The effectiveness of smoking cessation, physical activity/diet and alcohol reduction interventions delivered by mobile phones for the prevention of non-communicable diseases: A systematic review of randomised controlled trials

Background We conducted a systematic review to assess the effectiveness of smoking cessation, physical activity (PA), diet, and alcohol reduction interventions delivered by mobile technology to prevent non-communicable diseases (NCDs). Methods We searched for randomised controlled trials (RCTs) of mobile-based NCD prevention interventions using MEDLINE, EMBASE, Global Health, CINAHL (Jan 1990–Jan 2016). Two authors extracted data. Findings 71 trials were included: smoking cessation (n = 18); PA (n = 15), diet (n = 3), PA and diet (n = 25); PA, diet, and smoking cessation (n = 2); and harmful alcohol consumption (n = 8). 4 trials had low risk of bias. The effect of SMS-based smoking cessation support on biochemically verified continuous abstinence was pooled relative risk [RR] 2.19 [95% CI 1.80–2.68], I2 = 0%) and on verified 7 day point prevalence of smoking cessation was pooled RR 1.51 [95% CI 1.06–2.15], I2 = 0%, with no reported adverse events. There was no difference in peak oxygen intake at 3 months in a trial of an SMS-based PA intervention. The effect of SMS-based diet and PA interventions on: incidence of diabetes was pooled RR 0.67 [95% CI 0.49, 0.90], I2 = 0.0%; end-point weight was pooled MD -0.99Kg [95% CI -3.63, 1.64] I2 = 29.4%; % change in weight was pooled MD -3.1 [95%CI -4.86- -1.3] I2 0.3%; and on triglyceride levels was pooled MD -0.19 mmol/L [95% CI -0.29, -0.08], I2 = 0.0%. The results of other pooled analyses of the effect of SMS-based diet and PA interventions were heterogenous (I2 59–90%). The effects of alcohol reduction interventions were inconclusive. Conclusions Smoking cessation support delivered by SMS increases quitting rates. Trials of PA interventions reporting outcomes ≥3 months showed no benefits. There were at best modest benefits of diet and PA interventions. The effects of the most promising SMS-based smoking, diet and PA interventions on morbidity and mortality in high-risk groups should be established in adequately powered RCTs.


Methods
We searched for randomised controlled trials (RCTs) of mobile-based NCD prevention interventions using MEDLINE, EMBASE, Global Health, CINAHL (Jan 1990-Jan 2016). Two authors extracted data.

Introduction
The World Health Organization (WHO) estimates that 38 million deaths occur each year due to non-communicable diseases (NCDs)-principally cardiovascular diseases, cancer and chronic respiratory diseases. Approximately 42% of NCD deaths are premature (i.e. occur before the age of 70 years) [1]. Although the number of NCD deaths has increased in every world region since 2000, the burden is greatest among people of low socio-economic status. Nearly three-quarters of NCD deaths occur in low and middle income countries [1]. These inequalities also exist within countries, with higher NCD mortality among people with lower education, income, or social class [2]. Physical inactivity, unhealthy diet, tobacco use and the harmful use of alcohol all increase the risk of developing and dying from NCDs. The Global Burden of Disease Study estimated that in 2010, 12.5 million deaths were attributable to dietary risk factors and physical inactivity, over 6 million deaths were attributable tobacco smoking (including second hand smoke), and over 2.5 million deaths were attributable to alcohol use [3]. Encouraging health care consumers to adopt healthy behaviours can prevent the onset or progression of NCDs and reduce mortality [4,5].
Recent systematic reviews have concluded that there are benefits of interventions delivered by mobile phone targeting smoking cessation, physical activity and diet [6][7][8][9][10][11][12][13]. However, the meta-analyses reported in existing reviews include self-reported outcomes [6][7][8][9][10]13]. Self reported outcomes in trials of behaviour change interventions where participants are not blind to allocation can be prone to bias and overstated benefits [14]. Some reviews of diet and physical activity interventions included non randomised studies, which are prone to bias [9][10][11][12]. In some reviews the effects of interventions delivered partly by mobile phone have been pooled with those delivered wholly by mobile phone, making it impossible to judge the effects of the mobile phone based components [10][11][12]. Our previous systematic reviews of interventions delivered by mobile phone relied on objective measures of outcomes reported in randomised trials, but the searches for this review were completed in September 2010 [15,16].
We aimed to provide an updated review of the evidence base for interventions delivered by mobile phone for the prevention of non-communicable disease.

