Introduction

Oral pre-exposure prophylaxis (PrEP), either daily or event-driven regimen1,2, is an evidence-based HIV prevention strategy for high-risk populations, including men who have sex with men (MSM)3. However, low adherence to oral PrEP directly undermines its efficacy4. The protection diminishes at below 40% adherence5, leading to increased risks of HIV transmission and drug resistance6. Prior researches have demonstrated that suboptimal adherence to oral PrEP regimens is associated with multifactorial determinants, including non-adherence due to forgetfulness, safety concerns regarding pharmacological agents, inadequate health literacy, socioeconomic constraints, and psychosocial challenges7,8. Therefore, optimizing oral PrEP adherence is critical for its implementation.

The COVID-19 pandemic and the Mpox outbreak highlighted the need for digital health solutions9,10. Digital adherence technologies, including short message service (SMS) reminders, geosocial health platforms, and dose-confirmation monitoring systems, show clinically validated efficacy in optimizing therapeutic compliance among individuals with HIV/AIDS or chronic conditions11,12. These interventions mediate adherence improvements through multiple mechanisms: mitigating dose-timing inaccuracies, delivering tailored therapeutic education, and implementing reinforcement learning frameworks. However, previous studies on SMS-based oral PrEP adherence interventions yielded inconsistent results. A study in Chicago showed that an interactive text-messaging intervention improved PrEP adherence among young MSM13. Conversely, SMS-based interventions did not improve PrEP adherence among black MSM and transgender women in New York14, vulnerable young people in Los Angeles and New Orleans15, young women in Kenya16, MSM in China17, or male sex workers in the US Northeast18. Thus, SMS reminders alone may not adequately engage participants19, necessitating updated, multifaceted digital interventions to improve oral PrEP adherence.

Despite heterogeneous behavioral trends observed across various studies, high-risk sexual behaviors, e.g., condomless anal intercourse (CAI) and multiple partners, persist among PrEP users, significantly hindering HIV/STI prevention. Seattle PrEP users reported a 46% increase in condomless sex after 12 months20, whereas a Dutch AMPrEP trial documented declining CAI with casual partners21. Digital interventions may enhance PrEP use15 or reduce HIV risk behaviors22 through real-time monitoring and algorithm-driven behavioral nudges. Emerging evidence suggests anonymized peer networks mitigate stigma-related barriers23, while simulated clinical engagement extends intervention effects to sexual health decision-making24. However, rigorous quantification of direct behavioral impacts (e.g., partner concurrency, condomless acts) remains lacking.

While oral PrEP effectively prevents HIV, challenges in uptake, persistence, and adherence persist. Long-acting PrEP may enhance adherence and provide more stable protection25,26. However, cost, availability, and provider training barriers impede the widespread implementation of long-acting PrEP. Thus, interventions are required to enhance adherence to oral PrEP, a less demanding and lower-cost option to foster better health outcomes. Most studies on digital interventions focused on daily oral PrEP in high-income countries3,14,27, with limited data on event-driven oral PrEP or low- and middle-income countries. Therefore, we conducted a randomized controlled trial (RCT) in China to evaluate the efficacy of digital intervention for real-time enhancement of oral PrEP adherence (DIRECT-PrEP) over 6 months and assess their impacts on high-risk behaviors.

Results

Participant characteristics

Participants were recruited from Sept. 20, 2019, to Jul. 31, 2020. Follow-ups were completed on Nov. 27, 2020. Of the 516 invited individuals on PrEP, 442 (85.7%) were included and randomly assigned to the intervention (247) or control (195) group. The intervention and control groups included comparable proportions of daily (150 vs. 105, respectively) and event-driven (97 vs. 90, respectively) PrEP users, respectively. The follow-up completion rate was 88.2% (390/442) at 3 months and 96.6% (427/442) at 6 months, with higher rates in the intervention group (94.7% and 98.4%) than in the control group (80.0% and 94.4%) at both time points (Fig. 1).

