Survival and hematologic toxicity after whole-brain radiotherapy, photon craniospinal irradiation, and proton craniospinal irradiation for solid-tumor leptomeningeal disease: a systematic review

Abstract

Background

Leptomeningeal disease (LMD) from solid tumors carries poor prognosis and substantial neurologic morbidity. Radiotherapy is used for palliation and disease control, but outcomes across whole-brain radiotherapy (WBRT), photon craniospinal irradiation (CSI), and proton CSI remain incompletely defined.

Methods

We conducted a PRISMA-based systematic review of PubMed-indexed studies reporting radiotherapy outcomes for adult solid-tumor LMD. Studies were categorized as photon WBRT/non-CSI, photon CSI, or proton CSI. Primary outcomes were median overall survival (OS) and progression-free survival (PFS), including CNS-PFS or intracranial PFS when reported. Grade ≥ 3 hematologic and non-hematologic toxicities were extracted when available.

Results

Thirty-two unique studies/reports were included: 21 photon WBRT/non-CSI, 6 photon CSI, and 6 proton CSI studies. The median of reported median OS values were 4.5 months for photon WBRT/non-CSI, 4.1 months for photon CSI, and 10.7 months for proton CSI. Median PFS was 3.9 months for photon WBRT/non-CSI and 4.6 months for proton CSI; no extractable median PFS was available for photon CSI. Among studies with extractable denominators, crude pooled grade ≥ 3 hematologic toxicity was 2.8% for photon WBRT/non-CSI, 80.0% for photon CSI, and 59.4% for proton CSI; however, toxicity reporting was limited and heterogeneous.

Conclusion

Proton CSI cohorts had numerically longer reported survival, but cross-study comparisons are limited by heterogeneity, treatment-era and selection differences, and systemic therapy exposure. These findings do not establish superiority over photon CSI and warrant prospective comparative evaluation.

Introduction

Leptomeningeal disease (LMD) represents dissemination of malignant cells into the leptomeninges and cerebrospinal fluid (CSF) pathways and is an increasingly recognized complication of advanced solid tumors, particularly breast cancer, non-small cell lung cancer (NSCLC), melanoma, and gastrointestinal malignancies [1–4]. Although historically considered a terminal event with median survival often measured in weeks to a few months, outcomes are heterogeneous and increasingly influenced by tumor biology, performance status, extracranial disease control, molecular alterations, and availability of central nervous system (CNS)-active systemic therapies [1–5]. Patients may present with multifocal neurologic symptoms involving the cerebral hemispheres, cranial nerves, spinal cord, and cauda equina, including headache, nausea, cognitive changes, diplopia, gait instability, radiculopathy, seizures, and focal neurologic deficits [1, 2, 6]. Because LMD affects the entire neuraxis and may coexist with parenchymal brain metastases, bulky nodular disease, hydrocephalus, or systemic progression, treatment decisions are frequently complex and require multidisciplinary evaluation.

Current management of solid-tumor LMD remains individualized and is guided by clinical status, CSF cytology or molecular testing, imaging findings, primary tumor histology, systemic disease burden, and goals of care [1, 2, 7]. Treatment strategies may include CNS-active systemic therapy, intrathecal therapy in selected patients with linear or floating CSF disease, focal radiotherapy to symptomatic or bulky sites, whole-brain radiotherapy (WBRT), craniospinal irradiation (CSI), and best supportive care with or without CSF diversion procedures [1, 2, 7–9]. Contemporary intrathecal approaches, including HER2-directed therapy in selected patients with HER2-positive breast cancer, further emphasize the increasingly multimodal management of LMD and may be used alongside or, in selected settings, instead of neuraxis-directed radiotherapy.

Radiotherapy has historically played an important palliative role in LMD, particularly for symptomatic sites of disease, bulky nodular involvement, obstructive CSF flow, cranial neuropathies, spinal radiculopathy, and coexisting brain metastases [1, 7, 8, 10]. WBRT has commonly been used when intracranial symptoms, diffuse cranial meningeal enhancement, or multiple parenchymal brain metastases are present, but it does not treat the full craniospinal CSF compartment and is generally considered palliative in nature [7, 8, 10]. Photon CSI provides broader coverage of the entire craniospinal axis, theoretically addressing the compartmental nature of LMD; however, its use in adults has historically been limited by treatment-related toxicity, including myelosuppression, gastrointestinal toxicity, fatigue, and challenges with treatment tolerance in patients with advanced systemic disease [8, 11, 12].

