Clinical Radiobiology of Fast Neutron Therapy: What Was Learnt?

Abstract

Neutron therapy was developed from neutron radiobiology experiments, and had identified a higher cell kill per unit dose and an accompanying reduction in oxygen dependency. But experts such as Hal Gray were sceptical about clinical applications, for good reasons. Gray knew that the increase in relative biological effectiveness (RBE) with dose fall-off could produce marked clinical limitations. After many years of research, this treatment did not produce the expected gains in tumour control relative to normal tissue toxicity, as predicted by Gray. More detailed reasons for this are discussed in this paper. Neutrons do not have Bragg peaks and so did not selectively spare many tissues from radiation exposure; the constant neutron RBE tumour prescription values did not represent the probable higher RBE values in late-reacting tissues with low α/β values; the inevitable increase in RBE as dose falls along a beam would also contribute to greater toxicity than in a similar megavoltage photon beam. Some tissues such as the central nervous system white matter had the highest RBEs partly because of the higher percentage hydrogen content in lipid-containing molecules. All the above factors contributed to disappointing clinical results found in a series of randomised controlled studies at many treatment centres, although at the time they were performed, neutron therapy was in a catch-up phase with photon-based treatments. Their findings are summarised along with their technical aspects and fractionation choices. Better understanding of fast neutron experiments and therapy has been gained through relatively simple mathematical models-using the biological effective dose concept and incorporating the RBEmax and RBEmin parameters (the limits of RBE at low and high dose, respectively-as shown in the Appendix). The RBE itself can then vary between these limits according to the dose per fraction used. These approaches provide useful insights into the problems that can occur in proton and ion beam therapy and how they may be optimised. This is because neutron ionisations in living tissues are mainly caused by recoil protons of energy proportional to the neutron energy: these are close to the proton energies that occur close to the Bragg peak region. To some extent, neutron RBE studies contain the highest RBE ranges found within proton and ion beams near Bragg peaks. In retrospect, neutrons were a useful radiobiological tool that has continued to inform the scientific and clinical community about the essential radiobiological principles of all forms of high linear energy transfer therapy. Neutron radiobiology and its implications should be taught on training courses and studied closely by clinicians, physicists, and biologists engaged in particle beam therapies.

Keywords: RBE; hadron; high LET; neutron therapy; radiobiology; radiotherapy.

Figure 1

Approximate time frame for neutron research and therapy indicated by blue with x-ray/photon developments in yellow.

Figure 2

Plot of percentage hydrogen (by weight) and KERMA for a range of materials and tissues exposed to five different fast neutron beams. Standard errors are not shown for individual points since they are small (average 0.78, the largest being 2.1 for bone). Data obtained from Awschalom, Rosenberg & Mravca (10).

Figure 3

Schematic diagram of external neutron beams passing through a tumour (yellow) with progressive increase in RBE and dose reduction. Prescription of radiation used RBE of 3 at the tumour depth and assumes this to be operative at all other points within a patient, but the RBE is increasing where dose fall-off occurs beyond the tumour target as illustrated.

Figure 4

A 3-D plot based on the Batterman data set (35), assuming a 90% cell loss factor to convert the volume doubling time to an estimated potential doubling time (T_pot), which is itself inversely related to the α/β ratio by a function α/β = 48.9/T_pot. The RBE was then estimated using the linear-quadratic formulations given in the Appendix.

Figure 5

(A,B) Graphical displays of RBE_max (A) and RBE_min (B) plotted with respect to the reference low LET α/β ratio, as adapted from Jones et al. (38). Least squares fitting are (A) RBEmax = 2.43 + 4.97/(α/β)_L (no standard error weighting) and RBEmax = 2.29 + 4.81/(α/β)_L (with standard error weighting), (B) RBEmin = 0.76 + 0.22√(α/β)_L (no standard error weighting), and RBEmin = 0.73 + 0.19√(α/β)_L (with standard error weighting). Modified from Jones et al. (38).

Figure 6

(A,B) The relationships between (A) fast neutron dose per fraction and RBE for different α/β ratios, and (B) the transformation of the above to provide a near-flat response for α/β of 10 and 12 Gy to simulate proton data by modification of RBEmax and RBEmin. The red curves suggest how brain and spinal cord tissues may behave (α/β = 2 Gy), followed by a gradual change in α/β to faster-growing systems such as many rapidly growing tumour types and acute-reacting normal tissues (α/β = 10 Gy). Modified from Jones et al. (38).

Figure 7

Plots of α radiosensitivity with standard errors for 5-MeV x-rays and 64-MeV neutrons. The fitted curve follow the relationship α(neutron) = 5.37/3.68(1–e^{3.68α(−x−ray)}). The hatched line represents a linear fit, which would inappropriately lead to much higher α_H values.

Figure 8

Plots of the β parameter for megavoltage photons and neutrons. Outlying cell lines with β neutrons close to zero (labelled A and B) were excluded because of probable experimental error, and the extremely high value (C) was due to known radiation repair deficiency and was also excluded. The black line will be followed if there is no change in β with increasing LET, but most nonexcluded data points are above that line. The V-79 data point (labelled V) falls just above it. Application of the sign test indicates that the hypothesis that β is the same for neutrons and megavoltage x-rays can be rejected (p < 0.01). Adapted from Jones (42), with no error bars for ease of viewing.

Figure 9

The representative data in Figure 8, with error bars and fitted with a linear function and by a saturation curve described by β (neutron) = 2.29/23.57 (1–e^{−23.57β(x−ray)}). The later fit is preferable because the linear equation will continue to increase beyond acceptable limits.