Methods
This review includes eligible trials identified in the previous comprehensive systematic review that included studies published between 1990 and September 2010 and in further searches conducted to identify studies meeting the inclusion criteria that were published between September 2010 and Jan 2016 in MEDLINE, EMBASE, Global Health and CINAHL [15,16]. The search strategy for MEDLINE is provided in the Supporting Information (S1 Text). The search terms were adapted for use with other bibliographic databases in combination with databasespecific filters for randomised controlled trials, where these were available. Two reviewers independently scanned the electronic records to identify potentially eligible trials. We followed the protocol provided in S2 Text, however, we did not include trials targeting disease management due to resource limitations.
Participants were men and women of any age. We included all controlled trials employing any mobile phone interventions (mobile phones; PDA phones [e.g., BlackBerry, Palm Pilot]; Smartphone [e.g., iphone]) targeting behavioural risk factors for non-communicable diseases, i.e. tobacco use, harmful alcohol use, physical inactivity, and unhealthy diets.
We included studies in which the intervention delivered by mobile phone was the primary intervention component under evaluation. We excluded studies evaluating either mixed mobile phone technology and non-mobile phone technology-based interventions in which the treatment and control group both received the mobile phone technology-based component, or interventions in which treatments between the treatment and control groups differed in additional ways besides the components delivered by mobile phone, such as interventions involving face-to-face counselling with a text message intervention compared to a control group receiving information only. We excluded interventions which can be but do not need to be delivered by mobile phone such as websites, social media and email, except where these were provided in addition to an intervention delivered primarily through mobile phone technology. Interventions employing devices that linked to the mobile phones (e.g. phone-linked activity trackers) were included as these were considered an extension to the mobile phone technology.
For the purpose of this review, primary outcomes were defined as any objective measure of outcomes related to the specified NCD behavioural risk factors, including objective measures of the behaviour and the distal biometric or health effects of the behaviour. For example, objective measures of the behaviour would include salivary cotinine levels for smoking cessation, and step counts for physical activity; biometric measures of effect would include blood pressure, weight, and VO2 max (e.g. for fitness); and health effects would include incidence of diabetes or cardiovascular disease. Secondary outcomes were defined as self-reported measures relating to NCD-related health behaviours, health status, and cognitive outcomes. Studies reporting outcomes for any length of follow-up were included.
Two reviewers carried out the data extraction-this involved each reviewer extracting data independently from half of the studies, and then checking each other's data extraction against the original papers. The following data was extracted from eligible studies: number of randomised participants, intervention, intervention components, user involvement in intervention development, mobile devices employed, mobile technology functions used, sequence generation, allocation concealment, blinding of outcome assessors, completeness of follow-up, evidence of selective outcome reporting, any other potential sources of bias, and measures of effect using a standardised data extraction form. Where outcomes were measured at multiple time points, we extracted data for the final point of measurement. The authors were not blinded to authorship, journal of publication, or the trial results. All discrepancies were agreed through discussion, and involved a third reviewer when necessary. All analyses were conducted in STATA v 14. We calculated risk ratios and mean differences. We used random effects meta-analysis to give pooled estimates where there were two or more trials using the same mobile technology media (e.g. SMS messages) and targeting the same behaviour (e.g. physical activity) and reporting the same primary outcome. We examined heterogeneity visually by examining the forest plots and statistically using both the χ2 test and the I2 statistic.
The behaviour change techniques used in behaviour change interventions were classified according to Abraham and Michie's taxonomy of behaviour change techniques [17]. Risk of bias of each included study was assessed independently by two study authors according to the criteria outlined by the International Cochrane Collaboration [18]. Disagreements were resolved through discussion, and with input from a third author where necessary. We used a cut off of 90% complete follow-up for low risk of bias for completeness of follow-up. We applied the GRADE criteria [19] to assess the quality for evidence for all outcomes pooled in our meta-analyses.

Results
The combined search strategies identified 42,268 electronic records which were screened for eligibility (Fig 1). The full texts of 723 potentially eligible reports were obtained for further assessment. Out of the 723 potentially eligible reports, 72 met the study inclusion criteria and were trials delivered to health care consumers to improve health behaviours. Two papers report on the same trial involving an intervention targeting smoking cessation and an attention-matched control receiving messages promoting improved diet and physical activity. Ybarra (2013) reports on the smoking outcomes and Filion (2015) reports on the physical activity/diet outcomes. Therefore, in total, there were 71 unique trials. 18 interventions aimed to increase smoking cessation; 44 aimed to increase physical activity, improve diet, or a combination of both; 2 aimed to increase physical activity, improve diet, and increase smoking cessation; and 8 aimed to reduce harmful alcohol consumption.

Characteristics of studies
Smoking cessation. There were 18 randomised controlled trials with parallel groups which aimed to increase smoking cessation (Table 1). The smoking cessation trials included a total of 17857 participants, with sample sizes ranging from 31 to 5800. Twelve of the smoking cessation trials were delivered by SMS, three were delivered by voice calls, one by interactive voice response, one by a combination of SMS and video messages, and one by a mobile application combined with voice calls.
With the exception of one study carried out in Turkey [20], all trials were conducted in high-income countries.
Physical activity. There were 15 randomised controlled trials which aimed to increase physical activity ( Table 2). The physical activity trials included a total of 1416 participants, with sample sizes ranging from 36 to 174. In four trials the intervention was delivered though a mobile application, seven trials were delivered by SMS, one through SMS and voice calls, one via fitbit with linked smartphone app, one via SMS with a linked pedometer, and one through audio files uploaded to mobile phone. All physical activity trials were conducted in high income countries.
Diet. We identified three trials which aimed to improve diet (Table 3), including a total of 906 participants with sample sizes ranging from 41 to 808. In one trial the intervention was delivered through a mobile application, in one trial through SMS with a web-based tool, and one was delivered through personalised emails sent to smartphones based on participants' salt intake. All three trials focusing on diet were carried out in high income countries.
Physical activity and diet. There were 26 trials which targeted both physical activity and diet (Table 4). These trials included a total of 4092 participants, with sample sizes ranging from 24 to 502. Eleven trials tested interventions delivered via SMS, six trials were of interventions delivered through mobile applications, eight involved interventions delivered by a  group also received studyrelated reminder texts via SMS, particularly in the 2 weeks prior to each followup survey.

Borland 2013 [22]
Parallel group RCT Country: Australia; Device: Mobile telephone; Media: SMS 3530 adult smokers and recent quitters (quit within past 2 weeks) aged ! 18 y. Mean age 42.1 y. Females 60% Participants recruited from callers to smoking quitlines (n = 1335), and from an internet panel survey (n = 2195) composed of a mixture of known smokers and those of unknown smoking status. Participants recruited from quitlines were more highly motivated to quit.
To test the population impact of offering automated smoking cessation interventions via the internet and/or by mobile telephone. Four intervention groups: (1) received a personalised, automated tailored cessation program based on cognitive-behavioural principles that generates 2-4 page letters of advice with suggestions about strategy; (2) received a stream of SMS that mix snippets of advice on strategy and motivation; (3) participants could use either or both parts of interventions 1 and 2 which were offered as a package; (4) participants explicitly offered either or both interventions 1 and 2, with the person encouraged to make an upfront choice although they could subsequently change their minds, and take up whatever aspects they wanted. Duration: 7 months User involvement in development: not stated.
Participants received brief information on web-and telephone-based assistance available in Australia.

