A Systematic Intracranial Exposure‐Response Evaluation of Brain‐Penetrant Inhibitors: Informing Dose Selection for Brain Metastases Authors: Yuqian Zhao, Weize Huang Published: 2026 ABSTRACT ABSTRACTBrain metastases (BMs) represent a major clinical challenge due to their high incidence in cancers such as non‐small cell lung cancer (NSCLC) and breast cancer, and the limited efficacy of systemic therapies resulting from inadequate drug penetration across the blood–brain barrier. Although several brain‐penetrant tyrosine kinase inhibitors (TKIs) have been developed, the quantitative relationship between intracranial drug exposure and intracranial clinical response has not been well established, leaving an important gap in dose selection strategies. In this study, we systematically evaluated the intracranial exposure‐response relationship across brain‐penetrant small‐molecule TKIs. Fourteen TKIs were identified based on guideline‐recommended therapies for BMs. For each drug, plasma and cerebrospinal fluid (CSF) pharmacokinetic profiles were simulated using published population pharmacokinetic models, with CSF exposures extrapolated using known unbound plasma fractions and CSF‐to‐plasma unbound concentration ratios (K p,uu,CSF) in humans. Key intracranial exposure metrics, including the CSF steady‐state trough and average concentrations relative to IC50 (C min,ss/IC50 and C ave,ss/IC50) and the percentage of the steady‐state dosing interval during which CSF concentrations remained above IC50 (T% > IC50), were derived and correlated with intracranial outcomes from published clinical studies. Overall, significant positive correlations were observed between intracranial exposure metrics and both intracranial objective response rate (ORR) and complete response rate (CR). Greater exposure coverage above IC50 was associated with improved intracranial efficacy. This work demonstrates a positive intracranial exposure‐response relationship across brain‐penetrant TKIs and provides a quantitative benchmark to inform dose selection beyond traditional plasma exposure metrics. These findings highlight the importance of achieving sufficient CNS drug concentrations to optimize the therapeutic effect in patients with BMs.Keywords: brain penetration, intracranial response, pharmacokinetics Introduction Brain metastases are the most prevalent type of intracranial tumors in adults, significantly more common than primary brain tumors [1, 2]. An estimated 67%–80% of brain metastases originate from lung, breast, or melanoma primary malignancies [1]. Brain metastases are often found at the time of lung cancer diagnosis: among patients with metastatic disease, 15%–26% of those with non‐small cell lung cancer (NSCLC) and 24% of those with small‐cell lung cancer (SCLC) already have brain metastases [2]. Of these lung cancer brain metastases, about 2%–4% come from tumors with EGFR mutations and about 5% from tumors with ALK translocations [2]. In breast cancer, brain metastases are estimated to develop in 10%–16% of patients with Stage IV disease, while autopsy studies report rates as high as 30% [1]. About 8% of patients already have brain metastases when metastatic breast cancer is diagnosed, and the risk is highest in the HER2‐positive and triple‐negative subtypes [2]. Additionally, in women with localized or locally advanced HER2‐positive disease, the cumulative incidence of brain metastases can reach 11.8% by 10 years and 14.3% by 15 years [1]. Prognosis for metastatic brain disease remains poor, with median survival averaging 4 months in NSCLC and 10 months in breast cancer following the onset of brain metastases [3]. The pathophysiology of brain metastases is characterized by the interaction between metastatic cancer cells and the unique brain microenvironment [4]. This includes the formation of metastatic niches that support tumor colonization, interactions with astrocytes and microglia, and the establishment of an abnormal vasculature to sustain tumor growth [4]. These mechanisms underscore the aggressive behavior of brain metastases and contribute to their significant impact on patient survival [4]. Despite advancements in cancer therapies, effective treatments capable of penetrating the central nervous system to address brain metastases remain a critical unmet need [1, 2]. Therapeutic approaches for systemic (i.e., peripheral, non‐CNS) malignancies often fail to achieve sufficient drug concentrations in the brain, largely due to the restrictive nature of the blood–brain barrier (BBB) and the heterogeneous permeability of the blood‐tumor barrier (BTB) [5, 6]. Large molecules, such as monoclonal antibodies and antibody‐drug conjugates, face significant challenges, as the BBB effectively blocks nearly all of these therapies from entering the CNS under normal conditions [7]. Small molecules, while capable of crossing the BBB through passive diffusion, are often limited by active efflux mechanisms mediated by transporters like P‐glycoprotein (P‐gp) and breast cancer resistance protein (BCRP) [4, 8]. Additionally, the variability in BBB and BTB disruption across brain metastases further complicates drug delivery, resulting in inconsistent drug exposure and efficacy [5, 9]. These factors