Day 022 - 12 Sep 94 - Page 51


     
     1        3 and tab 5, because they reach opposing conclusions, do
     2        they not?
     3
     4   MR. MORRIS:  One is about breast cancer.
     5        A.  And one is about colon cancer.
     6
     7   MR. MORRIS:  And one is about colon cancer.  I am going to come
     8        on to that, if I may.
     9
    10   MR. JUSTICE BELL:  There are slightly more women in the -----
    11        A.  This was a study of women.
    12
    13   Q.   Yes, but there were slightly more in the breast cancer
    14        than the colon cancer one?
    15        A.  I think it is probably just that the follow-up period
    16        of time was longer.
    17
    18   Q.   About 700 or so.
    19
    20   MR. MORRIS:  Can I just continue with the results from the
    21        abstract. It concludes:  "No association was found for
    22        vegetable fat. The relative risk of colon cancer in women
    23        who ate beef, pork or lamb as a main dish every dish every
    24        day, was 2.49 at a 95 per cent confidence interval, 1.24
    25        to 5.03, as compared with those reporting consumption less
    26        than once a month".  Is that a very high finding?
    27        A.  You can see that the confidence limits are very wide;
    28        they go from 1.24 to 5.03.  Basically what that means is
    29        that there might be a 24 per cent increased risk or there
    30        might be a fivefold increased risk.  You know, they have
    31        averaged it out.  The average is two-and-a-half fold risk.
    32
    33   Q.   I see.
    34        A.  But one worries about confidence intervals being so
    35        wide.
    36
    37   Q.   Is this a problem with cohort studies, though?
    38        A.  No, it does not only apply to cohort studies.  If
    39        anything, it is more likely to occur with case-controlled
    40        studies where you get very wide confidence limits.  The
    41        wider the confidence limits that you see, the more one is
    42        concerned about, you know, the validity of the results.
    43        I think this just emphasises the fact that over 88,000
    44        women have been investigated and still there are wide
    45        confidence intervals.
    46
    47   Q.   But the confidence intervals would be due to methodology,
    48        would they not?
    49        A.  No, that is due to statistics.  Basically, when they
    50        analyse the results what they are trying to say is, could 
    51        this event have occurred by chance?  Right?  That is what 
    52        the 95 per cent confidence interval means; it means that 
    53        in 5 per cent of people investigated it could have arisen
    54        by chance, but in 95 per cent of people these are the
    55        risks of an increased risk of colon cancer varying from 24
    56        per cent up to fivefold. It is really quite an enormous
    57        variation.
    58
    59   MR. JUSTICE BELL:  Is that related by some route or other to
    60        the fact that during over half a million person years of

Prev Next Index