#
Research and Critical Skills

#
Solution to MCQ Self Test 3

1. TTTTF 2. TFTTT 3. TTTFF 4. TTFTT

Score +1 for a correct answer, -1 for an incorrect answer, and 0 if you did not
answer. This will give a total mark out of 20. You can rate your performance
as follows:- below 7: work much harder, 8-9: work harder, 10-13: adequate,
14-16: very good, 17-18: excellent, 19 or more: take over the class.

## Notes on the questions

1. (a) The pooled estimate is the meta-analysis odds ratio summarising the
three studies. (b) The difference between 0.1 and 1 is the same as the
difference between 1 and 10, i.e. a factor of 10 is represented by a constant
difference on the scale (c) This is the convention for this style of figure.
(d) The difference is significant if the confidence interval does not include
1.0, the null hypothesis value of the odds ratio (e) On the contrary, we should
try to get all studies which have been carried out, whether published or not,
to avoid publication bias.

2. (a) We can have three treatments in random order, or four, or even more,
though it gets difficult to organise and to keep the subjects interested. (b)
The order is randomized. (c) because the variation between patients is removed,
we have a much more accurate treatment comparison and so need fewer patients.
(d) the treatment should not change the disease, because then we could not
apply the second treatment, so this design is best for chronic diseases rather
than acute. (e) We compare the results on both treatments for the same patient.

3. (a) We need this so that we know which studies to analyse. (b) This is
the usual starting point (c) We need to get all the studies, whether published
or not, otherwise we may have a biased sample of the available. Studies are
more likely to be published if they show significant differences (publication
bias) (d) Unpublished studies are less likely to include significant treatment
differences than published ones, so these differences are likely to be smaller.
(e) Papers published in English are more likely to include significant
differences, another aspect of publication bias.

4. (a, b) Each circle and horizontal line represent the observed difference
and its confidence interval. (c) Studies with long horizontal lines have wide
confidence intervals and so should be given least weight (d) The confidence
interval for the combined estimate does not include zero (e) All the confidence
intervals for the individual trials include zero.

**Back to MCQ and EMI Self Test 3**.

This page maintained by Martin Bland.

Last updated: 29 June, 2004.