Sanders et al address an important but neglected topic in their paper
(1), however we have concerns about the measurement of the trial’s primary
outcome. From the information given it appears that perineal pain was
measured using a direct estimation method based on one item, probably a
visual analogue scale.
Our first concern is that a one item measure is prone to measurement
error as the reliability of a scale is related to the number of items it
contains (2). Secondly, direct estimation methods have a response bias
problem known as end-of-scale aversion, where respondents are reluctant to
put responses at the end of the scale (2). These two problems are of
concern in the context of a clinical trial: the effect of a reduced scale
makes it harder to detect differences in scores between groups and the
inherent unreliability of one item compounds the problem.
Thirdly, it is not stated whether any procedures to establish the
validity of the measure were carried out prior to the trial. The most
obvious danger is that the scale is measuring the pain of second stage
labour generally as opposed to perineal pain specifically.
Our final concern is with the estimate of a clinically significant
reduction in pain as 13 scale points or 0.5 standard deviations (1). A
difference of 0.5 standard deviations has previously been suggested as an
estimate of minimally clinically important difference in relation to
health-related quality of life (3), however this is not a universally
accepted rule of thumb (4). Applied to a measure of undocumented validity
and reliability, it may be spurious.
Given the problems outlined above, it is entirely possible a true
difference in perineal pain existed between the two groups but was not
detected.
1) Sanders J, Peters TJ, Campbell R. Effectiveness and acceptability
of lidocaine spray in reducing perineal pain during spontaneous vaginal
delivery: randomised controlled trial BMJ doi:10.1136/bmj.38878.833241.7C
2) Streiner and Normal Streiner DL, Norman GR. Health measurement
scales: A practical guide to their development and use. 2nd Ed. Oxford:
Oxford University Press, 1995.
3) Norman GR, et al. Interpretation of changes in health-related
quality of life: the remarkable universality of half a standard deviation.
Med Care 2003, 41:582-92.
4) Beaton DE. Simple as possible? Or too simple? Possible limits to
the universality of the one half standard deviation. Med Care 2003, 41:593
-96.
Competing interests:
None declared
Competing interests:
No competing interests
20 July 2006
Geraldine Barrett
Senior Lecturer
Sally Spencer
Brunel University, School of Health Sciences and Social Care, Uxbridge UB8 3PH
Rapid Response:
Concern about measurement of the primary outcome
Sanders et al address an important but neglected topic in their paper
(1), however we have concerns about the measurement of the trial’s primary
outcome. From the information given it appears that perineal pain was
measured using a direct estimation method based on one item, probably a
visual analogue scale.
Our first concern is that a one item measure is prone to measurement
error as the reliability of a scale is related to the number of items it
contains (2). Secondly, direct estimation methods have a response bias
problem known as end-of-scale aversion, where respondents are reluctant to
put responses at the end of the scale (2). These two problems are of
concern in the context of a clinical trial: the effect of a reduced scale
makes it harder to detect differences in scores between groups and the
inherent unreliability of one item compounds the problem.
Thirdly, it is not stated whether any procedures to establish the
validity of the measure were carried out prior to the trial. The most
obvious danger is that the scale is measuring the pain of second stage
labour generally as opposed to perineal pain specifically.
Our final concern is with the estimate of a clinically significant
reduction in pain as 13 scale points or 0.5 standard deviations (1). A
difference of 0.5 standard deviations has previously been suggested as an
estimate of minimally clinically important difference in relation to
health-related quality of life (3), however this is not a universally
accepted rule of thumb (4). Applied to a measure of undocumented validity
and reliability, it may be spurious.
Given the problems outlined above, it is entirely possible a true
difference in perineal pain existed between the two groups but was not
detected.
1) Sanders J, Peters TJ, Campbell R. Effectiveness and acceptability
of lidocaine spray in reducing perineal pain during spontaneous vaginal
delivery: randomised controlled trial BMJ doi:10.1136/bmj.38878.833241.7C
2) Streiner and Normal Streiner DL, Norman GR. Health measurement
scales: A practical guide to their development and use. 2nd Ed. Oxford:
Oxford University Press, 1995.
3) Norman GR, et al. Interpretation of changes in health-related
quality of life: the remarkable universality of half a standard deviation.
Med Care 2003, 41:582-92.
4) Beaton DE. Simple as possible? Or too simple? Possible limits to
the universality of the one half standard deviation. Med Care 2003, 41:593
-96.
Competing interests:
None declared
Competing interests: No competing interests