e-Monitoring of Asthma Therapy to Improve Compliance in children using a real-time medication monitoring system (RTMM): the e-MATIC study protocol

This study is a one year, multicenter, randomized controlled trial in children who use ICS for asthma. All children receive an RTMM-device which registers time and date of administered ICS doses. Children in the intervention group receive “time-tailored” text–messages that are only sent when a dose is at risk of omission. Patients in the control group do not receive such text-messages.

The medical ethics committee of the Erasmus Medical Center has approved the study protocol (protocol number MEC-2011-143, Netherlands Trial Registry code NTR2583, http://​www.​trialregister.​nl).

Study data are coded in order to guarantee privacy of participants. Before entering the study, all participants are asked for written informed consent.

Patients are recruited from five outpatient clinics in the Netherlands: St Lucas Andreas Hospital, Academic Medical Center, BovenIJ Hospital (all in Amsterdam), Erasmus MC (in Rotterdam) and Groene Hart Ziekenhuis (in Gouda). The inclusion criteria for participants are:

We aim to include 44 children per hospital into the study. From the hospital administrations of each participating hospital, records are randomly selected of children aged 4–11 years and diagnosed with asthma at least 6 months ago. After verification of the other inclusion criteria, a patient information leaflet is handed out or sent to the parents of the potential participants. Parents are contacted and invited to visit the paediatric outpatient department for an intake interview. In case of participation, at least one of the childs’ parents has to give written informed consent. If a hospital is unable to include the required number of patients, they will be recruited from one of the other participating hospitals.

Children are randomly assigned to the intervention group or the control group. We use block randomization per hospital with blocks of 16 patients. Initially, physicians, researchers, and patients are blinded for randomisation. However, randomisation is unblinded after start of the study period, when patients find out whether they receive text-message reminders or not.

A flowchart of patient selection, randomisation and data collection is shown in Figure 1.

All children, both in the intervention and in the control group, receive an RTMM device for one year. ICS inhalations are registered by the RTMM-device which operates as follows: each time the pMDI is fired a data message containing patient-identification and time and date of administration is sent to the study database using the mobile telephone network. In order to prevent incomplete registration caused by insufficient network connection, the RTMM-device is designed to use two different networks: the mobile data-network and the regular mobile telephone network. If both networks are unavailable at the time of inhalation, a data message is prepared for sending at a later moment.

Only in the intervention group, time-tailored text-message reminders are sent to the parents and, if the child has a mobile phone, also to the child, in order to warn that a dose is at risk of being forgotten. Parents and children are not always together, for example if children are at school. As such, there may be a problem if the child has missed its dose and is already at school when the text-message is received by its parents. In that case, parents will not be able to verify if the child takes its medication. To ensure that text-messages are sent before the child goes to school (morning dose) or to bed (evening dose), a text-message is sent automatically if no ICS dose has been registered within 15 minutes after the planned time of inhalation. Such a short time interval may be less important when children are not at school, for example during weekends. For that reason, time intervals are individually defined (‘time-tailored’) for each patient for each day of the week. This is also thought to improve patient-acceptance of RTMM, reducing the so called “alert fatigue”.

Patients in the intervention group who live in an area with a very poor mobile network connection may occasionally receive an unnecessary text-message reminder if an ICS administration can not be reported to the RTMM-database in time. When the RTMM-network connection is re-established, the data message is sent. This phenomenon is closely monitored during the study period.

Data are initially collected on a case report form (CRF) and on questionnaires (hard copy). After each patient interview data are manually copied to a digital CRF. Data entry errors are minimized by using multiple choice options and fixed data fields. At the end of the study 10% of entered data are checked by a second person. If data entry errors are found, additional portions of 10% of the data are checked until no errors are found within a portion. Also, a periodic back-up of the study database of each hospital is made and checked for missing data. Access to the research databases is secured by passwords. Changing the format of the study documentation or study databases is restricted to the primary investigator. New versions are distributed from the central study location.

