CLINICAL RESEARCH
Europace (2013) 15, 60–65
doi:10.1093/europace/eus197
Sudden death and ICDs
High incidence of implantable cardioverter
defibrillator malfunctions during radiation
therapy: neutrons as a probable cause of soft errors
Jan Elders 1*, Martina Kunze-Busch 2, Robert Jan Smeenk2, and Joep L.R.M. Smeets 3
1
Department of Cardiology, Canisius Wilhelmina Hospital, B16. PO Box 9015, 6500 GS, Nijmegen, The Netherlands; 2Department of Radiation Oncology, Radboud University
Nijmegen Medical Centre, Nijmegen, The Netherlands; and 3Department of Cardiology, Radboud University Nijmegen Medical Centre, Nijmegen, The Netherlands
Received 12 December 2011; accepted after revision 25 May 2012; online publish-ahead-of-print 29 July 2012
Aims
To investigate the behaviour of the implantable cardioverter defibrillator (ICD) function during actual radiotherapy
sessions.
.....................................................................................................................................................................................
Methods
Fifteen patients with an ICD underwent 17 radiation treatments for cancer [cumulative dose to the tumour was
and results
between 16 Gray (Gy) and 70 Gy; photon beams with maximum energies between 6 megaelectronvolt (MeV)
and 18 MeV were employed]. During every session, the ICD was programmed to a monitoring mode to prevent inappropriate therapy delivery. Afterwards, the ICDs were interrogated to ensure proper function. Calculated radiation
dose at the ICD site was ,1 Gy in all patients. In 5 out of 17 radiation treatments (29%) the ICDs showed 6 malfunctions (35%). We noticed four disturbances in the memory data or device resets during radiation treatment and
one case of inappropriate ventricular fibrillation detection due to external noise. In one case a late device data error
was observed. All malfunctions occurred at 10 and 18 MeV beam energies.
.....................................................................................................................................................................................
Conclusion
Despite the fact that all recommended precautions were taken to minimize the damage to the ICDs during radiotherapy and the calculated dose to the ICDs was ,1 Gy, in 29% of the treatments a malfunction occurred. We
observed a possible correlation between the beam energy and the malfunctions. This correlation may be due to
an interaction between neutrons produced in the head of the linear accelerator at beam energies ≥10 MeV, and
boron-10 which is present in the integrated circuit.
----------------------------------------------------------------------------------------------------------------------------------------------------------Keywords
Radiotherapy † Implantable cardioverter defibrillator † Malfunctions † Neutrons † Reset
In the western world the population is ageing progressively. With
ageing the incidence of malignancies is increasing rapidly and is one
of the major causes of death. Radiation therapy (RT) may be used
as a treatment both with curative and palliative intent. The goal of
RT is to deliver a high dose to the tumour while minimizing radiation damage to the surrounding healthy tissue. In addition to
the increase in malignancies, coronary artery disease remains the
major cause of death in the western world. To prevent sudden
cardiac death in cardiovascular disease implantable cardioverter
defibrillators (ICDs) are implanted frequently. Hence, more
patients with an ICD will be treated with RT because of malignancies. Although the American Association of Physicists in Medicine
published a guideline for RT in patients with an implantable
pacemaker in 1994, no recent guideline is available for patients
with an ICD undergoing RT.1 The purpose of our study is to investigate whether ICD behaviour and function is influenced by RT.
Methods
In this single-centre study ICD function and behaviour was studied in
15 patients undergoing 17 RT treatments consisting of a total of 290
sessions. The patient and clinical characteristics including the device information are given in Tables 1 and 2, respectively.
The following ICD parameters were measured before and after
every RT session: stimulation threshold, lead and shock impedance,
sensing characteristics and monitored events.2 During the RT
session, the ICD was temporarily programmed to ‘monitor only’ and
* Corresponding author. Tel: +31 24 3659726; fax: +31 24 3566118. Email: [email protected]
Published on behalf of the European Society of Cardiology. All rights reserved. & The Author 2012. For permissions please email: [email protected].
61
ICD malfunctions during radiation therapy
Table 1 Patient characteristics
Pt.