underscore the need for innovative therapeutic strategies that can reliably overcome these barriers and achieve effective drug concentrations in the CNS. Over the past decade, significant efforts have been made to develop brain‐penetrant tyrosine kinase inhibitors (TKIs) to overcome the challenges of treating brain metastases. The development of successive generations of TKIs targeting the same molecular pathways aimed at improving intracranial drug delivery and addressing the unmet needs of patients with brain metastases. First‐generation TKIs, such as crizotinib, exhibited limited CNS penetration and efficacy due to poor BBB permeability and susceptibility to efflux transporters like P‐gp [9]. In contrast, second‐generation TKIs, such as alectinib, were specifically optimized to enhance CNS penetration by evading efflux transporter recognition, thereby achieving greater therapeutic efficacy within the brain [9]. Indeed, the clinical efficacy of brain penetrants is highlighted by differences between first‐generation and second‐generation ALK inhibitors in the treatment of ALK‐positive NSCLC [10, 11]. Intracranial objective response rate is typically assessed using the Response Assessment in Neuro‐Oncology Brain Metastases (RANO‐BM) criteria, which incorporates MRI‐based measurements of target lesions, neurological function, and corticosteroid use to provide a more comprehensive evaluation of CNS tumor response [12]. Crizotinib, a first‐generation ALK inhibitor, shows poor CNS penetration, with an unbound CSF‐to‐plasma partition coefficient (K p,uu,CSF) of approximately 0.029 [13]. In patients with baseline CNS metastases, its intracranial objective response rate ranged from 23% to 50% [14]. In contrast, alectinib, a second‐generation ALK inhibitor specifically designed to improve CNS penetration, has a K p,uu,csf of approximately 0.943 [15] and a much higher intracranial objective response rate of 81% [10]. These data emphasize the critical role of CNS penetration in improving therapeutic outcomes for patients with brain metastases. Although the ALK+ TKI example shows that better brain penetration may lead to better CNS efficacy, the intracranial exposure‐response relationship for brain penetrants and the general dose selection strategy have not been consistently established. Achieving brain concentrations above the potency threshold is a critical consideration by the pharmacology principle in dose selection for brain penetrants. However, there remains a significant knowledge gap, as no existing data have definitively demonstrated how CNS exposure above potency thresholds translates into improved efficacy and therapeutic outcomes for patients with brain metastases. Typically, the dose levels of TKIs are selected to maintain unbound plasma drug concentrations above the IC50 for as much of the dosing interval as possible, with higher drug‐concentration/IC50 ratios expected to result in more sustained tyrosine kinase inhibition. For instance, in patients with R/R B‐cell malignancies, zanubrutinib maintains unbound C trough/IC50 ratios of 7 for 160 mg BID and 2 for 320 mg QD regimens, keeping free drug concentration above its IC50 throughout the dosing interval and sustaining complete, consistent BTK inactivation across target tissues under both regimens [16]. In our study, we identified 14 drugs to treat brain metastases and simulated their steady‐state exposure in plasma and CSF using published population pharmacokinetic models. Using simulated steady‐state PK profiles, intracranial drug exposure metrics were calculated. We aimed to investigate how intracranial drug exposure coverage above IC50 correlates with clinical efficacy and intracranial response, to explore the translatability of exposure metrics and to guide optimal dose selection. Methods Data Collection Based on the National Comprehensive Cancer Network (NCCN) Guidelines Version 1.2024 for Brain Metastases, twenty‐five small‐molecule brain penetrants were identified across various cancer types with brain metastases, including tumor‐agnostic cancers, breast cancer, melanoma, non‐small cell lung cancer (NSCLC), lymphoma, and renal cell carcinoma. For each drug identified, comprehensive literature searches were conducted using PubMed and regulatory New Drug Application (NDA) databases from the U.S. Food and Drug Administration (FDA) and the European Medicines Agency (EMA) to gather published information. Through this systematic approach, four crucial categories of data were gathered: (1) in vitro potency metrics (IC50 values); (2) brain penetration parameters including partition coefficients (K p,uu,CSF) representing the ratio of unbound drug concentration between cerebrospinal fluid (CSF) and plasma; (3) clinical central nervous system (CNS) or intracranial efficacy data [12], including intracranial objective response rate (iORR), intracranial complete response rate (iCRR), and median duration of response (mDoR); and (4) published population pharmacokinetic (PopPK) models for simulations of PK exposure metrics and subsequent exposure‐response analyses. Of the twenty‐five drugs initially identified, eleven were excluded due to missing data in one or more of these categories: seven lacked reported