Fig. 1: Study flowchart.
figure 1

Not completed signifies that no data were gathered due to the complete absence of the visit, primarily because participants were unable to travel amid the COVID-19 pandemic. While some individuals missed a visit, they eventually resumed participation in subsequent visits. PrEP pre-exposure prophylaxis.

Full size image

Baseline characteristics showed that 53.8% (238/443) of the participants were under 30 years old, and 81.7% (361/443) held a college or higher degree (Table 1). Of the 1659 missed doses detected, 81.9% (1358/1659) were effectively monitored, and 18.1% (301/1659) in the event-driven PrEP subgroup were excluded due to lack of sexual activity. Among the monitored missed doses, 99.1% (1346/1358) were in the daily PrEP subgroup, 39.4% of which (530/1346) were remedied by system interventions and 20.4% (275/1346) by research assistants, while 39.8% (541/1346) remained as missed doses. Among the 12 monitored missed doses in the event-driven PrEP subgroup, 10 were remedied by system interventions (Supplementary Table 1). All smart pillbox issues experienced by 34 participants were resolved (e.g., changing reminders from sound to vibration and adjusting volume settings).

Table 1 Baseline characteristics of participants
Full size table

Primary and secondary outcomes

During follow-ups at 3 and 6 months, the intervention group demonstrated significantly higher PrEP adherence (83.3% and 83.1%) than the control group (60.9% and 59.8%), with adjusted net differences of 21.1% (95% CI: 12.5% to 29.7%) and 21.0% (95% CI: 12.9% to 29.2%; p < 0.001 for both). At the 3-month follow-up, both the daily PrEP and event-driven PrEP subgroups in the intervention group demonstrated significantly higher adherence (86.9% and 77.5%) compared to the control group (64.3% and 56.9%), with adjusted net differences of 21.1% (95% CI: 10.4% to 31.9%) and 17.5% (95% CI: 3.1% to 31.8%; p < 0.05 for both). Consistent trends were observed in the daily and event-driven PrEP subgroups at the 6-month follow-up. No significant between-group differences (all p > 0.05) were detected in the number of male sexual partners, CAI prevalence, or substance use disorder over 3 and 6 months (Table 2).

Table 2 Effectiveness of telemonitoring and timely interventions on oral PrEP adherence among 442 MSM
Full size table

Post-intervention adherence and behavior trends

Figure 2 displays the adherence and sexual behavior trends. Over 6 months, the event-driven PrEP subgroup in the intervention group exhibited a notably increased adherence from 70.1% to 82.3% (p = 0.042). In contrast, the adherence in the control group declined from 74.4% to 59.8% (p = 0.030), particularly in the daily PrEP subgroup (decreasing from 84.8% to 58.0%; p < 0.001). No statistically significant changes in the number of male sexual partners (from 4.4 to 3.3), the proportion of CAI (from 53.8% to 45.7%), and the proportion of substance use disorder (from 42.1% to 35.0%) were observed in the intervention group (all p > 0.05). According to stratified analyses by PrEP regimen, the intervention and control groups showed no statistically significant differences in sexual behaviors (number of partners and CAI) and substance use disorder at either time point (all p > 0.05) (Supplementary Table 2), nor did the trends over time (Supplementary Fig. 1).

Fig. 2: Cochran–Armitage test for trend of PrEP adherence and high-risk behaviors.
figure 2

Cochran–Armitage test of the trend of total PrEP adherence rate (a); Daily PrEP adherence rate (b); Event-driven PrEP adherence rate (c); the no. of sex partners in the past 3 months (d); the proportion of CAI with male partners (e); the proportion of substance use disorder (f). substance includes rush (poppers or alkyl nitrites), MDMA (3,4-methylenedioxymethamphetamine; ecstasy), ice, amphetamines, tramadol hydrochloride, or ketamine hydrochloride. pI the p value derives from the trend analysis in the intervention group, p C, the p value derives from the trend analysis in the control group, PrEP pre-exposure prophylaxis, CAI condomless anal intercourse.