Gender
(M/F)
Age
ICD
indication
Prior MI
(Y/N)
CABG
(Y/N)
LVEF
(%)
NYHA
(class)
CMP
(Y/N)
1
M
2
M
3
4
PM
dependant
Prior
arrhythmia
70
Primary
Y
N
32
III
N
No
None
78
Primary
Y
Y
18
III
Y
CRT
None
M
M
67
71
Primary
Secondary
Y
Y
Y
N
30
20
II
III
Y
Y
CRT
No
None
NSVT
5
M
72
Secondary
N
N
12
III
Y
CRT
VT
6
7
M
F
60
73
Secondary
Primary
N
Y
N
Y
15
20
III
III
Y
Y
CRT
CRT
NSVT
None
8
M
72
Secondary
Y
N
11
III
Y
CRT
VF
9
10
M
F
75
74
Primary
Primary
Y
N
N
N
30
20
II
II
N
Y
No
No
None
None
11
M
72
Primary
N
N
25
II
Y
No
None
12
13
M
M
76
69
Primary
Primary
Y
Y
N
Y
35
30
II
II
N
N
No
No
None
None
14
M
65
Primary
N
N
25
II
Y
No
None
15
M
65
Primary
Y
N
15
II
Y
CRT
None
...............................................................................................................................................................................
MI, myocardial infarction; CABG, coronary artery bypass grafting; LVEF, left ventricular ejection fraction; NYHA, New York Heart Association classification for heart failure;
CMP, cardiomyopathy; NSVT, non-sustained ventricular tachycardia; VT, ventricular tachycardia; VF, ventricular fibrillation.
reprogrammed after the radiation exposure. In ‘monitor only’ mode
detection of tachycardia is stored but no therapy is given in order to
avoid inappropriate shocks.
According to Hurkmans et al.,2 modern ICDs seem to be more sensitive to RT than pacemakers. Therefore we adopted the recommendations for RT treatment of ICD patients.3 We did not alter the pacing
parameters during RT. Eight patients (54%) did not have a pacing indication and were programmed to dual-chamber pacing and sensing, but
inhibited mode only with a pacing rate of 40 b.p.m. Seven patients
(46%) had cardiac resynchronization therapy (CRT). In order to maintain CRT no reprogramming was performed. The electrocardiogram
was monitored during each session and trained personnel were
present. Standard cardiopulmonary resuscitation equipment was directly available. The ICD location was outside the direct irradiation
field. The dose to the ICD due to scattered radiation was estimated
with the treatment planning system Pinnacle (Philips Medical Systems,
Andover, MA, USA), with the collapsed cone convolution superposition
algorithm. (The Pinnacle convolution superposition dose model is a
three-dimensional dose computation and is based on the work of
Mackie et al.4).
In addition, the dose to the ICD was measured in two patients,
during the first two RT sessions, with a thermoluminescent dosimeter
(TLD-100, Harshaw/Bicron, Solon, OH, USA) which was placed on the
skin next to the ICD.
Three months after the completion of the radiation treatment all
ICDs were checked for any late device malfunctions. Regular ICD controls were planned every 6 months or in case of an event.
Results
Fifteen patients (13 males, 2 females) received a total of 17 radiation treatments. The median age was 72 years (minimum 60
years, maximum 78 years). Eleven patients (73%) had a primary
ICD indication and seven patients (46%) had a CRTD device.
The median left ventricular ejection fraction was 20% (range 11–
35%). Seven patients (46%) had a New York Health Association
functional classification for heart failure (NYHA) Class III and
eight patients (54%) a NYHA Class II. Ten patients (66%) had a
prior myocardial infarction and four patients (26%) had coronary
artery bypass grafting. Four patients (26%) had prior ventricular
arrhythmias (Table 1).
The prescribed dose varied between 16 Gray (Gy) and 70 Gy,
with fraction doses varying between 2 and 8 Gy. Polyenergetic
photon beams with maximum energies of 6 megaelectronvolts
(MeV), 10 MeV, and 18 MeV were used (referred to as 6, 10,
and 18 MeV photon beams, respectively). (In the field of radiotherapy it is established to express the energy of therapeutic photon
beams in megavolt.)
The total dose was delivered in several treatment sessions
(ranging from 2 to 33 sessions).
Radiation therapy was completed in all 15 patients without any
major cardiac event. The dose measured with a TLD placed close
to the ICD was negligible (,0.1 Gy).
We did not observe any permanent loss of function, increase of
threshold, fluctuations in sensing, or impedances in any device. We
noticed one inappropriate tachycardia sensing but no inappropriate therapy.
