Introduction

Malignant melanoma (MM) originates from melanocytes or melanin precursor cells and is commonly found in the skin and mucosal tissues. Notably, up to 60% of individuals with stage IV melanoma develop brain metastases (BMs). The incidence of BMs has increased in recent years due to improved imaging technologies and advancements in systemic therapies that typically do not have brain efficacy1,2. Nevertheless, the outlook for individuals with malignant melanoma brain metastases (MBMs) is grim, as their overall survival (OS) typically falls within the range of 4–9 months post-diagnosis3. Consequently, a thorough analysis of the clinical features of patients with MBMs enables healthcare professionals to assess the prognosis more comprehensively and objectively, thereby offering valuable scientific support for clinical interventions and treatment strategies.

Farris et al. (2017) established the notion of brain metastasis velocity (BMV) as a predictive biomarker specifically applicable to the salvage therapy of newly emerging BM after stereotactic radiosurgery (SRS) alone4,5. However, the application of BMV is restricted to individuals who have experienced their first intracranial failure. Because patients with MBMs have a poor OS, a part of them die as a result of lesion progression before intracranial failure. This group of patients failed to determine their BMV. In response to this limitation, Soike et al. (2018) introduced a novel metric, initial brain metastases velocity (iBMV), derived from the ratio of BMs at initial SRS to the duration since the initial MM diagnosis6. The study revealed that iBMV is correlated with OS, and shows promise for guiding treatment decisions. However, the index validation is limited to specific cancer types, such as non-small cell lung cancer, breast cancer, and renal cell carcinoma7,8,9,10, with a lack of validation in patients with MBMs. In addition, while iBMV has been validated in patients undergoing SRS for BMs, its generalizability to all patients with BMs remains uncertain.

Consequently, this study aimed to assess iBMV utility in individuals with MBMs and construct a prognostic model for patients with MBMs by incorporating this novel metric to guide clinical decision-making.

Materials and methods

Patients

This study was conducted in a cohort of 1072 patients diagnosed with MM at Fujian Cancer Hospital between December 2010 and February 2023. Patients included in the study were those who developed BMs after their MM diagnosis, including those who developed BMs following adjuvant or first-line treatment. Patients without BMs or with BMs at the time of initial cancer diagnosis were excluded. The Ethics Committee of the Fujian Cancer Hospital, Fuzhou, China (No. K2023-030-01) approved this study.

Variables

The data collection procedure consisted of two steps. Basic patient information included age, sex, Karnofsky Performance Status (KPS), time to diagnosis of melanoma, and BMs development. Tumor information included the primary tumor site, number and maximum diameter of BMs, extracranial metastases, BM with midline shift, meningeal metastasis, cerebral hemorrhage, clinical symptoms, serum lactate dehydrogenase (LDH) level, serum albumin level, lymphocyte-to-monocyte ratio (LMR), neutrophil-to-lymphocyte ratio (NLR), platelet-to-lymphocyte ratio (PLR), prognostic nutritional index (PNI) and systemic immune-inflammation index (SII).

Initial brain metastasis velocity

iBMV was computed by dividing the number of BMs by the time interval (measured in years) between the original MM diagnosis and the dates when magnetic resonance images were collected to identify BM6,7. For instance, if a patient develops four brain metastases at the time of the initial brain event, which is 6 months (0.5 years) post-diagnosis of cancer, the iBMV is 8 BM/year.

Statistical analysis

The study’s primary endpoint was the OS, calculated from the time of the initial BM to the endpoint. Figure 1 shows the study’s flow diagram. Kaplan–Meier (KM) estimators with log-rank tests were employed to assess the differences in OS. Cox regression analysis was used to identify potential OS predictors. Second, based on the multivariate analysis results, a nomogram was constructed using the 1-, 2-, and 3-year OS rates. The C-index, area under the curve (AUC) values, and the calibration curve were calculated to validate and evaluate the model regarding discrimination and consistency.

Fig. 1
figure 1

Flowchart depicting the overall design of the study.

Full size image

Based on the results of the multivariate analysis, specific values for the different prognostic factors were obtained by multiplying the β coefficients by the same multiplier. The study’s score was the β coefficients × 311. A novel prognostic scoring system combined with Graded Prognostic Assessment (GPA) was developed. Finally, by selecting optimal cutoff values, all MBM patients were classified into low-risk and high-risk groups. The model’s discriminatory accuracy was assessed by calculating the time-dependent receiver operating characteristics (ROC) curve.

All statistical analyses were performed with the statistical software package R (v4.2.0; http://www.R-project.org). All cut-off points were generated using X-tile software (version 3.6.1 (https://x-tile.software.informer.com/3.6/). Statistical significance was considered at P < 0.05.

Results

Baseline characteristics

The baseline patient characteristics are presented in Table 1. A total of 111 eligible patients (55 men and 56 women) were screened. The mean and median survival times were 18.54 months and 11.80 months. MM occurred more frequently in the limbs (48.6%, 54/111). The median iBMV score for the entire study cohort was 2.27. KM survival analysis showed that patients with an iBMV above the median had a significantly shorter OS than patients with an iBMV below the median (p < 0.001, log-rank test) (Fig. 2a). The optimal cutoff values for LMR, NLR, PLR, PNI, and SII were obtained, and the KM curves are plotted in Fig. 2b,c,d,e,f.

