The results of this pilot study suggest that real-time adherence monitoring is feasible in a resource-limited setting. The initial 3 months of data collection revealed significant technical problems; however, most were resolved during the course of the study. Minor modifications to the sleep mode extended the battery life to nearly 4 months for almost all issued devices, and SMS signal transmission became reliable through upgrades to the forwarder. The SIM card expiration after the conclusion of this pilot study occurred despite monitoring the level of airtime available. Research staff were unaware that the network coverage provider automatically inactivates the SIM cards if no airtime is added during a 6-month period, and contract pre-payment plans are not available. This finding underscores the importance of in field-testing for novel technologies.
Although the adherence levels assessed with Wisepill were similar to those made with prior electronic monitoring through MEMS, they were lower than the adherence levels determined by unannounced pill counts and self-report. It is not known which measure is correct, as there is no gold standard for adherence measurement. Moreover, the sample size in this pilot study is too small to determine the statistical significance of any differences among the measurements. The apparent discrepancy between unannounced pill count and electronic monitoring, both of which are objective measures, may be due to patients returning unused medications to the pharmacist during refills and other inconsistencies during medication disbursement. Alternatively, participants may have removed more than one pill at a time or further technical problems with Wisepill could have gone undetected, thus leading to falsely low adherence measurements. Undetected technical problems may also explain some of the difference in Wisepill adherence in the first and second 3-month periods. Additionally, the adherence levels as measured by both MEMS and Wisepill were lower than similar data from other studies [9
], which may reflect the relatively long duration of treatment of the participants in this study. The decline in adherence is consistent with data showing that highly adherent individuals may experience a modest decline in adherence over time [9
Both 3-month time periods had similar rates of lapses in the Wisepill signal that could not be classified as a specific technical or behavioral cause. These unclear lapses may reflect missed doses that were not reported due to social desirability bias against disclosure of missed doses. This type of bias is common and can lead to under-reporting of incomplete adherence [12
]. Interventions to address the lapses detected through Wisepill will therefore need to consider culturally acceptable means of determining both the cause of missed doses and appropriate responses. It is also possible that lapses of unclear cause could indicate undetected technical failures, despite the ongoing checks of signal strength, airtime, and battery power, as well as transmission of signals stored during temporary loss of the cellular network. To further decrease the likelihood of technical failures, the manufacturer is currently adding an automated daily signal, which will serve as a positive control of the device’s functionality.
Interventions targeted to individuals with known adherence lapses have been shown to be more effective than untargeted interventions [13
], and real-time adherence monitoring could be used to direct interventions to specific missed doses prior to virologic rebound. Such a “just-in-time” approach may increase effectiveness further [15
]. For example, automated SMS reminders could be triggered by missed or late doses, or a community health worker could call or personally visit a patient during an adherence lapse to intervene before viral rebound.
The pilot data showed that real-time adherence monitoring was acceptable to all participants. Many found Wisepill desirable, easy to use, and/or thought it helped them remember to take their doses. One limitation to these findings is that the study population consisted of individuals already accustomed to the use of MEMS caps, which may select for acceptance of the Wisepill device as well. The generalizability and sustainability of this initial enthusiasm will have to be determined in larger populations over time, and these assessments will be critical for the scalability of this technology. Several theories and models of technology acceptance have been developed to best understand the experience of the end user, such as the Technology Acceptance Model and the Unified Theory of Acceptance and Use of Technology [17
]. These models build upon the theory of reasoned action [20
] and include elements such as perceived usefulness, perceived ease of use, output quality, social influence, attitude, behavioral intention to use, and actual use. Moreover, standard electronic medication containers have known facilitators and barriers to uptake, such as the container acting as a reminder and incompatibility with pill boxes, respectively [21
]. While this study was too small in scope to inform a theoretical model, future efforts to implement real-time adherence monitoring should consider underlying theory and previous research.
Several individuals in this study experienced transient low-level viremia, consistent with blips [22
]. While transient low-level viremia has been associated with transient periods of incomplete adherence [23
], we were unable to assess such an association in our small pilot study. This study does, however, provide proof-of-concept that participants can be contacted during potential adherence lapses. In future studies, HIV RNA could be assessed during the adherence lapse to determine how quickly viral rebound occurs in this setting and to what extent it can be re-suppressed without the development of drug resistance.
Studies have shown that ART adherence predicts viral rebound, drug resistance, and morbidity and mortality from HIV/AIDS [24
]. Furthermore, relatively crude pharmacy refill adherence monitoring may be a better predictor of viral rebound than the standard practice of CD4 response for detecting treatment failure [27
]. More accurate adherence monitoring, such as can be achieved through a real-time approach, may help prioritize limited resources for HIV RNA monitoring to those at increased risk for viral bound.
The findings of this study reflect a small sample in a research context. Future research should aim to expand the number of participants and establish the data and management structure necessary to test the efficacy of detecting virologic rebound in real time, and ultimately how such real-time detection of virologic rebound can be linked to interventions that improve patient outcomes in clinical practice. Cost is another concern. Although the Wisepill device currently costs US$185; mass production could lower the price to US$20–30. Significant cost savings could be achieved through economies of scale, as well as targeted use of HIV RNA monitoring and reduced need for second line therapies. Such savings could bring real-time adherence monitoring into the same cost range as CD4 monitoring.
In sum, real-time adherence monitoring defines a feasible and acceptable strategy to proactively prevent rather than reactively respond to virologic treatment failure and thereby extend the duration of first-line therapy. Further studies should build on this pilot data to determine the efficacy of real-time intervention for sustained adherence and its potential to act as a surrogate for HIV RNA testing.