(Continued)
Smoking cessation, physical activity/diet and alcohol reduction interventions delivered by mobile phones To assess feasibility of a SMS-based smoking cessation intervention for pregnant smokers, to determine the acceptability of a SMS-based SGR plus support messages intervention, and to obtain preliminary efficacy data for SMS-based SGR to promote cessation during pregnancy. Participants received up to 5 SMS per day for 5 wks.
Participants received "alert messages" at scheduled times that instructed participants to smoke and were designed to help them gradually cut down to 0 cigarettes by the end of the fourth week. If participants smoked at an unscheduled time it was requested that they message back "s" so the next alert would correspond with the scheduled interval. If participants messaged 3 times "off schedule," the alert messages were stopped and resumed the next day. If this pattern continued for more than a day, the participant was called and asked if the schedule needed adjustment. Participants were asked to sign a contract stating they will not smoke more cigarettes during the study than they did before they entered the study. Participants were informed that if they responded to 80% of the alert texts, they would be entered into a $25 gift card raffle. Duration: 6 wk. User involvement in development: not stated prior to this. However, this was a pilot study which was also collecting feedback on intervention from users.
Participants were asked to choose a quit date within 2-3 wks post randomisation and then sent an SMS stating that quit date. Participants received up to 5 SMS per day for 5 wks. Each week, there was a new "theme" for a subset of the messages, such as reasons for quitting, getting ready for the quit date, partner smoking, and handling slips.

(Continued)
Smoking cessation, physical activity/diet and alcohol reduction interventions delivered by mobile phones To determine whether a video-based smoking cessation intervention delivered via mobile telephone was effective at increasing smoking cessation rates. Participants nominated a quit day between 1 and 3 wks from randomisation and 2 time periods during which they wished to receive SMS. Participants were directed to an online brief description and photograph of the 6 role models and asked to select a role model from whom they would receive messages. The video messages were filmed as video diaries during a quit attempt, with the role models discussing issues they had found difficult and the techniques and coping strategies they used to remain smoke-free. Participants received a pedometer and a walking instructional manual.

(Continued)
Smoking cessation, physical activity/diet and alcohol reduction interventions delivered by mobile phones (but were not asked to form any).

(Continued)
Smoking cessation, physical activity/diet and alcohol reduction interventions delivered by mobile phones and Actigraph GT3x+ concurrently for 7 days to access baseline PA levels. Valid day: 600 min/day. They also received 3 automated text messages a day for 6 weeks and were asked to upload their data everyday throughout the study. Duration 6 wks. User involvement in development: not stated.
Control group were also asked to continue to wear the Fitbit One tracker and upload data every day for the duration of the study. https://doi.org/10.1371/journal.pone.0189801.t002 Smoking cessation, physical activity/diet and alcohol reduction interventions delivered by mobile phones To explore the combined impact of a Web-based, fully automated planning tool and SMS reminders on intention to change saturated fat intake, self-reported saturated fat intake, and portion size changes over 4 wk. 2 intervention groups: (1) Participants used a web-based planning tool to select from a list of 13 situations in which they were tempted to eat unhealthily and then choose an approach to change their behaviour from a list of 13 solutions. Participants were asked to complete at least 3 situation-solution pairs. Once these pairs were chosen and saved, participants were not able to revisit the program to change them during the 4-week period. ( To assess the use of a popular smartphone app for dietary self-monitoring and weight loss compared to traditional diet counselling and entry methods. Participants randomized into 3 groups: the app group (AP), trained to use the diet-tracking ''Lose It!" app daily; the memo group (ME), trained to track dietary intake daily through use of the memo function on their smartphones; and the paper group (PA), trained to record dietary intake daily using a traditional paper-andpencil method.
The AP group recorded dietary intake using the ''Lose It!" app interface, which provided a large database of commonly consumed foods for users to search and add to a diary at each eating occasion. It also provided immediate feedback in the form of a daily calorie gauge graphic that increased in real-time as foods were entered. The ''Lose It!" app calculated the daily energy allotment for the user based on a pre-identified weight loss goal (1 lb/wk) and individual anthropometric data. No dietary advice was provided to the AP group; however, these participants received immediate feedback regarding calorie intake when dietary data were entered into the ''Lose It!" app. Duration 8 wks User involvement in development: not statedcommercially available app.
The paper group (PA), trained to record dietary intake daily using a traditional paper-andpencil method. https://doi.org/10.1371/journal.pone.0189801.t003 Smoking cessation, physical activity/diet and alcohol reduction interventions delivered by mobile phones To test the acceptability and feasibility (recruitment, dropout, and adherence) of weight loss application software for mobile phones with a view to informing a larger trial. Participants were given a HTC Desire mobile phone with the weight loss application software pre-downloaded. Participants received standardized training in the equipment. Participants were instructed to use the study equipment every day for a week and then to use it as much as they desired over the trial period. Participants were given access to an Internet forum for social support. Duration: 6 months User involvement in development: not stated.
Two control groups: (1) Participants were given a paper food diary, a calorie-counting book, and a calculator; (2) Participants were given a voucher providing 6 months access to the Weight Loss Resources website. Participants in Control groups 1 and 2 were instructed to use the study equipment every day for a week and then to use it as much as they desired over the trial period and were given access to an Internet forum for social support. (Continued) Smoking cessation, physical activity/diet and alcohol reduction interventions delivered by mobile phones (Continued) Smoking cessation, physical activity/diet and alcohol reduction interventions delivered by mobile phones Smoking cessation, physical activity/diet and alcohol reduction interventions delivered by mobile phones Research assistants told control group patients to "choose any activities you'd like to lose weight," without specifying any particular interventions. Control group participants were aware that they were participating in a study of a weight-loss app but were blinded to the name of the app. To minimize contamination of the control group, providers and clinic staff were also blinded to the name of the app and to group assignment. At the 3-month follow-up visit, all participants received a 1-page educational handout on healthy eating from www.myplate.gov.

