The justification of a new prognostic system for BM patients with low KPS scores
SRS is an established treatment for BM and is preferred over open surgery and WBRT21. With the recent introduction of the fractionated SRS system, SRS application has been extended to large tumors that previously required open surgery22,23. Additionally, SRS allows concurrent TT/IT, unlike open surgery, facilitating the treatment of systemic disease while minimizing the interruption of chemotherapy; this may have contributed to improvement in the survival of BM patients along with the advancement of chemotherapy24,25. Moreover, it has been demonstrated that even multiple BMs over 10 can be effectively treated with SRS, which is advantageous for conserving cognitive function compared to WBRT 4,21. Active treatment for BM used to be recommended in patients with KPS ≥ 704; however, advances in chemotherapy and the expansion of SRS application have increased the treatment options for BM patients with KPS < 70, which may further intensify the dependence on SRS for BM treatment26,27. Therefore, it is essential to identify the appropriate candidates for SRS among BM patients with KPS < 70. This consideration is vital to ensure the adequate allocation of limited medical resources to those with a higher likelihood of prolonged survival, while also enabling patients in a dire state to prepare for their end-of-life rather than enduring ineffective treatments.
A novel 3-month survival probability model in patients with a KPS score of ≤ 70
In this study, we defined patients with a life expectancy of over 3 months as needing active treatment. We investigated the factors associated with 3-month survival in patients with a KPS of 60–70, generally representing poor and borderline prognosis by dichotomous classification, who underwent GKS for BMs from NSCLC. As a result, controlled ECD, the presence of FND, and small ∑ TV were significantly associated with a survival time of over 3 months after GKS for BM. These results are consistent with previously published literature. ECD is a well-known prognostic factor of BM patients, and uncontrolled ECD is the leading cause of death in BM, as shown in this study28,29. KPS generally reflects the ECD status, and those with a KPS that worsens with ECD progression do not seem to recover well. This may be why the KPS has been a robust prognostic factor in cancer patients in many studies to date. However, patients presenting FND due to BM have a poor KPS scale but can recover after SRS for BMs. Notably, this study was confined to patients with a KPS of 60–70, and the presence of FND was given a higher weight than ECD for 3-month survival as 5 versus 4, which had the most substantial impact on survival among the three demonstrated factors until the ∑ TV was < 15.3 cm3. Similar to our study, Chernov and colleagues studied the prognostic factors of patients with KPS ≤ 50 who underwent GKS for BM and demonstrated that low KPS due to BM-derived FND was associated with favorable survival, while low KPS resulting from ECD was related to a poor prognosis26. Several studies have demonstrated ∑ TV as another important prognostic factor; the larger ∑ TV is, the more negative the influence on survival2,30,31,32. This study showed that the effect of ∑ TV was the greatest among the three prognostic factors when ∑ TV was ≥ 15.3 cm3. At first, these results seem to conflict with the finding that most BM patients die from uncontrolled ECD, not from BM itself4,33,34. Based on this, we speculate that ∑ TV reflects the severity and progression rate of the disease independently of KPS and is, therefore, closely related to OS. We categorized the patients into four groups based on their ECD and FND statuses and calculated the cut-off ∑ TV corresponding to the 3-month survival probability ranging from 10 to 90% based on the statistical model (Table 3). According to the model, while a large ∑ TV is acceptable for SRS in category 1, a smaller ∑ TV is treatable in categories 2 and 3, and in category 4, a poor prognosis is predicted in most cases. Therefore, we suggest active SRS for patients in category 1 unless their ∑ TV is enormous, whereas supportive care for patients in category 4. Although the 3-month survival probability model was developed based on dichotomized survival data, the result value is continuous data calculated in the context of the weight of each variable. Also, a notable correlation was demonstrated that a higher 3-month survival probability was linked to a more prolonged time to death (Pearson correlation r = 0.54, p < .0001) (Fig. 4). Hence, the suggested 3-month survival probability model effectively captures the actual life expectancy, even considering the limited sample size, and we believe this model can assist clinicians in determining whether the patient is amenable to GKS.
