A 68-year-old chronic smoker presented with a 1-month history of progressive dizziness, memory decline, and appetite loss, initially misdiagnosed as Alzheimer’s disease at a local hospital. Upon admission to our institution in May 2022, neurological examination revealed lethargy and slowed cognition requiring wheelchair assistance, though physical examination showed no focal deficits. Diagnostic workup demonstrated significant abnormalities: serum CEA was markedly elevated (32.1 μg/L, normal ≤5.0 μg/L), while chest CT identified a 1.8 cm spiculated right upper lobe nodule with adjacent ground-glass opacity, highly suspicious for malignancy (Figures 1A,B). Brain MRI initially showed only nonspecific white matter changes (Figures 2A,B). Pathological analysis of CT-guided lung biopsy confirmed moderately differentiated adenocarcinoma with PD-L1 expression (TPS 10%) (Figure 1C) and a KRAS p. G12V mutation at 27.84% variant allele frequency. The lumbar puncture confirmed leptomeningeal involvement with an opening pressure of 220 mmH₂O and markedly abnormal CSF parameters including elevated albumin (278.4 mg/dL, normal ≤45.0 mg/dL), hypoglycorrhachia (2.43 mmol/L, normal 2.5–4.5 mmol/L), and elevated lactate dehydrogenase (60 U/L, normal 8–32 U/L) and lactate (5.8 mmol/L, normal 0.5–1.7 mmol/L). The cerebrospinal fluid smear examination showed clustered atypical cells, indicating a high possibility of malignancy (Figure 1D). These findings established the definitive diagnosis of stage IVB (pT1cN0M1b) KRAS p. G12V-mutated lung adenocarcinoma with leptomeningeal metastasis.

Initial systemic therapy beginning 7 May 2022 with pemetrexed (500 mg/m²), bevacizumab (7.5 mg/kg) and carboplatin (AUC 5) paradoxically exacerbated neurological symptoms, prompting urgent multidisciplinary evaluation. This led to placement of a ventriculoperitoneal shunt (Medtronic) and implantation of an Ommaya reservoir (RE-2021, SOPHYSA UBC) on 19 May 2022 to relieve intracranial hypertension and facilitate subsequent intrathecal therapy.

From May to September 2022, induction therapy comprised intrathecal pemetrexed (20 mg on days 1,8) with dexamethasone plus systemic tislelizumab (200 mg on day 1) and carboplatin (AUC 5 on day 1) in every 3 weeks (Q3W). This regimen stabilized pulmonary lesions (Figures 1E,F) and improved neurological status, though grade II thrombocytopenia emerged. Maintenance therapy from October 2022 onward utilized higher-dose intrathecal pemetrexed (40 mg) with intravenous tislelizumab, maintaining response without new toxicity until June 2023.

After 6 months of maintenance treatment, the patient developed worsening neurological symptoms, including unsteady gait and dizziness. A head MRI revealed multiple abnormal signal foci (Figures 2C,D). Due to confirmed intracranial metastasis, the patient underwent whole brain radiotherapy (WBRT) (3000 cGy/10 fx), but neurological symptoms persisted. CSF cytology detected suspicious cells (Supplementary material online, Supplementary Figure S1), and follow-up MRI showed progression of intracranial lesions, particularly in the left frontal lobe (Figures 2E,F).

Given disease progression, the patient received intrathecal chemotherapy with thiotepa (10 mg) and dexamethasone (5 mg) on days 1, 8, and 15 of a 4-week cycle, alongside intravenous tislelizumab and bevacizumab on day 1. After two cycles, CSF cytology showed no abnormal cells, symptoms improved significantly, and the intracranial lesions slightly regressed (Figures 2G,H). Genetic testing of CSF confirmed a KRAS p.G12V mutation, consistent with the prior lung pathology.

Following 10 cycles of intrathecal thiotepa, the patient developed thrombocytopenia (platelets: 42, normal range: 125–350 × 10^9/L)). Due to concerns over bone marrow suppression, the regimen was switched to intrathecal methotrexate at a reduced frequency—10 mg on days 1 and 4 every three weeks—while continuing intravenous tislelizumab and bevacizumab. The patient tolerated this adjusted regimen well, with stable disease and no new neurological symptoms after two cycles.

To date, the patient has survived for 29 months since LM diagnosis, maintaining a performance status of 1 and a good quality of life with stable pulmonary and intracranial disease on imaging (Figures 1G,H; Figures 2I,J). The treatment summary of the patient is shown in Figure 3.

