The CHAT trial: Time for a Heart to Heart Talk

Mobile technology is pervasive and increasingly used for healthcare delivery given its accessibility and affordability. Mobile health (mHealth) interventions may be delivered in the form of text-messages, which have the advantages of broad delivery at low cost, or through mobile applications and wearable devices, which provide rich contextual and physiologic information. Digital interventions have the potential to revolutionize healthcare delivery, both for primary disease prevention and to reinforce behavior modification following a clinical event. While digital interventions offer great promise for shaping health behavior and improving clinical outcomes, these interventions have at times outpaced the medical community’s ability to systematically study them. Thus, there is an urgent need to determine which digital interventions are necessary and sufficient to mediate change and which populations benefit from which particular interventions.

In this issue of Circulation Cardiovascular Quality and Outcomes, Zheng et al report the results of their multicenter, randomized controlled trial (RCT) of a culturally-sensitive text message intervention for the secondary prevention of coronary heart disease (CHD) in China.¹ In this study, patients with a history of an acute myocardial infarction (MI) or percutaneous coronary intervention (PCI), but without co-morbid diabetes mellitus, were randomized to receive six text messages weekly or to receive two thank-you messages monthly, in addition to usual care. Text messages were developed iteratively by experts in medicine and behavioral theory and provided education to participants on the importance of blood pressure control, medication adherence, physical activity, and smoking cessation. The primary outcome was the change in systolic blood pressure (SBP) between baseline and six-months. Acceptability and utility surveys were administered to subjects in the intervention group at the end of the trial.

The study enrolled 822 participants from 37 hospitals across China between August 2016 and March 2017. Mean age was 56.4 years, and 14.1% were female. The majority of subjects had completed junior or senior high school and were from an urban environment. Mean SBP on enrollment was 131.1 mmHg and approximately 73% of study participants had a SBP less than 140 mmHg, making most subjects at or near their goal blood pressure at enrollment. Mean body mass index (BMI) was approximately 26 kg/m² and mean LDL cholesterol (LDL-C) was 96.8 mg/dL. At 6-months, SBP had declined in both the intervention and the control groups, and there was no statically significant difference in the change in SBP between the two groups (−1.3 [95% CI, −3.3,0.8], p = 0.221). There were also no significant differences between the two groups with respect to change in LDL-C, BMI, physical activity, or smoking status between baseline and 6-months. Despite finding no differences between groups with respect to the primary and secondary clinical endpoints, the study did demonstrate that health-oriented text messages can be effectively delivered to a diverse population living in a resource limited-setting.

The CHAT trial has many strengths, and the authors should be commended for using a rigorous study design to assess the impact of an mHealth intervention on cardiovascular outcomes. While the CHAT trial was designed to mirror the TEXT ME trial, which was conducted in a tertiary care-center in Australia, the CHAT trial reached a strikingly different conclusion. The TEXT ME trial demonstrated significant differences between groups across its primary and secondary clinical endpoints at 6-months that included LDL-C and SBP.² The marked differences between study results are instructive, however, in that they highlight the importance of studying mHealth interventions in each population in which they will be introduced. This is analogous to the study of established pharmaceuticals applied to new populations or for the treatment of new diseases. Secondly, the intervention design of the CHAT trial was strongly grounded in behavioral theory. Prior work evaluating mHealth interventions for secondary prevention of CHD have suggested that the most effective interventions are those which utilize theoretical foundations for behavior change and which have high usability.³ Finally, the CHAT trial differentiates itself from prior studies in its scale, broad inclusion criteria, and implementation in a resource poor setting. This last point is important since many mHealth interventions have been discussed in the global health context.

While the CHAT trial makes important contributions to our understanding of how to most effectively use, or not use, digital health interventions, it perhaps raises more questions than it answers. Why did the CHAT trial fail when the TEXT ME trial was positive across all cardiovascular endpoints? Herein, we offer our speculation as to the limitations of the CHAT trial specifically and our insights as to next steps for digital health technology as a whole. First, the trial enrolled patients who had already achieved at least modest control of secondary risk factors for cardiovascular disease. Thus, text messages as a passive intervention may be effective in populations with uncontrolled cardiovascular risk factors. Secondly, most patients had a history of relatively remote MI or PCI. As a result, individuals may have been less motivated to institute behavioral changes and reached a relative plateau with respect to further improvements. Third, text messages allowed for minimal customization aside from inserting the patient’s name and limiting smoking cessation counseling to active smokers. Finally, and perhaps most importantly, the study highlights limitations of large RCTs for mHealth interventions. Such trials are designed to assess the average efficacy of interventions that typically represent static features.⁴ Randomized controlled trials are less well-suited to ascertain complex mHealth-type interventions or their components – and importantly which time-varying contextual and psychosocial factors may moderate their observed effects.

In order to develop patient-centered digital interventions that can be consistently and effectively translated into clinical care, we believe additional innovation is needed. This includes (1) use of rigorous alternatives to the traditional randomized controlled design to optimize interventions prior to a confirmatory RCT; (2) emphasis on development of patient-first rather than technology-first interventions; and (3) further integration of mHealth technology with clinical care.

First, alternative study designs are needed to efficiently and effectively optimize and interrogate digital interventions.^5–7 Micro-randomized trials are one such alternative and involve the random assignment of an individual at each decision point of a multicomponent intervention.⁴ These trials, rooted in behavioral theory, examine how a participant’s prior behavior and current context influence the efficacy of individual intervention components over time in an effort to optimize intervention delivery to maximize effectiveness and minimize user burden. If designed effectively, such a strategy can reduce the number of notifications participants receive, enable study of each intervention component separately and in combination, and allow for evaluation of the time-varying effects of an intervention. Secondly, we believe there needs to be a shift to the development of patient-first rather than technology-first interventions.⁵ With such a strategy, interventions will be developed that more precisely match patients’ needs and barriers to change. A menu of theory-based intervention components could then be parameterized for patients based on their unique challenges. Such interventions would be highly usable and afford greater integration across disease states.³ In the current delivery model, patients with multiple disparate medical problems need to seek different digital interventions.⁸ Finally, we advocate for greater integration of digital technology with clinical care. On the most basic level, this requires coordination with patients’ primary care physicians or subspecialists. The TEXT ME trial, unlike the CHAT trial, was conducted out of a single tertiary care center in Australia. Thus, study participants may have associated the messages with their clinical care team, in part mediating the observed effects. In the case of wearable technology with its rich data streams, this will require further investment in analytics and data visualization to ensure that digital information from wearables is rapidly and efficiently merged with patients’ existing health information and then distilled to clinically relevant information so that patients and providers can participate in shared decision making.⁵

In conclusion, Zheng et al present the results of their rigorous, multicenter RCT of culturally-sensitive text messages for the secondary prevention of CHD. While there were no significant differences between study groups with respect to the primary and secondary clinical endpoints, the study demonstrated convincingly that an intervention grounded in behavioral theory could be effectively and broadly delivered in a resource limited setting in a manner acceptable to study participants. This is a critical step for digital health technology and shows the potential to disrupt healthcare delivery. Yet further development is obviously needed before it can deliver personalized and effective care for the masses. Evaluations of mHealth interventions like the CHAT trial help us shift from just “talking the talk” to “walking the walk.”