Secondary Logo

Journal Logo

Original Articles

Clear Cell Sarcoma of the Kidney

A Review of 351 Cases From the National Wilms Tumor Study Group Pathology Center

Argani, Pedram M.D.; Perlman, Elizabeth J. M.D.; Breslow, Norman E. Ph.D.; Browning, Nancy G. M.S.; Green, Daniel M. M.D.; D'Angio, Giulio J. M.D.; Beckwith, J. Bruce M.D.

Author Information
The American Journal of Surgical Pathology: January 2000 - Volume 24 - Issue 1 - p 4
  • Free


Clear cell sarcoma of the kidney (CCSK), an uncommon renal neoplasm of childhood, nonetheless represents one of the most common “unfavorable histology” tumors entered on National Wilms Tumor Study Group (NWTSG) clinical protocols. Approximately 20 new cases of CCSK are diagnosed each year in the United States. 3 CCSK was initially recognized as a distinct clinicopathologic entity by Kidd in 1970, who noted its propensity to metastasize to bone. 15 The distinctive histopathologic features of CCSK were reported simultaneously in 1978 by Morgan and Kidd, 22 Marsden and Lawler, 19 and Beckwith and Palmer. 5 These reports confirmed the propensity to metastasize to bone, the poor clinical outcome, and the sarcomatous, nonepithelial nature of the tumor. The tremendous morphologic diversity that CCSK can show, ranging from epithelioid to spindle cell patterns, has been emphasized in later writings. 4,23

Since its initial description, several small series of CCSK have been published. Early studies emphasized that even low-stage tumors had a poor prognosis, adding to the tumor's fearsome reputation. 30,32 Subsequently, results from the first three NWTS trials suggested that the addition of doxorubicin (Adriamycin) to vincristine and dactinomycin improved the 6-year relapse-free survival for patients with CCSK, although the difference did not reach statistical significance. 10 Nonetheless, all patients with CCSK on NWTS trial 5 are now treated with doxorubicin regardless of stage. While this approach has decreased the importance of accurate staging, the essential role of doxorubicin in therapy emphasizes the need for pathologists to accurately identify CCSK. Failure to recognize a tumor as a CCSK could deprive a child of optimal chemotherapy.

Immunohistochemical studies have been performed with the hope of establishing a unique profile that distinguishes CCSK from other lesions in the differential diagnosis. These have generally shown that CCSKs are reactive for vimentin but little else. 17,18,25 However, the possibility of cytokeratin expression in epithelioid areas of CCSK has not been addressed. This issue is highlighted by isolated reports of epithelial differentiation in CCSK, both in primary 16,31 and cultured tumors. 13 In addition, one study found a high frequency of p53 protein overexpression in these tumors as determined by immunohistochemistry, suggesting frequent mutations in this gene. 6

Diagnostic criteria for the renal tumors of childhood have evolved over the years. Recognition of new entities, such as metanephric stromal tumor, 2 embryonal sarcoma of the kidney, 7 primary renal peripheral neurorectodermal tumor (PNET), 29 and cystic hamartoma of the renal pelvis, 27 has allowed previously recognized tumors, such as congenital mesoblastic nephroma, to become more sharply defined. Staging criteria have also changed over the course of the NWTS. Finally, immunohistochemical studies have become more reliable and adaptable to archival material using the heat-induced epitope retrieval method. With these considerations in mind, we have reviewed all cases of CCSK available at the NWTSG Pathology Center, which holds the largest collection of pediatric renal tumors in existence. The goals of this study were to define more sharply the clinicopathologic and immunohistochemical spectra of CCSK, and to search for clinical and pathologic variables of prognostic import.


Clinicopathologic Correlative Study

The initial study population included 214 cases entered as CCSK on NWTS trials 1 through 4, running from August 1969 to August 1995. Patients on these trials were treated by standardized protocols and most had complete clinical follow up of at least 3 years. All histologic slides and the institutional pathologic report, with the operative note when available, were reviewed by one of the authors (P.A.). All tumors for which an alternative diagnosis was suggested were reviewed by two authors (J.B.B., P.A.) using a dual-observer microscope. Twenty-two cases were excluded; the revised diagnoses were blastemal Wilms tumor (6 cases), PNET (6 cases), stromal tumor not otherwise specified (3 cases), congenital mesoblastic nephroma (2 cases), metanephric stromal tumor (2 cases), sarcoma not otherwise specified (2 cases), and rhabdoid tumor of the kidney (1 case). Five additional cases were excluded because slides from the nephrectomy specimen were unavailable for review. This left 187 cases of CCSK representing 2.8% of 6650 malignant pediatric renal tumors entered on the NWTS over this time. From this group, five cases were excluded as a result of insufficient clinical follow up. This left 182 patients (the study cohort, or prognostic group) for clinicopathologic correlation analysis and evaluation of prognostic factors.