In four patients we noticed a temporary ICD malfunction.
During radiation treatment two patients had a device reset and
the ICDs of two other patients had invalid data retrieval. In one
patient we observed a second data error 9 months after RT in a
device which already had a reset situation during RT. It was not
clear whether this late device error was caused by RT. An overview is given in Table 2.
62
J. Elders et al.
Table 2 Patient characteristics, type of implantable cardioverter defibrillator, therapy dose, and implantable
cardioverter defibrillator malfunction. (Patient 1 was treated with photon and electron beams)
RT
no.
Pat
no.
Type of ICD
Treated
carcinoma
Total dose/dose
per fraction (Gy)
to the tumour
Radiation beam
maximum energy:
megaelectron volt
(MeV)
Observation
(text as annotated
by the programmer)
6; 12
6; 9
None
None
...............................................................................................................................................................................
1
2
1
1
Medtronic Marquis
Medtronic Marquis
Right ear
Thumb
66/2
66/2
3
2
Medtronic Virtuoso
Prostate
70/2.5
10
Invalid data retrieval
4
5
3
4
Boston Scientific Contak renewal
Boston Scientific Vitality II
Rectum
Prostate
25/5
67.5/2.25
18
18
None
None
6
5
St Jude Medical Atlas II
Prostate
70/2.5
18
7
6
Medtronic Insync Sentry
Oesophagus
60/2
18
Device reset trend data
error (after 9
months)
Device reset
8
7
Boston Scientific Contak Renewal
Rectum
25/5
18
Invalid data retrieval
9
10
8
9
Medtronic Concerto
Medtronic Entrust
Prostate
Thoracic
vertebra
64.4/2.3
25/5
10
10
None
None
11
12
10
11
Medtronic Marquis
Boston Scientific Contak Renewal
Lung right
Cerebrum
50/2
20/4
6
6
None
None
13
12
Medtronic Secura
Groin right
39/3
6; 10
None
14
15
13
13
St Jude Medical Atlas II
St Jude Medical Atlas II
Lung left
Cerebrum
16/8
20/4
10
6
None
None
16
14
Medtronic Virtuoso
Oesophagus
30/3
18
None
17
15
St Jude Promote Quadra
Femur left
20/5
10
Noise (inappropriate
tachycardia sensing)
Discussion
In our study, in 5 of the 17 (29%) patients treated with RT, malfunctions of the ICDs were seen despite the fact that the ICDs
were situated outside the direct radiation field and the calculated
cumulative radiation dose to the ICDs was as recommended
below 1 Gy.3 In one ICD a late data error was found. In total six
device errors (acute and late errors) were observed (35%). The
reset errors needed intervention by the manufacturer to restore
normal ICD function. In case of the data errors the data could
not be retrieved and were permanently lost.
Disturbances in the random access memory and read only
memory will trigger a ‘reset’ routine that attempts to overcome
the error and restore normal operation of the ICD. Sometimes
data will be lost permanently and in cases of unsuccessful reset
the device will restore as many permanently programmed settings
as possible. It should be realized that the electrical circuits for ICD
function and memory are working independently from each other.
This means that in case of a reset situation ventricular fibrillation (VF)
detection and defibrillation should remain available. However, this
has not been investigated in daily clinical practice.
In our study we did have one episode of inappropriate sensing of
VF due to external noise. The ICD would have delivered a shock if
the programmed therapy would have been enabled. A possible explanation for this error may be electromagnetic noise interference
produced by the linear accelerator.
Another cause for errors, as demonstrated in in vitro studies,2,5 – 7
is that the ICD can be influenced by radiation if it is positioned
in the direct radiation beam. This included sensing and charging
problems in different brands of ICDs. The problems occurred infrequently (8 : 1000)8 and it is believed that a high radiation dose
is the main cause of the problem. It is recommended to position
the ICD outside of the direct beam and to avoid a cumulative
dose to the ICD of .1 Gy.3 All of our ICDs were located
outside the radiation field.
The question remains as to why we saw so many malfunctions of
the ICDs during RT. The ICDs were located outside the direct radiation beam and the calculated total radiation dose to the ICDs
was ,1 Gy, so the dose to the ICDs should not to be the cause
of the problems. We did, however, observe a possible correlation
between the beam energy and the occurrence of the malfunctions:
all of the malfunctions occurred at a beam energy ≥10 MeV
(Table 2).
