Review of epidemiological associationsReview of epidemiological associations between low/moderate doses of ionizing

radiation and circulatory disease risks andradiation and circulatory disease risks, and their possible mechanisms

Late Health Effects of Ionizing Radiation. Bridging the Experimental and Epidemiologic Divide

Georgetown University Conference Center, Washington DC, May 4-6 2009

MP Little, EJ Tawn, I Tzoulaki, R Wakeford, G Hildebrandt, F Paris, S Tapio and P Elliott

Outline of talkOutline of talk

IntroductionIntroductionCirculatory disease in A-bomb survivorsCirculatory disease in medically exposedCirculatory disease in medically exposed groupsCirculatory disease in occupationallyCirculatory disease in occupationally exposed groupsRole of other risk factors: confoundinggMechanismsConclusions

Epidemiological evidence forEpidemiological evidence for associations of circulatory disease with

di ti tt d hi tradiation: potted historyHigh dose radiotherapy (>5 Gy): circulatory disease risks known for some time and seen in many studies (e.g. patients treated for y ( g pHodgkin disease, breast cancer)High dose circulatory disease mechanisms known (acute tissue damage (deterministic effect))g ( ))~15 years ago excess risk for circulatory disease and many other non-cancer endpoints seen in A-bomb survivors (Life Span Study, Adult Health Study)Excess circulatory risks seen in Chernobyl recovery workers, some radiation workers (US, Canadian, BNFL, Mayak) etcGenerally no excess risks of other non-cancer disease in groups y g papart from A-bomb survivors

Non-cancer mortality disease in A-bombNon cancer mortality disease in A bomb survivors (Preston et al Radiat Res 160:381-407;2003)

Endpoint ERR /Sv (90% CI)p ( )

Heart disease 0.17 (0.08, 0.26)

Stroke 0.12 (0.02, 0.22)

Respiratory disease 0.18 (0.06, 0.32)Respiratory disease 0.18 (0.06, 0.32)

-Pneumonia 0.16 (0.00, 0.32)

Digestive disease 0.15 (0.00, 0.32)

-Cirrhosis 0 19 (-0 05 0 50)Cirrhosis 0.19 ( 0.05, 0.50)

Infectious disease -0.02 (<-0.2, 0.25)

-Tuberculosis -0.01 (<-0.2, 0.4)

Other diseases 0 08 ( 0 04 0 23)Other diseases 0.08 (-0.04, 0.23)

-Urinary diseases 0.25 (-0.01, 0.60)

All non-cancer 0.14 (0.08, 0.20)

Uniformity of excess risk: possible bias?

A-bomb survivor cause of deathA bomb survivor cause of death misclassification (Sposto et al Biometrics 48:605-17;1992)

I i t lit d tIncrease in non-cancer mortality due to death certificate misclassification?Autopsy study finds 22% of non-cancer deaths misclassified as cancer deaths.Statistical adjustment reduces the ERR/Gy for non-cancer mortality from 0.06 to 0.05, ybut risk coefficient remains statistically significant.g

A-bomb survivor potential confoundingA bomb survivor potential confounding factors for non-cancer mortality (1) (Shimizu et al Radiat Res 152:374-89;1999)

No confounding suggested but subgroup analyses basedNo confounding suggested, but subgroup analyses based on mailed respondents

A-bomb survivor potential confounding p gfactors for non-cancer mortality (2) (Shimizu et al Radiat Res 152:374-89;1999)

A-bomb survivor potential confounding p gfactors for circulatory disease morbidity (Yamada et al Radiat Res 161:622-32;2004)

DiseaseDisease(quadratic dose-

response)

ERR at 1 SvERR at 1 Sv (95% CI)No adjustment for

smoking and drinking

ERR at 1 SvERR at 1 Sv (95% CI)Adjusted for smoking

and drinking

Hypertension(incidence 1958-98)

0.03 (0.00, 0.06) 0.03 (0.01, 0.06)(incidence, 1958 98)

Myocardial infarction(<40 year of age ATB)

0.25 (0.00, 0.69) 0.17 (-0.03, 0.56)

(incidence, 1968-98)

No confounding suggested based on these lifestyleNo confounding suggested, based on these lifestyle factors

Circulatory disease excess relative risk Sv-1C y Sin A-bomb data (Preston et al Radiat Res 160:381-407;2003, Yamada et al Radiat Res

