Вмешательства с использованием мобильных телефонов (текстовых сообщений и приложений) для отказа от курения
Abstract
Background
Mobile phone‐based smoking cessation support (mCessation) offers the opportunity to provide behavioural support to those who cannot or do not want face‐to‐face support. In addition, mCessation can be automated and therefore provided affordably even in resource‐poor settings. This is an update of a Cochrane Review first published in 2006, and previously updated in 2009 and 2012.
Objectives
To determine whether mobile phone‐based smoking cessation interventions increase smoking cessation rates in people who smoke.
Search methods
For this update, we searched the Cochrane Tobacco Addiction Group's Specialised Register, along with clinicaltrials.gov and the ICTRP. The date of the most recent searches was 29 October 2018.
Selection criteria
Participants were smokers of any age. Eligible interventions were those testing any type of predominantly mobile phone‐based programme (such as text messages (or smartphone app) for smoking cessation. We included randomised controlled trials with smoking cessation outcomes reported at at least six‐month follow‐up.
Data collection and analysis
We used standard methodological procedures described in the Cochrane Handbook for Systematic Reviews of Interventions. We performed both study eligibility checks and data extraction in duplicate. We performed meta‐analyses of the most stringent measures of abstinence at six months' follow‐up or longer, using a Mantel‐Haenszel random‐effects method, pooling studies with similar interventions and similar comparators to calculate risk ratios (RR) and their corresponding 95% confidence intervals (CI). We conducted analyses including all randomised (with dropouts counted as still smoking) and complete cases only.
Main results
This review includes 26 studies (33,849 participants). Overall, we judged 13 studies to be at low risk of bias, three at high risk, and the remainder at unclear risk. Settings and recruitment procedures varied across studies, but most studies were conducted in high‐income countries. There was moderate‐certainty evidence, limited by inconsistency, that automated text messaging interventions were more effective than minimal smoking cessation support (RR 1.54, 95% CI 1.19 to 2.00; I2 = 71%; 13 studies, 14,133 participants). There was also moderate‐certainty evidence, limited by imprecision, that text messaging added to other smoking cessation interventions was more effective than the other smoking cessation interventions alone (RR 1.59, 95% CI 1.09 to 2.33; I2 = 0%, 4 studies, 997 participants). Two studies comparing text messaging with other smoking cessation interventions, and three studies comparing high‐ and low‐intensity messaging, did not show significant differences between groups (RR 0.92 95% CI 0.61 to 1.40; I2 = 27%; 2 studies, 2238 participants; and RR 1.00, 95% CI 0.95 to 1.06; I2 = 0%, 3 studies, 12,985 participants, respectively) but confidence intervals were wide in the former comparison. Five studies compared a smoking cessation smartphone app with lower‐intensity smoking cessation support (either a lower‐intensity app or non‐app minimal support). We pooled the evidence and deemed it to be of very low certainty due to inconsistency and serious imprecision. It provided no evidence that smartphone apps improved the likelihood of smoking cessation (RR 1.00, 95% CI 0.66 to 1.52; I2 = 59%; 5 studies, 3079 participants). Other smartphone apps tested differed from the apps included in the analysis, as two used contingency management and one combined text messaging with an app, and so we did not pool them. Using complete case data as opposed to using data from all participants randomised did not substantially alter the findings.
Authors' conclusions
There is moderate‐certainty evidence that automated text message‐based smoking cessation interventions result in greater quit rates than minimal smoking cessation support. There is moderate‐certainty evidence of the benefit of text messaging interventions in addition to other smoking cessation support in comparison with that smoking cessation support alone. The evidence comparing smartphone apps with less intensive support was of very low certainty, and more randomised controlled trials are needed to test these interventions.
PICOs
Резюме на простом языке
Могут ли программы, предоставляемые с помощью мобильных телефонов, помочь людям бросить курить?
Актуальность
Курение табака является ведущей причиной предотвратимой смерти. Мобильные телефоны могут быть использованы для поддержки людей, которые хотят бросить курить. В этом обзоре мы сосредоточились на программах, которые используют текстовые сообщения или приложения для смартфонов для этих целей.
Дата поиска
Мы провели поиск опубликованных и неопубликованных исследований в октябре 2018 года.
Характеристика исследований
Мы включили 26 рандомизированных контролируемых исследований (с участием более 33 000 человек), в которых сравнивали частоту отказа от курения у людей, которые получали текстовые сообщения или использовали приложения для смартфонов, чтобы помочь им бросить курить, с людьми, которые не участвовали в этих программах. Мы были заинтересованы только в тех исследованиях, в которых оценивали эффект (курение) в течение шести месяцев и более.
Основные результаты
Мы обнаружили, что программы обмена текстовыми сообщениями могут быть эффективными в поддержке людей бросить курить, увеличивая частоту отказа от курения на 50‐60%. Это было в тех случаях, когда их сравнивали с минимальной поддержкой или когда их тестировали в качестве дополнения к другим формам поддержки для отказа от курения. Не было достаточно доказательств, чтобы определить влияние приложений для смартфонов.
Качество и полнота доказательств
Большинство исследований были высокого качества, хотя в трех исследованиях была высокая частота выбываний из исследований. Мы в средней степени уверены в результатах касательно вмешательств с использованием текстовых сообщений, но были некоторые проблемы с необъяснимыми различиями между результатами исследований, и для некоторых сравнений было не так много данных. У нас низкая уверенность в результатах, касающихся приложений для смартфонов, и необходимы дополнительные исследования в этой области.
Authors' conclusions
Summary of findings
| Text messaging compared to minimal support for smoking cessation | ||||||
| Patient or population: people who smoke | ||||||
| Outcomes | Anticipated absolute effects* (95% CI) | Relative effect | № of participants | Certainty of the evidence | Comments | |
| Risk with minimal SC support | Risk with text messaging | |||||
| Long‐term abstinence (all randomised) Measured with self‐report and biochemical validation at 6 to 12 months | Study population | RR 1.54 | 14,133 | ⊕⊕⊕⊝ | ||
| 6 per 100 | 9 per 100 | |||||
| *The risk in the intervention group (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). | ||||||
| GRADE Working Group grades of evidence | ||||||
| aDowngraded one level due to inconsistency: substantial unexplained heterogeneity (I2 = 71%). | ||||||
| Text messaging in addition to other smoking cessation support compared to other smoking cessation support alone for smoking cessation | ||||||
| Patient or population: people who smoke | ||||||
| Outcomes | Anticipated absolute effects* (95% CI) | Relative effect | № of participants | Certainty of the evidence | Comments | |
| Risk with other SC support alone | Risk with text messaging + other SC support | |||||
| Long‐term abstinence (all randomised) Measured as self‐reported and biochemical validation at 6 to 12 months | Study population | RR 1.59 | 997 | ⊕⊕⊕⊝ | ||
| 8 per 100 | 12 per 100 | |||||
| *The risk in the intervention group (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). | ||||||
| GRADE Working Group grades of evidence | ||||||
| aDowngraded one level due to imprecision: fewer than 300 events overall. | ||||||
| Smartphone app compared to lower‐intensity support for smoking cessation | ||||||
| Patient or population: people who smoke Intervention: smartphone app | ||||||
| Outcomes | Anticipated absolute effects* (95% CI) | Relative effect | № of participants | Certainty of the evidence | Comments | |
| Risk with lower intensity SC support | Risk with Smartphone app | |||||
| Long‐term abstinence (all randomised) Measured with self‐report and biochemical validation at 6 months | Study population | RR 1.00 | 3079 | ⊕⊝⊝⊝ | ||
| 8 per 100 | 8 per 100 | |||||
| *The risk in the intervention group (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). | ||||||
| GRADE Working Group grades of evidence | ||||||
| aDowngraded one level due to inconsistency: considerable unexplained statistical heterogeneity (I2 = 59%). | ||||||
Background
Description of the condition
Tobacco remains one of the most important risk factors for poor health across the globe (IHME 2018). Many countries are looking for sustainable options for the provision of smoking cessation support on a large scale.
