Skip to Content
U.S. National Institutes of Health
Last Updated: 05/17/13

NCI Radiation Research Program Meeting Report — Clinical Research in Neutron Capture Therapy*

Running Title: Neutron Capture Therapy

by The National Cancer Institute Radiation Research Program
Radiation Oncology Sciences Program
Radiation Research Program
Division of Cancer Treatment and Diagnosis
National Cancer Institute
EPN-6000
Bethesda, MD 20892-7440
United States of America
(240) 276-5690 voice
(240) 276-5827 fax
corresponding author: rc148m@nih.gov (Richard L. Cumberlin)

Article Outline

*International Journal of Radiation Oncology · Biology · Physics

A one and one-half day workshop to assess the current state of the science in neutron capture therapy (NCT) was convened at the request of the Radiation Research Program, Division of Cancer Treatment and Diagnosis, NCI, and the U.S. Department of Energy. The topics were primarily clinical with physics, chemistry, and biology relevant to immediate trials discussed. The morning of the first day was directed to updates on epithermal neutron sources, chemistry of medicinal boron compounds, and preclinical studies. In the afternoon, participants from Europe, Asia, and North America were invited to present their clinical experience with NCT. The participants then separated into breakout sessions. The following morning session included presentation and discussion each breakout session. These are presented below. Appendix I includes the Workshop participants.

Background

Neutron capture therapy was first proposed by in 1936, just four years after the neutron itself was discovered. NCT is a unique form of radiation therapy that carries a potential for a significant improvement in therapeutic gain. The classification of neutrons into energy categories is somewhat arbitrary, but for the present purpose they mean: "thermal neutrons" have an energy less than 1 ev, "epithermal neutrons" have an energy between 1 eV and 10 KeV, and "fast neutrons" may be considered, for therapy purposes, to be in the megavoltage range. NCT is a form of binary therapy, similar to photodynamic therapy, in which neither the thermal neutrons nor the boron carrier molecule has significant cytotoxic effect but produces highly radiobiologically effective particles when the two interact. This is in contrast to the more familiar "fast neutrons" which are highly radiobiologically effective by themselves. These reactions between a neutron carrier molecule such as boron-10 (B-10) and thermal neutrons produce He-4 and Li-7 ions of very high LET but very short (less than or equal to 10 microns) range. This therapy was first studied in glioblastoma in 1951 using crude thermal neutron beams and B-10 enriched boric acid. Since then, NCT has been studied in several countries and usually with glioblastoma. The difficulties with NCT were, and continue to be, difficulty in finding appropriate neutron beams, a scarcity of suitable boron carrier agents, uncertain dosimetry, and lack of rigorous and reproducible clinical trials. This workshop was convened to address these issues. These issues were divided into three broad topics: carrier agent development, preclinical studies, and clinical studies.

Boron Agent Development

Design Requirements

Boron and carbon are the only two elements that are capable of generating effective families of chemical compounds by bonding with themselves or each other. Organoborane chemistry, however, is subject to unique rules governing chemical structure and reactivity. These rules must be applied to agent design and synthesis. Consequently, the design and synthesis of boron agents for NCT differs significantly from the very active field of drug discovery based upon classical organic chemistry. Operational differences between boron agent design and conventional pharmaceutical discovery include the absence of sufficient biological and structural data with which to approach the computational design of boron agents. In addition, boron agent design, synthesis and evaluation, by it's very nature, cannot be adapted to the combinatorial synthesis methods which drive many of today's pharmaceutical discovery programs. This is not to say that computational design methods cannot be employed in the future as more stereoelectronic parameters are defined for boron-rich structural groups.