Full size image

Sensitivity analysis

Sensitivity analysis showed consistent intervention efficacy on primary and secondary outcomes between pre- and post-imputation data. At the 3- and 6-month follow-ups, the intervention group also demonstrated significantly higher PrEP adherence (83.8% and 83.4%, respectively) compared to the control group (64.6% and 61.0%; p < 0.05 for both time points). These effects were robust across both the daily PrEP and event-driven PrEP subgroups. However, no significant intervention effects were observed on high-risk behaviors or substance use disorder (Supplementary Table 3). Sensitivity analyses of the trends in PrEP adherence, high-risk behaviors, and substance use disorder are detailed in Supplementary Fig. 2.

Discussion

To the best of our knowledge, this is the first assessment of the impact of combined telemonitoring and timely interventions on oral PrEP adherence and risk behaviors among MSM in low- and middle-income countries. The results suggested that this combination enhanced oral PrEP adherence without affecting high-risk behaviors or substance use disorder. This real-time intervention strategy fills a research gap in oral PrEP adherence and addresses barriers such as forgetfulness and irregular dosing through real-time reminders, dose tracking, and automated alerts. Smart pillbox holds significant public health value, particularly for high-risk populations, by maximizing the preventive efficacy of PrEP and reducing HIV transmission.

This study demonstrates that combining telemonitoring and timely intervention improves adherence to both oral PrEP regimens (daily and event-driven) among MSM. A study in Thailand indicated that enhanced risk recognition can affect PrEP adherence among young MSM28. Real-time monitoring of oral PrEP adherence can potentially improve risk awareness and daily medication habits among MSM. Furthermore, electronic medication monitors identified suboptimal adherence, thereby enabling precise interventions. However, SMS reminders derived from electronic medication monitors failed to improve oral PrEP adherence due to potential message overload16. Thus, research assistants provided more individualized interventions via WeChat or phone calls in the case of text message intervention failures.

In the early stages of this study, both groups showed high adherence. However, adherence declined in the control group while remaining stable or improving in the intervention group, a pattern consistent with previous PrEP studies. For instance, a study in America targeting young MSM across 12 cities revealed that most participants adhered to four or more tablets weekly during the initial 12 weeks, but adherence fell to 34% by week 4829. Similar decreases were observed among black MSM and transgender women in New York14 and young women in Kenya20. These findings highlight the role of digital remote interventions in maintaining oral PrEP adherence, as system reminders and healthcare provider support foster a supportive environment for sustained adherence.

In addition, this study demonstrated fundamentally distinct adherence trajectories between daily PrEP in the control and event-driven PrEP in the intervention group, shaped by their unique behavioral engagement patterns. The progressive decline in daily PrEP use reflects challenges in sustaining preventive behaviors for infrequent risk exposures29,30, while event-driven PrEP’s improvement stems from its natural synchronization with sexual activity patterns31. Notably, digital adherence tools showed regimen-specific effectiveness, performing best when their design complemented the intrinsic motivation structure of each approach. This evidence underscores the need for differentiated implementation strategies that account for these behavioral determinants, moving beyond one-size-fits-all approaches in HIV prevention.

Despite electronic medication monitors and research assistants’ interventions, missed doses persist, underlining the multifaceted influence on adherence. Thus, a holistic approach that considers social support, economic status, and education level is required to promote oral PrEP adherence32. Such a comprehensive approach combining monitoring, timely feedback, and a collective analysis of adherence barriers can motivate behavior change. In Kenya, community outreach and partner engagement enhanced oral PrEP adherence33. Telemonitoring and timely intervention may appeal more to those facing obstacles to oral PrEP adherence, particularly those with occupational constraints requiring irregular schedules or extended work hours. It is also vital to examine the characteristics of PrEP users receiving no benefit from real-time interventions. Future research will analyze non-recipient characteristics to guide PrEP implementation.