Description of the intervention
'mHealth' describes the use of mobile communications technologies and mobile phones to support health care. In this review, we are specifically interested in the use of text messaging and smartphone applications (apps) to support smoking cessation.
How the intervention might work
The benefits of mobile phone‐based smoking cessation support (mCessation) interventions are: the ease of use anywhere at any time; cost‐effective delivery and scalability to large populations, regardless of location; the ability to tailor messages to key user characteristics (such as age, sex, ethnicity); the ability to send time‐sensitive messages with an 'always on' device; the provision of content that can distract the user from cravings; and the ability to link the user with others for social support.
A key benefit of the use of mobile phones for health programmes is their widespread uptake in those areas where health services are not easily accessible or used. In 2018, the number of mobile phone subscriptions globally topped 8 billion, with the developing world now having more mobile phone subscriptions than population (population penetration of 102%; ITU 2018). There is evidence to suggest that people from lower socioeconomic groups may prefer mCessation interventions due to the greater feeling of control associated with the ability to decide when and where they engage with messages, and the perception of around‐the‐clock support (Boland 2017). Focusing mCessation efforts on the populations in greatest need, could help to address the health inequalities that come about from high use of tobacco and lack of accessible health promotion and prevention services in low‐resource settings globally.
Furthermore, initial research suggests that the use of text messaging for smoking cessation is cost effective. Guerriero 2013 found that the cost of text message‐based support was GBP 278 per quitter. When the future health service costs saved (as a result of smoking cessation) were included, with 0.5 quality‐adjusted life years (QALYs) gained per quitter, text‐based support was considered to be cost saving.
Why it is important to do this review
Smartphones (mobile phones with a computer operating system) are fast becoming the computer of choice, or at least the most accessible computer, in many countries. According to the International Telecommunications Union only 36.3% of low‐and middle‐income countries has a computer in the household, but 61% have mobile broadband subscriptions (allowing mobile phones to access to the Internet; ITU 2018). Therefore, it was important to update this review to include studies on the effectiveness of smartphone apps, as well as text messaging interventions, for smoking cessation.
Objectives
To determine whether mobile phone‐based smoking cessation interventions increase smoking cessation rates in people who smoke.
Methods
Criteria for considering studies for this review
Types of studies
Randomised or quasi‐randomised trials. Cluster‐randomised trials were eligible for inclusion.
Types of participants
People who smoked at study enrolment.
Types of interventions
We included studies that examined any intervention that could be considered predominantly a mobile phone‐based programme (such as text messaging or smartphone apps) for smoking cessation. We excluded interventions where mobile phones were seen as an adjunct to a predominantly face‐to‐face or Internet programme, such as to remind participants of appointments, or where the effects of the various components of a multi‐faceted programme could not be separated. We also excluded interventions that could be performed via any type of telephone such as telephone counselling. We did not exclude any studies based on comparator, but instead grouped studies by comparators in the analyses.
Types of outcome measures
The primary outcome was smoking abstinence at longest follow‐up, and at least six months from baseline. Where multiple measures were available, we preferred sustained abstinence to point prevalence abstinence, and biochemically validated results to self‐report.
There is no obvious risk of adverse events for text messaging or smartphone app interventions, and so we have not included this as an outcome in this review.
Search methods for identification of studies
For the present update of the review, we searched the Cochrane Tobacco Addiction Group's Specialised Register on 29 October 2018 using the terms 'mobile phone', 'cell phone', 'txt', 'pxt', 'sms', or 'mms' in the title, abstract or keyword fields. The Specialised Register includes reports of possible controlled trials of smoking cessation interventions identified from sensitive searches of databases. At the time of the search, the Register included the following results of searches
-
Cochrane Central Register of Controlled trials (CENTRAL; 2018, Issue 1)
-
MEDLINE (via Ovid, to 26 October 2018)
-
Embase (via Ovid, to 28 October 2018)
-
PsycINFO (via Ovid; to 22 October 2018)
See the Cochrane Tobacco Addiction Group website for full search strategies and a list of other resources searched. We also searched the World Health Organization International Clinical Trials Registry Platform (WHO ICTRP; apps.who.int/trialsearch/) and ClinicalTrials.gov trials registers for ongoing or recently completed studies. We searched through the reference lists of identified studies for any additional eligible studies and attempted to contact the authors of ongoing studies.
We placed no restrictions on publication language or date.
Data collection and analysis
Selection of studies
The Cochrane Tobacco Addiction Group's Information Specialist ran the searches and provided the results. Two review authors (YG, HM) independently pre‐screened the titles and abstracts of records identified in duplicate to exclude reports that had no relevance to the topic and to provide a list of potentially relevant citations. A third reviewer (CB) resolved any differences in initial screening. Two review authors (from RW, YG, CB, RD) independently reviewed full‐text manuscripts in duplicate for the final eligibility screen.
We resolved any disagreements by discussion or by obtaining further information through contacting study authors. We recorded reasons for exclusion of studies in the Characteristics of excluded studies table. We contacted authors of unpublished, registered studies, which could potentially have been completed, to determine ongoing status or to request unpublished data.
Data extraction and management
We extracted the following methodological details from the included study reports and presented them in the Characteristics of included studies table. Two review authors (from RW, YG, RD, CB, HM) independently extracted data using the standardised Covidence data extraction form. A third review author provided a review of the quality assessment and a consensus check.
-
Funding source
-
Authors' declarations of interest
-
Country and context of the study
-
Study design
-
Number of participants
-
Age and other relevant recorded characteristics of study participants
-
Inclusion criteria
-
Exclusion criteria
-
Intervention details
-
Control details
-
Definition of abstinence outcome
-
Smoking cessation rates at six months (self‐reported abstinence or biochemically verified abstinence, or both)
-
Smoking cessation rates at final follow‐up (if follow‐up greater than six months and where these data were available)
Assessment of risk of bias in included studies
Two review authors (from RW, YG, RD, CB, HM) independently assessed the risk of bias for included studies, based on the guidance of the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2017), and the Cochrane Tobacco Addiction Group. For each study, we assessed the following domains.
-
Random sequence generation
-
Allocation concealment
-
Blinding of outcome assessment
-
Incomplete outcome data
-
Other sources of bias
Specific 'Risk of bias' guidance developed by the Cochrane Tobacco Addiction Group to assess smoking cessation studies states that performance bias (relating to the blinding of participants and providers) should not be assessed for behavioural interventions, as it is impossible to blind people to these types of interventions. We graded detection bias as low where there was biochemical verification of abstinence, or where abstinence was self‐reported with no difference in face‐to‐face contact between control and intervention arms. We considered bias due to incomplete outcome as low risk where numbers lost to follow‐up were clearly reported for each group, the overall loss was not greater than 50%, and the difference between groups was not greater than 20%, or sensitivity analysis showed that the direction of effect was not sensitive to different imputation methods for loss to follow‐up.
Each review author recorded information in study reports relevant to each relevant domain and then judged each domain as either at low, high, or unclear risk of bias. We resolved disagreements through discussion with a third review author.
