Page 42 - Journal of Special Operations Medicine - Fall 2016

The thigh was laid on a laboratory benchtop and was Results were summarized by outcome and by tourni-
operated in accordance with the manufacturer’s instruc- quet model. The critical, or primary, outcome was effec-
tions. The thigh did not bleed, but bleeding was repre- tiveness (yes-no, bleeding control). Another important
sented by red lights that transilluminated the wound. outcome was absence of palpable pulse distal to the
The number of lights illuminated represented the bleed- tourniquet (yes-no). Secondary outcomes included time
ing rate: all 26 lights illuminated indicated uncontrolled to stop bleeding (seconds); pressure (mmHg) applied to
bleeding; few lights blinking indicated intermediate con- the skin by the tourniquet to achieve hemorrhage con-
trol; no lights illuminated indicated bleeding had stopped. trol, and the calculated volume of simulated blood loss
Arterial pulses were palpable in the popliteal area distal (mL). Pump counts were the number of squeezes of the
to the site of tourniquet application. Touchpad readouts bulb by iteration. Effectiveness and pressure were mea-
for each iteration included bleeding control, the time to sured by the manikin, whereas pump counts and pulse
stop bleeding, the total time of use, the pressure exerted stoppage were determined by the user.
under the tourniquet, and the simulated blood loss vol-
ume. Total time was the sum of time to stop bleeding Descriptive statistics were used to portray results. Cat-
and the “after time”; after time was the additional time egorical data (bleeding control and pulse stoppage
taken for more pumps of the bulb and user assessment of in 2-by-2 contingency tables) were analyzed with a
the pulse by palpation and bleeding by inspection. The chi-square test and the likelihood ratio p values were
measure of time to determination of bleeding control ex- reported. For pairwise comparison of models, a non-
tended from the start of the iteration until the manikin parametric Wilcoxon method was used. For pairwise
detected that no more blood was lost. Effectiveness was comparison of model means, the Tukey method was
determined by the cessation of blood loss (i.e., bleeding used. Significance for results was established at p < .05.
control [yes-no] at the time of determination). All statistical analysis was conducted using SAS version
9.2 (SAS Institute, http://www.sas.com/) and MS Excel
Iterations began with a tourniquet laid out flat and un- 2003 (Microsoft, https://www.microsoft.com). Parame-
done on the benchtop. Iterations ended when the user ters of reported data, such as differences between means,
touched the touchpad button, assessing that the bleed- were rounded to the decimal place as used routinely in
ing was stopped. The user tightened tourniquets until he either care or research.
perceived that simulated bleeding stopped. The casualty
had a medium build and the setting was Care Under Fire, Results
a setting resembling emergency care when under gunfire.
Each model had one device used for the entire experi-
The manikin settings also included a constant simulated ment, because no device needed to be replaced for any
hemorrhage rate (635mL/min). The touchpad reported reason such as breakage, wear, or tear. No safety issue
simulated blood loss volume as calculated from arterial occurred. Of note, the quality assurance information of
flow and time. With such a rate, the resulting bleed-out an imprint of manufacture date faded or wore off on the
time was 4 minutes, and in the absence of any hemor- TPT2 after use such that only the first two digits, 06,
rhage control, simulated death would occur then. If par- remained legible at the end of the experiment.
tial hemorrhage control occurred, then longer survival
would occur. Neither tourniquet effectiveness (p = 1; likelihood ratio,
0) nor pulse cessation (p = 1; likelihood ratio, 0) differed
Additional data were derived from results that were cal- among tourniquet models: all three models had 100%
culated from other data; one such example was bleeding (30 of 30 tests) for both outcomes.
rate (mL/s) as blood volume lost divided by the time to
stop bleeding; and another example was after time, the Differences in mean time to stop bleeding by model were
total time minus the time to stop bleeding. Two compos- statistically significant in one-way analysis as TPT2 (90
ite outcomes were made of existing data from five pa- seconds) was slower than the other models (64 seconds
rameters: effectiveness, time to stop bleeding, pressure, and 47 seconds for TPT3 and EMT, respectively; Fig-
blood loss volume, and breakage. The first composite ure 2). In pairwise comparison, the mean differences of
was set as a binary outcome (good-bad) and was good TPT2-TPT3 (25 seconds) and TPT2-EMT (42 seconds)
only if all five parameters had a favorable outcome (yes were statistically significant (p = .0027 and p < .0001,
effectiveness, <61 seconds, application pressure between respectively), but the mean difference of TPT3-EMT (17
151 and 499mmHg inclusive, blood loss of <500mL, seconds) was not (p = .055).
and no breakage of the tourniquet). The second com-
posite outcome was simply the count (0 to 5) of the five Differences in mean total time by model were statisti-
parameters if they were favorable. Tourniquets, itera- cally significant in one-way analysis: TPT2 (110 sec-
tions, and outcomes were uniquely identified. onds) was slower than the other models (77 seconds and

24 Journal of Special Operations Medicine Volume 15, Edition 3/Fall 2016