Randomised controlled trial and economic evaluation of a targeted cancer awareness intervention for adults living in deprived areas of the UK

The study addresses persistent socioeconomic inequalities in cancer outcomes in the UK, where lower symptom awareness and greater practical, emotional, and attitudinal barriers to help-seeking contribute to delayed presentation and later-stage diagnoses in deprived communities. Mass media campaigns can improve awareness but often fail to reach lower socioeconomic groups. The research tests whether a targeted, tailored, community-delivered cancer awareness intervention (ABACus Health Check) can improve recognition of common and lesser-known cancer symptoms and reduce anticipated time to presentation among adults aged 40+ living in socioeconomically deprived areas of South Yorkshire and South-East Wales. The purpose is to evaluate effectiveness and cost implications compared with usual care, thereby informing strategies to expedite cancer diagnosis in underserved populations.

Prior work links low cancer symptom awareness with delayed help-seeking and later-stage diagnosis, particularly in deprived groups. Barriers include practical and emotional constraints, worry about wasting doctors’ time, and fatalistic beliefs. While large-scale campaigns (e.g., Be Clear on Cancer) can raise awareness, they may exacerbate inequalities by insufficiently engaging lower socioeconomic groups. Community-based, peer-supported behavioural interventions show promise for engaging underserved populations and normalising earlier help-seeking. Existing trials of awareness interventions often face limitations in design, intensity, dose, and follow-up, with mixed outcomes and challenges related to selecting comparators and avoiding contamination. The ABACus Health Check was co-produced with stakeholders, grounded in behaviour change theory, and previously demonstrated feasibility and potential for scalability and reach in deprived community settings, warranting a rigorous evaluation through a randomised controlled trial.

Design: Unblinded, individually randomised controlled trial with 1:1 allocation, conducted in socioeconomically deprived areas of South and West Yorkshire (Sheffield, Barnsley, Rotherham, Wakefield, Doncaster) and South-East Wales (Merthyr Tydfil, Newport). Randomisation used permuted blocks of sizes 2, 4, and 6. Follow-up at two weeks and six months post-randomisation. Participants: Adults aged ≥40 years recruited from community and healthcare venues located in the most deprived quintiles by IMD/WIMD. Exclusion: non-English speaking, unable to give written informed consent, or prior participation in Phase 2. Recruitment/settings: Trained lay advisors approached local delivery organisations in community (e.g., job shops, libraries, sheltered housing) and healthcare settings (GP surgeries, pharmacies), securing private spaces for intervention delivery. Sample size: Target n=246 (90% power, two-sided alpha 0.05, effect size 0.5, SD=2.2, assuming 30% attrition) to detect an average increase of one symptom in recognition. Intervention: ABACus Health Check—an online interactive touchscreen questionnaire covering common cancer symptoms, screening, and risk factors (smoking, alcohol, diet, activity), with tailored feedback and personalised behavioural advice delivered by trained lay advisors. The intervention employed seven behaviour change techniques: information about health consequences; prompts/cues; credible source; restructuring the social environment; social support; goal setting (behaviour); and action planning. Three lay advisors (UG/Masters-level, health promotion background) delivered the Health Check. Control: Usual care (standard support via GP/community organisations). Data collection: Baseline face-to-face data entry via iPad; two-week and six-month follow-up by telephone or post per participant preference. Incentives: £10 at baseline, £5 after six-month follow-up. Outcomes: Primary—total cancer symptom recognition at two weeks using an adapted Awareness and Beliefs about Cancer (ABC) measure. Secondary—anticipated time to presentation for selected symptoms (unusual lump, rectal bleeding, persistent cough, unexplained weight loss); perceived barriers to presentation; beliefs about cancer; state anxiety; implementation costs; subsequent healthcare resource use (Client Service Receipt Inventory adapted for deprived populations; costs in 2018 GBP). Statistical analysis: Intention-to-treat with complete cases. Linear models for primary outcome at two weeks, mixed-effects two-level partial cluster regression to adjust for stratification (lay advisor) and clustering. Distributional checks, transformations, and bootstrapping as needed. Adjustments for baseline scores. Pre-specified subgroup analyses (age, gender, recruitment setting, education). Sensitivity analyses for missingness and timing outside the two-week window. Psychometric testing (Cronbach’s alpha, factor analysis). Health economics: Within-trial cost differences (independent samples t tests with Holm-Bonferroni), paired t tests for within-group changes. Incremental cost per point improvement in awareness at six months. Model-based cost-utility analysis (decision tree + Markov model) over five years for a simulated cohort of 100,000 people in deprived areas, NHS and PSS perspective, 3.5% discounting, deterministic and probabilistic sensitivity analyses.

