The medical records of patients with NSCLC carrying EGFR mutations and diagnosed with BMs admitted to Cangzhou Hospital of Integrative Traditional Chinese and Western Medicine (TCM-WM) of Hebei between January 2018 and December 2022 were meticulously collated. The inclusion criteria were as follows: patients (1) with histopathologically verified lung adenocarcinoma; (2) with BMs, ascertained by magnetic resonance imaging (MRI); (3) with either exon 19 deletion (19DEL) or exon 21 points mutation (21L858R mutation), confirmed by genetic testing; and (4) who received osimertinib as an initial treatment regimen.

The exclusion criteria were as follows: patients (1) who previously used EGFR-TKI without receiving osimertinib during cranial RT or within one week after RT; (2) who underwent surgical resection at an initial BM diagnosis; and (3) with other serious illnesses that could affect their survival or cognitive function, such as stroke, Parkinson’s disease or severe liver or kidney dysfunction.

A total of 213 patients fitting these criteria were recruited for the study, and their baseline characteristics, namely age, gender, Karnofsky performance score (KPS), smoking history, EGFR mutation type, number of BMs, maximum BM diameter, BM symptoms and extracranial metastases, were documented. All the patients were evaluated using the diagnostic-specific grading prognostic score (DS-GPA) for BMs (15). The data collection cut-off date was 1 December 2022. The study received approval from the ethics committee of Cangzhou Hospital of Integrative TCM-WM of Hebei, and all patient-involved procedures complied with the principles of the Declaration of Helsinki. Informed consent was obtained from all the patients.

Patients were assigned to either the upfront cranial RT group (ucRT) or the non-RT group based on cranial RT application timing. The ucRT cohort comprised patients receiving cranial RT concurrent with a third-generation EGFR-TKI or patients commencing third-generation EGFR-TKI therapy after undergoing cranial RT (with an interval not exceeding one week). The term ‘upfront’ refers to treatment given as the first-line therapy for patients with BMs before the occurrence of any evidence of intracranial or systemic progression. The non-ucRT group included patients receiving osimertinib as their sole first-line treatment; all the patients were administered osimertinib at a standard daily dose of 80 mg. To reduce neurotoxicity and preserve cognitive function, cranial RT involved hippocampal-avoidance WBRT (HA-WBRT, 30 Gy/10 f; hippocampi doses: Dmax ≤ 17 Gy, Dmean ≤ 10 Gy and D40 ≤ 9 Gy) and and SRS (dose determined in line with the Rt Oncology Group 90-05 and 95-08 institutional guidelines) (16, 17). The former was predominantly used in patients with multiple BMs (BM number > 3) and patients not eligible for SRS (n = 36), whereas SRS was primarily utilised in patients with 1–3 BMs (n = 38).

The RT dose and plan were recorded as follows: Among the 213 patients in this study, 96 had oligometastatic brain disease (ucRT group: n = 47; non-ucRT group: n = 49). All 47 patients in the ucRT group underwent SRS treatment, with the following clinical data:

In patients with 1 BM (n = 18), the median BM diameter was 0.85 cm (0.7–3 cm), with a mean of 1.3 cm. The brain metastasis volume was <1 cm³ in 12 patients, and 1.2 cm³, 5.9 cm³, 9.5 cm³, 10.4 cm³, 10.8 cm³ and 15.8 cm³ in the remaining 6 patients, respectively. The SRS dose fractionation scheme was: 21–23 Gy/1 F in the 12 patients with a BM diameter of <1 cm³, 18 Gy/1 F in the 1.2 cm³ lesion, 27 Gy/3 F in the 5.9 cm³ lesion and 30 Gy/5 F in the rest.

In patients with 2 BMs (n = 23), the median BM maximum diameter was 1.5 cm (0.6–2.9 cm), with a mean of 1.6 cm. The brain metastasis volume and the corresponding SRS dose fractionation scheme were: <1 cm³ in 19 lesions (20–23 Gy/1 F), 1–5 cm³ in 14 lesions (18–20 Gy/1 F or 27–30 Gy/3–5 F), 5–10 cm³ in 10 lesions (27–30 Gy/3–5 F) and >10 cm³ in 3 lesions (30 Gy/5 F).

In patients with 3 BMs (n = 6), the median BM maximum diameter was 1.2 cm (0.6–2.8 cm), with a mean of 1.5 cm. The brain metastasis volume and the corresponding SRS dose fractionation scheme were: <1 cm³ in 10 lesions (22–23 Gy/1 F), 1–5 cm³ in 3 lesions (20 Gy/1 F, or 27 Gy/3 F), 5–10 cm³ in 2 lesions (27–30 Gy/3–5 F), and >10 cm³ in 3 lesions (30 Gy/5 F).