Measures of treatment effect
We recorded the information below.
-
Smoking cessation rates at six months or longer using the most stringent measure available
-
Biochemically verified abstinence, where available
We calculated risk ratios (RR) and 95% confidence intervals (CI) for the smoking cessation outcome for each included study. We calculated outcomes on an intention‐to‐treat basis, including all participants randomised to a trial arm and assuming that participants lost to follow‐up had continued to smoke or relapsed.
Dealing with missing data
If we found any important study characteristics or outcome data to be missing, we followed up with study authors where possible.
Assessment of heterogeneity
In order to assess whether it was appropriate to pool studies and conduct meta‐analyses we assessed the characteristics of included studies to identify any clinical or methodological heterogeneity. Where we deemed studies homogeneous enough to be combined meaningfully, we conducted a meta‐analysis, and we assessed statistical heterogeneity using the I2 statistic; we deemed an I2 value greater than 50% to indicate substantial heterogeneity (Higgins 2003).
Assessment of reporting biases
We planned to use funnel plots to assess reporting bias for any comparisons where we identified and analysed abstinence rates from at least 10 studies. Only the 'text messaging versus minimal smoking cessation support' comparison met this criteria in this review; therefore a funnel plot was generated for this comparison only. Funnel plots illustrate the relationship between the effect estimates from individual studies against their size or precision. The greater the degree of asymmetry, the greater the risk of reporting bias.
Data synthesis
We conducted meta‐analyses of the included studies, using the Mantel‐Haenszel random‐effects method to pool RRs and 95% CIs calculated for the smoking abstinence outcome, across the following comparisons.
-
Text messaging versus minimal smoking cessation support (including standard self‐help materials, as is standard practice in the Cochrane Tobacco Addiction Group)
-
Text messaging in addition to another form of smoking cessation support
-
Text messaging versus other smoking cessation support
-
Higher‐ versus lower‐frequency text messaging
-
Smartphone app versus less intensive smoking cessation support
Where studies had multiple intervention arms relevant to a single meta‐analysis, we split control arm data to avoid double‐counting.
Subgroup analysis and investigation of heterogeneity
We carried out the following subgroup analyses.
-
We split the 'smartphone app versus less intensive smoking cessation support' comparison into two subgroups to reflect the different comparators used across studies; either minimal non‐app smoking cessation support (e.g. self‐help materials, information on existing stop‐smoking services) or a less intensive smartphone app.
Sensitivity analysis
We conducted the following sensitivity analyses.
-
We calculated pooled RRs and 95% CIs for all analyses using complete case data to calculate quit rates. People may drop out of studies for reasons other than still smoking, and these reasons may differ between groups. For example, people who successfully stop smoking may withdraw from receiving an intervention if the text messages remind them of smoking. Therefore, this analysis tests whether assuming that all people lost to follow‐up are smoking (as in our primary analyses of all participants randomised) is potentially biasing our results.
-
Removing any studies judged to be at high risk of bias from all comparisons
-
Removing the only cluster‐RCT (Haug 2013), as information was not available to adjust for any potential clustering effect
-
Removing the two studies carried out in a pregnant (Abroms 2017), or postnatal population (Yu 2017), as these populations differ substantially from those recruited in the other studies.
'Summary of findings' tables
Following standard Cochrane methods (Schünemann 2017), we created a 'Summary of findings' table for the primary outcome (smoking abstinence), for the following comparisons.
-
Text messaging versus minimal smoking cessation support
-
Text messaging in addition to other smoking cessation support
-
Smartphone app versus less intensive smoking cessation support
Also following standard Cochrane methodology (Schünemann 2017), we used the five GRADE considerations (risk of bias, inconsistency, imprecision, indirectness and publication bias) to assess the certainty of the body of evidence for the abstinence outcome for each comparison, and to draw conclusions about the certainty of evidence within the text of the review.
Results
Description of studies
See Characteristics of included studies; Characteristics of excluded studies; and Characteristics of ongoing studies for further details.
Results of the search
The previous version of this review (Whittaker 2016), included 12 studies (Abroms 2014; Bock 2013; Borland 2013; Ferguson 2015; Free 2009; Free 2011; Gritz 2013; Haug 2013; Naughton 2014; Rodgers 2005; Tseng 2017; Whittaker 2011). Gritz 2013 was excluded at this update, as their intervention (telephone counselling and help line) was significantly different from the other interventions included in this review. A telephone help line intervention does not need to be carried out using a mobile phone specifically. Therefore, 11 of the previously included studies were included at this update, as well as one previously 'ongoing' study that was changed to 'included' as the study is now complete and data was available (Danaher 2019).
For this update of our review, the new literature search identified 370 studies (Figure 1). Many were duplicates, or unrelated and were immediately excluded at the title and abstract screening phase. We screened the full‐text of 71 reports of 62 studies, excluding 16 studies, and leaving 14 new studies eligible for inclusion at this update (Abroms 2017; Alessi 2017; Augustson 2017; Baskerville 2018; BinDhim 2018; Chan 2015; Cobos‐Campos 2017; Garrison 2018; Herbec 2019; Liao 2018; Peiris 2019; Squiers 2017; Wilson 2016; Yu 2017). Data were supplied by the authors for two studies (Danaher 2019; Herbec 2019). Reasons for excluding studies included: intervention that was not predominantly a mobile phone programme; not a randomised controlled trial; relapse prevention only; or no abstinence outcome measured at ≥ 6 months follow‐up (see Characteristics of excluded studies table for further details).
Study flow diagram for this update
We also identified 32 ongoing studies at this update. When added to the previously identified ongoing studies there was a total of 34 ongoing studies (for further details see the Characteristics of ongoing studies table).
Included studies
Context and participants
The settings and recruitment methods, and therefore the participants, varied considerably across studies. Where previously this review had included only studies from a small range of high‐income countries, the new studies included in this update provided greater variation in settings, including China (Augustson 2017; Chan 2015; Liao 2018).
Bock 2013 (USA) found usual in‐person recruitment methods slow and shifted to online recruitment methods during the study. Baskerville 2018 (Canada), Borland 2013 (Australia), Danaher 2019 (USA), Garrison 2018 (USA), Squiers 2017 (USA), Herbec 2019 (UK), and Abroms 2014 (USA) also used online recruitment via Internet advertisements. In Abroms 2014 this initially led to some fraudulent participants who were discovered and disqualified, and extra procedures were put in place to prevent this from happening again. Free 2009 and Free 2011 recruited via advertisements at UK primary care centres, smoking cessation clinics, pharmacies, newspapers, websites, bus billboards and on the radio in the UK, and Liao 2018 used similar advertising methods in China. Rodgers 2005 also used direct advertising via websites, email, and posters at tertiary institutions across New Zealand. Similarly Whittaker 2011 (New Zealand) used a wide range of advertising media, including Māori‐specific media, and targeted young people. Alessi 2017 recruited through email, flyers, and print advertisements and Ferguson 2015 (Australia) used advertisements in papers, radio, and Facebook. Abroms 2017 was embedded in the Text4Baby text message (three messages a week) health information programme for pregnant women in the USA. Women who had smoked at least one puff in the past two weeks were eligible to also receive Quit4Baby text messages (between one to eight messages a day) to support smoking cessation. Augustson 2017 recruited through Nokia Life Tools, a service providing more than 100 million users with tools pre‐installed on their Nokia mobile phones, in urban and rural areas of China’s Zhejiang, Heilongjiang and Shaanxi provinces. BinDhim 2018 recruited through the Apple App Store in several countries (Australia, Singapore, UK, USA). Participants were advised that by downloading the app they would be participating in a study. Chan 2015 recruited through a Quit and Win competition in Hong Kong that was promoted in shopping malls and other public areas. Wilson 2016 mailed letters to potential participants in the US Veterans Administration health system. Naughton 2014 was set in primary care practices in the UK with trained smoking cessation advisors providing smoking cessation advice; Cobos‐Campos 2017 in two health clinics in Spain with health advice provided by a doctor or nurse; and Tseng 2017 in large urban HIV clinics. Haug 2013 recruited in vocational schools and differed from the other studies by allowing the inclusion of occasional smokers (at least four cigarettes in the past month or at least one in the preceding week). Peiris 2019 (Australia) recruited via an Aboriginal Community Controlled Health Service, a regional community event, and the New South Wales Government telephone coaching service. Yu 2017 recruited in maternal‐child health centres in China after asking mothers about household second‐hand smoke exposure. The intervention included messages on both the harms of second‐hand smoke (to the mother and her husband) and additional messages to the husband to encourage quitting.
