Evaluating interventions to enhance public awareness of cancer symptoms: Protocol for a systematic review and network meta-analysis

Background

Globally, there are approximately 20 million new cases and nearly 10 million deaths due to cancer each year¹. The GLOBOCAN database highlights that in countries with a very high Human Development Index (HDI), the reported cancer mortality rates are 118.3 per 100,000 for men and 78.5 per 100,000 for women, figures that are higher than in countries with a lower HDI¹. Moreover, cancer is the second-leading cause of premature death in 127 countries and is expected to overtake cardiovascular disease as the leading cause of death within this century^2,3.

Cancer detected at an earlier stage is associated with improved outcomes, including reduced mortality^4,5, lower morbidity⁴, enhanced patient quality of life⁴, and reduced cost^6–10. Interventions to promote earlier detection of cancer can be classified into two broad categories: interventions to detect cancer in an asymptomatic population—(i.e. “cancer screening”) and interventions to shorten the time from symptom onset to diagnosis (often referred to as “early diagnosis” initiatives)¹¹.

While cancer screening attempts to detect cancer before symptoms develop, only a minority of cancers are detected through screening, despite the presence of multiple screening programmes in breast, cervical, colorectal and most recently lung cancer^12–14. In Ireland, between 2017–2019, 25% of breast, 32% of cervical and 6% colorectal diagnoses in Ireland were screen-detected, which equates to approximately 5% of overall invasive cancer cases⁵. Since the route to diagnosis will start with a symptom in over 90% of patients, interventions to promote the early detection of symptomatic cancer are of paramount importance.

Early detection initiatives, which aim to reduce the time from symptom-onset to diagnosis, can be subdivided into those that raise symptom awareness in the public, those that alter clinician practice, those that focus on tackling any of several barriers to early help-seeking (including symptom awareness, fear, stigma, negative beliefs about cancer), and those that optimise clinical care pathways. Initiatives focused on raising symptom awareness encourage individuals to recognise potential signs of cancer and seek medical advice promptly¹⁵. Efforts targeting healthcare professionals, particularly general practitioners, aim to enhance their awareness of cancer symptoms and facilitate timely referrals. Lastly, the effective design of clinical care pathways, which may include intermediate testing or referral to secondary care, streamlines the process from initial suspicion to diagnosis, ensuring that patients are seen by specialists within an appropriate timeframe. Each of these broad intervention categories aims to reduce unnecessary delays and improve outcomes via earlier diagnosis and treatment initiation¹⁶.

Existing literature

The existing literature on increasing cancer symptom awareness in the general public is extensive and comprises a vast array of approaches¹⁷. These approaches range from large-scale national campaigns to more targeted efforts focused on specific communities or opportunistic interventions during healthcare visits. At the national level, strategies include mass media campaigns, public service advertisements, advertisements¹⁸ and large-scale educational initiatives¹⁹. Community-level efforts often involve distributing informational materials such as booklets²⁰, decision aids^6,21, hosting local workshops, or collaborating with community leaders to promote awareness. At the individual level, interventions may include opportunistic conversations during healthcare visits, telephone consults^22,23, or personalized decision aids.

Many studies have reported positive findings that these interventions increase cancer awareness^{15,17,24–26}. However, raising awareness does not consistently lead to behaviour change²⁷. Help-seeking behaviour is influenced by a range of other factors beyond awareness, such as socioeconomic limitations, concerns about burdening health providers, fears about symptoms not being taken seriously, difficulty getting an appointment, and anxiety about potential diagnoses^3,28,29. Thus, while awareness is a crucial first step, it alone may not be sufficient to prompt individuals to seek medical attention.

Even when behaviour does change, it may not always lead to earlier diagnosis. Awareness campaigns such as "Be Clear on Cancer" (BCoC) in the UK successfully increased primary care visits and urgent referrals for suspected cancer, as well as influenced diagnosis patterns, including incidence and stage at diagnosis^28,30. However, changing help-seeking behaviour and increasing referrals do not automatically translate into a higher rate of early diagnoses, as factors like access to diagnostic services and healthcare provider response also play a role^31,32.

