Therapeutic Monitoring of Orally Administered, Small-Molecule Anticancer Medications with Tumor-Specific Cellular Protein Targets in Peripheral Fluid Spaces—A Review

Orally administered, small-molecule anticancer drugs with tumor-specific cellular protein targets (OACD) have revolutionized oncological pharmacotherapy. Nevertheless, the differences in exposure to these drugs in the systemic circulation and extravascular fluid compartments have led to several cases of therapeutic failure, in addition to posing unknown risks of toxicity. The therapeutic drug monitoring (TDM) of OACDs in therapeutically relevant peripheral fluid compartments is therefore essential. In this work, the available knowledge regarding exposure to OACD concentrations in these fluid spaces is summarized. A review of the literature was conducted by searching Embase, PubMed, and Web of Science for clinical research articles and case reports published between 10 May 2001 and 31 August 2022. Results show that, to date, penetration into cerebrospinal fluid has been studied especially intensively, in addition to breast milk, leukocytes, peripheral blood mononuclear cells, peritoneal fluid, pleural fluid, saliva and semen. The typical clinical indications of peripheral fluid TDM of OACDs were (1) primary malignancy, (2) secondary malignancy, (3) mental disorder, and (4) the assessment of toxicity. Liquid chromatography–tandem mass spectrometry was most commonly applied for analysis. The TDM of OACDs in therapeutically relevant peripheral fluid spaces is often indispensable for efficient and safe treatments.


Introduction
The past two decades have seen the rise of a new era of targeted oncological pharmacotherapy. The novel treatment options have led to a tremendous increase in success rates since the first market approval of the now generic imatinib (Gleevec ® , 2001), an inhibitor of the BCR-ABL oncogenic tyrosine kinase protein, and the first representative of orally administered, small-molecule anticancer drugs with specific tumor-associated cellular protein targets (OACDs). These synthetic molecules bind to proteins that are expressed excessively or even exclusively in cancer cells, resulting in the inhibition of the functions of cancer cells with a limited impact on non-malignant cells. Most OACDs are found in the subgroup L01E of the Anatomical Therapeutic Chemical (ATC) classification system (level 1: "Antineoplastic and immunomodulating agents", level 2: "Antineoplastic agents", level 3: "Protein kinase inhibitors"), and are further classified at level 4 as BCR-ABL tyrosine kinase inhibitors, epidermal growth factor receptor (EGFR) tyrosine kinase inhibitors, B-raf serine-threonine kinase (BRAF) inhibitors, anaplastic lymphoma kinase (ALK) inhibitors, mitogen-activated protein kinase (MEK) inhibitors, cyclin-dependent kinase (CDK) inhibitors, mammalian target of rapamycin (mTOR) kinase inhibitors, human epidermal growth factor receptor 2 (HER2) tyrosine kinase inhibitors, Janus-associated kinase (JAK) inhibitors, vascular endothelial growth factor receptor (VEGFR) tyrosine kinase inhibitors, Bruton's tyrosine kinase (BTK) inhibitors, phosphatidylinositol-3-kinase (Pi3K) inhibitors, Scientific works published between 10 May 2001 and 31 August 2022 were evaluated. Since, to the best of the authors' knowledge, no reviews have been previously written in the same topic, the searched time range was selected to cover the entire period OACDs have been available on the market. No filtering or limiting settings were applied. In the Embase and Web of Science databases, the search was conducted in the titles and in the abstracts ("Title or Abstract").
Each database record was evaluated by two reviewers (Z.K. and G.B.K.) who also conducted the manual research. Duplicate publications were removed by Z.K. before screening. No automation tools were employed for evaluating the eligibility of the records.

Screening Eligible Database Records
The workflow of retrieving research articles for full evaluation is shown in Figure 1. The evaluation of the records was performed by Z.K. and G.B.K.  First, duplicates of the PubMed records were removed from the results of the Embase and Web of Science database search. The remaining records were subsequently assessed individually for meeting basic requirements. Only peer-reviewed full manuscripts written in English, assigned an individual digital object identifier, and made available online by the publisher within the searched period were considered for further screening. Level 2 screening was based on the contents of the title and the abstract. Only records with an explicit evidence of ineligibility were removed at this level. The type of the article was the first object of assessment. Articles presenting randomized and nonrandomized registered clinical studies, non-registered, researcher-initiated clinical studies, retrospective observational studies, case series (describing 2 cases or more with the individual assessment of subjects), and individual case reports were included for further evaluation. Book chapters, comment articles, editorials, meta-analyses, practical guidelines, research protocols, scoping reviews, and systematic reviews were not considered. Second, articles describing experiments in which the subjects were not humans, i.e., in vitro experiments or in vivo animal studies, were removed. Subsequently, studies performed with the participation of human subjects, but without the aim to evaluate or to support decisions related to their medical treatment, i.e., without direct therapeutic relevance (e.g., with the inclusion of healthy volunteers, or conducted with the only aim to deliver pharmacokinetic data), or including medical intervention which, by current understanding, would not be part of the clinical practice (e.g., monitoring drug levels in cord blood or in amniotic fluid to evaluate the exposure of the fetus) were eliminated. Finally, studies explicitly performed without the monitoring of any of the drugs listed in Table 1 in a peripheral fluid space were also excluded.
In the phase of full manuscript screening, the first object of assessment was the ethical review board approval. Case reports and case series were exempt from this requirement. Articles continued to be retained if explicit evidence was found in the main text confirming that the study had been performed with direct therapeutic relevance, as described for level 2 screening. The presentation of the results of monitoring at least one drug displayed in Table 1 in a peripheral fluid space was a further requirement for inclusion. Articles not excluded in this phase were subject to full evaluation. The full manuscripts retrieved by manual search were screened in an identical manner before inclusion.

Data Evaluation and Visualization
All descriptive information on the database records and the contents of the manuscripts found were stored and processed using Microsoft Excel. The year-normalized number of publications on each drug was calculated as n publ /n year , where n publ is the number of included publications on the drug, and n year is the number of years the drug had been available on the market. The latter was defined as the period starting with the day of the first approval by the American Food and Drug Administration, and ending on 31 August 2022. Visualization was carried out using Microsoft Office applications.