(Continued)
Smoking cessation, physical activity/diet and alcohol reduction interventions delivered by mobile phones Both groups were given a glucometer and asked to monitor their glucose levels. The control group was examined initially and after 4 months physically in the clinic and there were no mobile phone contacts with these patients.
(Continued ) To determine the effect on weight of mobile technologybased (mHealth) behavioural weight loss interventions in young adults. The intervention was delivered through an investigatordesigned smartphone app which included goal setting, challenge games, and social support through a "buddy system" that allowed exchange of pre-determined messages to a randomly assigned buddy participant. Self-management behaviours for CP were regularly and frequently prompted by the app according to a protocoldriven schedule; participants did not have a choice in the timing or frequency of prompts. Tailoring within the CP intervention occurred mainly via setting personal goals. To examine whether a combination of podcasting, mobile support communication, and mobile diet monitoring can assist people in weight loss. Participants received the same Podcasts as in the comparison group. Participants were also instructed to download a diet and physical activity monitoring application and a social networking site's (Twitter) application to their mobile device. Participants created a user account on Twitter, were told to log on to Twitter at least once daily to read messages posted from the study coordinator, and were encouraged themselves to post at least daily to Twitter. Participants could choose any user name they wanted (to protect their identity) and were instructed on how to make their Twitter account private (if they chose to do so). During months 0-3, participants were divided into 4 groups to create Twitter cohorts of 11-12 people. They were sent a list of everyone's Twitter user names within their cohort, were instructed to follow everyone in their cohort, and were reminded to send follow requests to participants and to accept requests until everyone in each cohort was following one another. During months 3-6, participants were asked to follow everyone in the study, and similar procedures were used to allow everyone within the group to follow one another. The study coordinator sent out 2 messages per day to the group, which reinforced messages from the podcasts, posed questions to the group to facilitate discussion, and encouraged participants to share tips and recipes with one another that would assist in weight loss. Such messages were prompts to attend to weight-loss behaviour, and encouraged communication but were not individualized.

(Continued )
Smoking cessation, physical activity/diet and alcohol reduction interventions delivered by mobile phones combination of media such as SMS, MMS, mobile apps, podcasts, and/or voice calls, and one trial was of an intervention delivered through voice calls alone. Twenty-four of physical activity and diet trials were conducted in high income countries. The other two trials were carried out in India [55] and Pakistan [56]. Physical activity, diet, and smoking. There were 2 trials [82,83] which targeted physical activity, diet, and smoking cessation as part broad lifestyle interventions (Table 5). One, conducted in Australia, included 710 respondents and trialled an intervention delivered by SMS [82], and the other, conducted in Iran, involved 180 participants and assessed the effectiveness of an app-based intervention [83].
Alcohol. We identified eight randomised controlled trials which aimed to reduce alcohol intake ( Table 6). The alcohol reduction trials included a total of 4782 participants with sample sizes ranging from 18 to 1929. In five trials the intervention was delivered by SMS, in two trials the intervention was delivered by mobile phone application and in one trial the intervention was delivered through interactive voice response.
All of the alcohol reduction trials were conducted in high income countries. Behaviour change techniques. According to our behaviour change technique coding of the studies (Table 7), smoking cessation studies included between 1 and 13 BCTs (median: 8), physical activity/diet studies included between 0 and 9 BCTs (median: 5), the two combined physical activity, diet and smoking trials included 1 and 11 BCTs, and alcohol studies included between 5 and 13 BCTs (median: 8).

Outcomes
Smoking cessation. The smoking cessation trials reported up to fifteen outcomes. Primary outcomes were as follows: eight trials reported at least one biochemically verified measure of abstinence from smoking (measured using salivary-cotinine testing and/or exhaled carbon monoxide testing)-outcomes included continuous abstinence, 7-day point prevalence of smoking cessation, and 24-hour point prevalence of smoking cessation. Secondary outcomes included self-reported quitting, self-reported intentions and attempts to quit, selfreported use of nicotine replacement therapy, and cognitive and mediator outcomes such as depression and self-efficacy.
Physical activity. The physical activity trials reported up to fifteen outcomes. Nine studies reported objectively measured physical activity outcomes (e.g. step counts, activity time) using pedometers and/or smartphone integrated accelerometers; five reported objectively measured anthropometric outcomes such as BMI, body weight, and waist-to-hip ratio; and four studies reported objectively measured medical outcomes such as systolic blood pressure, diastolic blood pressure, heart rate, peak oxygen uptake, blood sugar control, insulin resistance, high density lipoprotein cholesterol, low density lipoprotein cholesterol, total cholesterol, and triglycerides. Secondary outcomes included self-reported measures of physical activity, body weight, and cognitive and mediator outcomes including quality of life, depression, anxiety, satisfaction, and self-efficacy.
Diet. The diet trials reported on up to five outcomes. Two trials reported on anthropometric outcomes such as BMI, waist circumference and body weight; one reported on medical outcomes specifically systolic blood pressure and diastolic blood pressure. Secondary outcomes included self-reported dietary behaviour.
Physical activity and diet. The trials targeting physical activity and diet reported up to twenty outcomes. Nineteen studies reported on anthropometric outcomes including body weight, waist circumference, hip circumference, BMI, body fat percentage. Eight studies included measure of medical outcomes including systolic blood pressure, diastolic blood Smoking cessation, physical activity/diet and alcohol reduction interventions delivered by mobile phones any gender differences for these 2 applications in terms alcohol outcomes. Participants were emailed a link to the smartphone app and instructed to use the app for the following 8 weeks.
One group received the Promillekoll app which predicts real-time estimated blood alcohol concentration (eBAC) calculation and offers a number of strategies to maintain alcohol consumption at a level that is not harmful. The second group were given access to the PartyPlanner app, which allowed people to simulate a drinking event beforehand and comparing the simulation to real-time events. Duration: 7 wks. User involvement in development: not stated. One app was commercially available, one app was developed by research team.
The control group did not receive access to any intervention or feedback on risky drinking.
(Continued ) to report their weekend drinking plans. If they reported anticipating a heavy drinking day, they were then asked whether they were willing to set a low-risk drinking goal (<5 drinks per occasion for men or <4 drinks per occasion for women). Depending on the response to each query, participants were provided with real-time text feedback to either strengthen their low-risk drinking plan or goal, or to promote reflection on their drinking plan or decision not to set a low-risk goal. Then, on Sunday, participants were sent a text asking them to report the most drinks they had during a single occasion during the weekend. Depending on their response, they were provided with text feedback to either support their lowrisk drinking behaviour or promote reflection on their binge-drinking behaviour. Participants in the SA group did not receive any preweekend text message assessments but received identical text drinking assessments each Sunday for 12 weeks without receiving any alcohol-related feedback. Duration:  pressure, high density lipoprotein cholesterol, low density lipoprotein cholesterol, total cholesterol, triglycerides, heart rate, incident diabetes, HbA1C, and glucose levels. Secondary outcomes included self-reported measures of physical activity, diet, body weight, and cognitive and mediator outcomes including quality of life, knowledge, depression, self-esteem, stress, anxiety, and self-efficacy. Physical activity, diet and smoking. The physical activity, diet and smoking trials reported up to ten outcomes. Primary outcomes included medical outcomes such as systolic blood pressure, diastolic blood pressure, resting heart rate, high density lipoprotein cholesterol, low density lipoprotein cholesterol, total cholesterol, and triglycerides and biochemically confirmed smoking status. Secondary outcomes were self-report measures of physical activity and dietary intake.
Alcohol. The alcohol reduction trials reported up to ten outcomes-all of which relied participants to self-report their alcohol consumption.

