VOLUME
29
NUMBER
7
MARCH
1
2011
JOURNAL OF CLINICAL ONCOLOGY
ORIGINAL
REPORT
Genetic Factors Underlying the Risk of Thalidomide-Related
Neuropathy in Patients With Multiple Myeloma
David C. Johnson, Sophie L. Corthals, Brian A. Walker, Fiona M. Ross, Walter M. Gregory, Nicholas J. Dickens,
Henk M. Lokhorst, Hartmut Goldschmidt, Faith E. Davies, Brian G.M. Durie, Brian Van Ness,
J. Anthony Child, Pieter Sonneveld, and Gareth J. Morgan
See accompanying editorial on page 783
From the Institute of Cancer Research,
ABSTRACT
London; Cytogenetics Group, Wessex
Regional Cytogenetic Laboratory, Salis-
bury; University of Leeds, Leeds,
Purpose
To indentify genetic variation that can modulate and predict the risk of developing thalidomide-
United Kingdom; Erasmus Medical
Center, Rotterdam; University Medical
related peripheral neuropathy (TrPN).
Center, Utrecht, the Netherlands;
Patients and Methods
University of Heidelberg, Heidelberg,
We analyzed DNA from 1,495 patients with multiple myeloma. Using a custom-built single
Germany; Cedars Sinai, Los Angeles,
CA; and University of Minnesota.
nucleotide polymorphism (SNP) array, we tested the association of TrPN with 3,404 SNPs. The
Minneapolis, MN.
SNPs were selected in predicted functional regions within 964 genes spanning 67 molecular
pathways thought to be involved in the pathogenesis, treatment response, and adverse effects
Submitted January 13, 2010; accepted
June 22, 2010; published online ahead
associated with myeloma and its therapy. Patient cases and controls were derived from two large
of print at www.jco.org on January 18,
clinical trials that compared thalidomide with conventional-based treatment in myeloma patients
2011.
(Medical Research Council Myeloma-IX and HOVON-50/GMMG-HD3).
Supported by International Myeloma
Results
Foundation as part of Bank on a Cure
We report TrPN associations with SNPs--ABCA1 (rs363717), ICAM1 (rs1799969), PPARD
(BOAC) initiative, Myeloma UK, and the
(rs2076169), SERPINB2 (rs6103), and SLC12A6 (rs7164902)--where we show cross validation of
National Institute for Health and Biologi-
cal Research at the Royal Marsden
the associations in both trials. To investigate whether TrPN SNP associations were related to
Hospital.
exposure to thalidomide only or general drug-related peripheral neuropathy, we performed a
Terms in blue are defined in the glos-
second analysis on patients treated with vincristine. We report SNPs associated with vincristine
sary, found at the end of this article
neuropathy, with a seemingly distinct underlying genetic mechanism.
and online at www.jco.org.
Conclusion
Authors' disclosures of potential con-
Our results are consistent with the hypothesis that an individual's risk of developing a peripheral
flicts of interest and author contribu-
neuropathy after thalidomide treatment can be mediated by polymorphisms in genes governing
tions are found at the end of this
repair mechanisms and inflammation in the peripheral nervous system. These findings will
article.
contribute to the development of future neuroprotective strategies with thalidomide therapy and
Clinical Trials repository link available on
the better use of this important compound.
JCO.org.
Corresponding author: Gareth J.
J Clin Oncol 29:797-804. © 2011 by American Society of Clinical Oncology
Morgan, PhD, MD, Section of
Haemato-Oncology, Institute of Cancer
Research, 15 Cotswold Rd, Belmont,
drawal of thalidomide, which can lead to symptom
Sutton, Surrey SM2 5NG, United King-
INTRODUCTION
resolution in up to 16 weeks9; however, in some
dom; e-mail: gareth.morgan@icr.ac.uk.
Peripheral neuropathy is a significant adverse event
cases, neuropathy is irreversible.
© 2011 by American Society of Clinical
Oncology
in patients with multiple myeloma treated with tha-
At a pathologic level, TrPN is a length-
0732-183X/11/2907-797/$20.00
lidomide. Thalidomide-related peripheral neu-
dependent, predominantly sensory axonal neurop-
athy affecting large and small fibers,1,5 but has also
DOI: 10.1200/JCO.2010.28.0792
ropathy (TrPN) typically consists of symmetric
paresthesias, with loss of tactile and pain response as
been classified as a ganglionopathy.5,10 Reduction in
well as numbness and muscle cramps.1 Rates of neu-
amplitude or absence of sensory nerve action poten-
ropathy after thalidomide treatment vary from 15%
tials is the most common electrophysiologic altera-
to 70%, with the risk of neuropathy being related to
tion that can precede symptoms or worsen after
the cumulative dose and duration of therapy.2-6 Fac-
thalidomide withdrawal and often does not re-
tors influencing the risk of neurotoxicity include
solve.6,11 Proposed mechanisms to explain TrPN in-
prior neuropathy, age,7 previous chemotherapy,
clude antiangiogenesis, direct toxic effects on the
and vitamin B
posterior root ganglia, and dysregulation of neu-
12 and/or folate deficiency.8 The main-
stay of TrPN prevention is dose reduction or with-
rotrophin activity through nuclear factor- B.
© 2011 by American Society of Clinical Oncology
797
Information downloaded from jco.ascopubs.org and provided by at CEDARS SINAI MEDICAL CENTER MEDICAL LIBRARY
Copyright © 2011 on March
American 1, 2011
Society from
of
192.231.86.98
Clinical Oncology. All rights reserved.

