Devil’s Vassiios
Raptopoulos,
MD
Mitchell
P. Fink,
#{149}
Advocate
MD
CT Grading of Splenic Trauma In Adults: How the Same Statistics Can Be Interpreted Differently’
W
with interest the article by Umlas and Cronan (1) in which previously described computed tomographic (CT) grading systems for splenic trauma (2-4) were retrospectively evaluated. We agree with their conclusion that although CT is useful, the “ultimate decision for laparotomy
the upper right side of the abdomen (perihepatic fluid) and the pelvis. Half grades can be given, the maximum score is 6, and patients with a score of 2.5 or greater are at higher risk for delayed splenic rupture than those with a
E read
should be based on clinical status and not radiographic findings” (1). Unfortunately, the initial clinical course of most patients with delayed splenic rupture is innocuous, and only eventually do they
experience terioration It is the
sudden,
usually
stormy,
of their hemodynamic majority of these patients
the grading
systems
In regard
and
the
presence
of
of Massa-
is graded
from in the
perisplenic
system
was modified.
comments
sity
0 to 3 upper
tomography (CT), Computed tomography Computed tomography
(CT), preoperative (CT), utilization #{149}Devil’s Advocate CT, 775.1211 #{149}Spleen, injuries, 775.41 Trauma, Radiology
Spleen, #{149}
775.41 1991;
180:309-311
‘From the Department of Radiology, Beth Israel Hospital, 330 Brookline Aye, Boston, MA 02215 (V.R.); and the Department of Surgery, University of Massachusetts Medical Center, Worcester, Mass (M.P.F.). Received March 21, 1991; accepted May 3. Address reprint requests
to yR. ©
RSNA,
1991
that
the
of Massachusetts
this misinterpretation, nan
in
Later,
system
made
things
was
the revised
not substantiated from the Univer-
(3,4). Because Umlas and more complicated
frying to differentiate sular and pericapsular
These
of Croby
between subcapfluid. This may
in the
first
in none
frequencies
of 23
are
not
sig-
nificantly different. Similarly, if patients with a safe score are considered, the occurrence of false-negative predictions with CT scoring in the three studies is not
very
different:
two
of 15, none
of
eight, and one of 33 patients (1,3,4). Umlas and Cronan noted that their patients with delayed rupture had a lower mean score (1.66) than those who required early celiotomy (estimated score from the presented data, 3.44). In latter
salvage
of
and
(3) this occurred
patients.
the
Scatamac-
the University
Therefore,
(1) are erroneous and by reading the articles
Computed
effectiveness
(1,3).
of 64 patients,
study
group,
splenectomy than those
Massachusetts, tested this same scoring system, retrospectively and prospectively. Nowhere in this later article do the authors indicate that the original
those
who
had a lower who underwent
operation
required
score (3.34) a splenic
(3.66).
Although
we
do not have access to all their data, it is almost certain that the difference between a score of 3.34 (splenectomy) and a score of 3.66 (splenic salvage operation) is not statistically addition, it does not
significant. In seem appropriate
to select a subgroup of patients in whom the initial decision not to operate was wrong and compare their mean CT score
of 1.66
with
the decision mated
score,
data
that
of those
in whom
to operate was correct (esti3.44). Instead, the same in our view, be analyzed as
should,
follows: Among 29 adult patients studied by Umlas and Cronan, the mean score of those who required surgery
(early or late) was 3 while were successfully treated
those who nonsurgically
weakness of the system. and Cronan are skeptical of the utility of splenic CT because of two patients with delayed splenic rupture, one of whom had a normal spleen at CT
had
This
and the other splenic injury.
nally,
increased
should terms:
fluid”
chia et a! (4), also from
have Index
In that
is “any intrafluid, except
left side of the abdomen (perisplenic fluid) is graded from 0 to 1, as is fluid
clinical
Cronan.
score equals 1 when there abdominal intraperitoneal
or a combination
of fluid
and
to identify,
to the University
injury
by Umlas
CT scorin the
that false-nega(5-8). This fact
chusetts grading system (3,4), Umlas and Cronan may have expected a bit more from the scoring than the system was intended for. Grading splenic trauma should be kept simple: The pa-
renchymal
article
and presented the as a table, reproduced
score equals fluid fluid”
mandates close monitoring of patients for whom nonsurgical treatment is chosen, whether this is based on clinical criteria, CT grading, the two.
parameters ing only
table, the “splenic capsule” 1 when there is “perisplenic present” and the “abdominal
attempt
with the understanding tive CT results do occur
destate. that
score of less than 2.5. We agree with Umlas and Cronan that it is hard to differentiate peritoneal-perisplenic from subcapsular fluid. Resciniti et al (3) eva!uated various clinical and laboratory
(2%)
not,
their
as they
frustration
claim,
but
be considered
an inherent
Umlas
currence; reported
of whom had minimal This is not a unique oc-
similar situations (5-8). In addition,
have
been
with different interpretations their results may not be as disappointing, nor are they different from ours (3,4): In the study by Umlas and Cronan of 29 adult patients, two (7%) received a “safe” score of less than 2.5 and experienced a delayed rupture. In the second University of Massachusetts study (4) this occurred in one
a mean
score
of 1.16.
should be significant. we would question paring mean scores; would
have
been
more
by not including
positive
cases
difference
Parenthetically, the validity of cornuse of medians
with
appropriate.
the nine a score
Fi-
true-
of 2.5 or
greater
in the scattergram
(1, fig 1), Urn-
las and
Cronan
skewed
data adversely. In summary, chusetts
scoring
may
have
the University system
was
the
of Massadeveloped
not to replace clinical evaluation but to help identify the majority of patients who although initially considered candidates for nonsurgical treatment may eventually
need
surgery.