Proton craniospinal irradiation has renewed interest in neuraxis-directed radiotherapy for selected patients with solid-tumor LMD. Because of the physical dose distribution of proton therapy, proton CSI can reduce exit dose to anterior organs and potentially spare bone marrow, gastrointestinal structures, heart, lungs, and other normal tissues compared with conventional photon CSI [11–13]. Early prospective and retrospective studies suggest that proton CSI may be feasible in selected patients and may provide durable CNS disease control in some cohorts [13–15]. More recently, randomized phase II data demonstrated improved CNS progression-free survival and overall survival with proton CSI compared with photon involved-field radiotherapy in patients with breast cancer or non-small cell lung cancer and LMD, without a significant increase in serious treatment-related adverse events [15]. Importantly, these data do not establish superiority of proton CSI over modern photon CSI, for which direct comparative evidence remains limited.

Despite this evolving evidence base, the literature remains heterogeneous. Published studies differ in patient selection, tumor histology, performance status, systemic therapy exposure, radiotherapy technique, dose and fractionation, outcome definitions, follow-up duration, and toxicity reporting. Prior reviews have evaluated radiotherapy for LMD broadly, or CSI specifically, but contemporary modality-stratified synthesis across photon WBRT, photon CSI, and proton CSI in adult solid-tumor LMD remains limited [8, 11, 16]. A focused systematic review is therefore needed to contextualize survival, intracranial or CNS disease control, and treatment-related toxicity across these radiotherapy approaches. In this systematic review, we synthesize the available literature on radiotherapy for solid-tumor LMD, stratifying outcomes by photon WBRT/non-CSI approaches, photon CSI, and proton CSI with emphasis on median overall survival (OS), progression-free survival (PFS), CNS PFS, hematologic toxicity, and non-hematologic adverse events.

Methods

Study design and reporting framework

This systematic review was conducted and reported in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses 2020 statement [17]. The objective was to synthesize published clinical outcomes after radiotherapy for LMD from solid tumors, with outcomes stratified by radiation modality. The modalities of interest were WBRT, photon CSI, and proton CSI. Given the anticipated heterogeneity in study design, patient selection, tumor histology, performance status, radiotherapy technique, dose and fractionation, systemic therapy exposure, and outcome definitions, findings were summarized descriptively rather than through formal quantitative meta-analysis, consistent with guidance for evidence synthesis when statistical pooling is not appropriate [18]. Cross-modality comparisons were considered descriptive and hypothesis-generating rather than estimates of comparative treatment efficacy.

Search strategy and information sources

A systematic PubMed search identified studies evaluating radiotherapy for solid-tumor leptomeningeal disease (LMD) using terms related to LMD, craniospinal irradiation (CSI), proton/photon CSI, whole-brain radiotherapy (WBRT), and radiation therapy. Reference lists and relevant online records were also manually screened.

The search identified 106 records; after removing 15 duplicates, 91 were screened and 79 underwent full-text review. Forty-seven were excluded for unavailable extractable survival outcomes, mixed adult/pediatric populations without adult-specific data, non-solid-tumor LMD, or dosimetry-only designs. Overall, 32 studies/reports met eligibility criteria, representing 33 modality-specific contributions because one comparative study included both proton and photon CSI arms (Fig. 1).

Fig. 1.

PRISMA 2020 flow diagram. The PubMed search identified 106 records, of which 15 duplicate records were removed. Ninety-one records were screened by title and abstract, and 79 full-text reports were assessed for eligibility. Forty-seven full-text reports were excluded because of no extractable survival outcomes, mixed adult and pediatric populations without adult-specific survival data, non-solid tumor leptomeningeal disease cohorts, or dosimetry-only study design. Overall, 32 unique studies/reports were included in the systematic review. The included evidence comprised 21 photon WBRT/non-CSI studies/reports, 6 photon CSI studies/reports, and 6 proton CSI studies/reports. One included report contributed both proton CSI and photon CSI arms; therefore, modality-specific counts sum to 33, although 32 unique studies/reports were included

Eligibility criteria

Eligible studies included adults aged ≥ 18 years with solid-tumor leptomeningeal disease treated with photon WBRT/non-CSI, photon CSI, or proton CSI and reporting at least one extractable survival or disease-control outcome. Prospective trials, retrospective cohorts, institutional series, and comparative observational studies were included. Reviews, editorials, case reports without cohort-level outcomes, abstracts without sufficient data, dosimetry-only studies, pediatric cohorts, hematologic malignancies, and non-extractable mixed populations were excluded. Toxicity-only studies were excluded; toxicity was analyzed secondarily among otherwise eligible studies.

