| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Original Research |
1 Present address: Department of Medical Imaging, Rabin Medical Center,
Beilinson Campus, Sackler School of Medicine, University of Tel-Aviv,
Jabutinsky St., Petach Tikva, Israel, 49100.
2 Department of Medical Imaging, Sunnybrook and Women's College Health Science
Centre, 2075 Bayview Ave., M5G 2C4, University of Toronto, Toronto, ON,
Canada, M4N 3M5.
3 Department of Medical Imaging, University Health Network, University of
Toronto, Toronto, ON, Canada, M5G 2C4.
4 Statistical Division, University Health Network, Toronto, ON, Canada M5G
2C4.
Received November 19, 2004;
accepted after revision January 10, 2005.
Address correspondence to M. Atri.
Abstract
 Top Abstract IntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
MATERIALS AND METHODS. Two radiologists blinded to pathology results retrospectively reviewed CT scans of 36 consecutive cystic renal masses in 30 patients (19 men and 11 women; age range, 28-76 years; mean age, 59 ± 13 years) who had undergone surgery. The study population included only masses with a cystic component on gross pathology and imaging. All patients underwent contrast-enhanced CT. The reviewers recorded the CT features of each cystic mass, including the presence of enhancing components. Accuracy values and odds ratio to predict malignancy were calculated for each CT feature. Weighted kappa was used to measure interobserver agreement.
RESULTS. There were 21 cystic renal cell cancers and 11 benign
cystic lesions. All cystic renal cell carcinomas showed an enhancing septal or
nodular component. The mean sensitivity and specificity of the two reviewers
in predicting malignancy for the presence of septal enhancement were 83% (95%
confidence interval [CI], 65-93%) and 82% (95% CI, 56-94%); for nodular
enhancement, 67% (95% CI, 49-81%) and 96% (95% CI, 75-99%); and for either
septal or nodular enhancement, 100% (95% CI, 86-100%) and 86% (95% CI,
67-95%), respectively. The interobserver agreements for septal and nodular
enhancement were good (
= 0.67) and moderate ( = 0.57),
respectively.
CONCLUSION. The presence of either nodular or septal enhancement shows the highest sensitivity for predicting malignancy with moderate to good interobserver agreement.
Keywords: Bosniak • cancer • CT • cyst • kidney
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
The original Bosniak classification for cystic renal lesions categorizes them into four groups [1]. Bosniak category I and II lesions are considered benign and do not require further evaluation. Bosniak III and IV lesions are treated surgically because of the high incidence of malignancy among these lesions. However, several investigators have challenged this classification [2-4]. Although category I and IV lesions are generally straightforward from a diagnostic perspective, category II and III lesions are often more problematic, with low interobserver agreement for these two categories reported [2-4]. Both categories include cystic masses with features that may be seen in malignancy, such as septations or calcifications with the degree of thickening, a subjective measure, being the difference between the two categories. Lesions in category III may ultimately be benign or malignant, and this is the most difficult category to accurately classify. In a revision of the original classification, Bosniak added category IIF, which consists of lesions with overlapping features from categories II and III [5]. Bosniak suggests that these lesions could be followed up with imaging to assess for lesion stability.
To our knowledge, there is no previous blinded study that has investigated the value of individual CT features in predicting malignancy. Specifically, little data on the significance of enhancing components within complex renal cysts are available [6, 7], and no blinded study comparing benign and malignant cystic masses with respect to enhancing components has been reported to our knowledge. Our hypothesis was that the presence of an enhancing component is the best feature with which to differentiate benign from malignant renal cysts. Also, interobserver agreement for individual CT features of a cystic mass has not been evaluated.
The purpose of our study was to evaluate the presence of an enhancing component, among other individual CT features of cystic renal masses, with respect to interobserver agreement and prediction of malignancy.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Study Population
One of the authors searched the medical records of one teaching hospital
for surgical pathology specimens that had been coded as "nephrectomy,
partial/radical nephrectomy" between January 1993 and March 2000. The
lesions that were described as complex renal masses with a cystic or necrotic
component on pathology and imaging were identified. This search yielded 47
masses described as renal masses with cystic or necrotic components in 45
patients. Over this period of time, 340 nephrectomies or partial nephrectomies
were performed in the same institution for a suspicious mass lesion. Three
hundred sixteen turned out to be renal cell carcinoma (RCC) (solid and cystic
combined), and 24 were benign masses (including both cystic and solid
masses).
