Day 017 - 25 Jul 94 - Page 24


     
     1        diagnosis and, in addition, a dietary history covering
              frequency and quantity of fat consumption six months
     2        before interview, which for most of the cases would have
              been close to the diagnosis of breast cancer.  The 24 hour
     3        recall (which gave the lowest estimates of intake of the
              three dietary methods) alone gave significant case-control
     4        differences in the study overall, with consumption of
              total calories and saturated fat being greater among the
     5        cases: total fat intake, however, though higher, was not
              significantly higher than among the controls. No
     6        appreciable dose-response relation with fat was found by
              any method.
     7
              The study by Lubin and his colleagues (1981), also carried
     8        out in Canada, reported a higher intake of beef, pork and
              butter among women with breast cancer than population
     9        controls. However, evaluation is made difficult by the
              fact that a large number of controls failed to cooperate
    10        (28%) and those who did, unlike the cases" -- those are
              the people who have the disease, are they not?
    11        A.  That is correct.
 
    12   Q.   "...  were interviewed at home, thereby providing
              opportunities for bias".  May I pause there?  Does that
    13        illustrate one of the problems with case controlled
              studies, that you cannot guarantee that the controls are
    14        going to do what is asked of them?
              A.  No.  As I said earlier on, often the controls are
    15        found by mailshot and these people are derived from a
              register of the population, and what they try to do is to
    16        select patients or people who are of a similar age
              distribution.  For example, in breast cancer they are all
    17        women, of course, but similar age distribution to those
              who actually have the disease.
    18
              Now, in a study where 28 per cent of people did not
    19        respond or did not participate, one, therefore, wonders
              how relevant the results can be in the remaining 72 per
    20        cent of patients of people when you actually have a
              sizeable proportion actually not participating in the
    21        study whom you have matched to the patients who actually
              have the disease.  So, there is a big slice of information
    22        missing from such a study.  It, therefore, inevitably
              casts doubt on the validity of the conclusions which may
    23        be drawn from that particular study.
 
    24   Q.   Without anything sinister in mind, I would like to leave
              out the rest of that paragraph, except if one reads it,
    25        one can see how difficult it is to interpret it.  At the
              start of the paragraph:  "Over a period of several years", 
    26        can I start there? 
              A.  Yes. 
    27
         Q.   Have you got that?
    28        A.  Yes. "Over a period of several years all women
              admitted with suspected cancer to the Rosewell Park
    29        Memorial Institute in Buffalo, New York, were routinely
              questioned about their diet.  Graham and his colleagues
    30        (1982) analysed the data collected on fat consumption from
              a large group of women with a confirmed diagnosis of

Prev Next Index