Clinical Impact of Electronic Monitoring Devices of Inhalers in Adults with Asthma or COPD: A Systematic Review and Meta-Analysis

We conducted a systematic review and meta-analysis to gain insight into the characteristics and clinical impact of electronic monitoring devices of inhalers (EMDs) and their clinical interventions in adult patients with asthma or COPD. The search included PubMed, Web of Science, Cochrane, Scopus and Embase databases, as well as official EMDs websites. We found eight observational studies and ten clinical trials, assessing a wide range of clinical outcomes. Results from the meta-analysis on adherence to inhalers in a period over three months were favourable in the EMD group (fixed effects model: SMD: 0.36 [0.25–0.48]; random effects model SMD: 0.41 [0.22–0.60]). An exploratory meta-analysis found an improvement in ACT score (fixed effect model SMD: 0.25 [0.11–0.39]; random effects model: SMD: 0.47 [−0.14–1.08]). Other clinical outcomes showed mixed results in the descriptive analyses. The findings of this review highlight the benefits of EMDs in the optimization of adherence to inhaled therapy as well as the potential interest in other clinical outcomes.


Introduction
Among chronic respiratory diseases, asthma and chronic obstructive pulmonary disease (COPD) stand out for their high prevalence, impact on quality of life and clinical repercussion. Both pathologies share some characteristics, such as airflow obstruction measured as forced expiratory volume in 1 s/forced vital capacity (FEV1/FVC) < 0.7, and the importance of inhaled treatment to achieve the best possible control, which will depend on the stage or severity of the patient, along with other strategies such as control of risk factors or smoking cessation. However, these pathologies are far from being controlled in multiple cases [1][2][3].
Adherence to pharmacological treatment is poor in chronic pathologies, close to 50% according to estimates by the World Health Organization [4]. Moreover, administration of inhaled therapy is correct in a third of the patients [5]. Likewise, adequate adherence to inhalers is associated with a reduction in asthma exacerbations, greater control of symptoms, lower systemic cortico-steroid requirements and lower disease-related mortality [6][7][8][9][10]. In COPD, several studies associate poorer adherence to inhalers with a higher number of

Search Strategy
Firstly, an exploratory phase was conducted to detect the terminology used for electronical interventions by means of reviewing original articles presented in some of the aforementioned websites. The following four domains were found and developed with their corresponding keywords: sensor, monitoring, innovative connected technology and devices ( Figure 1). The full search strategy is available in File S1. The search review process was assessed with the Peer Review of Electronic Search Strategies (PRESS) checklist [37]. randomized clinical trials. We included studies regardless of the nature of the comparator group in the case of randomized trials. We included studies in the English and Spanish language, with no time restrictions. Unpublished articles and conference abstracts were excluded. In addition, studies were considered ineligible when the outcomes of interest were not measured or reported.

Information Sources
The search was conducted on 1 June 2022 on the following bibliographic databases: PubMed, Web of Science, Cochrane, Scopus and Embase, according to the eligibility criteria, with no time restrictions.

Search Strategy
Firstly, an exploratory phase was conducted to detect the terminology used for electronical interventions by means of reviewing original articles presented in some of the aforementioned websites. The following four domains were found and developed with their corresponding keywords: sensor, monitoring, innovative connected technology and devices ( Figure 1). The full search strategy is available in File S1. The search review process was assessed with the Peer Review of Electronic Search Strategies (PRESS) checklist [37].

Selection and Data Collection Process
The data from searches in each database were exported to RefWorks, and an initial phase for detection of duplicates was performed. A subsequent duplication detection phase was conducted in the resulting database by DOI identifier and manual assessment.

Selection and Data Collection Process
The data from searches in each database were exported to RefWorks, and an initial phase for detection of duplicates was performed. A subsequent duplication detection phase was conducted in the resulting database by DOI identifier and manual assessment. The results of the refined database were screened (title/abstract) by two independent investigators, with disagreements resolved by a third researcher. This method was replicated for the following full-text assessment and final inclusion of articles. Additionally, identification of studies via websites and cites in systematic reviews and included studies (see Pharmaceuticals 2023, 16, 414 4 of 25 above) was performed by an independent researcher, with a subsequent validation by another researcher. Once selected, data were collected and validated by two independent investigators from each report.

