Day 022 - 12 Sep 94 - Page 55


     
     1        results the analysis was restricted to 846 cases, so they
     2        excluded a number of cases, and 862 controls.  In other
     3        words, less than 80 per cent of patients actually
     4        responded to the questionnaires.
     5
     6        Now, you wonder, what about the other 20 per cent?  The
     7        difficulty with these studies, as I say, is the
     8        methodology and the number of people who actually do not
     9        respond.    With a 20 per cent drop-out rate in this way,
    10        you do wonder what influence that 20 per cent would have
    11        if their results were available for the overall data.  If
    12        I run a clinical trial and could not collect 20 per cent
    13        of the data, my colleagues would not think very much of
    14        that clinical trial I was running.
    15
    16   Q.   I will move on from that one.
    17
    18   MR. JUSTICE BELL:  Is there a line at which someone carrying
    19        out an investigation like this, a percentage which is
    20        thought to be necessary to give reliability or not?  You
    21        say 80 per cent is not; that seems reasonably high to me.
    22        It is more than anyone ever votes in an election in this
    23        country, for instance.
    24        A.  That is true.
    25
    26   Q.   Is there an accepted level which is the water-shed for
    27        reliability?
    28        A.  I think most people would like to see 90 per cent
    29        responses.  You like to see 10 per cent or less of
    30        non-respondents.  If you are not getting evidence from a
    31        fifth of the population you are trying to examine, that 20
    32        per cent, if their results were different, could have a
    33        tremendous influence on the overall results of the
    34        study.
    35
    36        So, one wants to try to reduce that as low as possible,
    37        preferably to no more than 10 per cent.  This is one of
    38        the major problems with case-controlled studies, that
    39        often the controls who actually participate are people who
    40        are enthusiasts.  They may be people who are interested in
    41        diet, therefore, feel they would like to respond to the
    42        questionnaire.  They may have other factors which single
    43        them out.  Therefore, although they may ostensibly appear
    44        of the same age, the same sex and so on, one just wonders
    45        are there other factors which may be different about them
    46        which may explain some of the results that one finds?  It
    47        is a real dilemma in these studies.
    48
    49   MR. MORRIS:  Can we move on from that study?  Can I back track,
    50        you remember the nurses study? 
    51        A.  Yes. 
    52 
    53   Q.   I am sure you are very well aware of it.  One point you
    54        mentioned about the identification of breast cancer was
    55        that the fat content, the levels in the whole population
    56        of the study were very high.  In fact, I believe that the
    57        whole study, the medium fat consumption at the start of
    58        that study was 39 per cent of total daily calories,
    59        indicating that it was about -- the population all had
    60        high fat content in their diet.  Is that correct?

Prev Next Index