The most important, but often overlooked, difference between boron agents and conventional pharmaceuticals is the relative therapeutic concentrations of the two types of compounds and the resulting limitations that this places upon the use of molecular targeting with boron agents. Therapeutic doses of boron-10 in tumor cells generally require 3-30 µg/gm of boron in tumor tissue or 109 atoms per cell. Traditional pharmaceuticals normally function at concentrations orders of magnitude lower than this. Put another way, traditional pharmaceuticals are typically administered in the milligram dose range, whereas most current boron agents are administered in the gram range, i.e., about a 1000-fold more of the boron agent is required to reach a therapeutic concentration.

Boron agents and their attendant delivery features may be classified on the bases of target recognition mechanisms and agent size. Thus, four targeting categories are apparent: (1) global agents with no targeting (2) agents that recognize malignant cells (3) agents that bind to nuclei of cycling malignant cells, and (4) agents that both recognize malignant cells and which bind to their nucleus.

Global, or category 1, agents have no targeting features, low toxicity, low persistence and if present in very high systemic concentrations (100-300 µg/gm boron) must be maintained by constant infusion. Global agents are small, hydrophilic and incapable of crossing protective CNS membranes.

Category 2 agents generally target receptors present in the cell's plasma membrane followed by translocation into the cell or by other processes which provide internalization such as passive diffusion, transport proteins, metabolic processes and tumor specific membrane charges. Low molecular weight agents of this type are represented by the boronated porphyrin derivatives and boron-containing amino acids. High molecular weight agents in this category include boron rich conjugated immunoproteins. Category 2 nanoparticulate agents are represented by boron containing unilamellar liposomes and carborane-loaded low-density lipoproteins (LDLs).

Agents found in category 3 are useful only if they accumulate in cells and have ample access to cell nuclei. Such agents are represented by low molecular weight boron-containing nucleosides and nucleotides. An example of cell incorporation for these agents is provided by the hyperactivity of malignant cells and the rapid utilization of the nucleic acid precursors in DNA synthesis.

Agents of category 4 are highly valued for their dual ability to select tumor cells and localize to the tumor cell nucleus. Specific cell-targeting may be provided to low molecular weight agents which enter the cell nucleus and bind to DNA. This can be achieved, in principle, by encapsulation of the agent in a tumor-targeting unilamellar liposome (ULL), or by attaching a cell-recognition ligand, such as folate, to the agent. Thus those Category 3 agents which have no organelle specific recognition properties may be upgraded to category 4 by the addition of a specific delivery system. Low molecular weight agents which fall into this group are represented by acridines, phenanthridines, polyamines and oligomeric phosphate diesters (OPDs). Evidence suggests that OPDs may have the ability to selectively target tumor cells and translocate to the cell nucleus in therapeutic quantities without additional cell-targeting modifications.

Of growing importance in boron agent design is the recognition of nanoparticles (1-100 nrn diameter entities), such as unilamellar liposomes (ULLs), unimolecular micelles (OPDs) and low-density lipoproteins (LDLs) as boron-rich delivery vehicles directed to cellular receptor sites which are hyperactive and over-expressed by malignant cells. Immaturity of the vasculature and the resulting looseness of endothelial cell joining characteristic of rapidly expanding tumor structures allows nanoparticulates to leak into the intercellular volume through capillaries unique to the tumor site. The use of small ULLs (<100 nm diameter) for the selective delivery of boron agents to the interior of malignant cells has been extensively developed and long-circulating so-called "stealth" liposomes and liposomes carrying external immunoprotein attachments would be possible variants. Boron containing nucleosides and nucleotides may also be encapsulated in ULLs and preferentially delivered to tumor cells. Other species which may be useful for ULL encapsulation are boron-rich DNA intercalators (acridine dyes and phenanthridium cations), groove binders (bibenzimidazoles, netropsin and distamycin), poly amines (spermidine and spermine) and di- and oligonucleotides capable of antisense hybridization with DNA.

