RBE of the MIT clinical epithermal neutron beamDocument Transcript
RADIATION RESEARCH 164, 805–809 (2005)
2005 by Radiation Research Society.
All rights of reproduction in any form reserved.
RBE of the MIT Epithermal Neutron Beam for Crypt Cell
Regeneration in Mice
J. Gueulette,a,1 P. J. Binns,b B. M. De Coster,a X-Q. Lu,c S. A. Robertsd and K. J. Rileyb
Universite catholique de Louvain, Radiobiologie et Radioprotection (RBNT-5469), Brussels, Belgium; b Nuclear Reactor Laboratory, Massachusetts
Institute of Technology, Cambridge, Massachusetts 02139; c Beth Israel Deaconess Medical Center, Harvard Medical School, Boston, Massachusetts
02215; and d Biostatistics Group, Division of Epidemiology and Health Sciences, University of Manchester, Manchester, M13 9PT, United Kingdom
an element possessing a large cross section for thermal neu-
Gueulette, J., Binns, P. J., De Coster, B. M., Lu, X-Q., Rob- tron absorption—in conjunction with beams of low- or in-
erts, S. A. and Riley, K. J. RBE of the MIT Epithermal Neu- termediate-energy neutrons (1). The boron is nonradioac-
tron Beam for Crypt Cell Regeneration in Mice. Radiat. Res. tive and is chemically bound to a delivery agent that is
164, 805–809 (2005). nontoxic and accumulates preferentially in the tumor during
The RBE of the new MIT ﬁssion converter epithermal neu- its administration prior to irradiation. A therapeutic dose is
tron capture therapy (NCT) beam has been determined using achieved only if the boron atoms in the tumor absorb suf-
intestinal crypt regeneration in mice as the reference biolog- ﬁcient thermal neutrons from the beam. Fission fragments
ical system. Female BALB/c mice were positioned separately created by the 10B(n, )7 Li reaction possess a high linear
at depths of 2.5 and 9.7 cm in a Lucite phantom where the energy transfer and have very short ranges in tissue that
measured total absorbed dose rates were 0.45 and 0.17 Gy/ deposit dose adjacent to the target boron atoms on a scale
min, respectively, and irradiated to the whole body with no approximately equivalent to the dimensions of a mamma-
boron present. The -ray (low-LET) contributions to the total
absorbed dose (low- high-LET dose components) were 77%
(2.5 cm) and 90% (9.7 cm), respectively. Control irradiations In BNCT, the absorbed dose at any point in irradiated
were performed with the same batch of animals using 6 MV tissue results from the addition of four components: the -
photons at a dose rate of 0.83 Gy/min as the reference radi- ray dose (both incident and induced), D ; the fast-neutron
ation. The data were consistent with there being a single RBE dose arising mainly from interactions with hydrogen nuclei,
for each NCT beam relative to the reference 6 MV photon DN; the thermal neutron dose, due principally to the
beam. Fitting the data according to the LQ model, the RBEs 14
N(n,p) 14C reaction, DP; and the boron capture dose, DB,
of the NCT beams were estimated as 1.50 0.04 and 1.03 which is directly proportional to the product of the boron
0.03 at depths of 2.5 and 9.7 cm, respectively. An alternative concentration and the thermal neutron ﬂux in the tissue of
parameterization of the LQ model considering the proportion interest. Each of these radiation types has a different rela-
of the high- and low-LET dose components yielded RBE val-
tive biological effectiveness (RBE), and the total effective
ues at a survival level corresponding to 20 crypts (16.7%) of
5.2 0.6 and 4.0 0.7 for the high-LET component (neu- BNCT dose is expressed as the sum of the RBE-weighted
trons) at 2.5 and 9.7 cm, respectively. The two estimates are components. Extensive preclinical research was performed
signiﬁcantly different (P 0.016). There was also some evi- to determine appropriate values for the individual RBEs of
dence to suggest that the shapes of the curves do differ some- the various dose components, and these were applied by
what for the different radiation sources. These discrepancies the clinical facilities where they were measured as well as
could be ascribed to differences in the mechanism of action, by others in the speciﬁcation of a treatment prescription (2–
to dose-rate effects, or, more likely, to differential sampling 5).
