Psychological impact of risk-stratified screening as part of the NHS Breast Screening Programme: multi-site non-randomised comparison of BC-Predict versus usual screening (NCT04359420)

The NHS Breast Screening Programme (NHSBSP) aims to reduce mortality by early detection, but screening entails harms such as false positives and overdiagnosis. Risk-stratified screening could improve the benefit-harm balance by tailoring screening and prevention according to individual risk. NICE recommends enhanced surveillance and risk-reducing medication for high-risk women, yet most eligible women are unidentified in routine practice. Prior work (PROCAS) suggested no major psychological harms from communicating risk estimates, but had limitations: delayed communication relative to baseline, exclusion of highest-risk women, comparison to women awaiting results, and no baseline measures. This study aimed to determine whether offering BC-Predict as part of NHSBSP affects psychological harms (general anxiety, cancer worry) and benefits (knowledge, intentions) at 3 and 6 months, and how responses vary by risk category. The key research questions were: (1) Are there differences in self-reported harms/benefits between women offered BC-Predict versus NHSBSP, controlling for baseline? (2) Among those offered BC-Predict, do accepters differ from decliners, controlling for baseline? (3) Do changes in outcomes differ by communicated risk category (high, moderate, average, below average)?

Previous research (PROCAS) using Tyrer-Cuzick-based multi-factor risk estimates (self-report risk factors, mammographic density, and polygenic SNPs) found no major harm from receiving risk information: slightly higher cancer worry but lower general anxiety compared to women awaiting results, with overall low scores and risk perceptions shifting in line with communicated risk. However, that work had key limitations (timing, exclusion of highest-risk women, control group awaiting results, lack of baseline measures). Broader literature indicates routine screening can transiently affect anxiety and that personalized risk communication often has limited impact on behavior change. Evidence on whether risk-stratified screening improves knowledge and informed choice is sparse.

Design: Nested questionnaire study within PROCAS2, a multi-site non-randomised controlled comparison of offering BC-Predict versus standard NHSBSP at participating sites in North-West England. Ethical approvals: Harrow REC (18/LO/0649) and North West–Greater Manchester East REC (18/NW/0856). Sites: five sites offered BC-Predict; two sites offered standard NHSBSP. Recruitment: November 2020–July 2021 (one site from February 2021). Participants: Women scheduled for mammography (prevalent screens: first-time at any age; incident screens: invited in the round reaching age 60, typically 57–63 years). Inclusion: biologically female, able to consent. Exclusion: prior breast cancer or bilateral mastectomy. Procedure: All women received NHSBSP invitations (COVID-19 context: open-invite letters). Those in BC-Predict areas received an additional invitation 1–2 working days later to complete online risk assessment (self-reported risk factors; mammographic density automatically derived for most; optional saliva kits for SNPs). Women accepting BC-Predict who had clear mammograms received postal feedback ~6–8 weeks post-mammogram with 10-year risk category: high (≥8%), above average/moderate (≥5% to <8%), average (≥2% to <5%), below average (<2%); from March 2021 “below average” label changed to “low,” with note about potential extended screening intervals. High and moderate groups were encouraged to attend a Family History and Risk Prevention Clinic. Questionnaires: Administered at baseline (~7 days after first offered mammogram appointment) and at 3 and 6 months post initial mammogram date. Follow-up invites sent only after women had a clear mammogram result. Non-responders received a second invite ~2 weeks later. Paper questionnaires available. Measures: (1) Perceived comparative risk (single item comparing risk to peers over 10 years; 1=“much lower” to 5=“much higher”). (2) State anxiety: 6-item short-form STAI; 4-point response; internal consistency α=0.86; clinical case threshold set at 14.7 equivalent on 6-item scale (based on 49 on 20-item scale). (3) Cancer worry: 6-item Lerman Cancer Worry Scale; 4-point response; α=0.87. (4) Knowledge of breast cancer screening: 11 conceptual items from standard measure; α=0.30. (5) Attitudes toward screening: 3 items (good/beneficial/important); α=0.67. (6) Intentions to attend future screening: single item (5-point agree scale). (7) Satisfaction with information (BC-Predict recipients at 3 and 6 months only): 4 items (clarity, confusion, feeling informed, satisfaction with amount); 7-point scale; α=0.90. Demographics and clinical info: from BC-Predict questionnaire or aggregate data via CAG approval for non-consenting groups; deprivation via Index of Multiple Deprivation (IMD) deciles from postcodes (1 most deprived–10 least deprived). Sample and flow: Of 5901 invited (BC-Predict n=3093; NHSBSP n=2808), 662 consented (11.2%): BC-Predict n=358; NHSBSP n=304. Follow-up response: 3 months 77.2% overall (BC-Predict 80.7%; NHSBSP 73.0%); 6 months 71.5% overall (BC-Predict 72.6%; NHSBSP 70.1%). Among BC-Predict accepters with risk estimates (n=184 baseline): high n=22, moderate n=27, average n=114, below average n=21. Data analysis: Group comparisons of demographics via t tests and chi-squared. Main outcomes analyzed via ANCOVA comparing BC-Predict vs NHSBSP at 3 and 6 months, adjusting for baseline value of each outcome, age, and IMD decile; primary outcome prespecified as 6-month state anxiety. Sensitivity analysis for 6-month state anxiety using last observation carried forward. Comparisons of BC-Predict accepters vs decliners used similar covariate-adjusted analyses. Among BC-Predict recipients, ANCOVA compared changes across risk categories (high, moderate, average, below average) with age, IMD, and baseline measure covariates; significant omnibus tests followed by post hoc pairwise comparisons. Satisfaction compared across risk groups using ANOVA (no baseline available). Two-sided tests, α=0.05. Planned sample size: n=1054 at 6 months (527 per arm) for 90% power to detect d=0.2; achieved n=473 at 6 months.

