608 lung cancer patients who received at least one cycle of ICIs therapy were collected between January 2018 and May 2024 at the First Hospital of China Medical University. Data was extracted from 12 months after the use of ICIs. All patients underwent cranial MRI imaging examination before receiving ICIs treatment, and we subsequently excluded cases meeting the following criteria: (a) age under 18 years, (b) a history of hematologic, (c) a history of primary brain cancer and cerebrovascular disease, (d) absence of brain MRI data (At least two cranial MRI images for analysis: pre-ICI and within 12 months post-ICI). 419 patients were included with cranial MR imaging before and after ICIs treatment; then, 19 patients were excluded for lack of response assessment. Ultimately, 400 patients were analyzed in this study. For further analysis, all patients were divided into two groups: those with CSVD and those without CSVD, based on cranial MR imaging. The flow chat of our study design is shown in Figure 1 . Demographic, clinical, and survival data were retrieved from electronic medical records. All procedures performed in this study were in accordance with the Declaration of Helsinki (as revised in 2013). This study was approved by the Ethics Committee of the First Hospital of China Medical University (Project number: 2023-544-2).

In this study, CSVD was diagnosed using cranial MR imaging (3.0T) to obtain axial T1, T2-weighted, fluid-attenuated inversion recovery (FLAIR), and T2-weighted gradient echo (GRE) images. All MRI images were independently reviewed by two vascular neurologists; a third vascular neurologist adjudicated discordant findings. The imaging features of CSVD are white matter hyperintensity (WMH), lacunar infarction (LI), enlarged perivascular space (EPVS), cerebral atrophy (CA) and cerebral microbleed (CMB) (18). By comparing with pre-immunotherapy cranial MRI images, patients who had no baseline CSVD but developed new CSVD after treatment, as well as patients with baseline CSVD who exhibited new CSVD lesions or significant worsening of existing lesions post-treatment, were included in the CSVD group, while the remaining patients were assigned to the non-CSVD group.

The demographic and clinical characteristics of lung cancer patients were collected from patient’s electronic medical records, including age (at initiation of ICIs), gender, body mass index (BMI), tumor histology type, initial cancer stage, immunotherapy regimens, line of ICIs therapy (e.g., first line, second line), cranial MR imaging history, and vascular risk factors (smoking, drinking, hypertension, diabetes, coronary heart disease and hyperlipidemia). Peripheral blood parameters included white blood cell count (WBC), neutrophilic count (NE), lymphocyte count (LY), hemoglobin count (HGB), platelet count (PLT), systemic immune-inflammation index ratio (SII: neutrophil count × platelet count)/lymphocyte count absolute monocyte count), neutrophil to lymphocyte ratio (NLR), platelet to lymphocyte Ratio (PLR), creatinine (Cr), cystatin C (Cys-c), estimated glomerular filtration rate (eGFR), uric acid (UA), C-reactive protein (CRP), lymphocyte subsets (CD4/CD8), free triiodothyronine (FT₃), free thyroxine (FT₄), thyroid stimulating hormone (TSH), cortisol (Cor) and adrenocorticotropic hormone (ACTH), fibrinogen (Fg) and D-dimer (D-D).

Due to subsequent requirements, the timeline of hypothyroidism in all patients and the progression of CSVD in those with hypothyroidism were also documented. Among patients with CSVD, we collected peripheral blood parameters at two time points: baseline (before ICI treatment) and at the time of CSVD diagnosis. In the non-CSVD group, these parameters were recorded at two time points: baseline data prior to the initiation of ICI therapy, and the final data within 12 months after ICIs treatment. The progression free survival (PFS) was calculated from the date of first administration of the ICIs until the progression of disease. The overall survival (OS) was calculated from the date of first administration of the ICIs until death or the last follow-up date (31 May 2025).