Table 1 Characteristics of patients with malignant melanoma brain metastases.
Full size table
Fig. 2
figure 2

Clinical outcomes according to different risk factors. Kaplan–Meier curve and log-rank test for OS. (a) OS stratified by patients with an iBMV above and below the median. (b) OS stratification of patients according to the optimal cut-off value for LMR. (c) OS stratification of patients according to the optimal cut-off value for NLR. (d) OS stratification of patients according to the optimal cut-off value for PLR. (e) OS stratification of patients according to the optimal cut-off value for PNI. (f) OS stratification of patients according to the optimal cut-off value for SII. OS, overall survival; iBMV, initial brain metastasis velocity; LMR, lymphocyte-to-monocyte ratio; NLR, neutrophil-to-lymphocyte ratio; PLR, platelet-to-lymphocyte ratio; PNI, prognostic nutritional index; SII, systemic immune-inflammation index; HR, hazard ratio; CI, confidence interval.

Full size image

Nomogram construction and validation

Univariate and multivariate Cox regression analyses were conducted to examine factors influencing OS that were identifiable at initial BMs diagnosis. The univariate analysis revealed that the time to melanoma diagnosis and BMs development (p < 0.001), number of BMs (p = 0.022), iBMV (p < 0.001), KPS (p < 0.001), LDH level (p = 0.017), serum albumin level (P = 0.002), LMR (P = 0.041), NLR (P = 0.010), PLR (p < 0.001), PNI (P = 0.014), and SII (P = 0.001) were significantly associated with OS (Table 2). Subsequently, the significant factors identified in the univariate analysis were further examined using multivariate analysis. iBMV (Hazard Ratio [HR] 1.65, 95% Confidence interval [CI] 1.04–2.62, P = 0.035), KPS (HR 5.68, 95% CI 3.01–10.73, P < 0.001), LDH level (HR 1.61, 95% CI 1.05–2.46, P = 0.028), serum albumin level (HR 0.56, 95% CI 0.32–0.98, P = 0.041), and SII (HR 1.72, 95% CI 1.02–2.92, P = 0.042) were independent prognostic factors, as shown in Table 2.

Table 2 Univariable and multivariate Cox regression analysis of characteristics associated with OS.
Full size table

Based on the multivariate models, we integrated the risk factors into a nomogram to predict patient survival at 1, 2, and 3 years (Fig. 3a). The C-index, a measure of the predictive accuracy of the nomogram, was 0.723. The calibration (Fig. 3b–d) and ROC (Fig. 3e) curves suggested that the nomogram was calibrated and discriminated appropriately. In Fig. 3e, the areas under the ROC curves at 1, 2, and 3 years were 0.802 (95%CI 0.720–0.885), 0.839 (95% CI 0.757–0.921), and 0.812 (95% CI 0.703–0.920), respectively. The precision-recall ROC curves based on 1, 2, and 3 years were shown in the Supplementary Fig. 1.

Fig. 3
figure 3

A nomogram and its validations for MBMs. (a) Nomogram predicting the OS rates at 1, 2 and 3 years for patients with MBMs. The nomogram summed the points identified on the scale for each variable. The total points projected on the bottom scales indicate the probabilities of OS rates at 1, 2 and 3 years. (b) Calibration of the nomogram for predicting 1-year OS rate. (c) Calibration of the nomogram for predicting 2-year OS rate. (d) Calibration of the nomogram for predicting 3-year OS rate. (e) The nomogram’s ROC curves and AUC values for predicting 1-, 2-, and 3-year OS rates. MBMs, malignant melanoma brain metastases; OS, overall survival; iBMV, initial brain metastasis velocity; KPS Karnofsky performance status, Serum LDH, serum lactate dehydrogenase; Serum Alb, serum albumin; SII, systemic immune-inflammation index; ROC, Receiver operating characteristic; AUC, Area under the curve; CI, confidence interval.

Full size image

A novel melanoma-GPA based on initial brain metastasis velocity

Based on the multivariate regression analysis results and the GPA score, we established a novel prognostic scoring model, the Melanoma-GPA, which predicted the outcome better than the GPA score. The risk factors and distribution of points were as follows: serum albumin within the normal range: 3 points, LDH within the normal range: 1.5 points; iBMV ≤ 2.27, SII ≤ 610, age < 50, KPS between 90 and 100, number of BMs:1 and no extracranial metastases: 1 point each, age between 50 and 59, KPS between 70 and 80, number of BM 2–3:0.5 points each. The total Score ranges from 0 to 10, with a lower score indicating worse prognosis. A cut-off value of five was determined to be optimal, resulting in the categorization of patients into low-risk (56, 50.45%) and high-risk (55, 49.55%) groups. The median OS of the low- and high-risk groups were 17.7 and 7.9 months. The AUC was calculated at each time point and aggregated to generate a Time-dependent ROC plot (Fig. 4). As illustrated in Fig. 4, compared to GPA, Melanoma-GPA was significantly superior for each time point, including the AUC of 1-year (Melanoma-GPA, 0.779 (0.694–0.863) vs. GPA, 0.709 (0.617–0.802)), the AUC of 2-year (0.794 (0.690–0.899) vs. 0.664 (0.548–0.779)), and the AUC of 3-year (0.729 (0.585–0.874) vs. 0.591 (0.436–0.746)).