Summary of the Findings of the Literature Review
The database search yielded a set of 1503 potentially relevant articles (732, 305, and 466 hits in PubMed, Web of Science, and Embase, respectively). Four-hundred and seventyfour duplicates were removed. Of the remaining 1029 papers, 258 were presentation abstracts, and 31 were not written in English. In a single case, a record was listed with false authors. These records were also excluded. The manual search yielded five additional hits which were subsequently found in the PubMed database, but had not been listed by the automatic search. The assessment of the remaining 739 articles based on title and abstract resulted in the exclusion of further 546 publications. Forty-three articles were excluded based on their type. Three hundred and ninety works described in vitro experiments or in vivo animal studies, and 35 were conducted in humans, but without a direct therapeutic goal. Seventy-eight studies were excluded based on evidence retrieved from the title and/or the abstract that drug concentrations were not monitored in any peripheral fluid space.
All of the 193 publications retained for full evaluation could be retrieved from the websites of the publishers. An in-depth study of these manuscripts resulted in the elimination of 113 publications. The authors of two papers failed to present evidence of the approval of an ethical review board for conducting research on humans. Thirty-five studies were not performed in humans, and five were conducted without direct therapeutic relevance. Seventy-one works were excluded because drug concentrations were not monitored in any peripheral fluid space. The remaining 80 publications were selected for the detailed review ( Figure 1, Table 2). Overall, 34% of the included publications were individual case reports, 31% were registered clinical studies, 19% were case series, 13% were non-registered, researcher-initiated studies, and 3% were retrospective observational studies.     August 2020 to December 2020 Zanubrutinib in CSF and in plasma [87] Thirty-two small-molecule, orally taken anticancer medications with specific cellular protein targets were monitored with a clinical indication in at least one peripheral liquid space on at least one occasion in the investigated period. This comprises 38.6% of the OACDs which had an ATC code on 31 August 2022. CSF was the most frequently monitored peripheral space (82% of all publications). The share of manuscripts on all other peripheral spaces (breast milk-8%, pleural effusate-4%, ascitic or peritoneal dialysis fluid-2%, intracellular fluid-2%, other (saliva and semen, 1 record for each)-2%) was low. In 61% of the cases, the indication for monitoring an OACD in a peripheral fluid compartment was to control a secondary malignancy. Other indications were the treatment of a primary malignancy (19%), controlling toxicity (14%), and treatment of a mental disorder (6%). In a single manuscript, an additional indication was the prevention of graft rejection [81]. Registered clinical studies, non-registered, researcher-initiated studies, case series, and case reports comprised 29.4%, 14.1%, 22.4%, and 34.1% of the included publications, respectively ( Figure 2).   Based on the number of relevant publications available on a specific OACD, normalized to the number of years it has been marketed, tepotinib (Tepmetko ® ) triggered the largest interest, followed by erlotinib (Tarceva ® ), ribociclib (Kisqali ® ), imatinib, and osimertinib (Tagrisso ® ), while the OACDs receiving the least attention were lapatinib (Tyverb ® ), vorinostat (Zolinza ® ) and sunitinib (Sutent ® , Figure 2, Table 3). Table 3. Clinical background of the monitoring of OACDs in peripheral fluid spaces. ALL, acute lymphoid leukemia; AML, acute myeloid leukemia; CNS, central nervous system; CSF, cerebrospinal fluid; CLL, chronic lymphoid leukemia; CML, chronic myeloid leukemia; NSCLC, non-small cell lung cancer; PBMC, peripheral blood mononuclear cells; PCNSL, primary central nervous system lymphoma. CSF Secondary malignancy AML with extramedullary and meningeal relapse [27] CSF Secondary malignancy ALL and CNS leukemia prophylaxis [28] [20] Breast milk Risk of toxicity CML during pregnancy [21] Breast milk Risk of toxicity CML during pregnancy and breastfeeding [22] Breast milk Risk of toxicitiy CML during pregnancy and breastfeeding [23] Breast milk Risk of toxicity CML in early pregnancy and breastfeedng [24] Breast  Currently, the most common indication of monitoring OACDs in a peripheral fluid space is to improve the treatment of primary and secondary CNS tumors in adults and in pediatric patients by performing measurements in the CSF. Early examples for such efforts included the assessment of erlotinib in pediatric glioblastoma and in CNS hemangioblastoma with von Hippel-Lindau disease, and of vismodegib (Erivedge ® ) in pediatric recurrent or refractory medulloblastoma [47,48,94]. Broniscer et al. investigated the pharmacokinetics of erlotinib in a pediatric patient by measuring the concentrations of erlotinib along with its O-demethylated, pharmacokinetically active metabolite OSI-420 in plasma and in CSF. Six time-matched pairs of specimens were collected. The CSF/total plasma concentration ratio (CSF-TPR) of erlotinib was 7.0%, while the ratio of drug exposure was 6.9% based on 24-h areas under the concentration-time curves. This evaluation was based on total plasma levels. Since the fraction of erlotinib bound to plasma proteins is approximately 93%, it is reasonable to assume that the unbound fraction equilibrated between plasma and CSF at a 1:1 ratio [47,97]. In a single paired measurement performed in an adult patient, a median erlotinib CSF level corresponding to 21.6% of median total plasma concentrations was found, which would be equivalent to 309% of the unbound plasma fraction [48]. In a phase 1 study conducted with pediatric patients, a total of nine paired CSF and plasma samples were collected from three subjects to evaluate vismodegib concentrations. The CSF/unbound plasma concentration ratios (CSF-UPR) attained a median of 53% (26-78%) [94].
The monitoring of OACDs in this context has gained more attention only very recently. The concentrations of regorafenib (Stivarga ® ) as well as its active N-oxide and demethylated N-oxide products were assessed in recurrent malignant glioma. All three substances attained detectable levels in CSF. While the concentration values were not explicitly provided by the authors, visual plots showed that the CSF-TPR's were 0.01 or higher. Approximately 99.5% of circulating regorafenib is bound to proteins, indicating that the CSF levels exceeded unbound plasma concentrations [89].
The monitoring of ceritinib (Zykadia ® ) and ribociclib in patients diagnosed with recurrent glioblastoma was performed [74,78,79]. The unbound fraction of ceritinib, determined using equilibrium dialysis with a 5 kDa regenerated cellulose membrane, corresponded to 1.4% (0.6-2.6%) of total levels. The unbound CSF concentrations were comparable to concentrations measured in nonenhancing tumor regions, and were tenfold higher than unbound plasma levels [74].
The ratio of ribociclib CSF/unbound plasma concentrations was 1.29 in one study and 0.6-4.4 in another. Equilibrium dialysis was employed in both works to determine the unbound fractions directly. The ratios increased over time [78,79]. Ribociclib CSF concentrations were evaluated in recurrent or refractory malignant pediatric brain tumor. The CSF-TPRs were 0.0-42.9% [80].
Dasatinib (Sprycel ® ), imatinib, nintedanib (Ofev ® ), panobinostat (Farydak ® ), regorafenib, ribociclib, and vorinostat were assayed in 42 CSF samples obtained from nine pediatric brain tumor patients. Nintedanib and panobinostat were undetectable in the samples. There was a correlation between blood protein levels and imatinib concentrations. In addition, imatinib and regorafenib proved to bind to CSF proteins as well, resulting in unbound fractions of 88% and 65%, respectively. These data indicate that both plasma and CSF protein concentrations may have an impact on detectable drug levels, and that the elevation of drug availability can be expected in CSF when the blood-brain barrier is not intact and CSF protein levels increase [12].
Ibrutinib (Imbruvica ® ) was measured in CSF in primary CNS lymphomas [85,86]. In one study, hemodialysis was conducted every other day. Six-hour post-dose CSF ibrutinib levels were about tenfold higher on hemodialysis-free days than those observed on hemodialysis days. In addition, the CSF-UPR's (with an assumed protein-bound fraction comprising 97.3% of circulating ibrutinib) were 78% and 8%, respectively [85].
Zanubrutinib (Brukinsa ® ) concentrations were assayed in 23 time-matched plasma and CSF samples of 13 patients, 8 of whom were diagnosed with primary CNS lymphoma, and 5 with diffuse large B-cell lymphoma. The CSF-TPR was 2.39±1.71%. With an assumed 94% protein binding rate, the authors calculated CSF-UPR's of 42.7±27.7%, and concluded that zanubrutinib was successfully transported through the blood-brain barrier [87].