One hundred eleven additional cases of CCSK from the outside consultation (OCWT) files of the NWTSG Pathology Center and 53 cases of CCSK entered on NTWS 5 were also reviewed for purposes of the demographic and morphologic portions of the study. These cases were not included in the evaluation of prognostic factors because of the lack of uniform therapy or adequate clinical follow up.

All cases of CCSK featured either the classic pattern of histology, characterized by an arborizing vasculature separating cords of plump cells with indistinct cell borders and open chromatin, or multiple characteristic variant patterns of CCSK (see Results). The predominant histologic pattern and the presence of any variant pattern in each case were recorded. Tumor necrosis, whether identified microscopically or noted in the institutional gross pathologic or microscopic description, was also recorded. Specimen weight and maximal tumor diameter, when measured, were obtained from the institutional pathology report. On the basis of our examination of the histologic slides as well as the institutional pathologic and operative reports, tumors were restaged using updated NWTS 5 definitions as follows:

  • Stage 1: Tumor confined to the kidney and completely resected. No penetration of the renal capsule or involvement of renal sinus vessels.
  • Stage 2: Tumor extends beyond kidney but completely resected. Tumor penetrates renal capsule, invades sinus vessels, was biopsied before removal, or spilled locally during removal, but margins are negative.
  • Stage 3: Gross residual tumor, positive surgical margins, massive tumor spill or lymph node metastases.
  • Stage 4: Hematogenous metastases.
  • Stage 5: Bilateral renal tumors.

When the stage on review differed from the stage recorded by the NWTSG Data and Statistical Center, the relevant slides, gross pathologic findings, or operative findings were reviewed by two authors (J.B.B., P.A.) to resolve the discrepancy. Among the 187 verified CCSK cases in NWTS 1–4, the stage of 42 (23%) was revised. Twenty-three tumors were changed from stage 1 to stage 2, 10 from stage 2 to stage 3, 6 from stage 1 to stage 3, 2 from stage 3 to stage 2, and 1 from stage 3 to stage 4. The number of staging changes primarily reflects the evolution of staging criteria over the course of the study, and accounts for any discrepancies related to tumor stage between this report and previous NWTS studies of CCSK. 10

All patients in the prognostic group underwent unilateral nephrectomy. Adjuvant therapy varied depending on when they entered the NWTS. Children enrolled on NWTS 1 and 2 (August 1969 through May 1979) were treated according to tumor stage, regardless of histology. Hence, the 36 CCSK patients entered on these two studies received regimen A (dactinomycin), B (vincristine), C (dactinomycin and vincristine), or D (dactinomycin, vincristine, and doxorubicin). All patients with CCSK on NWTS 3 received unfavorable histology therapy regimens DD (dactinomycin, vincristine, and doxorubicin) or J (dactinomycin, vincristine, doxorubicin, and cyclophosphamide). Given the lack of demonstrated effectiveness of cyclophosphamide in NWTS 3, all patients on NWTS 4 received a regimen containing dactinomycin, vincristine, and doxorubicin, many receiving more intensive vincristine and doxorubicin therapy than on previous studies. Only four patients received preoperative chemotherapy before nephrectomy. Eighty-five percent of patients in the prognostic group received radiation therapy to the tumor bed.

Six-year survival and relapse-free survival were used as the prognostic end points. Differences between survival curves were evaluated by the log rank test. Variables proving significant by univariate analysis were evaluated in multivariate analysis using the Cox proportional hazards model.

Immunohistochemical Stains

Paraffin tissue blocks were available from 45 cases for immunohistochemical studies. Five-micron sections were deparaffinized with xylene for 30 minutes and rehydrated using graded ethanol concentrations. Antigen retrieval was performed using either protease digestion or a heat-induced epitope retrieval method. Immunohistochemical staining was performed using the avidin-biotin-peroxidase technique with 3,3´-diamino-benzidine as chromogen on the automated Bio-Tek-1000 system (Ventana/Biotek Solutions, Inc, Tucson, AZ, USA). The antibodies used, vendors and dilutions are listed in Table 1.

