold, except for the study of Jackson et al. [44]. We could argue that this threshold is rigid and may lead to dichotomous MA results, whereas assessing all levels of MA would be more relevant (high, medium and low).

Factors associated with MA

In chronic heart failure, MA was largely explored using a cutoff MA of 80% and was reported as nonoptimal at 1 year (around 17% to 39% for nonadherent patients, depending on the therapeutic class) [51, 52]. In those studies, MA worsens with the passing years and the number of medications (42% and 5% were adherent to dual and triple heart failure therapy) [51, 52]. Another well-described factor impacting MA in cardiovascular disease is the high drug intake frequency [51, 53, 54]. In our study focusing on PAH and CTEPH, the main factor associated with MA was increased daily drug intake frequency, especially observed with sildenafil [19, 32, 42, 45]. Other factors, such as combination therapy, concomitant treatment, younger age, a longer delay since diagnosis and co-payment, appear to negatively affect MA. This has yet to be confirmed since conflicting results between studies resulted in an unclear assessment of their impact [40–42, 45]. The risk of bias, inconsistency and imprecision of studies could explain these conflicting results and highlights the need for a good methodology [55, 56]. Analysing factors associated with either a low MA or poor outcomes may be subject to selective outcome reporting; these results should thus be interpreted cautiously.

Surprisingly, only three studies explored the link between quality-of-life measurements and MA, despite established knowledge that they could positively impact each other [17, 39, 40]. They revealed a significant positive association between a high quality of life score and MA. In future studies, it could also be beneficial to take patient-reported outcome measures into account to better capture all factors that could explain poor levels of adherence, as was done in heart failure [51, 57].

We also encountered challenges in providing a comprehensive assessment of economic factors associated with low levels of MA due to the heterogeneity of economic conditions and medical financial assistance between countries. The criteria used by health insurance for reimbursing the costs of medical treatment were not consistently detailed in the studies, although they could significantly impact MA.

It is evident that there is a need for improvement in MA and the identified baseline factors serve as crucial indicators of the risk of poor MA. These factors could help identify a global patient profile at risk of low MA, aiding in screening hospitalised patients who may benefit from adherence counselling. The heterogeneity of MA and its identified factors highlight the complexity of MA evaluations. Known barriers to MA are multifactorial and depend on the patient, socioeconomic factors, structure of the healthcare system, prescribed therapies and the condition for which they are prescribed. Therefore, assessing the reasons behind MA should be studied in dedicated prospective studies through semi-structured interviews, for example.

Clinical and socioeconomic impact of MA

This work also studied clinical and economic outcomes associated with levels of MA. It seems that low levels of MA did not significantly impact survival [39]. However, they may result in increased hospitalisation rates and AE, more extended inpatient hospital stays, more days of outpatient visits, and higher healthcare costs [28, 32, 39]. Similar findings were reported by Hood et al. [58] in heart failure patients. Finally, we observed that only one study examined the association between adherence and AE occurrence [28]. Even if Shah et al. [28] reported AEs in the study period in which MA was assessed, it seems unclear whether AEs were the consequence or the reason for lower MA. Exploring MA according to prostacyclin type (PAG or PAN) would have been relevant because of the different administration routes (oral or intravenous/subcutaneous). Shah et al. [28] described all concomitant PH-specific treatments (including PAN and PAG) but only measured MA for oral PDE5i. Knowing the potential AEs associated with oral PAG, we could hypothesise a lower MA for patients treated with selexipag [59].

Study limitations

This systematic review has some limitations, including 1) significant heterogeneity across studies (in terms of sample size, healthcare reimbursement and drug payment systems, design, population, and MA assessment method), making it impossible to perform a meta-analysis, 2) several included studies restricted to articles published in English and French, and 3) the presence of unmeasured confounders in most of these studies.

To our knowledge, only one other systematic review provides data on MA and treatment discontinuation of PAH patients [60]. Qadus et al. [60] included 14 studies and reported an overall pooled proportion of patients adherent to their PAH-specific treatments of 60.9% (95% CI 52.3–69.1%). However, it does so by pooling different MA thresholds and assessment methods (MPR and PDC). Although this review is primarily focused on patients with PAH, two studies also enrolled patients with CTEPH, potentially influencing the outcomes. Finally, none of the investigations concerned nonspecific treatments. Diuretics play a crucial role in managing patients with PH, where they are prescribed to reduce RV preload, enhance RV performance and limit interdependence between the RV and left ventricle (class of recommendation I, level of evidence C) [8]. Diuretics are essential for preventing episodes of right heart decompensation and may improve patients’ quality of life [61, 62]. They are the most used PH nonspecific treatment identified in our systematic review, potentially influencing MA levels. This underscores the need for future studies to explore MA in conjunction with nonspecific PH treatments.