Ch. 10 — The Question of Risk • 203 
the c'oiK'lusion is that noi'mal hormone le\els 
would change by less than 10 percent. Similar 
coiulitions toi’ interferon production could 
release a[)pro.\imately 70/ig or the ma.ximum 
dail\ dose currently used in cancer therapy, 
l.ong-term effects of such e.xposure ai'e current- 
ly unknow ti: therefore, experiments using high- 
producing strains (10® molecules per cell or 
more) aiv likely to he monitored if such strains 
e\ er hecome a\ ailahle. 
The .\IH program of risk assessment, which 
was formalK started in 1979, continues to iden- 
tity possible consetiuences of rDN'.A research. 
L'nder the aegis of the National Institute of 
AllergN’ and Infectious Diseases, the progi'am 
supports research studies designed to elucidate 
the likelihood of harm.’ In addition, it collates 
general data from other experiments that might 
he rele\ant to risk assessment. Other risk as- 
sessments are being conducted by European 
organizations” and by the L'.S. Environmental 
Protection .-\genc\' to assess the consec|uences of 
releasing micro-organisms into the en\iron- 
ment. 
Thus far. there is no compelling ex idence that 
E. coli K-12 bacteria carrying rDN.A will be more 
' hazardous than any of the micro-organisms 
I which serxed as the source of D\,A. Nexer- 
! theless, all the experiments hax e dealt with one 
I genus of bacterium. Unless the conclusions 
about £. coli can be extended to other organisms 
likely to be used in experiments (such as Bacillus 
subtilis and yeast), other assessments may be ap- 
propriate. 
’E.\tramural efforts were first conceived in the summer of 1975 
to develop and test safer host-vector systems based on £. coli, the 
interagency agreement entered into with the ,\a\al Biosciences 
Laboratory tested £. coli systems in a series of simulated 
accidental spills in the laboratory. .Xt the Uni\ersity of .Xlichigan 
the survival of these systems was tested in mice and in cultural 
conditions simulating the mouse gastrointestinal tract. Tufts 
Lniversity tested these systems in both mice and human 
volunteers. Finally, the surv ival of host-\ector systems in sewage 
treatment plants was tested at the Unh ersity of Te.xas. The peak 
year for costs of supporting research contracts was 1978; over a 
half-million dollars were required. Currently, the cost of 
maintaining the high containment facility at Frederick, Md., is 
between S200.000 and S250.000 annually. 
••First Report to the Committee on Genetic E.\perimentation . a 
scientific committee of the International Council of Scientific 
Unions, from the Working Group on Risk .Assessment, July 1978. 
Perception of risk. 
Tbe probability of damage can be estimated 
for xarious exents. Tbe entire insurance in- 
dustry is based on the fact that unfavorable 
exents occur on a regular basis. The number of 
people dying annually from cancer, or automo- 
bile accidents, or homicides can he predicted 
fairly accurately. These estimates depend on 
the ax ailahility of data and the assumptions that 
the major determinants do not change from 
year to year. 
But ex en if the probability of damage is fairly 
well knoxxn, a gap often exists between this 
"real” probability of occurrence and the "per- 
ceixed” probability. Txxo factors that tend to af- 
fect perceptions are the magnitude of the possi- 
ble damage and the lack of individual control 
ox er exposure to the risk. Both of these are sig- 
nificant factors in the fears associated xvith 
rDN.A and the manipulation of genes. Because 
intuitixe exaluations can contradict analytical 
exaluations, the question of risk cannot be re- 
solx ed strictly on an analytical basis. Its resolu- 
tion xx'ill have to come through the political 
process. 
BURDEN OF PROOF 
The possibility of inadvertently creating a 
dangerous organism does exist, but its prob- 
ability is lower than was originally thought. 
Nevertheless, an important principle emerges 
from the debate. Society must decide whether 
the burden of proof rests xvitb those who de- 
mand evidence of safety or with those who de- 
mand evidence of hazard. The former would 
halt experiments until they are proved safe. The 
latter xvould continue experiments until it is 
shown that they might cause harm. 
A significant theoretical difference exists be- 
txveen the tw o approaches. Evidence can almost 
alxx^ays be provided to show that something 
causes harm— e.g., it can be demonstrated that a 
poliovirus causes paralysis, that a Pneumococcus 
causes pneumonia, that a rhinovirus causes the 
common cold. However, it cannot be demon- 
strated that a poliovirus can never cause the 
common cold. It cannot be demonstrated that 
rDNA molecules will never be harmful. It can 
