A randomised controlled trial of the effect of a connected inhaler system on medication adherence in uncontrolled asthmatic patients

Discussion

This study demonstrated that the CIS significantly improved adherence to asthma maintenance therapy when participants and HCPs were provided with data on maintenance use compared with the control arm (no data feedback). All CIS intervention arms resulted in an improvement in participant adherence compared with the control group; however, provision of data on rescue medication use in addition to maintenance therapy use, or inclusion of HCPs in data feedback, did not result in further adherence benefit.

This study has some limitations reflecting the difficulties of carrying out randomised, controlled, clinical studies evaluating adherence. Although we experienced some difficulties in collecting baseline adherence at some sites, the results for the primary analysis (with imputed baseline data) and the sensitivity analysis (with these participants excluded) were consistent, suggesting that the approach taken for the primary analysis was sufficiently robust. The high level of adherence in all study arms, including the control arm, may have limited the potential for demonstrating differences between the active study arms. This was possibly a result of a change in participant behaviour in response to having sensors on their inhalers even in the absence of data feedback, as was recently reported in an observational study [26]. The high levels of adherence may also result from the recruitment by interested centres of asthma patients who are engaged in their disease and therefore may not be completely representative of asthma patients in general. Another explanation for the high adherence may be the switch to once daily ICS/LABA from a twice daily combination, which was the case for the majority of participants. Furthermore, although symptomatic patients were randomised in this study, as defined by ACT score, mean F_eNO was relatively low at screening (mean 29.3 ppb) and further decreased between screening and randomisation (mean 21.9 ppb), indicating a “biomarker low” population with little residual ICS-responsive inflammation and therefore little room for clinical improvement with better adherence to maintenance therapy. This was borne out by the lack of change demonstrated in F_eNO and PEF over time. This suggests that perhaps not all participant symptoms were due to asthma and may have been related to other factors including being overweight/obesity (mean BMI ≥30 kg·m⁻² across groups). Finally, the CIS sensors recorded the time that the maintenance inhaler was opened and closed as a surrogate for drug delivery, but did not capture data on a participant's inhalation technique, the impact of which is unknown. Although common errors with inhaler technique have been reported for both DPIs [27, 28] and MDIs [27–29], the rate of errors with the ELLIPTA inhaler is reportedly low [30], suggesting that this would have been of relatively low impact in the present study. However, these results may not be generalisable to other inhalers. Sulaiman et al. [31] reported relatively good agreement between adherence measured by dose counter and attempted adherence (attempted inhalation) in patients with good adherence but less so in patients with poor adherence. In our study the baseline adherence data showed relatively good adherence amongst participants and therefore results may not be applicable to a broader population.

Nonetheless, the primary aim of this study was to assess the effect of an adherence intervention and we demonstrated that the CIS was associated with a significant improvement in adherence to asthma controller therapy. A spike in total adherence observed during the first month for all active arms, but not seen in the control group, probably reflects the “new toy” effect of the asthma app. In addition, whilst adherence dropped off over time in the control arm, as participants reverted to their normal behaviour, adherence improvements were maintained in the CIS arms, demonstrating the utility of the system. The primary comparison exceeded the minimum detectable difference of 9% but did not meet the pre-specified difference of 15%. Considering the high mean adherence in the control arm, there was still an increase in absolute adherence of 12% in the primary intervention arm, equivalent to one additional dose of maintenance medication per week and a relative increase of approximately 17%. Adherence increased in all four CIS arms with no observed differences between arms. Although the largest change from baseline was observed in the “maintenance and rescue data to participants and HCPs” arm (arm 3), this group had the lowest baseline adherence and most room for improvement. As baseline adherence was included as a covariate in the model, this is taken into account and adjusts the least squares mean values accordingly, resulting in the largest difference in adherence over months 4–6 between a CIS arm and the control arm being seen in the “maintenance data to participants and HCPs” arm (arm 1). The baseline characteristics of age and ACT score at randomisation both had an impact on improvement in adherence to maintenance medication, with younger (18–48 years old) versus older participants and those with an ACT score ≤15 (worse asthma control) showing the greatest improvements. Consistent with our findings, the relationship between poor asthma control and non-adherence is well documented [2, 5, 6]. The positive response in our study shown by younger participants and those with worse asthma control is encouraging and may indicate subsets of patients who would particularly benefit from the CIS. These findings may also be pertinent when designing future studies to evaluate the CIS. Of note, these participants also had the greatest room for improvement. Identifying an ICS responsive population in terms of biomarkers such as F_eNO would also more likely result in the observed effects on adherence translating to improved clinical outcomes, as recently shown in a severe asthma population [32].

Although covert monitoring with the CIS could not be achieved practically in this study, several measures were incorporated to try and mimic real-world conditions. Minimising the number of clinic visits over the 6-month study period, study-independent medication dispensing visits that were initiated by the participant and the primary measurement period being over months 4–6 were all designed to result in participants reverting to their usual behaviour, as previously noted [33]. The largest improvements in adherence in the active CIS arms versus the control arm were observed over the last 3 months of treatment, compared with the first 3 months, supporting the benefit of this approach. The failure to demonstrate a relationship between higher adherence to maintenance treatment and lower short-acting β-agonist (SABA) use may be due to relatively low mean use of SABAs in the study or because patients with greater disease burden may use both more controller medication and more rescue medication. Whilst the provision of data on rescue medication use in addition to maintenance therapy use, or inclusion of HCPs in maintenance use data feedback, did not lead to additional benefits from the CIS in terms of maintenance adherence, results of mean data should not preclude the fact that, for some individual patients and/or HCPs, these features could provide further advantages in optimising patient adherence with treatment. On the other hand, it also suggests that a basic CIS, providing only feedback on maintenance treatment use to the patient, is a good workable system and HCPs need not view the dashboard system as an extra burden.

Clinically relevant improvements in asthma control scores were shown in all study arms, but no differences were observed between groups. This may have been due to the switch to once daily fluticasone furoate/vilanterol; however, the study was not powered on this or other secondary endpoints. The increase in the number of rescue-free days in participants who received data on their amount of rescue use suggests that their behaviour changed in response to seeing their rescue medication usage, possibly due to increased engagement in their asthma self-management. Although patients often overestimate how much maintenance medication they use [11, 34, 35], they may also underestimate how much rescue medication they use (due to misperceptions about their airway obstruction) [36]. In our study, seeing objective data for their rescue use appeared to affect this. Numerous surveys and reports have highlighted the serious consequences of under-using asthma maintenance therapy and over-using rescue medications [4, 37–39] and both have been identified as contributory risk factors for asthma deaths [38]. The problem of non-adherence is also significant in more severe disease. Approximately 65% of “difficult to control” patients with asthma referred for specialist care in the UK filled less than 80% of their inhaled maintenance medication prescriptions [40, 41]. In a US study of over 10 000 omalizumab users, 49% had very low adherence to ICS or ICS/LABA treatment (a medicine possession ratio of ≤50%) prior to initiation of biologic therapy [42]. The use of technology such as the CIS could provide a valuable tool for physicians to diagnose, monitor and support the way patients use their inhalers and ensure appropriate treatment escalation (including biologic therapy) by confirming adherence to background treatment.

In conclusion, in this first study to evaluate the impact of different components of a CIS (by comparing four different types of feedback on both maintenance and rescue medication use in participants with uncontrolled asthma), data feedback on maintenance therapy led to increased maintenance adherence and data feedback on rescue medication usage resulted in a greater number of rescue-free days, but not in additive benefits in controller adherence.