161:622-32;2004 Tatsukawa et al Radiat Res 170:269-74;2008)161:622 32;2004, Tatsukawa et al Radiat Res 170:269 74;2008)

LSS heart mortality 0.17 (90% CI 0.08 – 0.26)LSS stroke mortality 0.12 (90% CI 0.02 – 0.22)AHS hypertension morbidity 0 05 (95% CI -0 01 – 0 10)AHS hypertension morbidity 0.05 (95% CI -0.01 0.10) AHS hypertensive heart morbidity 0.01 (95% CI -0.09 – 0.09) AHS ischaemic heart morbidity 0.05 (95% CI -0.05 – 0.16) AHS myocardial infarction morbidity 0 12 (95% CI 0 16 0 60)AHS myocardial infarction morbidity 0.12 (95% CI -0.16 – 0.60) AHS stroke morbidity 0.07 (95% CI -0.08 – 0.24)AHS stroke/MI morbidity childhood AE 0.72 (95% CI 0.24 – 1.40)

LSS circulatory mortality risks modest but highly statistically significant

AHS circulatory morbidity risk lower and generally non-significant

A-bomb survivor exposure in utero + pchildhood (Tatsukawa et al Radiat Res 2008; 170: 269-74)

AHS used to investigate circulatory disease incidence during 1978- 2003 g y gamong 506 survivors exposed in utero in comparison with incidence among 1053 survivors exposed as childrenGenerally no significant effects among in utero exposedy g g p

Hypertension ERR = 0.20 Sv-1 (95% CI -0.39, 1.38)Stroke/MI ERR = -0.91 Sv-1 (95% CI -1.00, 79.3)

Generally significant effects among childhood exposedGenerally significant effects among childhood exposedHypertension ERR = 0.15 Sv-1 (95% CI -0.01, 0.34)Stroke/MI ERR = 0.72 Sv-1 (95% CI 0.24, 1.40)

In utero exposed still aged <60 years

Dose response for circulatory disease in p yA-bomb survivors (Preston et al Radiat Res 160:381-407;2003)

“Noise” at low dose: increased risk only clear above ~1 Sv

Shape of dose-response uncertain

Dose response non-linearity for circulatory p y ydisease in A-bomb survivors? (Little Radiat Env Biophys 43:67-75;2004)(based on data of Preston et al Radiat Res 160:381-407;2003)

Linear threshold: ERR=α (D - Dt)+

Linear-quadratic threshold: ERR=α (D - Dt)+ + β (D - Dt)+2

Endpoint/dose response model Dose threshold (Sv) (95% CI)

Power of dose (95% CI)

Power of dose: ERR=α Dk

Heart disease linear threshold 0.14 (<0, 1.08) -

Heart disease lin-quadratic threshold 0.15 (<0, 1.22) -

Heart disease power of dose - 1.12 (0.17, 3.34)Heart disease power of dose 1.12 (0.17, 3.34)

Stroke linear threshold 0.30 (<0, 3.11) -

Stroke lin-quadratic threshold 0.45 (<0, >5.00) -

Stroke power of dose - 1 54 (0 02 >10 00)Stroke power of dose 1.54 (0.02, >10.00)

Shape of dose-response uncertain

Circulatory excess relative risk Sv-1 inCirculatory excess relative risk Sv in diagnostic+other medical (Little et al Radiat Res 169:99-109;2008 +

Little et al Int J Epidemiol 2009 in press)p p )

US peptic ulcer coronary 0.11 (95% CI 0.01 – 0.22)US peptic ulcer coronary (low) 0.10 (95% CI -0.12 – 0.33)US peptic ulcer heart other -0.11 (95% CI -0.40 – 0.17)US peptic ulcer heart other (low) -0.16 (95% CI -0.49 – 0.17)UK spondylitis stroke -2 43 (95% CI -4 29 – 0 71)UK spondylitis stroke 2.43 (95% CI 4.29 0.71)UK spondylitis other circulatory -0.01 (95% CI -0.12 – 0.13)Massachusetts TB circulatory -0.11 (95% CI -0.20 – 0.01)

Risks in medically exposed groups generally non-significant, consistent with A-bomb

Dose response for circulatory disease inDose response for circulatory disease in US peptic ulcer study (relative risk + 95% CI)

