Very Young Age Stratification and Development of a Multiple-Imputation Cox Nomogram for Disease-Free Survival in Breast Cancer Women Aged ≤40 Years
Xiaohong Lyu, Ying Zhong, Yidong Zhou, Qiang Sun, Feng Mao, Songjie Shen, Yan Lin, Xin HuangAbstract
Background
Breast cancer in women aged 40 years or younger is clinically heterogeneous, yet risk stratification within this already-young population and rigorously validated disease-free survival (DFS) tools remain limited.
Methods
We retrospectively analyzed 2,069 women aged 40 years or younger who underwent surgery between 2006 and 2021. Patients were compared as younger than 35 versus 35–40 years to assess the prognostic relevance of very young age. Age was additionally modeled continuously using restricted cubic splines, and the optimal cutpoint was examined using maximally selected rank statistics. DFS was evaluated with Kaplan–Meier estimates and multivariable Cox models. We developed a Cox nomogram for 3- and 5-year DFS using age, molecular subtype, pathologic T stage, pathologic N stage, histologic grade, Ki-67 category, and calendar period. Missing covariates were handled by multiple imputation by chained equations (20 imputations), with internal validation by 1,000 bootstrap resamples and temporal validation of a reduced model.
Results
Of 2,069 women, 779 were younger than 35 years. Very young patients had more aggressive features, including higher frequencies of HER2-positive and triple-negative disease, grade 3 tumors, high Ki-67, and advanced T/N stage. Overall 5-year DFS was 88.9%; patients younger than 35 years had worse 5-year DFS (85.0% vs 91.4%; log-rank P = 0.0005; absolute difference 6.4 percentage points). Spline modeling showed a continuous, monotonic increase in risk with younger age without evidence of non-linearity (P = 0.88), and maximally selected rank statistics identified 35 years as the optimal cutpoint (standardized log-rank z = 3.51, P < 0.001). After adjustment for molecular subtype and pathologic T and N stage, age younger than 35 years remained independently associated with worse DFS (hazard ratio 1.34, 95% CI 1.03–1.76; P = 0.032). The multiple-imputation nomogram had an apparent C-index of 0.660, a bootstrap-corrected C-index of 0.636, and a temporal-validation C-index of 0.617.
Conclusion
Very young age identifies a higher-risk subgroup among women aged 40 years or younger with breast cancer. A multiple-imputation Cox nomogram based on routinely available variables provided moderate discrimination and acceptable calibration and is best suited to broad risk-group stratification rather than precise individual-level prediction.