Childhood cancers near nuclear power plants: A possible radiobiological explanation Alfred Kรถrblein Nuremberg, Germany
The KiKK study: KiKK = Kinderkrebs um Kernkraftwerke (German) Epidemiologic study of childhood cancers near German nuclear power plants • Commissioned in 2002 by the Federal Office for Radiation Protection (Bundesamt für Strahlenschutz, BfS) (since 1999 headed by a Green Party member, Prof. König) • Conducted by the German Childhood Cancer Registry (GCCR) • Started in April 2003 • Results presented in December 2007 • External advisory expert commission, consisting of epidemiologists, statisticians, physicians and physicists
Study design • Case-control study (3 controls per case, matched by age, sex and reactor site) • All cancers and, as a sub group, leukaemias • All German commercial NPPs • Children below age 5 • Longest possible study period (1980-2003) • One-tailed statistical test • Proxy of radiation exposure: Inverse distance of place of residence at diagnosis
Study region = 3 counties next to each reactor site (16 sites) altogether 41 counties (some overlap)
Results •
Significant negative distance trend for all cancers (p=0.0034) and for leukaemia (p=0.0044)
•
Relative risk for r < 5 km vs. r > 5 km: RR=1.61 for all cancers and RR=2.19 for leukaemia
•
Significant excess in the 5-km zone: 29 excess cancer cases, 20 excess leukemia cases
•
To date, the KiKK study is the most powerful study worldwide of childhood cancers near NPPs
Leukemia in children < 5 years distance dependency of relative risk 3,0
relative risk
2,5
2,0
1,5
1,0
0,5 0
10
20
30
40
50
distance from NPP [km]
60
70
80
Conclusions of the KiKK authors • Significantly increased cancer risk, mainly for leukaemia, living in the proximity (r<5 km) of NPPs • Results not consistent with most international studies • Results unexpected given the level of radiation exposure • Causes unknown, but radiation can be ruled out. Unexplained Confounding? Chance result?
Epidemiologic studies in other countries Leukemia in children < 5 y, r < 5km vs. r > 5km • Switzerland (Spycher et al. 2011) cohort study (1985-2009) • Great Britain (Bithell et al. 2008) geographical study 1969-2004 • France (Laurier et al. 2008) geographical study 1990-1998 • Germany (Kaatsch el al. 2008) geographical study 1980-2003
Pooled evaluation of leukemia data Germany (D), Great Britain (GB), Switzerland (CH), France (F) O
E
SIR
p value
RR
p value
< 5 km
11
7,87
1,40
0,1711
1,46
0,1668
> 5 km
54
56,37
0,96
< 5 km
20
14,74
1,36
0,1108
1,41
0,0858
> 5 km
1579
1640,44
0,96
0-5 km
5
5,2
0,96
0,5939
0,91
0,6449
5-20 km
109
103,0
1,06
< 5 km
34
24,09
1,41
0,0328
1,45
0,0274
> 5 km
585
599,58
0,98
< 5 km
70
51,9
1,35
0,0095
1,39
0,0055
> 5 km
2327
2399,4
0,97
CH
GB
F
D
pooled data
Joint regression analysis of leukemia data Germany (D), Great Britain (GB), Switzerland (CH), France (F) model: SIR~exp(1/r) p=0.0174 SIR = standardized incidence ratio = O/E O=observed cases, E=expected cases 1,7 CH
1,6
D
GB
F
1,5
SIR=O/E
1,4 1,3 1,2 1,1 1,0 0,9 0,8 0,7 0
5
10
15
20
25
30
distance from NPP [km]
35
40
45
50
Biological plausbility? • KiKK found leukemia risks were ~ doubled (RR=2.19) in children under age 5 near NPPs • Doubling dose for childhood leukemia: some mSv after in utero exposure in first trimester see: Stevenson, Strahlenbiologisches Gutachten, page 92
• Official dose estimate for 1 year old children: some µSv per year , i.e. 1000-times smaller than doubling dose see: http://dip21.bundestag.de/dip21/btd/16/068/1606835.pdf
From: http://dip21.bundestag.de/dip21/btd/16/068/1606835.pdf
Official dose and risk estimates unrealiable • Official dose calculations use simplified propagation models: Two dimensional Gauss model might be in error (up to ± factor 10) • UK Government’s Committee Examining Radiation Risks of Internal Emitters (CERRIE, 2004) concluded uncertainties in INTERNAL radiation doses were considerably greater than in EXTERNAL radiation doses • Uncertainties accumulate: official dose estimates might be low by at least a factor of 10 which, however, would still be below allowed doses (German legal dose limit = 0.3 mSv/y near NPPs)
The official risk estimates might be low by a factor of 10-100, but we need to explain a factor of 1000!
How to close the gap? The usual assumption that risk is proportional to average dose is only correct if the dose-response relationship is linear. But: (1) do we know the shape of the dose-response curve for prenatal leukemia induction? (2) dose rates from NPPs show large variations in time and space, while dose rate from natural background radiation is rather constant at a given location.
Factor 10 000
Calculated isolines of the fallout dispersion coefficient for diffusion category D and stack height 150 meters. The calculated maximum fallout occurs at a distance of 2100 m from the NPP From: Otfried Schuhmacher, Portsmouth conference 1996, http://strahlentelex.de/PORTS_Schumacher.pdf
Prenatal leukemia induction - a teratogenic effect?