2.5Coronary heart disease

2 . 5

Other heart disease95% CI) (Carr et al Int J Radiat Oncol Biol Phys 61:842-50;2005)

1.5

2.0

tive

risk

1 . 5

2 . 0

0.5

1.0Rel

at

0 . 5

1 . 0

1 1

0 1 2 3 4Average heart dose (Gy)

0.00 1 2 3 4

0 . 0

ERR = 0.11 Gy-1 95% CI (0.01, 0.22) ERR = -0.11 Gy-1 95% CI (-0.40, 0.17)

Clear increased risk for coronary heart disease (but onlyClear increased risk for coronary heart disease (but only above 2 Gy), no excess for other heart disease

Circulatory excess relative risk Sv-1 in yoccupational groups (1) (Little et al Radiat Res 169:99-109;2008 + Little et al

Int J Epidemiol 2009 in press)

IARC 15 country circulatory 0.09 (95% CI -0.43 – 0.70)IARC 15 country stroke 0.88 (95% CI -0.67 – 3.16)Canadian workers circulatory M 2 3 (90% CI 0 9 3 7)Canadian workers circulatory M 2.3 (90% CI 0.9 – 3.7)Canadian workers circulatory F 12.1 (90% CI -0.4 – 24.6)Chernobyl recovery hypertens 0.26 (95% CI -0.04 – 0.56)y y yp ( )Chernobyl recovery isch heart 0.41 (95% CI 0.05 – 0.78)Chernobyl recovery other heart -0.26 (95% CI -0.81 – 0.28)Ch b l t k 0 45 (95% CI 0 11 0 80)Chernobyl recovery stroke 0.45 (95% CI 0.11 – 0.80)

Canadian workers exhibit large risks (10-50 x A-bomb): possible bias? pOther studies generally modest risk (≈A-bomb)

Circulatory excess relative risk Sv-1 inCirculatory excess relative risk Sv in occupational groups (2) (Little et al Radiat Res 169:99-109;2008 + Little et al Int J Epidemiol 2009 in press)`

Mayak ischaemic heart (external γ) 0.11 (95% CI 0.05 – 0.17)Mayak stroke (external γ) 0.46 (95% CI 0.36 – 0.57)NRRW-3 circulatory 0.25 (95% CI -0.01 – 0.54)NRRW 3 circulatory 0.25 (95% CI 0.01 0.54)BNFL circulatory 0.54 (90% CI 0.30 – 0.82)BNFL ischaemic heart 0.70 (90% CI 0.37 – 1.07)BNFL stroke 0 66 (90% CI 0 17 1 27)BNFL stroke 0.66 (90% CI 0.17 – 1.27)UK AWE circulatory 2.51 (95% CI 0.01 – 5.56)UKAEA ischaemic heart -0.66 (95% CI -1.46 – 0.23)US Oak Ridge ischaemic heart -2.86 (95% CI -6.90 – 1.18)German uranium miner circulatory -0.26 (95% CI -0.6 – 0.05)

Increased risk in Mayak NRRW-3 BNFL + AWEIncreased risk in Mayak, NRRW 3, BNFL + AWE Other risks negative, consistent with modest excess risk

Combined circulatory disease risk estimates y(excess relative risk (ERR)/Sv +95% CI)

Studies used ERR /Sv (95%Studies used ERR /Sv (95% CI)

Occupational + environmental +BNFL +Mayak -NRRW-3 0.19 (0.14, 0.24)a

Occupational + environmental –BNFL +Mayak +NRRW-3 0.19 (0.14, 0.23)a

Occupational + environmental –BNFL –Mayak +NRRW-3 0.10 (-0.05, 0.24)a

A b b + di l 0 03 ( 0 01 0 07)aA-bomb + medical 0.03 (-0.01, 0.07)a

All studies –BNFL –Mayak +NRRW-3 0.03 (-0.01, 0.07)a

All studies –BNFL +Mayak +NRRW-3 0.09 (0.06, 0.12)aAll studies BNFL Mayak NRRW 3 0.09 (0.06, 0.12)ap<0.005 for heterogeneity of ERR

Overall modest (but highly statistically significant) ERRBUT significant heterogeneityBUT significant heterogeneityDropping Mayak gives substantial reduction in ERR (+lacks significance)