K p,uu,CSF, four lacked available intracranial clinical efficacy data. This resulted in a total of fourteen small‐molecule drugs (nine NSCLC drugs, three breast cancer drugs, and two tumor‐agnostic drugs). All detailed references for these data are summarized in the Tables S1–S4. Simulation of Plasma and CSF Pharmacokinetic Profiles and Derivation of Intracranial Exposure Metrics for Brain Penetrants For the fourteen selected brain penetrants identified from the NCCN guidelines, the PopPK model structures and parameters were extracted, and mrgsolve (R package version 1.5.2) was used for the plasma simulations of approved regimens until steady‐state using 500 virtual subjects. Then, the unbound CSF steady‐state concentration‐time profiles were generated by extrapolating from the simulated plasma PK profiles, utilizing the collected plasma unbound fraction (f u,p) and the ratio of unbound drug concentration between CSF and plasma (K p,uu,CSF) for each drug. Finally, the CSF exposure metrics relating to in vitro potency (IC50) were derived, including the ratios of CSF steady‐state minimum concentration (C min,ss)/IC50, CSF steady‐state average concentration (C ave,ss)/IC50, as well as the percentage of the dosing interval at steady‐state during which CSF concentrations exceeded the IC50 (T% > IC50). C min,ss was calculated as the geometric mean across 500 simulated virtual subjects. C ave,ss was calculated as AUC0‐tau,ss divided by tau, where AUC0‐tau,ss was the geometric mean across 500 simulated virtual subjects. T% > IC50 was calculated as the arithmetic mean of the percentage of the dosing interval at steady‐state during which CSF concentration > IC50. Intracranial Exposure‐Response Analysis for Brain Penetrants To investigate the relationship between intracranial exposure metrics and intracranial clinical response in selected brain penetrants, non‐parametric Spearman's ρ was used to assess the strength and direction of monotonic relationships between CNS exposure metrics (C min,ss/IC50, C ave,ss/IC50, T% > IC50) and intracranial clinical outcomes (iORR, iCRR, mDoR). Given that pooling different disease types (i.e., lung cancer, breast cancer) could confound the results and the limited sample size of breast cancer drugs (n = 3) and tumor‐agnostic drugs (n = 2), only NSCLC brain penetrants were included in the final statistical analysis (n = 9). Data from breast cancer and tumor‐agnostic drugs are only visualized in the Figures 3, 4, 5 but excluded from statistical testing for consistency and homogeneity. Exposure‐response analysis between the CNS exposure metrics C min,ss/IC50 and intracranial objective response rate (ORR), complete response (CR), and median duration of response (mDoR) for brain‐penetrants to treat lung cancer. Breast cancer and tumor‐agnostic brain penetrants are displayed in the figure but were not included in the statistical analysis. Exposure‐response analysis between the CNS exposure metrics C ave,ss/IC50 and intracranial objective response rate (ORR), complete response (CR), and median duration of response (mDoR) for brain penetrants to treat lung cancer. Breast cancer and tumor agnostic brain penetrants are displayed in the figure but were not included in the statistical analysis. Exposure‐response analysis between the CNS exposure metrics percentage of the dosing interval at steady‐state during which CSF concentrations exceeded the IC50 (T% > IC50) and intracranial objective response rate (ORR), complete response (CR), and median duration of response (mDoR) for brain penetrants to treat lung cancer. Breast cancer and tumor agnostic brain penetrants are displayed in the figure but were not included in the statistical analysis. Results Drug Properties The overall analysis workflow is illustrated in Figure 1. Data collection and extraction of published PopPK model parameters were performed, and these parameters were subsequently used to simulate plasma and CSF steady‐state concentration‐time profiles. A total of 14 brain penetrants were identified, and their IC50, K p,uu,CSF, and published PopPK model parameters are summarized in Table 1. A wide range of IC50 values across the brain penetrants was observed, from 0.39 ng/mL (afatinib) to 541.67 ng/mL (erlotinib), indicating significant differences in potency across molecules (Table 1). Similarly, the K p,uu,csf value, representing the unbound CSF‐to‐plasma concentration ratio, varied widely from 0.029 (crizotinib) to 0.943 (alectinib), reflecting a wide spectrum of brain penetration (Table 1). K p,uu,csf values were obtained directly from human data for nine drugs, rat data for three drugs (capmatinib, larotrectinib, and entrectinib), and PBPK model predictions for two drugs (lapatinib and neratinib); the corresponding references are provided in the Supporting Information. The approved dosing ranged from 40 mg (afatinib) to 1500 mg (lapatinib), either once or twice daily, highlighting variability in approved therapeutic dosing regimens (Table 1). High variability was also observed in the drug PK, with apparent clearance (CL/F) ranging from 3.09 L/h (erlotinib) to 195 L/h (neratinib), apparent volume of distribution in the central compartment (V c/F) from 13.4 L (erlotinib) to 4430 L (neratinib), and apparent volume of distribution in the peripheral compartment (V p/F) from 91.0 L (capmatinib) to 3990 L (larotrectinib) (Table 1). These findings highlighted significant variability in physicochemical properties, in vitro potency, brain penetration ability, and pharmacokinetics among brain‐penetrant agents. Overall workflow diagram including data collection, simulation of steady‐state plasma and CSF PK profiles, derivation of intracranial exposure metrics, and intracranial exposure‐response analysis for identified brain penetrants. TABLE 1Data summary for brain penetrants.Brain penetrantsIC50 a (ng/mL) K p,uu,csf b Dose (mg) regimenCL/F c (L/h) V c/F d (L) V p/F e (L)Tucatinib6.940.83300 BID131157573Lapatinib63.330.311500 QD3946.5338Neratinib3.120.80240 QD19544301440Alectinib0.920.943600 BID81.94016NA f Osimertinib3.100.31780 QD14.2986NA f Lorlatinib0.530.77100 QD14.5 g 120.5154.9Gefitinib3.580.377250 QD28.61540NA f Afatinib0.390.3440 QD42.3456NA f Adagrasib90.600.47600 BID35.9776208Erlotinib541.670.396150 QD3.0913.4456Crizotinib244.980.029250 BID67.488.4917Capmatinib1.010.34400 BID18.472.891.0Larotrectinib5.270.03100 BID52.1553990Entrectinib1.120.22600 QD20.5549NA f Note: The references of each parameter for each drug are included in Tables S1–S3. a Half‐maximal inhibitory concentration. b Partition coefficient that represents the ratio of unbound drug concentration between CSF and plasma. c Apparent clearance. d Apparent central volume of distribution. e Apparent peripheral volume of distribution. f Peripheral volume of distribution is not available since the PopPK model is one‐compartment. g Lorlatinib has initial clearance of 9.0 L/h and steady‐state clearance of 14.5 L/h due to time‐dependent induction of clearance. Data summary for brain penetrants. Note: The references of each parameter for each drug are included in Tables S1–S3. Half‐maximal inhibitory concentration. Partition coefficient that represents the ratio of unbound drug concentration between CSF and plasma. Apparent clearance. Apparent central volume of distribution. Apparent peripheral volume of distribution. Peripheral volume of distribution is not available since the PopPK model is one‐compartment. Lorlatinib has initial clearance of 9.0 L/h and steady‐state clearance of 14.5 L/h due to time‐dependent induction of clearance. Simulation of CSF Pharmacokinetic Profiles and Derivation of Intracranial Exposure Metrics for Brain Penetrants Using published PopPK model parameters, drug steady‐state plasma PK profiles were simulated at approved regimens (Figure S1). These profiles were then scaled to CSF PK profiles using the plasma unbound fraction and K p,uu,CSF, and were compared to the corresponding IC50 of each drug (Figure 2). The simulated CSF concentration‐time profiles demonstrated substantial variability in brain exposure coverage. For example, lorlatinib maintained steady‐state CSF concentrations consistently above IC50 for all 500 simulated subjects (T% > IC50 = 100%). In contrast, eight drugs, including tucatinib, alectinib, osimertinib, gefitinib, afatinib, capmatinib, larotrectinib, and entrectinib only achieved partial intracranial IC50 coverage, whereas five drugs (i.e., lapatinib, neratinib, adagrasib, erlotinib, and crizotinib) showed no intracranial IC50 coverage at all (T% > IC50 = 0%) (Figure 2 and Table 2). These simulated steady‐state CSF concentration‐time profiles indicated a wide spectrum of CSF exposures in relation to IC50 within the class of brain penetrants. Simulated steady‐state CSF PK profiles for brain penetrants. A total of 500 virtual subjects were used in the simulation. The solid dark blue line represents the median CSF concentration, the light blue shaded area indicates the 90% prediction interval. The purple horizontal dashed line represents the IC50 value. TABLE 2Summary of simulated intracranial exposure metrics and observed clinical responses.Brain penetrantsCSF C min,ss/IC50 a CSF C ave,ss/IC50 b T% > IC50 c (%)IC ORR d (%)IC CR e (%)IC mDoR f (months)Tucatinib0.831.5672.147.35.56.8Lapatinib0.060.11015.40NA g Neratinib0.050.09026.35.3NA g Alectinib1.651.8082.8813817.3Osimertinib0.981.2163.57078.9Lorlatinib52.3109.61008271NA g Gefitinib0.761.0657.56311.1NA g Afatinib1.151.5471.482.255.6NA g Adagrasib0.170.19042.115.812.7Erlotinib0.0590.090100NA g Crizotinib0.0020.0032033010.2Capmatinib5.7524.8597.64618NA g Larotrectinib0.0440.253.4630NA g Entrectinib1.562.498.379.212.512.9 Note: The references of clinical intracranial response parameters for each drug are included in Table S4. a Geometric mean of CSF steady‐state trough concentration in relation to in vitro potency IC50. b Geometric mean of CSF steady‐state average concentration in relation to in vitro potency IC50. c Average percentage of the dosing interval at steady state during which CSF concentrations exceeded the IC50. d Clinical intracranial objective response rate. e Clinical intracranial complete response rate. f Clinical intracranial median duration of response. g Data not available. Summary of simulated intracranial exposure metrics and observed clinical responses. Note: The references of clinical intracranial response parameters for each drug are included in Table S4. Geometric