SUPPLEMENTARY MATERIAL
For manuscript “Increased stomach cancer risk following radiotherapy for testicular cancer” by
Hauptmann et al.
PATIENT SELECTION
Of the 111 observed cases of second primary stomach cancer in the population-based cohort, medical
records could not be obtained for 41 cases, including the only observed case from Iowa
(Supplementary Table 1). Further, we did not include 18 cases and 36 controls from the extension of
the Norwegian study population to 1,713 TC survivors who received primary treatment in a hospital
other than the Radium Hospital, based on the fact that radiation doses could not be estimated for a
large fraction of patients treated with radiotherapy due to insufficient information for dose
reconstruction in the medical records (33% versus 4% among all other cases and controls). Compared
with all other patients, stomach cancer cases from Norway treated in a hospital other than the Radium
Hospital received radiotherapy substantially less often than controls [odds ratio (OR)=0.25], which is
in contrast with patients from all other remaining centers (OR=5.9). Also, for those whose doses could
be reconstructed, stomach cancer cases had on average lower doses than controls (22.9 Gy versus 24.0
Gy, respectively), whereas this was not the case among patients from the remaining study centers
(29.8 Gy versus 27.5 Gy, respectively). This resulted in a negative dose-response among patients from
the extension study in Norway (EOR/Gy= -0.01387, p=0.390), which was borderline significantly
different from the dose-response in the remaining patients (interaction p=0.069). The dose-response in
the total population of patients, including the extension study in Norway, was attenuated to an
EOR/Gy of 0.075 (95% CI, 0.014-0.27, p=0.003) with ORs for the dose categories in Table 2 of 1.7
(95% CI, 0.5-5.8), 1.1 (0.4-2.8), 3.6 (1.3-9.6), 3.3 (1.1-10.4), 10.5 (2.3-48.8), 2.1 (0.7-6.1)
(Supplementary Table 2).
SUPPLEMENTARY FIGURE: Illustration of the three stomach shapes used in the analyses, including
the typical J-shaped stomach used for all main analyses and two variant stomach shapes based on body
mass index (BMI) used for sensitivity analyses.
Typical stomach
Variant stomach #1
Variant stomach #2
Explanatory text for Supplementary Figure: Variant stomach #1 correlates with massive body build
and higher weight and variant stomach #2 correlates with thinner body build and lower weight (1). To
evaluate sensitivity of results with regard to stomach shape, we assigned shapes based on BMI. First,
we multiply imputed missing BMI for 114 of the 272 patients (42%) based on a Markov Chain Monte
Carlo method (2). We then assigned patients with a BMI above the 95th percentile of the observed 114
BMI values to the stomach shape that correlates with massive body build and higher weight (variant
stomach #1), and patients with a BMI below the 5th percentile to the stomach shape that correlates
with thinner body build and lower weight (variant stomach #2). For all others, we used the typical Jshaped stomach, which was used for all patients in the main analysis. Logistic regression on
categorical dose was then performed for each of the 5 imputed datasets and the resulting ORs and
standard errors were appropriately combined. SAS procedures MI and MIANALYSE were used. In
addition to using a 5% cutoff for stomach shape assignment, we also used 10%, 15%, 20% and 25%,
as well as absolute values of 20 and 30 kg/m2.
SUPPLEMENTARY TABLE 1: Exclusions among 111 eligible second primary stomach
cancer cases
Excluded* (%)
(N=41)
Included** (%)
(N=70)
Registry
Sweden
Denmark
Norway
Finland
Ontario
Iowa
1 (2.4)
36 (87.8)
0 (0.0)
3 (7.3)
0 (0.0)
1 (2.4)
20 (28.6)
20 (28.6)
18 (25.7)
7 (10.0)
5 (7.1)
0 (0.0)
Year of testicular cancer diagnosis
1959-1969
1970-1979
1980-1987
30 (73.2)
10 (24.4)
1 (2.4)
22 (31.4)
34 (48.6)
14 (20.0)
Age at testicular cancer diagnosis (years)
18-29
30-39
40-49
50-59
60-71
5 (12.2)
16 (39.0)
12 (29.3)
7 (17.1)
1 (2.4)
13 (18.6)
26 (37.1)
19 (27.1)
8 (11.4)
4 (5.7)
* Exclusion due to failure to obtain medical records was more common for patients
diagnosed before 1970 and therefore a larger percentage of cases could not be included
from registries which started earlier.
** Patients from the Dutch study are not included in this table since they were added to
the original study by pooling.
SUPPLEMENTARY TABLE 2: Radiation treatment-related risks for stomach cancer among patients
with testicular cancer and matched controls including 18 cases and 36 controls from Norway who
received primary treatment in a hospital other than the Radium Hospital
Any radiotherapy*
No
Yes
Unknown
Radiation dose (Gy)*
0-9.9
10.0-19.9
20.0-29.9
30.0-39.9
40.0-49.9
>=50.0†
Unknown
Number of
cases
Number of
controls
Odds
ratio
95% Confidence
intervals
8
101
2
33
181
1
1
2.7
2.1
Ref
1.1-6.7
0.2-28.4
21
9
18
30
12
8
56
16
55
43
22
6
1
1.7
1.1
3.6
3.3
10.5
Ref
0.5-5.8
0.4-2.8
1.3-9.6
1.1-10.4
2.3-48.8
12
18
2.1
0.003
0.7-6.1
p-trend‡
EOR/Gy=0.075 (95% CI: 0.014-0.27)
Abbreviations: Ref, reference; Gy, gray; EOR, excess odds ratio; CI, confidence interval
* Not adjusted for chemotherapy.
† Range: 50-59.1 Gy, median: 50.8 Gy.
‡ Based on continuous (linear) dose.
REFERENCES
1. Morton LM, Dores GM, Curtis RE, Lynch CF, Stovall M, Hall P, Gilbert ES, Hodgson DC, Storm
HH, Johannesen TB, Smith SA, Weathers RE, et al. Stomach cancer risk after treatment for
Hodgkin lymphoma. J Clin Oncol 2013; 31(27): 3369-3377.
2. Schafer JL. Analysis of Incomplete Multivariate Data, New York: Chapman and Hall 1997.