Furthermore, even when earlier diagnosis is achieved, it does not necessarily result in improved survival outcomes. For instance, despite the changes in diagnosis patterns following the BCoC campaigns, no clear improvements in survival rates were observed³⁰. This could be due to various reasons, such as the complexity of cancer progression, the time lag required to observe survival benefits, or the limitations of survival as a metric for campaign effectiveness³⁰. Some studies, such as an ecological comparison across geographical regions, have found an association between awareness and survival, but such findings are context-dependent and not directly tied to awareness campaigns³³.

Knowledge gap, rationale, and aim

If public awareness interventions drive timely help-seeking earlier diagnosis and improve cancer outcomes for symptomatic patients, health services could justify implementing such campaigns at a national level. However, the incumbent systematic review in this area was conducted in 2009 and highlighted the scarcity of studies rigorously evaluating the impacts of such campaigns¹⁵. Although a more recent systematic review focusing on lung cancer awareness and help-seeking was published in 2021²⁶ the aim of this review is to provide an updated synthesis of evidence on the effectiveness of interventions designed to increase public awareness and encourage help-seeking for earlier cancer diagnosis in general and improved outcomes.

Objectives

We will achieve this aim through the completion of the following specific objectives:

1.  To systematically identify relevant studies in the literature;

2.  To categorise the identified studies according to awareness-raising modality and, where available, the underpinning theoretical framework of behavioural change;

3.  To collate the outcome data from each study;

4.  To develop a schema for classifying the reported outcomes, utilising or modifying an existing schema, where possible;

5.  To publish a bibliometric analysis of the identified studies;

6.  To compare between interventions, where possible, by examining cancer mortality and cancer stage at diagnosis between the intervention and comparison groups;

7.  To produce a pooled estimate of relevant outcomes for comparable interventions, and if possible, to produce a network meta-analysis enabling comparison of multiple interventions.

Eligibility criteria

The following eligibility criteria will be used to operationalise our research question: Table 1: Eligibility Criteria.

Search strategy

A comprehensive search will be conducted in the following databases MEDLINE, EMBASE, PsycINFO, Scopus, Web of Science, ProQuest, Cochrane Library and Cumulative Index to Nursing and Allied Health Literature (CINAHL). The search strategy will include MeSH terms and keywords related to: (1) cancer awareness, (2) early presentation, (3) interventions. These are demonstrated in Table 2. The approach to this search will be customized to meet the unique specifications of each database, using filters to locate primary studies as needed. The search will be limited to studies published in English from November 2008 to April 22, 2024. References of included studies and relevant reviews will be hand searched to identify additional studies not captured by electronic searches. Full details of the search strategy, designed in consultation with an information specialist, will be outlined on the Open Science Framework.

Study selection

Screening of titles and abstracts for eligibility will be done by two reviewers independently, within “Rayyan”, an online tool for collaborative systematic literature reviews³⁸. The screening algorithm is outlined in Table 2. Discrepancies will be resolved through consensus or with an additional reviewer (B.J.). Full texts of potentially eligible studies will then be reviewed for inclusion based on the eligibility criteria.

Data collection

Relevant data items will be extracted from the full text of included articles using a pre-piloted pro forma. The extraction will be conducted independently by one reviewer and reviewed by another for accuracy. Data items to be extracted are outlined in Table 3.

Intervention Modality. Interventions will be classified into various categories based on their delivery method. These intervention categories include Community Education Sessions; Individual Health Education Sessions; Print Media Initiatives; Digital Media Interventions; Broadcast and Outdoor Media Campaigns; Dual-approach Interventions; Multi-faceted Intervention Programs; Miscellaneous Awareness Strategies. We will adapt this where necessary, and should the schema be inadequate we will develop a schema using thematic synthesis.