of a more complex dose–response relationship. 2005 by Radiation
While comparison of both the dosimetry and the physical
characteristics of different epithermal neutron beams has
begun (6–8), little information has been reported on the
radiobiological properties of the different beams now used
INTRODUCTION in BNCT research. Physical and radiobiological beam com-
parisons are required if experimental results derived at one
Boron neutron capture therapy (BNCT) is a binary form
institute are to be generalized and applied at other facilities.
of radiotherapy that uses a compound containing boron—
Moreover, if preclinical and clinical data from different cen-
Address for correspondence: UCL-RBNT 5469, Cliniques Universi-
ters are to be compared or combined, some degree of uni-
taire St-Luc, 54, avenue Hippocrate, 1200 Bruxelles, Belgium; e-mail: formity in specifying the biological quality of the different
firstname.lastname@example.org. beams must be established.
806 GUEULETTE ET AL.
As an extension of the previous successful studies of the
RBE of different clinical fast-neutron beams using the crypt
cell colony regeneration assay in mice (9), this assay is now
applied to the radiobiological characterization of the new
ﬁssion converter beam (FCB) at the Massachusetts Institute
of Technology (MIT) (10). To distinguish the characteristics
of this neutron capture therapy (NCT) beam from those
related to the capture agent, the assay was performed with-
out the infusion of a boron compound.
MATERIALS AND METHODS
The Biological System
Intestinal crypt regeneration in mice is an in vivo end point that is a
precursor to the intestinal syndrome and that can be scored before the
occurrence of the full functional effects (11). Early response scoring re- FIG. 1. Irradiation set-up (see the text). Irradiations were performed
duces the potential inﬂuence of environmental factors, which, together separately with the mice at a depth of either 2.5 or 9.7 cm in a solid
with the steepness of the dose–effect relationship, makes the system suit- Lucite phantom. The thickness of each mouse (cross section) is 1.5 cm.
able for RBE determinations. The system also has the advantage that the The ‘‘inner’’ pair of mice (in black on the front view) and the ‘‘outer’’
animals may be given whole-body irradiation. This allows the use of large pair (in gray) are displaced 1.8 cm and 5.3 cm from the central axis of
radiation ﬁelds whose dose distributions are considerably more easily the beam, respectively.
measured and prescribed.
Experimental Procedure beam. The 9.7-cm depth was chosen to provide an absorbed dose mixture
Conventional female BALB/c mice (11–13 weeks old) were supplied comprised predominantly of rays with only a small contribution from
by a commercial vendor (Charles River Laboratories) the week prior to thermal neutrons. Neutron ﬂux for each irradiation was monitored using
irradiation and were kept in quarantine facilities at the animal care facility the four ﬁssion counters housed in the collimator. These were positioned
of MIT. Mice were randomly assigned to beam type (control photon or in the periphery of the beam and formed part of the automated dose
epithermal neutron beam) and nominal dose levels. Mice were housed in delivery system for clinical irradiations. This monitoring system was cal-
groups of six in small plastic cages; each animal was identiﬁed by a color ibrated against dosimetry measurements performed at each of the mouse
code. The mice were killed humanely 84 1 h after the end of the locations using individual phantoms constructed from nylon (type 6) that
irradiations, and a 3-cm jejunal section was taken 1 cm below the pylorus. each had a mass of approximately 24 g, equivalent to that of the irradiated
The samples were immediately immersed in Bouin Hollande ﬁxative. mice. Thermal neutron doses were determined from measured differences
They were then embedded in parafﬁn, transversely sliced into 4- m-thick in the activation of bare and cadmium-covered gold foils while -ray and
sections, and stained according to the classical trichrome technique. This fast-neutron absorbed doses were measured with separate graphite and
study and its technical procedures were approved by the Committee on A-150 plastic-walled ionization chambers (13, 14). Total absorbed doses
Animal Care at MIT, for which animal welfare assurance number A-3125- are given for A-150 plastic with an uncertainty, expressed as one standard
01 was issued. deviation, that changed with depth, decreasing from 7% at 2.5 cm to 4%
The reported number of regenerated crypts per circumference is the at 9.7 cm. In the epithermal neutron beam, the mice received absorbed
average from two sets of three consecutive slices taken approximately 1 doses ranging from 2.6 to 12.3 Gy, for which the irradiation times were
mm on either side of the middle of the intestine segment, i.e. from six between 7 and 62 min. This dose range was determined after a pilot study
intestinal cross sections. No differences in crypt size between control performed using 10 mice at ﬁve different dose increments 1 month prior
photon and NCT beam samples were observed. Given this, and since this to the main experiment.
