Claims-Based Frailty Trajectories During Adjuvant Chemotherapy Predict Long-Term Mortality in Older Women with Stage I-III Breast Cancer

Background: Frailty assessed at diagnosis or cancer treatment initiation is predictive of outcomes in older women with breast cancer. Little is known on how frailty trajectories following cancer treatment initiation affect mortality.

Objectives: We evaluated the association between claims-based frailty trajectories during the year following adjuvant chemotherapy (ACT) initiation and five-year mortality in older women with stage I-III breast cancer.

Methods: We included women (≥65 yrs) with stage I-III breast cancer diagnosed from 2004-2017 in the linked Surveillance Epidemiology and End Results-Medicare database. We included women who underwent breast surgery without neoadjuvant therapy within 90 days of diagnosis and initiated ACT within 90 days of surgery. We used a landmark design to identify frailty trajectories during the year following ACT initiation. Women were required to be continuously enrolled in Medicare fee-for-service from 180 days prior to diagnosis through 360 days following ACT initiation (“landmark”). Women who died (3%) or disenrolled (3%) before the landmark were excluded. Frailty trajectories were identified using the Faurot frailty index, a validated claims-based frailty proxy based on demographics and healthcare billing data. The Faurot index was calculated every 30 days between ACT initiation and the landmark using 6-month assessment windows. Frailty trajectory clusters were identified using K-means clustering. Women were followed from the landmark for five years until death or end of data availability. We assessed associations between frailty trajectories and mortality using Kaplan-Meier analysis and Cox models, adjusting for age, race, ethnicity, surgery type, stage, region, year, and comorbidities.

Results: We included 20,774 women (median age 70 yrs; stage I: 23%, II: 54%, III: 24%) and identified 6 trajectory groups. Three groups (95%) had resilient trajectories, with a small increase in frailty during ACT followed by a return to baseline. Three groups (5%) had non-resilient trajectories, increasing after ACT initiation without returning to baseline. Five-year mortality risk was higher in women with non-resilient trajectories compared to those with resilient trajectories (52% vs. 20%, RD=32%, 95% CI 29-36%). In the Cox model, both baseline frailty and claims-based frailty trajectories were associated with 5-year survival.

Conclusions: Claims-based frailty trajectories over one year following ACT initiation were strongly associated with 5-year mortality in older women with stage I-III breast cancer. Improved monitoring for frailty during and after cancer treatment followed by appropriate supportive interventions may improve long-term survival in this population.