Study quality
Smoking cessation. The assessment of risk of bias for the smoking cessation trials is reported in S1 Table and the risk of bias summary is presented in Fig 2. Two trials targeting smoking cessation were at low risk of bias for all quality criteria [25,26].
Physical activity and diet. The assessment of risk of bias of the physical activity/diet trials is reported in S1 Table and the risk of bias summary is presented in Figs 3, 4 and 5 (for physical activity only, diet only, and physical activity and diet trials, respectively). Of 26 the trials targeting both physical activity and diet, one was judged to be at low risk of bias for all quality criteria [92]. None of the trials targeting only physical activity or only diet were considered to be at low risk of bias across all quality criteria.
Physical activity, diet and smoking. The assessment of risk of bias of the physical activity, diet and smoking trials is reported in S1 Table and the risk of bias summary is presented in Fig  6. Of the two trials, one was assessed as being at low risk of bias across all quality criteria [82]. Smoking cessation, physical activity/diet and alcohol reduction interventions delivered by mobile phones Alcohol. The assessment of risk of bias of the physical activity/diet trials is reported in S1 Table and the risk of bias summary is presented in Fig 7. None for the alcohol trials were assessed as being at low risk of bias for all quality criteria.

Effects
Smoking cessation-primary outcomes. Interventions delivered by SMS alone. SMSbased smoking cessation interventions providing support for a quit attempt more than  Smoking cessation, physical activity/diet and alcohol reduction interventions delivered by mobile phones doubled biochemically verified continuous smoking abstinence when measured between three and six months [20,25] 8). There was no evidence of between-study heterogeneity (I 2 = 0%). Pooled analysis showed smoking cessation interventions providing support for a quit attempt significantly increased biochemically verified 7 day point prevalence of smoking cessation [20,21,26,32] (pooled effect estimate RR 1.51 [95% CI 1.06-2.15]), with no evidence of between-study heterogeneity, when measured between three and six months (Fig 9). There was no evidence that SMS-based  Smoking cessation, physical activity/diet and alcohol reduction interventions delivered by mobile phones smoking interventions increased adverse events (car accident in which respondent was the driver pooled RR 1.01 [95% CI 0.71, 1.42], I 2 = 0.0%; thumb strain pooled RR 1.02 [95% CI 0.83, 1.25], I 2 = 33.5%) [25,26,32].
One trial of an SMS-based intervention promoting smoking cessation showed a statistically significant improvement in biochemically verified smoking cessation at 6 months (time frame of smoking abstinence not defined) [82] (Table 8).
Interventions delivered by other or mixed mobile technology media. A trial of mobile phone-based counselling sessions demonstrated a statistically significant improvement in biochemically verified 24-hour point prevalence abstinence at 3 months, but not for 7-day point prevalence at 3 months [35] (Table 8).
Smoking cessation-secondary outcomes. Fourteen of the 63 self-reported smoking outcomes showed statistically significant benefits and none showed statistically significant harms. There were eleven studies reporting outcomes relating to cognitive mediators of smoking behaviour change-none showed statistically significant benefits or harms. One study compared a smoking cessation intervention delivered by mobile application with one delivered by text message and found the text messaging arm to have statistically significantly higher selfreported quitting rates, although those in the application arm were more likely to have set a quit smoking date [23] (S2 Table).
Physical activity-primary outcomes. Interventions delivered by SMS alone: physical activity outcomes. Pooled analysis of three trials examining the effect of SMS-based interventions on physical activity showed a borderline statistically significant increase in objectively measured physical activity [44,45,51] (change in steps per day pooled MD 1256.9 [95% CI -159.7 to 2673.6, p-value = 0.081], with evidence of substantial between study heterogeneity (I 2 = 77.8%) (Fig 10). The outcomes were measured between 4 and 12 weeks. These trials reported 4 other objectively measured physical activity outcomes. Of these, 2 were not in a positive direction and 2 showed statistically significant benefits [44] (Table 9). One additional trial of an SMS-based intervention demonstrated a statistically significant effect on end-line number of steps per day measured at 6 weeks (MD 1750.8 [95% CI 157.4 to 3344.2]) [41]. Smoking cessation, physical activity/diet and alcohol reduction interventions delivered by mobile phones Interventions delivered by SMS alone: Anthropometric and bio-medical outcomes. Two trials of SMS interventions targeting physical activity reported 5 anthropometric outcome measures (waist-to-hip ratio, body weight, BMI), none of which showed benefits [45] [49]. Two of the physical activity trials included 5 bio-medical outcomes (such as blood pressure, blood sugar control, post-exercise breathlessness)-none of which demonstrated a benefit [45,92] ( Table 9).
Interventions delivered by other or mixed mobile technology media: physical activity outcomes. Two trials of mobile-application interventions reported 17 objectively measured physical activity outcomes-of these, one showed statistically significant improvement in change in number of steps per day measured at 8 weeks [40] (MD 2017 [95% CI 271.5 to 3762.5]). Five of the other 16 outcomes were in a positive direction but not statistically significant at 8 weeks [39]. A trial of a fit-bit versus standard pedometer reported increases in physical activity which were not statistically significant using 3 objective measures at 16 weeks [38]. A trial of an SMS based intervention plus an unblinded accelerometer reported statistically significant increases in 3 objective measures of physical activity at 4 weeks [44] (Table 9).
Interventions delivered by other or mixed mobile technology media: anthropometric and bio-medical outcomes. A trial of an app-based intervention found no effects on weight loss or Smoking cessation, physical activity/diet and alcohol reduction interventions delivered by mobile phones BMI at 8 weeks [40], and a trial of a fit-bit intervention found no effect on weight loss at 16 weeks [38].
A trial of an app found no effect on three medical outcomes (change in systolic/diastolic blood pressure or change in resting heart rate) at 8 weeks [40]. A trial involving multiple monitoring devices connected to a blackberry phone found a small but statistically significant effect on change in systolic blood pressure, but no effect on the other eleven medical outcomes reported at 12 weeks [47].
Diet-primary outcomes. Interventions delivered by other or mixed mobile technology media: anthropometric and bio-medical outcomes. One trial in which personal email advice was sent by mobile phone in response to self-monitoring of daily salt excretion observed no statistically significant effect of the intervention on blood pressure, BMI, waist circumference or body weight [52]. One trial assessing the efficacy of a diet tracking application versus instructing participants to track their dietary intake on their phone's memo function showed no statistically significant effect on weight loss [54] (Table 10).
