This was an observational, retrospective, cohort study using the Japanese medical claims database provided by Medical Data Vision Co., Ltd. (MDV; Tokyo, Japan). Data in this large-scale medical claims database are anonymized and contains unique Japanese diagnosis procedure combination (DPC)/per-diem payment system data, similar to diagnosis-related groups/prospective payment system (DRG/PPS) data. As of September 2021, MDV has accumulated data from more than 38 million patients and 23% of the acute care hospitals in Japan [12].

Patients with mBC were identified based on the following inclusion criteria: (1) having received a new prescription for any breast cancer treatment regimen, except for hormonal therapy, that were defined as drugs indicated for breast cancer in Japan (excluding adjuvant and neoadjuvant chemotherapy) between January 2017 and March 2022 and (2) diagnosis with breast cancer in the same month as the first-line index date or the month before. Among patients with mBC, the study population with mTNBC were identified by excluding those who did not have HER2, ER, or PR tests before the first-line index date, had a prescription of hormonal therapy or anti-HER2 drugs during the data period, or did not have a 3-month look-back period before the first-line index date.

The first-line regimen was defined as the first regimen administered after mTNBC diagnosis. The index date of the first-line regimen was defined as the start date of the first-line regimen and the first-line cohort was defined as the patient population followed up from the first-line index date; similar definitions were applied to the second- and third-line regimens, index dates, and cohorts. Treatment regimens were classified as either anthracycline-containing, taxane-containing [including taxane-based regimens containing ICI, e.g., atezolizumab/nab-paclitaxel, pembrolizumab/(nab-)paclitaxel], or other regimens. The ICIs were included in the taxane-containing category as the number of patients treated with these regimens was too small to be defined as a separate category. The details of the classification are defined in Online Resource 1.

The exposure period was defined from the first-/second-/third-line index date to the exposure end date. The exposure end date was defined as the earliest date out of the study end date (30 September 2022, the last date of available data), date of death, date of the last prescription of first-line regimen (or second/third-line regimen) considering the grace period, date of drug prescription for breast cancer not included in the first-line regimen (or second/third-line regimen), and date of the last hospital visit.

The grace period was 3 months, regardless of the regimen. The gap period was defined as the interval between the prescription dates for each regimen. If the gap period exceeded the grace period, the date of the grace period that elapsed from the previous prescription date was defined as the date of the last prescription. The follow-up period was defined as the period from the index date of each line until the end of exposure, including the censored date or hospital death. A minimum follow-up period of 180 days was allowed in this study.

Patient characteristics in the first-line cohort were summarized by the category of regimens: all, anthracycline-containing, taxane-containing, and other regimens. The information on age, sex, Charlson comorbidity index, medical history, and malignancy other than breast cancer was obtained from the database. The terms “Lymph node metastasis,” “Respiratory and digestive organ metastasis,” and “Other and unspecified site metastasis” were defined based on the ICD-10 diagnosis codes “Secondary and unspecified malignant neoplasm of lymph nodes,” “Secondary malignant neoplasm of respiratory and digestive organs,” and “Secondary malignant neoplasm of other and unspecified sites,” respectively.

TTNTD was defined as the time (in months) from the first-/second-/third-line index date until the next treatment or death, whichever occurred first. Patients who were still alive at the study end date or whose exposure ended when the gap period exceeded the grace period were censored at the exposure end date.

TTD was defined as the time (in months) from the first-/second-/third-line treatment start date until discontinuation due to any cause, start of next line, or death. Patients who were still on their first-/second-/third-line treatment, were alive at the study end date, or were lost to follow-up were censored at the last confirmed activity date.

The AEs of interest assessed in this study included nausea/vomiting, neutropenia/leukopenia, febrile neutropenia, anemia, diarrhea, interstitial lung disease (ILD), and peripheral neuropathy. The frequencies of these AEs were identified by the disease diagnosis codes (ICD-10 codes) in combination with the treatment prescribed for the AE. The ICD-10 codes and the treatments used are listed in the Online Resource 2. A diagnosis was considered an AE of interest if it occurred during the follow-up period in the same month as the start of the corresponding treatment.

Medical remuneration points are recorded in a medical claims database. Each cost during the first-/second-/third-line treatment and the cumulative costs from the first-line to the end of third-line treatment were calculated for each patient. Costs per day were also calculated and divided by the follow-up period. The three medical cost subtypes included cost limited to any prescribed drugs (drug cost), cost limited to inpatient claims (hospitalization cost), and cost limited to outpatient claims (outpatient cost).

Statistical analyses were performed using SAS, version 9.4 (SAS Institute, Inc., Cary, NC, USA). Continuous variables were reported as median and interquartile range. Categorical variables were summarized as the number and proportion of the total study population and subgroups. Missing data were not imputed.

For treatment patterns, number of patients and percentages per first-/second-/third-line regimen combination category were calculated. The results were illustrated in a Sankey diagram.

For TTNTD and TTD, the median time to event and 95% confidence intervals (CIs) plus 25% and 75% quantiles were estimated. The results were depicted graphically using Kaplan–Meier curves, plotting the survival rate against time.

Frequency distributions were provided for AEs of interest (nausea/vomiting, neutropenia/leukopenia, febrile neutropenia, anemia, diarrhea, ILD, and peripheral neuropathy) in overall and by age (< 65 or ≥ 65 years).

For medical costs (total costs, total costs per day, drug cost, drug cost per day, hospitalization cost, hospitalization cost per day, outpatient cost, and outpatient cost per day) during the first-/second-/third-line treatments, descriptive statistics (median and interquartile range) were calculated.

Subgroup analysis was performed on treatment pattern by age (< 65 or ≥ 65 years) and by medical history of adjuvant or neoadjuvant chemotherapy at the first-line index date.