Sensitivity of combined circulatory disease risks to y ystudy exclusion (ERR/Sv +95% CI)+contribution to χ2 heterogeneity (all excluding BNFL)χ g y ( g )

Studies excluded ERR /Sv (95% CI) χ2 heterogeneity (df)A-bomb 0.11 (0.07, 0.14) 7.85 (5)

US peptic ulcer 0.09 (0.06, 0.13) 2.29 (2)

UK ankylosing spondylitis 0.10 (0.07, 0.13) 6.49 (2)

Massachusetts TB 0.12 (0.08, 0.15) 17.54 (1)

IARC 15 country workers 0.09 (0.06, 0.12) 3.83 (1)

NRRW-3 0.09 (0.06, 0.12) 1.17 (1)

German uranium miners 0.10 (0.06, 0.13) 4.95 (1)

Chernobyl recovery workers 0.09 (0.06, 0.12) 0.68 (1)

Mayak workers 0.03 (-0.01, 0.07) 50.49 (2)

Three Mile Island 0.09 (0.06, 0.12) 11.62 (2)

None (All studies) 0.09 (0.06, 0.12) 103.08 (17)( p<10-13)

Major contributions to heterogeneity from Mayak, Massachusetts TB, TMIDropping most studies has little effect, but dropping Mayak reduces ERR lot

Can adjusting for fractionation explainCan adjusting for fractionation explain heterogeneity in risk (1)?Schultz Hector & Trott (IJROBP 67 10 18 2007) suggested adjusting forSchultz-Hector & Trott (IJROBP 67:10-18;2007) suggested adjusting for fractionation effects accounts for A-bomb vs RT risk discrepancyAssuming true dose response is αD+βD2, in RT study where dose D given in n fractions, biologically equivalent single (corrected) dose D' given by 2 2( / ) ( / ) ' 'n D n D n D Dα β α β⎡ ⎤+ = +⎣ ⎦

1 60 Smaller /β (larger

0.801.001.201.401.60

d do

se (G

y)

α/β = 1/β 3 7

Smaller α/β (larger quadratic term) gives bigger adjustment

0.000.200.400.60

0 0 5 1 1 5 2 2 5

Cor

rect

ed α/β = 3.7 (lowers dose, increases risk), particularly for large doses D0 0.5 1 1.5 2 2.5

Uncorrected dose (Gy)

large doses D

Can adjusting for fractionation explainCan adjusting for fractionation explain heterogeneity in risk (2)?

Consider A bomb studies (P t t l (R di t R 160 381 407 2003) Y d t lConsider A-bomb studies (Preston et al (Radiat Res 160:381-407;2003), Yamada et al

(Radiat Res 161:622-32;2004)) + 3 medical studies (Darby et al (BJC 55:179-90;1987), Davis et al (Cancer Res 49:6130-6;1989), Carr et al (IJROBP 61:842-50;2005)

Use central estimate of α/β=2 used by Schultz Hector & TrottUse central estimate of α/β=2 used by Schultz-Hector & Trott (IJROBP 67:10-18;2007), also lower limit (α/β=1) suggested by experimental data (Lauk et al Radiother Oncol 8:363-7;1987)

Adjustment ERR /Sv (95% CI) p-value of heterogeneityUnadjusted 0.03 (-0.01, 0.07) 0.00419α/β = 2 0 04 (-0 01 0 08) 0 00659α/β 2 0.04 (-0.01, 0.08) 0.00659α/β = 1 0.04 (0.00, 0.09) 0.00811

Adjusting for fractionation reduces statistical discrepancy j g p ybetween A-bomb and RT studies, but not by much

Does disease endpoint explainDoes disease endpoint explain heterogeneity in risk?Have considered all circulatory disease: difference between heart disease and stroke?