Diet and physical activity-primary outcomes. Interventions delivered by SMS alone: anthropometric outcomes. Random effects meta-analysis showed a borderline statistically significant effect of SMS-based interventions on BMI scores measured in three trials between 6 .01] p = 0.052), however there was evidence of substantial between-study heterogeneity (I 2 = 70.2%) (Fig 11). In the two other SMS based intervention trials, one observed no effect on change in BMI at 6 months [93]. The other observed a small but statistically significant reduction in BMI when measured in percentiles (MD -0.10 [95% CI -0.13, -0.07]) but no effect when BMI was measured using Z-scores at 12 months (MD 0.00 [95% CI -0.30 to 0.30]) [71].
Pooled analysis of four trials examining the effect of SMS-based interventions showed no statistically significant effect on waist circumference measured between 6 and 24 months [55,64,81,82] (pooled MD -2.19 [95% CI -4.88, 0.51] p = 0.112, I 2 = 82.8%) (Fig 15). However, one of these trials demonstrated a significant effect when the outcome was change in waist circumference at 12 months [64]. One trial reported a statistically significant reduction in hip circumference at 6 months [82]  Smoking cessation, physical activity/diet and alcohol reduction interventions delivered by mobile phones Interventions delivered by SMS alone: bio-medical outcomes. Pooled analyses of three trials [55,81,82] of SMS-based interventions aiming to increase physical activity and improve diet found evidence of a statistically significant reduction in triglyceride levels (pooled MD -0.19 mmol/L [95% CI -0.29, -0.08, p-value = 0.001], I 2 = 0.0%) (Fig 16). Pooled effects were heterogenous (I squared 59-90%), not statistically significant but in the direction of benefit for total cholesterol, high density lipoprotein cholesterol, systolic blood pressure and diastolic blood pressure, with one trial reporting statistically significant improvements (Figs 17,18,19 and 20). One trial also found a borderline significant effect in reducing LDL cholesterol and a significant reduction in heart rate [82]. Another of these trials showed no effect on glucose levels [81]. A trial of an SMS-based intervention vs information in a pamphlet also showed no effect on systolic or diastolic blood pressure [83]. Pooled analysis of two studies examining the effect of SMS based interventions for weight management showed a statistically significant reduction in the cumulative incidence of diabetes [55,81] (pooled RR 0.67 [95% CI 0.49, 0.90], I 2 = 0.0%) (Fig 21).
Interventions delivered by other or mixed mobile technology media: anthropometric outcomes. Pooled analysis of four studies [57,59,67,76] found no evidence that interventions delivered through mobile apps resulted in greater change in weight when measured between 6 Smoking cessation, physical activity/diet and alcohol reduction interventions delivered by mobile phones  (Fig 22). A trial of a mobile phone activity monitor intervention reported no statistically significant Smoking cessation, physical activity/diet and alcohol reduction interventions delivered by mobile phones changes in weight or waist circumference at 9 months [78]. A trial assessing an intervention delivered by SMS and app showed no significant effect on weight loss at 3 months (Hebden et al., 2014). There was no statistically significant change in BMI for two trials that used smartphone weight loss applications at 6 months and 24 months [57,59], or at 3 months in a trial of an intervention delivered by SMS and app [65]. A trial examining the effect of a podcast plus SMS intervention versus podcast-only found no effect on percentage change in body weight [77]. A trial of a multi-component smartphone-based intervention involving text messages, audio and video files, pre-installed health diary and activity monitoring, a blood pressure monitor and weight scale reported no statistically significant change in weight or waist circumference [80]. In Martin and colleagues' [69] study involving overweight and obese adults, intervention participants were instructed to weigh themselves daily on a bathroom scale and to wear an activity monitor provided to them. This wirelessly transmitted data to a website accessible to a counsellor and feedback was delivered through a loaned smartphone via text, phone calls, email. This intervention had a significant effect on weight loss (MD -7.20 kg [95% ), in addition to reductions in waist circumference. A trial of a mobile-app had no effect on waist circumference among women, but there was evidence of a statistically significant reduction among men [57] (Table 11). Smoking cessation, physical activity/diet and alcohol reduction interventions delivered by mobile phones Interventions delivered by other or mixed mobile technology media: bio-medical outcomes. A trial of a mobile-app intervention showed no effect on blood pressure [67], while a study of a voice-call delivered intervention demonstrated an effect on lowering blood pressure, HbA1c, and LDL cholesterol [56]. An intervention delivered by mixed mobile phone media showed statistically significant reductions in diastolic blood pressure and a reduction which was not statistically significant in systolic blood pressure [69]. A trial of a multi-component smartphone-based intervention involving text messages, audio and video files, pre-installed health diary and activity monitoring, a blood pressure monitor and weight scale, showed no statistically significant effect on cholesterol, HbA1c, heart rate, or systolic blood pressure, but had a statistically significant effect in reducing triglycerides, and increasing diastolic blood pressure [80].
Physical activity, diet and physical activity, diet-secondary outcomes. Of 128 selfreported outcomes 18 demonstrated statistically significant benefits and none showed statistically significant harms (S2 Table).
Alcohol-primary outcomes. There were no objectively measured outcomes reported in trials of alcohol reduction interventions.
Alcohol-secondary outcomes. For trials targeting alcohol consumption, one trial delivering supportive SMS observed a statistically significant increase in the number of days to first drink after inpatient discharge but no statistically significant effect on drinking frequency or Smoking cessation, physical activity/diet and alcohol reduction interventions delivered by mobile phones cognitive outcomes [84]. A trial of an SMS-based drinking assessment intervention found a statistically significant reduction in the number of binge drinking days and number of drinks per drinking day among the intervention group receiving real time feedback, but no such effect in the intervention group who did not receive feedback [86]. A second trial delivering a smartphone intervention observed a small but statistically significant effect on the number of risky drinking days but no effect on continuous abstinence [87]. Another trial observed no effect of SMS-based drinking assessments and brief interventions on drinking frequency in young adults discharged from the Emergency Department [90]. One trial found a small but statistically significant beneficial effect of an intervention delivered by interactive voice response on a multi-item scale measuring alcohol consumption, alcohol dependence and alcohol-related harm [85]. A study among Swedish students assessing the effect of an application-based intervention found that those in the intervention group had slightly increased alcohol consumption [86] (S2 Table).