work was not intended to determine the detailed shape of the crypt sur- Control photon irradiations were performed in a 6 MV clinical linear
vival curves or to make any inferences regarding crypt or cellular param- accelerator. As for epithermal neutrons, the mice were irradiated during
eters, no correction factor was applied to the number of regenerated the day, with half of the batch of animals being irradiated before and the
crypts (12). other half the day after neutron irradiations. Mice were placed in the same
jig as for NCT irradiations and were irradiated to the whole body at the
depth of the peak dose (8-cm Lucite backscatter). The dose rate was 0.83
Irradiation Conditions and Dosimetry Gy/min. Twelve dose levels (four mice per dose) were used. Absorbed
Irradiations were performed in the ﬁssion converter beam (FCB) at the doses were speciﬁed in water and measured according to the ICRU pro-
MITR-II reactor (10). The converter was operated at approximately 78 tocol for irradiation (15).
kW with the reactor at 4.7 MW to provide the nominal maximum beam
intensity for this experiment. The mice were irradiated simultaneously in Statistical Analysis
groups of four, held side by side in a custom-made jig that formed part
of a solid Lucite phantom, shown schematically in Fig. 1. The phantom Assuming a single RBE, s, for each source, s, and using the conven-
was positioned on the treatment couch with its front face at the (16-cm- tional linear-quadratic form for crypt survival with m clonogens, we rep-
diameter) circular collimator aperture. The height of the couch was ad- resent the three crypt survival curves by the following formula:
justed to align the abdomen of each mouse with the central axis of the
beam. Whole-body irradiations were carried out with the mice at depths
of 2.5 and 9.7 cm measured from the midline of the abdomen to the front
S(D) 1 1 exp s
D1 s D , (1)
surface of the phantom. The shallow depth was chosen as a reference where D is the dose and and / are the parameters of the LQ model.
point since this is the approximate depth of the dose maximum in the This formula works well providing the range of killing is not too great
RBE OF THE MIT EPITHERMAL NEUTRON BEAM 807
Absorbed Dose Rates Measured in Nylon (type 6) Mouse Phantoms Placed in the
Custom-Made Jig Forming Part of a Lucite Cube Used for Irradiating Mice with the
Reactor Operating at a Power of 4.7 MW
Absorbed dose rate (Gy/min)
2.5 cm 9.7 cm
Depth/location Photons Neutrons
Inner positions 0.38 (77%)b 0.11 (23%) 0.16 (90%) 0.02 (10%)
Total: 0.49 Gy/min Total: 0.18 Gy/min
Outer positions 0.32 (77%) 0.10 (23%) 0.14 (90%) 0.02 (10%)
Total: 0.42 Gy/min Total: 0.16 Gy/min
All positions Mean total dose rate: 0.45 Gy/min Mean total dose rate: 0.17 Gy/min
The proton recoil dose from nitrogen capture of thermal neutrons in A-150 plastic has been included with the
fast-neutron dose to give a combined or total neutron dose.
Percentage of each component relative to the total measured dose.
(16). The functional formula (Eq. 1) is ﬁtted to the data for all three shown in Table 1, where the data for the inner and outer
sources simultaneously using direct maximization of the binomial quasi- pairs of mice (see Fig. 1) have been averaged. The contri-
likelihood, which explicitly allows for overdispersion between animals.
The binomial denominators were based on the measured number of crypts bution from thermal neutrons to the total absorbed dose
per circumference (120) in control animals multiplied by the number of rates measured in A-150 plastic was determined from foil
sections scored (six sections). The parameter m was reparameterized in activation. At a depth of 2.5 cm, the absorbed dose rates
terms of its logarithm to improve numerical stability. from thermal neutrons were 0.092 0.004 and 0.083
Alternatively, the model was parameterized in terms of the speciﬁc - 0.004 Gy/min for the inner and outer pairs of mice, re-
ray and neutron components of the mixed beams, with proportions
and n, and RBEs, and n. Assuming a multitarget formulation, the spectively, and at 9.7 cm depth they were 0.013 0.001
survival function then becomes and 0.011 0.001 Gy/min. The contribution of the -ray
component to the total absorbed dose increased from 77 to
S(D) 1 [1 P(D)] m
90% when moving from a depth of 2.5 to 9.7 cm.