Data Items and Synthesis Methods
The outcomes of the review were divided into domains and their corresponding items to be assessed: All results compatible with each outcome domain were sought, regardless of the time frame of measurement. The data collected also included the possibility of "other" clinical variables as we were exploring a novel approach, and it was plausible that other relevant data may be available. Due to the nature of EMDs, we anticipated that adherence to medication may be the most frequent variable found in the studies.
The other variables collected were author, year of publication, sample size, age of participants, study design, number of centres, follow-up period, control group, intervention group, intervention details, recruitment setting, healthcare professional involved, circuit type (focused on the healthcare professional, the patient or both of them), type of interaction, type of inhaler assessed, type of EMD, data recorded and health outcomes.

Study and Report of Risk of Bias Assessment
The risk of bias was assessed with the Revised Cochrane risk-of-bias tool for randomized trials (RoB2) [38] and with the Methodological index for non-randomized studies (MINORS) [39]. All methodological components/items of the tools were applied. Assessment was performed by a researcher and validated by a second member of the team. Reporting of risk of bias was summarized in the text and in graphical figures for independent items and global results of each study.

Statistical Analysis
The descriptive data of the participants' characteristics were reported as a mean (SD). All meta-analyses' calculations were conducted with the R software (Vienna, Austria) with meta and metafor packages for meta-analysis (Version 3.5.1.). Descriptive analyses and figures of the risk of bias were performed using Microsoft Excel for MAC, version 16.29.1 (Microsoft, Redmond, WA, USA). The mean and standardized mean differences (Hedges' g) and 95% CI for each group were calculated. The analysis of pooled data was conducted using a random-effect model to estimate the change for each group at the same measurement time on primary and secondary outcomes. Standardized mean differences were weighted by the inverse of the variance to calculate the size of the effect and 95% confidence interval. Cohen's criteria were used to interpret the magnitude of the effect: <|0.50|: small; |0.50| to |0.80|: moderate; and >|0.80|: large. Heterogeneity was assessed using Cochran's Q statistics and its corresponding p-value as well as the I2 statistic, which describes the percentage of variability in effect estimates attributable to heterogeneity rather than chance when I2 was >30% (30-60% representing moderate heterogeneity). Publication bias was assessed with funnel plots and Begg's test. Significance was set at p < 0.05.