All analyses were performed using SPSS 26.0 (IBM, Armonk, NY, USA) and GraphPad Prism 9.0 (GraphPad Software, La Jolla, CA, USA). Two-side P values <0.05 were considered statistically significant. To describe general baseline characteristics, continuous variables data were expressed as mean ± standard deviation, and categorical variable data were summarized by frequency (%). The T-test, nonparametric test, or chi-square test were used to compare the baseline characteristics between groups, as appropriate. Logistic regression was performed to analyze the risk factors of CSVD. Selection of covariates in the multivariable models was based on univariate associations and biological relevance. An odds ratio (OR) with 95% confidence interval (CI) was reported for each covariate of interest. To address time-to-event outcomes, Cox proportional hazards regression was employed to identify CSVD risk factors. Univariable analyses of age, cancer stage, vascular risk factors and hypothyroidism yielded hazard ratio (HR) with 95% CI; significant predictors (P < 0.05) were retained in the final multivariable model with covariate adjustment. Proportional hazards assumptions were validated via Schoenfeld residuals (global P > 0.05), and multicollinearity was excluded (VIF < 2.0). The receiver operating characteristic (ROC) curve was performed to evaluate the diagnostic efficacy of data related to the occurrence of CSVD. The survival rates between the different groups were compared using the Kaplan-Meier method.

The representative MR imaging examples were demonstrated in Figure 2A . Figure 2B showed the percentages of each lesion when one lesion was present: WMH (67/104, 64.42%), LI (18/104, 17.31%), EPVS (12/104, 11.54%), CA (6/104, 5.77%), and CMB (1/104, 0.96%). Figure 2C showed the percentages of the combinations when two lesions were present. We could see the most common combinations was WMH and LI (51.43%). Secondly, WMH and CA (17.14%), LI and EPVS (17.14%) with same proportion. Then, EPVS and CA (5.71%), WML and CMB (5.71%) also with same proportion. A total of 11 cases were diagnosed with CSVD within 0–3 months after ICIs treatment, peaking at 6–9 months with 38 cases. Figure 2D clearly demonstrates the temporal distribution of CSVD following immunotherapy.

400 patients were included in this study and the incidence of CSVD in lung cancer patients treated with ICIs was 26% (104/400 patients). Slightly more than half were aged≥65 years, and 81.7% were male ( Table 1 ). In China, the incidence of lung cancer caused by smoking is significantly higher in male patients than in females (19). The demographic and clinical characteristics of the enrolled patients were shown in Table 1 . CSVD and Non-CSVD patients consisted of 26% (n = 104) and 74% (n = 296) of the entire lung cancer patients. 53.25% were aged≥65 years and 78.75% were male. There was no significant difference between groups with/without CSVD for age, gender, BMI, tumor histology, tumor stage, treatment data, vascular risk factors, baseline blood cell count, baseline eGFR and baseline TSH/FT₄.

In the univariate logistic regression analysis, the results indicated that age (OR, 1.97; 95% CI: 1.16-2.79, P=0.025), stage IV (OR, 1.69; 95% CI: 1.24-2.01, P=0.040), and hyperlipidemia (OR, 2.37; 95% CI: 1.37-3.21, P=0.008) were associated with an increased risk of CSVD. Variables with a P-value ≤ 0.05 from the univariate logistic regression analysis were included in the multivariate logistic regression analysis. The results indicated that age (OR, 1.86; 95% CI: 1.22-2.81, P=0.036), cancer stage IV (OR, 1.81; 95% CI: 1.33-2.38, P=0.047), and hyperlipidemia (OR, 2.12; 95% CI: 1.34-3.16, P=0.013) were significantly and independently associated with the risk of CSVD ( Table 2 ). Numerous studies have shown that age is a recognized independent risk factor for CSVD. Moreover, the severity and progression of CSVD increase with age (20, 21). Hyperlipidemia is a fatal risk factor for the development of stroke. Studies have shown that several common lipid abnormalities have a causal relationship with CSVD subtype infarction (22, 23). In this study, the results suggest that hyperlipidemia is an independent risk factor for CSVD in lung cancer patients undergoing immunotherapy.