---

Johnson et al
To date, there have been only a limited number of small studies
trols. To ensure homogeneity of allelic frequencies, only patients of European
investigating the genetic factors that associate with TrPN.12-13 In this
descent were included.
study, we have sought to address this by investigating the genetic
Clinical Trials
variation associated with risk of TrPN in two large clinical trials con-
The Myeloma-IX study comprised two random assignments. The first
sisting of patients treated with thalidomide or vincristine and geno-
random assignment included an intensive pathway for younger, fitter patients
typed with a custom targeted panel of 3,404 single nucleotide
comparing cyclophosphamide (500 mg orally weekly), vincristine (0.4 mg
polymorphisms (SNPs).
intravenous on days 1 to 4), doxorubicin (9.0 mg/m2 on days 1 to 4), and
dexamethasone (40 mg on days 1 to 4 and 12 to 15) with cyclophosphamide,
thalidomide, and dexamethasone (CTD) using the same doses of cyclophos-
PATIENTS AND METHODS
phamide and dexamethasone combined with a target dose of thalidomide 200
mg. The second random assignment, for older, less fit patients, compared an
attenuated dose of CTD (thalidomide 100 mg) with melphalan (7.0 mg/m2
Clinical Samples
once daily orally on days 1 to 4 every 28 days) and prednisolone. The HOVON-
Peripheral-blood DNA samples were obtained from 1,495 patients pre-
50/GMMG-HD3 study randomly assigned patients to either three cycles of
senting with myeloma derived from two randomized clinical trials (Medical
vincristine (0.4 mg intravenous rapid infusion on days 1 to 4), doxorubicin (9
Research Council Myeloma-IX, n
993; and Dutch-Belgian Hemato-
mg/m2 intravenous rapid infusion on days 1 to 4), and dexamethasone (40
Oncology Cooperative Group [HOVON] -50/German-Speaking Myeloma
mg orally on days 1 to 4, 9 to 12, and 17 to 20) or the same regimen but with
Multicenter Group [GMMG] -HD3, n
502) comparing standard treatment
thalidomide (200 mg, but could be escalated to 400 mg) replacing the
with thalidomide-containing regimens. Peripheral neuropathy was assessed
vincristine (Fig 1A).
using National Cancer Institute Common Toxicity Criteria version 2.0. Neu-
ropathy events were only considered after induction therapy and not during
Genotyping and SNP Selection
maintenance. In a discovery set analysis, we compared genotype results from
DNA was extracted from frozen WBC pellets using Flexigene kit (Qia-
194 Myeloma-IX patients with neuropathy after exposure to thalidomide with
gen, Crawley, United Kingdom) and quantified using a NanoDrop Spectro-
416 control patients treated with thalidomide who did not develop neuropa-
photometer (NanoDrop Products, Wilmington, DE). Genotyping was
thy. Patient cases and controls were matched for age and sex. As validation, we
performed using an Affymetrix Targeted genotyping True-tag (BOAC) array
compared 74 thalidomide-treated patients with neuropathy from the
(Affymetrix, Santa Clara, CA).14-16 SNPs were selected using a hypothesis-
HOVON-50/GMMG-HD3 trial with 176 thalidomide-treated controls. We
driven strategy targeting genes and SNPs with previously described associa-
then carried out a comparison using patients with neuropathy
grade 2 in the
tions or putative functional effects. There was no sample ascertainment bias
Myeloma-IX trial; 75 patients with neuropathy were compared with 297
between patient cases and controls because genotyping was performed before
matched controls; in the HOVON-50/GMMG-HD3 analysis, 49 patients with
access to demographic and phenotypic data.
neuropathy
grade 2 were compared with 176 controls. In the nonthalido-
mide arms, neuropathy events were also seen after vincristine exposure. We
Statistical Analyses
carried out a nested case-control comparison using 76 patients with
Departures from Fisher's exact Hardy-Weinberg equilibrium at a level of
vincristine-related neuropathy from Myeloma-IX compared with 307 con-
P
.0001 and bias in missing data were controlled for each SNP. A Cochran-
trols. In the HOVON-50/GMMG-HD3 cohort, we compared genotypes from
Armitage trend test and a Fisher's exact test were performed to evaluate odds
26 patients with vincristine-related neuropathy with genotypes from 226 con-
ratios. Genomic inflation factor was evaluated based on the median 2 value
A
MYELOMA - IX
HOVON - 50
Nonintensive
Intensive
All
Fig 1. (A) Simplified treatment arms of the
CTDa
MP
CTD
CVAD
TAD (3×)
VAD (3×)
Myeloma-IX and Dutch-Belgian Hemato-
Oncology Cooperative Group (HOVON) -50/
Neuropathy Events
German-Speaking
Myeloma
Multicenter
Measured at Induction
Group (GMMG) -HD3 studies. (B) Patient
case and control comparisons used in single
CAD + G - CSF
CAD + G - CSF
nucleotide polymorphism (SNP) analysis for
samples from the Myeloma-IX and HOVON-
50/GMMG-HD3 studies. CTDa, attenuated
HDM200
HDM 1 (2×)
cyclophosphamide, thalidomide, and dexa-
methasone; MP, melphalan and prednisol-
B
MYELOMA - IX
HOVON - 50
one; CTD, cyclophosphamide, thalidomide,
and dexamethasone; CVAD, cyclophosph-
amide, vincristine, doxorubicin, and dexa-
CTDa
CTD
CVAD
TAD (3×)
VAD (3×)
methasone; TAD, thalidomide, doxorubicin,
(n = 541)
(n = 443)
(n = 557)
(n = 400)
(n = 400)
and dexamethasone; VAD, vincristine, doxo-
rubicin, and dexamethasone; CAD, cyclophos-
phamide, doxorubicin, and dexamethasone;
G-CSF, granulocyte colony-stimulating
Thal
Thal
Vin
Vin
Thal
Thal
Vin
Vin
factor; HDM, high-dose melphalan; Thal,
194 cases
416 controls
76 cases
307 controls
74 cases
176 controls
26 cases
226 controls
thalidomide; Vin, vincristine.
Thal > grad 2
Thal
Thal > grad 2
Thal
75 cases
297 controls
49 cases
176 controls
798
© 2011 by American Society of Clinical Oncology
JOURNAL OF CLINICAL ONCOLOGY
Information downloaded from jco.ascopubs.org and provided by at CEDARS SINAI MEDICAL CENTER MEDICAL LIBRARY
Copyright © 2011 on March
American 1, 2011
Society from
of
192.231.86.98
Clinical Oncology. All rights reserved.

