This single-center retrospective cohort examined temporal and clinical changes in lung cancer leptomeningeal metastasis (LM) across a 14-year period. The investigation compared patients diagnosed with LM in two eras (2012–2016 vs. 2017–2026) to assess whether the timing of LM occurrence and post‑LM outcomes shifted in the contemporary targeted‑therapy era.
The study retrospectively analyzed 201 patients with cytologically confirmed LM from lung cancer diagnosed between November 2012 and May 2026. Patients were classified by the date of LM diagnosis into an earlier era (2012–2016) and a later era (2017–2026). Molecular paired profiling of primary tumor tissue and cerebrospinal fluid (CSF) was available for a subset of 80 patients. Survival analyses used Kaplan‑Meier methods and prespecified multivariable Cox proportional hazards models.
Time from initial lung cancer diagnosis to LM and time from Stage IV diagnosis to LM were both compared between eras using the Mann‑Whitney U test. The later era showed significantly longer intervals: the median interval from lung cancer diagnosis to LM was 23.7 months in the later cohort versus 13.8 months in the earlier cohort (p < 0.001). Similarly, the interval from Stage IV diagnosis to LM was longer in the later era (17.4 vs. 10.4 months; p = 0.002). These findings indicate a delay in LM occurrence in more recent years.
Adenocarcinoma accounted for the vast majority of cases (93.0%). Paired molecular profiling of primary tumor tissue and CSF was conducted in 80 patients. Tissue–CSF driver discordance was identified in 7 of those 80 patients (8.75%), demonstrating that CSF genotyping can reveal driver alterations not detected in primary tumor samples.
Unadjusted overall survival (OS) measured from the time of LM diagnosis did not differ significantly between the two eras: median OS was 9.8 months in one era versus 11.3 months in the other (p = 0.864). Despite the delay in LM onset observed in the later era, this did not translate into a statistically significant unadjusted improvement in post‑LM survival in this cohort.
Prespecified multivariable Cox regression identified several independent associations with post‑LM mortality. Later diagnostic era was associated with higher adjusted mortality (adjusted hazard ratio [HR] 1.826; p = 0.014). Smoking history was associated with higher mortality (adjusted HR 1.847; p = 0.014). An ECOG performance status of ≥2 was also linked to increased mortality risk (adjusted HR 1.674; p = 0.007). In contrast, exposure to third‑generation EGFR‑TKI after LM diagnosis was associated with lower mortality (adjusted HR 0.385; p < 0.001).
The observed tissue–CSF discordance in a subset of patients supports the clinical utility of CSF molecular profiling in LM. CSF genotyping may identify actionable driver alterations not captured by primary tumor testing and therefore can inform targeted therapy selection. The association between post‑LM third‑generation EGFR‑TKI exposure and lower adjusted mortality in this analysis suggests clinical benefit in appropriately selected patients, though causality cannot be established from this observational data alone.
The authors emphasize that adjusted associations reported in the multivariable analyses should be interpreted with caution because of potential residual confounding and post‑baseline treatment bias inherent to retrospective observational designs. Details on specific treatment regimens, duration of therapies, and other potential confounders were not provided in the abstract and therefore are not reported here. The single‑center nature of the cohort may limit generalizability.
In this 14‑year single‑center cohort of 201 patients with cytologically confirmed lung cancer LM, the later diagnostic era was associated with a longer interval from lung cancer and Stage IV diagnosis to LM, but no statistically significant improvement in unadjusted post‑LM overall survival. CSF profiling identified driver discordance in a minority of paired cases and may offer clinically relevant information beyond primary tumor genotyping. Smoking, poorer performance status, and later diagnostic era were associated with higher adjusted mortality, while post‑LM third‑generation EGFR‑TKI exposure was associated with lower adjusted mortality; these adjusted findings require cautious interpretation due to potential biases.