Study selection

Titles and abstracts were screened for relevance, followed by full-text assessment of potentially eligible studies. Eligibility was determined according to the predefined inclusion and exclusion criteria. Uncertain eligibility decisions were resolved through full-text review and discussion among the study team. The study selection process was summarized using a PRISMA 2020 flow diagram [19].

Data extraction

Data were extracted using a standardized data collection template. Extracted variables included first author, year of publication, study design, country or institution, tumor histology, LMD definition or diagnostic method, sample size, age, sex when available, performance status, presence of parenchymal brain metastases, radiotherapy modality, radiation dose and fractionation, use of systemic therapy or intrathecal therapy when reported, follow-up duration, survival outcomes, progression outcomes, and treatment-related toxicity.

Radiotherapy modality was categorized as photon WBRT/non-CSI, photon CSI, and proton CSI. For studies reporting more than one extractable treatment arm, histology-specific cohort, or modality-specific subgroup, each arm or subgroup was extracted as a separate analytic row when outcomes were reported separately. Unique studies were counted once in the PRISMA flow diagram, while modality-specific arms were extracted separately when outcomes were reported separately.

Outcomes

The primary outcome was median overall survival (OS), defined using each study’s reported index date. Secondary outcomes included PFS, CNS-PFS, intracranial PFS, leptomeningeal PFS, and time to local failure, summarized using the original study terminology. Because progression-related endpoints were variably defined and reported by different subsets of studies, these outcomes were summarized independently and were not directly compared with the modality-specific OS estimates.

Toxicity outcomes included grade ≥ 3 hematologic and non-hematologic adverse events. Hematologic toxicities included lymphopenia, anemia, neutropenia or leukopenia, and thrombocytopenia and were summarized both overall and by subtype when extractable. Crude rates were calculated by radiotherapy modality when numerator and denominator data were available. Studies lacking grade-specific counts or evaluable denominators were excluded from pooled calculations but retained descriptively. Toxicity was coded as absent when explicitly reported and as not reported when insufficiently described.

Data synthesis and statistical analysis

A descriptive synthesis was performed because of heterogeneity in study design, patient populations, tumor histology, radiotherapy technique, treatment era, systemic therapy exposure, and outcome definitions. Study-level median OS and progression outcomes were summarized as the median of reported medians with ranges by modality. Crude event rates for overall and subtype-specific grade ≥ 3 hematologic toxicity were calculated from studies with extractable denominators. Non-hematologic toxicities were summarized descriptively. Formal comparative testing was not performed because of substantial clinical and methodological heterogeneity and the absence of randomized comparisons across all modalities. Accordingly, numerical differences between modality groups were interpreted descriptively and not as evidence of comparative efficacy or safety.

Risk-of-bias assessment

No formal risk-of-bias scoring tool was applied. Methodological limitations were assessed descriptively, including patient selection, confounding, outcome ascertainment, follow-up completeness, and selective reporting. Particular attention was given to differences in these factors across radiotherapy modalities because they could influence observed survival and toxicity independently of radiation technique. Accordingly, cross-modality comparisons were considered descriptive and hypothesis-generating rather than confirmatory.

Results

Study selection

The PubMed search identified 106 records. After removal of 15 duplicate records, 91 records underwent title and abstract screening. Twelve records were excluded at the screening stage, and 79 full-text reports were reviewed for eligibility. Forty-seven full-text reports were excluded because they did not report extractable survival outcomes, included mixed adult and pediatric populations without adult-specific survival data, evaluated non-solid tumor leptomeningeal disease, or represented dosimetry-only studies without clinical outcomes. Overall, 32 studies/ reports met eligibility criteria and were included in the systematic review. Of these, 21 evaluated photon WBRT or non-CSI photon radiotherapy approaches [19–39], 6 evaluated photon CSI [40–45], and 6 evaluated proton CSI [13–15, 40, 46, 47]. One comparative report contributed both proton CSI and photon CSI arms; therefore, the modality-specific counts sum to 33 although 32 unique studies/reports were included.