---

Thalidomide-Related Neuropathy in Patients With Myeloma
for each set of analyses. To account for multiple testing, we carried out label-
was then assessed for its ability to correctly classify the remaining one third of
swapping permutation procedures on each SNP. Proxy association testing was
the Myeloma-IX and HOVON-50/GMMG-HD3 data sets.
performed, where multiple SNPs in a gene were found to be associated, to
In an alternate approach, a risk score classifier was generated based on the
resolve haplotypes. Association analysis was performed in the PLINK 1.07
summation of the associated at-risk and protective SNPs. Using a training set
program.17 Database for Annotation, Visualization, and Integrated Discovery
of two thirds of the Myeloma-IX data set, the log
(odds ratios) of the
10
Bioinformatics Resource18 was used to characterize pathways enriched in
associated SNPs for each patient was summed. This scoring system was as-
risk-associated genes. The functional impact of regulatory SNPs in association
sessed for its ability to correctly classify patients at risk in the remaining third of
with TrPN was investigated by ensuring that the gene was typically expressed in
the Myeloma-IX and HOVON-50/GMMG-HD3 data sets.
neurons19 and then subsequently examining myeloma tumor expression data
in relation to SNP genotype using a Wilcoxon-type test for trend.20,21
To generate a predictive strategy based on genetic variation, recursive
RESULTS
partitioning analysis was performed using the Willows program.22 Redundant
associated SNPs in linkage disequilibrium were filtered, and recursive parti-
tioning was performed on a training set consisting of two thirds of the
Clinical Results
Myeloma-IX data set to create an initial predictive tree; this tree was pruned at
The clinical characteristics and demographics of the
level of P
.001 to a smaller number of classifier SNPs. The final predictive tree
Myeloma-IX trial patients across the whole trial (Table 1) were com-
Table 1. Clinical Characteristics and Demographics of Patients in the Myeloma-IX study
Myeloma-IX Trial Arms
CVAD (n
548)
CTD (n
550)
MP (n
420)
CTDa (n
420)
Total (N
1,938)
Demographic or Clinical
No. of
No. of
No. of
No. of
No. of
Characteristic
Patients
%
Patients
%
Patients
%
Patients
%
Patients
%
Neuropathy
Any neuropathy
108
19.7
185
33.6
27
6.4
126
30.0
446
23.0
Sensory neuropathy
96
17.5
162
29.5
24
5.7
101
24.0
383
19.8
Motor neuropathy
24
4.4
46
8.4
11
2.6
51
12.1
132
6.8
Sensory/motor neuropathy
12
2.2
23
4.2
8
1.9
25
6.0
68
3.5
Whole trial, grade 2
Any neuropathy
35
6.4
51
9.3
9
2.1
57
13.6
152
7.8
Sensory neuropathy
28
5.1
44
8.0
7
1.7
43
10.2
122
6.3
Motor neuropathy
17
3.1
21
3.8
5
0.1
36
8.6
79
4.1
Whole trial, grade 3
Any neuropathy
14
2.6
16
2.9
5
0.1
19
4.5
54
2.7
Sensory neuropathy
10
1.8
10
1.8
2
0.05
11
2.6
33
1.7
Motor neuropathy
8
1.4
11
2.0
5
0.1
17
4.0
41
2.1
Sex
Male
348
344
231
242
1,165
Female
208
211
192
184
795
Median age, years
58.5
59
73
73
65
Median time to neuropathy, weeks
8
8
6
8
NA
ISS stage
I
110
138
53
46
347
II
161
149
120
128
558
III
134
125
121
118
498
Missing
151
143
129
134
557
Paraprotein
IgG
307
312
240
233
1,092
IgA
104
111
96
91
402
IgM
3
2
1
2
8
IgD
10
12
2
8
32
Light chain only
67
63
45
50
225
No paraprotein
11
9
4
6
30
Missing data
54
46
35
36
171
FISH, No./total No.
IGH translocation
136 of 305
149 of 316
88 of 210
90 of 222
463 of 1,053
Hyperdiploidy
161 of 290
159 of 294
127 of 193
121 of 213
568 of 990
t4;14
35 of 303
41 of 312
21 of 209
23 of 224
120 of 1,048
t11;14
46 of 302
46 of 311
24 of 209
30 of 222
146 of 1,044
del1p321
25 of 255
29 of 257
21 of 171
23 of 178
98 of 861
gain1q21
102 of 268
99 of 266
76 of 181
77 of 190
324 of 905
Abbreviations: CVAD, cyclophosphamide, vincristine, doxorubicin, and dexamethasone; CTD, cyclophosphamide, thalidomide, and dexamethasone; MP, melphalan
and prednisolone; CTDa, attenuated cyclophosphamide, thalidomide, and dexamethasone; NA, not applicable; ISS, International Staging System; Ig, immunoglob-
ulin; FISH, fluorescent in situ hybridization.
www.jco.org
© 2011 by American Society of Clinical Oncology
799
Information downloaded from jco.ascopubs.org and provided by at CEDARS SINAI MEDICAL CENTER MEDICAL LIBRARY
Copyright © 2011 on March
American 1, 2011
Society from
of
192.231.86.98
Clinical Oncology. All rights reserved.