Other Primary Malignancies
Other types of tumors in which OACD concentrations have been evaluated in peripheral fluid spaces include Philadelphia chromosome-positive (Ph + ) chronic myeloid leukemia (CML), non-small cell lung cancer, and gastrointestinal stromal tumors.
Imatinib concentrations were monitored in patients diagnosed with Ph + CML. In a follow-up study conducted with 15 adult patients, Nambu et al. found a weak correlation between imatinib levels determined in leukocytes (buffy coat cells) and in plasma (r = 0.281). While the intracellular concentrations of the drug were not associated with the cytogenic response, there was a significant difference between groups of patients with different genotypes (SLCO1B3 334TT and 334 TG/GG) [13]. In another study conducted with adult Ph + CML subjects, peripheral blood mononuclear cells (PBMC) were isolated from anticoagulated whole blood. Again, a weak yet statistically significant positive correlation was found between imatinib concentrations observed in plasma and in PBMC (r = 0.203) [14].
In both works, intracellular imatinib concentrations were about a magnitude higher than those found in plasma.
Malignant pleural effusion is a severe condition developing as a complication of lung or breast cancer in women [98]. Masago et al. investigated erlotinib and OSI-420 concentrations in the plasma and in the pleural effusate samples of nine adult patients diagnosed with advanced NSCLC. On days 1 and 8 of the treatment, 2-h post-dose (day 1) and trough pleural effusate levels (day 8) were compared to trough plasma concentrations. They found that erlotinib and OSI-420 pleural effusate concentrations had increased considerably, with larger than 100% pleural effusate/total plasma concentration ratios obtained by day 8 [49]. In an NSCLC patient, gefitinib concentrations in pleural effusates attained approximately 30% of those observed in plasma. The penetration of gefitinib into the peritoneal third-space fluid was, on the other hand, negligible [38].

Monitoring the Treatment of Malignant Tumor Metastases Central Nervous System Metastases of Myeloproliferative Malignancies
The involvement of the CNS presents a major challenge in the therapy of leukemias. Adult patients present with CNS leukemia in approximately 5% of acute leukemia cases, while CNS involvement occurs in about every third pediatric patient presenting with a relapse [99]. The risk of malignant cell penetration through the blood-brain barrier is especially high in Ph + B-cell precursor acute lymphoid leukemias (ALL) [100]. The prevention of CNS involvement in acute leukemias and the efficient treatment of established CNS leukemias are, therefore, of considerable importance and have an impact on the overall survival.
The poor penetration of the blood-brain barrier by imatinib, the first marketed tyrosine kinase inhibitor drug, was first mentioned in 2002 [15,16]. The total imatinib concentrations were 1.57 µg/mL and 0.017 µg/mL in the plasma and CSF samples of a young female adult diagnosed with Ph + ALL [16]. The size of the unbound fraction of imatinib was later established to be around 5% (4.3-6.5%) in healthy humans and in acute myeloid leukemia patients. By applying this percentage, the CSF-UPR of imatinib in this patient can be estimated as 21.7%. While the authors concluded that the distribution of imatinib into CSF was extremely poor, the consideration of the unbound fraction as the basis of the evaluation of blood-brain-barrier penetration delivers a more appreciable penetration rate [101].
On five separate days in an 11-day period, measurable imatinib concentrations were found in the CSF and plasma samples of a male Ph + CML patient who was in a lymphoid blast crisis after achieving complete cytogenic remission in the bone marrow following more than eight months of imatinib therapy, but had developed an isolated neoplastic meningitis. The authors concluded that imatinib CSF concentrations were not sufficient to inhibit 50% of BCR/ABL tyrosine kinase, and assume that the reason underlying the poor penetration of imatinib is its affinity to p-glycoprotein, a protein responsible for multi-drug resistance. Nevertheless, total imatinib concentrations were evaluated, and by calculating its unbound plasma concentrations, imatinib CSF-UPR's can be established as 7.7-56.2% [15]. Further investigations confirmed these findings. Imatinib CSF-TPR was 2.6% in a patient diagnosed with a CSF lymphoid blast crisis, while displaying a major cytogenic response in the bone marrow after 16 months of imatinib treatment. This corresponds to a calculated CSF-UPR of 52.6% [17]. In a randomized, multicenter phase 2 trial, plasma and CSF samples were collected from 17 BCR/ABL + ALL subjects with or without meningeosis and receiving imatinib. The CSF-TPRs were 1.8%. [18]. Imatinib CSF and plasma concentrations were further evaluated in parallel in four adult subjects of a multicenter clinical trial. One of the subjects was a biphenotypic Ph + CML patient, while the other three had been diagnosed with Ph + ALL. The CSF concentrations (mean: 0.044 µg/mL) were 74-fold lower than total plasma concentrations (3.27 µg/mL), corresponding to a CSF-UPR of 26.9% [19].
Dasatinib concentrations were below the detection limit in the CSF of a female adult patient treated with Ph + ALL and an extramedullary and meningeal relapse following bone marrow transplantation. The trough plasma dasatinib concentration was 32 ng/mL (CSF-TPR: 0.23-1.5%) [27]. Following the detection of large individual variability in the systemic exposure to dasatinib, Gong et al. measured pairs of the CSF and plasma concentrations of the substance in five Ph + ALL adult patients after giving doses of 100 mg or 140 mg. Only two pairs of samples contained dasatinib in quantifiable concentrations in both media. The CSF-TPRs were 0.75% and 1.42%, while the calculated CSF-UPRs were 18.7% and 37.2% [28].
Four leukemia patients (three diagnosed with Ph + ALL and one with Ph + CML and a blast crisis), all with CNS relapse after allogeneic stem cell transplantation, received nilotinib (Tasigna ® ). Seventeen matched pairs of CSF and plasma samples were collected. The CSF-UPRs were calculated by taking a 98% protein binding rate into account. The calculated concentration ratios were 12%, 20%, 30%, and 68%, pointing to large individual differences in the availability of the drug [29]. In a group comprising 30 Ph + ALL patients aged 15 years or older, only non-quantifiable traces of nilotinib were found in the CSF samples collected [30].
The penetration of the selective BCL2-inhibitor venetoclax (Venclyxto ® ) through the blood-brain barrier was also poor; however, it corresponded to the in vitro IC 50 of the drug in an adult, male chronic lymphocytic leukemia patient diagnosed with trisomy 12, IGHV unmutated (VH4L) chronic lymphoid leukemia and experiencing a CNS relapse. Time-matched pairs of plasma and CSF samples were assayed after their collection in steady state, after 2 h and 23 h of drug intake, with 0.23% and 2.89% concentration ratios obtained. The unbound fraction of venetoclax is smaller than 1% of the total circulating amount; therefore, the CSF concentrations corresponded to approximately 10-29% of the unbound plasma levels [95]. Venetoclax concentrations were evaluated 23, 30, and 37 days after initiating treatment in another male adult patient presenting with a complete remission in the bone marrow after hematopoietic stem cell transplant, but with a blast crisis detected in the CNS, and formerly receiving other chemotherapy. The CSF-TPRs were 0.32-0.40%, corresponding to CSF-UPRs of at least 32-40% [96].
An extremely low CSF concentration (0.1 ng/mL) of ponatinib (Iclusig ® ), another very heavily ( > 99%) protein-bound drug, was observed in a 3-year old girl diagnosed with Ph + acute lymphoblastic leukemia which had been confirmed to have penetrated the CNS [37].

