The Development of Risk Analysis:
A Personal Perspective
Richard Wilson
Mallinkrodt Professor of Physics (emeritus)
March 19th 2012
2:30pm
Harvard School of Public Health
1800s– Try out the technology, modify if any
problems arise.
In 1833, when the first passenger railway in the
world Liverpool to Manchester opened,
an engine ran down a Member of Parliament,
who failed to get out of the way.
Was the outcome good or bad?
Depends on his Party affiliation!
What do we mean by Risk?
Measures of Risk
How do we Calculate Risk?
(a) History
(b) Event Tree
(c) Animal Analogy
(d) Other
What about public perception?
• Yet data show unequivocally that life is
getting better.
• We all know that!
• So what are the public saying?
What is Life Expectancy?
Both the specific death rate and the life expectancy at
birth have a dip at 1919 world wide influenza epidemic.
BUT anyone born in 1919 will not actually see this dip.
Peculiarity of definition of life expectancy.
Uncertainties and Perception
Types of Uncertainties
Role of Perception
Kahneman’s 2002 economics Nobel prize
(Tversky and Kahneman, 1981)
People are inconsistent
but the inconsistency has a pattern
Major differences between
Public and Expert perceptions
Tversky’s Analysis
Assuming that the U.S. is preparing for an outbreak
of a disease imported from Asia two alternative
programs:
(72%)
Program A: 200 people would be saved.
Program B: 1/3 chance that 600 will be saved and
2/3 that no one will be saved.
Tversky’s Analysis (cont.)
Alternatively…
Program C: 400 people will die.
Program D: a 1/3 probability that no one will die and
a 2/3 probability that 600 will die.
(78%)
Simple arithmetic tells us that options C and D have
the same effect as options A and B.
When the question was posed as saving lives
the students were risk averse.
BUT
When the question was posed as preventing
deaths they favored risk!
Stanford and Harvard students were the same
MEASURES of Risk
Simple risk of Death (assuming no other causes)
by age
by cause
Risk of Injury
•by cause
•by type
•by severity
Per
•year
•lifetime
•unit operation
•event
•ton
•unit output
RISK MEASURES (continued)
Loss of Life Expectancy (LOLE)
Years of Life Lost (YOLL)
Man Days Lost (MDL)
Working Days Lost (WDL)
Public Days Lost (PDL)
Quality Adjusted Life Years (QALY)
Disability Adjusted Life Years (DALY)
Different decisions may demand different
measures
LOLE from cigarette smoking
In USA 600 billion cigarettes made
(presumably smoked)
400,000 people have premature death
(lung cancer, other cancers, heart)
1,500,000 cigarettes per death
Each death - 17 years (8,935,200 minutes) off life
or
6 minutes per cigarette
ABOUT THE TIME IT TAKES TO SMOKE ONE
What questions are you asking?
IF YOU DON’T KNOW…
THE CHANCE OF A SENSIBLE ANSWER IS
REDUCED.
The Importance of Models
Every risk calculation involves a model.
The simplest is:
NEXT YEAR WILL BE LIKE LAST YEAR.
NEXT YEAR WILL BE LIKE LAST YEAR WITH
IMPROVEMENTS.
EVERY MODEL HAS ASSUMPTION
WHICH MUST BE CLEARLY STATED
Wigner: “whenever there is a lot of energy in one
place and a lot of people in the same place, there
is a potential for disaster.
1848– no petroleum products brought up the Thames
River closer than 30 miles east of London Bridge
(Canvey Island) (VERY WISE)
1978– 120 seventeen million gallon tanks in Canvey
(for LNG each is 20 Hiroshimas)
Plus 500 vacation cottages
With one access road
EASY TO IMAGINE DISASTER
Even ignored historical data
includes big risks
65 million years ago - large meteor impact may
have destroyed the dinosaur
Risk per year 1/65 million for everyone
Lifetime risk 1 in a million
including smaller impacts we get
lifetime risk 1/100,000
Yet EPA pretended to regulate at 1 in a
million!