Small apoprotein-mediated LDL nanoparticles (<20 nm in diameter) become possible boron agent carriers following their isolation from the patient and replacement of their cholesterol package with a hydrophobic carborane derivative. The resulting very boron-rich LDLs are, in principle, targeted to their original receptor sites by the presence of their apoprotein marker. Thus, reinjection of the patient with viable boron-laden LDLs should selectively deliver the agent to tumor cells which over express LDL receptors.

Very small unimolecular micelles (<10 nm in diameter) may offer an additional approach to nanoparticle delivery of boron to malignant cells and at the same time increase solid tumor penetration due to their small size. While little is known with regard to boron agent delivery by micelles of any type, very small amphiphilic species represented by polyanionic oligmoeric phosphate diesters (20–30% boron) appear to be effective. Very low injected doses of OPDs in tumor bearing mice gave time-course biodistributions which closely resembled those of ULLs many times their size. Other experiments demonstrated translocation of OPDs from extracellular buffer to the cell nucleus.

Recent modeling of lethality as a function of subcellular boron distribution indicates that absolute boron concentrations of 20–30 µg/gm confined to the nucleus and/or the cytoplasm will assure cell death under normal irradiation conditions. Since the nucleus of a cell may constitute only 10–20% of its total mass, an analytical boron content, obtained from tumor tissue analysis, of only 3 µg/gm would represent an absolute boron content of up to 30 µg/gm (3 µg/gm x cell mass/nuclear mass) if boron was largely confined to the nucleus. A lethal absolute boron content of 30 µg/gm in cytoplasm would correspond to an analytical tissue analysis of about 25 µg/gm boron. Thus, in the absence of subcellular biodistribution data many agents investigated in the past which performed poorly (say 5 µg/gm) on the basis of the accepted 30 µg/gm tumor boron criterion, or were limited to a low boron value by systemic toxicity issues, were discarded. If the analytical boron content of these cells was actually localized in the nucleus or other critical organelle, the performance of the agent could be quite useful if it were judged on the basis of its likely performance in actual neutron irradiation experiments with the boron-10 enriched agent. Consequently, a low analytical boron content in gross tissue analyses cannot necessarily be used to eliminate otherwise promising agents from further consideration. A test based upon the radiobiological response of the new boron agents might be more appropriate. Specifically, if the radiobiological studies show that cell killing is much greater than would be expected from the analytical boron content, this would provide a very good indicator of intracellular localization whether accomplished in vivo or in vitro. Knowledge of subcellular boron distribution is essential to understanding the relationship of agent performance and should become a major research focus.

Agent Production Requirements

The production of boron target agents spans a range of requirements which include the type of agent, quantity, purity and the extent of boron-10 enrichment. All agents which enter clinical trials must be boron-10 enriched (>95%) and manufactured in accord with FDA-approved GMP procedures. The quantity of an agent required at any time is dependent upon where the agent is located in its development cycle. Initial evaluation only requires gram quantities, depending upon the agent's boron content. Advanced evaluation in large animals could require less than a kilogram while clinical trials would require several kilograms of boron-10 enriched GMP product.

Two important features of agent synthesis must be emphasized. First, regardless of the detailed synthesis sequence employed, agent syntheses rest upon the availability of boron-10 enriched inorganic precursors which must eventually be available in large quantities of GMP purity. All of these precursors are ultimately derived from boron-10 enriched boric acid which is commercially available. Small commercial production facilities exist for the synthesis of sodium borocaptate (BSH), boronophrnylalanine (BPA) and GB10, which are agents involved, or soon to be involved, in clinical trials. The second important feature regarding agent synthesis is the fact that the chemicals, chemical reactions, compound characterization, criteria of purity and purification procedures are unique to polyhedral organoborane chemistry. While in many respects this chemistry resembles and very often makes use of organic synthesis procedures, the reactions involved in organoborane syntheses often have no counterpart in organic chemistry and additional safety difficulties (fire and chemical toxicity) may arise without the proper training of personnel. These are surmountable problems, but they speak to the fact that dedicated facilities and specially trained personnel are required to assure success.