Study Selection
A total of 10483 articles were identified through computer searches in the selected databases, with 5264 duplicates removed through electronical or manual methods. A total of 5219 studies were screened by title and abstract and 146 were assessed for eligibility by full-text assessment. An additional search was performed using websites, and citation searching of articles and systematic reviews, with 177 articles assessed for eligibility. Some of the retrieved studies might appear to meet the inclusion criteria, but were eventually discarded due to several reasons, such as: not providing separate results from adults and paediatric patients [40,41], using EMDs to confirm the validity of the response to the Fractional Exhaled nitric oxide (FeNO) outpatient test, not as an intervention itself [42], or assessing patient satisfaction with EMDs rather than its clinical impact on disease as study outcomes [43]. Eighteen studies were finally included in the systematic review ( Figure 2). Seven out of eight observational studies were conducted in the USA, while one study was conducted in the Netherlands. The distribution of clinical trials was the USA (n = 5), Ireland (n = 2), New Zealand (n = 1), Switzerland (n = 1) and one multi-country study (Canada, Germany, Italy, Netherlands, Spain, UK and the USA).
Regarding observational studies, six of them focused on patients with COPD, one study on asthma and one study on patients with asthma and/or COPD (Table 1). Their sample size ranged from n = 11 to 2292 patients, with an average age ranging from 36.8 to 68.6 years and being over 60 years in five out of the eight studies. Clinical outcomes assessed in observational studies included adherence to maintenance medication, use of rescue medication, disease control, inhalation technique, quality of life, self-management variables (e.g., behaviour, knowledge and adherence to the intervention), steroid use, health-care utilization variables and mental health scores. Variables were differently operationalized across studies (e.g., mean daily short-acting beta-agonist [SABA] use, change in mean of SABA use, percentage of SABA-free days, etc.). Six of the studies performed statistical analyses of the clinical outcomes [28,44,46,47,49,50], while two studies presented descriptive analyses only [45,48]. Adherence to maintenance inhalers improved in two studies [49,50] whereas it was similar or decreased in the other two studies [44,46]. SABA use was found to be decreased in all variables assessed in two different studies [28,44]. One study focused on symptom control and found a clinical improvement in ACT score, days with asthma symptoms and nights with asthma symptoms. However, no difference was found regarding activity limitations [47]. Finally, one study found a decrease in COPD-related healthcare utilization and high accuracy of rescue alerts in predicting moderate-severe exacerbations but found no differences in variables such as all-cause healthcare utilization, number of pulmonary or primary care visits and antibiotic or steroid use [46]. Table 1. Summary of results. ACQ (Asthma Control Questionnaire); ACT (Asthma Control Test); AQLQ (Asthma Quality of Life Questionnaire); Asthma-related quality of life (AQOL); AE (asthma education); C (Control group); CAT (COPD assessment test); CI (confidence interval); CCQ (Clinical COPD questionnaire); COPD (chronic obstructive pulmonary disease); DPI (dry-powder inhaler); ED (Emergency Department); FeNO (Fractional Exhaled nitric oxide); FEV1 (forced expiratory volume in 1 s); HCP (healthcare professionals); I (intervention group); ICS (inhaled corticosteroid); NA (not available); ND (no difference); PEF (peak expiratory flow); pMDI (pressurised Metered Dose Inhaler); POEMS (Polymedication Electronic Monitoring System); SABA (short acting beta agonist); SGRQ (St. George's Respiratory Questionnaire).
points -ACT score from <20 to ≥20 -Asthma exacerbations ->1 course of oral corticosteroids ->1 ED visit or hospitalization 3 months Routine care Self-management for patients. Access to information, related interventions and periodical review with patients for HCP.
• C: −17% (p < 0.01) • I: +2% (p = 0.40) -Asthma control: ND -10% increase SABA-free days: ND -ACT improvement ≥3 points: ND -ACT score from <20 to ≥20: ND -Asthma exacerbations: ND ->1 course of oral corticosteroids: ND ->1 ED visit or hospitalization: ND Nides (1993) [50] COPD         As for the clinical trials, seven studies focused on asthma, one study on COPD and two studies on asthma and/or COPD (Table 1). Their sample size ranged from 19 in a pilot clinical trial to 437 participants, with an average age ranging from 36 to 70 years. Clinical outcomes assessed in clinical trials comprised adherence to maintenance medication, use of rescue medication, disease control, inhalation technique, quality of life, self-management variables, self-reported symptoms, exacerbations, lung function tests, hospitalizations, steroids use, ED visits and composite clinical variables. Variables were differently operationalized across studies resulting in great variability across studies (e.g., time to next exacerbation, exacerbation rate, severe exacerbations leading to hospitalization, etc.) (complete list of variables assessed in clinical trials: File S2). Adherence to maintenance inhalers was assessed in eight out of the ten clinical trials. The intervention group showed a statistically significant improvement in adherence compared with the control in six of the trials at the end of the study analysis [51,[53][54][55][56][57][58][59][60][61]. Two studies found no differences in adherence results [27,52]. Regarding SABA use, mixed results were found in three studies, as improvement or no difference was found depending on the variable or subgroup assessed (e.g., SABA use vs. SABA-free days) [23,27,51]. No differences in SABA use were found in one study [61]. Studies assessing disease control with ACT or ACQ questionnaires showed mixed results: three studies found no differences [27,51,52], one study showed better results in the intervention group [62] and one study showed improvement for the previously uncontrolled subgroup only [23]. One study assessed independently additional symptoms (cough, breathlessness, nocturnal symptoms, wheeze, etc.). The intervention group improved all the symptoms over the study period, but no direct comparison across groups was performed [58]. In relation to quality of life, four trials found no differences between groups [52,55,60,62] and one study found better results over time in the intervention group [58]. As for other variables, none found statistical significance: FeNO, PEF, exacerbations, time to exacerbations, steroids use, ED visits, hospitalizations, FEV1, technique error, critical technique error rate and composite clinical variable criteria [27,[51][52][53][54][55][56][57][58][59][60][61].