Antibodies used in immunohistochemical staining


Clinical Features

The age at presentation for NWTS cases ranged from 2 months to 14 years, with a mean age of 36 months. Among 351 total patients in this study, the highest incidence of CCSK was in years 2 and 3 of life in which 50% of cases were diagnosed. A sharp decline occurred thereafter (Fig. 1). A male predominance was noted: there were 231 males and 113 females (ratio: 2.04:1) in the study. Only one CCSK among the 351 reviewed was associated with perilobar nephrogenic rests. No case was associated with intralobar nephrogenic rests. Only a single CCSK was associated with renal dysplasia. No familial or syndrome-associated CCSKs were identified. However, the brother of one patient with CCSK subsequently developed a nephrogenic (metanephric) adenofibroma, 12 another rare renal neoplasm with a stromal component.

FIG. 1.
FIG. 1.:
Age distribution of the 334 patients for whom age and sex were recorded.

The revised tumor stage distribution at presentation for NWTS 1–4, using NWTS 5 criteria, was as follows: 25% of patients had localized stage 1 tumors, a majority of patients presented with stage 2 (37%) or 3 (34%) disease, whereas only 4% of patients presented with distant metastasis (stage 4). No true bilateral primary tumors were identified. In one case, a 1-cm tumor was found in the kidney contralateral to a kidney harboring a 13-cm CCSK. Because the patient presented with widely disseminated stage IV disease, we interpreted the contralateral tumor as a metastasis.

The distribution of all sites of metastases at presentation and relapse identified among 351 total patients is illustrated in Table 2. Ipsilateral renal hilar lymph nodes were by far the most common site of metastasis at the time of presentation. Of the 159 NWTS cases in which lymph nodes were sampled, 46 (29%) revealed metastatic disease. Of note, three NWTS cases showed “skip” metastases to periaortic lymph nodes in the face of histologically confirmed negative hilar lymph nodes. As expected from previous studies, 20,21 bone metastases was the most common mode of relapse, followed closely by lung metastases. Abdominal/retroperitoneal relapses, indicating local recurrence, were the next most frequent, followed by brain and liver metastases. Noteworthy is the fact that a significant number of unusual soft tissue (scalp, epidural, nasopharynx, neck, paraspinal, abdominal wall, axilla) and other (orbital) sites of metastasis were also encountered.

Sites of metastasis at presentation and sites of recurrence in the overall study group

The time interval to relapse among NWTS cases ranged from less than 116 months to 4 years. A trend toward a longer interval to relapse was noted among patients treated with doxorubicin compared with those who did not receive this drug (Fig. 2). Longer intervals to relapse were encountered in the OCWT files, in which the exact therapy given was not always known to us. Among these cases, three patients relapsed after 3 years, four at 4 years, and one at 5 years. In addition, two patients relapsed at extraordinarily long intervals of 9 and 10 years, respectively. One of these patients had a stage 1 tumor that was treated by nephrectomy alone.

FIG. 2.
FIG. 2.:
Cumulative incidence of relapse (1 minus the relapse-free survival curve) among CCSK patients treated with or without doxorubicin on NWTS 1–4. Whereas the overall relapse rate is lower for patients treated with doxorubicin, the risk of relapse persists for a longer time.

Five other extremely unusual clinical presentations of CCSK were encountered in the OCWT files. Two were adult CCSKs presenting as primary renal tumors in a 28-year-old man and a 54-year-old woman. In addition, we encountered three extrarenal tumors which presented histologic features identical to those of CCSK. One was a pelvic soft tissue tumor in a 13-year-old boy, 33 the second was an ovarian tumor in a 7-month-old girl, and the third was a retroperitoneal tumor in an 11-year-old boy. While we cannot prove that all of these tumors are extrarenal CCSKs, they represent, at the very least, perfect phenocopies that suggest the possible existence of this neoplasm in extrarenal sites.