Study characteristics

Included studies varied in design, histology, patient selection, radiotherapy technique, systemic therapy exposure, and outcome definitions. Most were retrospective. WBRT/non-CSI studies largely evaluated palliative cranial radiotherapy, whereas CSI cohorts used neuraxis-directed treatment; proton CSI studies were generally more contemporary and included more selected populations. Radiotherapy dose and fractionation varied across modalities and studies. Photon WBRT was most commonly delivered as 30 Gy in 10 fractions, although regimens ranged from 20 Gy in 5 fractions to more protracted schedules. Photon CSI generally ranged from approximately 30–36 Gy, often using 15–20 fractions in earlier series, whereas contemporary proton CSI studies predominantly used approximately 30 Gy(RBE) in 10 fractions (Table 1).

Table 1.

Characteristics of included studies/reports and modality-specific analytic cohorts

Ref.	Study	Year	Modality	Design	Standardized disease population	N analyzed	Key concurrent/associated therapy	Radiotherapy dose/ fractionation
[19]	Gani et al.	2012	Photon WBRT/non-CSI	Retrospective cohort	Solid-tumor LMD	27	None reported	NR
[20]	Sakaguchi et al.	2017	Photon WBRT/non-CSI	Retrospective cohort	Solid-tumor LMD	71	None reported	30 Gy/10 fx
[21]	Du et al.	2013	Photon WBRT/non-CSI	Retrospective cohort	Solid-tumor LMD	21	Systemic therapy varied	1.8–3 Gy/fx; total dose up to 40 Gy
[22]	Katano et al.	2025	Photon WBRT/non-CSI	Retrospective cohort	Solid-tumor LMD subgroup	17	NR	30 Gy/10 fx or 20 Gy/5 fx
[23]	Sause et al.	1988	Photon WBRT/non-CSI	Prospective cooperative-group study	Solid-tumor LMD	26	Intrathecal methotrexate	30 Gy/10 fx
[24]	Piña et al.	2025	Photon WBRT/non-CSI	Phase Ib prospective trial	Solid-tumor LMD	15	Avelumab	30 Gy/10 fx
[25]	Okada et al.	2020	Photon WBRT/non-CSI	Retrospective cohort	Breast cancer LMD	31	NR	30 Gy/10 fx or 37.5 Gy/15 fx
[26]	Niwińska et al.	2018	Photon WBRT/non-CSI	Retrospective cohort	Breast cancer LMD	45	None reported	30 Gy/10 fx
[27]	Niwińska et al.	2015	Photon WBRT/non-CSI	Prospective observational study	Breast cancer LMD	56	Intrathecal methotrexate or liposomal cytarabine	NR
[28]	Yust-Katz et al.	2013	Photon WBRT/non-CSI	Retrospective cohort	Breast cancer LMD	103	Systemic/intrathecal therapy varied	NR
[29]	Kingston et al.	2017	Photon WBRT/non-CSI	Retrospective cohort	Breast cancer LMD	182	Systemic/intrathecal therapy varied	NR
[30]	Morris et al.	2012	Photon WBRT/non-CSI	Retrospective cohort	NSCLC LMD	125	Systemic/intrathecal therapy varied	30–37.5 Gy/10–15 fx
[31]	Ozdemir et al.	2016	Photon WBRT/non-CSI	Retrospective cohort	NSCLC LMD	51	Systemic therapy varied	30 Gy/10 fx
[32]	Liao et al.	2015	Photon WBRT/non-CSI	Retrospective cohort	NSCLC LMD	212	EGFR-TKI therapy evaluated	NR
[33]	Xu et al.	2015	Photon WBRT/non-CSI	Retrospective cohort	NSCLC LMD	108	Systemic therapy varied	30 Gy/10 fx
[34]	Li et al.	2016	Photon WBRT/non-CSI	Retrospective cohort	EGFR-mutant NSCLC LMD	160	EGFR-TKI therapy evaluated	NR