mean of CSF steady‐state trough concentration in relation to in vitro potency IC50. Geometric mean of CSF steady‐state average concentration in relation to in vitro potency IC50. Average percentage of the dosing interval at steady state during which CSF concentrations exceeded the IC50. Clinical intracranial objective response rate. Clinical intracranial complete response rate. Clinical intracranial median duration of response. Data not available. Based on the simulated unbound CSF concentration‐time profiles, CSF exposure metrics in relation to in vitro potency were derived, including CSF C min,ss/IC50, CSF C ave,ss/IC50, and T% > IC50 for each drug (Table 2). The CSF C min,ss/IC50 values ranged from 0.002 (crizotinib) to 52.3 (lorlatinib), reflecting substantial differences in steady‐state CSF trough concentrations relative to IC50 among these drugs (Table 2). Similarly, CSF C ave,ss/IC50 varied widely, from 0.0032 (crizotinib) to 109.6 (lorlatinib), indicating significant variability in average CSF exposure relative to IC50 (Table 2). T% > IC50, representing the average percentage of time above IC50 during a dosing interval at steady‐state, also varied considerably, with lorlatinib achieving 100% and five drugs showing 0% above IC50 (e.g., lapatinib, neratinib, adagrasib, erlotinib, and crizotinib; Table 2). Intracranial Exposure‐Response Analysis for Brain Penetrants Clinical intracranial (IC) response outcomes (IC ORR, CR, and mDoR) from previous clinical trials using these brain penetrants to treat brain metastases were collected and summarized (Table 2). These intracranial outcomes varied markedly: IC ORR ranged from 10% (erlotinib) to 82.2% (afatinib), IC CR from 0% (erlotinib) to 71% (lorlatinib), and IC mDoR from 6.8 months (tucatinib) to 17.3 months (alectinib) (Table 2). These findings highlighted the significant differences in intracranial response outcomes, suggesting that drug‐specific brain penetration abilities may have influenced therapeutic efficacy in treating brain metastases. In the intracranial exposure‐response analysis for brain penetrants in non‐small cell lung cancer (n = 9), significant positive correlations were observed between CSF C min,ss/IC50 and intracranial ORR (ρ = 0.75, p = 0.02; Figure 3a) as well as intracranial CR (ρ = 0.86, p = 0.0028; Figure 3b). Out of the nine drugs, only four had intracranial median duration of response (mDoR) data, which limited the analysis for mDoR. A positive correlation trend is visually noted between CSF C min,ss/IC50 and mDoR as shown in Figure 3c. Similar to the CSF C min,ss/IC50 ratio, the ratio of CSF C ave,ss/IC50 also showed a significant positive correlation with intracranial ORR (ρ = 0.75, p = 0.02; Figure 4a) and intracranial CR (ρ = 0.86, p = 0.0028; Figure 4b). A positive correlation trend was noted between CSF C ave,ss/IC50 and mDoR (Figure 4c). Finally, the percentage of the dosing interval at steady state during which CSF concentrations exceeded IC50 (T% > IC50) was significantly correlated with intracranial ORR (ρ = 0.75, p = 0.021; Figure 5a) and intracranial CR (ρ = 0.82, p = 0.0072; Figure 5b). A positive trend between T% > IC50 and mDoR was also observed (Figure 5c). Across all derived intracranial exposure metrics (CSF C min,ss/IC50, CSF C ave,ss/IC50, and T% > IC50), drugs with higher exposure metrics (e.g., lorlatinib) consistently showed higher intracranial ORR and CR values, suggesting a positive exposure‐response relationship. This trend was particularly strong for ORR and CR, suggesting that both minimum and average steady‐state CSF concentrations, as well as continuous exposure above IC50, correlated with improved efficacy. Unfortunately, mDoR data are limited. Discussion Brain metastases present a critical challenge in cancer treatment, particularly for patients with NSCLC and breast cancer, partially due to the limited ability of therapies to penetrate the BBB and achieve effective concentrations in the CNS. Although brain‐penetrant TKIs like alectinib have demonstrated improved CNS efficacy compared to first‐generation inhibitors such as crizotinib, a significant knowledge gap remains regarding the relationship between intracranial drug exposure and intracranial clinical outcomes. This study addresses this gap by evaluating how intracranial exposure coverage above IC50 correlates with intracranial efficacy, using validated PopPK model‐based simulated PK profiles and exposure metrics for fourteen brain penetrants. These findings provide insights into the intracranial exposure‐response relationship, which can further support optimal dose selection for brain penetrants. Significant progress has been made in developing inhibitor drugs for the treatment of metastatic cancers, with extensive research exploring exposure‐response (ER) relationships based on plasma PK and peripheral responses. In some cases, positive ER relationships have been reported, such as the correlation between steady‐state neratinib exposure and systemic objective response rate (ORR) in advanced/metastatic breast cancer patients [17]. In contrast, no correlation was observed between lorlatinib plasma exposure metrics and systemic ORR in patients with ALK‐positive metastatic NSCLC after