Miyazaki T, Kuranishi F, Ohno T. Additional Landmark and RMST Analyses Addressing Immortal Time Bias in the
Evaluation of Autologous Formalin-Fixed Tumor Vaccine for Metastatic Breast Cancer. Clin Breast Cancer 2026;26:80-81.
Letters to the Editor
Dear Editors,
We write regarding our recently published article in Clinical Breast Cancer reporting the survival benefit associated with autologous formalin-fixed tumor vaccine (AFTV) in patients with distant metastatic breast cancer. Following publication, we recognized that the original survival analyses may have been susceptible to immortal time bias, as patients were required to survive long enough after
detection of distant metastasis to receive the Optimum AFTV Set treatment.
To address this concern, we conducted additional landmark analyses using predefined landmark times of 1.0, 1.5, and 2.0 years after distant metastasis. A binary variable, landmark AFTV-point (LMAFTV), was defined according to whether a patient had received
AFTV before each respective landmark. Survival analyses were then restricted to patients who were alive at the corresponding landmark time.
Patient numbers at each landmark and the corresponding survival analyses are summarized in Table 1 . In the 1.0-year landmark analysis, the hazard ratio (HR) for meta-OS —overall survival after distant metastasis —was 0.4035 (95% confidence interval [CI],
0.1189-1.369; P = .145). The corresponding HRs were 0.4245 (95% CI, 0.1426-1.264; P = .124) in the 1.5-year analysis and 0.4802 (95% CI, 0.1579-1.460; P = .196) in the 2.0-year analysis. All landmark analyses consistently favored the AFTV-treated group
( Table 1 and Figure 1 ).
Because proportional hazards assumptions may not fully capture long-term survival differences in small cohorts, we additionally
performed restricted mean survival time (RMST) analyses. Positive RMST differences were observed across all landmark definitions. In the primary 1.5-year landmark analysis, RMST differences between the LMAFTV-positive and LMAFTV-negative groups were 0.267 years at 𝜏 = 3 years ( P = .002), 0.875 years at 𝜏 = 5 years ( P = .017), 1.950 years at 𝜏 = 8 years ( P = .018), 2.621 years at 𝜏 = 10 years ( P = .028), and 3.307 years at 𝜏 = 12 years ( P = 0.037). Similar patterns were observed in the 1.0-year and 2.0-year
landmark analyses ( Table 2 and Figure 2 ).
We appreciate the opportunity to provide these additional analyses and believe that they strengthen the interpretation of the findings
reported in our original article.
Sincerely yours,