[ D n n D 2 2
n D2 2
] (2) The dose–effect curves for intestinal crypt regeneration
after epithermal neutron or 6 MV photon irradiation are
There are three quadratic terms, ( / )n, ( / ) and ( / ) n, representing presented in Fig. 2. A total of 144 mice were irradiated.
‘‘dual-hit’’ terms for pairs of neutrons, pairs of rays, and -ray–neutron Due to the distortion of the ﬁeld proﬁles of the NCT beams,
interactions, respectively. Given the proportions of the dose contributed
by the two radiation types, and n, and deﬁning the RBE of the
the doses to the pairs of ‘‘inner’’ and ‘‘outer’’ mice (see
rays ( ) as 1, the reparameterized form was ﬁtted to the data in the same Fig. 1) differed by about 17% and 13% at depths of 2.5
way, and the apparent RBE of the neutron component ( n) was deter- and 9.7 cm, respectively. The data were ﬁtted using the LQ
mined. Since this formula has a large number of parameters, and since model (Eq. 1). Since there was no evidence that the /
there was no evidence for a non-zero interaction term, ( / ) n , this in- ratios differed for the three beams, a model with a common
teraction term was set to zero. The RBE of the -ray component, , is
deﬁned as 1. As the / ratios of the two components are a priori dif-
/ ratio was ﬁtted to the data, allowing a single RBE
ferent, the RBE will be dependent on dose, and the parameter estimate, estimate to be obtained; the parameters of this model are
n, represents the extrapolation of the RBE to zero dose (100% survival). shown in Table 2. The RBEs of the NCT beams were es-
The RBE at speciﬁc survival levels was obtained trivially from the pa- timated as 1.50 0.04 and 1.03 0.03 at depths of 2.5
rameter estimates, and the standard errors of these estimates were ob- and 9.7 cm, respectively. These estimates are little affected
tained from these and the parameter covariance matrix using a parametric
bootstrap method. We present the RBE at a survival level corresponding
if the shoulder region is excluded. However, note that the
to 20 crypts per circumference (i.e. a surviving fraction of 0.167). Com- combination of biological and dosimetry uncertainties de-
parisons between models and inferences concerning differences between creases the precision of the RBE estimates, potentially in-
parameters were performed using quasi-likelihood deviance-ratio F tests, creasing the standard errors by a factor of two.
comparing the relevant model ﬁts.
Estimates of the RBE of the MIT epithermal neutron
beam have been obtained at two depths. The determination
The relative fractions of the total neutron and -ray dose of the RBE estimates is dependent on the ﬁt to the LQ
components did not vary appreciably across the ﬁeld, as model. Visually, the ﬁts presented in Fig. 2 seem reason-
808 GUEULETTE ET AL.
Fitted Parameters of the LQ Model for the Three
Mixed Beams (Common Estimates of the
Parameters) and Estimates for the RBE ( )
Relative to the Reference 6 MV Photon Beam
Parameter Estimate (SE)
(Gy) 0.022 (0.006)
/ (Gy) 1.03 (0.27)
log (m) 1.07 (0.08)
RBE ( ), at depth of 2.5 cm 1.50 (0.04)
RBE ( ), at depth of 9.7 cm 1.03 (0.03)
eters ﬁtted to this model are shown in Table 3. At a survival
level corresponding to 20 crypts per circumference (16.7%
survival), the RBE estimates are 5.2 0.6 at a depth of
2.5 cm and 4.0 0.7 at a depth of 9.7 cm. The neutron
component of the dose at 9.7 cm is small (10% of total
dose), and the dose rate is low (0.17 Gy/min) compared to
the control photon (0.83 Gy/min) irradiation. The neutron
FIG. 2. Dose–effect curves for intestinal crypt regeneration in mice RBE estimates differ signiﬁcantly (P 0.016), again sug-
after photon (open circles, broken line) or epithermal neutron irradiations
at depths of 2.5 cm (open triangles, lower solid line) and 9.7 cm (closed
gesting that not all the radiobiological mechanisms are ac-
triangles, upper solid line). The points represent individual mice; thus for counted for in this analysis. More data from a wider range
neutrons there are pairs of points at the doses for the ‘‘inner’’ and ‘‘outer’’ of beam compositions are required to conﬁrm whether this
positions (see the Results). The error bars represent binomial 95% con- discrepancy is real and whether it can be accounted for by
ﬁdence intervals of the values for individual mice (six sections). For repair mechanisms or by the use of more complex survival
clarity, the upper half of each error bar is shown for the photon data and
the lower half for the 2.5-cm data, and no error bars are shown for the
models, such as the dual killing model (16).
