Smartphone Health Apps for Tobacco Cessation: A Systematic Review

Abstract

Background

Given the low retention and lack of persistent support by traditional tobacco cessation programs, evidence-based smartphone app-supported interventions can be an important tobacco control component. The objective of this systematic review was to identify and evaluate the types of studies that use smartphone apps for interventions in tobacco cessation.

Methods

We conducted a systematic review of PubMed (1946–2019), EMBASE (1974–2019), and PsycINFO (1806–2019) databases with keywords related to smartphone-supported tobacco cessation. Included articles were required to meet 3 baseline screening criteria: 1) be written in English, 2) include an abstract, and 3) be a full, peer-reviewed manuscript. The criteria for the second level of review were: 1) primary outcome of tobacco cessation, 2) intervention study, and 3) smartphone app as primary focus of study.

Results

Of 1973 eligible manuscripts, 18 met inclusion criteria. Most studies (n = 17) recruited adult participants (18+ years); one included teens (16+ years). Tobacco cessation was usually self-reported (n = 11), compared to biochemical verification (n = 3) or both (n = 4). There were 11 randomized controlled trials, 4 of which reported statistically significant results, and 7 single-arm trials that reported a mean abstinence rate of 33.9%.

Discussion

The majority of studies that use tobacco cessation apps as an intervention delivery modality are mostly at the pilot/feasibility stage. The growing field has resulted in studies that varied in methodologies, study design, and inclusion criteria. More consistency in intervention components and larger randomized controlled trials are needed for tobacco cessation smartphone apps.

1. INTRODUCTION

Tobacco continues to be the leading cause of preventable death in the United States (WHO, 2017). Within 10 years, smokers who successfully quit reduce their risk of death from lung cancer by half (Yamaguchi et al., 1991); within 15 years, risk of coronary heart disease is equal to that of a nonsmoker (World Health Organization, 2016). Although 30 million smokers contact a healthcare provider each year (Fiore et al., 2008), and more than half attempt to quit, most are not successful (Shiffman et al., 2008). After one year, the success rates of cessation were 5% when patients tried to quit without support, 16% with behavioral intervention, and 24% when the patient used pharmacological treatment in addition to behavioral support (Castaldelli-Maia et al., 2013; Khan et al., 2012; Laniado-Laborín, 2010). Limitations in traditional cessation interventions include low utilization (Fiore et al., 2008; Soon et al., 2014) and intermittent interactions between treatment experts (e.g., cessation counselors) and smokers (Martinez et al., 2018). It is also difficult to provide real-time responses to smoking urges and related cues, which are known to be important factors in relapse (Fiore et al, 2008).

Smartphones are ubiquitous in the United States with 77% of Americans owning a smartphone (Pew Research Center, 2018). Providing real-time help and information to patients can be achieved with the use of mobile health (mHealth) smartphone apps. While short message service (SMS), or text-messages, successfully provide support for cessation (Scott-Sheldon et al., 2016), apps afford the functionality (e.g., visual aids, interactive components, advanced multimedia) to develop more complex interventions based on health behavioral theories that might not be implementable in an SMS-only platform. Given that smartphone apps are able to reach and engage a larger population of individuals, it is necessary to identify evidence-based tobacco cessation apps, understand the theories that support the intervention components they deliver, and collect information on their effectiveness.

Therefore, the objective of this systematic review was to identify and evaluate the current evidence on smartphone apps used as an intervention for tobacco cessation. The most important factors considered were a primary outcome of tobacco cessation and the explicit use of a smartphone app as the source of intervention. Previous related reviews were content analyses that focused on apps that were commercially available in Android, Apple, Blackberry, and Windows online stores (Abroms et al., 2011; Hoeppner et al., 2016; Ubhi et al., 2016); they reported on the lack of scientific evidence for most commercially available apps, and all tobacco cessation outcomes were self-reported. Those reviews rated the apps based on strength of evidence, which showed that only one commercially available app was supported by high quality evidence. Rather than identify potential apps through commercial venues (e.g., Apple app store), this review searched exclusively through several scientific databases. Previous reviews are also out-of-date; given the speed of changes in mobile technologies, there have been many new advances, developed by both industry and academic institutions, since 2016. This review focused on all smartphone apps reported in the scientific literature, including those outside commercial app stores, and include more information on the effectiveness of each study.

