Clinical and pathologic insights into early and late breast cancer relapses: a real-world analysis from the El Álamo IV registry
摘要
Finding prognostic factors for early and late relapses in operable breast cancer (BC) could improve relapse risk stratification; but there is a lack of real-world data with long enough follow-up.
ObjectiveTo investigate clinical and pathological features related to early versus late relapses in a real-world cohort of BC patients.
MethodsTo identify factors related to early (≤ 2 years) and late (≥ 5 years) relapse in women with stage I–III BC, hormone receptor (HR)-positive and human epidermal growth factor receptor 2-negative (HER2-), and HR-negative (HR −)/HER2 − by immunohistochemistry from El Álamo IV registry, logistic regression model was performed. To explore relapse dynamics, multivariate Cox regression models and annual hazard rates of relapses were reported.
ResultsOf the 1493 relapses, early, intermediate, and late relapses represented 28.3%, 29.7%, and 42.1% of all relapses, respectively. In 1050 patients, of both early and late relapses, age > 70 years (odds ratio [OR] 5.13; 95% confidence interval [CI] 3.23–8.15), stage III (OR 3.22; 95% CI 1.90–5.46), histological grade 3 (OR, 2.93; 95% CI 1.69–5.10), and HR − /HER2 − subtype (OR 10.26; 95% CI 6.55–16.05) were associated with an increase of early relapse risk. Annual hazard rate of loco-regional relapses (LRR) steadily increased over time in patients with HR + /HER2 − tumors, while those with HR − /HER2 − tumors exhibited a fluctuating pattern. For distant relapses (DR), the hazard rate peaked at 2 years and then rise steadily in HR + /HER2 − tumors whereas remained variable in HR − /HER2 − tumors. In multivariate Cox model, LRR was associated with stage, subtype, and histological grade, while DR was additionally influenced by age.
ConclusionAge at diagnosis, stage, histological grade, and tumor subtype were associated with distinct relapse timing patterns. These findings support further research into relapse dynamics and predictors by analyzing real-world data.