Central Nervous System Metastases of Non Small-Cell Lung Cancer
The first manuscript discussing the quantitation of OACDs in CSF for monitoring their efficacy regarding the treatment of the CNS metastases of NSCLC was published on the epidermal growth factor receptor inhibitor gefitinib (Iressa ® ). This was a case report presenting a Japanese male patient diagnosed with NSCLC and developing carcinomatous meningitis. Ten days after the initiation of gefitinib treatment, the drug was assayed in serum and in CSF before and 2 h after the intake of 250 mg drug. At both time points, the observed CSF concentrations were negligible, 0.9 nmol/L, while serum concentrations of 117 and 132 nmol/L were attained. Assuming a 97% protein binding rate, this corresponds to CSF/unbound serum concentration ratios of 22.7% and 25.6% [39,102]. Interestingly, significant positive linear correlations of gefitinib CSF and plasma levels were revealed in multiple research works (Figure 3) [40][41][42]. In contrast, the results of a phase 1 open-label trial of a novel, high-dose gefitinib treatment conducted with the involvement of seven patients diagnosed with leptomeningeal metastases of NSCLC showed that this approach did not result in an improved penetration of gefitinib into the CSF [43]. Evidence exists for supporting that the low penetration rate of gefitinib may be increased by whole-brain radiotherapy, an intervention considered to be an efficient strategy to improve blood-brain barrier permeability [41]. However, contrasting results have also been published [44]. A direct comparison of the concentrations of gefitinib and erlotinib, which have similar chemical structures, in the CSF of patients diagnosed with leptomeningeal metastases, resulted in the conclusion that erlotinib attained higher molar concentrations and a higher rate of penetration into the CNS [45].
CSF [43]. Evidence exists for supporting that the low penetration rate of gefitinib may be increased by whole-brain radiotherapy, an intervention considered to be an efficient strategy to improve blood-brain barrier permeability [41]. However, contrasting results have also been published [44]. A direct comparison of the concentrations of gefitinib and erlotinib, which have similar chemical structures, in the CSF of patients diagnosed with leptomeningeal metastases, resulted in the conclusion that erlotinib attained higher molar concentrations and a higher rate of penetration into the CNS [45]. Two years after the publication of the first measurement of gefitinib concentrations in CSF, erlotinib concentrations were evaluated in three lung adenocarcinoma patients developing leptomeningeal metastases during gefitinib therapy. Twenty-eight days after switching to erlotinib, clinical improvement was observed, accompanied by 2.5-13.3% CSF-TPRs, corresponding to CSF-UPRs of 36-190% [50,97]. Four cases of Asian female adult NSCLC patients who had developed adenocarcinoma as a CNS metastasis and started to receive 150 mg erlotinib once daily were described by Togashi et al. Matched Two years after the publication of the first measurement of gefitinib concentrations in CSF, erlotinib concentrations were evaluated in three lung adenocarcinoma patients developing leptomeningeal metastases during gefitinib therapy. Twenty-eight days after switching to erlotinib, clinical improvement was observed, accompanied by 2.5-13.3% CSF-TPRs, corresponding to CSF-UPRs of 36-190% [50,97]. Four cases of Asian female adult NSCLC patients who had developed adenocarcinoma as a CNS metastasis and started to receive 150 mg erlotinib once daily were described by Togashi et al. Matched pairs of CSF and plasma samples were collected on day 8 of the treatment. Similar penetration of the drug and its active metabolite OSI-420 into the CSF was found. The authors provided the CSF-TPRs and the CSF concentrations, which allows the calculation of total and unbound plasma concentrations, as well as CSF-UPRs (45.7-110%). The efficiency of erlotinib to penetrate the blood-brain barrier was concluded to be higher than that of gefitinib, and allows the effective treatment of EGFR wild-type cases as well [51]. Yet another study involving six adult NSCLC patients with brain metastases confirmed that erlotinib could reach a mean penetration rate of 4.4%, corresponding to a CSF-UPR of 47.2%. The CSF concentrations of the drug were associated with the outcome, with the highest levels attained in patients showing partial response to therapy, and the lowest seen in those with progression [52]. At steady state, the CSF penetration rate of erlotinib was determined as 5.6% (corresponding to a CSF-UPR of 77.0%) in a female patient diagnosed with stage IV lung cancer and stage I breast cancer, and receiving a combination of erlotinib and bevacizumab [53]. A considerably lower ratio of 1.15% (corresponding to a CSF-UPR of 16.4%) was observed, however, in a woman with stage IV NSCLC and leptomeningeal metastasis and receiving 1500 mg erlotinib weekly [54]. A similarly low penetration rate of erlotinib (1.6-2.6%) was identified in six Chinese adult NSCLC patients with leptomeningeal metastasis refractory to gefitinib treatment. Three patients received premetrexed and cisplatin in addition to erlotinib, while the other three received only erlotinib. There was no difference in the penetration rates between the two patient groups. The calculated CSF-UPRs were 22.8-36.6% [55]. A very strong linear correlation was identified, at the same time, between plasma and CSF erlotinib concentrations [56]. This finding was also confirmed by another study (Figure 3) [57]. A phase 2 single arm trial was conducted to reveal the efficacy of erlotinib in stage IV NSCLC with leptomeningeal metastasis (LOGIK11001) by Nosaki et al. The primary endpoint was the cytological clearance rate, and the secondary endpoints were time to disease progression, overall survival, toxicity, and quality of life. Plasma and CSF concentrations of erlotinib were determined in single steady-state samples collected from 12 participants. The mean penetration rate was 2.9-12.1%, corresponding to CSF/unbound concentration ratios of 41.9-173%. Again, a good correlation was observed between the plasma and the CSF concentrations (R 2 = 0.6247), regardless of the cytological response Figure 3 [58].
In a comparative study conducted to evaluate the penetration rate of standard (150 mg/die and 250 mg/die, respectively, administered for seven days) versus pulsatile high-dose erlotinib (1500 mg on day eight and fifteen) and gefitinib (2500 mg/die from day eight to fifteen) in NSCLC patients with brain metastases who progressed on standard doses, both drugs attained higher concentrations in the CSF as a result of high-dose administration, with a constant CSF-TPR of 2% in the case of erlotinib, and a saturable penetration rate of gefitinib with no increases in CSF levels predicted for doses of 839 mg or higher (Figure 3). In addition, those undergoing whole-brain radiotherapy attained disproportionately higher CSF concentrations of the drugs. Adverse effects were more prevalent in patients receiving erlotinib, with the high doses of gefitinib being well tolerated [46]. The next drug assayed in the CSF was crizotinib (Xalkori ® ), with negligible bloodbrain barrier penetration rates observed. A CSF-TPR of 0.26% (corresponding to a CSF-UPR of 2.89%, assuming a 9% unbound fraction of the drug) was found in a 29-year old Caucasian male diagnosed with stage IV NSCLC and treated first with cisplatin plus pemetrexed, then with erlotinib, and finally with crizotinib. The attained CSF concentration was substantially lower than the established 50% inhibitory concentration (IC 50 ) required to inhibit mutant cell lines against which crizotinib had been tested [70,103]. In two ALKpositive male adult NSCLC patients developing brain metastases, crizotinib CSF/total serum concentration ratio was 0.06% and 0.1%, corresponding to CSF/unbound serum concentration ratios of 0.66% and 1.1%, respectively [71]. Three CSF samples of a 60-year-old male patient diagnosed with ALK-rearrangement-positive NSCLC and receiving 250 mg crizotinib twice daily after developing brain metastases were assayed for crizotinib at oneweek intervals following whole brain radiotherapy (an additional sample was processed before conducting WBRT). Crizotinib was undetectable in the samples collected before and one week after WBRT, while 6.2 and 6.3 ng/mL concentrations were found after two and three weeks, respectively, accounting for 3.5% and 2.2% of the total, and for 39.0% and 24.5% unbound plasma concentrations [72]. In another male patient diagnosed with stage IIA lung adenocarcinoma and brain metastasis, a CSF-TPR of 2.6% (corresponding to a CSF-UPR of 30.4%) was achieved at a single sampling point following WBRT. The CNS symptoms diminished, and the negativity of CSF to malignant cells was confirmed. Comparing this result to earlier findings yielded the conclusion that WBRT may enhance the CNS penetration and the clinical efficacy of crizotinib [73].
