Introduction

Intraarterial (IA) drug delivery to GBM is a physiologically appealing strategy as it enables drug delivery throughout the extensive tumor capillary network and yet the delivery occurs within a limited volume of distribution which can be precisely controlled by careful injection. Tissue drug perfusion in this manner is theoretically very efficient as it follows the path of nutrient diffusion at very high regional concentrations. Notwithstanding, the pharmacokinetics of IA drug delivery is exceedingly complex. For effective IA delivery, drugs must be taken up rapidly and, preferably irreversibly, during their first pass through tissue circulation. In human subjects this occurs on a time scale of approximately 5 s. Full appreciation of IA pharmacokinetics therefore requires an understanding of both physiological and anatomic factors that influence arterial concentrations. These include blood flow hydrodynamic and injection parameters, endothelial-drug interactions, the kinetics of blood brain barrier (BBB) transit, and site-specific pharmacokinetics.

Despite past failures and persistent hurdles to effective GBM treatment, IA drugs are currently being used to treat recurrent gliomas and other brain tumors. This review highlights recent pre-clinical studies and ongoing trials utilizing IA drugs for the treatment of GBM. We describe how emerging technologies may offer novel insights into the complex pharmacokinetics of IA drugs, how concurrent reduction of blood flow enhances the regional effectiveness of IA delivery, and how new tumor targeting strategies and nanoparticle delivery could further improve IA delivery to GBM and other brain tumors. Emerging paradigms of IA drug delivery herald more effective and safe chemotherapy than was possible in the past.

Brief history

Modern chemotherapy emerged from toxicological studies with nitrogen mustards that were used during the Second World War [1]. Calvin Klopp at George Washington University used IA chemotherapy for head and neck tumors including GBM in 1950 [2]. In the 1960s, Charles Wilson systematically investigated IA chemotherapy for glioma treatment [3]. In 1972 Stanley Rapport demonstrated hyperosmotic disruption of the BBB [4]. Significant advances were made at the NIH under Oldfield et al. in the 1980s [5]. Robert Dedrick presented some of the key pharmacokinetic concepts at that time [6]. By the 1990s, interest in IA chemotherapy for brain tumor treatment generally started to wane in part due to the neurological complications reported with use of IA carmustine. Despite this trend, a number of individuals have championed the use of IA drugs. Edward Neuwelt, in particular, has kept the subject alive through the Blood Brain Barrier disruption program [7]. Unfortunately, a general loss of interest in IA treatments for brain cancer could not have been more untimely. Modern endovascular technology has rapidly advanced and safer therapeutic compounds are now available. In the last decade, significant advances have also been made in optical imaging and nanoparticle engineering—both of which may have an immense impact on IA drug delivery [8].

The case for IA drug delivery for GBM treatment

Local drug delivery

Effective IA injections rely on drug delivery through capillary networks that in turn are governed by the ability of nutrients to diffuse to the target cells. The diffusion distance for drugs to reach the target site should be short. Furthermore, drug delivery is ideally restricted to the site of pathology. Optimization of these delivery characteristics favorably alters the risk–benefit profile of IA intervention [9].

Success of IA chemotherapy in other cancers

IA chemotherapy has been used in treating many peripheral malignancies such as retinoblastoma [10], head and neck tumors [11], advanced liver [12, 13] and breast cancers [14], pancreatic cancer [15], penile [16] and other urogenital cancers [17]. The success of IA treatments—especially in cases of retinoblastoma and head and neck tumors—indicates that infusing drugs in proximity of the cerebral circulation can be safe. Success in the treatment of liver cancers additionally demonstrates the relevance of high first pass extraction of drugs delivered intraarterially [18].

Dose advantages

The net advantage of IA drug injection over IV injection depends on a number of factors, including the method of injection, the rate of injection, and the duration of infusions. PET measurements in human subjects have revealed a 50-fold increase in tumor tissue concentrations after IA versus IV injections [19].