Beware of risk assessments that
make assumptions that an
important parameter is precise.
(US EPA)
Large Accidents in New Technologies
After World War II, the old paradigm was
inadequate:
“try it and if it gives trouble, fix it”
Society now demands evidence, in advance, that
the technology is safe.
The first major example was nuclear energy.
WHY WAS THAT THE FIRST EXAMPLE?
A number of reasons have been suggested:
1) The new technology was in hands of
fundamental scientists from start: 1946
The US Atomic Energy Commission (AEC)
2) The new technology used new physical
principles.
3) The new technology arose simultaneously with
a new deadly form of war.
4) The new technology posed unprecedented
hazards.
Atomic Energy Commission
Established an Advisory Committee on Reactor
Safeguards (ACRS) to advise on safety.
Outstanding Scientists:
Glenn Seaborg, John Von Neuman,
Robert Bacher, Edward Teller,
Eugene Wigner and Richard Feynman
Defense in Depth
(a name borrowed from the military)
Imagine the worst thing that can reasonably go wrong
in the reactor,
the “Maximum Credible Accident”
devise an engineered safeguard to prevent it
Large reactors, particularly first in a series,
in unpopulated areas, following Wigner’s principle.
The Reactor Safety Study
(Rasmussen et al. 1975)
calculated beyond the maximum credible accident
HSPH March 21st 2011
ASSUMPTIONS
(1)We have drawn all possible trees with
consequences
(2)The probabilities are independent (design
to make them so; look very carefully at
correlation
(3)Consider carefully - with some
confidentiality - actions that can artificially
correlate the separate probabilities
INDEPENDENCE IS CRUCIAL
• DO NOT break down detail in event tree
beyond what independence allows
• Fukushima: Earthquake and Tsunami are
NOT independent
• But effect of radiation and accident
calculation are independent
BUT! A TERRORIST GROUP CAN
DESTROY INDEPENDENCE:
INITIATE A PIPE BREAK (SMALL BOMB)
AND
BOMB ON CONTAINMENT AT THE SAME TIME
SET OFF WHEN WIND IS
BLOWING TOWARD CITY
A saboteur (terrorist) can couple the
different steps
Rasmussen said in 1978:
“There is nothing a terrorist can do that
those clowns (operators at TMI ) did
not do on their own”.
My response:
A terrorist can increase the probability!
Especially a group of terrorists
HSPH March 21st 2011
What about chemicals
Paracelsus “The dose makes the poison”
More is worse
Less is better
IS there a threshold?
BUT beware of exceptions
Epidemiology
Associate Death (or other Measure)
to Postulated Cause
Is it statistically significant?
Are there alternative causes (confounders)?
No cause is generally accepted unless
there is a group where death rate has doubled.
Risk Ratio (RR) > 2
But beware of stupid answers
Correlation of
Number of
Brooding sSorks
with Newborn
Babies
A contribution to epidemiology....
Associations vs. Cause-Effect
Sies, H. (1988) Nature 332, 495
• Obviously storks bring babies.
• BUT direction of causality may be wrong
• Babies bring storks
Correlation, by itself, does not imply causality
Yet in March 2012 : NH legislature
correlation between women’s cancer rate and
prostate cancer implies causality
Issues in Low Dose Linearity
• Is the observed effect identical to one that occurs
naturally?
– (pathologists should not be able to tell difference)
• Is DNA the same?
– IF so: consider natural processes at same time as pollutant
• Probable that somewhere in the cancer causing
mechanism a pollutant acts like a natural process and a
bit more cancer (linear)
(radiation cancers, chemical cancers, lung from air pollution)
– But slope extrapolated down from high levels
Issues in Low Dose Linearity
• Implicit in Armitage and Doll (1954)
• Explicit in Crump et al. (1976)
extended to any outcome including air pollution
• Crawford and Wilson (1996)
Response
Figure 2-7
Alternative Dose-Response Models That Fit the Data
Datum
Super Linear
Linear
Datum
Hockey Stick
Hormesis
Threshold
Dose
We contrast two types of medical
response to pollutants.