Seven nontumor deaths were identified among 182 patients in the prognostic group: 3 patients died of second malignant neoplasms, 3 of drug toxicity, and 1 of an infectious process. These patients were included in the survival analyses as survivors until the date at which they died, when they were withdrawn (censured) from the analyses. The two patients who died of a combination of relapsed tumor and drug toxicity were included among the 47 tumor-related deaths. One hundred twenty-six patients (69%) survived and are free of disease, whereas another two are alive with disease. One hundred ten patients (60%) experienced relapse-free survival.

Gross Pathologic Features

CCSK usually presented as a large, unicentric mass markedly distorting or nearly completely replacing the kidney; the mean diameter of measured tumors was 11.3 cm (range, 2.3–24 cm). The mean of the more consistently measured combined kidney–tumor specimen weight was 661 g (range, 43.5–1950 g). When an epicenter could be determined, the renal medulla was the most common location. No case of multicentric origin was identified.

On cut section, tumors were most commonly described as tan–grey, soft, and mucoid. Cystic foci were nearly universal and occasionally represented the dominant feature such that a radiologic and gross pathologic diagnosis of multilocular renal cyst (cystic nephroma) was made. The microscopic appearance of such lesions also mimicked cystic nephroma (see below). Within these typically large and relatively homogeneous tumors, discrete foci of necrosis (73% of cases) and hemorrhage were common. A subset of tumors had a more firm, whorled appearance. Tumors usually appeared grossly well-circumscribed with a sharp kidney–tumor border. Gross extension into the renal vein (tumor thrombus) was evident in 5% of cases.

Light Microscopic Features

The classic pattern of CCSK was defined by nests or cords of cells separated by regularly spaced, arborizing fibrovascular septa (Fig. 3). The regularity of the septal spacing resulted in cord widths of between four and 10 cells. While usually plump and ovoid, the cord cells not uncommonly assumed a spindle shape. The septa ranged from thin, regularly branching “chicken-wire” capillaries highly reminiscent of the architecture of myxoid liposarcoma to capillaries enmeshed by sheaths of fibroblast-like cells set in a collagenous matrix. These “cellular septa” measured up to 50 μm in width and are a characteristic feature of CCSK (Fig. 4). Rarely, dilation of the thin septal capillaries created a staghorn, hemangiopericytomatous appearance.

FIG. 3.
FIG. 3.:
Intermediate power view of the classic pattern of CCSK characterized by cord cells with fine chromatin and indistinct borders within an arborizing capillary vasculature.
FIG. 4.
FIG. 4.:
Intermediate power view of CCSK showing cellular fibroblastic septa and myxoid pools.

The cord cells featured nuclei that were overall uniform in shape, although fine irregularities in contour could be appreciated at high magnification. The chromatin was characteristically of a fine, dusty texture without prominent nucleoli or coarse condensations. Empty-appearing “Orphan Annie” eye nuclei were frequent. These characteristic nuclear features were usually not evident on either original frozen sections or permanent sections prepared from previously frozen tissue, and hence are fixation-dependent.

The cord cells were usually loosely spaced, separated from their neighbors by optically clear material that has proven to be extracellular mucopolysaccharide matrix, and gives rise to the clear cell appearance. 11 Nuclear overlap was less common than in Wilms tumor, PNET, or other more cellular pediatric renal neoplasms. The cytoplasm was generally sparse with indistinct cell borders. However, well-demarcated pink cytoplasm was occasionally evident, particularly in areas of acinar-type epithelioid pattern (see below) or following fixation in mercuric fixatives such as B5. The combination of focal cord cell crowding and acidophilic cytoplasm imparted a “dark cell” appearance. Rarely, rhabdoid cytoplasmic inclusions were identified, but these cells lacked the prominent nucleolus characteristic of true rhabdoid tumors.

While tumors appeared circumscribed both grossly and under low-power microscopic examination, on closer inspection they entrapped individual tubules of the medulla and occasionally cortical glomeruli. Entrapped tubules often acquired a metaplastic “embryonal” appearance with occasional mitoses, simulating the neoplastic tubules of Wilms tumor. Alternatively, the entrapped tubules dilated extensively, giving rise to grossly evident cysts with paucicellular septa. In six cases, an exaggerated cystic appearance microscopically simulated cystic nephroma in the majority of the slides (Fig. 5). Only focally did areas diagnostic of CCSK emerge.