Unique pharmacokinetics

As a general rule, IA drugs are beneficial when there is high local extraction, low regional blood flow, and high systemic clearance [6]. These parameters can be manipulated to improve IA drug delivery (Fig. 1 ). One such technique is to transiently reduce blood flow during delivery. Preclinical and clinical studies indicate that this can have a profound effect on IA drug delivery [20–23]. Flow arrest methods have been used to enhance the delivery of chemotherapeutic drugs to treat retinoblastomas, liver cancer, and breast cancer [24, 25].

Fig. 1
figure 1

Pharmacokinetic and hydrodynamic factors involved in IA drug delivery. Dedrick’s mathematical model of intracarotid drug delivery is visually demonstrated (left panel). The model shows that IA infusions are useful with reduced Q and rapid regional extraction. However, when IA drugs are injected during transient cerebral hypoperfusion (TCH) the arterial blood concentrations increase approximately 10-fold, the transit time is increased 10 to 50-fold, and contact with blood cells and serum proteins is avoided—as evidenced by the absence of blood in vessels during video-microscopy of a rabbit brain during IA-TCH injection (grayscale center panel). The Color center panel shows the effect of IA-TCH on IA delivery of cationic liposomes to the brain. These post-mortem, multi-spectral images (MSI) show that cationic liposomal uptake is significantly improved by utilizing intra-arterial delivery with TCH. Corresponding concentration–time curves obtained by diffuse reflectance spectroscopy (DRS, 612 data points each) are shown in the right panel. Optical measurements using DRS enable one to track drug and tracer delivery while imaging methods such as MSI can map drug/tracer distributions. Such novel optical technologies have rapidly advanced our understanding of the pharmacokinetics of IA drugs. C concentration, V volume, CL clearance, Q regional blood flow

Full size image

Emerging real time assessment of tissue drug concentrations

As technology has advanced in recent years, optical methods have emerged that can track tissue tracer concentrations of drugs on a sub-second time scale (Fig. 1 ). Experimental data tracking of tissue concentrations and blood flow changes are able to generate more accurate pharmacokinetic models that include hydrodynamic factors that are critical to effective and safe IA drug delivery [8, 26, 27].

Systemic rescue

Another advantage of IA drug delivery is the possibility of removing recirculating drugs in order to decrease the systemic side effects. Extracorporeal hemoperfusion has been used in the past for this purpose [5, 28]. Alternately, an antidote that can neutralize recirculating drugs may be effective. For example, ototoxicity is a potential complication of IA carboplatin therapy that can be mitigated with concurrent thiosulfate infusion [29].

The case against IA drugs for GBM treatment

Track record of partial successes

Over 2000 patients have been treated with IA chemotherapy, mostly in Phase I and II trials. Unfortunately, nitrosoureas and cisplatin are the only IA drugs that have been tested in Phase III trials and their failure to significantly improve survival has stymied further clinical research. However, proper case selection, safer drugs, and better drug delivery protocols could improve outcomes [29].

Biological hurdles to IA chemotherapy for GBM

Examination of GBM tissue reveals that certain regions of the tumor are hypo-perfused while others are highly vascular. This vascular heterogeneity within the tumor microenvironment is a major hurdle to effective IA drug delivery. Ongoing research within this arena promises to improve drug delivery protocols and drug selection [30].

Lack of reliable pharmacokinetic models

Many models of IA delivery are overly simplified and ignore background blood flow (Fig. 1) [6]. The relevance of these simplistic models is further challenged by the evolution of nanotechnologies where particles are larger and hydrodynamic forces are considerably greater (Fig. 2 ). Hydrodynamic models of drug delivery are being developed and they have been tested in computer simulations and in vitro [31].

Fig. 2
figure 2

Nanoparticle-endothelial interaction. The retention of nanoparticles and drugs by vascular endothelium is determined by two forces: (1) the sheer stress due to flow and (2) additional forces affecting the probability of adhesion (Pa). Simulation studies show that discoid particle uptake is relatively size-independent. Nano-spheres and nano-rods have a biphasic effect; initially, increasing size increases the Pa but simultaneous hydrodynamic stress counteracts this. Nano-spheres at higher sheer stress demonstrate an axial flow pattern that decreases the Pa while nano-rods that display a tumbling movement have a higher Pa. Such hydrodynamic complexities have a profound effect on IA drug delivery. See [31] and [33] for additional details

Full size image

Inconsistent goals of targeted drug delivery

The primary goal of IA drug delivery is to selectively eradicate neoplastic tissue and, ultimately, to extend life. However, this ideal goal is often difficult to achieve. Without a clear improvement in survival, IA chemotherapy is often evaluated by surrogate end points such as a reduction of tumor size, suppression of tumor metabolism, or safe delivery of drugs. Such differences in therapeutic endpoints make it difficult to compare the results of clinical trials.