We excluded 15 lesions in 15 patients: three patients had only MRI; four patients had poor quality CT examinations; four patients had only enhanced CT examinations; three masses were completely solid on preoperative CT; and one mass was clearly a simple cyst on preoperative CT, as indicated by both reviewers. Neither of the two reviewers was involved in the initial search. The remaining lesions qualified as Bosniak category II or higher. Therefore, our study population included 32 renal masses in 30 patients. One patient with von Hippel-Lindau disease had three cystic masses removed. All patients had undergone preoperative imaging with CT. Surgery of all lesions had been performed because of the complex nature of these lesions on preoperative imaging based on the initial report. There were 19 men and 11 women in this cohort with an age range of 28-76 years (mean, 59 years; SD, 13 years).
The maximum diameter of the lesion in the axial plane ranged from 1.8 to 17 cm (mean, 5.3 cm; SD, 3 cm) as averaged by the two reviewers' measurements.
CT Examination
The CT studies were performed on a single-detector CT (n = 11) or
MDCT (n = 26) scanner (HighLight Advantage and LightSpeed Qxi, GE
Healthcare). Twenty-two patients had CT examinations with 5-mm collimation,
six with 7-mm collimation, and six with 10-mm collimation. The pitch was 1:1
for single-detector CT and 3:1 for MDCT examinations. Each patient received
80-100 mL of iohexol (Omnipaque 300, Amersham Health) IV at a rate of 2-3
mL/sec. Scan delays were 20-30 sec for the corticomedullary phase and 90-100
sec for the nephrographic phase. All patients had both unenhanced and enhanced
CT examinations. Corticomedullary phase images were available for the
evaluation of 11 masses and nephrographic phase images, for all 32 masses. The
technical parameters (peak kilovoltage and tube current) were maintained for
both unenhanced and enhanced studies.
Data Collection
The following features of each cystic mass were recorded by the two
reviewers: size of the mass; multilocularity of the cyst; diffuse wall or
septal thickness of 2 mm or greater; presence of focal wall or septal
thickening; presence and size of mural or septal nodules; presence and size of
calcifications; density of the cyst; and enhancement of wall, septum, or
nodule. A nodular component was considered enhancing if the same-sized region
of interest (ROI) on unenhanced and enhanced scans showed an increase in
attenuation of more than 15 H
[8,
9]. The four examinations
without unenhanced study were included because there was either no enhancing
component or a large enhancing component. ROI measurement was in general not
feasible in the wall or septum. A septum or wall was considered enhancing when
it was not visible on the unenhanced scan but was seen as a distinct structure
on the enhanced scan. We could not compare different phases of enhancement
because all patients did not undergo imaging during all phases. Cyst
attenuation was measured in Hounsfield units as the mean measurement of the
two reviewers on the unenhanced examination when an unenhanced examination was
available and on the enhanced examinations in four patients with no unenhanced
CT. The reviewers were asked to classify the lesions according to the modified
Bosniak classification [10].
At the time this study was conducted, our institution did not require ethics
committee approval for the retrospective review of the data.
Statistics
Sensitivity, specificity, negative predictive value (NPV), positive
predictive value (PPV), and accuracy were computed for the various CT signs as
predictors of malignancy. An exact form of the McNemar test was used to
compare the accuracy values
[11,
12]. The odds ratio was also
used as a measure of the strength of the relationship between the CT signs and
malignancy. A profile likelihood-based confidence interval for the odds ratio
was computed. Weighted kappa () was used to measure interobserver
agreement. It has been suggested that a kappa statistic of less than 0.40
indicates poor agreement; 0.40-0.59, moderate agreement; 0.60-0.74, good
agreement; and greater than or equal to 0.75, excellent agreement
[13]. In comparing the
malignant and benign groups, the Student's t test was used to test
equality of means of continuous variables, and Fisher's exact test was used
for the comparison of categoric variables. Generalized estimating equations
were used to estimate the confidence intervals (CIs) for the average
sensitivity and specificity of the two reviewers
[14]. A p value of
less than 0.05 was considered statistically significant. We used S-Plus 6.2
(MathSoft) for statistical analysis.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Table 1 summarizes the odds ratio, confidence limits of the odds ratio, and kappa statistics for the individual CT features in the malignant group compared with the benign group for both reviewers. Most complex features were significantly more common in the malignant group except calcification, which was more significantly seen in the benign group.