To clarify the specific changes in laboratory indicators during the occurrence of CSVD, we extracted laboratory data at the time of CSVD diagnosis and conducted a comparative analysis with baseline data. The results indicated no significant alterations in WBC, NE, LY, HGB, PLT, NLR, Cr levels, eGFR, CRP, CD4/CD8 ratio, FT₃, Cor, ACTH, Fg, or D-D. However, there were notable changes in PLR, SII, Cys-c, UA, FT4, and TSH levels from baseline to the onset of CSVD, as shown in Table 3 . A comparison of these six biomarkers was made between the CSVD group and the non-CSVD group. All six indicators showed no statistically significant differences when compared to baseline and final medication data (within 12 months after ICIs treatment) in the non-CSVD group ( Figure 3 ). Apart from UA, the remaining five data points in the non-CSVD group did not show any statistically significant differences when comparing the baseline data with the final data (12 months after initiating immune checkpoint inhibitor therapy) ( Figure 3 ).

Although no statistically significant differences were observed in inflammatory cell counts and non-specific biomarkers of inflammation (CRP and D-D), the composite biomarkers of systemic inflammation indices PLR and SII both exhibited an upward trend. Specifically, PLR increased from 164.93 ± 27.86 to 171.27 ± 32.29, with a P-value of 0.035; while SII rose from 774.28 ± 53.57 to 790.65 ± 68.34, with a P-value of 0.009. The final results indicated that FT₄ levels decreased significantly from the baseline to when CSVD was diagnosed, from 13.21 ± 4.56 pmol/L to 11.01 ± 2.11 pmol/L, with a P-value of 0.012. A corresponding tendency in TSH levels was observed in the CSVD group, increasing from 4.12 ± 0.46 pmol/L to 4.78 ± 1.13 pmol/L, with a P-value of 0.043. Cys-c levels gradually increased from 1.01 ± 0.98 mg/L to 1.29 ± 0.86 mg/L (P = 0.049) in the CSVD group.

To clarify the correlation between thyroid function and CSVD, hypothyroidism was further incorporated as a time-dependent covariate for Cox regression analyses. The results demonstrated that age ≥65 years (HR=1.31; 95% confidence interval CI: 1.05-1.81, P=0.043), stage IV cancer (HR=1.82; 95% CI: 1.42-2.96, P=0.015), hyperlipidemia (HR=1.65; 95% CI: 1.20-2.25, P=0.012), and hypothyroidism (HR, 2.38; 95% CI: 1.89-5.04, P=0.005) all exhibited significant independent correlations with CSVD ( Table 4 ).

Observations suggest that thyroid dysfunction may contribute to the occurrence of CSVD. Consequently, we compiled the timeline of hypothyroidism in all patients and analyzed the ROC curves to evaluate the predictive performance of FT₄ and TSH levels as individual indicators. The optimal cutoff value for FT₄ to distinguish the occurrence of CSVD was determined to be 11.84 pmol/L [AUC= 0.775 (95% CI 0.749-0.891), sensitivity = 70.4%, specificity = 80.5%, P = 0.028, Figure 4A ]. The optimal cutoff value for TSH was determined to be 4.23 pmol/L [AUC = 0.548 (95% CI 0.672-0.805), sensitivity = 45.2%, specificity = 60.8%, P = 0.247] ( Supplementary Figure S1 ).

During the period of ICIs treatment, we found 78 patients (19.5%) with confirmed hypothyroidism, and among these, 34 patients (43.6%) were confirmed to have CSVD. We observed that the incidence of CSVD was higher in the hypothyroidism group compared with the non-hypothyroidism group (43.6% versus 21.7%, P<0.05) ( Figure 4B ). The onset of hypothyroidism in all patients occurred from 3 to 24 weeks, with a peak at 12 weeks ( Figure 4C ). The incidence of CSVD in patients with hypothyroidism ranged from 0 to 12 months, peaking at 3 to 6 months ( Figure 4D ).

Among the 400 lung cancer patients, we compared the PFS and OS between the CSVD group and non-CSVD group. The Kaplan-Meier curve analysis revealed no significant difference in median PFS between the CSVD group and the non-CSVD group (14.52 months vs. 13.12 months, P = 0.882, Figure 5A ), and similarly, no significant difference in median OS was observed (27.74 months vs. 23.69 months, P = 0.068, Figure 5B ).