Risk of Bias in Studies
A separate assessment of risk of bias was conducted for observational and clinical trials (Figure 4). Regarding observational studies, we obtained a range from 7 to 11 points with the MINORS tool (0 = lowest risk of bias; 16: higher risk of bias) [28,[44][45][46][47][48][49][50]. One observational study scored 16 points in the additional assessment of comparative projects (0 = lowest risk of bias; 24: higher risk of bias) [50]. As for the clinical trials, the assessment was performed with the RoB2 tool. Five studies were assessed as having low risk-of-bias criteria [23,27,51,52,60], four studies as high risk-of-bias criteria [53,58,61,62] and one study was classified as having some concerns criteria [55]. For studies with high risk of bias, the main reason was potential bias arising from the randomization process criteria due to the inherent use of electronic devices and consequent interventions [53,58,61,62]. Due to the nature and small number of studies, no exclusions for this reason were pre-established in the meta-analysis.

Impact of Interventions on Adherence to Maintenance Inhaled Therapy
A total of six trials were found to measure adherence in percentage as a clinical outcome.
The impact of interventions in studies assessing adherence up to 3 months was assessed in three studies, including 318 participants [53,58,61]. Figure 5 shows the results of the forest plot analysis. Although adherence results were favourable in the intervention group, no significant differences were found in both the fixed effect model analysis (

Impact of Interventions on Adherence to Maintenance Inhaled Therapy
A total of six trials were found to measure adherence in percentage as a clinical outcome.
The impact of interventions in studies assessing adherence up to 3 months was assessed in three studies, including 318 participants [53,58,61]. Figure 5 shows the results of the forest plot analysis. Although adherence results were favourable in the intervention group, no significant differences were found in both the fixed effect model analysis (SMD:

Impact of Interventions on Adherence to Maintenance Inhaled Therapy
A total of six trials were found to measure adherence in percentage as a clinical outcome.
The impact of interventions in studies assessing adherence up to 3 months was assessed in three studies, including 318 participants [53,58,61]. Figure 5 shows the results of the forest plot analysis. Although adherence results were favourable in the intervention group, no significant differences were found in both the fixed effect model analysis (SMD: 0.07 [−0.16 to 0.31]] or the random effects model analysis (SMD: 0.77 [−0.26 to 1.80]]. The estimate of the between-study variance (heterogeneity) was considered high (I 2 = 91%; τ 2 = 0.69; p = 0.01).  The impact of interventions in studies assessing adherence in a period of 3 months or longer was assessed in five studies, including 1223 participants [27,[51][52][53]58]. The impact of interventions in studies assessing adherence in a period of 3 months or longer was assessed in five studies, including 1223 participants [27,[51][52][53]58].

Impact of Interventions on Other Clinical Outcomes
The variability of other variables in terms of definition, operationalization and study timeline duration prevented us to perform additional meta-analysis. However, two of the trials were found to measure control with the ACT questionnaire as a clinical outcome, leading to an exploratory assessment with 799 individuals [23,51]. Figure 7 shows the results of the forest plot analysis. ACT results were favourable in the intervention group. Significant differences were found in the fixed effect model analysis (SMD: 0.25 [0.11 to 0.39]] but not in the random effects model analysis (SMD: 0.47 [−0.14 to 1.08]]. The estimate of the between-study variance (heterogeneity) was considered high (I 2 = 94%; τ 2 = 0.46; p < 0.01).

