[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

HP Newsletter - Effects of Low Doses



My copy of the September Newsletter just arrived. It contains an
interesting piece by Dr. Lars Perrson of Sweden defending the linear-no
threshold model. I really don't want to restart the debate on LNT again. I
have a couple of questions about epidemiological studies in general.

Dr. Perrson refers to the atomic bomb data, indicating that a "substantial"
risk increase is evident at 20 rem and above. He also says that a more
recent study (Pierce et al., 1996) found "substantial" risk at 5 to 10 rem.
Later he says that the Oxford Study of Childhood Cancer (1970) found
increased cancer among children up to age 15 following in utero exposure at
dose between 1 and 2 rem.

When I teach dosimetry techs about background subtraction for personnel
dosimeters, I tell them that, because we subtract an average background
value from each individual TLD measurement, we get both positive and
negative results. Since most of the people we monitor receive no measurable
dose, the net readings for a large group of measurements (e.g., end of
month or end of quarter) should look like a normal distribution centered
about zero. There should be about as many small positive values as small
negative values. The techs can look for this pattern as a "sanity check" on
the data. Finding no negative values or too many negative values would be
indicative of a bias of some type in the measurement system.

Doesn't this idea apply to the aggregate of low-dose epidemiological
studies? I have not (personally) compiled a summation of study results, but
it seems to me that I've seen about as many negative coreletion results as
positive, and most can't find any corelation. If no relationship existed
between low doses and cancer risk, that's exactly what I would expect from
a variety of studies where the ambient cancer rate is significant.

It seems to me that a citation of one particular study from the multitude
of studies is selecting only the data that fit the hypothesis, and excludes
the data that don't. Am I over-simplifying?

In general use, the definition of what constitutes a "substantial" risk of
cancer depends on the listener (how afraid is the person of dying at all,
how afraid of cancer, etc). So, when one uses the term "substantial" in
desribing the findings of a study, does this have a specific definition in
the epidemiology world?


Bob Flood
Stanford Linear Accelerator Center
(415) 926-3793     bflood@slac.stanford.edu
Unless otherwise noted, all opinions are mine alone.