In a single-arm, open-label, multicenter phase 1/2 study conducted with the involvement of adult subjects with histologically confirmed, locally advanced or metastatic NSCLC with crizotinib-resistant ALK-positive rearrangement and receiving 600 mg or 900 mg alectinib (Alecensa ® ) twice a day in the fixed dose phase, five matched alectinib CSF-plasma concentration pairs were obtained. The CSF concentrations not only showed positive correlation with the unbound plasma fraction of alectinib (which corresponded to 0.3% of the total amount), but were also equivalent or higher. The extrapolated trough CSF concentration exceeded the reported in vitro IC 50 of alectinib for ALK inhibition [75]. In an institutional case series comprising eleven adult subjects diagnosed with histologically confirmed ALK-positive NSCLC and receiving 600 mg alectinib twice daily until disease progression, unacceptable toxicity or withdrawal of consent, matched CSF-serum concentration pairs were obtained in two patients in the second month of alectinib therapy. The total serum concentrations were 694 ng/mL and 707 ng/mL, both corresponding to 2.1 ng/mL unbound serum concentrations. The calculated CSF/unbound serum concentration ratios were, therefore, 100% and 30% in the two patients [76].
The evaluation of afatinib (Giotrif ® ) CSF levels was first described in a woman diagnosed with stage IV adenocarcinoma of the lung with an underlying mutation of the EGFR gene. Two CSF samples were assayed, and afatinib was found to attain a penetration rate lower than 1%, with a calculated CSF-UPR of 13.9% when 95% protein binding rate of the drug is assumed [59,104]. A remarkable case of a female patient diagnosed ten years earlier reporting with stage IV adenocarcinoma of the lung with an EGFR mutation was also described. Afatinib (40 mg/die, deescalated to 30 mg/die after four months) was administered as the eighth line of treatment following interchanging periods of progression and remission. Trough plasma and CSF concentrations were assayed at three, four and five months following the initiation of afatinib dosing. The CSF-TPR's were 0.28-0.40%, while the calculated CSF-UPRs are 7.5-8.8%. The total plasma concentrations were 19.0-33.4 ng/mL, which can be measured with relative convenience using liquid chromatography-tandem mass spectrometry (LC-MS/MS), but the obtained CSF levels of 0.05-0.14 ng/mL clearly indicate that assaying afatinib in the CSF is a major analytical challenge [60]. Further, a prospective multicenter trial was conducted with the involvement of 11 patients diagnosed with EGFR mutation-positive NSCLC with leptomeningeal carcinomatosis and with the aim of evaluating the CSF penetration rates and the clinical efficacy of afatinib. Participants received 40 mg afatinib once a day. On day eight, the trough concentrations were assayed in plasma and in CSF. Afatinib could be quantitated in the CSF samples of eight subjects (72.7%). The CSF-TPRs were 0.1-3.1%, with a single case of 9.3% which resulted from an unusually low plasma concentration (corresponding to 44.4% of the next value in the ranked series of the measured concentrations), accompanied by the second-highest CSF concentration. This corresponds to CSF-UPRs of 2.1-185%. It was concluded that the ability of EGFR tyrosine kinase inhibitors to penetrate the CSF should be assessed along with the efficacy of the drug against tumors with particular mutation types [61].
The penetration of icotinib (Conmana ® ), an OACD currently approved in China, into CSF was first evaluated in a phase 2 clinical study involving ten patients following the administration of 125 mg in a three-times-per-day regime. Meanwhile, WBRT was delivered in 3-Gy fractions once per day, five days per week, to a total dose of 30 Gy. The mean total plasma concentrations were 940.6±503.8 ng/mL (corresponding to 47.0±25.2 ng/mL unbound concentrations), while the mean CSF concentrations were 11.6±9.1 ng/mL in samples collected two hours after drug intake. The CSF-TPR was 1.4±1.1%, and the mean CSF-UPR can be calculated as 24.7% [66]. The impact of WBRT on the CSF penetration of icotinib was directly investigated in fifteen patients receiving escalating dose levels (125-352 mg) three times a day. Blood and CSF samples were collected immediately before beginning the WBRT treatment (applied in fixed doses of 37.5 Gy, five times a week, lasting for three weeks), immediately after terminating WBRT therapy, and four weeks into the follow-up period. The CSF-TPR's of icotinib were 2.4-3.7% in a dose range of 125-500 mg (peculiarly, 6.1% at 375 mg), while the CSF-UPR can be calculated as 52.0-58.0% (130% at 375 mg) [67].
The CSF concentrations of osimertinib were first measured in an NSCLC patient with leptomeningeal metastases and EGFR-TKI resistance. A poor penetration rate (1.47%) was observed [62]. In an open-label, single-arm, multicenter, prospective study (APOLLO), twelve adult patients donated matched blood and CSF samples. The evaluation of osimertinib concentrations was based on the unbound drug fractions. A strong linear correlation was found between blood and CSF levels (r = 0.8306). Based on these calculations, the median CSF-UPR of osimertinib was 31.7% (19.8-57.8%) after six weeks of treatment [63]. In a phase 2 study involving radiotherapy-naive adult patients diagnosed with T790M EGFR mutation-positive NSCLC and CNS metastasis, who had been previously treated with EGFR tyrosine kinase inhibitors, the plasma and CSF concentrations of osimertinib and its pharmacologically active metabolite were assessed in seven participants on day twenty-two of osimertinib therapy. The CSF-TPRs of the drug and the metabolite were 0.79% (0.43-1.32%) and 0.53% (0.31-0.64%), respectively, corresponding to 15.8% (8.6-26.4%) in the case of the parent drug by assuming 99% plasma protein binding rate [64].
Tepotinib plasma and CSF concentrations were evaluated in a male adult patient diagnosed with stage IIIA lung adenocarcinoma. EGFR mutation and ALK fusion gene were not detected. Following right lung pneumonectomy, a brain metastasis was identified in the left cerebrum which later progressed to leptomeningeal metastasis and hydrocephalus in spite of treatment with cisplatin and pemetrexed. A tepotinib regimen (500 mg/die) was started. On day 20 of therapy, the tepotinib CSF-TPR achieved 1.83% in the matched samples collected four hours post-dose. The attained concentration was judged to have exceeded the IC 50 [90]. In a female patient diagnosed with NSCLC with MET exon 14 skipping mutation and with brain metastases, and having received WBRT, remarkable clinical improvement was achieved after a 1-month treatment with tepotinib (500 mg/die). The penetration rates of tepotinib into the CSF at two, four and eight weeks of therapy were 1.19%, 1.42%, and 1.73%, respectively. By taking the 98% protein binding rate of tepotinib into account, the CSF-UPRs can be calculated as 60.0%, 71.1%, and 86.6%, respectively, based on the data described by the authors [91].
Lorlatinib (Lorviqua ® ) was monitored in the CSF in an ongoing, open-arm, multicenter phase 1/2 trial with the aim to further investigate the penetration of the drug into the CNS. Five patients with suspected or confirmed leptomeningeal carcinomatosis not visualized on magnetic resonance imaging, or carcinomatous meningitis, were included. Samples were collected at baseline and a later yet undefined point of the study. The CSF/plasma unbound lorlatinib concentration ratios were 61-96%, and showed very strong correlation (adjusted r 2 = 0.96). The CSF/total plasma lorlatinib concentration ratios were 21-33%. The results indicated that lorlatinib concentrations exceeded the minimum efficacy concentrations in all of the patients regarding wild-type anaplastic lymphoma kinase (ALK) and the L1196M ALK resistance mutation. The authors concluded that this supported the broad coverage of these mutations, and, in approximately one-third of patients, the coverage of the G1202R ALK resistance mutation [77].
The CSF concentrations of pralsetinib (Gavreto ® ) and osimertinib were investigated in an adult patient with an EGFR-mutant NSCLC with acquired RET fusions and meningeal metastasis after four months of co-treatment with pralsetinib and osimertinib. Pralsetinib attained concentrations of 91.3 µmol/L and 0.705 µmol/L in plasma and CSF, respectively (ratio: 0.77%, corresponding to a CSF-UPR of 15.4%). Osimertinib concentrations were 2.149 µmol/l and 0.0237 µmol/L, respectively (ratio: 1.10%, corresponding to a CSF-UPR of 110%). Despite the lower CSF/unbound concentration ratios, pralsetinib levels were judged to be sufficiently high both in plasma and in CSF to inhibit the CCDC6-RET-mutated protein, indicating that pralsetinib is more efficient than osimertinib to treat this mutation [65].