Impact of Interventions on Other Clinical Outcomes
The variability of other variables in terms of definition, operationalization and study timeline duration prevented us to perform additional meta-analysis. However, two of the trials were found to measure control with the ACT questionnaire as a clinical outcome, leading to an exploratory assessment with 799 individuals [23,51]. Figure 7 shows the results of the forest plot analysis. ACT results were favourable in the intervention group. Significant differences were found in the fixed effect model analysis (SMD: 0.25 [0.11 to 0.39]] but not in the random effects model analysis (SMD: 0.47 [−0.14 to 1.08]]. The estimate of the between-study variance (heterogeneity) was considered high (I 2 = 94%; τ 2 = 0.46; p < 0.01).
trials were found to measure control with the ACT questionnaire as a clinical outcome, leading to an exploratory assessment with 799 individuals [23,51]. Figure 7 shows the results of the forest plot analysis. ACT results were favourable in the intervention group. Significant differences were found in the fixed effect model analysis (SMD: 0.25 [0.11 to 0.39]] but not in the random effects model analysis (SMD: 0.47 [−0.14 to 1.08]]. The estimate of the between-study variance (heterogeneity) was considered high (I 2 = 94%; τ 2 = 0.46; p < 0.01).

Discussion
To the best of our knowledge, this is the first systematic review with meta-analysis exploring the characteristics and assessing the impact of clinical interventions derived from data of EMDs in adult patients with asthma or COPD. The most remarkable result to emerge from our study is the positive impact of EMD-based interventional programmes on adherence to inhaler treatment and the tendency to positive or mixed results in other outcomes such as symptom control, with potential benefits for daily practice in the optimization of inhaled therapy management.

