The study was designed as a 2-arm (intervention using the SPA vs wait-list control) parallel RCT. Participants were recruited between November 2020 and January 2021. All follow-up was completed by March 2021. An email for recruitment was sent to more than 3000 workers of 2 companies and 1 university in Japan. The institutional affiliations (Kyoto University; HealthTech Laboratory, Inc; and OKI Electric Industry Co, Ltd) were displayed in the emails as the organizations conducting the study. To avoid the risk of divulging information to competing research institutions, we selected companies and universities affiliated to our research group for recruitment of participants. However, we excluded workers of departments involved in the study. Both eligible employees and their acquaintances were allowed to participate in the study. Written informed consent was obtained through participation in face-to-face information sessions or online information sessions using a web conference system. After the information sessions, all subsequent communications were conducted via email.

The inclusion criteria were workers aged 20 years or older, ownership of a smartphone, and an affirmative answer to the question “Do you have any sleep problems?” that was asked during the application for participation in this study.

There were 6 exclusion criteria: (1) individuals whose version or model of smartphone did not support the SPA, (2) shift workers, (3) individuals likely to be subjected to significant time zone changes such as overseas travel during the study period, (4) those who had a total score of 10 or more on the Patient Health Questionnaire-9 at the time of application for participation in the study, (5) those who responded that the frequency of “thoughts that you would be better off dead or of hurting yourself in some way” was more than half of the days in the previous 2 weeks, and (6) those who were judged to be inappropriate for the study.

Random assignment into the SPA group or nonintervention control group was performed by stratified block allocation with an allocation ratio of 1:1, a block size of 4 using sex and 3 groups of total Insomnia Severity Index (ISI) scores obtained at the time of study entry: <8, 8-14, and ≥15. Central randomization with computer-generated tables was used for the allocation. One investigator (TS) prepared the list, and another investigator (KK) independently performed the allocation. Due to the nature of the intervention, the participants were not blinded to their allocation. The researchers included statisticians (KK), and data analysts (RF and YN) who were blinded to the assignment of the participants to their respective groups.

All online questionnaires were distributed using Google Forms. The email which included information of the questionnaire and the URL of the questionnaire were sent to all participants.

We administered the Japanese version of the ISI questionnaire [13,14]. This is a 7-item questionnaire with scores ranging from 0 (no insomnia) to 28 (severe insomnia). Individuals with scores ≥8 were defined as ISI-8-insomniacs [13].

The online questionnaire included questions regarding sociodemographic information (sex and age), height, weight, caffeine consumption (calculated per day as 95 mg per cup of coffee, 55 mg per cup of tea, and 45 mg per cup of cola), alcohol consumption (never, sometimes, or every day), smoking habits (yes or no), living with others or alone, hypnotic medication (more than once a week or not), and use of alcohol as an aid to sleep (more than once a week or not). BMI was calculated using height and weight, with 2 groups being created using a cutoff of 25: obese (BMI≥25 kg/m²) or normal.

Sample-size calculations were based on our pilot study (not published) using a prototype app; the intervention produced a 1.3-point reduction (SD 2.3) on the ISI. We needed 50 participants per intervention group (100 in total) to detect 80% power using a 2-sided α of .05. Accounting for 20% potential attrition, we estimated the need to enroll 120 participants.

Descriptive statistics were computed for participant characteristics. We computed the differences in clinical and demographic variables between the intervention and control groups by original assigned groups using the independent samples t test for continuous variables and chi-square tests for categorical variables. As a primary analysis, linear mixed-effects models were applied to longitudinal data to measure the effectiveness of the intervention, to account for repeated measurements over time, and to avoid the imputation of missing data. These models included the ISI scores of the 2 groups, 5 time points (baseline, and 1, 2, 3, and 4 weeks later), and the time × intervention interaction. As secondary measures, we assessed the changes in absolute presenteeism score, relative presenteeism score, CFS total 11, CFS mental 11, and CFS physical 11 with an independent samples t test. A subgroup analysis was conducted for ISI-8-insomniacs. No interim analysis was conducted. Statistical significance was defined as P<.05 using 2-tailed tests. All analyses were performed using SPSS (version 27.0, IBM Corp) for Windows (Microsoft Corp).