Endpoint ERR /Sv (95% CI) p-value of heterogeneityHeart disease 0.07 (0.04, 0.11) 0.00085Stroke 0.27 (0.20, 0.34) 0.00004All circulatory 0.09 (0.06, 0.12) <10-13

Stroke ERR substantially higher than heart disease ERR, but heterogeneity remains

Confounding factors for circulatoryConfounding factors for circulatory disease (1)W ll k i d d t i k f t f i l t di i kWell known independent risk factors for circulatory disease risk

– cigarette smoking– diabetes– obesity– high blood pressure– high LDL cholesterol (and total cholesterol)

Each capable of about 2 x increase in riskStress also associated with circulatory disease: particularly important for occupational studies (shift work etc) +A bombimportant for occupational studies (shift work etc) +A-bomb (survivors know position in cities at time of bomb, so ~dose)

Inverse relation of suicide with dose in A-bomb (Shimizu et al Radiat Res(152:374-89;1999)

Confounding factors for circulatoryConfounding factors for circulatory disease (2)Many risk factors correlated with socioeconomic status (SES):Many risk factors correlated with socioeconomic status (SES): limited adjustment for SES in some occupational studies (IARC 15-country, BNFL, NRRW-3), none in others Few studies adequately control for risk factors

– A-bomb morbidity study (Yamada et al Radiat Res 161:622-32;2004) controls for smoking, drinking– A-bomb non-cancer mortality adjusted for highest education level, house size, occupation,

marital status, smoking (Shimizu et al Radiat Res 152:374-89;1999)

– Mayak study (Azizova & Muirhead 2009) controls for smoking and drinking

Will they confound (i.e., are they correlated with radiation dose)?Cannot rule out (possibly negative) confounding by stress in A-bomb (suicide)

Mechanisms for low dose cardiovascular disease risk (1) (Little et al Radiat Res 169:99-109;2008)

R l f ti t ti i th l ll (SMC)Role for somatic mutation in smooth muscle cells (SMC) (Benditt & Benditt

PNAS 70:1753-1756;1973) (based on clonality of plaques) – would allow for low dose effects (but is DNA damage seen in atherosclerotic plaques simply consequence of ROS damage via inflammation?)p y q g )However, evidence for clonality in SMC somewhat discredited: arterial wall is normally clonally “patchy” (Chung et al Am J Pathol 152:913-23;1998)

Mutations in SMC giving increased promotion more likely to be g g p ybeneficial – now clear that SMC proliferation makes lesion development and rupture less likely (Clarke et al Circ Res 102:1529-38;2008), so role for SMC mutations unclearD t d th li l ll d th li l it ll dDamage to endothelial cells, or endothelial progenitor cells, and subsequent inflammatory response: would this extend to very low dose?

Mechanisms for low dose cardiovascular disease risk (2) (Little et al Radiat Res 169:99-109;2008)

Biological data (in vivo + in vitro) suggests moderate doses (0.1-1 Gy) down-regulate adhesion of leukocytes to the endothelium – anti-inflammatory effect – contrast with up-regulation (pro-inflammatory) of adhesion at higher (> 1 Gy) doses (Hosoi et al Int J Cancer 96:270-6;1973, Hildebrandt et al Int J

Radiat Biol 74:367-78;1998): could explain inconsistency of epidemiologic evidence (linear model fitted)Role for immune system in A bomb survivorsRole for immune system in A-bomb survivors (Kusunoki et al Radiat Res 150:227-36;1998)

suggest bone marrow/spleen may be target

Conclusions (1)Conclusions (1)Excess risk of circulatory diseases (circulatory respiratoryExcess risk of circulatory diseases (circulatory, respiratory, digestive) in A-bomb survivors, Mayak and a few othersPossibility of confounding in A-bomb + occupational studies (e.g., due to stress)?due to stress)?Taking all studies together, modest (but highly significant) ERR ~0.09 Sv-1

BUT entirely driven by Mayak study: knock this out and ERR is much smaller ~0.03 Sv-1 + non-significantHighly significant heterogeneity between studies (p<0 005) risksHighly significant heterogeneity between studies (p<0.005), risks vary by >100Heterogeneity not much altered by subdividing endpoints (heart disease vs stroke) or adjusting for dose fractionation indisease vs stroke) or adjusting for dose fractionation in radiotherapy studies

Conclusions (2)Conclusions (2)

Circulatory disease risk at high dose (>5 Gy) in many radiotherapyCirculatory disease risk at high dose (>5 Gy) in many radiotherapy studies well established, mechanisms understood (acute tissue damage?)“Noise” in A-bomb and other studies: possibly result of confounding by well known risk factors?Mechanisms for low dose circulatory disease unclear (probably notMechanisms for low dose circulatory disease unclear (probably not mutational)Picture at low dose (<5 Gy) unclear: need for better understanding of mechanisms