The NIH Rapid Access to Intervention Development (RAID) program is relatively new and effective in the proper context. Among the RAID program functions is the synthesis of pharmaceuticals or their precursors which allow an academic synthesis and evaluation program to go forward rather than falter for lack of support at a critical moment. This program would be ideal to assist in the synthesis or evaluation of a new agent. However, the boron-10 enriched precursors, manufactured under GMP, are not commercially available and any synthesis work undertaken under RAID auspices would require beginning with boron-10 enriched boric acid and proceeding under GMP to the product. As pointed out above, this is not organic chemistry and the work involved is unique in the pharmaceutical field. The same situation arises in the possible evaluation of agents using RAID expertise since the performance milestones for boron target agents are quite different from those of traditional pharmaceuticals. The boron agent development community could make use of RAID if the necessity arose and cost effectiveness were evident.

Each research group pursuing the validation of new agents cannot independently devote resources to the synthesis of boron-10 enriched decaborane from enriched boric acid solely to fulfill its own needs. Since this is a relatively efficient task in the hands of experienced personnel, it follows that all synthesis-oriented research groups (including appropriate RAID projects) may benefit from the establishment of a medium scale GMP synthesis facility. A facility of this sort could provide moderately priced agents and agent precursors by building a stockpile of GMP boron-10 enriched reagents for general use. At present, no such facility exists.

Preclinical Studies

Animal Tumor Models

Animal tumor models provide the means for the evaluation of boron compound biodistribution. Some NCT experiments can provide information about the degree of tumor cell kill, which may be an indirect measure of the boron distribution at the cellular level. Analytical techniques do exist that can detect boron in individual cells, for instance secondary-ion mass spectroscopy (SIMS). It is critical that animal experiments are carried out with tumor models that mimic the relevant properties of human tumors, and be coupled with analytical techniques of sufficient sensitivity to determine biodistribution.

The particular boron compound may influence the choice of an animal tumor if there is a biochemical basis for tumor selectivity. For example, a boronated melanin precursor analog would be screened in melanoma. Therefore because of the critical dependence of NCT effectiveness on the delivery of the compound, a bank of tumor models that differ in vascular density or in vascular supply would be very useful. Subcutaneous tumors are generally considered adequate for in vivo screening, whereas an orthotopic tumor model may be better for testing therapeutic gain under neutron irradiation conditions. The standard endpoints used in animal tumor model radiobiology such as tumor re-growth delay or cure probability are perfectly adequate for NCT investigations.

Normal Tissue Studies

In NCT radiobiology, measured biological effectiveness factors for the component of the dose are a combination of the intrinsic RBE of the high LET ions and the biodistribution of the particular boron compound. This combined effect has been termed compound biological effectiveness (CBE) factor. The calculation of the dose delivered to normal tissues in NCT requires estimates of three basic parameters: (1) the boron concentration (2) the CBE factors for that particular boron compound in the tumor and in all normal tissues within the treatment field, and (3) the RBE of the beam itself for the tumor and for the normal tissues involved (which depends on accurate beam dosimetry). Animal models play an important role in providing information on all three of these parameters.

Because the low energy thermal or epithermal neutron beams cannot be easily focused or collimated, irradiated volumes are large. Thus normal tissue response may be more of an issue with NCT than with conventional photon therapy. As with conventional photon irradiation, normal tissue response in NCT is related to the particular tumor site being addressed. Animal irradiations for normal tissue response studies must be as close as possible to the clinical situation under consideration . The normal tissues in the treatment field will depend on the tumor site, but normal tissue response studies in animals have been done, and clinical data exists for most of the tissues involved (e.g., skin, brain, the lung, oral mucosa). Much of this normal tissue response work has been carried out in rats. If NCT is ever applied to the treatment of prostate carcinoma, the critical normal tissue involved, the rectum, would require a larger animal model such as the dog.

