Revival: Biological Effects of Low Level Exposures to Chemical and Radiation (1992)

120 BIOLOGICAL EFFECTS OF LOW LEVEL EXPOSURES

Table 7.1. Possible Results for a Study

Actual Effect

Study Result

Harmful Neutral Beneficial

Harmful TH FN FL

Neutral FH TN FL

Beneficial FH FN TL

Unspecified H N L

study, measure of outcome employed, spontaneous rate of the disease, and

the number and importance of other potential confounding variables that

may be present.4 Some of these that are under the investigators’ control will

be examined for the effects that their manipulation prior to beginning a

study may have on the probability of correctly classifying a study outcome.

For any study the possible results may be arrayed as in Table 7.1. H, N,

and L denote harmful, neutral, and beneficial, respectively, and T and F

denote true and false. Thus, if a study of a substance whose effects are

actually beneficial results in a statistically significant harmful result, the

study result is a FH (false high); if the actual effect is unspecified, then the

result is simply H. Whatever the actual effect of a substance, the result of

the study will be either H, N, or L. The probabilities of these three events

thus sum to unity for each of the four scenarios in the rows of the table.

In testing for beneficial effects, it is essential that there be a high sponta­

neous rate in the unexposed group, so the beneficial effect can manifest

itself. For if the disease outcome were extremely rare, then even a zero

disease rate in the exposed subjects would not be unusual, and the sensitiv­

ity of the study to detect beneficial effects would be greatly impaired. In

such cases either more sensitive subjects might be found, or case-control

studies might be employed since these may assure sufficient numbers of

disease cases. Failing these or other techniques to increase sensitivity to

detect a TL, serious consideration should be given to abandoning the

study.

Data Layout: Two-by-Two Tables

In many studies the relevant study data can be put in the form of a two-

by-two table, as shown by Table 7.2. The study subjects are presumed to

have been randomly allocated to the exposure groups: nl in the exposed

group and n0 in the nonexposed comparison group. These are then exposed

Table 7.2. Sample Two-by-Two Table

Exposure

Disease

Yes No Total

Yes a b ni

No c d no

Total mi m0 n