---

Johnson et al
parable to those of the patients in the case-control comparison (Data
Supplement) and included 970 patients treated with CTD and 550
A
Response to treatment
120
patients treated with cyclophosphamide, vincristine, doxorubicin,
No change
and dexamethasone (Fig 1B). In this analysis, 31.8% of patients devel-
Minimal response
100
Partial response
oped neuropathy (grade
1, 31.8%; grade
2, 11%; and grade
3,
Very Good Partial response
3.6%) after exposure to thalidomide. The median time to neuropathy
Complete response
80
was 8 weeks. A higher frequency of grade
3 neuropathy was seen in
older patients (5%). In patients exposed to vincristine, 33.6% devel-
Patients 60
oped neuropathy in a median of 8 weeks. In contrast, 6.4% patients
of
not exposed to thalidomide or vincristine developed grade 1 neurop-
No. 40
athy in a median of 6 weeks (Data Supplement). We did not find
significant differences between patients with and without neuropathy
20
for previously described risk factors such as immunoglobulin type and
0
age.7 Assessing the rates of TrPN based on the ISS stage, we found that
Non-Neuropathy
Thalidomide-Related
tumor burden is not related to risk. Similarly, because fluorescent in
Controls
Neuropathy Cases
situ hybridization variants are distributed evenly between patients
B
with and without neuropathy, the pathogenic subtype does not seem
1.2
to contribute to risk. Interestingly, we identified increased rates in men
compared with women, the cause for which is uncertain. To address
1.0
the hypothesis that patients who respond best to thalidomide are more
sensitive to TrPN, we examined the relationship of neuropathy to
0.8
(proportion)
response during induction in the Myeloma-IX study. Although we
saw a greater percentage of complete responses and very good partial
0.6
Cases
responses in patients lacking neuropathy, the difference did not reach
CR
0.4
statistical significance (Pearson 2 test, P
.55; Fig 2A). Additionally,
VPGR
MR
we did not see a significant relationship between response and time to
PR
0.2
neuropathy (Fig 2B). These clinical data show the variability of risk
NC
between patients and the lack of any simple clinical variants that can
Neuropathy
0
predict risk.
3
4
6
8
9
12
15
16
20
21
24
28
36
Time to Thalidomide-Related
Neuropathy (weeks)
Genotyping Results
To address the hypothesis that genetic variation mediates the risk
Fig 2. (A) Response status versus neuropathy during induction therapy in the
of TrPN, we compared the genetic contribution of patients who de-
Myeloma-IX study. (B) Time to neuropathy in relation to response during
veloped neuropathy with that of patients who did not develop neu-
induction therapy with thalidomide in the Myeloma-IX study. CR, complete
ropathy. The initial analysis included all patients with TrPN, and a
response; VGPR, very good partial response; MR, minimal response; PR, partial
response; NC, no change.
secondary analysis was carried out using only patients with neuropa-
thy of grade
2. SNPs significantly associated with TrPN (P
.01) in
the Myeloma-IX data set are listed in Table 2. The associated SNPs
gene reveals a four-SNP haplotype, with three SNPs additively
were similar in grade
1 and grade
2 analyses, indicating a com-
contributing to risk of TrPN (Data Supplement).
mon mechanism for thalidomide neuropathy across the groups.
To investigate whether the observed associations with thalido-
Findings from the Myeloma-IX study were validated in a second
mide were drug specific, we sought to contrast findings with a parallel
comparable study, the HOVON-50/GMMG-HD3 trial, and signifi-
study in patients treated with vincristine. Significantly associated SNPs
cantly associated SNPs in this study are provided in the Data Supple-
associated with vincristine neuropathy in the Myeloma-IX data set are
ment. To cross validate the findings between the studies, we
listed in Table 4 and the Data Supplement.
investigated whether a significantly associated SNP in either study
Using inherited genetic variation as a clinical means of risk strat-
alone (permutated P
.05) was associated with risk in the second
ification to adjust thalidomide dose is an important clinical goal, and
study. This included 103 SNPs from the Myeloma-IX study and 82
in an effort to build a predictive model based on a limited number of
SNPs from HOVON-50/GMMG-HD3. Using this approach on TrPN
predictive SNPs, we examined two classification methods. In the first
grade
2, we found five cross-validating SNPs in different genes
method, we used recursive partitioning to create a predictive tree. This
(Table 3). It should be noted, however, that the size of the validation
tree predicted patient cases and controls with a sensitivity of 38% and
set may be underpowered to validate all associations in the discovery
a specificity of 31% in the Myeloma-IX validation set and a sensitivity
set. Associated regulatory region SNPs were analyzed for a relationship
of 30.7% and a specificity of 81.6% in the HOVON-50/GMMG-HD3
with gene expression (Data Supplement).
data set (Data Supplement). The second method of cumulative risk
Haplotype analysis (see haplotype-tagging SNPs)of ABCC1
score predicted patient cases of neuropathy with a sensitivity of 100%
variants with TrPN reveal that the main effect can be attributed to
and a specificity of 77.8% in the Myeloma-IX validation set and
a two-SNP haplotype involving SNPs rs246217 and rs246218 (Data
