## Results

### Main analysis

The model predicted that the DTx + TAU strategy produced 18.778 QALYs and was associated with ¥3,924,075 ($34,122) expected costs compared to 18.686 QALYs and ¥3,813,358 ($33,160) generated by TAU-only treatment over a lifetime horizon. The introduction of DTx would increase costs by ¥110,717 ($962) and prolong QALYs by 0.092. The ICER of DTx + TAU against TAU would be ¥1,199,880 ($10,434)/QALY gained, which was well below the hypothetical threshold value in Japan of ¥5 million/QALY gained (Table 2).

## Discussion

In this study, we assessed the cost-effectiveness of prescription DTx for essential hypertension, in addition to TAU, compared to the TAU-only strategy. We found that the ICER of the DTx + TAU strategy was ¥1,199,880 ($10,434)/QALY against the TAU-only strategy in the lifetime horizon, which was well below the threshold value of ¥5 million/QALY. Moreover, the probability of the cost being below the threshold value was 87.8% in the probabilistic sensitivity analysis. This model analysis demonstrated that DTx + TAU is cost-effective for patients with hypertension. Several conclusions can be drawn from this study. First, this is the first study to evaluate the cost-effectiveness of DTx for patients with hypertension using the results of an RCT. Nordyke et al. reported the cost-effectiveness of DTx for hypertension with person-to-person health coaching according to data from a retrospective study over a 3-year time horizon [35, 36]. Other than hypertension, DTx for opioid use disorder and that for low back pain also revealed cost-effectiveness in a relatively short time horizon (12 weeks to 3 years) without considering mortality rates [13,14,15]. Although the time to half DTx usage rate was approximately 15 months in this model, the benefit of the BP-lowering effect persisted for more than three years, resulting in prevention of the trajectory of CVD from hypertension, left ventricular hypertrophy, and atherosclerotic events to HF and CVD deaths. Therefore, a long-term cost-effectiveness analysis would be valuable when considering the practical effects of DTx for chronic diseases such as hypertension. The monthly cost of DTx for hypertension had the greatest impact on ICERs in the one-way deterministic sensitivity analysis. Lewkowicz et al. pointed out a similar conclusion by evaluating the cost-effectiveness of DTx for low back pain [15]. At a DTx cost of ¥5,000/month, DTx dominated conventional therapy or was a less costly and more effective option. Even when we set the cost at ¥20,000/month, the ICER was approximately ¥3.5 million/QALY, which is still under the threshold value of ¥5 million/QALY (Fig. 2). Recently, sacubitril/valsartan (angiotensin receptor-neprilysin inhibitor), a novel antihypertensive drug, was approved for hypertension management in Japan [37], with a medication cost of ¥6,057 ($52.67) to ¥12,114 (\$105.34)/month. Thus, the DTx cost we set to approximately ¥10,000/month as the default might be a reasonable value according to the current simulation model.

The attrition rate of the DTx program is a crucial factor for cost-effectiveness. A smartphone mHealth app usage rate attenuated over time, reporting that approximately 40% of attrition occurred from 6 months to 1 year [38]. Previous DTx studies used for cost-effectiveness analyses also used similar attrition rates [15, 35]. Thus, it could be reasonable to set the attrition rate to 25% for every 6 months in this model, which was approximately 15 months as time to half usage rate since the initiation of DTx. In another scenario, when we changed the rate to 10% (lower than the predefined 25%), the time to half exceeded approximately 40 months, and the ICER was increased to approximately ¥4.3 million/QALY in a lifetime horizon, which was close to the threshold value (Fig. 2). Lewkowicz et al. previously reported that lowering the DTx app attrition rate was essential for cost-effectiveness in a relatively short-term (3-year) horizon [15]. Although the model inputs and target disease were different, it might be possible that high attrition rates along with DTx actual usage duration could substantially impact DTx cost-effectiveness. Achieving good cost-effectiveness for DTx might require sensitive handling to balance the appropriate DTx app usage duration with DTx costs and expected attrition rate.

The degree of the gained ICER using DTx hypertension management surely depends on the degree of BP reduction by quality of the DTx app algorithm, the impact of BP reduction, and the cost of cardiovascular events in each country. A significant difference in home BP reduction (−4.3 mmHg of morning home SBP) was clearly found between the DTx + TAU and TAU-only groups from the previous phase III clinical trial. Notably, this TAU-only group followed the ideal guideline-driven management of hypertension, which used home BP monitoring [7]. Compared with office BP, home BP is more closely associated with cardiovascular event risk [39, 40]. In addition, in Asian countries, the benefit of BP reduction is greater than that in Western countries, especially for stroke [41]. The comparison of cost-effectiveness studies using different DTx apps in different contigs is also needed in the future.

The strength of this study is that, to our knowledge, this is the first study to evaluate the cost-effectiveness of DTx for patients with hypertension using the results from an RCT. Furthermore, we demonstrated that DTx + TAU is cost-effective for patients with hypertension. This study has some limitations. First, we used middle-aged, digital-friendly grade I or II hypertensive patients who were not taking antihypertensive medication for the simulations according to previous RCT results. Thus, caution should be taken when applying our results to a broader population. Second, we modeled the DTx effect as permanent, although the effect estimates were derived from the 6-month duration of the HERB-DH1 trial results. Thus, it might be possible that the treatment effect could attenuate or diminish over time, especially six months after the intervention. Although we set the BP reduction effects even for the TAU-only strategy and age-related annual SBP increases in both groups, this could lower the mortality and complication occurrence rates in the DTx + TAU strategy, resulting in a higher ICER than we concluded in this study. Third, we did not estimate the additive effects of having multiple comorbidities in this model, which could impact lifetime ICER. Fourth, we could not strictly assess the DTx cost-effectiveness in addition to antihypertensive medications since the model patients derived from previous trials were all antihypertensive drug-naïve. Finally, we did not consider reusing DTx after finishing the initial DTx usage according to the attrition rate, which could account for additional costs for the DTx strategy.

In conclusion, DTx + TAU demonstrated good cost-effectiveness in hypertensive patients compared to the TAU-only strategy. DTx cost and its attrition rate largely influence cost-effectiveness, and we need to explore the balance among the DTx application cost, app attrition rate, and effectiveness for target hypertensive patients in clinical practice. DTx is generally reimbursed by Japan’s national universal health insurance system. Thus, it is becoming increasingly essential to evaluate the cost-effectiveness of DTx, as was done in this study, to consider the appropriate value-based reimbursement pricing.