309
computed
References 1. Umlas SL, Cronan JJ. Splenic trauma: CT grading systems enable prediction
4.
successful
nonsurgical treatment? Radiology 1991; 178:481-487. Buntain WL, Gould HR, Maul KI. Predictability of splenic salvage by computed tomography. J Trauma 1988; 28:24-34. Resciniti A, Fink Ml’, Raptopoulos V, Day-
2.
3.
5.
idoff A, Silva WE. Nonoperative treatment of adult splenic trauma: development of a
Shari-Lynn
Umlas,
MD
John
#{149}
J.
scoring system that candidates for expectant management. J Trauma 1988; 128:828-831. Scatamacchia SA, Raptopoulos V, Fink Ml’, Silva WE. Splemc trauma in adults: impact of CT grading on management. Radiology 1989; 171:725-729. Federle Ml’, Griffiths B, Minagi H, Jeffrey RB Jr. Splenic trauma: evaluation with CT. Radiology 1987; 162:69-71. Pappas D, Mirvis SE, Crepps JT. Splenic trauma: false-negative CT diagnosis in cases of delayed rupture. AJR 1987; 149:727-728. detects
can of
6.
Cronan,
tomographic
7.
Taylor
8.
computed tomography delayed splenic rupture. Tomogr 1984; 8:1205-1207. Fagelman D, Hertz MA, development of splenic matoma: CT evaluation. Tomogr 1985; 9:815-816.
appropriate
CR, Rosenfield
AT.
Limitations
of
in the recognition J Comput Assist Ross AS. subcapsular J Comput
of
Delayed heAssist
MD
Reply thank Drs Raptopoulos and Fink for their interest in our article (1). We agree with Drs Raptopoulos and Fink that computed tomography (CT) is useful as an adjunct to evaluation of clinical sta-
We
tus
when
tient
deciding
how
to treat
the
pa-
with
splenic injury. However, we believe that no grading system has yet been suggested that is reliable enough to be used as a screening test in patients with sp!enic injury. False-negative results are far too prevalent with use of these systems. Our fear is that physicians might come to rely too heavily on these systems, and this might prove to be dangerous or even lethal in patients at risk for delayed splenic rupture. Raptopoulos and Fink state that we implied that the grading system described in the second study from the University of Massachusetts had been “revised” since the first study in regard to the scoring of perisplenic fluid. We were referring to the fact that in the first study (2), a score of 0 was given for splenic capsule “intact” and a score of 1 for “perisplenic fluid present.” With this distinction, the presence of only subcapsular fluid would imply that the capsule was still intact, thus generating a 0 capsular score. In the later study by Scatamacchia et a! (3), a statement was made that “no distinction is made between subcapsular splenic fluid and perisp!enic fluid,” thus implying that they also found this a cumbersome distinction to make. In our study, we noted that a distinction between perisplenic and subcapsular fluid when grading 2From
the Department
of Diagnostic
Imaging,
Rhode Island Hospital, 593 Eddy St, Providence, RI 02903. Received April 29, 1991; accepted May 3. Address reprint requests tOJ.J.C.
310
#{149} Radiology
sp!enic injury resulted in significant differences in the total scores of our cases. In addition, when one attempts to eva!uate the integrity of the splenic capsule, the importance of this distinction is clear; subcapsular fluid without the presence of perisplenic fluid implies that the capsule is still intact, whereas perisplenic fluid around an injured spleen would imply that the capsule had been violated. If these two types of fluid are graded equally, the score might reflect the presence or absence of fluid, but it would not necessarily reflect the status of the capsule. Raptopoulos
and
Fink
note
our
skep-
ticism of their system, which we based on the rate of false-negative results that we encountered, and state that interpreting our data differently might prove to be less disappointing. It is always true in scientific research that there are many ways to interpret a set of results, and in our entire study group the rate of delayed rupture overall was, in fact, low. However, if this particular comparison is to be made accurately, the rate of delayed rupture among cases of splenic injury in the general population must be taken into account. Perhaps a review of the natural history of the injured spleen might reveal that without any intervention,
the
overall
rate
of delayed
rupture is low as well. As for their next area of criticism, we agree with Raptopoulos and Fink that the mean total scores of patients undergoing early splenic salvage operations versus early splenectomy in our study were not statistically different. However, the type of procedure chosen by the surgeon during early celiotomy is not what the scoring systems are attempting to predict. With use of this
grading system, the mean scores of those patients who experienced delayed rupture versus those who were successfully treated nonoperatively (ie, those in whom the initial decision to treat nonoperatively was incorrect vs those in whom it was correct) were also not statistically
different,
and
it is this
frighten-
ing finding that we believe makes system unacceptable as a predicting tool. Raptopoulos and Fink compare data
in our
successfully those who Raptopoulos
study
early
the
who
were
treated nonoperatively and “required” early surgery. and Fink are making the
assumption
derwent
for patients
this
that
those
surgical celiotomy
patients
intervention “required”
who
un-
during surgery
and
thus that the “decision to operate was correct.” On the contrary, recent studies indicate that the patient with splenic injury shown on a CT scan and no other indications for surgery may be quite successfully treated nonoperatively (4). This appears often to be the case in patients aged
found Merely
with even the most severely and shattered spleens, as we
was
true
operate early not necessarily was a correct many
in our
because
of these
the
own
surgeon
dam-
study. decided
to
and repair the injury does mean that the decision one; it is possible that patients
might
have
re-
covered just as well had they merely been observed. Parenthetically, we used mean scores in our study simply because
mean
scores
were
what
was
used
to evaluate the data in the second University of Massachusetts study (3) and we wanted to be as consistent as possible with data interpretation so that our findings could most accurately be compared with the findings in this study. August1991