2. METHODS

2.1. Design

The PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) guidelines were followed to report the results of this systematic review (see Appendix A). The protocol was registered with PROSPERO International Prospective Register of systematic reviews (CRD42018096567).

2.2. Data sources

Studies were located from PubMed/MEDLINE (1946–2019), EMBASE.com (1974–2019) and PsycINFO, Ovid® (1806–2019) using formulated search strategies developed by a medical librarian. Search strategies used appropriate and broad controlled vocabulary, phrases and keywords representative for the concepts of smartphones (e.g., smartphone, cellphone, iPhone), smoking (e.g., smoking, tobacco, nicotine), and smoking cessation (e.g., abstinence, quitting, intervention). Searches have no specific restrictions and include all languages and publication years (see Appendix B for search strategy details). Downloading database records, importing into EndNote, and de-duplicating records occurred on February 26, 2019. The web-based software DistillerSR aided in screening and data extraction.

2.3. Eligibility criteria

This systematic review proceeded through two levels of eligibility screening. First, articles were required to meet the following three baseline criteria: the articles 1) were written in English, 2) included text in the abstract, and 3) published as a full manuscript from a peer-reviewed journal. These requirements disqualified any dissertations, abstracts without manuscripts, book chapters, and other similar publications.

Articles then underwent a second level of review and had to meet the following criteria: the study must 1) have had tobacco cessation as a primary outcome, 2) was an intervention, and 3) primarily focused on a smartphone. For the purposes of this review, an app-based intervention was defined as using a smartphone application to support the user in stopping the use of tobacco. Studies that only used tablets (e.g., iPads) or websites accessed by smartphones—techniques that provide a different user experience (e.g., smartphone interventions might support outdoor locations where smokers feel the urge to have a cigarette, addressing a scenario that is not suitable for tablets)—were not considered. Given the ubiquity of smartphones, this systematic review excluded tablet apps in favor of smartphone apps that reach a more widespread audience.

2.4. Analysis

Before conducting this review, 5 researchers independently reviewed a preliminary group of 500 abstracts to establish a baseline understanding of the screening criteria. One adjudication meeting was held per level of review to better establish definitions and guidelines for including and excluding articles. If coders could not agree, a final determination was made by the lead author. Following adjudication, researchers began to independently review articles from the final dataset. An assessment of bias was also conducted for each study. Because there were a wide range of study designs as well as an expectation of small or underpowered studies, we applied several metrics from STROBE and CONSORT guidelines that could be considered for multiple types of studies (e.g., single and multiple arms). The data assessed included: 1) setting and location of study, 2) eligibility criteria, 3) recruitment source, 4) study size justification, 5) baseline participant demographics, 6) discussion of loss of participants, 7) discussion and justification on generalizability, and 8) funding source. Inter-rater reliability (Cohen’s Kappa) for the basic data extraction from two coders (CGEV and SJM) ranged from κ=0.78 to κ=1.0.

3. RESULTS

The searches retrieved 2835 database records containing 826 duplicate records, leaving 1973 records to screen (duplicates can occur more than once). In the initial screening, if the text was not written in English, did not have an abstract, or was not a full peer-reviewed manuscript, the record was excluded (n = 219). In the second screening, if tobacco cessation was not an outcome, the study was not an intervention, or if a smartphone app was not used, the record was excluded (n = 1733). Finally, after 21 manuscripts were selected for full text review, two additional manuscripts were deemed ineligible because they did not use an actual smartphone app (e.g., use of a smartphone to access a website); a third was disqualified because it was a secondary analysis of an app already included. A total of 18 manuscripts were included in the final review (Figure 1). There were two studies that used a different version of a similar app, but as the trial designs were also different, both studies were independently considered. Almost every study (n = 17) recruited adult participants (18+); 1 app included teens as young as 16 years old. Confirmation of cessation was usually self-reported (n = 11), compared to biochemical verification (n = 3) or both (n = 4).