sensitivity of 60% and a specificity of 70.5% in the HOVON-50/
Supplement). Analysis of the 12 genotyped SNPs in the ABCC2
GMMG-HD3 data set (Data Supplement).
800
© 2011 by American Society of Clinical Oncology
JOURNAL OF CLINICAL ONCOLOGY
Information downloaded from jco.ascopubs.org and provided by at CEDARS SINAI MEDICAL CENTER MEDICAL LIBRARY
Copyright © 2011 on March
American 1, 2011
Society from
of
192.231.86.98
Clinical Oncology. All rights reserved.

---

Thalidomide-Related Neuropathy in Patients With Myeloma
Table 2. SNP Association With Thalidomide-Related Neuropathy (grade
2) in the Myeloma-IX Study Using a Trend Test for Association (permutated P
.01)
Fisher's Exact Odds Ratio
SNP
Chromosome
Allele
Odds Ratio
95% CI
Permutated P for Trend
Gene
SNP Type
rs246220
16
G C
0.49
0.33 to 0.73
.001
ABCC1
Intron, TagSNP:ABCC1
rs7164902
15
A G
0.60
0.44 to 0.80
.001
SLC12A6
Coding-synon
rs1805386
13
G A
0.56
0.39 to 0.79
.001
LIG4
Coding-synon
rs3740066
10
T C
1.51
1.18 to 1.93
.001
ABCC2
Coding-synon
rs4752904
11
G C
0.69
0.54 to 0.88
.002
PTPRJ
Coding-nonsynon
rs8192341
8
A C
1.48
1.14 to 1.93
.003
SFTPC
Coding-nonsynon
rs2292334
6
A G
0.71
0.55 to 0.91
.004
SLC22A3
Coding-synon
rs246
11
A G
1.54
1.15 to 2.07
.005
CYP2C9
Intron
rs4839469
1
A G
1.56
1.13 to 2.15
.005
VANGL1
Coding-nonsynon
rs3735481
7
A C
1.50
1.12 to 2.00
.006
PPIA
Intron
rs2272037
15
T C
1.40
1.10 to 1.78
.007
IGF1R
Intron
rs231775
2
G A
1.38
1.08 to 1.77
.007
CTLA4
Coding-nonsynon
rs914959
1
T C
0.70
0.54 to 0.90
.007
DPYD
Intron, TagSNP:DPYD
rs3136794
8
G A
1.62
1.15 to 2.29
.009
POLB
Intron
NOTE. The genomic inflation factor
is 1.0.
Abbreviation: SNP, single nucleotide polymorphism.
the group who develop neuropathy associated with prolonged
DISCUSSION
exposure, which is dependent on the total dose of thalidomide
received and was not studied here. We did not identify any simple
Using a hypothesis-driven candidate gene approach, we show for the
clinical parameters that were associated with early-onset neuropa-
first time that there is a significant genetic contribution to the risk of
thy, and patients were not unduly sensitive to thalidomide, with
TrPN. Thalidomide is used in the treatment of a number of diseases
response rates being similar in patients and controls. We cannot be
where the increased production of TNF- is thought to be pathogen-
certain about associations with other comorbid conditions such as
ically important, but the benefits of thalidomide can be limited by the
diabetes, diet, and alcohol consumption because the data were not
onset of TrPN. This study, one of the largest pharmacogenomics
studies of its kind to date,22a,22b has allowed us to identify genetic
systematically collected. Patients with pre-existing neuropathy
variants that contribute to TrPN risk. These findings provide impor-
were excluded.
tant insights into the therapeutic management of these patients and
A number of ATP-binding cassette (ABC) transporter genes
will stimulate further research into protective strategies for patients
were linked with TrPN. Both ABCC1 SNPs rs246217 and rs246218
treated with thalidomide. Although we considered taking a genome-
are intronic but lie within transcription binding domains and may
wide approach to identify such genetic variation, it was not applicable
mediate levels of expression of ABCC1. Two SNPs in ABCC2
because of issues with statistical power. In addition, the candidate gene
(rs3740066 and ­24 C T rs717620) are also associated with neu-
approach allowed us to examine our candidate regions in greater detail
ropathy risk. rs717620 has been associated with decreased ABCC2
than could be achieved with mapping arrays such as the Affymetrix
function in vitro23-25 and toxicity in other systems,26 whereas
500k (Affymetrix) or Illumina 550k (Illumina, San Diego, CA).
rs3740066 may modulate substrate specificity via codon usage,
The doses of thalidomide used in the study were moderate at
therefore influencing the translation rate.27 Also, in ABCA1,
100 to 200 mg, and the neuropathy occurred early, after a median
rs363717 is a cross-validating SNP and is found in the binding
of three cycles of treatment. This group of patients is distinct from
domain of the microRNA hsa-miR-299. Weaker associations were
Table 3. SNP Association With Thalidomide-Related Neuropathy (grade
2) Associated in Both the Myeloma-IX and HOVON-50/GMMG-HD3 Studies Using a
Trend Test for Association (permutated P
.05)
Myeloma-IX Trial
HOVON-50/GMMG-HD3
Fisher's Exact
Permutated P
Fisher's Exact
Permutated P
SNP
Chromosome
Allele
Odds Ratio
95% CI
for Trend
Odds Ratio
95% CI
for Trend
Gene
SNP Type
rs7164902
15
A G
0.60
0.44 to 0.80
.001
0.47
0.25 to 0.87
.023
SLC12A6
Coding-synon
rs6103
18
G C
0.70
0.52 to 0.95
.018
0.56
0.30 to 1.07
.054
SERPINB2
Coding-nonsynon
rs2076169
6
G A
0.60
0.38 to 0.95
.026
0.27
0.08 to 0.90
.025
PPARD
Intron
rs363717
9
C T
0.71
0.52 to 0.98
.041
0.46
0.22 to 0.97
.045
ABCA1
Untranslated
rs1799969
19
A G
0.67
0.44 to 1.03
.050
0.40
0.15 to 1.04
.046
ICAM1
Coding-nonsynon
Abbreviations: SNP, single nucleotide polymorphism; HOVON, Dutch-Belgian Hemato-Oncology Cooperative Group; GMMG, German-Speaking Myeloma
Multicenter Group.
www.jco.org
© 2011 by American Society of Clinical Oncology
801
Information downloaded from jco.ascopubs.org and provided by at CEDARS SINAI MEDICAL CENTER MEDICAL LIBRARY
Copyright © 2011 on March
American 1, 2011
Society from
of
192.231.86.98
Clinical Oncology. All rights reserved.