Targeted Behavioural Change Construct. To capture and understand how interventions are designed to modify behaviour, we will categorise interventions based on underlying behavioural theories in two ways. Firstly, in the bibliometric analysis, we will report whether the authors reported, and frameworks used. Secondly, we will conduct a framework analysis of the methods section of the included papers, utilising the Theoretical Domains Framework to categorise the aspects of behavioural change targeted by the intervention³⁹. This categorisation will help in understanding the theoretical basis of the interventions and their potential mechanisms of action.

Outcomes. The outcomes are structured in a hierarchy to reflect their logical progression from intervention to impact. Clinical outcomes (1–6) are at the top, as they represent the ultimate measures of success in terms of mortality and cancer stage. Healthcare utilization outcomes (7–10) follow, indicating how the intervention changes healthcare-seeking behaviour, which is necessary for clinical improvements. Measures of intent, attitude, or knowledge (11–12) come next, as these are precursors to behavioural change. Finally, campaign reach (13) is positioned at the base, as it represents the foundation upon which all other changes depend. This order is necessary to demonstrate the pathway from awareness to measurable health outcomes.

The outcomes of interest will be:

It is anticipated that most studies will not report clinical outcomes, and thus all reported outcomes will be further synthesized in our systematic review. However, our synthesis will specifically mention the outcomes listed above wherever possible (Figure 1).

Figure 1. A working theoretical model for cancer awareness interventions, illustrating the causal flow from campaign reach through increased awareness, behavior change, earlier diagnosis, and improved clinical outcomes.

This approach draws on behaviour change frameworks like the Health Belief Model and CRUK's "Waterfall" diagram for early diagnosis initiatives.

Inductive Thematic Analysis of Outcome Data. Due to the anticipated large number of included studies, an inductive thematic analysis of the reported outcomes will be utilised to produce a classification schema which will be employed in the systematic review and meta-analysis results, leading to data that are more interpretable and manageable for analysis. This will be completed using Braun & Clarke’s approach to thematic analysis³⁴, whereby each step will be conducted by the lead author and at least one other co-author. During the data extraction phase, the authors involved read and extract excerpts from the “Results” sections of included studies. Each excerpt will be comprehensive enough to cover all reported outcomes. This ensures that all relevant data from the studies are captured accurately. To enable the thematic synthesis, the excerpt will be loaded into either “NVivo”, a qualitative data analysis software⁴⁰ or “Taguette”, an open-source alternative⁴¹.

Step 1 (Familiarisation with data): During this step, all authors involved in the thematic synthesis will familiarise themselves with these excerpts by reading and re-reading them to gain an understanding of the content.

Step 2 (Generating initial codes): As outlined above, for each paper, two authors will code each reported study outcome in the excerpt.

Step 3 (Searching for themes): Collaborative discussions between all authors will take place to review and segregate the initial codes into larger groups that share common features.

Step 4 (Reviewing themes): This step will involve refining the themes synthesised in the previous phase. Authors will review each theme against the coded data to ensure it accurately represents the data. This process may involve combining, splitting, or discarding themes as necessary to ensure they comprehensively and accurately capture the nuances of the data.

Step 5 (Defining and naming themes): An attempt will be made to define each theme clearly and, where applicable, identify sub-themes. This process will involve developing detailed descriptions for all themes to ensure they comprehensively capture the essence of the coded data. Authors will write concise definitions and descriptions for each theme and sub-theme, highlighting their scope and distinguishing characteristics. Consensus discussions among all authors will be conducted to resolve any discrepancies and to ensure agreement on the definitions and descriptions. This step ensures that the themes are well-defined and distinct, facilitating a clearer and more coherent analysis.