Figure 1.

PRISMA Flow Diagram.

The length of study ranged from 1 to 12 months from enrollment to the final survey or assessment (mean=4.18), participants from 11 to 28,112 (mean=1953, median=235), and retention from 8% to 100% (mean=71.55%, median=81%). Participant eligibility criteria was highly variable in the inclusion criteria which comprised a mixture of smoking habits, location, smartphone ownership, gender, health status, and literacy. The behavioral theories applied in each study also showed heterogeneity, with 12 different theories or techniques applied (Table 1). Study designs split between randomized 2-arm (n = 11) or 1-arm (n = 7) trials. In the 2-arm trials, most had a different control condition: relapse training, counseling and medication, a different app, a text-messaging intervention, a self-help guide, experience training, brief advice, and a yoked condition (participants in the experimental arm were compensated specifically for low carbon monoxide (CO) readings, while participants in the control/yoked arm were compensated for any CO reading). In 3 of the 2-arm trials, the control arm was a reduced or downgraded version of the app (Table 2). Primary outcomes of point prevalence abstinence (PPA) for each study differed (e.g., 7-day PPA or 30-day PPA) based on study design as listed in Table 2. In the trials with a comparison group, 4 reported significantly higher rates of abstinence in the experimental app arm, 5 reported non-significant higher abstinence than control, and 2 reported no difference. In the 1-arm trials, the mean abstinence rate was 33.9%. In randomized-control trials, the mean abstinence rate was 22.9% for the app group. In total, the 18 studies reported an average of 6.1 out of the 8 categories that were assessed for bias (Table 3). The most common underreported categories were justification of sample size (39% of studies reported) and discussing loss of participants (39%).

Table 1.

Tobacco Cessation Smartphone App Studies, Overall Descriptions.

Author & Year	App Name	Theories or Techniques	Phone Platforms	Participants	Completed	% Retained	Smoking Status	Location	Other Conditions
BinDhim 2018	Smartphone Smoking Cessation Application (SSC App)	Ottawa decision support	Apple	684	583	85%	Daily smoker	US, UK, Australia, Singapore
Bricker 2014	SmartQuit	Acceptance and commitment therapy (ACT)	Apple + Android	340	196	58%	5 cigs/day >1 year willing to quit in next 30 days		English speaking
Bricker 2017	SmartQuit 2.0	Acceptance and commitment therapy (ACT)	Apple + Android	99	84	85%	5 cigs/day >1 year willing to quit in next 30 days		Not using other cessation
Buller 2014	REQ_mobile	Quit Coach theoretical framework	Android + Windows	102	68	67%	Smokers		English proficient
Businelle 2016	Smart-T	EMA/tailored messages	Android	61	59	97%	5 cigs/day current smokers willing to quit within 7 days		6th grade literacy, clinic visit once/week for 6 months
Gordon 2017	See me smoke-free	Guided imagery	Android	151	73	48%	Smoked in last 30 days	US	Female, English speaking
Hassandra 2017	Physical Activity over Smoking (PhoS)	Multiple	Android	44	34	77%	Regular smokers (10+ years) motivated to quit		No mental health conditions
Hertzberg 2013	Mobile Contingency Management (mCM)	Contingency management	Apple + Android	22	22	100%	10 cigs/day >1 year, no other tobacco products		PTSD, not pregnant or schizophrenic
Hicks 2017	Stay Quit Coach	Contingency management	Apple + Android	11	11	100%	10 cigs/day >1 year willing to quit	US	VA PTSD patients, speak/write English
Iacoviello 2017	Clickotine	Motivational interviewing	Apple	416	365	88%	5 cigs/day wants to quit	US	English speaking
Ubhi 2015	SmokeFree28 (SF28)	PRIME theory	Apple	1977	1170	59%	Smoked at time of registration, set quit date
Tombor 2019	SmokeFree Baby	Behavioral Change Theories	Apple + Android	565	247	44%	Weekly smoker		Pregnant, 18+, interested in cessation and set a quit date in the app
Masaki 2019	CureApp Smoking Cessation (CASC)	Cessation Counseling	Apple + Android	56	51	91%	Nicotine dependence patients
Marler 2019	Pivot	Multiple	Apple + Android	319	272	85%	5+ cigs/day current smokers		18–65 years old
Baskersville 2018	Crush the Crave	Contingency Management	Apple + Android	1599	725	45%	Smoker considering quitting in the next 30 days	Canada
Crane 2019	Smoke Free	Behavioral change techniques	Apple + Android	28112	2114	8%	Current smokers		App users, 18+, set one quit date
Krishnan 2019	Coach2Quit	Biomarker Feedback	Android	102	89	87%	Daily smoker	US	Willing to set a quit date within 2 weeks
Garrison 2020	Craving to Quit	Mindfulness	Apple + Android	505	325	64%	5+ cigs/day current smokers		18–65, less than 3 months past-year abstinence, 8 or great on the contemplation ladder and 4 or greater on the readiness to change questionnaire