[35]	Hirano et al.	2020	Photon WBRT/non-CSI	Retrospective cohort	Solid-tumor LMD	22	Steroids and/or intrathecal methotrexate in selected patients	Most commonly 30 Gy/10 fx; range included 3–40 Gy/1–20 fx
[36]	Zhen et al.	2020	Photon WBRT/non-CSI	Retrospective cohort	NSCLC LMD	38	Targeted/systemic therapy varied	30 Gy/15 fx or 36 Gy/18 fx
[37]	Yan et al.	2019	Photon WBRT/non-CSI	Retrospective comparative cohort	EGFR-mutant NSCLC LMD	26	EGFR-TKI and/or chemotherapy varied	30 Gy/10 fx
[38]	Lee et al.	2013	Photon WBRT/non-CSI	Retrospective cohort	NSCLC LMD	149	Systemic/intrathecal therapy varied	Mostly 30 Gy/5–10 fx
[39]	Oyoshi et al.	2023	Photon WBRT/non-CSI	Retrospective cohort	Lung adenocarcinoma LMD	80	EGFR/ALK-directed therapy evaluated	30 Gy/10 fx (72 patients) or 20 Gy/5 fx (8 patients)
[41]	Hermann et al.	2001	Photon CSI	Retrospective cohort	Solid-tumor LMD	6	None reported	Median 36 Gy; 1.6–2.0 Gy/fx
[41]	Hermann et al.	2001	Photon CSI	Retrospective cohort	Solid-tumor LMD	10	Intrathecal methotrexate	Median 36 Gy; 1.6–2.0 Gy/fx
[42]	El Shafie et al.	2019	Photon CSI	Retrospective cohort	Solid-tumor-predominant LMD	25	Systemic therapy after CSI in selected patients	Median 35.2 Gy; most commonly 36 Gy/20 fx
[43]	Devecka et al.	2020	Photon CSI	Retrospective cohort	Breast cancer LMD subgroup	5	No concomitant systemic therapy during RT	Median 30.6 Gy (range 3–36 Gy); median 1.6 Gy/fx
[43]	Devecka et al.	2020	Photon CSI	Retrospective cohort	NSCLC LMD subgroup	5	No concomitant systemic therapy during RT	Median 30.6 Gy (range 3–36 Gy); median 1.6 Gy/fx
[44]	Takeda et al.	2024	Photon CSI	Retrospective cohort	Solid-tumor LMD	25	Systemic therapy varied	Median 30 Gy (IQR 24–31); most commonly 30 Gy/15 fx
[45]	Keit et al.	2024	Photon CSI	Retrospective cohort; abstract	Solid-tumor LMD subgroup	9	Intrathecal chemotherapy in selected patients	IMRT-based CSI; dose/fractionation NR
[40]	Lamanilao et al.	2025	Photon CSI	Retrospective comparative cohort; abstract	Solid-tumor LMD	16	Systemic therapy varied	30 Gy/10 fx
[15]	Yang et al.	2025	Proton CSI	Randomized clinical trial	Breast cancer or NSCLC LMD	42	Systemic therapy status used for stratification	30 Gy in 10 fractions
[46]	Gal et al.	2025	Proton CSI	Retrospective cohort	Solid-tumor LMD	38	Concurrent systemic therapy in selected patients	30 Gy(RBE)/10 fx; modified 18–25 Gy(RBE)/6–10 fx in previously irradiated regions
[13]	Yang et al.	2021	Proton CSI	Prospective phase Ib trial	Solid-tumor LMD	24	Systemic therapy generally held during treatment	30 Gy(RBE)/10 fx
[14]	Lam et al.	2025	Proton CSI	Retrospective cohort	Solid-tumor LMD	45	Limited concurrent systemic therapy	30 Gy(RBE)/10 fx in 39/45 patients; other regimens in 6/45
[47]	Ehret et al.	2025	Proton CSI	Retrospective cohort	Solid-tumor LMD	9	Systemic therapy varied	Median 30 Gy(RBE)/10 fx
[40]	Lamanilao et al.	2025	Proton CSI	Retrospective comparative cohort; abstract	Solid-tumor LMD	21	Systemic therapy varied	30 Gy/10 fx