Four studies deliberately targeted young adults (Baskerville 2018 in Canada, Haug 2013 in Switzerland; Squiers 2017 in USA; Whittaker 2011 in New Zealand). Most studies had similar proportions of men and women or slightly more women than men. The exceptions were Abroms 2017, as the intervention was targeted at pregnant women (100% women), Wilson 2016, which recruited 89% male veterans, and the studies in China, where the rates of smoking in women are low (Chan 2015 > 80% men, Liao 2018 94.6% men, Yu 2017 100% men).
Intervention programmes
Text messaging
All studies tested automated text messaging interventions. Eighteen of the included studies used text messaging (SMS) as a central component of the intervention (Abroms 2014; Abroms 2017; Augustson 2017; Bock 2013; Borland 2013; Chan 2015; Cobos‐Campos 2017; Ferguson 2015; Free 2009; Free 2011; Haug 2013; Liao 2018; Naughton 2014; Rodgers 2005; Tseng 2017; Squiers 2017; Whittaker 2011; Yu 2017). Whittaker 2011 sent text messages containing links to theoretically driven video messages from 'ordinary' role models coping with quitting. Several studies paired text messages with in‐person visits or assessments (Bock 2013; Cobos‐Campos 2017; Haug 2013; Naughton 2014).
The text message interventions varied in length from one week (Chan 2015), to five weeks (Ferguson 2015), six weeks (Augustson 2017; Yu 2017), eight weeks (Bock 2013;Squiers 2017), three months (Abroms 2017; Haug 2013; Naughton 2014; Tseng 2017), and six months (Abroms 2014; Cobos‐Campos 2017; Free 2009; Free 2011; Liao 2018; Rodgers 2005; Whittaker 2011), or were variable (Borland 2013).
Eight studies did not state that text messages were tailored to the individual (Abroms 2017; Augustson 2017; Chan 2015; Cobos‐Campos 2017; Liao 2018; Tseng 2017; Squiers 2017; Yu 2017). In other studies using text messages, the degree of individual tailoring varied:
-
Abroms 2014 tailored messages to include first name, quit date, top three reasons for quitting, money saved by quitting, and use of quit‐smoking medications;
-
Bock 2013 and Haug 2013 tailored messages to the stage of readiness to quit;
-
Borland 2013's programme could be interacted with by reporting changes in smoking behavior (e.g. a quit attempt, relapse), so that appropriate stage‐specific messages could be sent;
-
Ferguson 2015 tailored their intervention text messages to contain advice and encouragement tailored to participants' current quit status (preparing to quit, first week of the quit attempt, second week of attempt etc.)
-
Free 2009 and Free 2011 tailored the messages to information collected at baseline about the individual;
-
Naughton 2014 individually tailored messages using 24 items from the iQuit questionnaire and information on smoking status at three and seven weeks;
-
Rodgers 2005 matched participant characteristics to messages by keyword to create an individualised programme;
-
Whittaker 2011's participants selected the role model from whom they wished to receive messages.
A number of text messaging interventions included interactive components such as:
-
the ability to text for more support in the instance of cravings or lapses (Abroms 2014; Bock 2013; Free 2011; Liao 2018; Naughton 2014; Rodgers 2005);
-
an optional Quit Buddy in Rodgers 2005 and Free 2011;
-
a Quit support network in Bock 2013;
-
polls and quizzes (Rodgers 2005);
-
regular checking in on smoking status (Haug 2013).
Borland 2013 was the only study to include some degree of choice. Participants received offers of support via a personalised tailored Internet programme, a text message programme, both programmes, a choice of all three, or a minimal control. For the purposes of meta‐analyses, we compared the text message group with the control group.
Some of the included interventions were somewhat related to each other. The text messaging intervention in Rodgers 2005 was developed in New Zealand, and later adapted and tested in a UK pilot study (Free 2009), and then a large randomised controlled study (Free 2011). The intervention in Abroms 2017 was developed for pregnant women from the same group’s previous intervention for adult smokers in Abroms 2014. The Augustson 2017 intervention in China was adapted from the smoke‐free text programme that was evaluated in Squiers 2017. For further details of the messaging interventions across individual studies see the Characteristics of included studies table.
The control conditions used in the text message studies could be categorised into four groups.
-
Minimal smoking cessation support (13 studies): the control programmes across the studies in this category varied from no smoking cessation support (Haug 2013; Yu 2017), to non‐smoking‐related text messages sent two‐weekly (Free 2009; Free 2011; Rodgers 2005; Whittaker 2011), or weekly (Liao 2018), to written or Internet untailored materials (Abroms 2014; Chan 2015; Ferguson 2015), to links to smoking cessation support (Borland 2013; Rodgers 2005), or regular general health advice provided by a clinician (Cobos‐Campos 2017). Abroms 2017's control group participants received standard non‐smoking‐related Text4Baby text messages (three a week) without the extra smoking cessation‐related Quit4Baby text messages.
-
Another form of smoking cessation support (matched to support received by the intervention group, but without the text messaging intervention; four studies): support varied across studies and included a single session of smoking cessation counselling plus non‐smoking‐related text messages (Bock 2013); smoking cessation behavioural support and pharmacotherapy (Naughton 2014), and behavioural support and pharmacotherapy (Tseng 2017). Participants in the comparison arm of Yu 2017 received in‐person counselling and materials on establishing a smoke‐free home.
-
Another form of smoking cessation support (not matched in the intervention arm; two studies): an Internet‐based interactive smoking cessation programme (Borland 2013), and a five‐minute smoking cessation counselling session (Chan 2015).
-
Higher‐ versus lower‐frequency text messaging. Three studies examined the effect of higher‐ versus lower‐frequency text messages (Augustson 2017; Liao 2018; Squiers 2017). In Augustson 2017 this was comparing 91 messages over six weeks (three a day initially, followed by two a day, then one a day), with one text message a week for six weeks. In Liao 2018 this was three to five messages per day compared with three to five messages per week. Squiers 2017 compared smoking assessment and quit date messages only, with those messages plus motivational preparatory messages for two weeks prior to quitting, and with all of those messages plus six weeks of follow‐up post‐quit messages.