9.7-cm data. Note the axis break to show data with zero surviving crypts; The purpose of this experiment was to study the suit-
some of these points represent two identical superimposed values. The ability of the regeneration of intestinal crypt cells in mice
lines represent the ﬁt to the LQ model (Eq. 1 and Table 2). as a possible transfer measurement to relate the biological
properties of different NCT beams. In this respect, it is
able, but there is some evidence of a lack of ﬁt, and it can important to note that the stated RBEs do not represent the
be shown statistically that if we allow the biological param- actual RBE values that should be applied in the clinical
eter m to take different values for the three curves, a sig- prescription of absorbed dose, which, due to the intrinsic
niﬁcantly better ﬁt is obtained (P 0.001). The formal lack variability of RBE, requires the use of clinically relevant
of ﬁt suggests that the shapes of the curves do differ some- biological systems and irradiation conditions (17). Conse-
what for the different radiation sources, an observation that quently, the results should be interpreted as a ‘‘beam
could be ascribed to differences in the mechanism of action weighting factor’’ or ‘‘RBEbeam’’ that is primarily useful in
or to dose-rate effects. Crypt survival curves are not always
well represented by the LQ model (16), and the discrepancy TABLE 3
is perhaps more likely to be accounted for by the difference Fitted Parameters of the LQ Model for the Beam
in survival range sampled in the measurements of the two Components, with Separate / Ratios for
beams, which differentially sample different components of Neutrons and Photons
the survival curve. More data are required to investigate Parameter Estimate (SE)
this further, but from Fig. 2 it is clear that a single RBE is
( ) (Gy) 0.050 (0.011)
a good practical summary of the differences between ( / ) (Gy) 0.45 (0.11)
beams. ( / )n (Gy) 0.11 (0.07)
The NCT beams actually contain a mixture of rays and log(m) 1.43 (0.12)
neutrons (the neutron components comprise only 23% and RBE ( )a at depth of 2.5 cm 9.7 (2.5)
10% of the dose at 2.5 and 9.7 cm, respectively; see Table RBE ( )a at depth of 9.7 cm 7.5 (2.0)
RBE20b at depth of 2.5 cm 5.2 (0.6)
1). An alternative parameterization of the LQ model is to RBE20b at depth of 9.7 cm 4.0 (0.7)
consider the RBE and the proportions of rays and neu-
trons (Eq. 2) and to obtain estimates for the RBE of the Note. Since the RBE estimates for neutrons relative to the photon beam
were signiﬁcantly different for the two depths, separate estimates are
neutron component of the beams. Since the / ratios provided.
would be expected to differ for the two types of radiation, a
The neutron RBE estimate extrapolated to zero dose.
the RBE estimate will be dependent on dose. The param- b
The neutron RBE at a survival level of 20 crypts per circumference.
RBE OF THE MIT EPITHERMAL NEUTRON BEAM 809
comparing the relative biological potency of a beam under 7. P. J. Binns, K. J. Riley and O. K. Harling, Dosimetric comparison of
six epithermal neutron beams using an ellipsoidal water phantom. In
well-deﬁned irradiation conditions. This will enable varia- Research and Development in Neutron Capture Therapy (W. Sauer-
tions in irradiation conditions within a single beam (e.g. as wein, R. Moss and A. Wittig, Eds.), pp. 405–410. Monduzzi, Bolo-
a function of depth) or between different beams at the same gna, 2002.
depth to be quantiﬁed and if necessary incorporated in the 8. K. J. Riley, P. J. Binns, D. D. Greenberg and O. K. Harling, A phys-
ical dosimetry intercomparison for BNCT. Med. Phys. 29, 898–904
speciﬁcation of the clinical dose. (2002).
