Day 252 - 20 May 96 - Page 24


     
     1        number of people to study.  What is, then, the minimum that
     2        you consider to be a satisfactory number?
     3        A.   Well, nowadays, if one is thinking are carrying out an
     4        investigation on human subjects one has to look at the
     5        variable that you are measuring, which would be blood
     6        cholesterol.  You have to look at the standard deviation
     7        within the population, which is a measure of the
     8        variability and statisticians then work out how many people
     9        you would need to use in the study to achieve the level of
    10        significance what was acceptable.  This is a rather
    11        complicated procedure which I usually get somebody else to
    12        do for me, but they will end up by saying you will need 103
    13        people in your study, minimum, or 22, depending on how
    14        variable the factors that you are measuring. Obviously, the
    15        greater variability in the population the larger number of
    16        people you will have to use to demonstrate a difference.
    17
    18   Q.   Because one of your studies that was disclosed was on ten
    19        people was it not?
    20        A.  Yes.
    21
    22   Q.   But you would regard that as satisfactory?
    23        A.   Yes, indeed.  I think, if I could explain that, we
    24        were not comparing these ten people with another ten
    25        people.  We were looking at these ten people.  Making
    26        measurements on them, and doing something to them and
    27        making more measurements.  Rather like the one with 20
    28        subjects in it that you saw here, but ten people was more
    29        than adequate to demonstrate what we were doing in that
    30        particular piece of work.
    31
    32   Q.   The one with 5 people, that was not comparing them with
    33        other people, was it?  That was just within the 5?
    34        A.   Yes, yes, it was looking at their response to being
    35        given something.
    36
    37   Q.   So, that would fall in the same category as your own
    38        research?
    39        A.   If you are looking specifically at that paper, we were
    40        comparing two breakfasts, one very high in fat and one
    41        virtually fat free.  In doing a study like this what would
    42        have made this a better study -- if I am looking at the
    43        right one -- yes, is -- no, the number was really too
    44        small.  Ideally, if you are carrying out two different
    45        procedures on a group of people what you do is split the
    46        group and carry out both procedures simultaneously and then
    47        at the end of your measurements you swop them over and
    48        those that have been on a high fat diet would go on low fat
    49        diet and those that had been on the low fat diet would go
    50        on the high fat diet.  This is called crossover study.  The 
    51        advantage of doing that is that you can take account of 
    52        factors which could influence your results.  For example, 
    53        if you are doing a study on soups and you were you given a
    54        high fat diet and made measurements, and then after that
    55        give them a low fat diet, it could be that the low fat diet
    56        coincides with when they are sitting examinations or when
    57        environmental temperature has gone up by ten degrees, and
    58        that could influence their behaviour and it could influence
    59        the result.
    60

Prev Next Index