Study design

A systematic review of effectiveness, being regarded as the gold standard for identifying evidence-based practice¹⁶, was best suited for this review. Given that the review also aimed to identify the critical features of the successful interventions, it also included an intervention component-level analysis (ICA) as outlined by Sutcliffe et al.¹⁷. The ICA by Sutcliffe et al.¹⁷ is a pragmatic but formal and rigorous approach for analyzing the characteristics of the studied interventions, which may be associated with successful outcomes. This method has been particularly recommended when the studies’ interventions differ from one another, which limits the ability to explore meaningful numbers of mediators and moderators of intervention effect through other formal methods of analysis and synthesis¹⁷. In view of the significant diversity in the interventions utilized by the study authors (as remarked below), the ICA by Sutcliffe et al.¹⁷ was found to best suited for this review to provide more information about the critical features of the successful interventions for providing practice recommendations. This review is presented in compliance with the Preferred Reporting Items for Systematic Reviews and Meta-Analysis PRISMA statement (Supplementary file Table 1).

Inclusion and exclusion criteria

The studies included in this review had to assess the effectiveness of intensive stand-alone smoking cessation interventions by comparing them to a less intensive intervention, such as brief tobacco cessation advice or usual care, specifically amongst individuals with diabetes. Intensive stand-alone interventions included pharmacological and/or non-pharmacological behavioral interventions for smoking cessation, which were not part of broader interventions for improving diabetes management, consisting of multiple and/or long (>20 min) smoking cessation support sessions). Studies in which the experimental smoking cessation intervention was part of a more extensive intervention for diabetes management, such as a lifestyle management intervention for improving diabetes, were thus considered ineligible. The studies included in this review had to include individuals living with (diagnosed) diabetes mellitus as their study population. Studies in which the participants had pre-diabetes or gestational diabetes were deemed ineligible. Furthermore, studies in which only a proportion of the participants had diabetes or reports of studies which were not specific to individuals with diabetes were also excluded. Only published articles, or unpublished reports, of randomized controlled trials were considered in this review. Non-randomized trials were deemed ineligible as these tend to produce higher effect estimates of the studied intervention when compared to randomized trials¹⁸. Systematic reviews were also not included. This is because none of the identified reviews^13-15 provided sufficient detail on the critical features of the successful interventions. No language or time limiters were set. The minimum requirement for non-English trials was that the title and/or abstract had to be available in English within the identified (below) bibliographic databases.

Search strategy

The search was carried out on the 28 May 2022 from inception using the following electronic literature databases: APA PsycInfo, CINAHL Complete, Cochrane Central Register of Controlled Trials, Cochrane Clinical Answers, Cochrane Database of Systematic Reviews, Cochrane Methodology Register, MEDLINE Complete, ProQuest Dissertations & Theses A&I, Public Health Database, PubMed, Scopus, System for Information on Grey Literature in Europe, and all the databases on Web of Science. Based on the review objective, the main keywords, ‘diabetes mellitus’ and ‘smoking cessation’ and their synonyms (Supplementary file Table 2), were combined using the Boolean operators ‘AND’ and ‘OR’, and searched in titles, abstracts, and subject headings/medical subject headings accordingly. The search strategy used for searching in Web of Science is outlined in Supplementary file Table 3.

Study selection

After carrying out the search, the identified records were collated on Mendeley® for de-duplication. The remaining records were screened by reading titles and abstracts. Potentially relevant articles were then read and assessed for eligibility basing decisions on the inclusion and exclusion criteria. The reference lists of these studies and those of the identified reviews^13-15 were also examined for the identification of other possible suitable studies for inclusion in this systematic review.

Data extraction

The following information was extracted from all the identified studies: authors; year of study; location; study duration; detailed information on the experimental smoking cessation intervention, its components and the control intervention; study sample; percentage followed up; smoking cessation outcome, time-points and reporting methods; information about the researchers’ reflections and accounts of their experience in evaluating the intervention for conducting the ICA as per Sutcliffe et al.¹⁷; and other relevant observations. Furthermore, the information required to assess risk-of-bias (as outlined in the Cochrane risk-of-bias tool, RoB 2)¹⁹ for each of the identified studies, was also extracted.