---

Johnson et al
Table 4. SNP Association With Vincristine-Related Neuropathy (grade
2) Associated in Both the Myeloma-IX and HOVON-50/GMMG-HD3 Studies
Using a Trend Test for Association (permutated P
.05)
Myeloma-IX
HOVON-50/GMMG-HD3
Fisher's Exact
Permutated P
Fisher's Exact
Permutated P
SNP
Chromosome
Allele
Odds Ratio
95% CI
for Trend
Odds Ratio
95% CI
for Trend
Gene
SNP Type
rs7242
7
G T
1.88
1.17 to 3.01
.006
2.15
1.15 to 4.00
.015
SERPINE1
Untranslated
rs2082382
5
G A
0.50
0.30 to 0.83
.009
0.52
0.27 to 1.02
.051
ADRB2
Promoter
rs1555026
1
C T
2.73
1.29 to 5.76
.009
2.45
0.95 to 6.30
.049
ID3
Locus
rs1042714
5
G C
0.51
0.31 to 0.85
.010
0.52
0.27 to 1.01
.047
ADRB2
Coding-nonsynon
rs1934951
10
T C
1.90
1.14 to 3.16
.012
2.36
1.20 to 4.65
.016
CYP2C9
Intron
rs7214723
17
C T
1.70
1.06 to 2.71
.020
1.90
1.03 to 3.51
.038
CAMKK1
Coding-nonsynon
rs1058932
10
A G
1.74
1.04 to 2.89
.027
2.18
1.11 to 4.30
.024
CYP2C8
Untranslated
rs2301157
13
A G
0.61
0.38 to 1.00
.039
0.52
0.27 to 0.98
.035
SLC10A2
Untranslated
rs228832
20
T C
0.51
0.27 to 0.96
.040
0.39
0.16 to 0.94
.030
NFATC2
Intron
Abbreviations: SNP, single nucleotide polymorphism; HOVON, Dutch-Belgian Hemato-Oncology Cooperative Group; GMMG, German-Speaking Myeloma
Multicenter Group.
also seen with ABCB1, and its role in mediating peripheral neurop-
nel activity may prove useful in reducing neurotoxicity caused by
athy in response to taxane exposure has been previously
axonal membrane ion channel variation.39,40
reported.28-29 Interestingly, thalidomide can modulate the func-
The genetic mechanisms underlying thalidomide and vincristine
tion of ABCC2 and ABCB1,30 both of which are active in neuro-
neuropathy seem distinct. SNP associations with vincristine neurop-
nal function.
athy include rs2082382 and rs1042714 in ADRB2. Association with
We saw a number of associations with other ADME (absorption,
vincristine neuropathy was observed in a highly conserved promoter
distribution, metabolism, excretion) genes, including a promoter
SNP (rs7214723) in CAMKK1, which is expressed at higher levels in
SNP in FMO6 (rs1736565) and the ion channel gene SLC12A6
neurons resistant to oxidative stress. We also found an association
(rs7614902), previously associated with peripheral neuropathy.31 It
with nonsynonymous SNPs in CYP2C9 (rs1934951) and NFATC2
has been suggested that CYP2C19 plays a role in thalidomide metab-
(rs228832), conserved intronic SNPs in ID3 (rs1555026) and
olism32-34; however, it is now known that thalidomide breakdown is
SLC10A2 (rs2301157), and in CYP2C8 (rs1058932), which is located
dependent on spontaneous nonenzymatic degradation.35 Consistent
in the binding site for the microRNA hsa-miR-96. Interestingly, we
with this, we found no association of neuropathy risk with functional
saw little overlap of the genes associated with TrPN and those associ-
variants in CYP2C19 or other genes in the CYP2C subfamily. No
ated with vincristine-related neuropathy. We propose that this is in-
consistent association was seen with common N-acetyltransferase 2
dicative of a fundamentally different pathologic mechanism between
(NAT2) variants, as previously reported.13
these neuropathies. In contrast, similar TrPN-associated genes have
Pathway analysis of the TrPN-associated SNPs highlights the
been seen in studies investigating peripheral neuropathy in response
association with genes involved in the development of the central and
to paclitaxel, bortezomib, and docetaxel, which result in sensory pe-
peripheral nervous system. A conserved promoter SNP (rs1611753),
ripheral neuropathies as opposed to the sensorimotor neuropathy
which affects the expression of the gene SPRR1A, was significantly
associated with exposure to the vinca alkaloids and paclitaxel.
associated with TrPN in HOVON-50/GMMG-HD3 samples. The
We show that simple clinical factors do not allow the identifica-
SPRR1A gene is expressed by axotomized neurons and promotes
tion of patients at greater risk of neuropathy after thalidomide expo-
axonal outgrowth.36 Variation in neurologic genes may dictate the
sure, for whom dose adjustment or cessation of therapy would be
ability of a damaged neuron to undergo repair and may mediate risk
appropriate. Consequently, we have attempted to define a limited
of neuropathy.
number of SNPs that can identify high-risk patients suitable for more
We also describe associations with SNPs in genes mediating
intensive clinical monitoring. In this approach, we have used two risk
neuro-inflammation, with approximately 35% of the TrPN-
classification methods, both of which can identify patients at increased
associated SNPs having an inflammatory role. We found associations
risk, but the predictive value of these is not adequate to totally avoid
with SERPINB2 (rs6103), a gene induced in injured neurons,37 and
thalidomide usage. However, this approach can be used to identify
PPARD (rs2076169), which may also indicate a role of neuro-
patients requiring greater clinical vigilance and suitable counseling.
inflammation in the pathology mechanism.
The poor risk prediction in our classifiers is a result of a
We found 20% of the TrPN-associated genes in this study to
number of limitations, including naive assumptions in modeling
have some transmembrane transporter activity, consistent with a
methods (the risk score method assumes interactions to be solely
hypothesis where variants in transmembrane transporter genes
additive, whereas recursive partitioning method interactions are
mediate their effects at the level of the peripheral nerve.38 At this
assumed to be unidirectional); the fact that hypothesis-driven ap-
site, they negatively impact on the ability of the dorsal root gangli-
proaches do not consider all variation contributing to an outcome;
ons to repair themselves after exposure to the toxic effects of
population-specific effects; and the impact of rare variation. De-
thalidomide and/or promote neuro-inflammatory change. Pro-
spite these drawbacks, the limited number of SNPs identified in
phylactic pharmacologic therapies aimed at modulating ion chan-
this study can be used as a simple and useful method for identifying
802
© 2011 by American Society of Clinical Oncology
JOURNAL OF CLINICAL ONCOLOGY
Information downloaded from jco.ascopubs.org and provided by at CEDARS SINAI MEDICAL CENTER MEDICAL LIBRARY
Copyright © 2011 on March
American 1, 2011
Society from
of
192.231.86.98
Clinical Oncology. All rights reserved.