Metastases of Other Malignancies in the Central Nervous System
Lapatinib inhibits both EGFR and HER2; therefore, it has activity against brain metastases developing from HER2-positive metastatic breast cancer. This activity may be enhanced by combining lapatinib with capecitabine. Nevertheless, 0.9-1.3% of CSF-TPRs of lapatinib were observed in two adult female patients diagnosed with HER2-positive (one HR-negative and one HR-positive) ductal carcinoma yielding CNS metastases. The CSF-UPRs can be calculated as 8.6-12.9% in these two patients [83,105]. Neratinib, another HER2 tyrosine kinase inhibitor, was absent (<1.50 ng/mL) in the CSF samples of three adult HER2-positive breast cancer patients [84]. Vemurafenib was, on the other hand, quantitated successfully in the matched CSF and plasma samples of patients treated with the drug in a dose of 960 mg, given twice daily, for brain metastatic BRAF-V600 mutated melanoma. The CSF-UPRs were 28-250%, assuming a 99% protein binding rate [68,106].

Malignant Ascites
Malignant ascites is a rare condition secondary to abdominal malignancies [107]. In an elderly adult patient diagnosed with papillary renal cell carcinoma and undergoing treatment first with pazopanib (Votrient ® ), then with sunitinib, concentrations of the administered OACD were monitored in plasma and in ascitic fluid. The concentrations measured in the ascitic fluid were equivalent to or higher than those assayed in the systemic circulation, and, following an early phase with sufficient plasma levels, systemic concentrations became subtherapeutic [88]. The ascited fluid concentrations of the drugs remained high after discontinuation of treatment. While the underlying reason of the accumulation of these drugs in the ascitic fluid is not evident, it was proposed that it acted as a sink of the administered OACDs, while the strong binding of pazopanib and sunitinib to albumin may have facilitated the extravasation of the drugs.

Monitoring OACDs to Control Toxicity
Monitoring the Exposure of the Infant to the Drug during Breastfeeding CML occurs very rarely during pregnancy, at an estimated rate of 1:750 000. Imatinib is employed for treating Ph + cases developing during pregnancy, an approach which may cause harm to the fetus and the newborn. Assaying the drug in breast milk is valuable for characterizing the exposure of the infant. The first appearance of the measurement of imatinib in breast milk was the description of a case with the imatinib concentrations being approximately 60% of the lower limit of the currently accepted blood reference range (1000-3000 ng/mL). Its pharmacologically active metabolite, however, displayed accumulation in breast milk [20]. Another patient on 400 mg once-daily imatinib donated blood and breast milk samples on a single day, 1, 2, 3, 4, and 9 h after drug intake. The concentrations of imatinib and its active metabolite in milk reached 0.5 and 0.9 of those found in plasma, respectively. The authors concluded that the maximum intake of the infant was 3 mg imatinib/day, and should be considered safe [21]. A case described two years later described a Ph + CML patient receiving the same dose resulting in complete hematological and cytogenetic remission. Blood was drawn on day 2, while breast milk was collected on days 7, 14, 15, and 16 postpartum. The imatinib concentrations measured both in plasma (2385 ng/mL) and in breast milk (1430-2623 ng/mL) were in the therapeutic range. The authors concluded that, since the long-term effects of imatinib on infants are unknown, breastfeeding is not advisable when imatinib is administered [22]. This conclusion was confirmed by the presenters of another case when imatinib treatment was initiated immediately after delivery. While the concentrations of imatinib were relatively low in breast milk, those of the active metabolite attained threefold concentrations of those measured in plasma, clearly displaying accumulation [23]. Yet in another patient, the concentrations of imatinib and the active metabolite were measured in breast milk 99 h after the last intake. The attained concentrations were 19 ng/mL and 600 ng/mL, respectively, pointing to a very significant accumulation of the metabolite in breast milk. Neonatal urine was also evaluated, with 90 ng/mL imatinib and 165 ng/mL active metabolite concentrations detected. These results indicate that the infant was exposed to the drug, and, to an even greater extent, to the metabolite. This case raises the clinical relevance of assessing the concentrations of oral anticancer medications taken during pregnancy in neonatal urine for evaluating the potential impacts on the newborns [24]. In the most recently published case report the milk/plasma ratio of imatinib attained 0.35 at 5 days postpartum. Blood was also collected from the infant on the same day to reveal a 27-ng/mL concentration of imatinib, which was considered to be safe by the authors [25].
Everolimus is primarily administered as an immunosuppressant based on its ability to inhibit the mammalian target of rapamycin (mTOR) functional complex mTORC1. In a heart-transplanted patient, everolimus therapy was continued during pregnancy and following delivery. At 48 h postpartum, the drug was not detectable in the colostrum, indicating that the evaluation of the immunosuppression of the newborn had to be based on its prepartum administration [81].

Monitoring Other Types of Toxicity
Oral anticancer medications have serious adverse effects, including low blood cell counts, resulting in an increased susceptibility to infections and, potentially, bleeding, as well as dermal and gastrointestinal symptoms. Several of these may prompt the discontinuation of therapy. Efforts have therefore been made to identify the relationships between the presentation of the drugs in non-targeted organs and fluid compartments, and the development of adverse symptoms.
Pleural effusion may be induced by tyrosine kinase inhibitors. In a young male adult patient who had developed pleural effusion from dasatinib earlier, nilotinib therapy again led to the formation of the effusate. The measured nilotinib concentrations were 927 ng/mL and 2092 ng/mL in plasma and in the pleural effusate, respectively, clearly indicating the accumulation of nilotinib in the latter medium. Other possible causes, including malignancy, were excluded. The severity of this adverse effect is shown by the fact that, eventually, performing endotracheal intubation and left thoracic drainage was required [31].
The relationship between the occurrence of stomatitis and everolimus (Afinitor ® ) levels in saliva was investigated in 11 cancer patients receiving everolimus in a once-daily (10 mg) or twice-daily (2 × 5 mg) regime. Both the plasma and saliva concentrations of the drug were higher in patients with stomatitis than in those who did not develop this condition. While the statistical significance of this difference was low, this result may indicate the utility of everolimus saliva assays concerning the prevention of the occurrence of stomatitis. Of note, the rate of the penetration of everolimus into saliva was extremely low (0.8%) with high interindividual variability (67.7%) [82].
Imatinib has been demonstrated to cross the brain-testis barrier and to reach equilibrium. Imatinib concentrations reached concentrations of 1471 ± 570 ng/mL and 1397±425 ng/mL in the plasma and the semen of eleven male CML patients, respectively. The clinical relevance of the assay was confirmed by the finding that the number, the survival rate, and the activity of sperms were reduced in these patients. Reproductive hormone structures and sex hormone concentrations were unaffected [26].
Panobinostat was not detected in the CSF of patients diagnosed with human immunodeficiency virus (HIV) infection [92]. In addition, it was not present in the CSF samples of pediatric patients with refractory hematological malignancies [93]. It was concluded in these works that panobinostat did not cause CNS symptoms.