The VPS and Ommaya reservoir have significantly facilitated intrathecal chemotherapy, eliminating his fear of lumbar puncture. Furthermore, the attentive care provided by the medical staff has greatly alleviated his anxiety. Consequently, he is overjoyed to see an improvement in his symptoms.

The clinical manifestations of LM patients are highly diverse, and onset is often insidious. Initial symptoms may include weakness, headaches, back pain, nausea, vomiting, and so on (Ozcan et al., 2023; Wang et al., 2022). In our case, the patient presented with were dizziness, loss of appetite, and memory decline, initially misdiagnosed as Alzheimer’s disease. Due to its nonspecific symptoms, LM is frequently misdiagnosed, delaying optimal treatment. While MRI abnormalities may support LM diagnosis in typical cases (Wang et al., 2022), CSF cytology remains the gold standard. Repeated CSF analyses can achieve up to 90% sensitivity (Thakkar et al., 2020). Our patient exhibited malignant cells in CSF, elevated ICP, and abnormal CSF profiles (elevated protein, low glucose, lymphocytic pleocytosis). These findings—high protein, hypoglycorrhachia, pleocytosis, and positive cytology—are hallmark features of LM (Ozcan et al., 2023).

Patients with LM face a poor prognosis, with median survival typically ranging from 1 to 4 months. However, certain subgroups, particularly those eligible for molecularly targeted therapies or with radiation-sensitive disease, may achieve better outcomes (Lamba et al., 2021). Current treatment strategies aim to stabilize neurological symptoms and prolong survival (Palmisciano et al., 2022), though standardized protocols remain lacking. The therapeutic arsenal includes whole neuraxial radiation therapy, systemic chemotherapy, and intrathecal administration of traditional chemotherapeutic agents. Notably, molecular targeted therapies have emerged as preferred first-line options for patients with specific actionable mutations. Osimertinib has demonstrated significant survival benefits in NSCLC patients with epidermal growth factor receptor (EGFR) mutations who develop LM (Yang et al., 2020; Lee et al., 2020). Similarly, the third-generation anaplastic lymphoma kinase (ALK) inhibitor Lorlatinib shows excellent central nervous system (CNS) penetration (Bauer et al., 2020). However, for NSCLC patients lacking targetable mutations, treatment options remain limited, highlighting an urgent need for more effective therapeutic approaches.

The KRAS mutation occurs in approximately 30% of human cancers (Hung et al., 2020), with most variants located at codon G12. Among these, the G12V subtype accounts for roughly 20% (Harris and Thawani, 2024). In gastric cancer, KRAS G12V is associated with poorer patient survival (Fu et al., 2019). Notably, Fukuda et al. demonstrated in a LCM mouse model that KRAS-G12V-driven, EGFR-mutant lung cancer cells developed Osimertinib resistance within the leptomeningeal space (Fukuda et al., 2021). Currently, no approved targeted therapies exist for KRAS G12V. However, preclinical studies suggest immunotherapy may hold promise (Zhang et al., 2024). In our case, both the patient’s tumor tissue and CSF tested positive for KRAS G12V, leaving no viable targeted treatment options.

The management of LM from NSCLC includes IC(4), which can be administered either through lumbar puncture or via an Ommaya reservoir. While lumbar puncture remains an option, it often results in poor patient compliance due to procedural discomfort and significant side effects. In contrast, the Ommaya reservoir - a subcutaneous implant with a catheter extending into the lateral ventricle - offers several advantages, including avoidance of repeated lumbar punctures, more consistent drug distribution, and the ability to both administer chemotherapy and sample CSF through a single access point (Zubair and De Jesus, 2024). However, many LM patients develop intracranial hypertension, which can lead to CSF leakage around the Ommaya device and clinical deterioration. To address this, VPS has been employed to effectively reduce ICP. Studies by Kim et al. demonstrate that VPS not only alleviates ICP-related symptoms but may also improve overall survival (Kim et al., 2019). The versatility of VPS allows its use in both communicating and non-communicating hydrocephalus (Kim et al., 2019). Recent evidence from Chen et al. suggests that aggressive shunt placement may be particularly beneficial for LM patients (Chen et al., 2023). The patient’s CSF opening pressure, measured via lateral decubitus lumbar puncture, was 220 mm H₂O, approximately equivalent to 16.18 mm Hg. A systematic review indicates that the normal range of lumbar CSF opening pressure is between 6.3 and 15.9 mmHg (Norager et al., 2021). Thus, the patient’s pressure clearly exceeded the upper limit of normal. Furthermore, several studies investigating intrathecal chemotherapy and CSF shunting have utilized an inclusion criterion of lumbar puncture opening pressure greater than 20 cm H₂O (approximately 14.7 mmHg) (Woo et al., 2022; Su et al., 2022). Therefore, based on the available evidence, we considered the measured CSF pressure of 220 mm H₂O clinically significant and indicative of elevated intracranial pressure, supporting the decision for shunt insertion. A critical consideration in treatment planning is that while VPS effectively controls ICP, it lacks a dedicated channel for chemotherapy administration. Clinical data from Huntoon et al. highlight the importance of combined approaches, showing that patients receiving both VPS and multiple IC sessions achieved significantly longer survival (11.7 months) compared to those without IC (2.8 months) (Huntoon et al., 2024). In the presented case, the combination of VPS with Ommaya reservoir implantation provided dual benefits of ICP control and preserved access for intrathecal therapy, illustrating the potential advantages of this integrated surgical approach.

