University of York

Department of Health Sciences

Applied Biostatistics

Exercise: Nasal diamorphine

A multicentre randomised controlled trial was carried out to compare the effectiveness of nasal diamorphine spray with intramuscular morphine for analgesia in children and teenagers with acute pain due to a clinical fracture (BMJ 2001; 322: 261-265).  In the emergency departments of eight UK hospitals, patients aged between 3 and 16 years presenting with a clinical fracture of an upper or lower limb were randomised to one of the treatments. The main outcome measures were patients’ reported pain using the Wong Baker face pain scale, ratings of reaction to treatment of the patients and acceptability of treatment by staff and parents, and adverse events.  404 eligible patients completed the trial (204 patients given nasal diamorphine spray and 200 given intramuscular morphine).

The authors stated that ‘The distribution of pain scores for the spray group was lower than that for the intramuscular group at 5 (P=0.04), 10 (P=0.003), and 20 minutes (P=0.002) after treatment, but no different after 30 minutes (P=0.20)’.

(a)   What can we conclude about pain score after 10 minutes?

The pain scores after 30 minutes were as follows:

 Pain score Nasal diamorphine group Intramuscular morphine group n (%) n (%) 1 54 (28) 48 (25) 2 65 (34) 61 (32) 3 36 (19) 42 (22) 4 24 (13) 23 (12) 5 7 (4) 11 (6) 6 5 (3) 8 (4)

(b)        Is it correct to conclude that pain scores were no different after 30 minutes?

The following graph was used to present the data on reaction to the treatment:

(c)   What kind of variable is reaction to the treatment?  What does this graph actually show?

(d)   How could the graph be improved?

Solution

(a)   What can we conclude about pain score after 10 minutes?  The difference is highly significant, so we conclude that there is strong evidence that pain score is reduced when nasal administration is used.

(b)   Is it correct to conclude that pain scores were no different after 30 minutes?  We cannot correctly conclude that pain scores were no different after 30 minutes.  The difference is not significant, but this means that we have not shown that there is a difference, not that we have shown that there is no difference.  In fact a very small difference can be seen in the data, the intramuscular group reporting more pain than the nasal group.

(c)   What kind of variable is reaction to the treatment?  We divide the children into separate categories, ‘no discomfort’, ‘mild reaction’, ‘winced’, ‘cried’, and ‘screamed’.  These have no numerical value, we therefore have a qualitative or categorical variable.  The categories are clearly ordered, from ‘no discomfort’ to ‘screamed’.  We therefore have an ordered categorical variable, or ordinal variable.  What does this graph actually show?  The graph shows the cumulative reaction.  The first pair of bars shows the proportions reporting no discomfort, the second pair shows the proportions reporting no discomfort or mild reaction, the third pair shows no discomfort or mild reaction or winced, and so on.  The fifth pair shows all children and is 100% by definition.

(d)   How could the graph be improved?  What we want to see is the proportions showing each reaction:

This shows much more clearly that about 20% of the intramuscular group screamed and none of the nasal group.  If we want a cumulative graph, which I don’t think we do, we should start with screamed, then screamed or cried, then screamed or cried or winced, etc.:

Now the third pair of bars show the percentage who winced or worse.