Devil's Advocate James
A. Brink,
MD
•¿ Joseph
T. Ferrucci,
MD
Use of CT for Predicting A Dissenting View' HE use of computed tomography (CT) for determining gallstone composition has not gained widespread application because of its variable accu racy, cumbersome interpretive algo rithms, and cost. CT of the gallstone is performed to identify gallstones that are predominantly cholesterol in com position and, therefore, most amenable to nonsurgical therapy. Patients with significantly calcified gallstones and patients with predominantly pigment @ stones are generally excluded from nonsurgical therapies. CT is extremely sensitive to minute amounts of calcium within gallstones. Sixty percent of gallstones scanned in vivo demonstrate calcification (1), whereas only 15%-20% of gallstones are radiopaque at plain radiography (2). However, the added benefit of in creased sensitivity to calcium is not clear. Gallstones with rimmed and lam inated calcification patterns have been shown to dissolve in methyl tert-butyl ether (3), and calcified gallstones have been shown to fragment well with in vitro extracorporeal shock wave litho tripsy (4). In addition, most clinical protocols for extracorporeal shock wave lithotripsy have used plain radi ography to exclude heavily calcified gallstones. Stones with a central nidus of calcium or a peripheral rim of calci um were acceptable in the recent proto
Gallstone
Composition:
T
Index terms: Computed tomography (CT), tis sue characterization •¿ Gallbladder, calculi, 762.1211, 762.289 •¿ Opinion Radiology
1991; 178:633-634
‘¿Fromthe Department chusetts General Hospital
of Radiology, and Harvard
Massa Medical
School, Boston, Mass. Received August 16, 1990;acceptedNovember 12.Addressreprint requests to JAB., Mallinckrodt Institute of Ra diology, Washington University Medical School, 510 S Kingshighway Blvd. St Louis,
MO 63110. ©RSNA,
1991
a. b. c. (a)Globalattenuationmeasurementof 15-mmgallstone with 2-mmperipherallycalcified rim. The mean attenuation of the entire stone (90 HU) is measured with a single region of interest attenuation value without preservation of zonal architecture. (b, c) Regional attenu ation measurements of homogeneous regions of stone in a. The attenuations of the rim (b) (169 HU) and center (c) (10 HU) are measured
with specific,
localized
regions
of interest.
Al
though regional techniques are theoretically more accurate in predicting gallstone composi tion, global techniques have been more successful in vitro.
col for U.S. Food and Drug Administra tion approval of the MPL 9000 litho triptor (Dornier Medical Systems, Marietta, Ga) (5). Thirty percent of all gallstones are predominantly pigment gallstones, and 50% of those stones are calcified at plain radiography. Conversely, 85% of gallstones are radiolucent at plain radi ography, and 15% of those are pigment stones (2). Therefore, the only gall stones that are clinically insoluble in dissolution agents and that are not de tectable by means of plain radiography are the radiolucent pigment stones, which comprise 12% of all gallstones. CT characterization of gallstones in this small fraction of patients is important. However, gallstones must first be iden tified before they can be characterized, and CT is only 79% sensitive in the de piction of gallstones (1). Therefore, even under the best of circumstances, characterization of gallstone composi tion may be beneficial in only 10% of gallstone patients. Investigators have used both quanti tative and qualitative interpretive algo
rithms to identify stones that will be amenable to nonsurgical therapy. Re cently, successful prediction of gall stone composition with CT was report ed in both in vivo and in vitro analyses (6). The technique involves measuring the mean attenuation of the entire gall stone with use of a single region of in terest (Figure). Although this “¿global― technique reduces the complex layers of gallstones to a single mean attenua tion value, the reported accuracy was good. The correlation coefficient be tween cholesterol content and mean at tenuation was —¿0.83.This is in contrast to a technique described by Baron et a! (7), in which “¿regional― analysis was used to measure the attenuation of ho mogeneous regions within gallstones by using specific, localized regions of interest. This approach permits correla tion of CT attenuation with specific, more compositionally pure layers in the stones. Although regional analysis
Abbreviation:0CC = oralcholecystography.