Dose response relationship comparison of stochastic and teratogenic radiation effects
Combined effect of nonlinear dose-response and discontinuous releases Assumptions: 1. Nonlinear dose-response: effect ~ dose^3 2. Annual dose from NPPs: 10% of natural background dose 3. Discontinuous releases: NPP-releases only during 3 weeks per year (annual revision) 4. Doses from NPP releases add to background dose Relative risk: RR = 3/52*(1+0.1*52/3)^3 + (52-3)/52*1 = 2.12 i.e.: 10% increase of annual dose ď&#x192; 112% increase of risk!
annual dose duration of revision power of dose relative risk exess relative risk
5% background dose 4 weeks 3 1,27 27%
annual dose duration of revision power of dose relative risk exess relative risk
10% background dose 4 weeks 3 1,86 86%
annual dose duration of revision power of dose relative risk exess relative risk
5% background dose 4 weeks 4 1,49 49%
annual dose duration of revision power of dose relative risk exess relative risk
10% background dose 4 weeks 4 3,08 208%
annual dose duration of revision power of dose relative risk exess relative risk
5% background dose 2 weeks 4 2,04 104%
annual dose duration of revision power of dose relative risk exess relative risk
10% background dose 2 weeks 4 7,42 642%
The calculation needs to be modified if we allow for non-zero releases during routine operation. Then a 4th parameter is needed for the share of releases during revision (defined as percentage of the annual releases).
annual dose duration of revision releases during revision power of dose relative risk (RR) excess relative risk (ERR)
10% of background dose 3 weeks 60% of annual releases 4 2,11 111%
RR: B5 = 1/52*(B2*(1+B1*B3/B2*52)^B4+(52-B2)*(1+B1*(1-B3)/(52-B2)*52)^B4) ERR: B6 = B5-1
Noble gas emissions from NPPs NPP Gundremmingen (Bavaria), revision in Sept/Oct 2011 linear plot
noble gas concentration [kBq per m続]
1600 1400 1200 1000 800 600 400 200 0
36th
37th
38th
39th
40th
calendar week
41st
42nd
43rd
44th
Noble gas emissions from NPPs NPP Gundremmingen (Bavaria), revision in Sept/Oct 2011, semilogarithmic plot
noble gas concentration [kBq per m続]
10000
1000
100
10
1
36th
37th
38th
39th
40th
calendar week
41st
42nd
43rd
44th
Noble gas emissions from NPPs NPP Gundremmingen, 38th calendar week 2011, linear plot
noble gas concentration [kBq per m続]
1600 1400 1200 1000 800 600 400 200 0 9-19
9-20
9-21
9-22
date
9-23
9-24
9-25
Noble gas emissions from NPPs NPP Gundremmingen, 38th calendar week 2011, semilogarithmic plot lower curve: 37th calendar week
noble gas concentration [kBq per m続]
10000
1000
100
10
1 9-19
9-20
9-21
9-22
date
9-23
9-24
9-25
Shape of the dose response relationship A cumulative lognormal distribution follows from the assumption that radio-sensitivities and doses are randomly distributed in a population 0,5
effect
0,4
0,3
0,2
Cumulative lognormal distribution with median Âľ=4 and standard deviation=0,5
0,1
0,0 0
1
2
3
annual dose (multiples of background dose)
4
Step 1: From concentrations to releases 1. Releases proportional to concentrations: ½ hourly releases: air stream times ½ hourly concentrations air stream in exhaust stack: 130 000 m³ per half hour 2. Releases in Sept-Oct 2011: sum of reported ½ hourly releases 3. Releases during rest of the year (10 months): Assumption: release rate = before revision (37th calendar week) 4. Total annual release: ~16.9 TBq/a releases on 22nd September: 3.1 TBq (18% of total) releases on 22-23 September: 5.4 TBq (32% of total) releases on 38th calendar week 7.8 TBq (46% of total)
Dr. Alfred Körblein
KiKK Study
27
Step 2: From releases to doses 1. Assumption: Doses proportional to releases: Proportionality factor = annual dose from NPP releases (e.g. 50 ÂľSv), divided by total annual releases (16.9 TBq) 2. Doses from NPP releases add to doses from background radiation (1 mSv per year without radon contribution)
Dr. Alfred KĂśrblein
KiKK Study
28
Dose, in units of background dose NPP Gundremmingen, 38th calendar week 2011 13
relative dose
11
9
7
5
3
1 9-19
9-20
9-21
9-22
9-23
9-24
9-25
date
Dr. Alfred Kรถrblein
KiKK Study
29
Step 3: From dose to relative risk 1. ½ hourly doses (including background dose) are converted into risks using a lognormally shaped dose-response relationship 2. ½ hourly risks are divided by risk at background dose to obtain relative risks 3. Annual average of relative risk (RR) is overall annual RR 4. The calculation yields RR = 2,18 i.e. 5% increase of dose 118% increase of risk!
Relative Risk NPP Gundremmingen, 38th calendar week 2011 400 350
relative risk
300 250 200 150 100 50 0
9-19
9-20
9-21
9-22
date
9-23
9-24
9-25
Summary and Conclusion • Leukemia in children < 5y near German NPPs more than doubled • Official dose from NPP releases is some µSv/y, doubling dose for leukemia after in-utero exposure is some mSv, 1000-times greater! • Aerial radioactive discharges from NPPs are highly episodic, mainly during refueling and inspection • If dose-response is curvilinear, effect increases disproportionately with dose • Conclusion: The present result challenges the assertion of official Radiation Protection Agencies that radioactive releases must be ruled out as a possible cause of increased leukemias near NPPs.
Greetings from Nuremberg!