Step 6: (Creating the classification schema): In this step, we will develop a classification schema for the reported outcomes. First, we will review the themes and sub-themes from previous steps and group similar ones into higher-order categories. We will then arrange these categories into a logical framework to highlight patterns and relationships within the data. Consensus discussions among all authors will refine the schema, ensuring agreement on its structure. Finally, we will finalize and describe the schema, ensuring it accurately reflects the data and enhances the interpretation of study findings.

Risk of Bias Assessment. We will critically assess and report the risk of bias in each study using established tools including ROBINS-E, ROBINS-I, and Cochrane RoB2^36,42,43. ROBINS-E is used for observational studies of exposure, whereas ROBINS-I is used in observational studies of intervention^42,43, and Cochrane RoB2 is used for assessing bias in randomised trials³⁶. This assessment will help in understanding the reliability and validity of the findings reported in the individual studies.

Data synthesis

The data will be synthesised in three phases. The first phase involves a bibliometric analysis, supported by a thematic synthesis of outcome data into a classification schema. In the second phase, a systematic review will be conducted to catalogue key outcome measures, enabling comparisons between interventions. The final phase will involve a network meta-analysis to further analyse the data.

Stage 1 (Bibliometric analysis): For our bibliometric analysis, we will collect and analyse data in several key categories. We will gather publication details (year, journal, authors, and affiliations) to understand research trends. We will examine the study context, focusing on geographical locations, settings, and populations. Intervention details, including types, duration, and modalities, will be documented to compare different approaches. We will categorise the behavioural theories underlying the interventions to understand their theoretical foundations. Study details such as design, sample size, and methodology will be compiled to assess research robustness. Finally, we will classify the reported outcomes to aid thematic synthesis and create a classification schema for the data. This systematic approach will help identify trends, gaps, and patterns in the research.

Stage 2 (Systematic review of efficacy): Our systematic treatment of the literature will allow us to catalogue key outcome measures, enabling detailed comparisons between different interventions. This process will provide a comprehensive overview of the effectiveness of various approaches, and if possible, a subgroup analysis of the impact of these interventions across sociodemographic groups, which will highlight the most effective interventions and areas needing further investigation.

Stage 3 (Network meta-analysis): We plan to conduct a network meta-analysis to compare multiple interventions simultaneously. This technique allows us to estimate the relative effectiveness of different interventions by integrating data from various studies. The analysis will help identify the most effective interventions and rank them in terms of efficacy and safety, providing a comprehensive understanding of their performance. However, this is only possible if the included studies are adequate in terms of quality, consistency, and comprehensiveness, as these factors ensure the reliability and validity of the meta-analysis findings.

Stage 4 (Overarching synthesis): To conclude, we will attempt to interpret the findings from the bibliometric analysis, systematic review, and network meta-analysis. This synthesis will consider the strengths and limitations of each methodology to generate concise and practical conclusions which are nonetheless rooted in the available evidence.

Network meta-analysis

Network meta-analysis (NMA) differs from standard meta-analysis as it incorporates direct and indirect comparisons of interventions across multiple studies, forming a network of evidence. The technique provides insights into the comparative effectiveness of all available treatments within the network. To perform any meta-analysis effectively, certain conditions should be met⁴⁴: literature quality should be of high standard; risk of bias of trials as well as publication bias should be analysed; heterogeneity of studies should be identified as well as relevance of studies to the primary research question. Specifically, to perform an NMA, additional considerations are required: homogeneity of evidence, consistency between direct and indirect estimates for the same intervention comparison, and transitivity between studies.

We will conduct all analyses using R, specifically the netmeta package^44,45, with a significance level set at p < 0.05. Data will be formatted using the pairwise () function, converting it into a contrast-based format suitable for NMA. The NMA will be conducted with the netmeta() function, integrating direct and indirect evidence. Both fixed-effect and random-effect models will be considered, with between-study variance estimated via restricted maximum likelihood. Meta-regression techniques using the metareg() function will adjust for covariates such as cancer type and intervention type.