Quality assessment

In this review, the RoB 2 for randomized parallel-group trials and the RoB 2 tool specific to cluster-randomized trials were utilized. Studies found to be at high risk-of-bias were not excluded from the review or from the analysis. As recommended by Higgins et al.²⁰, the information obtained in carrying out the risk-of-bias assessment was presented as part of the review’s findings and was also considered in the analysis and conclusions of the review. The robvis tool²¹ was used to visualize the risk-of-bias assessments.

Synthesis of results

A meta-analysis of effect estimates was initially the preferred method of synthesis. However, given the significant diversity in the interventions (both experimental and control interventions) utilized by the study authors, and the incomplete reporting of outcomes/effect estimates in some of the identified studies^22-25, this was not recommended^18,26. Thus, following data extraction and charting of data in table format, a narrative approach using vote counting based on the direction of effect¹⁸, was adopted to analyze the results. In essence, the studies showing a statistically significant increase in smoking abstinence were compared to the studies which did not. Additionally, a narrative component analysis, in which the interventions’ critical features were identified by following the method outlined by Sutcliffe et al.¹⁷, followed. This included the mapping of the characteristics of the interventions, taking into consideration the effectiveness of the interventions for establishing the components which appeared to be of significance, and the coding of informal data on the researchers’ reflections in evaluating the intervention (using inductive thematic analysis), to help understand the association between the identified characteristics and the studies outcomes.

Search results

The PRISMA 2020 flow diagram²⁷ was utilized to outline the selection process, providing details on the exclusion reasons at the full-text level of screening (Figure 1). A total of 15442 records were retrieved of which 6007 were found to be duplicates. After removing duplicates, 9365 were screened by reading titles and abstracts. A total of 135 reports were found to be possibly relevant and were screened at full-text level. On matching these publications to the inclusion criteria, 122 reports were found to be ineligible as they did not report on the effectiveness of stand-alone smoking cessation interventions for individuals with diabetes from randomized controlled trials. Conversely, 13 reports were deemed eligible. An additional six reports, which were obtained from citation searching, were assessed for eligibility. Two reports were found to be eligible and were included. This led to a final selection of 15 reports.

Figure 1

PRISMA flow diagram

Some of the identified articles reported the same study, resulting in a smaller number of studies. Both Lam et al.²⁸ and Thankappan et al.²⁹ published the findings from the randomized control trials by Li et al.³⁰ and Thankappan et al.³¹, respectively, in conference proceedings. Furthermore, in the publications by Thankappan et al.²⁵ and Nichter et al.²⁴, the authors followed up participants from the Thankappan et al.³¹ trial for a total of one and two years, respectively, without providing any additional interventions. Additionally, Mini et al.²³ reported on the biochemical verification of a sample of non-smokers who participated in the Thankappan et al.³¹ trail at follow-up at 1 year, in their publication. Thus, the total number of studies included was 10.

Characteristics of the selected studies and relevant findings

The characteristics and the relevant findings of the identified studies are outlined in Table 1. Except for the study by Pérez-Tortosa et al.³², who reported the findings from a cluster randomized parallel-group trial, all the remaining publications reported the findings from individually randomized parallel-group trials. All studies were published in journals except for the study by Albaroodi et al.³³ which was available as a preprint.

Most studies included individuals with both type I and type II diabetes as study participants^33-36, who were mostly men^30-33,35-38, and in their fifties^{30-32,35,37,38}. Sample sizes varied across the studies; from n=34²² to n=948³²; however, only three studies^30,32,35 reported a priori power calculations to detect a significance in smoking cessation outcome.

Most authors based their intervention on the 5As algorithm (Ask, Advise, Assess, Assist and Arrange)^30,31,33,38, followed by motivational interviewing^32,37; however, the frequency and the number of sessions provided, and their duration greatly varied across the studies. In two studies^30,32, the authors also took into consideration the participants’ current stage of behavior as per the trans-theoretical model of change³⁹ when delivering the intervention. Additionally in four of the identified studies^32,35-37, NRT or buproprion were provided depending on the set eligibility criteria. Information material, such as general or diabetes-specific smoking cessation leaflets were also provided in four studies^30,31,34,38, while in the studies of Fowler et al.²², Ng et al.³⁸ and Thankappan et al.³¹, visual aids of diabetes-related complications were also utilized. The identified interventions were mostly delivered by nurses^{30,32,33,35,37}, and doctors^31-34,38, and delivered in diabetes centers/clinics^{22,30,31,33,36-38}.