Smartphone apps
Five studies tested the effectiveness of smoking cessation smartphone apps alone (Baskerville 2018; BinDhim 2018; Garrison 2018; Herbec 2019; Peiris 2019). These apps varied considerably in intervention content and components. The app in Baskerville 2018 was described as comprehensive and evidence‐informed, including components such as a quit plan, contingency reinforcement, a link to an online Facebook community, supportive messages through the app, web‐based distraction, information and performance feedback, access to evidence‐based cessation services. BinDhim 2018 described their app as a decision aid (based on the Ottawa Decision Support Framework drawing from a number of psychological and behavioural theories) with additional support with push notifications, messages, diary and benefits tracker. The Garrison 2018 app training modules taught mindfulness for smoking cessation and how to work through cravings. Herbec 2019 included craving management tools within an app that supported smokers to be smoke free for 28 days.
The control conditions used in these smoking cessation app studies could be categorised into two groups:
-
minimal non‐app smoking cessation support that included: a printed self‐help guide (Baskerville 2018), and encouragement to access available smoking cessation services (Peiris 2019);
-
less intensive app support that included an app that provided only basic information. In BinDhim 2018 this included information only on quitting (no structured process or support). Garrison 2018 delivered experience sampling to query smoking, craving, and mindfulness in real time, and the control app in Herbec 2019 was designed to be a minimally credible intervention that resembled the intervention but without key intervention components.
Carbon monoxide (CO) monitoring and contingency management
Alessi 2017 and Wilson 2016 used mobile phone technology slightly differently to the above studies, by specifically using mobile phones to monitor the concentration of carbon monoxide in end‐expiratory air (CO levels). In Alessi 2017 interactive voice response calls would prompt the participant to conduct a CO test using a CO monitor. This was video recorded on the mobile phone and submitted using multimedia messaging. The CO result was provided via interactive voice response call. In the reinforcement arm of the trial, this was supplemented by negative CO test results (not smoking), which were rewarded with chances to win prizes. Therefore, the study had two arms that received mHealth CO monitoring as well as counselling and nicotine replacement therapy (NRT), with one of the arms also receiving rewards for smoking abstinence. Wilson 2016 combined cognitive behavioural telephone counselling and access to NRT with a mobile app for CO monitoring and contingency management in one study arm, and compared this to the same intervention without the CO monitoring and contingency management app. Participants provided CO readings twice a day by video through the app and received payment for abstinence in the intervention arm.
Smartphone app plus text messaging
Danaher 2019 tested an intervention that used both an integrated mobile web app and text messaging. Text messages were prompts and motivations to visit parts of the web programme as well as information, motivation and smoking questions (290 messages over six months). The control group received a PC‐based web intervention with interactive and multimedia features based on phases of quitting, the main difference to the intervention app being that it was not adapted for the small screen and did not include text messaging. Emails were sent as prompts if there were periods of inactivity.
Outcome
The included studies provided a range of abstinence outcome measures. Five studies (Cobos‐Campos 2017; Free 2009; Free 2011; Liao 2018; Peiris 2019), reported the strictest outcome as biochemically verified sustained/continuous abstinence, and Abroms 2014 and Alessi 2017 defined abstinence as biochemically confirmed repeat point prevalence at six months.
Seven additional studies reported self‐reported continuous abstinence at six months, without biochemical verification Baskerville 2018; BinDhim 2018; Borland 2013; Herbec 2019; Naughton 2014; Rodgers 2005; Whittaker 2011).
Two studies used self‐reported four‐week or 30‐day point prevalence abstinence at six‐month follow‐up (Abroms 2017; Haug 2013), three studies used self‐reported seven‐day point prevalence at six months (Augustson 2017; Bock 2013; Danaher 2019), one used self‐reported point prevalence abstinence at 32 weeks (Squiers 2017), one at 12 months (Yu 2017), and an additional four studies used six‐month biochemically verified measures of seven‐day point prevalence (Chan 2015; Ferguson 2015; Garrison 2018; Tseng 2017). Chan 2015 also provided biochemically verified seven‐day point prevalence abstinence rates at 12‐month follow‐up. Wilson 2016 reported six month follow‐up data in their trial registry entry; however they do not specify whether these rates were validated or not.
Risk of bias in included studies
The Characteristics of included studies table provides details of 'Risk of bias' judgements for each domain of each included study. Figure 2 illustrates judgements for each included study. Overall, we judged 13 studies to be at low risk of bias (judged at low risk for all domains), and three to be at high risk (judged to be at high risk in at least one domain). We judged the remaining studies to be at unclear risk (judged to be at unclear risk of bias for at least one domain, but with no judgements of high risk).
'Risk of bias' summary: review authors' judgements about each 'Risk of bias' item for each included study
Selection bias
The majority of studies (17 of 26) appeared to have adequate procedures for random sequence generation and allocation concealment, so we judged them to be at low risk of bias for these domians; however Alessi 2017; Augustson 2017; Chan 2015; Ferguson 2015; Garrison 2018; Haug 2013; Squiers 2017; Wilson 2016 and Yu 2017 did not provide sufficient description of either randomisation, concealment procedures, or both. Therefore, it is impossible to know whether the lack of information is due to actual bias or simply because it has not been reported, and we judged them to be at unclear risk of bias for at least random sequence generation or allocation concealment.
Detection bias
Blinding of participants is not possible in studies of behavioural interventions. In this case participants knew if they were receiving text messages or using an app. Therefore, we did not assess performance bias, and instead judged the likelihood of detection bias. We did not deem a study to be high risk for this domain where there was biochemical verification of abstinence, or where both arms received the same amount of face‐to‐face contact (or none).
In most cases, studies collected outcomes electronically and remotely (Abroms 2014; Augustson 2017; Baskerville 2018; BinDhim 2018; Bock 2013; Danaher 2019; Free 2009; Free 2011; Garrison 2018; Squiers 2017). Chan 2015; Herbec 2019; Liao 2018; Haug 2013 and Wilson 2016 all collected outcomes by phone, and Naughton 2014 by mailed questionnaire or in person. Cobos‐Campos 2017 collected outcomes in person in the clinic and this was not blinded, however this was mitigated by biochemical verification of quitting.
A number of the trials sought biochemical verification of long‐term abstinence with salivary cotinine (Abroms 2014; Free 2009; Free 2011), urinary cotinine (Liao 2018), or expired CO (Cobos‐Campos 2017; Garrison 2018). Chan 2015 assessed both CO and cotinine concentrations. Abroms 2017 biochemically validated their primary outcome at three months, but not at six months, and Rodgers 2005 validated abstinence at six weeks but not long‐term follow‐up. Similarly, Naughton 2014 used verification at four weeks only. Wilson 2016 stated that they planned to verify abstinence at all follow‐up points using salivary cotinine; however it is not stated whether the abstinence rates reported in their trial registry entry were the validated rates or not. However, as data was collected remotely this study was still deemed to be at low risk of bias for this domain. In fact, we deemed all studies to be at low risk of detection bias.
Attrition bias
We judged three studies to be at high risk of bias due to greater than 50% of participants lost to follow‐up at six months (Augustson 2017; Cobos‐Campos 2017; Herbec 2019). Several other studies had moderately high loss to follow‐up but the numbers were clearly reported. The difference between groups was not greater than 20%, and overall loss was not greater than 50%. Ferguson 2015 did not report loss to follow‐up and so we judged it to be at unclear risk of attrition bias.
Other
In Abroms 2014 there were some issues with fraudulent enrolment at the outset of the study, although this was corrected once detected. In Haug 2013, although clustering is adjusted for in this study's analysis the authors do not report the clustering effect, making it impossible to adjust for this in our analysis. Therefore, it is not clear how much the clustering adjustment influences the result from this study and our meta‐analyses. Rodgers 2005 suggested that some participants in their control group may have thought their incentive at follow‐up (a month of free text messaging) depended on reporting quitting. This could account for an unexpected increase in control group participants reporting quitting from six weeks (109 participants) to six months (202 participants reporting no smoking in the past seven days), which could have led to an underestimation of the effect of the intervention.
