Worldwide, chronic diseases are the main cause of death and years lived with disability [1,2]. Cardiovascular disease (CVD) and type 2 diabetes mellitus (T2DM) are globally among the top 5 chronic conditions in terms of incidence and prevalence [2]. The behavioral risk factors for these conditions, such as smoking, harmful use of alcohol, poor diet, and physical inactivity, are highly correlated with the disease progression [3]. For example, Gellert et al [4] observed in their meta-analysis a dose-response relationship between the number of cigarettes smoked and premature death. They also found an inverse correlation between time since cessation and all-cause mortality. Wood et al [5] reported that all-cause mortality was positively associated with the level of alcohol intake, based on data from over half a million current drinkers. Chudasama et al [6] found a negative dose-response relationship between physical activity levels and all-cause mortality in their analysis of almost half a million people. Regarding low whole-grain intake, which is the highest risk factor related to poor diet, in their meta-analysis, Zhang et al [7] showed an inverse dose-response relationship between whole-grain intake and all-cause mortality. Hence, targeting these with preventive measures could significantly reduce people’s chronic disease risk [8], and behavior change interventions are well suited for preventing CVD and T2DM [2,3].

To stop noncommunicable diseases from rising further, the World Health Organization (WHO) developed the Global Action Plan 2013-2020 [8]. In this report, the WHO emphasized the importance of early screening and the implementation of preventive programs. Further, the WHO recommended the use of information and communication technologies, such as the internet and mobile phone technologies, to deliver health education and promotion programs. In 2019, the WHO released a guideline with recommendations on digital interventions for health system strengthening [9]. This report outlined how the implementation of technology could overcome current challenges in health care systems and help to achieve the goal of universal health coverage. Health apps have promising potential. Wilson [10] pointed out that digital health interventions have the advantage of being easily accessible and cost-effective. According to the Pew Research Center [11], many people use their smartphones daily. Riley et al [12] reported that new advancements allow apps to be tailored to personal needs and preferences, as well as the integration of dynamic feedback systems. Despite the promising potential of health apps, there is still ambiguity about their effectiveness, as outlined by the WHO guideline [9].

The aim of this systematic literature review was to assess the current evidence regarding the effectiveness of mobile health–based interventions in reducing the risk for CVD and T2DM. The focus was on multiple behavioral risk–factor interventions, rather than single risk–factor interventions, because of the lack of evidence on their combined effectiveness compared with substantial evidence on single risk–factor interventions [13,14].

We conducted this systematic review in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement [15] and registered it with International Prospective Register of Systematic Reviews (PROSPERO; registration number CRD42019135405).

We searched the following medical and bioengineering databases to retrieve all relevant articles regarding preventive mobile health intervention for CVD and T2DM: EMBASE (via Ovid), Scopus, ScienceDirect, CINAHL (via EBSCOhost), MEDLINE (via Ovid), ProQuest science and technology databases, and Ei Compendex and Inspec (both via Engineering Village 2). The search strategy (Multimedia Appendix 1) included terms relating to the 2 conditions under study and the intervention; we combined the terms using Boolean operators [16] and adapted the terms to the database-specific requirements. The search included articles published from the inception of the databases until March 25, 2019. We limited the search to English- and German-language publications because these languages were proficiently spoken by the review team. We excluded review articles, conference abstracts, comments, editorials, letters to the editor, and theses. Additionally, we identified studies using “snowballing” techniques by reviewing the reference lists of articles included in the initial search and searching for other publications by authors included in the initial search [17].

Relevant data (study objective, study design, study population, comparator, description of the intervention, duration of the intervention or follow-up, outcomes, main results, and methodology for the assessment of the study’s quality) were extracted by 1 researcher (VHB) using a standardized form in Excel 365 (Microsoft Corporation). This was reviewed by all the other researchers. We synthesized the main results of the included studies in a narrative manner focusing on the intervention delivery and reported outcomes. A meta-analysis was not possible due to the small number of identified studies and the heterogeneity in interventions and outcomes.