Table 2.

Tobacco Cessation Smartphone App Studies, Study Designs, and Outcomes.

Author & Year	Study Design	Comparison or Control	Length	Abstinence Report	Primary Outcome	Difference with Control	Other Outcomes
BinDhim 2018	Automated double blind RCT	Downgraded app (no motivational messages)	6 months	Self-report	28.5% abstinent	Significant difference (p<.001)	Quit attempts, app feasibility
Bricker 2014	Double blind RCT pilot trial	QuitGuide	2 months	Self-report	13% abstinent	Significant difference (p<.001)	Receptivity to app, smoking reduction
Bricker 2017	Single arm pilot	NA	2 months	Self-report	21% 7-day point-prevalence abstinent (PPA)	NA	Receptivity to app, smoking reduction
Buller 2014	Randomized pre/post feasibility	onQ text messaging	3 months	Self-report	37% 7-day PPA	Significant difference (p=.04)	App usage
Businelle 2016	Nonrandomized feasibility	NA	12 weeks	Biochemical	20% abstinent	NA	Acceptability
Gordon 2017	Single arm pilot	NA	90 days	Self-report	47% 7-day PPA	NA	Weight, diet, exercise, app satisfaction and acceptability
Hassandra 2017	RCT feasibility	Relapse prevention training	6 months	Both	36% 7-day PPA	No significant difference	App engagement, physical activity
Hertzbeg 2013	Randomized feasibility	Yoked condition	3 months	Biochemical	50% abstinent	No significant difference
Hicks 2017	Two arm RCT feasibility	Counseling and medication	6 months	Both	40% 7-day PPA	100% abstinent in Combined Contact 60% abstinent in QUIT4EVER	Participant satisfaction, app compliance
Iacoviello 2017	Single arm feasibility	NA	8 weeks	Self-report	45% 7-day PPA	NA	App engagement, medical monitoring of other health events
Ubhi 2015	Single arm observational prospective cohort	NA	28 days	Self-report	19% 28-day PPA	NA	App usage
Tombor 2019	RCT	Different versions of the app	6 weeks	Self-report	N/A	No difference	App engagement
Masaki 2019	Single-arm	N/A	52 weeks	Both	58% continues abstinence rate at 52 weeks	NA	Withdrawal and craving symptoms
Marler 2019	Open-label single arm	N/A	18.5 weeks	Both	27.6% 30-day PPA	NA	Module completion rate, increased confidence to quit
Baskersville 2018	Parallel double-blind RCT	Evidence informed self-help guide OnRQ (usual care)	6 months	Self-report	14.4% 30-day PPA at six months	No significant difference	Frequency of use
Crane 2019	Two-arm exploratory RCT	Reduced version (no daily ‘missions’)	3 months	Self-report	19.3% abstinent	Significant difference (p<.001)	CPD, quit date designation
Krishnan 2019	RCT	Brief advice only	30 days	Self-report	3% 30 day PPA	No difference	Carbon monoxide levels, CPD
Garrison 2020	RCT	Experience training	6 months	Biochemical	11.1% 7 day PPA	No significant difference	Craving strength and frequency, mindfulness

Table 3.