- Participants: 234 randomised (117 intervention; 117 control); 82 (35%) from South-East Wales, 152 (65%) from South/West Yorkshire. Mean age 61.3 years; 63.2% female; 97.9% White; high deprivation representation. Retention: 90.5% at two weeks; 85.0% at six months. - Primary outcome at two weeks: No statistically significant difference in total cancer symptom recognition (Adjusted Mean Difference [AMD] 0.6; 95% CI -0.03 to 1.17; p=0.06). - Secondary outcomes: • At six months, increased total symptom recognition in intervention (AMD 0.8; 95% CI 0.18 to 1.37; p=0.01). • Shorter anticipated time to presentation at six months in intervention (AMD -2.0; 95% CI -3.02 to -0.91; p<0.001). • No significant group differences for anticipated presentation at two weeks (AMD -0.8; 95% CI -1.79 to 0.29), perceived barriers (two weeks AMD 0.1; 95% CI -0.23 to 0.41; six months AMD 0.3; 95% CI -0.36 to 0.91), beliefs about cancer (two weeks AMD -0.02; 95% CI -0.50 to 0.45; six months AMD -0.03; 95% CI -0.45 to 0.38), or state anxiety (two weeks AMD 0.03; 95% CI -0.11 to 0.18; six months AMD 0.81; 95% CI -0.10 to 1.72). • Recognition subscales: Large ceiling effect for well-known symptoms at baseline. Higher recognition of lesser-known symptoms in intervention at two weeks (AMD 0.5; 95% CI 0.03 to 0.97) and six months (AMD 0.7; 95% CI 0.16 to 1.17). No differences for well-known symptoms at two weeks (AMD 0.1; 95% CI -0.16 to 0.27) or six months (AMD 0.1; 95% CI -0.09 to 0.28). - Health economics: • Implementation cost per participant: £91.34 (training £12.51; delivery staff £21.05; equipment/hosting £57.78). • Six-month healthcare costs related to potential cancer symptoms: intervention had £11.64 lower mean total costs vs control (95% CI -£198.77 to £175.50; p=0.90), not significant. Including implementation costs, mean cost difference £79.70 higher vs control (95% CI -£107.43 to £266.84; p=0.40), not significant. • Incremental Cost-Effectiveness Ratio (ICER): £108.85 per one-point improvement in symptom recognition at six months (sensitivity £31.51–£205.97). • Model-based cost-utility (5-year horizon): Intervention dominated (on average £4.08 less costly and 0.005 more QALYs). Cost-effective or cost-saving across sensitivity analyses; 88.6% and 95.5% probability cost-effective at £20,000 and £30,000/QALY thresholds, respectively.

The trial tested a targeted, tailored, lay advisor–facilitated cancer awareness intervention in highly deprived communities and found no short-term improvement in total symptom recognition but did demonstrate longer-term gains at six months, including increased recognition overall and especially for lesser-known, non-specific symptoms, and reduced anticipated time to presentation. These findings address the research question by showing that a personalised behavioural approach can support knowledge retention and intention to seek help earlier over time without increasing anxiety or healthcare utilisation in the short term. The observed ceiling effect for well-known symptoms highlights measurement reactivity and high baseline awareness, which may have dampened detectable short-term effects. The results align with literature suggesting that community-based, tailored interventions may be more effective for underserved populations than mass media alone, particularly in improving recognition of non-specific symptoms that are often misattributed. Economically, the intervention was low-cost to implement and showed a high probability of being cost-effective or cost-saving over a five-year horizon, suggesting potential value to health systems aiming to reduce diagnostic delays and stage at diagnosis disparities. Overall, the intervention offers a scalable model to complement diagnostic pathways by encouraging earlier help-seeking for subtle symptoms in deprived communities.

The ABACus Health Check, a tailored, lay advisor–facilitated cancer awareness intervention, is a viable, low-cost approach that improved cancer symptom recognition at six months and reduced anticipated time to presentation among adults in socioeconomically deprived areas. While short-term total recognition gains were not significant, longer-term improvements—particularly for lesser-known symptoms—were observed without increased anxiety or healthcare costs. Economic modelling indicates the intervention is likely cost-effective or cost-saving over a five-year horizon. Future research should include larger-scale implementation with longer-term follow-up using routine data to assess impacts on actual help-seeking behaviour, diagnostic intervals, stage at diagnosis, and survival, and to evaluate adaptations for diverse ethnic and linguistic groups and broader health system integration.

- Measurement issues: Potential measurement reactivity and ceiling effects for well-known symptom recognition; reliance on self-reported knowledge, beliefs, and anticipated rather than observed presentation behaviour; low internal consistency of the adapted ABC cancer belief measure. - Generalisability: Predominantly White British, native English-speaking sample limits applicability to more diverse populations; adaptations may be needed for ethnic minorities and non-English speakers. - Potential contamination: Unknown exposure of control participants to other local/national awareness campaigns could attenuate observed effects. - Implementation context: Trained peer supporter model may face integration challenges within routine UK healthcare provision. - Sample and follow-up: Although retention was high and power adequate, actual behavioural outcomes (e.g., real-world presentations, diagnoses, stages) were not directly measured within the trial period.