In vivo data on the effect of RT in ICD patients is limited to a few
case reports or studies with small numbers of patients.6,7,9 – 13 In
these studies, a total of 63 ICDs were exposed to radiation. In
six cases malfunctions to the ICD were seen varying from resets
and one runaway pacemaker (Table 3). In three of these cases
the beam energy delivered to the patient was .10 MeV. In one
case the beam energy is not mentioned (Table 3).
Starting at 10 MeV beam energy, neutrons are produced in
the head of the linear accelerator. We hypothesize that these
ICD malfunctions during radiation therapy
Table 3 Overview of the in vivo studies of the influence of radiation therapy on ICDs
Author/year
(reference no.)
Manufacturers
Number
(ICD)
Total dose to
tumour: Gray
(Gy)
Fraction
dose (Gy)
3
,5
Unknown
20
4
Beam energy:
megaelectron
volt (MeV)
Direct/
scatter dose
to ICD
Unknown
Failure dose to
ICD
(Gy/MeV)
Malfunction
No. malfunctions
failure rate (%)
.............................................................................................................................................................................................................................................
Niehaus 20016
Unknown
Kapa 20087
Hoecht 20029
Medtronic, Boston
Scientific, St Jude
Medical
Boston Scientific
Nĕmec 200710
Boston Scientific
1
60
1.8
Thomas 200411
Medtronic
1
56
Lau 200812
Gelblum 200913
Medtronic
Medtronic, Boston
Scientific, St Jude
medical
1
33
74
81
4
Unknown
Unknown
None (0%)
4
6
Scatter
None (0%)
Unknown
6
Scatter
,0.5/6
Unknown
Scatter
5.4/beam energy
unknown
2
18
Scatter
Dose unknown/18
Reset
1 (100%)
2
2.9
23
15 (1 pt.) 6
Scatter
Scatter
4/23
2/15
Reset
Reset
1 (100%)
1 (3%), after
reduction
of beam energy
to 6 MV none
Reset (same patient,
ICD replacement)
Runaway pacemaker
resulting in
resuscitation
2 (50%)
1 (100%) (ICD
close to beam)
Kapa et al.7 did an in vivo as well as an in vitro study.
63
64
neutrons may cause malfunctions of the ICD. This was suggested
earlier by Gelblum and Amols13 who suggested a possible correlation between neutrons and malfunction of ICDs during RT.
To confirm the production of neutrons in the head of the linear
accelerator at beam energies ≥10 MeV, we irradiated a phantom
with 6, 10, and 18 MeV photon beams. The neutron dose was measured with a neutron detector (LB 6411 neutron dose rate detector, Berthold, Bad Wildbad, Germany) in the isocentre of the beam,
next to the phantom and 1 m away from isocentre. We noticed a
(relative) increase in dose with increasing beam energy in all three
locations.
The link between neutrons and soft errors is very well documented and accepted by the electronics industry.14 – 26 The beam
energy employed may have significant implications as neutrons
will be produced in the head of the linear accelerator at energies
≥10 MeV. Baumann et al. 25 first reported the potential for neutrons to cause soft errors in some integrated circuits (ICs).
These errors are caused by the production of a particles in the
(n, a) capture reaction with boron. The nuclear reaction that
occurs when boron-10 is irradiated with low-energy thermal neutrons is also used as a radiation treatment for cancer known as
Boron Neutron Capture Therapy.27
Wilkinson et al. state that neutrons interact with boron-10
found in the lower intermetal dielectric layers of an IC, resulting
in the production of lithium-7 and a-particles in the immediate
vicinity of the active circuitry. Subsequent interaction of these
charged species with the IC causes soft errors.14 Figure 1 shows
a cross section of a typical IC using time-of-flight secondary ion
mass spectrometry. The presence of boron is in the natural
isotope ratio of 80% boron-11 and 20% boron-10.