ACUTE TOXIC EFECT
A dose within a day causes death within a few days
(causality easy to establish)
CHRONIC EFFECT
lower doses repeated give chronic effects
(cancer, heart) within a lifetime.
(Causality hard to establish)
Characteristics
• One dose or dose
accumulated in a short time
KILLS
• 1/10 the dose repeated 10
times DOES NOT KILL
Early Optimism Based on Poisons
There is a threshold below which nothing happens
BUT
J.G. Crowther 1924
Probability of Ionizing a Cell by radiation
is Linear with Dose
Naively this would lead to many cancers each
second.
There must be a repair mechanism. Is that linear?
Incremental Risk can actually be greater
than the simple linearity assumption of a
non-linear biological dose-response is
assumed
ALSO
Possible to have 2 effects.
One benign or beneficial, other adverse
Can animals help us?
Especially laboratory animals,
rats and mice
Rat lung and human lung are
similar
These principles do not work for cancer
Arguments due to Richard Peto
Rat lungs look like people’s lungs
If a piece of tissue is equally likely to develop a cancer
whether attached to rat or man
Cancer in man/cancer in rat (for same dose)
=(
) x (mass-man/mass-rat) =millions
Actually closer to one
Test needed. Compare rats and mice use whatever human
data we have (20 or so chemicals)
Usual model:
A toxic chemical will be carcinogenic
highly toxic - highly carcinogenic
If carcinogenic in animals
carcinogenic in people.
What is the uncertainty in this
assumption?
What do data say?
What does EPA say?
Different End Points Behave Differently
• Alcohol in moderation reduces the incidence of
stroke
• Alcohol is an animal carcinogen and in conjunction
with cigarettes increase lip cancer drastically
• Alcohol is a narcotic and in large quantities has
dramatic side effects (auto accidents etc)
– Minimum is uncertain.
– Sir Richard Doll orally said he thought it is too low.
– What does that tell you?
What is the
Precautionary Principle?
Do not do anything unless you personally
understand the effects
Do not do anything unless all politicians agree
on the effects
Do not do anything unless a “responsible
committee” understands the effects
Do not do anything unless there is scientific
agreement.
Do not do anything unless one person says the
effects are small
ARSENIC A DISASTROUS EXCEPTION
Animals did not get cancer from
arsenic
(till Jack Ng in Bribane fed them an organic compound)
Why should people get cancer?
The world was not looking
1880 Hutchinson cancers from long use of Fowler’s
solution
1930 Angiosarcoma in farmers
(NOT DUE TO VINYL CHLORIDE)
1965 skin cancers in Taiwan threshold suggested
1976 data from Taiwan ignored for many years
Risk at present EPA regulatory level
100 times level of anything else EPA regulate
Many millions exposed in SE Asia
What is the next disaster we are missing?
What about Fibers? (Asbestos)
Pliny - Asbestos means does not burn
(That is why we liked it before 1970)
1970 Asbestos means one of 6 materials in commerce
Bulk mineral not measurably carcinogenic
cleavage fragments probably weakly carcinogenic
fibers very carcinogenic
A PHYSICAL CARCINOGEN NOT A CHEMICAL ONE
Animal data not helpful
Are all 6 the same? No.
No alternative to EPIDMIOLOGY
2 groups distinguishable by electron microscope
Amphiboles - Serpentine (Chrysotile)
Chrysotile does not stay long in lung
Crocidolite stays 40+ years
This explains why crocidolite causes mesothelioma
(latent period 40 years)
and Chysoltile at least 200 times less.
1986 Few Electron microscopes available:
EPA assumed all 6 the same; regulated fiber counts only
2006 EPA realizes that there are differences.
Maybe in 2012 the lawyers will understand.
In Order of Nastiness:
Crocidolite
Amosite
Tremolite?