receiving one or more ALK inhibitors [18]. Similarly, a flat ER relationship was reported between adagrasib plasma exposure and ORR in NSCLC patients with KRAS G12C mutation, also between capmatinib plasma exposure and ORR in NSCLC patients with MET mutations, potentially due to small sample sizes within the same regimen [19, 20]. Together, these contrasting patterns indicate that the systemic exposure‐efficacy relationship is not consistent across inhibitors. Beyond these systemic ER analyses, several studies have investigated whether higher drug exposures could improve overall clinical outcomes in patients with or without brain metastases. For instance, in NSCLC patients with baseline brain metastases, those started on afatinib 40 mg once daily achieved a longer median PFS than those started on 30 mg (13.3 vs. 5.3 months), which the authors attributed to the higher peak plasma concentration (C max) reached at the 40 mg dose [21]. Similarly, standard daily dosing of erlotinib (150 mg) yields CSF concentrations below IC50 for EGFR‐mutant cells (100 nM), whereas pulsatile high‐dose weekly erlotinib (1500 mg) reaches 130 nM and produces CNS responses in EGFR‐mutant NSCLC patients progressing on standard dosing [22]. In EGFR wild‐type NSCLC patients on the same 250 mg daily dose of gefitinib, those achieving higher steady‐state trough concentrations (C min,ss ≥ 200 ng/mL) had significantly longer median overall survival (16.8 vs. 4.1 months) than those with lower concentrations (C min,ss < 200 ng/mL) [23]. Additionally, a pharmacokinetic/pharmacodynamic meta‐analysis found higher steady‐state AUC of sunitinib (total drug including its active metabolite) significantly associated with prolonged overall survival and delayed tumor progression in gastrointestinal stromal tumor and metastatic renal cell carcinoma [24]. Collectively, these data suggest that higher exposure may improve clinical outcomes for some inhibitors. However, the supporting evidence is limited and heterogeneous, drawn from small or retrospective studies across different tumor types and outcome measures. These studies also assessed different exposure metrics, including steady‐state AUC, plasma peak and trough concentration, and CSF concentration relative to IC50, and the relative contribution of each to the exposure‐response relationship has not been systematically compared. Although most inhibitors were not initially indicated for the treatment of brain metastases, their clinical intracranial response has since been investigated in patients with brain metastases, underscoring the increasing need to understand the relationship between drug exposure and clinical intracranial outcomes. Despite these efforts to establish ER relationships for TKIs based on systemic exposure, no definitive ER relationships have been established between intracranial drug exposure and CNS response. In our study, we addressed this gap by simulating drug CSF exposure metrics in relation to IC50 and correlating them with intracranial clinical endpoints, such as IC‐ORR. By comprehensively evaluating a broad range of TKIs using a uniform approach, we systematically summarized their intracranial exposure and response analysis results. Additionally, our analyses highlight the importance of considering CSF‐to‐plasma concentration ratios as a surrogate for intracranial exposure, emphasizing their relevance in evaluating intracranial efficacy. Furthermore, by using exposure metrics normalized to IC50, our approach provides a benchmarked framework that is directly applicable for optimizing dosing strategies. Notably, drugs like lorlatinib demonstrated superior and continuous CSF coverage and consistently achieved concentrations above IC50, aligning with the observed relatively higher intracranial ORR and CR among all TKIs. In contrast, several TKIs, such as lapatinib and crizotinib, failed to sustain meaningful intracranial exposure and had relatively poor response. To assess whether this exposure gap could be addressed through increased dose level, we simulated the doses required for these drugs to achieve CSF C min,ss/IC50 = 1 and compared them with reported maximum tolerated doses (MTD) (Table S5). In all cases the required doses substantially exceeded the MTD or maximum administered dose (MAD), indicating that insufficient intracranial coverage due to limited brain penetration cannot be addressed by safely administering higher doses. Leveraging all TKIs, exposure‐response analysis revealed strong positive correlations between CSF exposure metrics (e.g., C min,ss/IC50, C ave,ss/IC50, T% > IC50) and intracranial ORR and CR, underscoring the importance of achieving sufficient intracranial drug concentrations to improve clinical outcomes. However, the relationship between exposure and mDoR showed higher variability, which may be due to limited data or additional factors beyond exposure that may influence long‐term response durability. This work addresses a critical knowledge gap by establishing the first comprehensive evaluation of brain exposure‐response relationships for TKIs, paving the way for improved dose selection and therapeutic strategies in patients with brain metastases. Several limitations merit consideration. First, intracranial