For the ucRT group, systemic and intracranial response assessments were undertaken 1 month after RT completion and every 2–3 months subsequently. For the non-ucRT group, tumour response was routinely evaluated every 2–3 months after the commencement of the osimertinib treatment. The cut-off date for the follow-up was December 2022. In cases of suspected intracranial or systemic progression, cranial MRI or computed tomography were swiftly performed. Treatment response was evaluated in accordance with the Response Evaluation Criteria in Solid Tumours version 1.1.

The primary endpoints comprised OS, PFS and IPFS. The first, OS, was defined as the duration from the initiation of osimertinib treatment to either the patient’s death or the last known follow-up; PFS referred to the period from the commencement of osimertinib treatment to systemic progression, death from any cause or the last known follow-up; and IPFS referred to the time from the onset of osimertinib treatment to intracranial tumour progression, death from any cause or the last known follow-up. All the patients were subject to data review at the final follow-up visit.

Data analysis was executed using the SPSS software 27.0 and GraphPad Prism software 8.0. The baseline characteristics of patients in both groups were contrasted using the chi-square test or Fisher’s exact test, and OS, PFS and IPFS were calculated using the Kaplan–Meier method. Differences in survival curves between the two groups were assessed using the log-rank test. The association of various factors with survival was analysed using the Cox proportional risk model. In the univariate model, variables linked with survival (P < 0.10) were incorporated into the multivariate model, with a two-sided P-value of <0.05 denoting statistical significance.

To account for potential selection bias arising from the retrospective nonrandomised study design, differences in OS, PFS and IPFS between the two groups were analysed using inverse probability of treatment weighting (IPTW). The presence or absence of BM symptoms was not included as a covariate in the IPTW, as this factor represents an obvious indication for determining whether to perform cranial RT. All other baseline characteristics were incorporated into the covariates of IPTW.

In the present propensity score matching (PSM) process, the authors utilised a 1:1 nearest-neighbour-matching algorithm with a calliper width of 0.02 to align patients from the ucRT (unconsolidated RT) group with those from the non-ucRT group. This strict matching criterion was employed to ensure a high degree of similarity between the matched pairs in terms of their propensity scores. Despite the authors’ efforts to closely match patients from both groups, the inherent variability within the patient cohorts meant that not all individuals could be paired. This discrepancy in the final matched numbers reflects the stringent calliper setting, which prioritises the quality of the match over the quantity.

This study involved a cohort of 213 patients, 96 of whom were placed in the ucRT group, while 117 were allocated to the non-ucRT group. Following IPTW, 74 patients were placed in each group. The baseline attributes of the patients in both groups are outlined Table 1 . It can be seen that there were no significant disparities between the two groups in terms of age, sex, smoking history, KPS, EGFR mutation type, number of BMs and DS-GPA score. Notably, the ucRT group had a higher prevalence of patients with symptomatic BM (P < 0.05). The median follow-up time was 18.6 months (range: 3.2–48.5 months) for the ucRT group and 16.4 months (range: 2.7–46.3 months) for the non-ucRT group.

For the entire cohort, the median OS was 36.4 months. The ucRT and non-ucRT cohorts had a median OS of 37.6 and 36.2 months, respectively (P = 0.0716; Figure 1A ). In terms of PFS, the median for all patients was 16.2 months, with the ucRT and non-ucRT groups demonstrating 17.35 and 14.4 months, respectively (P = 0.6884; Figure 1B ). As for IPFS, the overall median was 27.3 months, with the ucRT and non-ucRT groups reporting 29.65 and 21.8 months, respectively (P < 0.0001; Figure 1C ).

All the patients underwent IPTW, after which survival analysis was conducted on 74 patients in each group. Here, age, gender, smoking history, KPS, EGFR mutation type, number of BMs, maximum BM diameter and extracranial metastasis status were balanced. After IPTW, there were no significant deviations in OS and PFS between the two groups (P = 0.1481 and P = 0.4643; Figures 2A, B ). However, IPFS was markedly higher in the ucRT group than in the non-ucRT group (P < 0.0001; Figure 2C ).