J. Elders et al.
Gelblum and Amols reported a reset of an ICD during RT with a
beam energy of 15 MeV. The device was well located out of the
radiation portal. The patient completed therapy with a beam
energy of 6 MeV without any further issues. Since then, they
have treated all patients with ICDs with 6 MeV beams and have
not had any further events.13
In an in vitro study by Wilkinson et al. several ICs were exposed
to a radiation beam with an energy of 18 and 6 MeV. When
exposed to a beam with an energy of 18 MeV, a total of 93
errors were recorded. No errors were recorded when the
beam energy was adjusted to 6 MeV.14
In another study on scattered radiation no malfunctions with a
photon beam of 6 MeV and a relatively high dose of 4 Gy to the
ICD were observed,7 whereas we observed malfunctions with
high energy beams and a low dose to the ICDs.
Based on the in vivo and in vitro studies and our own observations,
we find that neutron interaction with boron should be added to the
list of possible causes for malfunction of the ICD during RT.
Limitations
This is a single-centre experience in 15 patients. Our observations
should be investigated in a multicentre prospective study. Until
further evidence is provided in this clinical important issue ICD
patients treated with radiotherapy should be closely monitored.
Conclusions
Our observation suggests a possible interaction between neutrons
with boron located in the internal circuitry of ICDs causing soft
errors and device malfunctions. Until further evidence is provided
Figure 1 Time-of-flight secondary ion mass spectrometry results of four sample integrated circuits from two equipment manufacturers.
Time-of-flight secondary ion mass spectrometry provides spectroscopy for characterization of chemical composition, imaging for determining
the distribution of chemical species, and depth profiling for thin film characterization. In all cases, the upper material (blue) is aluminium used for
interconnects and the lower material (green) is boron. The boron isotopes are in a natural ratio of 80 – 20%. The transistors are below the
boron layer. (Picture and subscription used with permission of the original author, Wilkinson.14 &[2005IEEE].).
ICD malfunctions during radiation therapy
we recommend careful monitoring of patients and ICDs during and
after RT with beam energies ≥10 MeV.
Acknowledgements
The authors are most grateful to J.D. Wilkinson, BA for his useful
advice during the preparation of this manuscript and for his permission to use his illustration. In addition we appreciate the comments of Y. Lokhandwala, MD and N. van Hemel MD, PhD in the
final preparation of the manuscript.
Conflict of interest: none declared.
References
1. Marbach JR, Sontag MR, Van Dyk J, Wolbarst A. Management of radiation oncology patients with implanted cardiac pacemakers: report of AAPM Task Group
No.34. American Association of Physicists in Medicine. Med Phys 1994;21:85 –90.
2. Hurkmans CW, Scheepers E, Springorum BG, Uiterwaal H. Influence of radiotherapy on the latest generation of implantable cardioverter-defibrillators.
Int J Radiat Oncol Biol Phys 2005;63:282–9.
3. Solan AN, Solan MJ, Bednarz G, Goodkin MB. Treatment of patients with cardiac
pacemakers and implantable cardioverter-defibrillators during radiotherapy.
Int J Radiat Oncol Biol Phys 2004;59:897–904.
4. Mackie TR, Reckwerdt PJ, Holmes TW, Kubsad SS. Review of convolution/superposition methods for photon beam dose computation. Proceedings of the Xth
ICCR 1990:20–3.
5. Rodriquez F, Filimonov A, Henning A, Coughlin C, Greenberg M. Radiation
induced effects in multiprogrammable pacemakers and implantable defibrillators.
Pacing Clin Electrophysiol 1991;14:2143 –53.
6. Niehaus M, Tebbenjohanns J. Electromagnetic interference in patients with
implanted pacemakers and cardioverter-defibrillators. Heart 2001;86:246–8.
7. Kapa S, Fong L, Blackwell CR, Herman MG, Schomberg PJ, Hayes DL. Effects of
scatter radiation on ICD and CRT function. Pacing Clin Electrophysiol 2008;
31:727 –32.
8. Maisel WH. Pacemaker and ICD generator reliability: meta-analysis of device
registries. J Am Med Assoc 2006;295:1929 –34.
9. Hoecht S, Rosenthal P, Sancar D, Behrens S, Hinkelbein W, Hoeller U. Implantable cardiac defibrillators may be damaged by radiation therapy. J Clin Oncol 2002;
20:2212 –3.
10. Nĕmec J. Runaway implantable defibrillator—a rare complication of radiation
therapy. Pacing Clin Electrophysiol 2007;30:716–8.
11. Thomas D, Becker R, Katus HA, Schoels W, Karle CA. Radiation therapy-induced
electrical reset of an implantable cardioverter defibrillator device located outside
the irradiation field. J Electrocardiol 2004;37:73 –4.