Although interventions using electronic medication monitors, high-risk behaviors among MSM using PrEP remained unchanged in this study. No statistically significant differences were detected in daily or event-driven subgroups. Researches on changes in sexual behavior post-PrEP use are inconsistent. A previous study found no effect of PrEP on the number of sexual partners, or sexual activity frequency20, while a PrEP study from Belgium reported decreased condom use with casual partners but increased with partners of unknown HIV status34. Current PrEP programs and adherence interventions do not focus on reducing high-risk sexual behavior or substance use disorder. Future strategies should consider integrating behavioral interventions with oral PrEP dispensation.

While long-acting medications represent a promising advancement in HIV treatment and prevention25,26, they are not the sole option. Meanwhile, future anti-HIV strategies may consider various mechanisms, treatment stages, and clinical scenarios. Thus, long-acting PrEP alone cannot eliminate AIDS despite its potential, and the needs of diverse populations necessitate innovative, people-centered approaches. Providing high-risk individuals with poor oral PrEP adherence with multiple options, including long-acting formulations, is crucial. Due to their user-centric, personalized, and adaptable features, low-cost, reminder-based daily oral PrEP options and event-driven oral PrEP can serve as essential alternatives. Accordingly, effectively combining various preventive strategies for different populations is critical, demanding stakeholder collaboration.

While smart pillboxes offer substantial public health potential, their implementation faces real-world challenges35,36,37: limited user acceptability due to privacy concerns and technical complexity, particularly among older or technologically inexperienced individuals; sustainability constraints stemming from waning engagement without behavioral reinforcement; structural barriers including cost and healthcare infrastructure limitations; and imperfect data accuracy from manual overrides or device errors. To maximize impact, future interventions should combine these technologies with behavioral support strategies while prioritizing user-centered design, ensuring both effectiveness and scalability across diverse populations.

This study has several limitations. Firstly, it confirmed that telemonitoring and timely interventions enhance adherence among MSM, but their efficacy in other high-risk groups warrants further investigation. Secondly, free electronic medication monitors may not reflect real-world scenarios where purchasing is more often. Thirdly, though most variables exhibited no significant differences between the China Real-World Oral Intake of PrEP (CROPrEP) participants and non-participants (Supplementary Table 4), COVID-19 restrictions did limit recruitment. Furthermore, the baseline PrEP adherence varied between the intervention and control groups, regardless of whether they were assigned to the daily PrEP or event-driven PrEP subgroup. Accordingly, we adjusted for the PrEP regimen variable in both primary and secondary outcomes. Additionally, the evident non-adherence among MSM who prefer nocturnal activities indicates a need for alternatives to manual reminders, such as smart pillboxes or potentially AI-driven 24/7 reminders38. Large language models could promote anonymous discussions about adherence challenges. However, evaluating the effectiveness of AI interventions requires larger samples and rigorous designs.

In conclusion, this trial found that telemonitoring and timely interventions enhance oral PrEP adherence among MSM, thus addressing prevention failures across medication modalities. However, these interventions exhibited no effect on high-risk behaviors or substance use disorder. Until long-acting PrEP is affordable and accessible, digital technologies like smart pillboxes, a key component of the people-centered HIV intervention strategy, can be leveraged for personalized support.

Methods

Study design and participants

The CROPrEP prospective interventional study assessed HIV prevention efficacy of emtricitabine-tenofovir disoproxil fumarate (F/TDF) among MSM in Shenyang, Beijing, Chongqing, and Shenzhen39. Participants were assigned to the PrEP users group (daily PrEP or event-driven PrEP group) and PrEP non-users groups depending on their personal decision at the baseline of CROPrEP. For the daily PrEP group, participants took one tablet (F/TDF) per day. The event-driven PrEP group followed the “2-1-1” dosing regimen. The “2-1-1” regimen involves taking 2 tablets 2 to 24 h before sex, followed by taking 1 tablet 24 and 48 h after the first dose. In case of additional sex within 48 h, participants switch to 1 tablet/day until 48 h after the final act. In case of the next sex occurring <7 days after that, the 1 tablet/day regimen continues. Otherwise, the “2-1-1” regimen restarts. Within follow-up, participants in the PrEP users groups could switch between daily and event-driven PrEP groups.