Discussion
To the best of our knowledge, this is the first systematic review with meta-analysis exploring the characteristics and assessing the impact of clinical interventions derived from data of EMDs in adult patients with asthma or COPD. The most remarkable result to emerge from our study is the positive impact of EMD-based interventional programmes on adherence to inhaler treatment and the tendency to positive or mixed results in other outcomes such as symptom control, with potential benefits for daily practice in the optimization of inhaled therapy management.
Firstly, we would like to highlight the impact of these interventions on adherence. Adherence on maintenance therapy was the most frequent clinical outcome assessed in both the observational studies and clinical trials. Mixed-favourable results were found in the observational studies, which principally focused on COPD patients. However, the clinical trials showed an improvement in adherence for the intervention group in the meta-analysis. Although most trials focused on asthma patients, those studies focusing exclusively on COPD patients or including both participants with COPD or asthma also showed an improvement in adherence, which suggests that COPD patients may also benefit from this type of interventions. Poor adherence is an acknowledged risk factor by GINA and GOLD guidelines to symptom burdens, exacerbations, and poor quality of life to be assessed and tackled by the multidisciplinary team [1,63]. Inhaler treatment involves factors contributing to suboptimal adherence at three levels: medication, unintentional and intentional issues [64]. One of the main relevant results of the review is the ability of these interventional programmes to include a wide range of aspects related to these factors, such as patient self-awareness and efficacy, disease and management education, inhaler technique, triggers, age-related factors such as comorbidities, misunderstood directions, forgetfulness, dissatisfaction or inappropriate expectations [64]. The methodology of interventions tackling adherence in the assessed studies was aligned with successful interventions to improve adherence in previous studies regardless of the use of monitoring devices, such as shared decision making, electronic reminders, visits and providing information to clinicians [65][66][67][68][69][70][71].
As for the rest of the variables, mixed results were found for SABA use and symptoms of asthma and COPD. Theoretically, positive results may be obtained if adherence is improved but these clinical outcomes could be shaped by additional factors such as weather, infections, environmental triggers, comorbidities, severity of illness and other factors. These aspects may have a higher impact on other clinical outcomes and be responsible for the lack of results in variables such as exacerbations, hospitalizations or spirometry results. However, some studies on children have found fewer exacerbations requiring oral corticosteroids at 12 months [67]. With this in mind, adherence management would be the primary goal of EMDs due to the nature of the data collected by the sensors, but there is a need for further research to assess their real clinical benefits in terms of other health outcomes or to what extent it would be useful to implement specific additional interventions linked to these programmes as a global strategy [35]. A high degree of variability is inherent in the development of new technologies. In our review, we found 10 different EMDs, in line with previous literature [29][30][31][32][33][34][35]. Technological differences between their inhaler sensors should not pose a differential problem as they are validated systems. However, the associated digital-engagement tools, dashboards, data available, clinical setting, specific professionals involved, the number of associated interventions, usability and acceptability could make a difference between them [29][30][31][32][33][34][35]72]. In our review we have found a great variability in terms of clinical setting, professionals involved, follow-up, interventions, availability of smartphone applications and whether the clinical approach was focused on the patient, the healthcare professional or both. To what extent these factors may impact the clinical results remain uncertain. However, the results of our review show that all clinical trials included healthcare professionals as key roles in the intervention, highlighting the importance of interventions conducted by healthcare professionals rather than focusing on self-management with apps only [23,27,[51][52][53][54][55][56][57][58][59][60][61][62] ( Table 2).
Cost-effectiveness and feasibility are essential concerns to be considered when translating interventional programmes based on data from monitoring devices into regular clinical practice. On one hand, there are direct costs associated with monitoring devices, e-health platforms, and related-products and interventions. Minimizing the cost of monitoring devices has been described as a relevant acceptability criterion to assess device characteristics [73]. On the other hand, suboptimal inhaler use results in poor clinical outcomes, with studies reporting increased direct and indirect costs that may be potentially reduced [74,75]. In this context, an economic analysis of monitoring devices and adherencerelated interventions showed it may be cost-effective and cost-saving [76]. In terms of feasibility, some inexperienced patients face a considerable number of challenges, such as sensory impairment, intellectual ability, motivation, reduction of fine motor control, low self-efficacy of technology, fear or dislike of electronic devices, inexperience with e-health or computers, lack of awareness of e-health opportunities, previous unmet expectations, fear or losing traditional health services or lack of smart phones [77]. With regard to healthcare professionals, apart from e-health literacy skills, there is a need of training on EMD functioning, checking alerts, dashboards and typical errors [78]. Healthcare professionals have highlighted a high degree of administrative burden and complexity of these interventions, that would require additional employees to handle the corresponding workload [79]. Furthermore, a study showed that providing adherence information to healthcare professionals did not improve adherence unless the professional deliberately decided to check the details of a specific patient [69]. Moreover, the existence of an increasing number of EMDs can complicate their implementation and management in regular practice. Thus, a careful selection of patients who are most likely to benefit from interventions and the development of a common framework of the platforms seem to be convenient options to be prioritized in the near future.
Our systematic review and meta-analysis have some limitations. Firstly, we found a great variability of clinical outcomes, operationalization of variables and follow-up length periods across studies, which could impact the results, their interpretation and comparability. In relation to the use of monitoring devices, patients in both the intervention and control groups can suffer from the Hawthorne Effect, which implies a change in behaviour since they know they are being monitored [80]. Another aspect to be considered is the decrease of user engagement to e-health interventions with time, its clinical impact and potential measures to minimize this effect [81,82]. In addition, some studies focused on severe cases while others focused on patients regardless of their severity, which could affect the results if specific subgroups benefited more from the intervention. Moreover, data on the use of biological agents in asthma were not provided and may represent a major cofounding factor. Furthermore, information on pharmacological-inhaled treatment and posology was not present in some studies. Moreover, some of the studies were pilot projects and may benefit from escalation and ulterior methodology improvements of this type of studies. With regard to the review process, some limitations may also arise. For example, innovative technologies do not count with a definite terminology, so some articles might not have been found in the title/abstract search if alternative nomenclature was used by some authors. This limitation may be minimized if we performed preliminary searches to find synonyms in potential studies. In addition, as this is a novel technology, it is possible that additional features and interventions may improve clinical results in future trials. Finally, data from real-life studies would be of interest.
The findings of our systematic review and meta-analysis confirm the positive impact of EMD-based interventions on adherence to inhalers in adults with asthma or COPD. Our data indicate that other clinical outcomes, such as symptom control, may also improve when using EMD interventional programmes, but further research is needed to confirm whether additional interventions would be necessary, as asthma and COPD clinical results also rely on additional aspects such as environmental issues, comorbidities, etc. The broad implication of the present research is that EMDs represent a valuable asset that should help healthcare providers to implement policies to address asthma and COPD management.