Because NCT radiobiology requires the experimental generation of correction factors for biological effectiveness in order to account for the high-LET beam components, animal model systems used for these studies should have easily measured endpoints. For example, in the CNS, the rat spinal cord model is used extensively. The endpoint is paralysis within seven months. Paralysis is a quantal response and the data can be fitted by logit or probit analysis and the LD50 with confidence limits can be determined. Comparison of the LD50 for a photon reference irradiation, the thermal neutron beam only, and the thermal neutron beam in the presence of boron compound allows the estimating of RBEs for the effects of the thermal beam and the specific CBE factors for the particular boron compound. However, paralysis is not the endpoint expected in the human clinical trials with brain tumors. A functional assay of CNS damage would be more appropriate. MRI changes are an indication of subclinical damage. Though clearly more relevant to the clinical situation, these endpoints are more difficult to quantify. Patients in the NCT clinical trials have exhibited a somnolence syndrome. There is no animal model for a somnolence syndrome. This is a general problem for any treatment for brain tumors where the anticipated adverse effects are cognitive in nature.

In the end, NCT radiobiology must always be viewed as a model system that provides a guide to the estimate of NCT doses in the clinical situation. NCT clinical trials will rely on dose escalation to conservatively approach tolerance doses in humans. Validation of the calculated photon-equivalent doses currently being used in NCT clinical trials can come from animal models, where the effects of Gy-Eq doses delivered during boron neutron capture irradiations can be compared with the known response of the tissue to photon irradiation or from the NCT clinical response data, if there are endpoints reached in the NCT dose escalation trials that can be related to the known response to photons of the tissue in question.

Dose-Volume Considerations

The recent initiation of clinical trials of NCT for glioblastoma multiforme using epithermal neutron beams capable of penetration through closed scalp and cranium, has required careful calculation of the dose to the normal brain, including measurement of the biological effectiveness factors for the high-LET components of the normal brain dose. Much of the work on the response of the central nervous system to NCT has been done in the rat spinal cord model using a thermal neutron beam. The epithermal beams required for adequate neutron penetration in the human brain are inappropriate for use in the rat. Epithermal neutrons require several centimeters of tissue, in which little neutron capture reactions occur, to dissipate the energy necessary to become the thermal neutrons needed for the neutron capture process. Also, in the clinical situation, the normal brain dose decreases as a function of depth, making volume a potentially critical parameter; a parameter that is difficult to incorporate in studies using the rat spinal cord. To more closely approximate the clinical situation, epithermal neutron beams have been used in normal tissue response studies in the canine brain but primate models may be necessary.

The lung is another site where dose volume relationships are critical. There have been reports of Adult Respiratory Distress Syndrome after NCT treatment to the brain. Whether this is due to uncollimated scatter, head leakage, or other factors, it is known that partial irradiation of the lung can trigger a more widespread pneumonitis reaction due to the release of cytokines. If NCT is to be applied to lung tumors, or for the analysis of the effects of the scattered dose to the lung from treatment of head and neck tumors, small animal models may be sufficient for initial studies, but larger animals such as the dog or the pig will be required for definitive studies.

Analysis of the results of the NCT clinical trials will involve dose-volume histograms (DVH). In this regard NCT dosimetry is unusually complex. The DVH for each of the beam components is different. For example, the DVH for the region of brain for the fast neutron component of the dose will be different from the DVH for the region of brain for the gamma component and likewise for the neutron capture components of the dose. This is not an insurmountable problem, but must be remembered when doing these analysis and will require accurate beam characterization and dosimetry.