The patients were segregated into the multiple-BMs subgroup and the oligometastases (OLOGO-BMs) subgroup based on the number of BMs, with OLOGO-BM characterised as 1–3 BM lesions with a maximum diameter of ≤3.5 cm (maximum diameter of ≤3.0 cm in the ucRT group and a maximum diameter of ≤3.5 cm in the non-ucRT group) (18). The remaining cases were classified as having multiple BMs. Survival analysis was then performed on patients from both groups; in the multiple-BMs subgroup before IPTW the ucRT and non-ucRT groups demonstrated a median OS of 27.3 and 34.35 months, respectively (P = 0.0710; Figure 3A ); a median PFS of 13.7 and 13.15 months, respectively (P = 0.0516; Figure 3B ); and a median IPFS of 26.4 and 21.35 months, respectively (P = 0.0028; Figure 3C ). Meanwhile, in the OLOGO-BMs subgroup, the ucRT and non-ucRT groups had a median OS of 44.5 and 37.3 months, respectively (P < 0.0001; Figure 3D ); a median PFS of 32.3 and 20.8 months, respectively (P = 0.0011; Figure 3E ); and a median IPFS of 37.8 and 22.1 months, respectively (P < 0.0001; Figure 3F ). Furthermore, the Cox multifactorial analysis of OS and PFS in the OLOGO-BMs subgroup demonstrated that ucRT was linked with better OS (heart rate [HR] = 0.392; 95% confidence interval [CI]: 0.178–0.863; P = 0.020) and PFS (HR = 0.558; 95% CI: 0.316–0.986; P = 0.044) than seen in the non-ucRT group ( Supplementary Tables 1 , 2 ). These findings validate the effectiveness of ucRT in patients with OLOGO-BM.

Following IPTW, 38 patients in the ucRT group and 29 patients in the non-ucRT group were successfully matched in the OLOGO-BM subgroup. In the multiple-BMs subgroup, 36 patients in the ucRT group were successfully matched with 45 patients in the non-ucRT group. The discrepancies in covariates between the two groups are outlined in Table 2 . In the OLOGO-BM subgroup, all 38 patients in the ucRT group received SRS. Conversely, in the multiple-BMs subgroup, all 36 patients in the ucRT group underwent HA-WBRT. There were no significant post-IPTW disparities in OS and PFS between the two groups in the multiple-BMs subgroup (27.3 months and 34.9 months; P = 0.0890; Figure 4A ; 13.7 months and 13.2 months; P = 0.1356; Figure 4B ). Notably, IPFS was significantly higher in the ucRT group than in the non-ucRT group, with medians of 26.35 and 21.0 months, respectively (P = 0.0009; Figure 4C ). In the OLOGO-BM subgroup, post-IPTW, the ucRT group reported a significantly higher OS, PFS and IPFS compared with the non-ucRT group, with a median OS of 44.45 and 37.5 months, respectively (P < 0.0001; Figure 4D ); a median PFS of 31.95 and 20.8 months, respectively (P = 0.0104; Figure 4E ); and a median IPFS of 37.7 and 24.3 months, respectively (P = 0.0003; Figure 4F ). As can be seen from the forest map of OS and PFS of patients with oligocerebral metastasis, extracranial metastases and patients with maximum BM diameter > 10mm can benefit from non-ucRT in terms of OS. Patients with age < 60 and L858R mutation may have prolonged PFS from non-ucRT ( Figure 5 ).

Throughout the follow-up period, 156 patients (73.2%) registered either intracranial or extracranial tumour progression. Of these, the ucRT group exhibited a predominant first site of progression extracranially, whereas the non-ucRT group showed a significant trend towards intracranial progression as the initial site (P < 0.001; Figure 6 ).

The EGFR-TKI-related adverse reactions: The incidence of adverse reactions in the ucRT group and the non-ucRT group was 75.0% and 73.5%, respectively (P = 0.804). Among them, diarrhoea was the most common (45.8% vs 44.4%, P = 0.839), followed by rashes (44.8% vs 43.6%, P = 0.860), paronychia (32.3% vs 30.8%, P = 0.812), dry skin (31.3% vs 29.9%, P = 0.833), etc. Most were grade 1–2 (97.9% vs 98.3%, P = 1.000; Supplementary Table 3 ). Adverse reactions related to brain RT: The incidence of adverse reactions was 51.3% and 85.7% in the OLOGO-BM group and multiple-BMs groups, respectively (P < 0.001). The most common adverse reactions were dizziness (31.9% vs 53.1%, P = 0.036), headache (27.7% vs 51.1%, P = 0.019), radiation dermatitis (17.0% vs 47.3%, P = 0.002) and neurocognitive dysfunction (32% vs 85.7%, P < 0.001). The incidence of grade 3 and above adverse reactions was 2.1% and 4.1% (P = 1.000; Supplementary Table 4 ).