As seen in Table 2, in most studies the authors assessed the impact of the studied intervention on smoking cessation for up to 6 months. Smoking abstinence was usually defined as a 7-day point prevalence abstinence^{25,30,31,37,38}, and biochemically verified (at the end of the study) by measuring exhaled carbon monoxide^30,32-34, and/or cotinine in saliva, urine or blood plasma^22,30,34-37.

Assessment for risk-of-bias

A risk-of-bias assessment was carried out on the endpoint reported smoking cessation outcome of each publication. Risk-of-bias assessments were carried out by following the guide by Higgins et al.²⁰. The assessments using RoB 2 are shown in Figure 2, while the risk-of-bias assessment using the RoB 2 tool for cluster-randomized trials is shown in Figure 3. Given that all studies were judged to be of concern of risk-of-bias in at least one domain, the overall risk-of-bias of most studies was also judged to be of some concern. Conversely, the overall risk-of-bias of the reports by Ng et al.³⁸, Thankappan et al.^25,31 and Nichter et al.²⁴ was judged ‘high’, as these were found to be at high risk-of-bias in the measurement of the smoking cessation outcome, being based on self-reported data.

Figure 2

Summary of risk-of-bias assessments performed using the RoB 2 for randomized parallel-group trials

Figure 3

Summary of risk-of-bias assessment performed using the RoB 2 for cluster-randomized trials

Narrative analysis of the findings of the studies

As outlined in Table 1, significant differences in smoking cessation between the intervention and control group were only reported in the studies by Canga et al.³⁵ and Thankappan et al.³¹ at follow-up at 6 months, and in the studies by Pérez-Tortosa et al.³² and Thankappan et al.²⁵ at follow-up at 1 year. Conversely, the other study authors^{22,24,30,33,34,36-38} did not report a significant improvement in the smoking cessation rate of the intervention group when compared to the control group; and although Ng et al.³⁸ reported a significant decrease in the self-reported smoking prevalence in both groups at follow-up at 6 months, in these studies the smoking abstinence rate in both the intervention and control group was also deemed to be insignificant. Given these inconsistent findings, and taking into consideration the high level of heterogeneity amongst the interventions of the studies, a narrative component analysis was conducted.

Narrative component analysis

Guided by Sutcliffe et al.¹⁷, the main features of the interventions were mapped out at the following levels: the framework/behavioral support method on which the intervention was based; its intensity, the length of each session and the number of sessions provided; the provision of additional information about tobacco associated diabetic complications, using visual aids or leaflets; the provision of pharmacotherapy; and the provision of general stop-smoking information material. Given that in most studies^22,31-34,38 two or more health professionals were involved in delivering the intervention, the interventions of the studies were not mapped at the provider level. However, the health professionals involved in providing the successful features of the interventions were then identified. Table 2 maps out the characteristics of the interventions of each study based on these categories.

It appears that intensive smoking cessation interventions may enhance smoking cessation success. Canga et al.³⁵, Thankappan et al.³¹ and Pérez-Tortosa et al.³², who provided 3–4 sessions of duration >20 min for their study participants, found that more smokers in the intervention group quit smoking when compared to those in the control group. Conversely, in the other studies, in which a non-significant smoking cessation outcome was reported, the experimental intervention was brief ≤20 min^30,33,37 or consisted of only one or two lengthier sessions in total^34,38. While for the studies of Fowler et al.²² and Sawicki et al.³⁶, whose experimental intervention was of an intensive nature, reported a non-significant outcome, it is worth noting that few participants adhered to the study protocol, possibly undermining the intervention’s effectiveness. Trained nurses^32,35, doctors³² and counsellors/diabetes educators³¹ provided the lengthier sessions as part of the successful smoking cessation interventions.

The provision of frequent smoking cessation support also seems to have been beneficial. In the studies of Canga et al.³⁵ and Pérez-Tortosa et al.³², smokers who were ready to quit were provided with frequent smoking cessation support (follow-up appointments at 1 to 2 weeks). Given that the frequency of the sessions varied across the identified studies, it is, however, difficult to establish the ideal total duration of the studied interventions in terms of months.