Effects of interventions
See: Summary of findings for the main comparison Text messaging compared to minimal support for smoking cessation; Summary of findings 2 Text messaging in addition to other smoking cessation support; Summary of findings 3 Smartphone app compared to lower‐intensity support for smoking cessation
Text messaging versus minimal smoking cessation support
We pooled those studies that compared a text messaging intervention with minimal smoking cessation support. This included 13 studies (Abroms 2014; Abroms 2017; Borland 2013; Chan 2015; Cobos‐Campos 2017; Ferguson 2015; Free 2009; Free 2011; Haug 2013; Liao 2018; Rodgers 2005; Whittaker 2011; Yu 2017). The analysis of all randomised participants, with those lost to follow‐up classified as smokers resulted in a RR of 1.54 (95% CI 1.19 to 2.00; I2 = 71%; 14,133 participants; Analysis 1.1) with minimal difference found in the result when we carried out a complete case analysis (RR 1.56, 95% CI 1.21 to 2.02; I2 = 72%; 11,969 participants; Analysis 1.2).
We conducted the following sensitivity analyses:
-
removing studies with very different populations from the main analysis (i.e. Analysis 1.1), pregnant women only in Abroms 2017 and postnatal families only in Yu 2017. This made very little difference to the overall result (RR 1.57, 95% CI 1.18 to 2.07; I2 = 68%; 13,408 participants);
-
removing the only cluster‐randomised trial, which we were unable to adjust for (Haug 2013). That again had minimal impact on the result (RR 1.57, 95% CI 1.19 to 2.07; I2 = 73%; 13,378 participants);
-
removing the only study judged to be at high risk of bias (Cobos‐Campos 2017), which again had minimal impact on the result (RR 1.49, 95% CI 1.13 to 1.96; I2 = 72%; 13,813 participants).
Text messaging versus other smoking cessation intervention
Only two studies (Borland 2013; Chan 2015; 2238 participants), compared text messaging with another smoking cessation intervention. When pooled these did not show a superior effect of either text message support to quit or the other forms of smoking cessation intervention in either an analysis including all randomised participants (RR 0.92, 95% CI 0.61 to 1.40; I2 = 27%; 2238 participants; Analysis 2.1) or a complete case analysis (RR 0.93, 95% CI 0.63 to 1.36; I2 = 20%; 1813 participants; Analysis 2.2).
Text messaging plus other smoking cessation support versus other smoking cessation support alone
Four studies (Bock 2013; Naughton 2014; Tseng 2017; Yu 2017; 997 participants), compared those who received both text messaging and another form of smoking cessation support with those only receiving the other form of smoking cessation support. The analysis of all randomised participants, assuming those lost to follow‐up were smoking, showed a benefit of adding the text messaging with RR of 1.59 (95% CI 1.09 to 2.33; I2 = 0%; 997 participants; Analysis 3.1). The result was comparable when we carried out a complete case analysis (RR 1.63; 1.12 to 2.37; I2 = 0%; 796 participants; Analysis 3.2).
We carried out a sensitivity analysis on Analysis 3.1 removing Yu 2017, as it had a substantially different population (postnatal families). The interpretation of the effect remained the same (RR 1.87, 95% CI 1.13 to 3.09; I2 = 0%; 769 participants).
High‐frequency versus low‐frequency text messaging
Three studies (Augustson 2017; Liao 2018; Squiers 2017; 12,985 participants), compared high‐frequency text messaging interventions with low‐frequency text messaging interventions. The pooled effect indicated no difference in cessation rates between groups in either the analysis of all participants randomised (RR 1.00, 95% CI 0.95 to 1.06; I2 = 0%; Analysis 4.1) or the complete case analysis (RR 1.04, 95% CI 1.00 to 1.09; I2 = 0%; 6798 participants; Analysis 4.2). A sensitivity analysis removing the one study judged to be at high risk of bias (Augustson 2017), led to no difference in the interpretation of the effect (RR 1.02, 95% CI 0.92 to 1.12; I2 = 0%; 4985 participants).
Smartphone app versus lower‐intensity smoking cessation support
We divided studies of smartphone apps according to the type of control. Two studies (Baskerville 2018; Peiris 2019; 1645 participants), compared a smartphone app with minimal non‐app smoking cessation support. There was no evidence of a favourable effect of smartphone apps in comparison with minimal non‐app smoking cessation support (RR 0.82, 95% CI 0.56 to 1.18; I2 = n/a as Peiris 2019 had no events; Analysis 5.1). Interpretation remained the same when we carried out a complete case analysis (RR 0.87, 95% CI 0.62 to 1.23; I2 = n/a; 771 participants; Analysis 5.2). Three studies (BinDhim 2018; Garrison 2018; Herbec 2019; 2175 participants), compared a smoking cessation smartphone app with a less intensive smoking cessation smartphone app. The analysis including all randomised participants resulted in an RR of 1.12 (95% CI 0.60 to 2.09; I2 = 68%; Analysis 5.1) with a very similar result in the complete case analysis (RR 1.18, 95% CI 0.67 to 2.09; I2 = 65%; 1003 participants). When we pooled all five studies, the resulting RR for all randomised participants was 1.00 (95% CI 0.66 to 1.52; I2 = 59%; 3079 participants; Analysis 5.1), providing no clear evidence of an increase in quit rates as a result of smart phone smoking cessation apps when compared to smoking cessation support of lower intensity. A sensitivity analysis removing the only study judged to be at high risk of bias (Herbec 2019), led to no difference in the interpretation of the effect (RR 1.10, 95% CI 0.60 to 2.00; I2 = 71%; 2654 participants).
Carbon monoxide monitoring + contingency management versus smoking cessation support
Neither of the studies that used mobile phones to monitor CO and provide contingency management provided evidence that these strategies were more effective than standard smoking cessation support.
Alessi 2017 compared messages prompting CO monitoring via video alone with the same CO monitoring plus reinforcement (with the chance to win prizes) for negative readings, and resulted in a RR of 0.88 (95% CI 0.35 to 2.21; 90 participants; Analysis 6.1).
Wilson 2016 compared CO monitoring and contingency management combined with smoking cessation telephone counselling and NRT, with the counselling and NRT alone, and resulted in an RR of 0.94 (95% CI 0.64 to 1.38; 310 participants; Analysis 6.1).
In both cases carrying out a complete case analysis resulted in a change in the direction of the effect estimate; however CIs still incorporated evidence of both considerable benefit and harm (Alessi 2017: RR 1.13, 95% CI 0.44 to 2.93; 81 participants; Wilson 2016: RR 1.06, 95% CI 0.74 to 1.54; 250 participants; Analysis 6.2).
Smartphone app + text messaging versus web‐based interventions
Danaher 2019 compared a smartphone app plus text messaging with a web‐based smoking cessation intervention and found evidence for a benefit of the app plus text messaging (RR 1.80, 95% CI 1.32 to 2.45; 1271 participants; Analysis 7.1). Complete case analysis resulted in a similar point estimate (RR 1.56, 95% CI 1.19 to 2.05; 463 participants; Analysis 7.2).
Discussion
Summary of main results
We found 26 randomised controlled trials of mobile phone smoking cessation interventions that met our inclusion criteria.
Whilst text messaging interventions tend to be very similar in design and content, the choice of control varied considerably. In this update, we separated out comparisons ensuring that only similar interventions and similar controls were pooled in meta‐analyses.