FIG. 5.
FIG. 5.:
(A) Dilation of entrapped renal tubules in the medulla simulates cystic nephroma. (B) Focal areas within the septa show the classic features of CCSK on high magnification.
Figure 5
Figure 5:

While most tumors (91%) had the classic pattern as either a predominant or a secondary morphology, a majority also demonstrated one or more variant patterns. These patterns usually blended smoothly with the classic pattern or another variant pattern, with two exceptions below. The variant patterns recognized and the percentage of cases in which they were seen are as follows:

  • 1. Myxoid pattern (50%)
  • 2. Sclerosing pattern (35%)
  • 3. Cellular pattern (26%)
  • 4. Epithelioid pattern (trabecular or acinar type) (13%)
  • 5. Palisading (verocay-body) pattern (11%)
  • 6. Spindle cell pattern (7%)
  • 7. Storiform pattern (4%)
  • 8. Anaplastic pattern (2.6%)

As has been previously conceptualized, 23 the variant patterns of CCSK can best be understood as alterations of either cord or septal cell morphology. This concept does not imply these cell types are biologically distinct.

The myxoid pattern featured pools of amphophilic extracellular material separating the cord cells. This material ranged from the minute deposits to large grossly evident expanses of mucin (Fig. 4) creating large pseudocysts that were difficult to distinguish from the true cysts formed from entrapped tubules at the tumor periphery. The extracellular mucoid material possessed the histochemical staining characteristics of hyaluronic acid; whether between closely spaced cells or in large pools, it was light blue on alcian blue stain, and staining was sensitive to hyaluronidase treatment. The appearance of the cord cells at the periphery of these mucoid pools was variable. Some remained similar to their neighboring cells closer to the septa. Others acquired a more epithelioid appearance, simulating gland formation around the smaller pools. Others clung to the septa, creating an anastamosing channel pattern that simulated a vascular neoplasm.

The sclerosing pattern usually featured deposition of acellular, osteoid-like material between the septa and the cord cells. In this configuration, the cord cells became compressed on both sides by the hyaline material resulting in an Indian-file pattern (Fig. 6). When the cord cell cytoplasm became attenuated and the nucleus was compressed against the extracellular hyaline material, the appearance mimicked the hyaline intracytoplasmic inclusions of RTK. Lesions acquired a paucicellular atrophic appearance when the sclerosis was less hyaline and more collagenous in nature. The sclerosing pattern appeared to arise from the myxoid pattern in some tumors in which the amphophilic extracellular material became progressively hyalinized and pink. This sometimes resulted in nodules of hyaline sclerosis separated by a background of loose myxoid material. The hyaline sclerotic material stained blue on Masson trichrome stains, indicating its collagenous nature.

FIG. 6.
FIG. 6.:
Hyaline sclerosing CCSK pattern in which osteoid-like material compresses residual cord cells.

The cellular pattern formed discrete, well-demarcated nodules within a less cellular background. In these areas, the intercellular material spacing the nuclei apart was diminished, and nuclear overlap was common. Mitotic activity in these areas often appeared increased, and the septa were limited to thin capillaries (Fig. 7). The appearance approached that of an undifferentiated small, round, blue cell tumor, raising the differential diagnosis of PNET and blastemal Wilms tumor.

FIG. 7.
FIG. 7.:
Cellular CCSK pattern showing small nodules of overlapping cord cells, simulating blastemal condensations of Wilms tumor.

The epithelioid trabecular pattern featured plump cord cells uniformly retracted away from surrounding septa such that the cord cells aligned themselves into ribbons one or two cells thick (Fig. 8). This appearance was highly reminiscent of the trabecular growth pattern of epithelial tumors such as hepatocellular carcinoma and thyroid follicular carcinoma. When viewed in cross section, the uniform alignment of these plump cord cells about the thin vascular septa created the appearance of a perivascular pseudorosette. When the central vessel was not visible, the appearance mimicked that of the Homer Wright rosettes of PNET or the true tubular differentiation of Wilms tumor. Like the cellular pattern, the trabecular pattern also tended to form discrete nodules within a tumor containing other CCSK patterns.

FIG. 8.
FIG. 8.:
Epithelioid patterns of CCSK. (A) Acinar epithelioid pattern simulates the tubules of Wilms tumor. (B) Trabecular epithelioid pattern simulates carcinoma. Neither pattern is immunoreactive for cytokeratin (not shown).
Figure 8
Figure 8:

An acinar epithelioid appearance (Fig. 8) occurred when the cord cells aligned their nuclei against the septa and their prominent eosinophilic cytoplasm toward the center of the cord cell clusters. In contrast to the true tubules of a Wilms tumor, such structures did not feature well-defined luminal borders.

