A retrospective analysis was conducted using anonymized German claims data provided by AOK PLUS covering the period from January 1, 2010, to December 31, 2021. AOK PLUS, the sixths-biggest German sickness fund, covers approximately 3.5 million insured persons in the regions of Saxony and Thuringia in central-eastern Germany. This corresponds to 50% of the local population and approximately 4.8% of the overall German Statutory Health Insurance (SHI) population. The analyzed dataset comprised anonymized patient-level data on claims for inpatient care, outpatient care, outpatient drug prescriptions, and additional relevant data, including demographic information and all-cause of death.

The study population consisted of newly diagnosed mGC cases identified using German Modification of the International Classification of Diseases Version 10 (ICD-10-GM) codes. A metastatic-free period of at least 12 months was established before the index date (to identify mGC incident cases). The index date was defined as the first documented diagnosis of metastasis on or after the date of GC diagnosis. The following selection criteria were applied to retrieve the mGC population:

First, the disease criterion: At least one inpatient diagnosis of GC (ICD-10-GM code: C16.-) and/or at least two confirmed outpatient GC diagnoses in two quarters within 12 months. Second, the metastatic criterion: Any inpatient or confirmed outpatient diagnosis of metastasis (ICD-10-GM codes: C77-80) on or after the date of GC diagnosis. Third, the incident metastatic criterion: A metastatic-free period of at least 12 months before the first observable claim related to metastatic diagnosis. Fourth, the insurance criterion: Continuously insured for at least 12 months prior to the first observable metastatic diagnosis. Fifth, the age criterion: Age ≥ 18 years on the date of the first metastatic diagnosis. Sixth, the non-breast cancer (BC) criterion: No breast cancer diagnosis (ICD-10-GM code: C50.-) before or on the date of the first metastatic diagnosis. Seventh, the non-GIST criterion: No treatment for malignant gastrointestinal stromal tumor with imatinib (Anatomical Therapeutic Chemical Classification System (ATC) code: L01EA01, L01XE01; operation and procedure classification system (OPS) code: 6–001.g) and/or sunitinib (ATC code: L01EX01, L01XE04; OPS code: 6–003.a) during the entire study period. In addition, HER2-positive mGC patients were identified by a treatment proxy [treatment with trastuzumab (ATC code: L01XC03; OPS codes:6–001.7, 6–001.k) at any time after the first observed metastatic diagnosis] for subgroup analyses.

Incident mGC patients were identified during a 10-year period (from 2011 to 2020) by applying the selection criteria. Ensuring a reasonable time to record therapy treatments and time-to-event outcomes, data were considered until December 31, 2021. An overview of the study period and inclusion period is shown in Supplementary Fig. 1.

mGC therapies were identified through inpatient and outpatient codes. The treatment codes are listed in Supplementary Table 1. The first line of treatment (1LOT) was established considering the first prescription/application of any of the agents listed in Supplementary Table 1. All therapies documented within 28 days after starting a LOT were considered part of combination therapy. A new line of treatment was initiated if an additional active substance or combination of substances (that was not part of the current LOT) was prescribed after 28 days or more or if therapy was restarted after a gap of more than three months. When an unspecified OPS code was documented for inpatients (chemotherapy with unspecified substances code, OPS 8–54), it was assumed that the preceding LOT was continued during the period of hospitalization. No new treatment line was established for switching between platinum compounds (carboplatin, cisplatin, or oxaliplatin).

Qualitative variables were described as frequencies with percentages, and quantitative variables as mean ± standard deviation (SD). Baseline characteristics of mGC patients were stated at time of first mGC diagnosis, including age, sex, comorbidities identified through ICD-10 codes (Charlson Comorbidity Index (CCI) score [14], Supplementary Table 2), and the number of all-cause hospitalizations.

The most frequent therapy regimens were described numerically overall and by treatment line. Real-world time-to-treatment initiation (rwTTI) and real-world time-to-next-treatment (rwTTNT) were assessed using Kaplan–Meier (KM) method. RwTTI was estimated from mGC diagnosis until the date of death or censoring at end of continuous insurance or end of the study, whichever occurs first. RwTTNT was estimated since the start of each treatment line among those who reached the next LOT.

Survival analyses were conducted from the date of mGC diagnosis until the date of death (all-cause) or censoring at end of continuous insurance or end of the study, whichever occurs first. KM method was used to estimate real-world overall survival (rwOS). Median survival time and survival probabilities were reported. KM curves were displayed along with the number of patients at risk. Analyses were performed in each treatment line and in the HER2-positive mGC patients subgroup.

All findings were reported using the guidelines of the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) [15] and Standardized Reporting of Secondary data Analyses (STROSA) [16]. Descriptive and analytical statistical analyses were conducted using STATA software version 17 (STATA Corp., College Station, Texas).