Our analyses found moderate‐certainty evidence (summary of findings Table for the main comparison), that text messaging interventions are more effective than minimal smoking cessation support (Abroms 2014; Abroms 2017; Borland 2013; Chan 2015; Cobos‐Campos 2017; Ferguson 2015; Free 2009; Free 2011; Haug 2013; Liao 2018; Rodgers 2005; Whittaker 2011; Yu 2017). Text messaging added to other smoking cessation interventions also appeared more effective than the other smoking cessation interventions alone (Bock 2013; Naughton 2014; Tseng 2017; Yu 2017; summary of findings Table 2).
However, when text messaging was compared with other smoking cessation interventions, the analysis did not find evidence that either the text messaging intervention or the other smoking cessation interventions resulted in superior quit rates. It is important to highlight that there were just two studies in this analysis and they each had slightly different contexts: Borland 2013 included people not seeking cessation support and participants were given 'suggestions about resources to use'; Chan 2015 was in the context of a Quit & Win contest.
We were also able to assess the effect of higher‐ versus lower‐intensity text messages on long‐term abstinence rates, using data pooled from three studies providing direct comparisons (Augustson 2017; Liao 2018; Squiers 2017). The frequency of messaging did differ somewhat between studies (e.g. Augustson 2017 used on average 15 versus 1 text message per week; Liao 2018 used 21 to 35 versus 3 to 5 messages per week; and Squiers 2017 used, on average, 16 versus 5 versus 1 text message per week), but overall, this analysis did not provide evidence that the intensity of the text messaging intervention impacted on abstinence rates. On average high intensity interventions resulted in abstinence rates of 26.6% versus 27.1% in low intensity interventions.
Studies of smartphone apps also included various control programmes. We found no evidence for a benefit of high intensity smartphone apps when compared with lower‐intensity smoking cessation apps (BinDhim 2018; Garrison 2018; Herbec 2019), or minimal non‐app smoking cessation support (Baskerville 2018; Peiris 2019), but we judged the evidence to be of very low certainty, meaning we have very little confidence in the effect estimate (summary of findings Table 3) .
Danaher 2019 was the only intervention that used both text messaging and a smartphone app and found that this combination resulted in higher quit rates than a web‐based smoking cessation intervention.
Overall completeness and applicability of evidence
Our review includes 26 studies with 33,849 participants. In comparison with previous reviews, there is now a much greater number of eligible studies, with increased sample sizes and including a greater diversity of settings and countries. We also found a large number of ongoing studies (n = 34), the results of which are likely to increase the diversity of contexts even further.
This is the first update of this review where there were randomised controlled trials of smartphone apps eligible to be included. In 2011, a review of available smoking cessation apps found them to be lacking in adherence to cessation guidelines or theory (Abroms 2011). In this review the included smartphone apps, although few in number, tended to be based on evidence or theory and were tested in high‐quality randomised controlled trials.
There has been criticism that smartphone apps may not be widely accessible to all, as they may rely on a certain degree of digital literacy and technology access that may not be widely dispersed in the population. It is important to note that in the included studies of smartphone apps there were reasonably high levels of education: 84% of participants in Garrison 2018 had greater than high school education; in BinDhim 2018, 53.7% had graduate level or higher education; in Baskerville 2018, 55.5% had post‐secondary education or higher; Danaher 2019 included 70% with a high school graduate education and higher; and in Herbec 2019, 68.7% had a post‐16 years qualification.
A common criticism of randomised controlled trials is that whilst they might provide evidence of effectiveness in a clinical trial setting, these data are not applicable to ‘real‐world’ settings. We are aware that many countries are implementing mCessation interventions and encourage routine monitoring and evaluation of these programmes, which will provide important ‘real‐world’ evidence for consideration alongside the research evidence.
Certainty of the evidence
There was moderate‐certainty evidence that text messaging increases quit rates by approximately 50% when compared to minimal support for smoking cessation (summary of findings Table for the main comparison). We downgraded the evidence by one level due to inconsistency as there was substantial unexplained statistical heterogeneity. This means the true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different.
There was also moderate‐certainty evidence that text messages increase quit rates by approximately 60% when tested as an addition to other smoking cessation support (summary of findings Table 2). We downgraded results by one level due to imprecision: there were fewer than 300 events overall, and confidence intervals encompassed minimal benefit and substantial benefit.
There was very low‐certainty evidence regarding the effect of smartphone apps compared to lower‐intensity support (summary of findings Table 3). This is due to inconsistency (considerable unexplained statistical heterogeneity) and very serious imprecision, with confidence intervals encompassing both clinically significant harm and clinically significant benefit.
Potential biases in the review process
A wide variability of control group programmes is potentially important in ensuring that the studies can provide the best information for decision makers who may want to compare mCessation with what already exists in their context. However, it could also lead to difficulties in the interpretation of the results. In some cases control groups received substantial smoking cessation support and the details of this were not always clear. This is supported by the fact that in some cases high quit rates, over what might have been expected, were observed in control groups (Rodgers 2005; Squiers 2017), with high rates in both the intervention and control groups in another study (Augustson 2017). This could indicate some degree of trial effect (everyone does better just through being involved in a research study), social desirability bias, or that minimal mobile phone interventions (just for reminders, prompts or data collection) may also be effective in producing behaviour change. High‐intensity control groups leading to high quit rates could have underestimated the relative effect of mobile phone interventions.
Though we searched trial registries, there remains a risk that there were eligible but unpublished studies we failed to identify. Reassuringly, a funnel plot (Figure 3), showed no evidence of asymmetry.
Funnel plot of comparison 1. Text messaging versus minimal smoking cessation support, outcome: 1.1 long‐term abstinence (all randomised))
Agreements and disagreements with other studies or reviews
This review agrees with other reviews of the benefits of text messaging to support healthy behaviour change (Armanasco 2017; Thakkar 2016; Scott‐Sheldon 2016). Several reviews have shown mixed results with respect to the effectiveness of smartphone apps for behaviour change, with significant issues relating to the size and quality of studies (Byambasuren 2018; Dirieto 2016; Lunde 2018; Schoeppe 2016; Zhao 2016).
Study flow diagram for this update
'Risk of bias' summary: review authors' judgements about each 'Risk of bias' item for each included study
Funnel plot of comparison 1. Text messaging versus minimal smoking cessation support, outcome: 1.1 long‐term abstinence (all randomised))
Comparison 1 Text messaging versus minimal smoking cessation support, Outcome 1 Long‐term abstinence (all randomised)).
Comparison 1 Text messaging versus minimal smoking cessation support, Outcome 2 Long‐term abstinence (complete case).
Comparison 2 Text messaging versus other smoking cessation intervention, Outcome 1 Long‐term abstinence (all randomised).
Comparison 2 Text messaging versus other smoking cessation intervention, Outcome 2 Long‐term abstinence (complete case).
Comparison 3 Text messaging + other smoking cessation support versus other smoking cessation support alone, Outcome 1 Long‐term abstinence (all randomised).
Comparison 3 Text messaging + other smoking cessation support versus other smoking cessation support alone, Outcome 2 Long‐term abstinence (complete case).
Comparison 4 High‐frequency versus low‐frequency text messaging, Outcome 1 Long‐term abstinence (all randomised).
Comparison 4 High‐frequency versus low‐frequency text messaging, Outcome 2 Long‐term abstinence (complete case).
Comparison 5 Smartphone app versus lower‐intensity smoking cessation support, Outcome 1 Long‐term abstinence (all randomised).