Effect of streaming

Streaming results in the mal-distribution of drugs when the infusion rate is less than 20 % of the background blood flow. It has been invoked to explain both treatment failures and neurotoxicity. Streaming can be decreased by injecting drugs as pulses, at rates exceeding 20 % of the background blood flow rate, injecting drugs during diastole, and by injecting through catheters with side port rather than tip exits. Nonetheless, the role of streaming is controversial and has been disputed [32].

Safety concerns

IA drug delivery carries a number of risks, some of which are unique to this delivery strategy. These may include: (1) complications related to catheter placement and vascular access, (2) local reactions to IA drugs such as, ocular erythema with cis-platin, (3) neurological complications such as seizures or focal deficits, and (4) systemic effects of chemotherapy (infections, marrow suppression etc.).

Methods to investigate the kinetics of IA drugs

Conventional approaches

The conventional methods of determining drug uptake such as brain to plasma partition ratio, in situ perfusion, brain uptake index, and autoradiography usually provide snap shot measurements of tissue uptake. Alternately, microdialysis, PET, and MRI provide real-time data. Microdialysis of lipid soluble compounds is challenging due to the difficulties encountered during extraction of said compounds. Similarly, PET and MR imaging are logistically demanding. In clinical trials, the effectiveness of IA chemotherapy is often judged by concentrations of the drug in the CSF. Incidentally, CSF drug concentration measurements are not reliable as they are both a function of drug uptake by the brain and secretion by the choroid plexus.

Novel optical approaches

In contrast to the conventional approaches for investigating IA delivery kinetics, novel optical tools offer tremendous advantages. Principally, they enable simultaneous measurement of blood flow and optical tracer and drug concentrations in a sub-second time domain [8, 27]. Optical techniques such as diffuse reflectance spectroscopy can provide rapid, site specific and tissue non-destructive tissue concentration measurements for certain drugs. These methods are cost-effective, relatively simple to execute, and do not carry the hazards of radiation and magnetic fields that conventional imaging methods often do. However, optical methods can only be applied to certain drugs. The use of tracers such as dyes and quantum dots that have a narrow light absorption spectrum and high quantum yield, promise to greatly advance pre-clinical research in this field [26].

Complex pharmacokinetics of IA drugs

Basic model of IA drug delivery

The significance of the Dedrick model lies in defining clearly when IA drug delivery really works. As described in Fig. 1, if the advantage of IA regional drug delivery (Rd) is defined as (C1/C2)IA/(C1/C2)IV then Rd can be represented as follows: Rd = 1 + CLTB/[Q × (1 − E)]. C1 and C2 are concentrations in the brain and rest of the body, respectively. CLTB is the total body clearance of the drug, Q is the regional blood flow, and E is the fraction of drug extracted in the first pass through the cerebral circulation.

This basic model reveals that IA drugs work when the regional blood flow (Q) is low, when extraction (E) is high, and when systemic clearance (CLTB) is high. This model has several limitations; it assumes uniform mixing of drug in the arterial blood, it ignores streaming and hydrodynamic factors that affect drug delivery, and it assumes no efflux of drugs [6].

Computational fluid dynamic models of regional drug delivery

New models of drug delivery use human magnetic resonance imaging data to acquire arterial dimensions and flow profiles. Such models assume Newtonian properties of blood. A parabolic wave front presents the nanoparticles to the vascular endothelium. The probability of adhesion of a nanoparticle in this model is determined by several factors as shown in Fig. 2. Such models have been tested in vitro and can predict local deposition of nanoparticles under different flow conditions [31]. Other models reveal how the shape and volume of the particles affect the probability of adhesion [33].