The sample size calculation was based on the primary outcome measure: adherence to ICS. We use the adherence data from our observational study [15] in which adherence to ICS was electronically measured with RTMM in children (<12 years old) with asthma. In this dataset, 4 subgroups with different adherence patterns could be distinguished: patients with very poor adherence (≤5%), poor adherence (mean 34%), good adherence (mean 78%) and excellent adherence (≥95%). We assumed that patients with very poor adherence would not show any relevant improvement, since it is likely that they deliberately stopped taking ICS. The adherence rate in this group is not likely to be improved by the text-message intervention. The group with excellent adherence is also not expected to show improvement, since adherence is already nearly 100%. Adherence in both intermediate groups (poor, good adherence) is expected to improve by 10-15%. This estimated effect size was based on an adherence improvement reported in a systematic review on the effect of (non-tailored) reminder systems on patient adherence to treatment [55]. Since the time-tailored text-message reminders used in our study are considered potentially more effective, we estimated the improvement at 15%.

Using these assumptions, we have simulated the adherence data of the control and treatment groups. Since the four adherence subgroups cannot be analyzed in one single regression model, we used a mixture of regressions (“mixture model”) in order to assess the effect of the intervention. We also calculated levels of statistical power at different group sizes. Requiring a power of at least 0.8 and assuming an adherence improvement of 15%, we calculated that a group size of 110 per arm is needed to detect the expected difference. Data analysis is based on an intention-to-treat principle. All patients with a follow-up of at least three months, regardless of whether they actually finish the intervention, are included in the analysis. The two groups are compared for baseline characteristics. Co-variables that may influence adherence levels, and therefore may confound the effect of the text-message intervention on adherence, are added to the multivariable model. In addition, a sensitivity analysis is carried out using a per-protocol approach. In this analysis the effect is studies of patients who complete less than three months of follow-up and of patients who appear to have stopped using ICS early in follow-up (less than 1% of total prescribed inhalations are administered during the complete follow-up). Data are analysed with SPSS for Windows.

For calculation of the cost-effectiveness a prospective economic evaluation from a societal perspective is performed alongside the clinical trial. The one-year costs of all relevant health care utilization are included as well the direct non-healthcare costs and the costs of productivity losses when parent stay home to take care of their children. Costs will be related to adherence, asthma control and asthma-specific quality of life to calculate the following incremental cost-effectiveness ratios (ICERs):

Subsequently, a decision-analytic model is developed that includes different levels or forms of adherence and the outcomes, both clinical and costs, attributed to each level or form of adherence as well as different price levels for RTMM. For the base-case, this model is filled with estimates of the relationship between adherence on the one hand and asthma control, symptoms, exacerbations, quality of life and healthcare utilization on the other hand. These estimates are obtained from the clinical trial, where potential associations between adherence and outcomes are studied. This model is used to run extensive one-way and multivariate sensitivity analyses to simulate the anticipated benefits of improved adherence in terms of health outcomes and costs.

A strength of this study is the use of RTMM as an objective and reliable method for adherence measurement. This method provides minimal room for bias, e.g. by socially acceptable patient response (patient self-report), misjudgement of patient behaviour (adherence questionnaires) and overestimation of adherence based on pharmacy refill data (refill rate, persistence) [13, 59, 60]. The RTMM device has been designed as a small add-on to the ICS inhaler. Since it does not need to be carried separately, it provides a patient-friendly way of measuring and stimulating adherence to ICS. This multi-centre study is the first to investigate RTMM with text-message reminders in a large sample of children with asthma. It is also the first to study the cost-effectiveness of RTMM in asthma. More data on cost-effectiveness are needed since the costs of this innovative technique are still substantial (approximately €750,= per patient per year) and are keeping physicians from using it in daily clinical practice. Health insurance companies also require more data on cost-effectiveness before covering costs for applying RTMM in asthma therapy.