The most commonly used IC drugs for LM include methotrexate, cytarabine, and thiotepa (Ozcan et al., 2023; Wang et al., 2022), although no single regimen has demonstrated clear superiority (Ozcan et al., 2023). Our review of clinical trials (Fan et al., 2024; Li H. et al., 2023; Fan et al., 2021; Miao et al., 2020; Ju et al., 2016; Pan et al., 2020; Pan et al., 2016) and case reports (Zhong et al., 2024; Li G. et al., 2023; Zheng et al., 2022; Ma et al., 2022; Li et al., 2020; Wang et al., 2018) (shown in Table 1) from the past decade reveals that pemetrexed has become the most frequently employed IC agent for LM secondary to lung cancer, with administered doses ranging from 10 mg to 50 mg. Dose-finding studies have provided important guidance: Pan et al. demonstrated that 10 mg of intrathecal pemetrexed given 1–2 times weekly with vitamin supplementation offered favorable efficacy with manageable toxicity in patients with recurrent LM from lung adenocarcinoma (Pan et al., 2019), while Fan et al.'s phase I study ultimately selected 50 mg as the optimal dose after evaluating a 15–80 mg range (Fan et al., 2024; Fan et al., 2021).

In our case, intrathecal pemetrexed (40 mg) demonstrated safety and efficacy, achieving >1 year of initial remission when combined with systemic therapy. However, at the time of recurrence, WBRT provided limited benefit. We subsequently transitioned to intrathecal chemotherapy with thiotepa, an alkylating agent known for its excellent CNS penetration and long-standing use since the 1970s. Recent evidence from Jamison et al. (2024) supports thiotepa as a viable alternative, given its favorable efficacy/safety profile. Our patient maintained nearly 1 year of remission following thiotepa treatment. However, moderate thrombocytopenia developed, necessitating a regimen change. The patient has now completed two cycles of intrathecal methotrexate, which was well-tolerated. Given its pharmacokinetic properties (including an 8–10.5 h half-life), methotrexate remains the first-line intrathecal agent for most solid tumors, particularly breast cancer (Jafari et al., 2024; Palmisciano et al., 2022).

We have invited international scholars to further discuss several issues regarding the diagnosis and treatment of the patient.

Question 1

Is early Ommaya reservoir implantation appropriate for newly diagnosed leptomeningeal metastasis(LM) patients without sensitive gene mutations?

Expert opinion 1: Dr. Rimas V. Lukas,MD, Lou & Jean Malnati Brain Tumor Institute, Robert H. Lurie Comprehensive Cancer Center, Northwestern University, United States.

It is a very reasonable consideration. A number of thought leaders in the field advocate for this, although there is no firm data to support superior outcomes. It can be a practical approach for patients with a good performance status who have a limited likelihood of deriving benefit from systemically administered targeted therapies.

Expert opinion 2: Koji Fukuda, PhD, Cancer Research Institute, Kanazawa University, Japan.

From a research perspective, early placement of an Ommaya reservoir at the time of leptomeningeal metastasis diagnosis is often considered appropriate in EGFR-wildtype NSCLC. Without actionable mutations, treatment options rely on chemotherapy (including intrathecal therapy) rather than targeted drugs. An Ommaya reservoir provides reliable intraventricular access for repeated intrathecal chemotherapy, showing modest efficacy in LM (response rates ∼50% and median survival of 4–6 months in treated patients). While no randomized trials exist, expert consensus suggests that prompt Ommaya placement can facilitate timely therapy and potentially improve symptom control in this setting.

Question 2

To prevent intracranial hypertension, we sometimes combine Ommaya reservoir implantation with ventriculoperitoneal shunt(VPS). What are your thoughts on this approach?

