Day 017 - 25 Jul 94 - Page 22


     
     1        value, which is just another letter for a statistical
              significance, where the P value is 0.05 meaning that the
     2        relationship is statistically related in the sense that it
              could not have arisen in more than five per cent of cases
     3        purely by chance.
 
     4        As far as correlation coefficients are concerned, they are
              measuring the strength of this relationship and they need
     5        to be under one to have a meaningful relationship.  The
              further they are below one the better.  So this
     6        correlation coefficient of .89 you see is significant.
 
     7   Q.   At face value, as Dr. Kinlen says, it is very striking, is
              it not?
     8        A.  It is indeed.
 
     9   Q.   One could be misled, could one not, if one looked at that
              in isolation and uncritically, into supposing that it was
    10        established that saturated fat or total fat was a cause of
              breast cancer, could one not?
    11        A.  One could say they are related to one another.  I do
              not think even from this that one could say one is
    12        providing evidence that there is a causal relationship.
              These studies are observational.  They are not
    13        establishing a cause and effect relationship.
 
    14   Q.   May I read on: "A crucial question is whether this
              impressive relationship is direct or indirect.  Countries
    15        with a high fat consumption are distinctive in other
              respects, including certain established risk factors for
    16        breast cancer.  In theory, therefore, these might be
              producing the geographic correlations with fat.  Hems and
    17        Gray and his colleagues explored this question using
              international data on marital status, childbearing,
    18        height, weight and age at menarche. After adjustment for
              these factors, however, there remained an appreciable
    19        residual correlation with fat".
 
    20        We did not mention height, I do not think, a moment ago
              when you were telling us what you saw as being, as it
    21        were, important factors to consider in the aetiology of
              breast cancer.  If you turn back a tab to tab 8,
    22        Dr. Arnott, one sees a study done on Norwegian women in
              1990 by Vatten and Kvinnsland.  Do you have that?
    23        A.  Yes.
 
    24   Q.   Which, tell me if I am wrong, I am not going to read it,
              appeared to come to this conclusion:  That there was a
    25        positive association between height and incidence of
              breast cancer in women? 
    26        A.  Yes. 
  
    27   Q.   But that this positive association was shown most clearly
              or was confined to, I should say, those women studied in
    28        the group who had been subjected to nutritional
              deficiencies before and at the time of and just after the
    29        first onset of menstruation.  Have I got that right?
              A.  Yes.
    30
         Q.   The peripubertal period, the reason being that they were

Prev Next Index