Quality of evidence assessment
The assessment of quality of evidence for pooled outcomes is reported in Table 12.
According to the GRADE [19] criteria (Table 12), there was high quality evidence of benefit for smoking cessation support delivered by text message and no evidence of harms. For SMS  Smoking cessation, physical activity/diet and alcohol reduction interventions delivered by mobile phones based physical activity interventions, there was low quality evidence of changes in physical activity which was not statistically significant. For SMS based diet and physical activity interventions there was low quality evidence suggesting benefit in reducing the incidence of diabetes in those with pre diabetes and modest or small benefits in change in weight (KG or %), BMI and triglycerides. The evidence of benefit for end point weight, waist circumference, total cholesterol, and blood pressure was very low, with one trial at low risk of bias conducted in those with coronary heart disease reporting statistically significant improvements. The evidence of benefit on HDL cholesterol was very low with one trial reporting statistically significant improvements. The effect of diet and physical activity interventions delivered by app was in the direction of a small benefit, but not statistically significant.

Key findings
We identified 71 trials of interventions delivered by mobile phone targeting prevention of NCD focussed on smoking cessation, physical activity, diet, and alcohol reduction. No trials reported effects on morbidity or mortality. Smoking cessation, physical activity/diet and alcohol reduction interventions delivered by mobile phones There is high quality evidence that smoking cessation support delivered by text message for smokers making a quit attempt increases smoking cessation in trials conducted in high income countries and no evidence for adverse effects of these interventions. In single trials there was no suggestion that text messages to prompt a quit attempt and link people with existing smoking cessation services are more effective than a leaflet with the same content [24]. There was low quality evidence that phone counseling by mobile phone increased smoking cessation at 3 months [35].
For physical activity interventions delivered by SMS, App or fit bit, trials reporting outcomes at 3 months or longer showed no evidence of benefit. The effects of multiple monitoring devices was mixed with statistically significant benefits in only one of 12 biomedical outcomes at 12 weeks.
For SMS based diet and physical activity interventions there was low quality evidence suggesting benefit in reducing the incidence of diabetes in those with pre diabetes and for modest or small benefits in change in weight (KG or %), BMI and triglycerides. The evidence of effects on end point weight, waist circumference, total cholesterol, and blood pressure were in the direction of benefits, but not statistically significant and highly heterogeneous in pooled analyses, with one trial at low risk of bias conducted in those with coronary heart disease reporting statistically significant improvements. The evidence of benefit on HDL cholesterol was Smoking cessation, physical activity/diet and alcohol reduction interventions delivered by mobile phones heterogeneous and very low quality, with one trial reporting statistically significant improvements. The effect of diet and physical activity interventions delivered by app on change in weight at 6 months or longer was in the direction of a small benefit, but not statistically significant. There was mixed evidence regarding benefits of interventions delivered by multiple smartphone media.
There were some promising but inconclusive self-reported effects from alcohol reduction trials.

Strengths and weaknesses of the review
This is a comprehensive review of all randomised controlled trials of interventions delivered by mobile phone designed to prevent non-communicable diseases. Our review has several strengths. The methods are reproducible, screening for trials and data extraction was conducted by two researchers, we used standard Cochrane tools for assessing risk of bias and used the GRADE criteria for assessing the overall quality of evidence. We only included randomised controlled trials, which are less prone to bias than other types of controlled study. We did not pool the results of interventions delivered by mobile phone with those delivered by mobile phone in conjunction with non mobile phone based components, as has been done in some Smoking cessation, physical activity/diet and alcohol reduction interventions delivered by mobile phones previous reviews [10][11][12]. Only six trials did not provide sufficient data to calculate effect estimates.
There are also a number of limitations to our review. It was not possible to contact authors for data in this review, due to time and funding constraints. It was beyond the scope of our review to include interventions delivered by PDA or hand held computer or to review all internet or video based interventions, which in principal can be viewed on many modern mobile phones. Our review aimed to examine the effects of interventions delivered by mobile technologies alone. We excluded interventions combining mobile technologies with additional interventions such as face-to-face counselling, which could be subject to a separate systematic review. We only pooled the results of trial where the trial aim, outcomes and mobile phone media used were the same (e.g. SMS, application software). Nonetheless some of the results of pooled analyses were heterogeneous. This is likely to be due to the wide range of factors which could influence the effectiveness of particular mobile technology interventions including: trial quality [95], participant factors, the setting (low/middle or high income country), intervention design, intervention components (e.g. the behaviour change techniques employed), intensity or intervention duration. We only pooled objectively measured outcomes in meta-analyses due to prior evidence that self-reported outcomes in behaviour change trials can be prone to overstated benefits. Some evidence in our review supports this, for example, a smoking cessation trial showed a null result for a biochemically-confirmed measure, but a benefit in the equivalent self-report outcome [21]. Our review provides no insight into the mechanism of action of interventions. The examination of funnel plots in exploring publication bias was limited as few trials contributed to some pooled analyses.
The scope of the review encompasses trials of novel non-communicable disease prevention interventions delivered by mobile phone. It is plausible that in lower and middle income countries the main benefits of mobile phones would not be evaluated by RCT for example simply owning a phone might afford the potential for the first time for people to gain information about and access to existing health promotion interventions and services.