One researcher (VHB) assessed the risk of bias using the following assessment tools: for RCTs, the Cochrane Collaboration’s tool for assessing risk of bias [21]; and for non-RCTs, the Risk of Bias In Non-randomized Studies - of Intervention assessment tool [22].

In total, we identified 941 articles using the search strategy for CVD and 732 articles using the search strategy for T2DM. In the validation of the 10% random sample of all retrieved articles, there was a 100% agreement (after initial disagreements were resolved by a third investigator) with the selection conducted by the researcher who screened all articles. Finally, 3 CVD articles [23-25] and 6 T2DM articles [26-31] fulfilled the inclusion and exclusion criteria; we identified no additional articles through the snowballing technique (Figures 1 and 2). We excluded many articles for several of the exclusion criteria.

There were 3 CVD [23-25] and 6 T2DM studies [26-31]. Table 2 provides details about the CVD studies and Table 3 provides details about the T2DM studies. For each study, the table includes the first author, year of publication, study design and duration, objectives, study population, interventions and comparators, outcomes, and the main results.

All RCTs used acceptable methods for randomization [25,27-31], but in none of the studies were the participants blinded to the design, which is an inherent problem with this type of intervention. Figure 3 summarizes the risk of bias for the RCTs. Of the 6 studies, 3 ensured blinding of the study personnel [25,27,28] and 3 ensured the blinding of the outcome assessors [25,28,31]. Apart from the study by Fukuoka et al [30], all RCTs published study protocols on the ClinicalTrials.gov database. Overall, due to performance bias, all studies were at high risk of bias.

Of the 3 non-RCTs [23,24,26], the study by Muntaner-Mas et al [24] was at moderate risk of bias, the study by Gore et al [23] was at high risk of bias, and the study by Arens et al [26] was at critical risk of bias. Figure 4 summarizes the risk of bias for the non-RCTs. The biggest issue with the study by Arens et al [26] was that missing data were not handled adequately, putting the study at critical risk of bias. We assessed the study by Gore et al [23] to be at high risk of confounding because some of the measurements that were used to control for confounding were based on nonvalidated questionnaires.

We identified only a small number (n=9) of articles that fulfilled the preset inclusion and exclusion criteria. We assessed most of the studies to be at high risk of bias. Additionally, 3 studies were underpowered (sample size <100), and 2 studies had short follow-up times (<6 months). Ideally, to show the effectiveness in reducing the risk of CVD or T2DM, the studies should have reported disease incidence rates. The only study that did this was that by Ramachandran et al [31], with their primary outcome being a decrease of T2DM incidence due to the SMS text messaging intervention over 2 years. Block et al [28] reported the percentage of people with metabolic syndrome as defined by the International Diabetes Federation Task Force on Epidemiology and Prevention [32]. Block et al [28] also measured change in the Framingham 8-year diabetes risk score [33], and Gore et al [23] measured change in the Framingham 10-year CVD risk score [20]. All other studies reported single risk factors rather than multivariable absolute risk of disease. None of the identified studies directly targeted tobacco smoking cessation or responsible alcohol intake. Rubinstein et al [25] mentioned in their article that their original protocol included both lifestyle factors, but they were later excluded. According to the authors, alcohol intake was considered a sensitive matter requiring face-to-face interactions, while tobacco smoking was excluded because supposedly, compared with physical activity and diet, it had less effect on the onset of hypertension and was more difficult to target via a mobile health intervention [25]. Overall, there were some positive findings suggesting that mobile health-based interventions can achieve at least small to moderate reductions in CVD and T2DM risk, although these were based on weak evidence.

The strength of this literature review was that it followed the PRISMA statement. We systematically searched several databases to identify all relevant published articles. Further, we conducted a manual search through the snowballing technique. For the title and abstract screening, a 10% random sample of all retrieved articles was validated by a second researcher, and 2 reviewers independently performed the full article selection. However, only 1 researcher conducted the database search, the data extraction, and the risk-of-bias assessment. Although we a priori restricted the search to English- and German-language articles, we did not exclude any articles because they were not available in these 2 languages. We did not perform a meta-analysis due to the small number of publications that met the inclusion criteria and the differences in their interventions and outcome measures.