Chrysotile
lung cancer with 20 year latency
mesothelioma 40+ years latency
Is Carbon Burning Causing Climate Change?
Standard model
• 1750 Sun heats earth and earth reradiates
– Then temperature -70 degrees Celsius (200K)
• 1827 Fourier suggested we are in a greenhouse
• 1870 Tyndall measured CO2 and H2O infrared spectra
– CO2 uniform in atmosphere H2O non uniform
• 1897 Arrhenius calculated effect of ice ages
• 1945 Dobson pointed out deep and surface oceans do not
mix
• 1960 + Keeling measured CO2 at Muona Loa, Hawaii
– In detail still very uncertain
Each Step Needs Verification
• Sun’s radiation – sunspot cycle now suggests less radiation
than usual
• Do carbon emissions stay in air? (50%)
– Ocean-air interaction crucial
– (different in Atlantic and Pacific)
• Is upper atmosphere disconnected from lower atmosphere?
• THESE QUESTIONS ASSUMED POLITICALLY IN 1981
– Models never run with all uncertainties
• Reopened in 2000 by
– aggressively of some scientists
– bad bill in Congress (Waxman-Markey)
– (pork barrel with a thin veneer of climate change)
Difficult though it is to get scientific
agreement
Hard though it is to get a detailed risk
assessment from EPA that I like
It is harder to get the main body
politic to face the truth
Why Invade Iraq?
• Iraq is close to having nuclear weapons?
– Few at Los Alamos believed that they were
• They have a lot of chemicals/nerve gases?
Less < 1/1000 of US storage in Umatilla, Oregon
• Bush Jr. did not like Saddam?
(Saddam tried to kill his father)
The next slide gives the answer.
It is the Oil, Stupid
Topical issue – Fukushima a year agoEvacuate or not?
What do we know from previous incidents?
Those who do not understand history are condemned to repeat it
Medical X rays
Radium Dial painters
Hiroshima-Nagasaki
Windscale (UK)
TMI
Chernobyl
Tokai (Japan)
Avoid Acute Radiation Sickness (ARS)
ARS can occur if dose in a week is > 200 Rems (2 Sv).
But chronic doses can accumulate over years
Chronic effects:
cancer
heart disease
genetic effects
No cancers specific to radiation
Radiation can cause an increase
in natural cancer rate
Perhaps 30% at 200 Rems (maybe much less)
My Estimate of Dose at Site Boundary
My Dose at Site Boundary
March 2012 calculation
– Peak probably noble gases (little effect) followed by
cesium. (134 and 137)
– Take peak and multiply by width
– Dose about 2 Rems (0.02 Sv) and falling
– My CAT scan doses last year 2.4 Rems
What about Ibaraki?
(on way to Tokyo)
It looks different with different scale.
Dose Negligible
Adverse Effects of Evacuation
– Stress can increase cancer rates 5% or more
– Direct loss of life due to lack of facilities
– NO ONE CONSIDERED THESE AT ALL
Subsequent study (IAEA, INPO, TEPCO, etc)
says:
• No one got Acute Radiation Exposure
• (highest exposure in a worker 30 Rems)
• If there are cancers the probability will be less
than 50% in all but a few cases.
These are the locations of measurement
BEIR report on effects of radiation
Deposition to NW of plant highest from
Tuesday/Wednesday releases
Comparison with other disasters
What should a decision maker have
done (my personal opinion)?
• Order an immediate evacuation for many km ? NO
• Make it easy for any VOLUNTARY evacuation
– (to avoid panic as happened after Katrina (YES)
• Do simple steps to reduce exposure to fall out
– including measurement (YES)
• ORDER all out clean up for years (NO)
• facilitate individual towns and people in voluntary steps
(YES)
• Richard Wilson paper in DOSE_RESPONSE (2012)
(circulated)
Should US troops stay in Afghanistan?
They seem to cause trouble.
If an Afghan is angry what are his
alternative courses of action?