This study was a multicenter, non-blinded, two-arm RCT. MSM community opinion leaders and volunteers at the CROPrEP sites in Shenyang, Beijing, Chongqing, and Shenzhen promoted the free use of smart pillboxes among participants of CROPrEP via WeChat. MSM who volunteered for this study were randomly assigned to intervention and control groups. Eligible participants were included based on the following criteria: assigned male at birth, aged 18 to 65, engaged in anal sex with males, self-reported at least one high-risk behavior in the past 6 months (CAI with male partners, two or more male sexual partners), any STI diagnosis, history of post-exposure prophylaxis but not in the previous month, HIV-1 negative, participated in the CROPrEP project, and capable of understanding the study procedures and willing to provide written consent. The trial protocol is provided in Supplementary (Supplementary Protocol). The first participant was enrolled on Sept. 20, 2019, and the final participant completed the study on Nov. 27, 2020.

Randomization and masking

Following baseline screening, participants of this trial were stratified based on their medication regimens and randomly assigned to either the intervention group or the control group (1:1) using envelopes containing random numbers. The researchers devised the randomization protocol, and the research assistants implemented the grouping at each site. Participant blinding was not feasible due to the nature of the distinct interventions.

Procedures

After randomization, participants attended follow-up visits at 3 and 6 months (plus or minus 7 days). At baseline and follow-ups at 3 and 6 months, participants provided blood samples for HIV (antibodies and pooling RNA) and active syphilis infection (Trepeonema pallidum particle assay and rapid plasma reagin) testing and completed an online questionnaire (https://jinshuju.net/home). The questionnaire included sociodemographic information (age, ethnicity, education level, marital status, monthly income, etc.) and sexual behavior characteristics (venues for seeking sex, types of sexual partners, sexual roles, number of sexual partners in the past 3 months, condom usage, substance use disorder, etc.).

Both groups received standard care at each follow-up, including tablet counts, medication supplementation, personalized adherence education, and face-to-face counseling on PrEP adherence. Meanwhile, participants discussed adverse events with experienced physicians.

The intervention group received smart pillboxes (Supplementary Fig. 3) from Shanghai Magnesium Health Technology Co., Ltd., which utilized narrowband IoT (NB-IoT) technology for medication real-time monitoring. The device integrates multimodal reminders (auditory, vibration, light, and WeChat push notifications) triggered by preset dosing schedules, with automatic cloud-based logging of box-opening events accessible via a WeChat Mini Program interface. The rechargeable 600 mAh lithium battery provides 45 days of runtime on a single 2-h charge (via Micro-USB). In case of a missed dose, the smart pillbox emits a reminder via vibration, sound, and light. If the missed dose is not rectified, the system sends real-time reminders to the participant’s mobile phone every 30 min (up to twice) and notifies research assistants. Upon notification, research assistants could review missed doses via a management platform, contact participants, and document information (Fig. 3). Doses missed before 22:00 were addressed immediately, while those after 22:00 were handled by 10:00 the next day. Repeated pillbox opening without medication intake triggers system alerts to research assistants for follow-up.

Fig. 3: The telemonitoring and timely intervention process diagram.
figure 3

Setting Medication Regimen: participants configured medication reminders in the smart pillbox according to their prescribed dosing schedule. Missed doses (initial occurrence): a missed dose event was recorded if the participant failed to take the medication at the scheduled time. Real-Time Reminders: the smart pillbox would automatically trigger a reminder 30 min after the scheduled time if a missed dose was detected, and again after 60 min if the initial reminder was ignored. Took the missed dose: the participant took the missed dose following the reminder from the smart pillbox. Research assistant intervention (phone/WeChat): if the participant did not respond to the pillbox’s secondary reminder, research assistants contacted them via phone or WeChat to deliver an adherence prompt. Missed doses (final determination): a missed dose was logged as missed if the participant failed to take the medication after two consecutive smart pillbox reminders and one research assistant intervention.