Discussion in relation to existing literature
We provide an updated systematic review of evidence regarding the effects of interventions for smoking cessation, physical activity, diet and alcohol. Based on short term outcomes, self reported outcomes and non randomised study designs previous reviews have concluded that there are benefits of interventions targeting physical activity and /or diet. Our review demonstrated there is no reliable evidence of benefit for physical activity interventions delivered by SMS, app or fit bit at 3 months or longer. Whilst some diet and physical activity interventions Smoking cessation, physical activity/diet and alcohol reduction interventions delivered by mobile phones have reported benefits, the overall quality of evidence for even modest benefits is low or very low. Our results are similar to the findings for existing systematic reviews of alcohol reduction interventions and show some promising findings but overall very low quality evidence for their benefits [96]. In contrast to the existing Cochrane review of smoking cessation interventions delivered by mobile phone we did not pool objective outcomes with self reported outcomes and we examined the effects of different mobile phone media (SMS, app, video, interactive voice and phone counselling) separately. The rational for this was that there are important differences between different mobile phone media. For example; SMS messages are sent directly to peoples phones, flash up on screens, and can be stored to be re-read at convenient times, whilst apps are dependent on motivated participants going to the app to view content (unless apps also use instant messages), voice messages are often only accessible when sent and cannot always be reviewed. This decision appears to be supported by the results of the review. Our resulting pooled analyses of the effects of SMS based smoking cessation support during a quit attempt demonstrated clear benefits without heterogeneity [13], but there was no suggestion of benefit for smoking cessation interventions delivered by video or interactive voice recording [13,29].
In comparison to the findings of our 2013 review [15], this review confirms that SMS based smoking cessation support increases quitting in those willing to make a quit attempt, with a Smoking cessation, physical activity/diet and alcohol reduction interventions delivered by mobile phones larger number of trials now included in the meta-analysis. Since our last review trials have been published showing that interventions designed to prompt a quit attempt have not shown clear benefits. This review includes recently published evidence of benefits in diet and physical activity interventions in preventing the onset of diabetes, in those with pre-diabetes. There is still no clear evidence of benefit of physical activity, diet and physical activity or dietary interventions for other populations at 3 months or longer. There are a larger number of alcohol reduction trials included in this review (8 compared to 1), but whilst the effects of some alcohol reduction interventions look promising the results are still inconclusive.

Meaning of the study, implications for clinicians or policy makers
Continuous abstinence is considered the gold standard for smoking cessation trials. Estimates of effect using biochemically verified point prevalence for smoking cessation support delivered by SMS were lower than continuous abstinence effect estimates. This would be expected as point prevalence estimates are diluted by quit attempts occurring randomly throughout the follow up period in both intervention and control groups but unrelated to the intervention. The interventions for smoking cessation included in our pooled analyses contained at least 8 behaviour change techniques used in effective face to face smoking cessation support but Smoking cessation, physical activity/diet and alcohol reduction interventions delivered by mobile phones adapted for delivery by text message. SMS support for smoking cessation is highly cost-effective [97] and has been implemented in New Zealand, UK, USA and India. In England between 60,000 and 90,000 people have registered for smoking cessation support delivered by text message each year since 2012. An evaluation of implementation showed that the 4 week quit rates achieved were similar or higher than in the UK trial [98]. An evaluation of implementation of an interventions delivered by SMS in India is ongoing. Health services should consider implementing smoking cessation support delivered by SMS with similar content to interventions found to be effective in trials.
Whilst some trials report benefits, there is currently insufficient evidence of beneficial effects on long term objective outcomes to warrant implementation of interventions for physical activity or diet and physical activity. The statistically significant effects of diet and physical activity interventions on change in weight, illustrates the greater power of these outcomes compared to absolute weight outcomes, which did not achieve statistical significance. The effects of behavioural support for weight loss delivered by text message is broadly consistent with the modest benefits achieved by behavioural support for weight loss in general [99,100].
Few trials target smoking diet and physical activity, but there is no evidence from existing trials to suggest that targeting multiple behaviours reduces the intervention effects for individual behaviours. Most interventions were delivered by text message. There was no good Smoking cessation, physical activity/diet and alcohol reduction interventions delivered by mobile phones Smoking cessation, physical activity/diet and alcohol reduction interventions delivered by mobile phones evidence of benefit of app based interventions for smoking cessation, physical activity, diet or alcohol reduction.

Unanswered questions and future research
Further evaluations of the impact and costs of smoking cessation interventions delivered by mobile phone in low and middle income settings, are needed. There is a considerable body of existing research regarding effective interventions for behaviour change and much of this literature suggests multifaceted interventions are required [101][102][103][104][105]. Interventions should be developed using established methods including: needs assessment, reviewing the evidence regarding factors influencing the target outcome, drawing on behaviour change theory and evidence based behavioural change techniques and adapting content based on user views regarding the acceptability, comprehensibility and relevance of intervention content [106][107][108]. However, the impact of even the most well developed interventions delivered by mobile phone for preventative behaviours such as diet and physical activity, which are strongly influenced by environmental factors, are likely to be modest. Nonetheless, intervention delivery costs are low, so adequately powered high quality trials of optimised interventions targeting diet and physical activity are required to reliably establish their effects, especially in high risk groups such as for diabetes prevention in those with pre-diabetes. The effects of interventions targeting diet, physical activity and smoking cessation on morbidity and mortality in populations with existing coronary heart disease should be established.
Further research is required to evaluate the mechanism of action of interventions. Understanding the mechanism of action of interventions could inform the content of future interventions. A range of questions regarding the effects of mobile technologies remain open to question including whether some intervention functions are more effective delivery tools than others (SMS, video, oral communication, application software), which behaviour change techniques are effective when modified for delivery by mobile phone and whether the effectiveness of interventions is influenced by setting or participant demographics.
Whilst some trials have been conducted in low and middle income countries, the majority of the research has been conducted in high-income countries. In view of the high coverage of mobile technologies in these settings, trials of interventions in low and middle-income countries are required.

Conclusion
In high income settings, SMS based smoking cessation support during a quit has been shown to result in increases in smoking cessation. Effective interventions included 8 or more behaviour change techniques. The effects, costs and impact of interventions in low income countries should be established. There is only weak evidence regarding the benefits of diet physical activity and alcohol reduction interventions delivered by mobile phone. Large scale high quality trials of the effects of optimised interventions on morbidity and mortality, especially in high risk groups, are warranted.