Although electronic monitoring such as Real Time Medication Monitoring is considered more sensitive for measuring non-adherence than other, subjective tools for adherence measurement [13, 61, 62] the actual adherence to ICS may still be overestimated. All participating patients are aware that they are being observed, so they may act more adherent than in average daily practice. A common critique of electronic medication monitoring based on the time and date the inhaler is fired, is that it cannot be confirmed that the medication is actually taken. Only drug assays can confirm ingestion. However, studies comparing the sequence of medication events with projected and periodically measured concentrations of the drug in plasma, confirmed the validity of medication event monitors. Mismatches between medication events and actual dosing were too rare to create substantial differences between projected and actual concentrations of the drug in plasma [63‐66]. Another concern with adherence measurement of ICS is the fact that registered doses may not have been administered correctly due to poor inhalation technique. This may have a negative influence on the effectiveness of ICS therapy [67] and therefore on asthma related quality of life and on patients’ motivation to adhere to therapy. This phenomenon is considered evenly distributed within intervention and control group, so we expect that the effect on the outcome measures adherence to ICS and “asthma control” is limited. A potential limitation of this study is the high quantity of outcome-measures and co-variables (Table 1). This may provoke partial non-response, leading to missing data. Another concern is the fact that both children using fluticasone and those using a combination of fluticasone and the long acting beta-agonist salmeterol are included into this study. It is well known that co-inhalation of a long acting beta-agonist causes bronchodilatation resulting in a relief from asthmatic symptoms. This may be rewarding for the asthma patient, possibly resulting in a better adherence. Besides, patients needing a combination of fluticasone and salmeterol may have more severe asthma than those who’s symptoms can be sufficiently controlled by ICS alone and may therefore be better motivated to adhere to their asthma therapy. To overcome this limitation we collect data on the type of ICS (fluticasone or fluticasone/salmeterol) as a co-variable, which enables us to include it as a confounder in the multi-variable analysis or to perform stratified analysis.

One of the inclusion criteria of this study is the use of ICS for at least three months. This is verified by checking medical records and by asking potential participants which drugs are used for asthma. This procedure, however, does not account for patients who have stopped using ICS without consulting a physician. If a part of the patients that are not included for not using an ICS still had an indication for taking ICS, they have a 0% adherence rate. Patients who, on the other hand, are included into the study, but in fact already have stopped using ICS, also have a 0% adherence rate. Since these phenomena are expected to be equally distributed among patients in the intervention and control group, the only potentially relevant effect is a decrease in statistical power. In order to quantify the effect of the latter (patients, who stopped using ICS but still enter the study) a sensitivity analysis is carried out in which the patients who took less than 1% of prescribed doses are excluded.

It is expected that RTMM has most value in patients with therapy resistant poor asthma control. In daily practice, it is often unclear whether the prescribed asthma treatment is suboptimal (e.g. dose is too low, inconvenient inhaler) or the treatment is adequate, but the patient does not adhere to it. In the current study, however, we have chosen not to make a pre-selection of patients with poor asthma control or (suspected) non-adherence. Instead, asthma control is measured during the entire follow up, which enables us to investigate if poor asthma control at baseline is associated with response to the RTMM intervention.

It is crucial for correct sending of text-messages and for correct adherence measurement that any changes in mobile telephone (used for receiving text-message reminders), ICS dose, ICS dosing frequency and type of ICS, is correct at any moment during the study period. To ensure this, patients are requested to inform the investigators about any relevant changes. In addition these data are verified in the patient interview each three months of the study period. If RTMM-devices are detected that have not been actuated for more than a month, patients are contacted once to check for technical failures. This intervention is documented.

The trial-based cost-effectiveness analysis proposed here aims to explicitly estimate the cost-effectiveness of RTMM. However, at this early stage of development of RTMM with text-message alerting, adherence in stead of asthma control was used as the primary outcome measure. Therefore, the trial may not allow definite conclusions on the impact of this intervention on the cost of asthma treatment. Hence, we will apply appropriate decision-analytic modelling techniques to simulate the anticipated benefits of improved adherence. Such a model needs to relate the different levels of exposure to ICS to levels of asthma control. A model like that allows extensive sensitivity analyses on both clinical outcomes and costs that are attributed to each level of adherence.