Page 136 - 2018_11-Haematologica-web

A. Nikkilä et al.
obtained the parameters used for dose assessment including year, body part, use of contrast medium and the number of sequences. Manufacturers and models of CT scanners in each hospital were acquired from the Radiation and Nuclear Safety Authority (STUK). For dose calculations, we assumed in the main analysis that each CT scan was performed using the latest CT scanner available at the hospital.
Data on a total of 80,783 pediatric CT scans were obtained and of those, 49 CT scans were performed on the study subjects, excluding the 2-year latency period (Table 1). Half (n=25) were head scans, and 19 were lung scans. Of the CT scans, 36 were per- formed on 15 (1.4%) cases and 13 scans on 10 (0.3%) controls.
The CT scan parameters were obtained based on expert opin- ion of an experienced hospital physicist (Online Supplementary Table S2). The doses were estimated using the NCICT software (v1.2).32 Age- and sex-specific pediatric software phantoms (for neonates, and children aged 1, 5, 10, and 15 years) were used. The input for dose calculation also included the scanner manu- facturer and model. If data were available only on the manufac- turer, a manufacturer-specific average was used. It was assumed that a head or body filter was used, based on the target body part. The cumulative absorbed red bone marrow (RBM) doses were obtained as the sums of absorbed RBM doses from all CT scans for each study subject. The dose from a scan was multi- plied by 1.5 if contrast medium was used, consistent with tissue- specific coefficients suggested for other tissues.33 Alternative dose estimates were obtained based on values reported in the lit- erature.34
We identified subjects with Down syndrome (40 cases and 2 controls) from the Congenital Malformation Register and Care Register for Health Care, and those with a previous malignancy (2 cases) from the Finnish Cancer Registry. They were excluded to avoid confounding by indication (reverse causation). We also collected information on birth weight (large for gestational age) and maternal smoking during pregnancy from the Medical Birth Register, as well as socioeconomic status and education of the parents from Statistics Finland. Residential exposure to back- ground gamma radiation, including natural terrestrial radiation and Chernobyl fallout, was estimated as described previously.8
Due to small frequencies, exact conditional logistic regression in SAS 9.4 was used for estimating odds ratios (OR), excess odds ratios and their confidence intervals (CI).35 Statistical power cal- culations indicated that the sample size is sufficient for detecting a linear dose-response with an OR of 1.05 or greater per 1 mGy increase in cumulative RBM dose with a statistical power of 80% using asymptotic conditional logistic regression.36
The ethical committee of Pirkanmaa Hospital district reviewed the study protocol (tracking number R14074) and, in accordance with Finnish regulations, no informed consent was required for this register-based study. In addition, each hospital approved our study protocol before delivering the data on CT scans. We obtained permission to use data from the Finnish Cancer Registry, the Medical Birth Register, Care Register for Health Care and Congenital Malformation Register from the National Institute of Health and Welfare (1774/5.05.00/2014), as well as census data on socioeconomic status from Statistics Finland (TK-52-306-16).
Results
In our nationwide register-based study, after excluding cases with an incorrect personal identification number or prohibition to use their data, we identified 1,093 cases of childhood leukemia diagnosed in 1990-2011. Most of the
cases were acute lymphoblastic leukemia (81.1%) or acute myeloid leukemia (13.0%). The median age at diag- nosis among cases was 4.52 years (interquartile range, IQR 2.72 – 8.23). Of the cases and controls, 52% were male (Table 2). The criteria for large for gestational age were met by 121 (13.3%) of the cases and 275 (9.9%) of the controls.
Table 2. The characteristics of cases and controls before any exclusions.
Gender Female Male
Large for gestational age No
Ys
Missing
Mother’s smoked during pregnancy
No
Yes Missing
Down syndrome No
Yes
Parents’ education Mother
Upper secondary Bachelor’s degree
Master’s or doctor’s degree Missing
Father
Upper secondary Bachelor’s degree
Master’s or doctor’s degree Missing
Parents’ socioeconomic status Mother
Self-employed
Upper level employee Lower level employee Manual worker
Other
Missing
Father
Self-employed
Upper level employee Lower level employee Manual worker
Other
Missing
Age at leukemia diagnosis, years 0–2
2–7
7–15
Leukemia type Pre-B-ALL
Pre-T-ALL
Unclassified ALL
Acute myeloid leukemia Other
48.0% (525) 52.0% (568)
86.7% (788) 13.3% (121) 184
83.1% (742) 16.9% (151) 200
96.3% (1053)
3.7% (40)
48.5% (530) 22.3% (244) 10.2% (112) 18.9% (207)
52.0% (568) 15.2% (166) 10.0% (110) 22.8% (249)
7.7% (84) 16.1% (176) 34.8% (380) 21.4% (231) 18.2% (199) 2.1% (23)
13.9% (152) 17.6% (192) 18.3% (197) 34.0% (372) 12.4% (135) 4.1% (45)
14.3% (156) 55.5% (605) 33.4% (332)
75.6% (826) 5.9% (64) 1.8% (20) 13.6% (149) 3.1% (34)
48.0% (1575) 52.0% (1704)
90.1% (2493) 9.9% (275) 511
84.5% (2296) 15.5% (420) 563
99.9% (3277)
0.1% (2)
50.6% (1659) 23.1% (756) 9.8% (321) 16.6% (543)
51.4% (1685) 16.2% (532) 10.2% (334) 22.2% (728)
8.3% (273) 15.7% (514) 34.5% (1130) 20.6% (674) 20.3% (664) 0.7% (24)
12.0% (395) 18.2% (596) 17.9% (587) 35.0% (1148) 14.3% (469) 2.6% (84)
Cases(n=1,093) Controls(n=3,279)
P
0.001
0.096
<0.001
ref 0.869 0.406
ref 0.423 0.880
ref 0.477 0.521 0.490 0.859
ref 0.178 0.273 0.170 0.036
The reported P-values are from an univariate conditional logistic regression model.The non- binary variables were treated as factors and the reference categories are marked with“ref”.ALL: acute lymphoblastic leukemia.
1876
haematologica | 2018; 103(11)