9. J. Gueulette, M. Beauduin, V. Gregoire, S. Vynckier, B. M. De Cos-
ter, M. Octave-Prignot, A. Wambersie, K. Strijkmans, A. De Schrijver
ACKNOWLEDGMENTS and P. Chauvel, RBE variation between fast neutron beams as a func-
tion of energy. Intercomparison involving 7 neutrontherapy facilities.
The authors thank Profs. Jeff Coderre and Otto Harling for their com- Bull. Cancer/Radiother. 83 (Suppl. 1), 55s–63s (1996).
ments during the preparation of this manuscript. This work was supported
10. O. K. Harling, K. J. Riley, T. H. Newton, B. A. Wilson, J. A. Bernard,
by IAEA (contract no. 11063/RO) and by the U.S. DOE (contracts DE- L. W. Hu, E. J. Fonteneau, P. T. Menadier, S. J. Ali and P. M. Busse,
FG02-97ER62489 and DE-FG02-96ER62193). The ﬁssion converter-based epithermal neutron irradiation facility at
the Massachusetts Institute of Technology reactor. Nucl. Sci. Eng.
Received: October 15, 2003; accepted: August 1, 2005 140, 223–240 (2002).
11. H. R. Withers and M. M. Elkind, Microcolony survival assay for
cells of mouse intestinal mucosa exposed to radiation. Int. J. Radiat.
REFERENCES Biol. 17, 261–267 (1970).
12. C. S. Potten, M. Rezvani, J. H. Hendry, J. V. Moore and D. Major,
1. G. L. Locher, Biological effects and therapeutic possibilities of neu-
The correction of intestinal microcolony counts for variation in size.
trons. Am. J. Roentgenol. Radium Ther. 36, 1–36 (1936).
Int. J. Radiat. Biol. 40, 321–326 (1981).
2. J. A. Coderre, M. S. Makar, P. L. Micca, M. M. Nawrocky, H. B. 13. K. J. Riley, P. J. Binns and O. K. Harling, Performance characteristics
Liu, D. D. Joel, D. N. Slatkin and H. I. Amols, Derivations of relative of the MIT ﬁssion converter based epithermal neutron beam. Phys.
biological effectiveness for the high-LET radiations produced during Med. Biol. 48, 943–958 (2003).
boron neutron capture irradiations of the 9L rat gliosarcoma in vitro
and in vivo. Int. J. Radiat. Oncol. Biol. Phys. 27, 1121–1129 (1993). 14. R. D. Rogus, O. K. Harling and J. C. Yanch, Mixed ﬁeld dosimetry
of neutron beams for boron neutron capture therapy at the MITR-II
3. P. R. Gavin, S. L. Kraft, R. Huiskamp and J. A. Coderre, A review: research reactor. Med. Phys. 21, 611–1625 (1994).
CNS effects and normal tissue tolerance in dogs. J. Neuro-Oncol.
33, 71–80 (1997). 15. ICRU, Determination of Absorbed Dose in a Patient Irradiated by
Beams of X or Gamma Rays in Radiotherapy Procedures. Report 24,
4. J. A. Coderre and G. M. Morris, The radiation biology of boron International Commission on Radiation Units and Measurements, Be-
neutron capture therapy. Radiat. Res. 151, 1–18 (1999). thesda, MD, 1976.
5. J. Benczik, T. Seppala, M. Snellman, H. Joensuu, G. M. Morris and 16. S. A. Roberts, J. H. Hendry and C. S. Potten, Intestinal crypt clon-
J. W. Hopewell, Evaluation of the relative biological effectiveness of ogens: a new interpretation of radiation survival curve shape and
a clinical epithermal neutron beam using dog brain. Radiat. Res. 159, clonogenic cell number. Cell Prolif. 36, 215–231 (2003).
199–209 (2003). 17. J. Gueulette, H. G. Menzel, P. Pihet and A. Wambersie, Speciﬁcation
6. O. K. Harling, International dosimetry exchange: Status report. In of radiation quality in fast neutron therapy: Microdosimetric and ra-
Research and Development in Neutron Capture Therapy (W. Sauer- diobiological approach. In Recent Results in Cancer Research (R.
wein, R. Moss and A. Wittig, Eds.), pp. 333–340. Monduzzi, Bolo- Engenhart-Cabillic and A. Wambersie, Eds.), pp. 31–53. Springer-
gna, 2002. Verlag, Berlin, Heidelberg, New York, 1998.