- Participation: Of 5901 invited, 662 (11.2%) consented (BC-Predict 358; NHSBSP 304), with similar participation likelihood by allocation (χ²=0.83, P=0.363). Participants were less deprived than non-participants (higher IMD decile; P<0.001). Groups were similar in age and mammography attendance; BC-Predict participants were less deprived on average. - Primary and secondary outcomes (BC-Predict vs NHSBSP): No statistically significant differences at 3 or 6 months in comparative risk perception, state anxiety, cancer worry, screening knowledge, attitudes, or intentions after adjustment for baseline, age, and IMD. Examples: state anxiety at 3 months F(1,502)=0.017, P=0.896; at 6 months F(1,462)=1.314, P=0.252. Sensitivity analysis for 6-month state anxiety using LOCF: no difference. Baseline levels were the strongest predictors of follow-up outcomes (e.g., state anxiety β=0.58, P<0.001). - Acceptors vs decliners of BC-Predict (baseline): Acceptors had higher IMD deciles (less deprived; t=3.23, P=0.001), lower comparative risk perceptions (t=2.39, P=0.017), lower cancer worry (t=2.01, P=0.045), and higher screening knowledge (t=2.29, P=0.023); no significant differences in age or baseline state anxiety. - Within BC-Predict by risk category: • Comparative risk perceptions changed in line with communicated risk (3 and 6 months omnibus ANCOVA P<0.001; e.g., F(3,146)=28.56, P<0.001 at 6 months). High and moderate groups rated higher risk; below-average group rated lower risk. • Cancer worry differed by risk category at 6 months (F(3,146)=4.15, P=0.008), increasing for higher-risk groups and decreasing for lower-risk group; no significant differences at 3 months. • State anxiety differences emerged at 6 months (F(3,146)=3.73, P=0.013), with the below-average/low-risk group showing a notable increase (mean 13.67 vs ~9.8–10.7 in other groups); not evident at 3 months. Proportion above clinical anxiety threshold did not differ significantly (P=0.088 at 6 months), and overall levels remained low. • Screening knowledge and intentions showed little change by risk category; attitudes showed small differences at 3 months (F(3,144)=3.55, P=0.016) but not a consistent pattern. - Overall, no evidence of increased general anxiety or cancer worry attributable to offering BC-Predict vs standard NHSBSP. Risk information was understood and influenced subjective risk perceptions appropriately without inducing clinically significant distress.

The study addressed whether embedding risk-stratified screening (BC-Predict) in NHSBSP leads to psychological harms or benefits compared with usual screening. Findings showed no differences in anxiety, cancer worry, knowledge, attitudes, or intentions between women offered BC-Predict and those offered standard NHSBSP at either 3 or 6 months, even after controlling for baseline, age, and deprivation. Among BC-Predict recipients, risk perceptions and cancer worry moved appropriately in line with communicated risk, indicating comprehension of risk information, but changes were modest and did not translate into clinically meaningful anxiety. The strongest determinant of follow-up distress was baseline distress, suggesting that offering risk-stratified screening does not create new psychological burden. An unexpected pattern was an increase in general state anxiety at 6 months among women told they were at below-average/low risk, despite concurrent decreases in cancer-specific worry, indicating that the increase in general anxiety may be unrelated to breast cancer concerns. The absence of effects on screening knowledge and intentions aligns with broader literature showing limited behavioral impact of personalized risk communication. Overall, results support the psychological safety of implementing risk-stratified screening within NHSBSP.

This multi-site, real-world comparison provides no evidence that offering BC-Predict as part of NHSBSP causes psychological harm, including among women informed they are at high risk. Communicated risk estimates appropriately adjusted women’s perceived risk and cancer worry without elevating clinical anxiety. The study strengthens prior evidence with more timely risk communication, inclusion of high-risk women, and baseline-controlled analyses. Implementation of risk-stratified screening should not be hindered by concerns over psychological harm. Future research should: (1) track actual screening attendance and informed choice in subsequent rounds; (2) explore reasons for the unexpected increase in general anxiety among low-risk women; (3) evaluate longer-term psychological and behavioral outcomes (as in MyPeBS) and potential benefits or harms over several years; and (4) develop and test strategies to improve screening knowledge alongside risk communication.

- Non-randomised design with allocation by site and resulting imbalance in deprivation (IMD) between groups, though analyses adjusted for IMD. - Low questionnaire response rate (11.2%), raising potential response bias; however, participation was similar across allocation groups. - Smaller achieved sample than planned (n=662 baseline; n=473 at 6 months vs planned n=1054), reducing power to detect small effects, though results consistently showed no meaningful differences. - COVID-19 disruptions altered NHSBSP operations (open invitations), impacted recruitment and uptake (NHSBSP uptake ~60.7%; BC-Predict uptake 16.1%). - Limited ethnic diversity data available (self-report ethnicity collected only for BC-Predict group).