N analyzed refers to the extractable LMD cohort, treatment arm, or subgroup used for the modality-specific synthesis, rather than the overall institutional cohort when these differed. For studies reporting multiple treatment arms, subgroups, or sensitivity cohorts, values shown in the table reflect only the prespecified eligible arm or subgroup used for the modality-specific analysis. One comparative report contributed both proton CSI and photon CSI treatment arms and is cited once in the reference list; therefore, modality-specific counts sum to 33, although 32 unique studies/reports were included. Abbreviations CSI, craniospinal irradiation; EGFR, epidermal growth factor receptor; LMD, leptomeningeal disease; NR, not reported; NSCLC, non-small cell lung cancer; RT, radiotherapy; TKI, tyrosine kinase inhibitor; WBRT, whole-brain radiotherapy; FX, fraction

Although 32 unique studies/reports were included, some contributed multiple extractable arms or subgroups. Outcomes were therefore summarized by unique studies and analytic rows. WBRT/non-CSI included 21 studies/21 rows, proton CSI 6/6, and photon CSI 6 studies/8 rows (Table 1).

Overall survival

Median survival (OS) was reported in 21 photon WBRT/non-CSI rows [19–39], 8 photon CSI analytic rows [40–45] and 6 proton CSI analytic rows [13–15, 40, 46, 47]. The median reported OS was 10.7 months for proton CSI, with reported study-level medians ranging from 4.0 to 13.7 months. In comparison, the median reported OS was 4.5 months for photon WBRT/non-CSI [19–39], with reported medians ranging from 1.9 to 14.5 months, and 4.1 months for photon CSI [40–45], with reported medians ranging from 1.9 to 4.8 months (Fig. 2). These values represent descriptive summaries of heterogeneous study-level cohorts rather than direct comparisons between radiotherapy modalities.

Fig. 2.

Median overall survival by study and radiotherapy modality. Forest-style plot showing reported median overall survival, in months, across studies evaluating radiotherapy for solid-tumor leptomeningeal disease. Points represent the reported median OS for each study or analytic cohort, and horizontal lines represent the corresponding 95% confidence intervals. Only studies or analytic cohorts with extractable median OS and corresponding 95% confidence intervals were included in this plot. Studies are grouped by radiotherapy modality: photon WBRT/non-CSI, photon CSI, and proton CSI

Progression-free survival and CNS disease control

Progression-related outcomes were reported less consistently than overall survival and varied in terminology across studies. Median PFS was reported in 3 photon WBRT/non-CSI rows and 2 proton CSI rows, with no extractable median PFS reported in the photon CSI group. Among studies reporting median PFS, the median of reported values was 3.9 months for photon WBRT/non-CSI and 4.6 months for proton CSI.

CNS-PFS, intracranial PFS, leptomeningeal PFS, or related CNS disease-control endpoints were also variably reported. The median of reported CNS-PFS or intracranial PFS values was 6.8 months for photon WBRT/non-CSI, 7.2 months for photon CSI, and 7.0 months for proton CSI. Because these estimates were derived from different study subsets and variably defined endpoints, they should not be directly compared with the modality-specific OS summaries.

Hematologic toxicity

Among studies with extractable toxicity denominators, crude pooled grade ≥ 3 hematologic toxicity was lowest with photon WBRT/non-CSI at 2.8% (4/145), compared with 80.0% (20/25) for photon CSI and 59.4% (38/64) for proton CSI. Lymphopenia was the dominant reported hematologic toxicity in CSI-based studies, occurring in 80.0% (20/25) of photon CSI patients and 58.6% (34/58) of proton CSI patients with extractable data. In contrast, grade ≥ 3 lymphopenia was reported in 26.7% (4/15) of evaluable photon WBRT/non-CSI patients, although hematologic toxicity reporting in this group was limited.

Grade ≥ 3 anemia occurred in 6.7% (1/15) of evaluable photon WBRT/non-CSI patients, 12.0% (3/25) of photon CSI patients, and 3.1% (2/64) of proton CSI patients. Grade ≥ 3 neutropenia or leukopenia occurred in 12.0% (3/25) of photon CSI patients and 6.9% (4/58) of proton CSI patients. Grade ≥ 3 thrombocytopenia occurred in 6.7% (1/15) of evaluable photon WBRT/non-CSI patients, 8.0% (2/25) of photon CSI patients, and 1.6% (1/64) of proton CSI patients (Fig. 3). Toxicity denominators varied substantially because of inconsistent reporting across studies.

Fig. 3.