Heterogeneity and inconsistency will be assessed using Cochran’s Q and I² statistics. The decomp.design() and netsplit() functions will be used to explore within-design and between-design heterogeneity and local inconsistency. Sensitivity analyses will include examining the impact of excluding studies with high risk of bias, using different methods for handling zero-event studies, and applying the Mantel-Haenszel method and continuity correction approaches to assess robustness of the findings. Subgroup analyses based on cancer type, stage, and patient characteristics will be performed by conducting separate NMAs for each subgroup.

Interventions will be ranked using SUCRA curves and P-scores, calculated with the netrank() function, and rankograms will visually display ranking probabilities. Results will be presented using netgraph() for network plots, forest.netmeta() for forest plots, and netleague() for league tables.

Forest plots generated with forest.netmeta() will display the estimated treatment effects and their confidence intervals, while league tables produced with netleague() will summarize all pairwise comparisons, providing a comprehensive overview of the relative effectiveness of each intervention. Publication bias will be assessed using comparison-adjusted funnel plots generated with the funnel.netmeta() function to detect any asymmetry indicative of potential publication bias.

Confidence in Cumulative Evidence. The GRADE (Grading of Recommendations Assessment, Development and Evaluation) approach will be used to assess and rate the strength of the body of evidence for each outcome examined in this review⁴⁶. GRADE provides a systematic method to evaluate the quality of evidence, initially categorising randomised controlled trials as high quality and observational studies as low quality. The evidence will be downgraded based on factors such as risk of bias, inconsistency, indirectness, imprecision, and publication bias. Conversely, it can be upgraded for factors like large effect sizes or a dose-response relationship. This process will result in a final rating of high, moderate, low, or very low quality for each outcome, providing a clear indication of the confidence we can place in the results.

Summary

The proposed evidence synthesis work aims to tackle the urgent need for evidence-based interventions to increase public awareness of cancer symptoms thereby facilitating earlier presentation and, consequently, improving cancer outcomes. With a rigorous methodology, this study ensures a thorough analysis of existing literature, providing valuable insights for policy and practice decisions. By synthesising evidence and evaluating interventions critically, it aims to contribute significantly to alleviating the burden of cancer on both individuals and society.

Limitations

Limitations exist that may affect the interpretation of our findings. First is subjectivity in data extraction in which reviewers analysing and extracting data from studies might introduce subjectivity, which could lead to biases in the synthesis process. The possibility of incomplete data should also be noted. The reliability of the synthesised results may be affected if there are missing or incomplete data due to variations in the quality and reporting of the included studies. For more recent papers the COVID-19 pandemic should also be taken into consideration, with its potential effects on factors such as patient willingness to seek physicians and attend routine GP consultations as well as shift to remote consultations⁴⁷. Publication bias also serves as a limitation as studies with more favourable outcomes are more likely to be published, leading to an overestimation of intervention efficacy. Language bias could also lead to overestimation of efficacy as studies published in languages other than English might be excluded.

Implications. The results of this systematic review will have an impact on the public health policy and practice, as well as future research. This review will provide public health professionals with collated evidence on the effectiveness of interventions targeting cancer awareness, early presentation and barriers to cancer seeking diagnosis. The goal of this review is to provide answers to key questions including the degree of positive impact of various interventions and cost-effectiveness of such interventions. This would provide a basis for the development of current health policies to allocate resources towards the chosen cancer awareness and early presentation initiatives. In addition, this review will highlight areas where further research is needed, ensuring the evidence based continues to develop.

Specifically, we will collate evidence for the National Cancer Control Programme (NCCP) in Ireland and thus give recommendations that will aid in shaping the field of increasing awareness in cancer. We will work with NCCP policymakers to create a policy brief summarising key findings and recommendations addressing early diagnosis of symptomatic cancer for the NCCP and other health policy stakeholders. Through academic publications, policy briefs, and stakeholder meetings, in addition to a summary report, the results of this systematic review are intended to augment efforts to raise awareness of cancer symptoms and presentations in Ireland.