Expert opinion 1: Rimas V. Lukas, MD, Lou & Jean Malnati Brain Tumor Institute, Robert H. Lurie Comprehensive Cancer Center, Northwestern University, United States.

In our clinical practice we do this in some scenarios where there is symptomatic elevation of intracranial pressure. There is clinical data to support superior outcomes with the use of VPS in patients with LM and elevated intracranial pressure. Huntoon, et al. Neuro Oncol Pract. 2023 is one recent example.

Expert opinion 2: Koji Fukuda, PhD, Cancer Research Institute, Kanazawa University, Japan.

Leptomeningeal spread of cancer can impair CSF flow and lead to hydrocephalus with raised intracranial pressure. Combining an Ommaya reservoir with a VP shunt (often with an on-off valve) addresses both issues by enabling intrathecal chemotherapy delivery and providing CSF diversion to prevent intracranial hypertension. Literature reports (including an extensive retrospective study) indicate that patients who receive both CSF diversion and intrathecal chemotherapy survive longer than those managed with a shunt alone. In practice, the shunt valve can be temporarily closed during intrathecal drug administration and opened afterwards, allowing treatment of the LM while controlling pressure-related symptoms. This combined approach has been associated with significant symptom relief and improved quality of life in LM patients with hydrocephalus.

Question 3

Have you used intrathecal chemotherapy for leptomeningeal metastasis patients? If so, what drug and dosage do you commonly use?

Expert opinion 1: Rimas V. Lukas, MD, Lou & Jean Malnati Brain Tumor Institute, Robert H. Lurie Comprehensive Cancer Center, Northwestern University, United States.

Yes. In our clinical practice, for solid tumors we likely most often use topotecan (Groves, et al., 2008). For HER2+ disease we often consider trastuzumab (Wu et al., 2023; Kumthekar et al., 2023).

Expert opinion 2: Koji Fukuda, PhD, Cancer Research Institute, Kanazawa University, Japan.

Common intrathecal chemotherapy agents in solid tumor LM include methotrexate, cytarabine, and thiotepa. Standard intrathecal dosing reported in the literature is typically around 10–15 mg for methotrexate and about 20–50 mg for cytarabine (with ∼50 mg used for the liposomal slow-release form) per administration. Treatment is usually given in an induction phase (for example, intrathecal injections twice weekly for 2–4 weeks) followed by a maintenance phase with dosing every 2 weeks or monthly. Newer intrathecal agents (such as pemetrexed in NSCLC LM) have also been explored in clinical trials, but methotrexate-based regimens remain a frequently reported approach (References: Zhou et al., 2023; Huntoon et al., 2024; Scott et al., 2016).

To our knowledge, this case represents one of the longest reported survival times in an NSCLC patient with LM harboring the KRAS-G12V mutation (or other non-targetable driver mutations). However, several limitations should be noted.

First, while most literature describes IC administration via lumbar puncture or Ommaya reservoir alone, limited data exist on IC combined with a VPS. Moreover critically, we did not perform therapeutic drug monitoring of pemetrexed, thiotepa, or methotrexate in the CSF or plasma for this patient. The CSF diversion effect of VPS complicates drug concentration monitoring, as therapeutic agents may drain into the peritoneal cavity, potentially reducing the exposure of the leptomeningeal space to the chemotherapy. This pharmacokinetic challenge is supported by the literature. Studies have demonstrated that standard VPS can siphon intraventricularly administered drugs away from the CSF space, leading to underdosing (Huntoon et al., 2024). However, the combination of VPS with an Ommaya reservoir, particularly one equipped with an on-off valve as used in our case, is specifically designed to mitigate this issue by allowing temporary occlusion of the shunt during drug administration (Huntoon et al., 2024). Here, we determined dosing based on published efficacy and safety profiles from studies in the literature, but the precise pharmacokinetic impact of the VPS with on-off valve configuration on the IC agents used warrants further study.

Second, the patient received multiple simultaneous interventions (intrathecal agents, systemic immunotherapy, bevacizumab, WBRT, and VPS), making it impossible to attribute survival benefit to any single component. Our report is hypothesis-generating and underscores the need for prospective studies to delineate the individual contribution of each modality.

Third, the initial diagnostic work-up was incomplete: only PD-L1 immunohistochemistry was performed on the lung biopsy, whereas a full immunohistochemical panel and molecular profiling were omitted for cost-saving reasons. This deficiency may have limited the precision of subtype classification and subsequent therapy selection, highlighting the need for more comprehensive pathological evaluation in future cases.