---

Thalidomide-Related Neuropathy in Patients With Myeloma
patients at high risk of TrPN, who in turn may benefit from greater
AUTHOR CONTRIBUTIONS
clinical vigilance.
Conception and design: David C. Johnson, Faith E. Davies, Brian G.M.
AUTHORS' DISCLOSURES OF POTENTIAL CONFLICTS
Durie, Brian Van Ness, Pieter Sonneveld, Gareth J. Morgan
OF INTEREST
Financial support: Brian G.M. Durie
Provision of study materials or patients: Henk M. Lokhorst, Hartmut
Although all authors completed the disclosure declaration, the following
Goldschmidt, J. Anthony Child, Pieter Sonneveld
author(s) indicated a financial or other interest that is relevant to the subject
Collection and assembly of data: David C. Johnson, Sophie L.
matter under consideration in this article. Certain relationships marked
Corthals, Brian A. Walker, Fiona M. Ross, Walter M. Gregory,
with a "U" are those for which no compensation was received; those
Nicholas J. Dickens
relationships marked with a "C" were compensated. For a detailed
description of the disclosure categories, or for more information about
Data analysis and interpretation: David C. Johnson, Sophie L. Corthals,
ASCO's conflict of interest policy, please refer to the Author Disclosure
Brian A. Walker
Declaration and the Disclosures of Potential Conflicts of Interest section in
Manuscript writing: David C. Johnson, Faith E. Davies,
Information for Contributors.
Gareth J. Morgan
Employment or Leadership Position: None Consultant or Advisory
Final approval of manuscript: David C. Johnson, Sophie L. Corthals,
Role: Brian G.M. Durie, Celgene (C), Millennium Pharmaceuticals (C);
Brian A. Walker, Fiona M. Ross, Walter M. Gregory, Nicholas J. Dickens,
Pieter Sonneveld, Celgene (C) Stock Ownership: None Honoraria: Faith
Henk M. Lokhorst, Hartmut Goldschmidt, Faith E. Davies, Brian G.M.
E. Davies, Celgene Research Funding: None Expert Testimony: None
Durie, Brian Van Ness, J. Anthony Child, Pieter Sonneveld,
Other Remuneration: None
Gareth J. Morgan
13. Harland CC, Steventon GB, Marsden JR:
on mRNA and protein expression in normal and
REFERENCES
Thalidomide-induced neuropathy and genetic differ-
cancerous kidney cortex. Pharmacogenomics J
ences in drug metabolism. Eur J Clin Pharmacol
7:56-65, 2007
1. Chaudhry V, Cornblath DR, Corse A, et al:
49:1-6, 1995
24. Hirouchi M, Suzuki H, Itoda M, et al: Charac-
Thalidomide-induced neuropathy. Neurology 59:
14. Van Ness B, Ramos C, Haznadar M, et al:
terization of the cellular localization, expression
1872-1875, 2002
Genomic variation in myeloma: Design, content, and
level, and function of SNP variants of MRP2/ABCC2.
2. Zomas A, Anagnostopoulos N, Dimopoulos
initial application of the Bank On A Cure SNP Panel
Pharm Res 21:742-748, 2004
MA: Successful treatment of multiple myeloma
to detect associations with progression-free sur-
25. Itoda M, Saito Y, Soyama A, et al: Polymor-
relapsing after high-dose therapy and autologous
vival. BMC Med 6:26, 2008
phisms in the ABCC2 (cMOAT/MRP2) gene found in
transplantation with thalidomide as a single agent.
15. Johnson DC, Corthals S, Ramos C, et al:
72 established cell lines derived from Japanese
Bone Marrow Transplant 25:1319-1320, 2000
Genetic associations with thalidomide mediated ve-
individuals: An association between single nucle-
3. Cavaletti G, Beronio A, Reni L, et al: Thalido-
nous thrombotic events in myeloma identified using
otide polymorphisms in the 5 -untranslated region
mide sensory neurotoxicity: A clinical and neuro-
targeted genotyping. Blood 112:4924-4934, 2008
and exon 28. Drug Metab Dispos 30:363-364,
physiologic study. Neurology 62:2291-2293, 2004
16. Durie BG, Van Ness B, Ramos C, et al: Ge-
2002
4. Mileshkin L, Prince HM: The troublesome
netic polymorphisms of EPHX1, Gsk3beta, TNFSF8
26. Choi JH, Ahn BM, Yi J, et al: MRP2 haplo-
toxicity of peripheral neuropathy with thalidomide.
and myeloma cell DKK-1 expression linked to bone
types confer differential susceptibility to toxic liver
Leuk Lymphoma 47:2276-2279, 2006
disease in myeloma. Leukemia 23:1913-1919, 2009
injury. Pharmacogenet Genomics 17:403-415, 2007
5. Isoardo G, Bergui M, Durelli L, et al: Thalido-
17. Purcell S, Neale B, Todd-Brown K, et al:
27. Kimchi-Sarfaty C, Oh JM, Kim IW, et al: A
mide neuropathy: Clinical, electrophysiological and
PLINK: A tool set for whole-genome association and
"silent" polymorphism in the MDR1 gene changes
neuroradiological features. Acta Neurol Scand 109:
population-based linkage analyses. Am J Hum
substrate specificity. Science 315:525-528, 2007
188-193, 2004
Genet 81:559-575, 2007
28. Sissung TM, Baum CE, Deeken J, et al:
6. Briani C, Zara G, Rondinone R, et al: Thalido-
18. Dennis G Jr, Sherman BT, Hosack DA, et al:
ABCB1 genetic variation influences the toxicity and
mide neurotoxicity: Prospective study in patients with
DAVID: Database for Annotation, Visualization, and
clinical
outcome
of
patients
with
androgen-
lupus erythematosus. Neurology 62:2288-2290, 2004
Integrated Discovery. Genome Biol 4:P3, 2003
independent prostate cancer treated with docetaxel.
7. Tseng S, Pak G, Washenik K, et al: Rediscov-
19. Su AI, Wiltshire T, Batalov S, et al: A gene
Clin Cancer Res 14:4543-4549, 2008
ering thalidomide: A review of its mechanism of
atlas of the mouse and human protein-encoding
29. Sissung TM, Mross K, Steinberg SM, et al:
action, side effects, and potential uses. J Am Acad
transcriptomes. Proc Natl Acad SciUSA 101:6062-
Association of ABCB1 genotypes with paclitaxel-
Dermatol 35:969-979, 1996
6067, 2004
mediated peripheral neuropathy and neutropenia.
8. Beckmann MJ, Hussein MA, Lichtin A, et al:
20. Walker BA, Leone PE, Jenner MW, et al:
Eur J Cancer 42:2893-2896, 2006
Low serum vitamin B12 in patients with plasma cell
Integration of global SNP-based mapping and ex-
myeloma is associated with plasma cell myeloma
30. Yang XX, Hu ZP, Xu AL, et al: A mechanistic
pression arrays reveals key regions, mechanisms,
with true functional cobalamin deficiency. Am J Clin
and genes important in the pathogenesis of multiple
study on reduced toxicity of irinotecan by coad-
Pathol 140:350, 1995
myeloma. Blood 108:1733-1743, 2006
ministered thalidomide, a tumor necrosis factor-
9. Clark TE, Edom N, Larson J, et al: Thalomid
21. Cheung VG, Spielman RS, Ewens KG, et al:
alpha inhibitor. J Pharmacol Exp Ther 319:82-104,
(thalidomide) capsules: A review of the first 18
Mapping determinants of human gene expression
2006
months of spontaneous postmarketing adverse
by regional and genome-wide association. Nature
31. Howard HC, Mount DB, Rochefort D, et al:
event surveillance, including off-label prescribing.
437:1365-1369, 2005
The K-Cl cotransporter KCC3 is mutant in a severe
Drug Saf 24:87-117, 2001
22. Zhang H, Wang M, Chen X: Willows: A mem-
peripheral neuropathy associated with agenesis of
10. Giannini F, Volpi N, Rossi S, et al: Thalidomide-
ory efficient tree and forest construction package.
the corpus callosum. Nat Genet 32:384-392, 2002
induced neuropathy: A ganglionopathy? Neurology
BMC Bioinformatics 10:130, 2009
32. Ando Y, Price DK, Dahut WL, et al: Pharma-
60:877-878, 2003
22a. Hirschhorn JN, Lohmueller K, Byrne E, et al:
cogenetic associations of CYP2C19 genotype with
11. Rao DG, Kane NM, Oware A: Thalidomide
A comprehensive review of genetic association
in vivo metabolisms and pharmacological effects of
neuropathy: Role of F-wave monitoring. Muscle
studies. Genet Med 4:45-61, 2002
thalidomide. Cancer Biol Ther 1:669-673, 2002
Nerve 23:1301-1302, 2000
22b. Daly AK: Genome-wide association studies
33. Ando Y, Fuse E, Figg WD: Thalidomide me-
12. Hess CW, Hunziker T, Ku¨pfer A, et al:
in pharmacogenomics. Nat Rev Genet 11:241-246,
tabolism by the CYP2C subfamily. Clin Cancer Res
Thalidomide-induced peripheral neuropathy: A pro-
2010
8:1964-1973, 2002
spective clinical, neurophysiological and pharmaco-
23. Haenisch S, Zimmermann U, Dazert E, et al:
34. Li Y, Hou J, Jiang H, et al: Polymorphisms of
genetic evaluation. J Neurol 233:83-89, 1986
Influence of polymorphisms of ABCB1 and ABCC2
CYP2C19 gene are associated with the efficacy of
www.jco.org
© 2011 by American Society of Clinical Oncology
803
Information downloaded from jco.ascopubs.org and provided by at CEDARS SINAI MEDICAL CENTER MEDICAL LIBRARY
Copyright © 2011 on March
American 1, 2011
Society from
of
192.231.86.98
Clinical Oncology. All rights reserved.