Crude pooled rates of overall and subtype-specific grade ≥ 3 hematologic toxicities are shown for photon WBRT/non-CSI, photon CSI, and proton CSI among studies with extractable denominators. Values were calculated from available study-level data; no formal meta-analysis was performed. WBRT, whole-brain radiotherapy; CSI, craniospinal irradiation; NR, not reliably pooled

Discussion

In this systematic review of 32 unique studies/reports evaluating radiotherapy for solid-tumor LMD, we found substantial heterogeneity in study design, patient selection, radiotherapy modality, systemic therapy exposure, endpoint definitions, and toxicity reporting. Despite these limitations, several important observations emerged. The median of reported study-level median OS values was 10.7 months for proton CSI, 4.5 months for photon WBRT/non-CSI, and 4.1 months for photon CSI. However, these cross-study differences should be interpreted cautiously because proton CSI cohorts generally represented more contemporary and selected populations. Progression-related endpoints were inconsistently reported, limiting comparisons of CNS disease control. Grade ≥ 3 hematologic toxicity was reported more frequently in CSI cohorts, although reporting was incomplete and heterogeneous.

LMD remains one of the most challenging complications of advanced solid tumors. It is associated with neurologic morbidity, impaired quality of life, and historically poor survival. Contemporary guidelines emphasize that management should be individualized according to performance status, neurologic burden, CSF flow, radiographic disease pattern, primary tumor biology, systemic disease status, and availability of CNS-active systemic therapies [1, 2]. In this context, radiotherapy has traditionally been used for symptom palliation, bulky or nodular disease, cranial neuropathies, spinal radiculopathy, obstructive CSF flow, and coexisting brain metastases rather than as a uniformly applied disease-modifying treatment [1, 8, 9]. Our findings align with this clinical reality: the WBRT/non-CSI literature was the largest and most heterogeneous group, reflecting decades of palliative cranial radiation practice in patients with variable disease burden and prognosis.

The numerically longer median OS observed in proton CSI cohorts is consistent with the growing interest in comprehensive neuraxis-directed radiotherapy for selected patients with LMD. Unlike WBRT or focal involved-field radiotherapy, CSI treats the entire craniospinal CSF compartment, which is biologically appealing in a disease characterized by diffuse CSF dissemination. These findings should be interpreted in the context of modern prospective data. In the randomized phase II trial by Yang and colleagues, proton CSI improved CNS PFS (8.2 vs. 2.3 months) and OS (11.3 vs. 4.9 months) compared with photon involved-field radiotherapy, with a multivariable OS HR of 0.50 in patients with non-small cell lung cancer or breast cancer LMD, without an increase in serious treatment-related adverse events [15]. This trial provides the prospective support for proton CSI in selected patients, although the comparator was involved-field radiotherapy rather than photon CSI. In contrast, Frechette et al. reported a median OS of 3.0 months with proton CSI in a real-world cohort, underscoring the substantial influence of patient selection and treatment context on observed survival [16]. Photon CSI remains an important modality but is represented by fewer and generally older studies with heterogeneous techniques, dose schedules, and patient populations. In this review, photon CSI had a median reported OS of 4.1 months, similar to the photon WBRT/non-CSI group and shorter than proton CSI. This should not be interpreted as definitive inferiority of photon CSI, because the photon CSI literature includes older treatment eras, less conformal techniques, and potentially more heavily pretreated or clinically fragile patients. Nevertheless, these findings reinforce why photon CSI has not been widely adopted for adult solid-tumor LMD: comprehensive neuraxis treatment may be conceptually attractive, but toxicity and tolerability remain major concerns when using photon-based techniques [11, 12].

The role of WBRT in LMD remains nuanced. WBRT can palliate intracranial symptoms and treat coexisting brain metastases, but it does not comprehensively address spinal CSF disease or the full neuraxis. In this review, the WBRT/non-CSI group had a median reported OS of 4.5 months, which is consistent with the historically poor prognosis of LMD reported in prior reviews [11, 12]. Our findings extend this prior literature by stratifying outcomes across WBRT/non-CSI, photon CSI, and proton CSI approaches and by incorporating more contemporary evidence across neuraxis-directed radiotherapy approaches.

