Re: Help needed in analyzing resuls of a clinical study

The follwing result is given:

Group Name N Mean Std Dev
----- ---------------- -- ----- -------
A Trauma severe BI 18 1.828 0.618
B Trauma mild BI 16 1.266 0.529
C Trauma no BI 12 1.471 0.741
D Control 19 1.003 0.247

I would interpret this result the following way (while mostly ignoring
p-values):
Group A have higher mean level of EnM than group D.
If we regard this comparison as the primary issue to resolve, we have a
relatively certain result. The only concern would be the question why A
has larger variation than D.
To examine this we could look at the distribution of EnM values in both
groups.
It might reveal wether cases in group A show a general drift towards
higher vaules,
or if the higher mean of in A are caused by a few extreme cases.
The former situation is suggested by the large difference in variance
and it would be
interesting to see a complete plot of all data.
Or a continguency table giving a categorized distrubution, for example
in quartiles or tertiles, might give valuable information in that
direction.

The difference between B+C and D or between B+C and A is harder to
evaluate as there seem to be a larger effect in EnM in the group
without injury than in the group with mild injury. However, the
difference is not large, and it is reasonable to assume that group B
and C are approximately on the same level.
If we assume that we can pool B+C, we have:


Group Name N Mean Std Dev
----- ---------------- -- ----- -------
A Trauma severe BI 18 1.828 0.618
B+C Trauma no/mild BI 28 1.354 0.6 (approx)
D Control 19 1.003 0.247


This would lead to the following two conclusions:
1. Severe TBI is associated with higher levels of EnM, although it is
not yet
decided if it is due to a general shift in mean or median level, or
an
increased risk of high EnM among certain cases, or both.

2. There seem to be an dose-response relation in this data so that
severe injury is associated with higher levels than non-servere
injury,
and non-servere injury is assocated with higher levels than
controls.

Power is always an issue when comparing groups of modest sizes, so I
would suggest that you perform a proper power analysis before
increasing sample size, and that may require that you consult a
professional statistician.

Valter Sundh


Bangali skrev:

Scenario and hypothesis: We believe that traumatic brain injury (TBI)
induces a strong systemic inflammatory response that can be measured by
the serum levels of enzyme M (EnM). To explore this hypothesis we
measeured the basal levels of EnM within 12 hours of injury in a group
of severe TBI , Patients with Trauma and no TBI , Mild-TBI and a
Control group of normal volunteers .

To see if the means were different we conducted a one-way ANOVA with
the following results for EnM:

Group Name N Missing Mean Std Dev
Severe TBI 18 0 1.828 0.618
Trauma no TBI 12 0 1.471 0.741
Mild TBI 16 0 1.266 0.529
Control 19 0 1.003 0.247

Normality Test: Passed (P = 0.257)
Equal Variance Test: Passed (P = 0.056)

There is a statistically significant difference among groups (P =
<0.001). Pairwise Multiple Comparison Procedures (Bonferroni t-test)
gave the following results:

Comparison Diff of Means t P
Severe TBI vs. CONTROLS 0.826 4.643
<0.001
Severe TBI vs. MILD TBI 0.562 3.025
0.022
Severe TBI vs. Trauma NO TBI 0.357 1.774
0.486
Trauma NO TBI vs CONTROLS 0.468 2.348
0.133
Trauma NO TBI vs MILD TBI 0.205 0.991
1.000
MILD TBI_B vs. CONTROLS 0.264 1.437
0.935

Questions:

1)Our conclusions are that severe TBI induces a severe systemic
inflammatory response that is significanly different than in controls
and mild TBI. Do you think this conclusion is correct according to
the data we present?
2)We believe that Trauma NO TBI also seems to induce a less strong
inflammatory response compared with controls and MILD TBI. However the
results are not significant becasue the goup is to small. Do you agree
with that statement?
3) If we want to increase the power of the group to test for this
differences would it be correct to increase the sample size just for
this group (trauma NO TBI).

Any help in interpreting this results, suggestions and criticisms
(even destructive criticims) would be greatly appreciated. I apologize
for the long post, but I have learned a lot from this forum.

.