Previous mobile health research has focused more on self-management of chronic diseases than on prevention. In their systematic review and meta-analysis, Wu et al [34] investigated the effectiveness of mobile phone apps for diabetes self-management (including prediabetes, gestational diabetes, type 1 diabetes, and T2DM). They identified 3 studies that targeted prediabetes, 2 of which we also included in this review. The overall conclusion of Wu et al [34] was that there was evidence for the effectiveness of app interventions in T2DM self-management, but not for prediabetes. Lunde et al [35] conducted a systematic review looking at various types of noncommunicable diseases and lifestyle advice. Most of the identified studies (8 out of 9) targeted T2DM patients for whom the authors measured improvements in lifestyle factors, particularly reduced glycated hemoglobin levels (in 5 of the 8 studies). For CVD patients, Lunde et al [35] found only 2 relevant articles, and these were without statistically significant improvements in any of the outcomes of interest (weight, BMI, waist circumference, physical activity, and quality of life). A systematic review by Coorey et al [36] focused on self-management of CVD via mobile apps, in which they concluded that short-term improvements in behavior and risk factors were possible but there was insufficient evidence for long-term effects. Alessa et al [37] reported from their systematic review of 21 studies that mobile apps could reduce blood pressure, although the evidence originated mainly from studies that had a high risk of bias.

Palmer et al [14] conducted a systematic review of noncommunicable disease prevention through smoking cessation, alcohol reduction, physical activity, and diet using mobile technology. In total, they found 71 articles, but only 2 of the studies were aimed at the combination of physical activity, diet, and smoking cessation, with both studies targeting secondary prevention of CVD. Among the studies they reviewed, 8 RCTs focused on alcohol reduction but did not include any other lifestyle advice, with these studies specifically targeting heavy drinkers [14]. In general, it appears that many interventions are designed to provide advice for 1 or 2 behavioral risk factors that are associated with increased chronic disease risk, whereas there were only a few evaluation studies of comprehensive mobile health interventions addressing the 4 common behavioral risk factors (ie, tobacco smoking, excessive alcohol consumption, physical inactivity, and poor diet) [14]. Noble et al [13] stated in their systematic review that there were clustering patterns between the 4 behavioral risk factors—tobacco smoking, excessive alcohol consumption, physical inactivity, and poor diet—which indicated similar or the same reasons causing these behaviors. Hence, the authors suggested that future interventions should apply a holistic approach instead of targeting single risk factors. Similarly, Geller et al [38] called for future research studies to focus on improved lifestyles, meaning a change in multiple health behaviors rather than 1, even if it might be harder to achieve. Meader et al [39] reported in their systematic review that targeting smoking simultaneously with other behaviors resulted in negative outcomes for diet and physical activity, suggesting that it might be more beneficial to apply a sequential approach. In a Cochrane review published in 2016, Whittaker et al [40] stated that studies have demonstrated that mobile phone-based interventions can be effective in achieving smoking cessation over 6 months, particularly SMS text messaging in high-income countries.

Most studies that were conducted according to the review’s inclusion criteria were at high risk of bias. This review only considered studies of multirisk factor interventions, which resulted in only 9 studies being included. There is a lack of research evaluating interventions that address the 4 common behavioral risk factors (ie, tobacco smoking, excessive alcohol consumption, physical inactivity, and poor diet) in a single mobile health intervention. Researchers may have preferred to focus on 1 risk factor at a time due to simplicity for participants and clarity of intervention-outcome relationships. Hence, future studies should further explore the use of mobile technology for primary disease prevention, by applying a rigorous study design.

According to the findings of this systematic review, evidence for the effectiveness of mobile health-based interventions in reducing the risk of CVD and T2DM is scarce due to the quality of the included studies and the small effects that were measured. This highlights the need for further high-quality research to investigate the potential of mobile health interventions.