response data were collected from multiple clinical studies that involved heterogeneous patient populations, varied lines of therapy, and different disease areas. This variability introduces potential confounding factors, making it challenging to draw definitive conclusions regarding the exposure‐response relationship across drugs. To try to alleviate this issue, in our analysis, we primarily focused on the nine drugs that were used as monotherapy only and the indications were all NSCLC (Figures 3, 4, 5); while data from other drugs in other disease areas are displayed in Figures but were not included in the ER analysis. Restricting the analysis in this way reduced heterogeneity but left a small dataset; therefore, leave‐one‐out sensitivity analyses were performed where each of the nine NSCLC drugs was excluded to assess whether the observed relationships were driven by any individual drug. This was applied to all three CSF exposure metrics (e.g., C min,ss/IC50, C ave,ss/IC50, T% > IC50) with intracranial ORR and CR, but not mDoR, which was reported for only four of the nine drugs (Table S6). The sensitivity analyses showed that the correlation between C min,ss/IC50 or C ave,ss/IC50 and intracranial CR remained statistically significant, while for the correlation between C min,ss/IC50 or C ave,ss/IC50 and intracranial ORR, removal of adagrasib, erlotinib, or crizotinib reduced the significance of correlations slightly. The small sample size limits the statistical power for the correlation analysis and should be interpreted with caution. Second, steady‐state CSF concentration‐time profiles for each brain penetrant were extrapolated directly from simulated steady‐state plasma PK profiles using parameters such as f u,p and K p,uu,CSF. This approach assumes a perfectly proportional relationship between plasma and CSF PK profiles, which may not fully reflect the brain tissue distribution, as the CSF PK profiles may exhibit less fluctuation than plasma PK profiles. For example, in a phenobarbital study in rats in which total plasma and brain concentrations were measured, the CSF concentration remained relatively stable despite plasma concentration changes [25]. CSF and plasma PK profiles may also be shifted in time relative to each other. In another example of rivastigmine, CSF T max was longer (1.4 to 3.8 h) compared to plasma T max (0.5 to 1.67 h), showing that peak CSF concentration can occur later than peak plasma concentration [26]. Damping and delay of the CSF PK profile would primarily affect exposure metrics defined by specific points on the concentration‐time profile, such as C min,ss. For this reason, we also evaluated C ave,ss, which reflects total exposure over the dosing interval and is therefore not sensitive to the shape or timing of the profile. Additionally, this study assumed that CSF exposure is the most representative indicator of brain exposure due to its feasibility for measurement and simulation. However, brain tissue or intracranial tumor drug concentrations would provide a more accurate representation for correlating with intracranial response. Using simulated CSF PK profiles as a surrogate may not precisely capture actual drug concentrations within brain tissue or tumors. Moreover, the limited availability of mDoR data for most NSCLC TKIs restricts the ability to fully evaluate the long‐term durability of intracranial responses. Another data limitation we encountered relates to the in vitro potency values used in our exposure metrics. IC50 values were compiled from published primary sources rather than generated under a single standardized protocol, and therefore differ across drugs in assay format (kinase versus cell‐based), cell line type, and readouts (Supporting Information Table S1). This heterogeneity introduces variability into the exposure‐to‐potency ratios that is independent of drug exposure itself. In addition, some selected drugs lacked published IC90 values, necessitating our reliance solely on IC50 values in our exposure metrics. Finally, our simulations were based on published PopPK models that primarily describe the parent drug, without accounting for active metabolites. Among the nine NSCLC TKIs evaluated, alectinib, osimertinib, and crizotinib have reported active metabolites. Because our simulations only captured parent drug exposure, total intracranial exposure of active drug‐related material is likely underestimated for these drugs. Nonetheless, metabolite exposure is generally in high correlation with parent exposure, and metabolites generally have higher polarity and poorer brain penetration than the parent drugs. Future research is warranted to address these limitations to improve the understanding of intracranial exposure‐response relationships for brain penetrants. For PK extrapolation from plasma to CSF, mechanistic models that incorporate physiological processes governing drug distribution across the blood–cerebrospinal fluid barrier should be considered. Models such as physiologically based pharmacokinetic (PBPK) models tailored for brain penetrants could better capture the dynamics of CSF and brain tissue drug distribution [15, 27]. To address the limited availability of mDoR data, larger sample