Factors associated the survival in NSCLC patients with BM in the literature
It has been reported that EGFR mutations and ALK rearrangements play a significant role in the invasion of circulating tumor cells through the blood–brain barrier (BBB) and tumor angiogenesis, resulting in BM in NSCLC patients35,36. Therefore, targeting these genetic abnormalities with drugs capable of penetrating the BBB has shown efficacy in treating BM of NSCLC, and some studies have reported a potential reduction in the occurrence of BM37. Recent studies have reported that EGFR tyrosine kinase inhibitors improved the median OS to > 16 months and ALK inhibitors significantly improved the median OS to > 40 months compared to 7.2–7.8 months of the conventional chemotherapy era38,39,40,41,42. This study demonstrated that the concurrent administration of TT/IT after GKS was significantly associated with OS in the Kaplan–Meier analysis (Supplement 2). On univariate logistic regression analysis with the endpoint of the 3-month survival, the concurrent use of TT/IT following GKS still had a significant association, while it was not determined as an independent factor on multivariate analysis. We conducted an additional analysis to ascertain the genuine impact of TT/IT by excluding three patients who continued the same treatment they had been receiving prior to the development of BM, considering the possible drug resistance. On multivariable Cox proportional regression analysis, concurrent TT/IT following GKS demonstrated significant improvement in OS (HR 2.030, 95% CI 1.087–3.789, p = .026), in line with previous studies. However, in logistic regression analysis with the endpoint of 3-month survival, concurrent TT/IT did not show a correlation on multivariable analysis, while it was significantly associated with 3-month survival on univariable analysis (HR 5.202, 95% CI 1.521–17.796, p = .009). The limited statistical power of TT/IT in the multivariate logistic analysis could be attributed to the small sample size. Otherwise, the impact of poor disease status, characterized by uncontrolled ECD and substantial ∑ TV, on premature mortality within 3 months might outweigh the potential advantages of TT/IT. Meanwhile, the use of TT/IT before GKS did not show any association with OS or 3-month survival. Among these patients, most patients (n = 20, 77%) were considered to develop drug resistance. There was no significant difference in 3-month survival between patients who demonstrated resistance to targeted TT/IT and those who did not. Given the limited sample size and the complexities of the individual treatment course, compounded by the various systemic chemotherapy regimens, this study could not conclude the significance of disease refractoriness to TT/IT on early mortality after GKS. In addition, age and number of BMs, which were included as prognostic factors in other predictive systems, such as RPA, GPA, and the score index for radiosurgery, were not associated with 3-month survival in this study12,43,44. We consider that the inconclusive results on other strong prognostic factors in this study might have been attributed to the small sample size and the narrow KPS range of the patients. Nevertheless, this study, which focused on patients with KPS ≤ 70, known as borderline-to-poor prognosis, provides insight into screening patients who are suitable to undergo active SRS despite the low KPS score by suggesting a statistical prediction model for 3-month survival.
This research was based on a retrospective single-center study with a relatively small sample size. Our hospital is among the leading tertiary referral hospital in South Korea, where a substantial number of cancer patients receive treatment, and the demand for GKS for BM at our hospital is exceptionally high, with approximately 300 patients undergoing GKS for BM per year. However, this study specifically focused on patients with NSCLC and poor KPS of ≤ 70. In addition, to mitigate any potential confounding influences stemming from improved survival with advances in chemotherapy over time, we recruited patients who were treated contemporaneously, excluding those who had undergone GKS more than 5 years prior to the study design. As a result, only 67 patients were finally included despite a high volume of BM patients receiving GKS. In addition, the proposed prediction model for 3-month survival has a limitation in terms of generalizability, as it has not undergone external validation. However, we believe this study can serve as a stepping stone to future prospective large-scale multicenter studies to develop a more robust survival prediction model for patients with BM.