Evaluation of New Agents

Thorough evaluation of new compounds must provide data on cytotoxicity, in vitro uptake of the agent by model cell lines, tumor and blood analysis for boron concentration and, if possible, time-course biodistribution. The increasing recognition of subcellular boron distribution in model cells as an important parameter in agent evaluation and molecular targeting has led to the development of new analytical procedures for this purpose. Microdistribution of boron has been studied using a-track autoradiography, electron spectroscopic imaging (ESI), electron energy loss spectroscopy (EELS) and secondary ion mass spectroscopy (SIMS). These are not routine measurements and they pertain to very few agent and cell types. Analytical boron concentrations that may not appear to be high enough for efficacy may be misleading if the observed boron were actually localized in the nucleus or other critical organelle.

Care must be exercised in the use of in vitro agent uptake by model cell lines since the most common method employed to monitor such experiments has employed the quantification of cellular death upon irradiation of boron-loaded cells with a thermal neutron source. False positive results can and have been obtained in the past due to the presence of non-specifically bound agents at the plasma membrane surface. This can normally be prevented by thorough washing of the cell suspension before irradiation. Other results obtained from in vitro experiments which employ nanoparticle agents may be unreliable because nanoparticle uptake often requires long periods to reach maximum values in vivo and the properties of the nanoparticles may be significantly different in serum and buffer suspensions. However, the in vivo evaluation of small molecules and nanoparticles using small animal models overcomes the uncertainties mentioned above and provides a clear view of competitive uptake by the reticuloendothelial system, blood clearance, tumor accretion and systemic toxicity under conditions resembling those encountered in humans. Thus, nanoparticle agents require evaluation in time-course experiments carried out over an extended time. The minimal tissue requirements for boron analysis in all time-course experiments are blood, liver, spleen, tumor and relevant normal structures.

Low molecular weight agents would advance to small animal evaluation, as described. If agents of any type displayed attractive properties through small animal evaluation, the next steps would involve biodistribution studies, systemic toxicity evaluation and dose escalation using large animals such as dogs or monkeys. Following the collection of these data, the large animals would then be subjected to neutron irradiation using boron-10 enriched agents. Sufficient large animal data would then be collected to move the new agent into the clinical trial protocols.

It must be emphasized that there can be no strictly uniform evaluation protocol for agents of all types created to serve all varieties of tumor. Evaluation protocols must remain sufficiently flexible to evaluate unusual agents. The tumor types of interest to the clinical community must be defined to guide agent design and establish agent evaluation standards.

One or more centralized facilities for biological evaluation may be helpful. Such a facility should be equipped with the proper laboratory equipment and instrumentation, a vivarium (or equivalent) and both thermal and epithermal neutron sources properly equipped for irradiation experiments. Such a facility should be staffed with personnel having extensive experience in the biological aspects of NCT. The centralization of biological evaluation work would provide validation of the models, and uniformity of biological standards, the capability to perform nonroutine evaluations, provide a selection of well-maintained cell lines as well as neutron sources and the means to accumulate a meaningful data base.

The no-cost ICP-AES boron analyses of tissue and other research samples now provided by the Idaho Nuclear Engineering and Environmental Laboratory (INEEL) under DOE auspices should be continued and offered to the scientific community as a service. Data obtained by this means has been invaluable for the determination of boron concentrations in nonbiological as well as biological research samples not directly related to biodistribution studies. This will not be sufficient for a thorough evaluation which will also require support of ESI, EELS, and SIMS facilities to determine boron microdistribution. Many compounds that are taken up in the cell, but are sequestered away from the nucleus or other critical organelle could be quickly eliminated from consideration if this were part of an initial evaluation process.

Clinical Studies

Neutron capture therapy was first investigated in patients with malignant glioma from 1951 to 1962 at Brookhaven National Laboratories and at MIT. These trials used non selective boric acid derivatives and inadequate thermal neutron sources. The Japanese began clinical trials in 1968 using the somewhat more tumor-selective compound BSH. The preponderance of experimental and clinical work on NCT has been directed toward glioblastoma multiforme, a tumor poorly responsive to any form of treatment. Phase I trials in this disease have been initiated at Harvard-Massachusetts Institute of Technology (Harvard-MIT), Brookhaven National Laboratory (BNL), The Netherlands (European Consortium) and most recently Helsinki, Finland.