The study protocol was approved by the Ethics Committee of Kyoto University (#C1478). The study was conducted in accordance with the Declaration of Helsinki. After identification at the information session and provision of written informed consent forms under real names, a correspondence chart between IDs and real names was created, and the correspondence chart was kept in strict confidence at the research department of Kyoto University. All data collected from participants' responses to the web questionnaire and the app were managed in a manner that was linked to their IDs, thus guaranteeing participants' privacy.

In this RCT, we examined the effectiveness of the new SPA in providing brief personalized therapy for relatively mild insomnia among Japanese workers. The results showed that the SPA significantly improved the primary outcome of ISI scores in both the total population and the ISI-8-insomniacs. In addition, the SPA also improved physical fatigue as measured by the physical CFS in ISI-8-insomniacs. To our knowledge, this is the first study to demonstrate the effectiveness of SPA in treating insomnia among workers.

There were several methodological limitations to our study. First, we used a subjective rating scale. The use of an objective assessment of sleep, such as polysomnography, would strengthen the results of this study. Although some studies have reported that self-reported data on sleep status concur with physiological data [18,19], using objective data (ie, physiologic measurements such as polysomnography) is desirable in future studies. Second, the participants were not blinded. The participants were aware of their assignment to the intervention or control group, and this difference could have affected the outcome. Because only those in the intervention group used the app and maintained a record in their sleep diaries, the placebo effect might have affected their behavior and influenced the results. Third, the selection of companies and universities from which participants were recruited was not random, and therefore sampling bias may exist. Fourth, the long-term prognosis of the intervention was not investigated.

Previous studies demonstrated that internet-based CBT-I improved ISI scores. One RCT compared the effectiveness of web-based CBT-I with telehealth-based CBT for chronic insomnia and found that both web- and telehealth-based treatment improved the ISI score [20]. Another RCT showed that internet-delivered CBT-I improved the ISI scores for patients with insomnia or depression [9]. In our study, the SPA group showed a statistically significant decrease in ISI score, but the difference in ISI scores between the 2 groups was smaller than that reported in previous studies [10]. In a meta-analysis of 2 previous studies, the group receiving internet-based CBT showed a decrease in the ISI score of 2.28 points (–2.89 to –1.67) compared with the control group [10]. There are some possible explanations for this. First, the low ISI scores of the study participants at baseline could have affected the results. We found a large change in the ISI scores of ISI-8-insomniacs. The fact that we were able to show the efficacy of the SPA for mild insomnia in this study is important for the establishment of future preventive methods through early interventions for sleep disorders among workers. Second, in some of the previous studies, email or phone support was required in addition to the app [9,20]. However, considering the need for widespread implementation, it is important that the SPA is effective independently, without the need of additional supportive measures.

In this study, the dropout rate was 3.2% (n=2), and most participants were able to complete the final questionnaire. In many previous studies, about 20% to 40% of the participants dropped out [11]. The app used in this study sends prompt messages to the participants at a convenient time for them, which might have contributed to maintaining their motivation and reducing the number of dropouts.

In this study, a significant improvement in physical fatigue was observed among the ISI-8-insomniacs. Fatigue symptoms have been reported as one of the most frequent daytime complaints of patients with insomnia [21], which has a major impact on day-to-day functioning and quality of life [22]. Clinical studies on patients with chronic obstructive pulmonary disease and on cancer survivors have shown that CBT-I significantly improves fatigue and insomnia in patients with insomnia [23,24]. Although no previous study has examined the effect of internet-delivered CBT-I or BTI on recovery from fatigue in workers, our intervention with the SPA showed, for the first time, the possibility of recovery from physical fatigue.

This RCT showed the effectiveness of the SPA in providing brief personalized therapy for treating relatively mild insomnia among Japanese workers. The physical fatigue score was also significantly improved in ISI-8-insomniacs. The SPA could play an important role in reducing the adverse effects of sleep problems in communities. To promote the use of SPA in the future, further studies are warranted to examine its effectiveness in other age groups and individuals with health problems. To increase generalizability, future studies with follow-up surveys are required to examine the long-term prognosis of the intervention.