Monitoring the Treatment of Mental Disorders
It is increasingly acknowledged that certain OACDs may be effective against neurodegenerative and autoimmune diseases [9,108]. Nilotinib, a BCR-ABL tyrosine kinase inhibitor, has been investigated in multiple cases as a medication against mental disorders, such as Parkinson's disease and Alzheimer's disease [109,110]. The rationale of these indications is that nilotinib leads to the degradation of misfolded α-synuclein by autophagy [111]. In addition, in preclinical studies, nilotinib increased dopaminergic neuron survival in the CNS, and improved motor and cognitive outcomes in in vivo models. Abl inhibition has been demonstrated to reduce oxidative stress, and to protect dopaminergic neurons [112].
In an open label pilot study conducted to investigate the safety and tolerability of two doses of nilotinib, the drug penetrated readily into the CSF, and remained detectable there for five hours, when administered to stage 3-5 Parkinson's disease patients in low doses (150-300 mg/die). This was accompanied by a steady increase in plasma concentrations. The CSF-TPR was higher when the lower dose was administered, with a comparable level of Abl inhibition [32]. Further research revealed that the penetration of nilotinib into CSF was dose-dependent in the dose range 150-400 mg, with 200 mg exerting optimal effects. Again, the CSF-TPR was very similar at various doses (0.5-1.0%). Nevertheless, avoiding higher doses was recommended since more side-and off-target effects were detected in the CNS [33]. A phase 2 randomized clinical trial was published with the involvement of 75 participants, 50 of whom received nilotinib. The CSF-TPR was considerably lower, 0.33% and 0.53% after applying 150 mg and 300 mg nilotinib, respectively [34].
The most recent evaluation of nilotinib delivered results which contradicted some key findings of the above works, although the safety and tolerability of low-dose nilotinib was still acceptable. The measured CSF penetration was in concert with previous findings. This was a six-month, multicenter, randomized, parallel-group, double-blind, placebo-controlled trial conducted with the involvement of 76 participants, 51 of whom received nilotinib (150 mg or 300 mg pro die). Based on the evaluation of the geometric means of measured drug concentrations, the administration of 150 mg or 300 mg resulted in 0.61-1.10 ng/mL and 1.10-1.90 ng/mL peak CSF nilotinib concentrations along with 343.3-524.4 ng/mL and 485.8-621.2 ng/mL peak serum concentrations, respectively, after three months of treatment. This corresponded to 0.16-0.23% and 0.20-0.32% CSF penetration rates, respectively. In contrast to the favorable outcomes of the earlier studies, this trial ended with the conclusion that the low penetration rates were associated with no treatment-related alterations of dopamine metabolites in the CSF. Therefore, the changes in the protein biomarkers (α-synuclein, phospho-α-synuclein, and phospho-tau) alone provided weak evidence of the clinical efficacy of nilotinib treatment [35].
In animal models of neurodegeneration, nilotinib promoted the degradation of proteins Aβ/amyloid protein and the microtubule-associated protein tau [113]. This result prompted a phase 2, randomized, double-blind, placebo-controlled study to evaluate the effects of nilotinib in mild to moderate Alzheimer's disease. The label arm received 150 mg nilotinib daily for 6 months, followed by 300 mg daily for another 6 months. The ratios of the mean CSF and plasma concentrations were 0.29% and 0.27% at 150 mg and 300 mg doses, respectively. The ratios of the areas under the concentration-time curves were 0.30% and 0.33%, respectively [36].
The above works included the detailed evaluation of quantitative changes in the pharmacodynamic variables, such as microtubule-associated protein tau or amyloid proteins. Although the penetration rate of nilotinib into the CSF was very low, these were associated with statistically significant pharmacodynamic improvement and measurable clinical efficacy.

Bioanalytical Methods of Monitoring OACD Concentrations in Peripheral Fluid Spaces
All of the described bioanalytical methods relied on chromatographic separation using high-performance or ultra-high performance liquid chromatography. Mass spectrometry was chosen for detection by most authors, but multiple examples of applying ultraviolet-visible (UV-VIS) light absorbance detection for the assessment of afatinib, erlotinib, gefitinib, imatinib, nilotinib, and vemurafenib were found.
Solid phase extraction with a polymeric reversed-phase sorbent was employed for extracting crizotinib from CSF [71] and imatinib from leukocytes [13]. Afatinib was recovered from CSF using an octadecyl silica loading [61]. The extraction of vismodegib from CSF was feased by employing a strong mixed-mode cation exchange sorbent [94]. Gefitinib was recovered from CSF using an unspecified cartridge [44]. Equilibrium dialysis was employed to assess the unbound concentrations of ceritinib, ribociclib, and vismodegib directly in plasma [74,78,79,84].
In addition, special pretreatment procedures were described by a few authors. Imatinib was recovered from peripheral blood mononuclear cells by sonicating the defrosted pellet in an ice-water bath, followed by centrifugation, counting cells in the supernatant, washing with acetonitrile-methanol 1:1 and solvent exchange [14]. Ribociclib was assayed in CSF after dilution with methanol-water 1:1, acidification with water containing 0.2% formic acid, and centrifugation [80]. Automated sample preparation was employed as part of the analysis of everolimus [81], imatinib [18], and nilotinib [29].
Detailed information on the methods employed for monitoring OACDs in peripheral fluid spaces is provided in Table 4. Eight of the eighty-five publications (9.4%) failed to provide any methodological information or a reference to another manuscript describing the methodology employed for the quantitation of OACDs in peripheral fluid spaces. Altogether, 29 methodological publications were cited in the included manuscripts. The work of Jones et al. was cited by most included works [114]. Only five methodological works described the analysis of OACDs in peripheral fluid spaces, namely CSF (three publications), colostrum (one publication), or PBMC (one publication). The rest of the cited methodological papers described the analysis of one or more OACDs in blood.