It is not necessary, nor even desirable, to confine NCT trials to glioblastoma multiforme. Many tumor types have shown sufficient uptake in order for NCT to be considered a possibility; among these are non-small cell lung carcinoma, metastatic colon adenocarcinoma, malignant melanoma, and anaplastic thyroid carcinoma. Before clinical trials can be initiated, many factors in addition to tumor uptake of a boron-containing compound need to be taken into consideration. It will be necessary to know the uptake of adjacent and potentially dose-limiting normal tissue and the detailed pharmacokinetics of boron in the tumor and irradiated normal tissue. Toxicity needs to be assessed well outside of the treatment field because of the poor colomnation of thermal neutron beams. Malignant melanoma is an especially attractive disease for a proof-of-principal phase II study in that BPA was developed specifically for melanoma and pharmacokinetic studies suggest a tumor to blood boron concentration of 3-4 to 1, translating in principle into a therapeutic gain of 1.4.

The availability of adequate sources of epithermal neutrons is an impediment to clinical research with NCT. Right now, there is only one available in the United States (MIT) and four others in Europe and Japan. There are presently two trials with glioblastoma in the U.S. and two others in Europe and Japan. One trial with melanoma and one with lung cancer are also underway in the United States. The neutron sources used in these trials are derived from nuclear reactors, each with its unique beam characteristics. The epithermal neutron is not monoenergetic, and it also contains high energy gamma radiation and fast neutrons, both of which damage tumor and normal tissue unrelated to the presence of boron. This makes it difficult to compare clinical results from different centers. If collaborative clinical trials are to be done, it will be necessary to develop a consistent procedure for dosimetry calculations and reporting, which may require that each beam component be analyzed separately. A uniform method of beam characterization with standardized phantoms will also be necessary. Such a phantom has yet to be developed.

Effective BNCT need not be based upon the use of a single agent. Multiagent therapy may overcome heterogeneity of tumor structure, cell type and dose-limiting systemic toxicity in much the way that combination chemotherapy has been shown to do. Agents effective against particular subcellular compartments might be used in combination with one another, but at lower individual doses, and in a complementary manner providing boron throughout the complete cell while minimizing any systemic problems arising from a particular agent. For example, cells in active DNA synthesis may preferentially incorporate bornated nucleosides while cells in active protein syntheses may show a preference for boronated amino acids. The timing of the administration of these "cocktail" components could be such that they simultaneously reach their maximum effectiveness at the time of neutron irradiation. The possibilities for customizing a combination of agents for a particular tumor type are attractive.

NCT has traditionally been given in a single treatment, largely for logistical reasons. Conventional photon treatments and fast neutron treatments are given in several fractions. It is unlikely that a single administration of boron agent will incorporate in enough tumor cells for optimal NCT. Several administrations may be necessary to assure uptake in the remaining tumor cells. In addition, fractionation may minimize the effects of the beam contaminants that damage normal tissue independent of boron. This is a prime area for clinical research.

A related area is the use of concomitant fractionated NCT boost, given with fast neutron irradiation. Thermal neutrons are produced spontaneously in tissue from fast neutrons at depth. This has the advantage of not requiring that every tumor cell take up boron since the fast neutrons are lethal by themselves. Fast neutron beams also collimate better that epithermal beams thus irradiating less normal tissue as the beam enters and exits the patient. Pre clinical studies suggest that a dose enhancement factor of 1.3 relative to fast neutrons alone can be achieved. This area of research is in the early stages and should be pursued.