Recent studies on IA GBM treatment

Table 1 summarizes important clinical studies that utilized IA drug delivery for the treatment of GBM and other brain tumors in the last few years [34–43]. The earlier studies have been previously reviewed [44]. Recent studies show noticeable shifts in clinical practice including the use of super-selective drug delivery and the use of less neurotoxic drugs [29, 45]. Interestingly, many investigators continue to use mannitol in an attempt to disrupt the blood brain barrier (BBB), despite reported complications with its use [46] and known variability in the degree of BBB disruption achieved [47]. In experienced hands BBB disruption seems to be safe [48]. Yet others have avoided routine osmotic disruption while acknowledging its capacity for increasing tumor penetration by chemotherapeutic drugs [37]. Recent studies with IA bevacizumab using concurrent hyperosmotic disruption show that such techniques can be well-tolerated. On the other hand, disruption of the BBB might be more relevant to the delivery of larger therapeutic molecules than to the delivery of smaller (<400 Daltons) ones. Use of modern micro-catheter techniques to focally deliver osmotic agents to disrupt the BBB could improve the safety of the method. Controlled disruption of the blood tumor barrier with focused ultrasound is another strategy that may improve IA delivery of large molecular weight agents.

Table 1 Recent publications describing the use of IA chemotherapy for brain tumors
Full size table

Also notable is a recent report by Riina et al. that describes the use of balloon occlusion methods to localize bevacizumab to posterior fossa tumors [49]. Such an approach utilizing flow reduction in conjunction with mechanical localization resonates well with pre-clinical efforts to improve IA drug delivery by transient flow arrest [50]. It is also consistent with the clinical observations of Chow et al. who show that tumors with low blood flow respond better to IA chemotherapy [35]. It is clear that reduction of blood flow during IA procedures is technically feasible and that it enhances drug delivery and could augment therapeutic effects. Flow reduction appears to be a reasonable alternative to the strategy of increasing IA drug doses in proportion to arterial flow [38].

Evolutionary advances in IA chemotherapy for GBMs

Catheter technology

Catheter-related complications occur in under 2–3 % of IA procedures [51]. Transient vasospasm and neurological symptoms are often seen but the incidence seems to have declined over the years. Single or double balloon catheters have been designed to facilitate drug delivery by isolating proximal as well as distal arterial irrigations to minimize local complications [25, 49, 52].

Patient selection

As experience increases and new biomarkers emerge, more nuanced patient selection should occur. For example, MGMT gene expression may be used as an indicator for IA temozolamide treatment [53, 54]. Similarly, highly vascular GBMs might be more responsive to bevacizumab than less vascular GBMs.

Drug selection

Due to their high lipid solubility, short biological half-life, and rapid onset of action, nitrogen mustards (BCNU, ACNU, HeCNU) were intensely investigated for the treatment of brain tumors. These compounds were associated with significant neurological complications. Recent studies have therefore focused on drugs that are safer for IA delivery, such as carboplatin and melphalan. These drugs have reduced local toxicity and their systemic complications seem to be manageable [36]. No significant neurological toxicities have been reported with IA bevacizumab although it has yet proven to be safer than conventional chemotherapy [37]. While bevacizumab and temozolamide are both promising investigational agents for IA treatment of GBM, the potential risks are significant and have yet to be fully appraised in large studies.

Systemic rescue

The concurrent administration of drugs that can neutralize recirculating chemotherapeutics promises to reduce the systemic toxicity of a number of IA agents [29]. For example, sodium thiosulfate is currently being evaluated as an agent that may decrease the otological toxicity of certain drugs (Table 2).

Table 2 Ongoing clinical trials with IA drugs for GBM treatment in the National Trial registry
Full size table

New paradigms in IA chemotherapy

Relatively non-toxic biological interventions

The development of anti-angiogenic compounds such as the VEGF-A antagonist, bevacizumab, have considerably increased the safety of IA drug delivery, however, there is no evidence that this has impacted survival. As noted in Table 2, clinical trials are underway using blood brain barrier disruption with mannitol and IA bevacizumab to determine its impact on survival. Although tumor pheno- or genotype may predict drug responsiveness, there is still much to be learned. Tumor delivery of genomic drugs for GBM is proving difficult in clinical trials as well. While these drugs are highly tumor specific and generally lack systemic side effects, their use as an alternative to conventional chemotherapy is a matter of future investigation [53].