|
Tables 2 and 3 outline the sensitivity, specificity, NPV, PPV, and accuracy of the CT features that were the best predictors of malignancy for reviewers 1 and 2. There was no statistically significant difference between the values of the two reviewers (p > 0.05). The presence of nodular or septal enhancement showed the best sensitivity for predicting malignancy. The mean sensitivities, specificities, NPV, PPV, and accuracies of the two reviewers for septal enhancement in predicting malignancy were 83% (CI, 65-93%), 82% (CI, 56-94%), 73%, 90%, and 83% (CI, 71-96%) and for nodular enhancement were 67% (CI, 49-81%), 96% (CI, 75-99%), 61%, 97%, and 77%, respectively. The corresponding numbers for septal or nodular enhancement were 100% (CI, 86-100%), 86% (CI, 67-95%), 100%, 94%, and 95%, respectively.
|
|
Wall and septal enhancement showed excellent and good interobserver agreement, respectively, whereas nodular enhancement showed moderate interobserver agreement (Table 1). The lower agreement for enhancing nodule was due to disagreement about focal thickening and a true nodule.
The mean size of the enhancing nodules seen by both reviewers in the malignant group was 7 mm (range, 2-18 mm).
The kappa value for Bosniak classification between the two reviewers was 0.52 (CI, 0.28-0.75) indicating moderate agreement (Table 4).
|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
In 1993, Bosniak revised his classification to include category IIF lesions. These lesions have slightly more complex imaging features than those of category II [6]. Lesions classified in category IIF require serial imaging follow-up to assess for stability [5]. The differentiation between categories IIF and III is important because category IIF lesions require follow-up, whereas category III lesions require surgery.
Since its publication, the Bosniak classification has been applied by several authors with conflicting results [2-4]. Although category I and II lesions are benign in most reports, Wilson et al. [2] found four of five lesions in category II to be malignant, Cloix et al. [3] found two of nine category I or II lesions to be malignant, and Siegel et al. [4] found one malignant lesion among eight categorized as Bosniak II (Table 5). The reported prevalence of malignant lesions in category III lesions range from 25% to 100% and in category IV lesions, from 67% to 100% [2-4, 15-17] (Table 5). More recently, Curry et al. [15] published their results on 116 renal cystic lesions, 82 of which were resected. Twenty-nine (59%) of the 49 lesions categorized as Bosniak III were malignant.
|
Most investigators agree on the usefulness of the Bosniak classification for categories I and IV. However, there are difficulties differentiating between II and III. Category III is the most difficult to classify with the widest range in terms of prevalence of malignancy.
Classification of cystic renal lesions remains subjective. Previous studies show significant interobserver variability partly because of variable interpretation of cyst wall thickness, irregularity, and type of calcification [4]. Whereas some reviewers classified some lesions as category II, others classified the same lesions as category III or IV [4]. The kappa for Bosniak classification between the two reviewers was 0.52 in our study, indicating moderate agreement. This variability results in unnecessary resection of benign renal lesions.
To our knowledge, no previous blinded study has investigated the significance of each CT feature in predicting malignancy or evaluated the interobserver agreement for each feature. In this cohort of cystic renal cell cancers, an enhancing septum was seen in 90% and 76%, wall enhancement in 76% and 86%, and an enhancing nodule in 57% and 76% of lesions evaluated by reviewers 1 and 2, respectively. Our sample size was too small to perform a multivariable analysis to determine the independent predictors of malignancy. However, 100% of the malignant lesions in our cohort had an enhancing septum or nodule identified by both reviewers (Figs. 1A and 1B). Reviewers 1 and 2 identified nine and five malignant lesions with septal enhancement without nodular enhancement (Figs. 2A, 2B, 3A, and 3B), and two and five lesions had only nodular enhancement without septal enhancement, respectively (Figs. 4A and 4B). Among the lesions in our study, there was no malignant lesion with the wall being the only enhancing component (Figs. 5A, 5B, 6A, and 6B).
|
|
|
|
|
|
|
|
|
|
|
|
The interobserver agreement for septal enhancement was good ( =
0.67), and that for nodular enhancement was moderate ( =0.57). The
reason for the lower interobserver agreement for nodular enhancement was the
disagreement with respect to the presence of a true nodule versus focal septal
thickening (Figs. 2A,
2B,
3A, and
3B). Siegel et al.