633
would be expected to yield a stronger correlation than global analysis, the op posite is true. Baron et al reported poor correlation (r = —¿0.64)between choles terol content and CT attenuation. The difference may be due to the relative lack of radiolucent pigment stones sub jected to global analysis (6). Presuming that stones with CT atten uation below 50 HU are radiolucent at plain radiography, current studies may be analyzed with respect to the fraction of pigment stones that are also likely to be radiolucent at plain radiography and, therefore, have clinical interest for CT scanning. Although one-half of all pigment stones are radiolucent, only 15% (two of 13) of the pigment stones analyzed by Brakel et a! (6) would be expected to be radiolucent at plain radi ography. Therefore, the successful dis crimination of predominantly choles terol and pigment stones probably re lates, in part, to limitations in the sample stone population. Secondly, the threshold CT attenuation for discrimi nating predominantly cholesterol and pigment gallstones is high—140 HU. As most of the pigment stones in this population were calcified, this thresh old actually discriminates the calcified from the noncalcified stones. In fact, this confusion of calcium and pigment extends into comparison with studies that predicted stone composition by means of analysis of features at oral cholecystography (OCG). When quanti tative CT is compared to radiolucency or opacity at plain radiography, CT cer tainly is more useful than plain radiog raphy in helping to predict whether gallstone composition is predominant ly cholesterol or pigment. However, both cholesterol and pigment stones may be calcified on the plain abdomi nal radiograph, and, therefore, this comparison is invalid. Comparison of the quantitative CT techniques to mul tivariate discriminate analysis of 0CC features—including stone size, num ber, surface characteristics, and calcifi cation pattern—does not reveal a dif ference between quantitative CT and OCG. In fact, predominantly cholester
634 •¿ Radiology
ol gallstones were predicted with a higher accuracy with use of 0CC (100%) (8) than with quantitative CT (84%) (6). Limitations of quantitative CT may best be illustrated with consideration of a hypothetical cholesterol gallstone 15 mm in diameter with a 2-mm pe ripheral rim of calcification. If a single region of interest encompassing the en tire stone is chosen, the mean attenua tion will reflect both the central choles terol core and the peripheral calcified rim. However, the 2-mm peripheral rim comprises 46% of the stone cross section, and the central portion com prises 54%. If the attenuation of the central cholesterol core is 10 HU and that of the peripheral calcified rim is 200 HU, the mean attenuation of this stone cross section would be 113 HU. Because the mean attenuation is less than the 140-HU threshold, the stone would be accurately classified as a pre dominantly cholesterol stone. This threshold is set high enough to be in sensitive to the laminated and rimmed calcification patterns often seen in large cholesterol gallstones (9) but is still low enough to correctly classify pigment gallstones that are heavily cal cified and easily identified on a radio graph of the abdomen. Again, the lack of relatively radiolucent pigment gall stones in the sample population proba bly elevates the accuracy of the tech nique. Significant composition-specific in formation can also be obtained by means of qualitative interpretation of gallstones at CT. Several different ap proaches to CT classification of gall stone patterns have been reported by several investigators (1,6,7). Baron et al (7) demonstrated a significant differ ence in the cholesterol content among stones with different CT patterns. Gen erally, stones that were isodense with bile and the center of stones with pe ripheral calcified rims were high in cholesterol. Although CT is extremely sensitive to small amounts of calcium within gallstones, the increased sensitivity rel ative to routine radiography is not clearly beneficial. To our knowledge,
the potential of CT in enabling identi fication of the 12% of gallstone patients with pigment stones that are radiolu cent at plain radiography has not been proved, and evidence suggests that quantitative measurements will be in accurate with current single energy techniques. Qualitative analysis of gall stone structure with CT does provide some insight into the chemical ele ments;
however,
the radiolucent
pig
ment stone is no better identified CT.
with
References 1.
Barakos JA, Rails PW, Lapin SA, et al.
2.
Cholelithiasis: evaluation with CT. Radiol ogy 1987;162:415-418. Trotman BW, Petrella EJ, Soloway RD. San
chez HM, Morris TA III, Miller WT. Eval
3.
uation of radiographic lucency or opaque ness of gallstones as a means of identifying cholesterol or pigment stones. Gastroenter ology 1975;68:1563-1566. Baron RU, Kuyper Si, Lee SP, Rohrmann
CA Jr. Shuman WP, Nelson JA. In vitro dissolution of gallstones with MTBE: corre lation with characteristics at CT and MR imaging. Radiology 1989;173:117-121.
4.
5.
Schachler R, Sauerbruch
safety of extracorporeal tripsy
of gallstones:
Biliary Lithotripsy 6.
7.
8.
9.
I, Wosiewitz U, et
al. Fragmentation of gallstones using ex tracorporeal shock waves: an in vitro study. Hepatology 1988; 8:925-929. Schoenfield U, Berci C, Carnovale RU, et at. The effect of ursodiol on the efficacy and
shock-wave
the Dornier
litho
National
Study. N Engi J Med
1990;323:1239-1245. Brakel K, Laméris JS, Nijs HGT, Terpstra OT, Steen C, Blijenberg BC. Predicting gallstone composition with CT: in vivo and in vitroanalysis.Radiology 1990;174:337341. Baron RU, Rohrmann CAJ. Lee SP, Shuman WP, Teefey SA. CT evaluation of gall stones in vitro: correlation with chemical analysis. AJR 1988; 151:1123—1128. Dolgin SM, Schwartz 5, Kressel HY, et al. Identification of patients with cholesterol or pigment gallstonesby discriminant analysisof radiographicfeatures.N Engi Med 1981; 304:808-811. Wolpers C. Gallenblasensteine: ihre mor phogenese und auswahl zur litholyse. Ba sel.Switzerland:Karger. 1987;83-88.
March 1991