Fig. 4
figure 4

Comparison of AUC values from 1–3 years for two prognostic scoring models. AUC, Area under the curve; GPA, Graded Prognostic Assessment.

Full size image

Discussion

As the incidence of MM continues to rise, diverse combination therapies, such as targeted therapy, immunotherapy, and SRS, are being actively explored and have resulted in improved OS for many patients12,13,14. Assessing a patient’s status at the initial presentation of BMs can facilitate the development of more holistic treatment decisions for MBMs, ultimately leading to further survival benefits.

Since iBMV was first proposed in 2018, several correlative studies have been conducted. Xu et al. retrospectively analyzed 132 patients with BM from lung adenocarcinoma and found a strong association between iBMV and OS. This suggests that iBMV may be a valuable prognostic indicator in patients with lung adenocarcinoma and BMs who underwent SRS9. Yamamoto et al. enrolled patients with BM originating from various primary tumors, including breast, lung, and kidney tumors. Their findings aligned with previous research, indicating that iBMV is a significant independent prognostic factor for patients following SRS10. A recent study examined the prognostic implications of three different types of BMV: “classic” BMV (cBMV), iBMV, and volume-based BMV (vBMV). These findings indicate that cBMV and iBMV are prognostic indicators of OS in patients with BMs treated with Stereotactic Radiosurgery, whereas vBMC has not been validated. Similarly, the study’s Cox regression analysis revealed that iBMV was associated with OS in patients with MBMs (HR 1.65, P = 0.035)15. Furthermore, we found that patients with higher iBMV exhibited a poorer prognosis than those with lower iBMV, regardless of whether they underwent SRS.

Systemic inflammatory markers, which indicate the body’s inflammatory and immune responses, are associated with increased cancer incidence and mortality16,17. In previous studies, the SII, derived from lymphocyte, neutrophil, and platelet counts, has emerged as a powerful prognostic indicator of unfavorable outcomes in patients with hepatocellular carcinoma, offering a potential tool for guiding treatment decisions18,19. Moreover, a meta-analysis of 8083 patients demonstrated a correlation between high SII and low OS in all ethnic groups, tumor types, and critical masses and increased SII was associated with lower progression-free survival (HR = 1.80, 95% CI 1.27–2.56, P = 0.001)20. Consistent with these findings, our study revealed that SII is an independent prognostic indicator of OS in patients with MBMs (HR 1.72, P = 0.042). Hence, using SII as a prognostic indicator can enhance the precision of evaluating disease progression and survival prognosis in patients, thereby facilitating the formulation of treatment strategies.

Numerous studies have been conducted on the prognostic factors of BMs, but limited research focuses specifically on MBMs due to their aggressive nature21,22,23. The GPA score, a commonly utilized clinical prognostic tool, incorporates age, KPS, the presence of extracranial metastases, and the number of BMs24. However, our findings indicate that LDH, serum albumin, iBMV, and SII significantly influence OS. Thus, a novel GPA scoring system was developed, and patients were categorized into low- and high-risk groups. The median OS of the high-risk group was notably inferior to that of the low-risk group (7.9 vs. 17.7 months). Furthermore, the predictive efficacy of the two scoring methods was evaluated using time-dependent ROC, which revealed that the AUC value of melanoma-GPA surpassed that of GPA across all time points within 1–3 years. Thus, melanoma-GPA has a better predictive value and can more comprehensively and objectively assess the prognosis of patients with MBMs. In clinical practice, physicians can utilize this scoring system at the time of a patient’s initial brain metastasis diagnosis. Without requiring complex calculations, the system provides an objective assessment of prognosis, enabling rapid risk stratification and providing valuable support for developing personalized treatment plans tailored to the patient’s condition.

This study has some limitations. First, it is a retrospective study conducted at a single center, which complicates the ability to control for potential confounding variables. Second, due to the low MM prevalence, the sample size of this study was limited. Third, future research should validate our findings in prospective studies to ensure model generalizability. And treatment-related factors should be incorporated to enhance the assessment of the effectiveness of various treatment approaches for MBMs. This will assist clinicians in selecting the optimal treatment modalities based on individual patient characteristics, ultimately leading to improved clinical outcomes in patients with MBMs.

Conclusion

In conclusion, iBMV is a prognostic prediction tool for patients with MBMs who develop such metastases after their initial cancer diagnosis. Consequently, we established Melanoma-GPA based on iBMV, which can more accurately assess patients’ prognosis compared to the GPA score. This new melanoma-GPA is expected to be a simple, convenient, and objective tool for predicting the survival of patients with MBMs in future clinical applications.