Stage-stratified prognostic impact of comorbidities on breast cancer-specific survival: A population-based flexible parametric modelling study.

BACKGROUND

Comorbidities are common among women with breast cancer, yet their independent contribution to prognosis remains poorly characterised, particularly across disease stages. Most prognostic models prioritise tumour-related factors and when comorbidities are included, they are often summarised using aggregate indices, limiting evaluation of specific conditions. We investigated whether individual comorbidities exert effects on breast cancer-specific survival stratified by stage.

METHODS

We analysed 3323 women aged 20 - 79 years diagnosed with invasive breast cancer in Queensland, Australia (2010-2013) with a follow-up to December 2020. Comorbidities were self-reported and mapped to ICD-10. Flexible parametric survival models were fitted separately for early (I-II) and late (III-IV) stages, adjusting for established clinical factors. Prognostic performance was quantified using Royston's D-statistic, RD2 and Harrell's concordance index.

RESULTS

Comorbidities had no prognostic impact on early-stage disease. In late-stage breast cancer, chronic bronchitis independently predicted poorer breast cancer-specific survival after adjusting for subtype, grade, age, family history and detection mode. The final late-stage model demonstrated good discrimination (C-statistic 0.71) and explained 26% of the variation in survival. Model-based predictions showed clinically meaningful reductions in 5- and 7-year survival among women with bronchitis, partially mitigated by a family history of breast/ovarian cancer.

CONCLUSION

Stage stratification revealed prognostic signals that were obscured in pooled analyses. Chronic bronchitis worsened survival in late-stage disease, highlighting the value of assessing individual comorbidities during treatment planning. These findings support guideline recommendations for comorbidity informed management and underscore the need for stage-specific prognostic modelling frameworks.