Flow arrest delivery of drugs

Reduction of blood flow radically alters the pharmacokinetics of IA drugs [20]. Guided by the principle that brain tissue can safely tolerate up to three minutes of ischemia. Flow arrest has been widely used during neurovascular surgery and during neuro-radiological treatments of high flow lesions. Two methods of achieving flow arrest are by systemic arrest (Fig. 3) and by local balloon occlusion [49, 55]. The benefits of transient flow (<1 min) arrest during IA drug delivery include: (1) better targeting of drugs to the tumor site, (2) achieving higher cerebral arterial concentrations, (3) achieving more consistent concentrations in the arterial distribution, (4) increasing transit time, (5) decreasing shear stress on drug molecules and carriers, and (6) avoiding binding with serum proteins and blood cells.

Fig. 3
figure 3

IA delivery in clinical practice. Translation of IA-TCH assisted delivery to the human brain has been achieved by both temporary cardiac arrest with adenosine and by balloon occlusion during endovascular procedures. The hemodynamic changes in EKG, mean arterial pressure (MAP), and transcranial Doppler (TCD) flow velocity (A–C) indicate that TCH is achieved by administration of adenosine during embolization of a high flow cerebral arteriovenous malformation (D–E) (see [55]). Similarly, temporary balloon occlusion within the posterior circulation (F) has been used to effect TCH and targeted IA drug delivery to the brainstem glioma seen on axial T1-weighted MRI (G–H) (see [49])

Full size image

Drug formulations optimized for IA delivery

Due to high resting brain tissue blood flow, the impact of any IA delivery strategy on GBM will be minimal unless rapid extraction during the drug’s first pass through the cerebral circulation occurs [56]. A number of strategies have been proposed to improve regional extraction after IA injections. The use of small molecule chemotherapeutics (<400 Daltons) that can diffuse across the BBB, such as Temozolamide, has been effective but treatment response and safety in larger patient samples remains unproven [41]. Increasing drug lipid solubility by adding methyl groups, replacing polar groups, or by adding halogenated alkane chains (ie. tributyl chlorambucil) also enhances brain uptake [57]. Identifying tumor specific receptors and transporters that are overexpressed in GBM allows the development of targeted immuno-conjugated drugs. Anti-transferrin receptor anti-body, OX-26, was shown to be an effective carrier for drug delivery across the BBB. Compounds such as OX26-methotrexate [58] or OX26-daunorubicin were developed but not tested for IA delivery [59]. Likewise, the use of cationic drugs and liposomes, cell penetrating peptides, cationic albumin or other cationic delivery platforms can increase tumor selective drug delivery due to the over-expression of anionic charges on tumor cell membranes [60].

Rationalizing the role of IA drugs for GBM treatment

Given the biological variability and complexity of GBMs, the current “one-size-fits-all” approach is unlikely to be an effective treatment strategy. Transient improvement as evidenced by GBM regression can be readily achieved with current IA treatments. In order to affect the duration of survival, IA drug selection needs to be based on the genotypic and phenotypic characteristics of the tumor. However, given the lack of a single target, even a genomic approach will have to be flexible, requiring several drugs and delivery methods to achieve sustained remission. As the search for greater tumor selective drugs and carriers continues, the unique characteristics of IA delivery will enable it to play a significant role in future GBM treatments.

Conclusions

The history of IA drug delivery to the brain when viewed over several decades reveals several missteps. The complexity of IA delivery with regards to hydrodynamics was not universally appreciated both in preclinical research and during clinical trials. The lessons from these past failures should guide us to new paradigms in IA drug delivery. Given the complexity of GBM, its treatment will require an individualized approach based on patient and tumor profiling. IA drug delivery could play a significant role in delivering novel pharmaceuticals including viral vectors, gene therapy agents, and emerging genomic drugs. These IA treatments may be curative, adjunctive, or palliative as needed. Improved patient and drug selection, advanced drug delivery protocols, and fast, high resolution imaging technology will enable IA therapies to play a significant role in GBM management in the future.