Rather than localizing the epithermal neutrons, it may be more helpful to localize the boron agent since without boron-10 in normal tissue, the neutrons would have no toxicity and the effectiveness of BNCT depends on having a sufficient concentration of boron in the tumor cells. Intra-arterial administration of boron compound is one method of achieving this, much as intra-arterial limb and hepatic perfusion is done with chemotherapeutic agents. For glioma, this administration may be enhanced using blood-brain barrier disruption with mannitol or similar agent. This has been demonstrated in laboratory animals but has not yet been evaluated in the clinic. Another promising method of boron delivery is to conjugate a suitable boron compound with a ligand for a cell membrane receptor that is preferentially overexpressed in malignant cells. This is similar to the use of conjugating radioactive iodine to monoclonal antibodies for specific surface receptors but would not have the systemic toxicity seen with radiosotopes. This has been done in vivo with epidermal growth factor (EGF) targeting the epidermal growth factor receptor (EGFR) which is known to be overexpressed in several malignancies. This area of research is also in the very early stages and should be pursued. Perhaps some combination of neutron localization and boron localization may show substantially less toxicity and greater efficacy than either localization method alone.

As mentioned, there are a limited number of neutron sources worldwide modified for clinical use. More than a dozen research reactors are located at universities and national laboratories near major medical centers in the U.S. and are suitable for conversion to clinical use provided there is sufficient justification for doing so. To this end, a definitive proof-of-principle clinical trial is essential. In addition, research groups at MIT, Lawrence Berkeley National Laboratory, Brookhaven National Laboratory, and elsewhere are designing linear accelerators with sufficient neutron intensity for NCT. Now, its greatest promise is in locoregional disease that other therapies cannot effectively treat. This may change considerably if agents can be developed that give a tumor to normal tissue concentration of 10:1 or better. This may be possible with further advances in molecular targeting technology. If NCT then proves to be successful and comes to enjoy an applicability that exceeds the capacity of a handful of reactor-based clinical units or fast neutron units, then hospital-based accelerators designed specifically for NCT could be produced and put into routine clinical use much like the accelerators in widespread use for photon radiation therapy.

For the present, however, the difference between encouraging well designed clinical trials with appropriate endpoints and advocating NCT as a proven therapeutic alternative must be made clear.

Appendix I. Workshop Participants

Rolf F. Barth, M.D., Ohio State University
James E. Boggan, M.D., UC Davis
Paul Busse, M.D, Ph.D., Beth Israel Deconess Medical Center
Jeffrey A. Coderre, Ph.D., Brookhaven National Laboratory
Aidnag Z. Diaz, M.D. , Brookhaven National Laboratory
Reinhart A. Gahbauer, M.D., Ohio State University
Patrick R. Gavin, Ph.D., Washington State University
M. Frederick Hawthorne, Ph.D., UCLA
Roger Henriksson, M.D., Umea University (Sweeden)
John W. Hopewell, Ph.D., Oxford
George Kabalka, Ph.D., University of Tennessee
Stephen B. Kahl, Ph.D., UCSF
Merja Kallio, M.D., Ph.D., Helsinki University
Irving D. Kaplan, M.D., Beth Israel Deconess Medical Center
Jody Kaplan, RN, BSN, Beth Israel Deconess Med Ctr.
Peter T. Kirchner, M.D., U.S. Department of Energy
George E. Laramore, M.D., Ph.D., Univ of Washington
Yoshinbu Nakagawa, M.D., National Kagawa Children's Hospital (Japan)
Matthew R. Palmer, Ph.D., Beth Israel Deconess Medical Center
Wolfgang Sauerwein, M.D., Ph.D., Universitaetsklinik und Strahlenklinik (Germany)
Raymond F. Schinazi, Ph.D., Emory University
Albert H. Soloway, Ph.D., Ohio State University
Werner Tjarks, Ph.D., Ohio State University
Nora Volkow, M.D., Brookhaven National Laboratory
Charles A. Wemple, Ph.D., INEEL
Richard J. Wiersema, Ph.D., Neutron Therapies, LLC.
Robert G. Zamenhoff, Ph.D., Harvard-MIT