Discussion
The rapid growth of the number of related publications reflects the increasing clinical interest in monitoring OACDs in therapeutically relevant extravascular fluids. Nevertheless, the range of substances that have been monitored in these compartments with the aim of supporting clinical decision making comprises the minor segment of marketed OACDs. Currently, imatinib is the most extensively studied drug, followed by erlotinib, gefitinib, and nilotinib. Interest in studying recently approved entities, such as dasatinib, osimertinib, panobinostat, and ribociclib, is also rising.
To date, frequently monitored peripheral fluid spaces have included cerebrospinal fluid, and, to a lesser extent, breast milk. Sporadic examples of monitoring OACDs in pleural effusion fluid, ascitic fluid, the intracellular space of peripheral blood mononuclear cells, semen, and saliva have been encountered. Collecting, handling, and processing samples originating from these fluid spaces requires expertise and, regarding CSF, pleural effusate, and ascitic fluid, specialized clinical infrastructure. Due to this limitation, as well as to the need to use specialized and resource-intensive analytical technology, it is likely that OACD monitoring in peripheral fluid spaces remains a competence of centers of excellence in oncology.
The attainment of very low OACD concentrations in CSF seems to have been unexpected by several authors. One explanation could be the poor permeability of the blood-brain barrier to these drugs, but this assumption has been contradicted by results showing that WBRT, an adjuvant intervention undertaken to increase this permeability, had not always led to increased penetration rates [67,72,91]. The application of WBRT is part of an effort to employ multimodal therapy against CNS malignancies, yet recent reports have shown that it may have detrimental adverse effects, and should not be considered as a stan-dard measure in the therapy of NSCLC patients developing brain metastases. Experience with WBRT is also controversial regarding the treatment of primary CNS lymphomas [141]. At the same time, it has been found effective in the therapy of brain metastases of breast cancer patients, especially when combined with carboplatin injected intravenously [142]. Conventional photon radiotherapy, a similar approach with a more favorable adverse effect profile, has also been proposed for increasing the penetration rate of OACDs through the blood-brain barrier [143]. Various options of using more focal radiotherapy have also been described [144].
Another interpretation is that the CSF concentrations of OACD substances could be associated with the unbound plasma fractions. Several examples of a correlation observed between unbound serum/plasma concentrations and CSF levels confirm this assumption ( Figure 3). Since the unbound fractions of various OACDs display considerable differences, it is indeed rational to judge CNS penetration based on these fractions instead of the total serum/plasma levels. The evaluation of the unbound fractions shows that the concentrations of some drugs attained in the CSF are equal to or even higher than unbound circulating concentrations. The negligible presence of everolimus in saliva, another medium accessed only by the unbound plasma fractions, confirms this rationale. No approved clinical approaches exist for establishing individual protein binding rates. Equilibrium dialysis has been used as an experimental sample pretreatment procedure for determining unbound plasma concentrations of OACDs [74,78,79,94,145]. Microdialysis has the potential to be employed for this purpose, but no examples of its application for the assessment of unbound OACD concentrations were identified. A promising sample pretreatment technology has recently become available for the rapid assessment of the extent of protein binding. The device fits into the sample preparation workflow employed by LC-MS/MS-based TDM laboratories, but there is still very limited experience regarding its use [146].
The extent of plasma protein binding may not be the only factor of the penetration of OACDs through the blood-brain barrier. Guntner et al. have shown with seven OACD substances that molecule size and the affinity of the molecule to p-glycoprotein (ABCB1 or MDR1, EC 7.6.2.2) are also key determinants. In accordance, the permeability of the bloodbrain barrier to dasatinib, imatinib, regorafenib, ribociclib, and vorinostat was higher than to nintedanib or panobinostat. The comparison of experimental results to those obtained using computer models nevertheless indicated that further variables, currently unidentified, are likely to play an important role in this process [12].
In sum, more research is needed to find dosages and monitoring approaches that result in the attainment of clinically sufficient CSF concentrations in all patients. Aggressive dosing, the artificial facilitation of the penetration of drugs through the blood-brain barrier, or the administration of drug combinations containing a component which inhibits p-glycoprotein or other drug-eliminating proteins relevant to a specific OACD are potential strategies for the more efficient therapy of CNS malignancies. A methodology is also emerging to predict OACD treatment efficacy by comparing the drug concentrations measured in the target peripheral fluid to the in vitro IC 50 established for the given malignant cell line, and based on this relationship, by creating a mathematical link between the pharmacokinetic and pharmacodynamic properties of the administered drug. In the future, this approach may prove useful in developing precision dosing schemes with pharmacokinetic-pharmacodynamic targets, in an analogy to those already employed for guiding antibiotic therapy.
In sharp contrast to the observations made in the CSF, high penetration rates or even the accumulation of OACDs were consistently described in exudates formed by pleural effusion and malignant ascites, and in excreta such as breast milk and semen. Imatinib showed considerable accumulation in buffy coat cells and in peripheral blood mononuclear cells. These findings indicate that the consideration of third spaces as pharmacokinetic compartments may be rational in patients treated with lung or breast cancer, as well as in leukemia patients.
There has been a solid consensus in relying on liquid chromatography-based analytical approaches for the therapeutic monitoring of OACDs in peripheral fluid spaces. Several early methods relied on the use of UV-VIS detectors, but LC-MS/MS has by now emerged as the primary analytical technique as a result of ensuring sufficient selectivity and sensitivity, requiring small sample volumes for the analysis, and allowing the high-throughput processing of peripheral fluid space samples. When applied for the clinical analysis of OACDs in these compartments, the main steps of these methodologies were reversed phase chromatographic separation followed by positive electrospray ionization and multiple reaction monitoring. The simple and rapid process of deproteinization was in most cases sufficient for the pretreatment of samples. A common weakness of the analytical methodologies employed in the reviewed records is that they had not undergone comprehensive validation, lowering the credibility of the presented results.
The application of equilibrium dialysis to retrieve direct clinical pharmacological information fits into a series of related emerging approaches, such as the rapid assessment of protein binding, or the partitioning between plasma and red blood cells. Such technologies are expected to facilitate the reporting of truly individualized, and, in a clinical sense, substantially more relevant information on the pharmacokinetic properties of drugs including OACDs in the future [146,147].
An important limitation of the performed evaluation is that only a minority of the retrieved publications described the outcomes of registered clinical trials. The majority of the works reported small-scale, researcher-initiated, unicentric studies, case series, or case reports. In addition, the methodologies employed for sample collection and analysis were uniquely developed by most investigators, limiting the comparability of results. Only a fraction of the subjects involved in the studies had given their consent for collecting CSF samples; consequently, the number of available CSF concentrations was small in several publications. Indeed, peripheral space drug monitoring was conducted as a collateral tool of diagnosis or patient status monitoring in several cases.
Malignancies are the leading causes of premature death worldwide, with breast and lung cancers underlying the largest number of new cases [148]. The importance of improving the treatment of these diseases is therefore beyond dispute. Therapeutic drug monitoring and research regarding model-informed precision dosing is currently based on the evaluation of drug concentrations in the systemic circulation, while evidence now shows that the monitoring of OACDs in therapeutically relevant extravascular fluid compartments can be equally important, especially for the better treatment of central nervous system malignancies. TDM laboratories providing service for large oncological centers can add a fundamental impetus by introducing suitable, validated, LC-MS/MS-based analytical methods for monitoring these drugs in peripheral fluid spaces, and in vitro approaches to determining unbound OACD concentrations. Establishing these competences is the first step for the introduction of therapy guidance based on highly relevant pharmacokinetic models and pharmacokinetic-pharmacodynamic indices, as well as for the early detection of suboptimal dosages and the risk of certain adverse effects. Since the number of available OACDs, as well as the range of their indications, is growing rapidly, the identification of further therapeutic goals and therapeutically relevant peripheral fluid spaces can be expected, maintaining a long-term need for the close cooperation of clinicians, clinical pharmacologists, and the TDM service in this field.

Conclusions
This review has revealed that the therapeutic monitoring of OACDs in peripheral fluid spaces is an important diagnostic tool for the assessment of the penetration of these substances into CSF and third space fluids, which is imperative for the optimization of drug administration, and of their appearance in excreta, which may convey important information on adverse effects and other forms of toxicity. LC-MS/MS is an established analytical technology for performing these measurements, with little effort required to transfer conventional, blood-based TDM methods. Nevertheless, dedicated centers of excellence are needed to perform such measurements routinely.
A range of indications has been identified for which the TDM of OACDs in peripheral fluid spaces can provide clinically powerful information. More systematic studies with rigorous quality control are needed, however, for elucidating the pharmacokinetic properties of OACDs, for setting quantitative therapeutic targets, and for establishing standard analytical methodology. Related research in pediatric populations still remains an unmet need.
After more than 20 years of using OACDs, an alarmingly small number of these substances has ever been investigated in a clinically important peripheral fluid space. In several malignancies, the administration of these medications cannot be optimized without knowledge regarding their quantities in these fluid spaces, especially in CSF; therefore, research should be focused on gathering information on all OACDs in this respect.