Assessment of bias, including: ^aselection bias, ^breporting bias, ^cattrition bias.

Author & Year	Study settinga	Eligibility criteriaa	Source of recruitmenta	Justify sample size	Baseline data in resultsb	Discuss loss of participantsc	Discuss generalizability	Funding source
BinDhim 2018	✓	✓	✓	✓	✓		✓	✓
Bricker 2014	✓	✓	✓	✓	✓	✓		✓
Bricker 2017	✓	✓	✓		✓			✓
Buller 2014	✓	✓	✓		✓	✓	✓	✓
Businelle 2016	✓	✓	✓		✓	✓		✓
Gordon 2017	✓	✓	✓		✓	✓		✓
Hassandra 2017		✓			✓	✓	✓	✓
Hertzberg 2013		✓			✓		✓	✓
Hicks 2017		✓			✓			✓
Iacoviello 2017	✓	✓	✓		✓		✓	✓
Ubhi 2015	✓	✓	✓		✓			✓
Tombor 2019	✓	✓	✓	✓	✓		✓	✓
Masaki 2019	✓	✓	✓		✓		✓	✓
Marler 2019	✓	✓	✓	✓	✓	✓	✓	✓
Baskersville 2018	✓	✓	✓	✓	✓		✓	✓
Crane 2019	✓	✓	✓	✓	✓		✓	✓
Krishnan 2019	✓	✓	✓		✓		✓	✓
Garrison 2020	✓	✓	✓	✓	✓	✓	✓	✓

4. DISCUSSION

This systematic review identified 18 articles that focused on smartphone apps for delivering theory-driven tobacco cessation intervention. Overall, there was heterogeneity in the design of testing each app, for example, many of the 2-arm trials had a different control condition. Additionally, studies varied in length of study duration, participant eligibility criteria, and behavioral theories. Unfortunately, the amount of variability in the number of theories and models applied by the different studies prevented identifying patterns to understand if any particular model or design can be more effective for tobacco cessation. Similarly, given the amount of variation in protocols, measurements of outcome, and control groups, a meta-analysis was not feasible. However, this review identified new apps and research developments since the most recently available analyses (Hoeppner et al., 2016; Ubhi et al., 2016). Additionally, this review searched exclusively for studies available through scientific databases rather than commericial app stores, allowing us to provide more detailed information on developing projects.

An important challenge to conducting mHealth studies is participant attrition (Lane et al., 2015; O’Connor et al., 2015; Pfammatter et al., 2017). This might be explained in part by app user fatigue or limited resources; unfortunately, only 39% (n=7) studies included any discussion regarding participant retention (Table 3). Gordan (2017) suggested the pragmatic nature of their trial, with little to no contact with participants, contributed to the high attrition; Hassandra (2017) found that the most common reason for dropping out was relapse back to smoking; Garrison (2020) believed their study was able to achieve high retention with proactive methods that included obtaining contact information for three additional referrals. Overall, there was no pattern of app features, participant compensation, or other retention strategies that demonstrated consistently high retention. It would also be beneficial to disentangle the issue of attrition with user engagement, leaving an opportunity for future research. For example, mHealth studies in other fields have found that gamification and interactive components help improve engagement and retention (Muessig et al., 2013), which could be applied for tobacco cessation. Another potential effect of scope and resources is that the majority of studies (61%) relied on self-report as the source of the cessation outcome, even though biochemical validation is the gold-standard for traditional tobacco cessation studies. However, mHealth studies are inherently based on remote use and/or monitoring. It is possible that participants could be more truthful reporting the amount they smoke to an app rather than a person that could be perceived as judgmental. Further studies are needed to understand whether the same reporting standards are necessary.