---

Johnson et al
thalidomide based regimens in multiple myeloma.
37. Yamanaka H, Obata K, Fukuoka T, et al: Induc-
39. Argyriou AA, Polychronopoulos P, Iconomou G,
Haematologica 92:1246-1249, 2007
tion of plasminogen activator inhibitor-1 and -2 in
et al: A review on oxaliplatin-induced peripheral
35. Lu J, Helsby N, Palmer BD, et al: Metabo-
dorsal root ganglion neurons after peripheral nerve
nerve damage. Cancer Treat Rev 34:368-377,
lism of thalidomide in liver microsomes of mice,
injury. Neuroscience 132:183-191, 2005
2008
rabbits, and humans. J Pharmacol Exp Ther 310:
38. Kodama T, Abe M, Iida S, et al: A pharmaco-
40. von Delius S, Eckel F, Wagenpfeil S, et al:
571-577, 2004
kinetic study evaluating the relationship between
Carbamazepine for prevention of oxaliplatin-related
36. Bonilla IE, Tanabe K, Strittmatter SM: Small
treatment efficacy and incidence of adverse events
neurotoxicity in patients with advanced colorectal
proline-rich repeat protein 1A is expressed by axo-
with thalidomide plasma concentrations in patients
cancer: Final results of a randomised, controlled,
tomized neurons and promotes axonal outgrowth.
with refractory multiple myeloma. Clin Lymphoma
multicenter phase II study. Invest New Drugs 25:
J Neurosci 22:1303-1315, 2002
Myeloma 9:154-159, 2009
173-180, 2007

Glossary Terms
ADME (absorption, distribution, metabolism, ex-
generation and in which genotype frequencies are a product of allele
cretion): Gene pathways important in the absorption, distribu-
frequencies. A randomly mating population tends toward a Hardy-
tion, metabolism, and excretion of a pharmaceutical compound
Weinberg equilibrium state if there are no mutations, migrations, or
within an organism. These genes influence the drug levels and
environmental factors favoring particular genotypes.
kinetics of drug exposure to tissues and can influence the perfor-
mance and pharmacologic activity of the compound as a drug.
Pharmacogenomics: The study of how a person's genome can
affect their reaction to medications.
Allele: an alternative form of a gene (in diploids, one member
of a pair) that is located at a specific position on a specific
Recursive partitioning: Multivariable analysis that generates a
chromosome.
clinically intuitive decision tree model in which the population is di-
vided into prognostic subgroups. This is achieved through multiple di-
FISH (fluorescent in situ hybridization): In situ hy-
chotomous divisions on the basis of a set of independent variables.
dridization is a sensitive method that is generally used to detect
specific gene sequences in tissue sections or cell preparations by
Risk score: A simplified version of a prognostic model, where scores
hybridizing the complementary strand of a nucleotide probe to
are assigned to each risk factor (eg, based on rounded regression
the sequence of interest. FISH uses a fluorescent probe to in-
coefficients).
crease the sensitivity of in situ hybridization.
SNP (single nucleotide polymorphism): Genetic poly-
Haplotype-tagging SNPs: Markers that are statistically
morphisms are natural variations in the genomic DNA sequence
associated with and are therefore considered to be representative
present in greater than 1% of the population, with SNP representing
of a set of SNPs within a particular region. Using this technique,
DNA variations in a single nucleotide. SNPs are being widely used to
the identification of a few alleles of a haplotype block allows for
better understand disease processes, thereby paving the way for
the identification all other polymorphic sites in this region of
genetic-based diagnostics and therapeutics.
linkage disequilibrium.
Validation set: Samples used in evaluating performance of a clas-
Hardy-Weinberg equilibrium: A state in which genotype
sifier. The validation set is formed by the units not used in develop-
frequencies and ratios remain constant from generation to
ing the classifier (ie, the training set and test set).
804
© 2011 by American Society of Clinical Oncology
JOURNAL OF CLINICAL ONCOLOGY
Information downloaded from jco.ascopubs.org and provided by at CEDARS SINAI MEDICAL CENTER MEDICAL LIBRARY
Copyright © 2011 on March
American 1, 2011
Society from
of
192.231.86.98
Clinical Oncology. All rights reserved.

---

Genetic factors underlying the risk of thalidomide related neuropathy in multiple
myeloma patients
Johnson et al
Supplementary table 1: Clinical demographics of neuropathy cases and controls in the
Myeloma-IX study.
© 2010 by American Society of Clinical Oncology
- 1 -

---

Supplementary table 2: SNP association with thalidomide related neuropathy (grades
2) in the HOVON-50/GMMG-HD3 study, using a Trend test for
association. Permutated P < 0.01. The genomic inflation factor is 1.023
© 2010 by American Society of Clinical Oncology
- 2 -

---

Supplementary table 3: SNP association with vincristine related neuropathy (grades
2) in the Myeloma IX study, using a Trend test for association.
Permutated P < 0.01. The genomic inflation factor is 1.076.
© 2010 by American Society of Clinical Oncology
- 3 -

---

Supplementary table 4: SNP association with vincristine related neuropathy (grades
2) in the HOVON-50/GMMG-HD3 study, using a Trend test for
association. Permutated P < 0.01. The genomic inflation factor is 1.046.
© 2010 by American Society of Clinical Oncology
- 4 -

---

Supplementary figure 1: Expression of associated genes related to genotype cal in the SNP. We
examined predicted regulatory SNPs where the median absolute expression probe intensity >100 and
MAF >0.05. Four SNPs achieved a p<0.1 when log2 (expression) was tested by using a Wilcoxon-type
test for trend with genotype call.
© 2010 by American Society of Clinical Oncology
- 5 -

---

Supplementary figure 2: a) p-values for BOAC genotyped SNPs across the gene ABCC1 for association with
thalidomide related neuropathy. b) Lnkage disequilibrium and recombination rate in the gene ABCC1 around
SNP rs246220. Figures were generated using the web-based tool SNAP. (Johnson, A. D, 2008)
© 2010 by American Society of Clinical Oncology
- 6 -

---

Supplementary figure 3: a) p-values for BOAC genotyped SNPs across the gene ABCC2 for association with
thalidomide related neuropathy. b) Lnkage disequilibrium and recombination rate in the gene ABCC2 around
SNP rs3740066.
© 2010 by American Society of Clinical Oncology
- 7 -

---

Supplementary figure 4: Recursive partitioning tree of the training set from Myeloma IX cases and controls stratified by thalidomide related neuropathy status. Tree
determined in the Willows program. Terminal nodes of increase risk are outlined in bold and nodes with decreased risk outlined with a broken line.
© 2010 by American Society of Clinical Oncology
- 8 -

---

Supplementary figure 5: Jitter plot of thalidomide related neuropathy risk score in the Risk score (SCORE) of developing thalidomide related neuropathy in myeloma patients
against observed phenotype (cases or controls).The cumulative risk TrPN risk score for an individual patient is calculated by the taking the sum of the Log10 (Odds ratio of
significant associated SNPs with thalidomide neuropathy) multiplied by the number of significant associated alleles per patient by multiplied by a transformation constant (104).
© 2010 by American Society of Clinical Oncology
- 9 -

---

Supplementary figure 6: Genes of associated SNPs with thalidomide related thalidomide analysed for known
physical and functional protein-protein interaction by the EMBL database STRING 8.1. (Jensen L, 2009)
© 2010 by American Society of Clinical Oncology
- 10 -

---