sizes and more comprehensive clinical datasets are necessary. Finally, incorporating both parent drug and metabolite PK data into future simulations will be critical for accurately assessing total intracranial exposure of drug‐related material. Recent advancements in molecular imaging techniques, particularly positron emission tomography (PET), have demonstrated significant potential in predicting whole‐body distribution and tumor drug uptake of TKIs. Radiolabeled tracers including 11C‐erlotinib, 18F‐afatinib, and 11C‐osimertinib have been investigated as predictive biomarkers for response to EGFR‐TKI treatment [28]. In these studies, tracer uptake was measured as the tumor‐to‐blood ratio (TBR), which reflects drug distribution and target binding [28]. For instance, 11C‐osimertinib has been shown to effectively penetrate an intact blood–brain barrier and distributed rapidly to the brain in healthy volunteers, reaching a whole‐brain total distribution volume of 14 mL/cm3 [29]. These findings highlight the utility of PET imaging for assessing the CNS distribution of brain penetrants. PBPK models were later developed to predict tissue TBR in the brain, lung, tumor, spleen, kidney, and bone for the same three tracers in advanced‐stage NSCLC patients [28]. The predicted tumor‐to‐lung contrast fell within threefold of the observed PET ratios for all three tracers, although tissue‐level TBR accuracy varied by drug [28]. Brain was among the less accurately predicted tissues, estimated within threefold for osimertinib but overpredicted for afatinib by about 190%, which the authors attributed to efflux transporters at the blood–brain barrier that the model did not capture [28]. Building on these models, future work may extend PBPK frameworks to incorporate CNS physiology and brain tumor compartments, together with the active influx and efflux transporters that impact drug distribution across the blood–brain barrier [28]. Extending such models across a broader range of brain penetrants, and from microdose tracer studies toward therapeutic dosing, could enable more accurate prediction of intracranial drug exposure. Linking these predictions to CNS clinical endpoints would help bridge the current gap between intracranial exposure and CNS response and ultimately improve dose selection and treatment strategies for patients with brain metastases. Author Contributions Y.Z. and W.H. wrote the manuscript. W.H. designed the research. Y.Z. performed the research. Y.Z. and W.H. analyzed the data. Funding The authors have nothing to report. Conflicts of Interest The authors declare no conflicts of interest. Supporting information Figure S1: Simulated steady‐state plasma PK profiles for brain penetrants. A total of 500 virtual subjects were used in the simulation. The solid dark blue line represents the median plasma concentration, the light blue zone indicates the 90% prediction interval. Table S1: References, assay format, and free‐drug conditions for the IC50 value used for each drug. Table S2: References and properties for the K p,uu,CSF value used for each drug. Table S3: References for the population pharmacokinetic (PopPK) model used for each drug. Table S4: References for clinical intracranial efficacy outcomes used for each drug, including intracranial objective response rate (ORR), complete response (CR), and median duration of response (mDoR). Table S5: Simulated doses required to achieve CSF C min,ss/IC50 = 1 compared with maximum tolerated doses for brain penetrants without CSF coverage above IC50. Table S6: Leave‐one‐out sensitivity analysis of Spearman rank correlations between CSF exposure metrics and intracranial efficacy endpoints in NSCLC brain penetrants (n = 9). Each drug was sequentially excluded and the correlation analysis was re‐performed. Intracranial median duration of response (mDOR) was not evaluated, as it was reported for only four of the nine drugs. References Associated Data Supplementary Materials Figure S1: Simulated steady‐state plasma PK profiles for brain penetrants. A total of 500 virtual subjects were used in the simulation. The solid dark blue line represents the median plasma concentration, the light blue zone indicates the 90% prediction interval. Table S1: References, assay format, and free‐drug conditions for the IC50 value used for each drug. Table S2: References and properties for the K p,uu,CSF value used for each drug. Table S3: References for the population pharmacokinetic (PopPK) model used for each drug. Table S4: References for clinical intracranial efficacy outcomes used for each drug, including intracranial objective response rate (ORR), complete response (CR), and median duration of response (mDoR). Table S5: Simulated doses required to achieve CSF C min,ss/IC50 = 1 compared with maximum tolerated doses for brain penetrants without CSF coverage above IC50. Table S6: Leave‐one‐out sensitivity analysis of Spearman rank correlations between CSF exposure metrics and intracranial efficacy endpoints in NSCLC brain penetrants (n = 9). Each drug was sequentially excluded and the correlation analysis was re‐performed. Intracranial median duration of response (mDOR) was not evaluated, as it was reported for only four of the nine drugs.