12. Lau DH, Wilson L, Stiles MK, John B, Shashidar P, Dimitri H et al. Defibrillator
reset by radiotherapy. Int J Cardiol 2008;130:e37 –8.
65
13. Gelblum DY, Amols H. Implanted cardiac defibrillator care in radiation oncology
patient population. Int J Radiat Oncol Biol Phys 2009;73:1525 –31.
14. Wilkinson JD, Bounds C, Brown T, Gerbi BJ, Peltier J. Cancer radiotherapy equipment as a cause of soft-errors in electronic equipment. IEEE Trans Device Mater
Reliability 2005;5:449 – 51.
15. Franco L, Gomez F, Iglesias A, Pardo J, Lobato R, Marı́n J et al. SEUs on commercial SRAM induced by low energy neutrons produced at a clinical linac facility.
European Congress RADECS 2005. Cap d’Agde, France; 2005 September 13 – 19.
16. Baumann RC, Smith EB. Neutron-induced boron fission as a major source of soft
errors in deep submicron SRAM devices. IEEE 38th Annual International Reliability Physics Symposium. 2000. San Jose, CA, USA; 2000 April 10 –13. pp152 – 7.
17. Warren KM, Wilkinson JD, Morrison S, Weller RA, Porter ME, Sierawski BD et al.
Modeling alpha and neutron induced soft errors in static random access memories. IEEE International Conference on Integrated Circuit Design and Technology,
2007. ICICDT ‘07. Austin, TX, USA; 2007 May 30 – June 1. pp. 1 –4.
18. Warren KM, Sierawski BD, Weller RA, Reed RA, Mendenhall MH, Pellish JA et al.
Predicting thermal neutron-induced soft errors in static memories using TCAD
and physics-based Monte Carlo simulation tools. IEEE Electron Device Lett 2007;
28:180–2.
19. Hajek M, Berger T, Schoner W, Vana N. Analysis of the neutron component at
high altitude mountains using active and passive measurement devices. Nucl
Instrum Methods Phys Res A 2002;476:69 –73.
20. Normand E, Vranish K, Sheets A, Stitt M, Kim R. Quantifying the double-sided
neutron SEU threat, from low energy (thermal) and high energy (.10 MeV) neutrons. IEEE Trans Nucl Sci 2006;53:3587 –95.
21. Ziegler J. Terrestrial cosmic rays and soft errors. IBM J Res Dev 1996;40:19.
22. Dyer CS, Clucas SN, Sanderson C, Frydland AD, Green RT. An experimental
study of single-event effects induced in commercial SRAMS by neutrons and
protons from thermal energies to 500 MeV. IEEE Trans Nucl Sci 2004;51:2817 – 24.
23. Olmos M, Gaillard R, Van Overberghe A, Beaucour J, Wen S, Chung S. Investigation
of thermal neutron induced soft error rates in commercial Srams with 0.35 mm to
90 nm technology. 44th Annual, IEEE International. Reliability Physics Symposium
Proceedings. San Jose, CA, USA; 2006 March 26 –30. pp. 212– 6.
24. Baumann RC, Hossain TZ, Murata S, Kitagawa H. Boron compounds as a dominant source of alpha particles in semiconductor devices. 33rd Annual, IEEE International Reliability Physics Symposium Proceedings. Las Vegas, NV, USA; 1995
April 4– 6. pp. 297 –302.
25. Baumann RC, Hossain TZ, Smith E, Murata S, Kitagawa H. Boron as a primary
source of radiation in high density DRAM’s. Symposium on VLSI technology,
Digest of technical papers. Kyoto, Japan; 1995 June 6– 8. pp. 81 –2.
26. Kobayashi H, Shiaishi K, Tsuchiya H, Motoyoshi M, Usuki H, Nagai Y et al. Soft
errors in SRAM devices induced by high energy neutrons, thermal neutrons
and alpha particles. Digest. IEEE International Electron Devices Meeting. San Fransisco, CA, USA; 2002 December 8–11. pp. 337 – 40.
27. Barth RF, Coderre JA, Graça M, Vicente H, Blue TE. Boron neutron capture
therapy of cancer: current status and future prospects. Clin Cancer Res 2005;
11:3987 –4002.