[9] showed good interobserver
agreement for the presence of enhancement in both solid and cystic renal
masses that measured more than 1.5 cm in diameter.
Calcification was significantly more common in benign lesions with odds ratios of 0.3 and 0.4 for the two reviewers. A recent study by Israel and Bosniak [10] also showed that the presence of calcification alone does not predict malignancy in a cystic lesion.
Our results show statistically significant higher attenuation values of the content of the malignant lesions. This indicates a more complex nature of the content of this group as compared with benign lesions in our cohort. However, because the values of these two groups overlapped, this information has no practical value.
Nuclear grade and pathologic stage are usually lower in cystic RCCs than in the solid variant [20]. Survival was also significantly longer in patients with cystic RCCs [20]. Therefore, a conservative surgical approach has been suggested as the treatment of choice whenever it was technically feasible. We propose that complex cystic lesions with thick septal or nodular enhancement be considered malignant. Lesions with no enhancing component or with only wall enhancement or questionable or thin enhancing septa can be followed up as category IIF lesions. MRI may play a role in the evaluation of lesions with questionable nodular or septal enhancement because of better contrast resolution than CT and its multiplanar imaging capabilities. MRI can potentially improve interobserver agreement for suspicious features because of more obvious enhancement on MRI.
Moderate agreement was present between the two reviewers for Bosniak classification. Five lesions called Bosniak II by one reviewer were called IIF (n = 3) or III (n = 2) by the other reviewer. This low interobserver agreement is likely the explanation for surgery on some of these lesions with a low Bosniak classification.
Lang et al. [21] advocated percutaneous aspiration of category IIF and III lesions to reduce the number of unnecessary surgeries. However, percutaneous biopsy had a sensitivity of only 71% to diagnose malignancies in their series and most urologists are reluctant to biopsy an RCC.
Our study has a number of limitations. The CT examinations were performed on a combination of single-detector CT and MDCT scanners with different scanning parameters. This was inevitable because of the retrospective nature of the study and rarity of cystic RCCs requiring a long time span to collect a sufficient number of cases from one institution. MDCT performed with a thin collimation can potentially improve the capability of CT in showing an enhancing component. However, all the lesions in this cohort showed an enhancing nodule or septum despite the variation in equipment and technical parameters. Improvement in equipment and technique can potentially influence the specificity of these features.
A multivariable analysis would have helped us to determine independent variables capable of predicting malignancy, but that would have required a larger study population [22]. A larger sample size would have also resulted in narrower CIs for accuracy and kappa values. We believe the sensitivity of enhancing components on CT to diagnose cystic RCCs in our study is accurate considering the long period for inclusion of our cases, which would have captured even slow-growing RCCs. A study including all cystic masses classified as Bosniak category II or higher would give a more accurate estimate of the specificity of an enhancing component. However, such a design would require a long follow-up to confirm the benign nature of a mass, considering that cystic RCCs could be slow growing and most complex masses are not managed with surgery.
We were not able to compare corticomedullary and nephrographic phase scans to determine which phase is more accurate in showing the enhancing components because a limited number of patients had corticomedullary phase scans in addition to the nephrographic phase scans.
The results of this study should be corroborated in a prospective study that includes nonsurgical cases to better assess the specificity of these CT features.
In conclusion, enhancement of septal or nodular components should be considered the main criterion for malignancy when evaluating cystic renal masses with CT. Complex lesions with nonenhancing or questionable enhancing components, heavy calcification only, or mural enhancement only can be managed as Bosniak IIF lesions.
Acknowledgments
We thank Mrs. Carole Leduc for assistance in the preparation of the
manuscript.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
CiteULike 
Complore 
Connotea 
Del.icio.us 
Digg 
Reddit 
Technorati What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