Full size image

To ensure proper functioning, the smart pillbox was tested by five staff members and ten PrEP users during the design phase. Based on feedback, training materials were improved. Custom videos were created for daily PrEP and event-driven PrEP. Prior to participant recruitment, on-site staff and research assistants underwent standardized training on the smart pillbox protocol. During implementation, researchers provided participants with detailed instructional materials, including customized demonstration videos, to facilitate personalized device setup. To ensure continuous adherence monitoring, real-time technical support was made available throughout the study period. Upon study completion, all participants were required to return the smart pillboxes as per the predefined protocol.

Outcomes

The primary outcome was overall PrEP adherence, measured by tablet counts during follow-ups at 3 and 6 months. Secondary outcomes included changes in adherence over the 3 and 6 months for both daily PrEP and event-driven PrEP subgroups. Additionally, the intervention’s impact on high-risk behaviors, such as the number of sexual partners, condom use, and substance use disorder, was evaluated over the 3 and 6 months.

Statistical analysis

Based on a previous study13, we hypothesized that 72% of the intervention group and 57% of the control group would achieve expected adherence. The sample size was calculated in PASS 15.0 (NCSS, LLC, Kaysville, Utah, USA), with each group requiring 156 participants to detect adherence differences (power = 80%, two-sided α = 0.05). Considering a 15% dropout rate, 184 participants per group (368 in total) were required.

Adherence was calculated as the ratio of tablets taken to the expected number and was deemed good if ≥90%40. The expected tablets number for the event-driven PrEP group (“2-1-1” dosing regimen) was calculated based on sexual activity dates, which were collected through WeChat-based online questionnaires and weekly messages. Sexual activity and medication intake data were obtained from participant-maintained weekly logs, with staff cross-verifying self-reports against pharmacy dispensing records and tablet counts during routine follow-up visits to ensure data reliability39.

Chi-square tests were conducted to compare baseline sociodemographic characteristics and high-risk behaviors. An intention-to-treat analysis was conducted, comparing outcomes based on randomization. Covariance analysis assessed the differences in average outcome changes, adjusting for age, education, income, and dosing regimen, as well as controls at baseline (number of male sex partners, CAI, and substance use disorder). Mixed linear models evaluated the effects of telemonitoring and timely intervention on PrEP adherence at 3- and 6-month follow-ups and the impacts on high-risk behaviors and substance use disorder. The Cochran-Armitage trend test was adopted to analyze the trends in high-risk behaviors and substance use disorder. Missing data during follow-ups were addressed via multiple imputations using the ‘mice’ package in R, applying the random forest method across 10 imputations and 50 iterations. Sensitivity analyses were performed by comparing results from complete-case analyses with MI-derived estimates to verify the robustness of our findings. A two-tailed p value < 0.05 indicated significance. IBM SPSS V26.0 (Armonk, NY, USA) and R 4.3.2 (R Foundation, https://www.r-project.org/) were used for the analyses.

Ethics

This study was approved by the Medical Science Research Ethics Committee of the First Hospital of China Medical University (No. 2019-253-3) and adhered to the CONSORT 2010 guidelines(Supplementary Table 5). The trial was registered before commencement (registration date: 2 September 2019; registration name: Real-time monitoring and just-in-time intervention for adherence to pre-exposure prophylaxis among men who have sex with men in China: a multicentre RCT study; registration number: ChiCTR1900025604; https://www.chictr.org.cn/showproj.html?proj=42628). The research protocol has been published41. Prior to the questionnaire survey, smart pillbox distribution, and specimen collection, all participants provided written informed consent and were informed of their right to withdraw without affecting their interests and privacy.