Comparison 5 Smartphone app versus lower‐intensity smoking cessation support, Outcome 2 Long‐term abstinence (complete case).
Comparison 6 CO monitoring + contingency management versus smoking cessation support, Outcome 1 Long‐term abstinence (all randomised).
Comparison 6 CO monitoring + contingency management versus smoking cessation support, Outcome 2 Long‐term abstinence (complete case).
Comparison 7 Smartphone app + text messaging versus web‐based intervention, Outcome 1 Long‐term abstinence (all randomised).
Comparison 7 Smartphone app + text messaging versus web‐based intervention, Outcome 2 Long‐term abstinence (complete case).
| Text messaging compared to minimal support for smoking cessation | ||||||
| Patient or population: people who smoke | ||||||
| Outcomes | Anticipated absolute effects* (95% CI) | Relative effect | № of participants | Certainty of the evidence | Comments | |
| Risk with minimal SC support | Risk with text messaging | |||||
| Long‐term abstinence (all randomised) Measured with self‐report and biochemical validation at 6 to 12 months | Study population | RR 1.54 | 14,133 | ⊕⊕⊕⊝ | ||
| 6 per 100 | 9 per 100 | |||||
| *The risk in the intervention group (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). | ||||||
| GRADE Working Group grades of evidence | ||||||
| aDowngraded one level due to inconsistency: substantial unexplained heterogeneity (I2 = 71%). | ||||||
| Text messaging in addition to other smoking cessation support compared to other smoking cessation support alone for smoking cessation | ||||||
| Patient or population: people who smoke | ||||||
| Outcomes | Anticipated absolute effects* (95% CI) | Relative effect | № of participants | Certainty of the evidence | Comments | |
| Risk with other SC support alone | Risk with text messaging + other SC support | |||||
| Long‐term abstinence (all randomised) Measured as self‐reported and biochemical validation at 6 to 12 months | Study population | RR 1.59 | 997 | ⊕⊕⊕⊝ | ||
| 8 per 100 | 12 per 100 | |||||
| *The risk in the intervention group (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). | ||||||
| GRADE Working Group grades of evidence | ||||||
| aDowngraded one level due to imprecision: fewer than 300 events overall. | ||||||
| Smartphone app compared to lower‐intensity support for smoking cessation | ||||||
| Patient or population: people who smoke Intervention: smartphone app | ||||||
| Outcomes | Anticipated absolute effects* (95% CI) | Relative effect | № of participants | Certainty of the evidence | Comments | |
| Risk with lower intensity SC support | Risk with Smartphone app | |||||
| Long‐term abstinence (all randomised) Measured with self‐report and biochemical validation at 6 months | Study population | RR 1.00 | 3079 | ⊕⊝⊝⊝ | ||
| 8 per 100 | 8 per 100 | |||||
| *The risk in the intervention group (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). | ||||||
| GRADE Working Group grades of evidence | ||||||
| aDowngraded one level due to inconsistency: considerable unexplained statistical heterogeneity (I2 = 59%). | ||||||
| Outcome or subgroup title | No. of studies | No. of participants | Statistical method | Effect size |
| 1 Long‐term abstinence (all randomised)) Show forest plot | 13 | 14133 | Risk Ratio (M‐H, Random, 95% CI) | 1.54 [1.19, 2.00] |
| 2 Long‐term abstinence (complete case) Show forest plot | 13 | 11969 | Risk Ratio (M‐H, Random, 95% CI) | 1.56 [1.21, 2.02] |
| Outcome or subgroup title | No. of studies | No. of participants | Statistical method | Effect size |
| 1 Long‐term abstinence (all randomised) Show forest plot | 2 | 2238 | Risk Ratio (M‐H, Random, 95% CI) | 0.92 [0.61, 1.40] |
| 2 Long‐term abstinence (complete case) Show forest plot | 2 | 1813 | Risk Ratio (M‐H, Random, 95% CI) | 0.93 [0.63, 1.36] |
| Outcome or subgroup title | No. of studies | No. of participants | Statistical method | Effect size |
| 1 Long‐term abstinence (all randomised) Show forest plot | 4 | 997 | Risk Ratio (M‐H, Random, 95% CI) | 1.59 [1.09, 2.33] |
| 2 Long‐term abstinence (complete case) Show forest plot | 4 | 796 | Risk Ratio (M‐H, Random, 95% CI) | 1.63 [1.12, 2.37] |
| Outcome or subgroup title | No. of studies | No. of participants | Statistical method | Effect size |
| 1 Long‐term abstinence (all randomised) Show forest plot | 3 | 12985 | Risk Ratio (M‐H, Random, 95% CI) | 1.00 [0.95, 1.06] |
| 2 Long‐term abstinence (complete case) Show forest plot | 3 | 6798 | Risk Ratio (M‐H, Random, 95% CI) | 1.04 [1.00, 1.09] |
| Outcome or subgroup title | No. of studies | No. of participants | Statistical method | Effect size |
| 1 Long‐term abstinence (all randomised) Show forest plot | 5 | 3079 | Risk Ratio (M‐H, Random, 95% CI) | 1.00 [0.66, 1.52] |
| 1.1 Comparison: minimal non‐app SC support | 2 | 1645 | Risk Ratio (M‐H, Random, 95% CI) | 0.82 [0.56, 1.18] |
| 1.2 Comparator: less intensive smartphone app | 3 | 1434 | Risk Ratio (M‐H, Random, 95% CI) | 1.12 [0.60, 2.09] |
| 2 Long‐term abstinence (complete case) Show forest plot | 5 | 1774 | Risk Ratio (M‐H, Random, 95% CI) | 1.06 [0.72, 1.54] |
| 2.1 Comparison: minimal non‐app SC support | 2 | 771 | Risk Ratio (M‐H, Random, 95% CI) | 0.87 [0.62, 1.23] |
| 2.2 Comparator: less intensive smartphone app | 3 | 1003 | Risk Ratio (M‐H, Random, 95% CI) | 1.18 [0.67, 2.09] |
| Outcome or subgroup title | No. of studies | No. of participants | Statistical method | Effect size |
| 1 Long‐term abstinence (all randomised) Show forest plot | 2 | Risk Ratio (M‐H, Random, 95% CI) | Totals not selected | |
| 1.1 Comparison: counselling, nicotine replacement therapy and monitoring only app | 1 | Risk Ratio (M‐H, Random, 95% CI) | 0.0 [0.0, 0.0] | |
| 1.2 Comparison: counselling and nicotine replacement therapy | 1 | Risk Ratio (M‐H, Random, 95% CI) | 0.0 [0.0, 0.0] | |
| 2 Long‐term abstinence (complete case) Show forest plot | 2 | Risk Ratio (M‐H, Random, 95% CI) | Totals not selected | |
| 2.1 Comparison: counselling, nicotine replacement therapy and monitoring only app | 1 | Risk Ratio (M‐H, Random, 95% CI) | 0.0 [0.0, 0.0] | |
| 2.2 Comparison: counselling and nicotine replacement therapy | 1 | Risk Ratio (M‐H, Random, 95% CI) | 0.0 [0.0, 0.0] | |
| Outcome or subgroup title | No. of studies | No. of participants | Statistical method | Effect size |
| 1 Long‐term abstinence (all randomised) Show forest plot | 1 | Risk Ratio (M‐H, Random, 95% CI) | Totals not selected | |
| 2 Long‐term abstinence (complete case) Show forest plot | 1 | Risk Ratio (M‐H, Random, 95% CI) | Totals not selected | |
