Role of Statisticians in Follow-Up of A-Bomb Survivors

1
Role of Statisticians in Follow-Up of A-Bomb
Survivors

• Donald A. Pierce
• Oregon Health Sciences Univ.
• Retired from Radiation Effects Res. Fndn.

Slides for talk, related things, at
www.science.oregonstate.edu/piercedo
2
(No Transcript)
3
My Talk Today

• Some brief history of ABCC/RERF, including role
  of statisticians
• General nature of the radiation-cancer dose
  response, including age-time variation
• (Note Is primary source of quantitative
  information on radiation effects in humans --
  medicine, workplace, environment)
• Why the continued research remains important
  after more than 50 years

4

• Bombs August 1945, Joint Commission of
  Occupation, October 1945
• Pres. Truman directive to NAS 1946, Atomic Bomb
  Casualty Commission (ABCC)
• Motivations leukemia, cancer, acute effects,
  inherited effects, others
• By 1950 Depts of Genetics, OBGYN, PEDS, Internal
  Med, Radiology, Pathology, Biochem/Micro,
  Biometrics

5

• Large-scale clinical and pathology programs
  examinations and autopsies
• Enormous efforts interviewing survivors within 2
  km for shielding histories
• More than 1500 employees at peak, now about 250
  with 40 scientists
• Americans Around 10-15 recently, with far more
  at peak (largely physicians military and
  jointly with Yale)

6

• Francis Committee (Jablon, Moore) 1955 profound
  effect establishing sound epidemiological study
• Fixed study cohort of around 100,000 that could
  be followed up (most importantly no addition of
  cases only also for F1 and in-utero)
• Became bi-national Radiation Effects Research
  Foundation (RERF) 1975
• Recurrent low ebbs, particularly in the late
  1970s

7

• Statisticians played increasingly major role from
  around 1950
• Gil Beebe, Seymour Jablon were the NAS contract
  officers during about 1955-85
• Charles Land (OSU 1970-75) was in Hiroshima about
  6 years, is still involved
• Many other US statisticians were there for 2
  years or so in that era
• Several Japanese statisticians highly involved,
  but

8

• In 1978 Jablon set up a major contract with UW
  Biostats (low ebb thing)
• Ross Prentice, Art Peterson, others, were there
  in 1980-81
• They recruited Dale Preston and me in 1981
  Preston stayed until 2004 and I was there for 16
  years during 81-04. Other OSU connections include
  students Ken Kopecky and Bob Delongchamp
• By 1987 we had a Stats Dept of 15 that would have
  done well in a US university

9

• Thanks to Beebe, Jablon Land, by 1975 stat
  methods were state of art in testing for effects
  (Mantel-Haenszel methods)
• These methods did not lend themselves to
  estimation, so Preston and I took this on
• Relative risk regression notions had just become
  available requiring adaptation for large study,
  suitable form of interactions, multiple time
  scales
• By 1986 we had this ready for use, with
  widely-used and general interactive software
  developed by Preston (Epicure)

10

• Possibilities richer than most applications, due
  to size of study and small chance of confounding
  (can estimate RRs of 1.1)
• Largely because the dose-distance gradient was
  very steep, so those with large and small doses
  differ little otherwise
• Also, the participation and follow-up rates were
  essentially 100 (interesting point)
• Finally, there is such a long-term strong
  interest, promoting continued efforts

11

• To proceed, we need some perspective on radiation
  dose Gray
• 1 Gy to major organs causes severe illness,
  although seldom fatal
• A CT scan, although usually localized, is about
  0.01 Gy GI series about half of that
• Occupational limits are about 0.02 Gy/yr,
  although cumulatively further limited
• Thus 0.10 Gy is a fairly large dose of
  considerable interest

12
General Summary
Dose Gy MeanDistance PersonsFollowed CA Cases1958-98 Est ExcessCases
lt 0.005 3680 60,800 9,600 3
.005 0.1 1990 27,800 4,400 80
0.1 0.2 1630 5,500 970 75
0.2 0.5 1500 5,900 1,100 180
0.5 1 1280 3,170 690 210
1 2 1110 1,650 460 44
gt2 900 564 185 61
Tot excl lt .0005 row 44,584
7,805 650 First row is a
sample of distal survivors 5-10 km --- thus
analyses are done totally within cohort
13
ERR is factor increasing baseline rates, here
sex-averaged FM ratio is 64 (offsets
baseline ratio) ---- EAR is absolute risk
14

• ERR is factor increasing baseline rates, here sex
  averaged and at age 70
• At 1 Gy rates are increased by about 50 over
  normal levels

15

• Why such long follow-up, and such extensive
  analysis, is needed
• Lifelong effect for cancer was not expected
• Even when this became apparent the age-decline in
  RR was confused with effect of exposure age
• Understanding of such things is only emerging
  with continued follow-up and analysis

16

• The left panel here shows the view of things
  until the late 1990s (still widely held) and the
  right panel shows our current understanding of
  the same data
• We now have a reasonable understanding of why
  the age-declining ERR should be expected

17

• Simply, cancer is caused by accumulation of
  somatic mutations, and
• The radiogenic mutations persist for all
  remaining lifetime, but become relatively less
  important as more accumulate
• For any mutational exposure (including smoking)
  with age-cumulative dose D(a) it is plausible and
  explains well the data that

18

• Continued follow-up and analysis is needed to
  clarify the effect of exposure age --- one of the
  most important remaining issues
• On another issue, some would like to believe that
  for small radiation doses, e.g. 0.05 Gy, there is
  no cancer risk at all
• But careful analysis based on the 30,000
  survivors in the low-dose range shows that this
  is implausible
• Statisticians also have clarified the (modest)
  effect of random errors in dose estimates

19

• Virtually no other data really bear strongly on
  the quantitative needs for radiation protection
• Less explicable effect on non-cancer mortality,
  much smaller ERR
• Possible that this is only for large doses, due
  to killing large proportions of marrow cells with
  immunological effects
• Virtually no evidence of inherited effects, where
  mechanisms seem mainly limited to gonadal
  mutations

20

• The needs and opportunities at RERF, along with
  the Golden Age of biostatistics, made all this
  incredibly attractive
• My OSU career spanned 25 years and was very good
  for me, forming the basis for that second
  career
• Am really grateful for what both places have
  meant for me and my family

21
SOME REFERENCES Preston, D.L., Shimizu, Y.,
Pierce, D.A., Suyama, A. and Mabuchi, K. (2003b).
Studies of mortality of atomic bomb survivors,
Report 13 Solid cancer and noncancer mortality
1950 1997. Radiation Research 160,
381-407. Pierce, D.A. and Vaeth, M (2003e).
Age-time patterns of cancer to be anticipated
from exposure to general mutagens. Biostatistics
4, 231-248. Pierce, D.A. (2002). Age-time
patterns of radiogenic cancer risk their nature
and likely explanations. Journal of Radiological
Protection 22, A147-A154. Pierce, D.A., Stram,
D.O., Vaeth, M., and Schafer, D.W. (1992b). The
errors-in-variables problem considerations
provided by radiation dose-response analyses of
the A-bomb survivor data. J. Amer. Statist. Assn.
87, 351-359. Pierce, D.A. and Preston, D.L.
(2000a). Radiation-related cancer risks at low
doses among atomic bomb survivors. Radiation
Research 154, 178-186.