Single-arm trials had a mean abstinence rate of 33.9%, showing promise compared to rates of existing recommended therapies; as a frame of reference, 24% of patients successfully quit smoking when using pharmacological treatment in addition to behavioral support (Laniado-Laborín, 2010). Of the 11 trials with a control group, only 4 demonstrated significant improvement in intervention group over the control group. Only the Bindhim study (BinDhim, McGeechan, & Trevena, 2018) was both sufficiently powered and detected a significant difference in cessation outcomes (80% power at alpha=0.05).

Most of the studies underreported information or data that could lead to potential biases (Table 3). As most of the studies were limited in sample size and power, acknowledgement and discussion of sample sizes and retention rates would be beneficial to inform future trials and larger studies. Overall, a stronger adherence to reporting guidelines such as CONSORT or STROBE could help smaller pilot studies have a larger impact.

Only 10 studies (56%) had developed versions of their app for both Android and Apple, limiting their potential reach. Researchers likely constrained by time and budget focused their development time on a single platform. These design decisions affect their reach because only releasing the app on a single major platform limits the generalizability of the study to only users of that platform (StatCounter Global Stats, 2018). Additionally, the average price of an iPhone is more than double an average Android phone (The Washington Post, 2018), potentially limiting the reach of different apps based on socioeconomics.

Although traditional tobacco cessation studies often report outcomes at 12 months (Lemmens et al., 2008), the study periods in this review had a mean of 4.18 months, with only a single study reaching 12 months; this is likely due to the nature of pilot studies and limited resources. While there was a variation in participant eligibility criteria, we did not find studies focused on specific groups. Research has identified important disparities in tobacco products use among racial, ethnic, and sexual and gender minorities (Bränström and Pachankis, 2018; Drope et al., 2018; Margerison-Zilko and Cubbin, 2013; McQuoid et al., 2018). Future mHealth research should have a focus on studying feasibility, acceptability, and effectiveness of apps for all tobacco users, with an awareness for design features that are tailored and culturally appropriate for minority groups.

5. LIMITATIONS

This work has several limitations. We included studies with different methods and settings; there was also no scoring or review of the methodological quality of the included studies. However, these differences were explicitly reported, and a meta-analysis was not conducted due to the heterogeneity in study designs. The studies chosen for review were limited by the selected search terms (Appendix B). We also did not include studies that were not published, possibly due to null findings, raising the risk of publication bias. Despite these limitations, this systematic review was able to identify and summarize the peer-review literature of smartphone apps for delivering theory-driven tobacco cessation interventions. Using DistillerSR also provided rigor and consistency to the review. These findings are important for the future of research in this field. There is a need for more consistency in intervention components, which could be supported by a registry or consistent set of guidelines for mHealth tobacco trials. Larger, randomized controlled trials are also needed to better elucidate how smartphone apps can be effective for tobacco cessation.

6. CONCLUSIONS

Studies have consistently shown a significant decrease in cancer and heart disease mortality for people who quit tobacco use (World Health Organization; 2016; Yamaguchi et al., 1991). However, given the low retention and lack of consistent support in traditional evidence-based treatment programs (Fiore et al., 2008; Martinez et al., 2018; Soon et al., 2014), theory-driven app-supported interventions serve as a promising component in tobacco control strategies. Over the past several years, tobacco researchers have leveraged the engagement of smartphone users to develop and deliver app-supported cessation efforts. This study shows that more effort is needed to design studies that can be generalizable as well as better retain participants. Mobile apps provide the real-time, persistent, and cost-effective ability to support tobacco cessation and cancer prevention. If effective, they can reduce various negative health outcomes by providing prevention practitioners with a valuable tool to reduce smoking.

Highlights.

• Identified 18 studies using apps for evidence-based tobacco cessation intervention

• Most interventions using tobacco cessation apps were pilot/feasibility studies

• Studies had varied methodology, study design, inclusion criteria

• More effort is needed to design studies that can be generalizable as well as retain participants

• Consistent app interventions and larger randomized controlled trials needed