Toxicity interpretation was limited by inconsistent reporting across included studies. Among studies with extractable denominators, crude pooled grade ≥ 3 hematologic toxicity was 2.8% (4/145) for photon WBRT/non-CSI, 80.0% (20/25) for photon CSI, and 59.4% (38/64) for proton CSI. Lymphopenia was the predominant hematologic toxicity in CSI-based studies, occurring in 80.0% (20/25) of photon CSI patients and 58.6% (34/58) of proton CSI patients with extractable data. Other grade ≥ 3 hematologic toxicities, including anemia, neutropenia or leukopenia, and thrombocytopenia, were reported less frequently and inconsistently across modalities but may be more clinically consequential when considering subsequent systemic therapy. The higher crude hematologic toxicity observed in photon CSI cohorts should not be interpreted as an intrinsic disadvantage of photon therapy, as these studies largely reflect older techniques, more protracted fractionation, and greater vertebral marrow exposure. Contemporary marrow-sparing photon CSI approaches may reduce this toxicity burden. However, interpretation remains limited by small denominators, variable grading criteria, incomplete reporting, and potential confounding from prior chemotherapy or concurrent systemic therapy. Although not eligible for quantitative inclusion because it did not report a prespecified survival or progression endpoint, Perlow et al. reported no grade ≥ 3 toxicity with vertebral body–sparing photon CSI, supporting the possibility that hematologic toxicity may be technique-dependent [48].

This review has several strengths. It focuses specifically on solid-tumor LMD and stratifies outcomes by clinically relevant radiotherapy modality. It separates unique study/reports counts from analytic rows, thereby avoiding inflation of PRISMA study counts while preserving modality-specific outcome extraction. It also summarizes both survival and toxicity, which is essential in LMD because treatment decisions often require balancing potential disease control against symptom burden, treatment time, and toxicity. Finally, by placing proton CSI, photon CSI, and WBRT/non-CSI literature in the same descriptive framework, this review provides a clinically useful benchmark for interpreting contemporary and future LMD radiotherapy studies. Ongoing prospective efforts will be important to validate these modality-stratified findings. In particular, NRG-BN014/RADIATE-LM (NCT06500481) [49], a phase III trial evaluating proton CSI versus involved-field radiotherapy in patients with breast cancer or NSCLC leptomeningeal metastasis, is expected to provide higher-level evidence regarding the role of proton CSI in selected solid-tumor LMD populations. However, because the comparator is involved-field radiotherapy rather than photon CSI, the relative efficacy of proton versus modern photon CSI will remain unresolved.

Several limitations should be noted. Only PubMed was searched, and no formal risk-of-bias assessment was performed. Most included studies were retrospective and vulnerable to selection bias, confounding, and incomplete follow-up. The absence of patient-level data prevented adjustment for key clinical factors. Progression endpoints were inconsistently defined, and toxicity reporting was incomplete, particularly in older WBRT and photon CSI studies. Accordingly, these modality-based comparisons should be considered descriptive and hypothesis-generating rather than definitive.

Conclusion

This systematic review highlights the evolving role of radiotherapy in solid-tumor leptomeningeal disease, with outcomes varying substantially across photon WBRT/non-CSI, photon CSI, and proton CSI approaches. Numerically longer survival was reported in proton CSI cohorts, although these cross-study differences do not establish superiority over photon CSI and are influenced by patient selection and treatment era. Among available studies, proton CSI showed the most favorable reported survival outcomes, while CSI-based treatment was associated with a greater hematologic toxicity burden than photon WBRT/non-CSI, likely influenced by treatment technique and reporting heterogeneity. The current evidence remains limited by non-randomized study designs, heterogeneous patient selection, variable endpoint definitions, and incomplete toxicity reporting. Future prospective trials should clarify the role of proton CSI in the modern treatment paradigm and incorporate standardized CNS disease-control endpoints, detailed modality-specific toxicity assessment, and integration with contemporary CNS-active systemic therapies.

Author contributions

M.M., K.S., and H.H. performed the literature search, study screening, data extraction, and data analysis. M.M. drafted the manuscript and prepared the tables and figures. M.H., M.P.M., J.T.Y., and R.K. contributed to the interpretation of the findings, provided clinical and methodological supervision, and critically revised the manuscript for important intellectual content. All authors contributed to the study conception and design, reviewed the manuscript, approved the final version, and agree to be accountable for the integrity of the work.

Funding

No specific funding was received for this study.

Data availability

No datasets were generated or analysed during the current study.

Declarations

Ethical approval

Institutional review board approval and informed consent were not required because this study is a systematic review of previously published literature and did not involve individual-level patient data.

Competing interests

The authors declare no competing interests.