The NeuroPoint Alliance SRS Registry is a multi-institutional collaboration prospectively collecting information from patients undergoing SRS since 2017, with the most common pathologies treated being brain metastasis, meningiomas, and schwannomas [12]. The current database abstracts patient data from 27 active contributing sites with over 5500 patients enrolled and more than 8400 treatment events captured [13]. The registry was being used for quality improvement processes, and as such, was deemed IRB-exempt.

Data were gathered from eligible patients within the NPA registry who underwent SRS for the treatment of brain metastases originating from lung cancer, breast cancer, or melanoma and for whom there was at least one pre-SRS and one post-SRS QOL endpoint captured as part of their disease treatment progression.

Baseline demographic data, functional information, disease characteristics, and treatment details of patients with brain metastases originating from lung cancer, breast cancer, and melanoma were included in the analysis. Baseline information included demographic information (e.g. age, sex, race); comorbidities and medical risk factors (e.g. diabetes mellitus, body mass index, coronary artery disease, smoking status); the Karnofsky Performance Status (KPS) score and patient-reported five-dimension Euro-QoL (EQ-5D-5 L in quality-adjusted life-years (QALYs – signifying how the patient’s quality of life compares to a full year in perfect health) as assessed by the clinical team [14, 15]. Disease characteristics included radiographic metrics (e.g. lesion maximum diameter, lesion volume, the cumulative intracranial tumor volume (CITV, defined as the sum of all intracranial lesions’ volumes)), number of brain metastases, the presence of intratumoral hemorrhage with or without extension into the brain parenchyma, and anatomic location of the lesions [16]. Treatment details consisted of dosimetry data (e.g. radiosurgical prescription dose, mean dose, margin dose, and number of fractions); pre-SRS treatments (e.g. surgical resection, whole-brain radiation therapy (WBRT), chemotherapy, immunotherapy, molecular therapy); the first metastatic site; and different neo-adjuvant treatments.

Outcomes of interest for the time-to-event analyses included all-cause mortality, local progression, out-of-field progression, and overall intracranial progression. Local progression was defined as an increase in the lesion volume by at least 72.8% compared to the baseline, corresponding to the volumetric equivalent of a 20% diameter increase of a spherical volume, by the Response Assessment in Neuro-Oncology Brain Metastasis (RANO-BM) criteria [17, 18]. Out-of-field progression was defined as the appearance of new metastasis beyond the previously irradiated field. Overall intracranial progression was a composite endpoint of the local and the out-of-field progression, considering the time point at which either local or out-of-field progression occurred first.

To assess QOL measures, we relied on the EQ-5D questionnaire responses. The EQ-5D is a self-reported patient survey designed to capture their description of five distinctive health states about their primary disease: mobility, self-care, usual activities, pain/discomfort, and anxiety/depression [19]. Patients are asked to indicate the level of problem they experience with each of the five dimensions, which collectively represent their EQ-5D profile. Subsequently, utilizing the “eq5d” package in RStudio, their profile data was converted to a single utility index value that ranges from 0, indicating death, to 1, indicating full health, using preference valuations weights for U.S. populations [19]. The SRS registry was queried for the baseline and follow-up EQ-5D questionnaire scores at 6–12 months and the last recorded follow-up, including the utility index scores given based on the patient’s responses for all five domains. Patients were included in the analytic sample with the availability of their SRS index scores pre-SRS and at the final follow-up. Regression analyses were then conducted to establish predictors of EQ-5D index and domain scores at final follow-up. We analyzed both the EQ-5D index and domain scores.

Categorical variables were reported as proportions and frequencies, normally distributed continuous variables as means and standard deviations (SD), and non-parametric continuous variables as medians with interquartile ranges (IQR). Cut-off thresholds for continuous variables were determined based on tertile and quartile distributions to ensure adequate representation across the groups. These distributions were calculated for age, body mass index, CITV, margin dose, and mean dose. Following SRS guidelines, the lesion number at baseline was grouped as 1, 2–4, and ≥ 5 [20]. Missing baseline data were imputed using multivariate imputation by chained equations with 20 iterations and via the “mice” package [21]. Statistical significance was defined as a p-value < 0.05 level. Time-to-event analyses for overall survival, out-of-field, local, and intracranial progression were expressed by Kaplan-Meier curves. Multivariable Cox regressions were computed to establish hazard ratios for the associated risk factors for specific endpoints of interest. Statistical analyses were conducted using the “survival” and “ggplot2” in the R statistical package [22].

The Paretian Classification of Health Change (PCHC) was utilized to analyze changes in patients’ EQ-5D scores over time [23]. PCHC categorizes changes in health status as better (improvement in at least one EQ-5D dimension), worse (deterioration in at least one dimension), mixed (both improvements and deteriorations in different dimensions), or unchanged. To visualize these changes, the Health Profile Grid (HPG) was used, which ranks health states from 1 (best) to 243 (worst) based on severity [23]. The HPG illustrates whether health has improved or worsened by plotting patients’ rankings at baseline and final follow-up: points above the 45° line indicate improvement, while points below indicate decline. The distance of a point from the 45° line reflects the magnitude of change, with patients on the line showing no change. Following the conversion of the health profiles to single indices, linear regression was conducted to identify predictors of EQ-5D at the final follow-up. In addition, the patients were categorized based on the direction of change of the EQ-5D single index. A paired t-test was conducted to compare pre-and post-SRS EQ-5D indices. Subsequently, logistic regression was performed to uncover predictors of EQ-5D change. These analyses were conducted using the “eq5d” and “stats” packages in R Fig. 1.

Health Profile Grid (HPG) depicting EQ-5D changes in individuals between baseline and final follow-up. The further a point is above the 45° line, the greater the improvement in an individual’s health. Conversely, the further below the line, the point is, the more their health has deteriorated. Those individuals on the line show “no change.”