There is, however, not enough evidence to recommend the use of a specific framework/method for smoking cessation. While both Ng et al.³⁸ and Thankappan et al.³¹, who based their intervention on the 5As framework, reported a significant decrease in smoking prevalence in their studies, Albaroodi et al.³³ and Li et al.³⁰, who also utilized the same framework, did not. Similarly, while Pérez-Tortosa et al.³², whose intervention was based on motivational interviewing (MI), reported significant findings, Hokanson et al.³⁷, who utilized the same technique, did not. There is also not enough evidence to recommend the tailoring of interventions based on the participants’ stage of change, as unlike Pérez-Tortosa et al.³², Li et al.³⁰, who also based their intervention according to the participants’ stage of change, did not report a significant smoking cessation outcome.

It is not clear whether the addition of pharmacotherapy to behavioral support helped increase smoking cessation success. While both Pérez-Tortosa et al.³² and Canga et al.³⁵, who provided pharmacotherapy for smoking cessation to those assigned to the intervention group observed a significant smoking cessation outcome, they did not report the smoking cessation rate of those who used it. It is also worth noting that in the study of Canga et al.³⁵, only 25 out of 105 participants utilized the provided NRT. On the other hand, it is even more difficult to ascertain the effect of the provision of pharmacotherapy on smoking cessation in the Hokanson et al.³⁷ and Sawicki et al.³⁶ studies. This is because in both studies the participants in the intervention and control groups made use of such treatment.

On the other hand, the use of visual aids of diabetes-related complications, may have been useful in supporting smoking cessation. When considering that in the study of Fowler et al.²² few participants adhered to the study protocol, possibly undermining the intervention’s effectiveness, in both Ng et al.³⁸ and Thankappan et al.³¹, who also used visual aids of diabetes-related complications, a significant decrease in smoking prevalence was reported. Nonetheless, the use of diabetes-specific leaflets, in which information on diabetes-related complications was also conveyed, is not clear, as unlike Ng et al.³⁸ and Thankappan et al.³¹, Li et al.³⁰ did not report a significant smoking cessation outcome.

There is not enough evidence to suggest the use of general stop-smoking leaflets. Unlike the Canga et al.³⁵ study, in both Ardron et al.³⁴ and Li et al.³⁰ (the latter of which provided leaflets to both the intervention and control group), non-significant findings were reported.

Analysis of informal evidence

As part of the ICA outlined by Sutcliffe et al.¹⁷, the researchers’ reflections, and accounts of their experience in evaluating the intervention, were coded to help understand the association between the identified intervention features and the success or failure of the interventions. Two major themes were identified: ‘intensive smoking cessation support’, and ‘strong warning messages on tobacco associated diabetic complications’, both of which were identified as being associated with smoking cessation success. Table 3 outlines these themes, providing examples of the underlying evidence, the number of studies contributing to these themes, and a brief explanation of the association between these themes and the studies’ outcomes.

Strengths and limitations

This review builds on the findings of the scoping review by Grech et al.¹⁵, assessing the effectiveness of intensive stand-alone smoking cessation interventions amongst individuals with diabetes. The strength of this review was that it comprised a systematic search of randomized trials of stand-alone smoking cessation interventions for individuals with diabetes, utilizing a wide range of databases. This review does not include trials in which smoking cessation was a part of a more extensive intervention for diabetes management. Furthermore, studies in which only a proportion of the participants had diabetes or reports of studies which were not specific to individuals with diabetes, were also excluded. While this limited the number of trials to be reviewed, it allowed us to specifically measure the effect of stand-alone smoking cessation interventions which were specifically designed for and delivered to individuals with diabetes.

While a meta-analysis of effect estimates was the initial preferred method of synthesis, given the significant diversity in the interventions utilized by the study authors, and the incomplete reporting of outcomes/effect estimates in some of the identified studies^22-25, this was not recommended. Nonetheless, the addition of an ICA to the systematic review proved insightful, as in analyzing the components of the identified diverse interventions some critical features of the successful interventions were identified. Furthermore, recommendations for further research were also provided.

Despite the utility of the ICA to this systematic review, it is worth noting that some of the findings of the studies were found to be possibly at risk-of-bias. Thus, further research to establish the validity of these findings is recommended. Sutcliffe et al.¹⁷ suggest carrying out qualitative research to explore the views and experiences of recipients and providers of the identified features. Apart from establishing the validity of the obtained findings¹⁷, in carrying out qualitative research with such stakeholders, the need for other smoking cessation intervention characteristics (specific to individuals with diabetes), may be identified. Given that healthcare interventions are very much dependent on patient involvement and their attitudes to them⁴⁹, the exploration of the views of individuals with diabetes on the identified promising smoking cessation components and their perceived needs to quit smoking, is thus recommended.