The palisaded “verocay-body” pattern resulted from the alignment of spindled cord cells in parallel linear arrays with nuclei perpendicular to the septa (Fig. 9). The septa were usually surrounded by collagenous material in these formations, creating a pink contrast to the aligned blue nuclei. The appearance raised the differential diagnosis of schwannoma, particularly in soft tissue metastasis.

FIG. 9.
FIG. 9.:
Palisading pattern of CCSK simulates the verocay bodies of schwannoma.

Spindle cell patterns resulted in a variety of appearances that suggested the diagnosis of other sarcomas. Spindled transformation of both cord and septal cells produced a storiform pattern of growth that mimicked fibrohistiocytic neoplasms (Fig. 10). More commonly, the cord cell/septal cell demarcation was completely lost, resulting in a monomorphic high-grade sarcomatous appearance. In a spindled nucleus the fine chromatin characteristic of CCSK was more difficult to appreciate. In several cases, most of the sections showed a high-grade spindle cell sarcoma appearance, with the diagnosis of CCSK becoming evident only in small characteristic foci or in recurrences (Fig. 11).

FIG. 10.
FIG. 10.:
Storiform pattern of CCSK, reminiscent of that of fibrohistiocytic neoplasms.
FIG. 11.
FIG. 11.:
(A) Purely spindle CCSK raises the differential of a variety of sarcomas, such as monophasic synovial sarcoma and cellular congenital mesoblastic nephroma. (B) The lesion's identity is apparent in its recurrence, which shows the characteristic cellular septa and myxoid pools.
Figure 11
Figure 11:

Anaplasia, defined by nuclear hyperchromasia, nuclear gigantism, and atypical mitoses, was identified in nine tumors (Fig. 12). Six of these tumors were untreated; in one the anaplasia was diffuse, whereas the other five showed only discrete foci in anaplasia. Anaplasia was identified in three of 53 histologically examined posttreatment specimens. In two of these, the anaplasia was identified in a nephrectomy specimen obtained after preoperative dactinomycin/vincristine therapy. The anaplasia was focal and no preoperative biopsy had been performed. In the other treated case, anaplasia appeared in a brain metastasis of a non-anaplastic primary tumor following intensive cranial radiation therapy.

FIG. 12.
FIG. 12.:
Anaplastic CCSK. (A) Anaplastic focus in this untreated tumor appears as a discrete nodule at low power. (B) High-power view shows atypical mitoses and monstrous nuclei diagnostic of anaplasia. (C) Immunohistochemical stain for p53 shows selective intense staining of the anaplastic focus.
Figure 12
Figure 12:
Figure 12
Figure 12:

No one histologic pattern was overrepresented in lymph node metastases at presentation. Occasionally, a pattern not seen in the primary tumor was evident in its metastasis. However, the entrapped renal tubules and the epithelioid patterns seen in primary tumors were not encountered in a metastasis.

Posttreatment relapses generally displayed the same range of patterns seen in primary tumors, with the notable absence of the epitheliod pattern. However, a trend toward more sclerotic, less cellular lesions was seen, possibly reflecting treatment effect. These non-pleomorphic, sclerotic relapses often entrapped native epithelial structures (bile ducts, alveoli) in their metastatic sites, mimicking benign bile duct or pulmonary hamartomas. Two unique patterns were identified among the metastases. The first was a hypocellular spindle cell pattern that simulated fibromatosis (Fig. 13). On diligent search, small residual foci of epithelioid cord cells could be identified, allowing a diagnosis of metastatic CCSK to be made with confidence. The second was a paucicellular myxoid pattern that featured pools of myxoid material bearing only rare free-floating tumor cells. 9 Confusion with myxoma was heightened by a focal absence of the characteristic CCSK vasculature.

FIG. 13.
FIG. 13.:
Recurrent CCSK presenting as a leg mass. Hypocellular spindled pattern, prominent vasculature, and entrapment of muscle all are reminiscent of fibromatosis. This lesion is the recurrence of the tumor having the classic pattern shown in Figure 3.

Immunohistochemical Findings

A paraffin-embedded tissue block was available for immunohistochemical staining from a total of 45 cases of CCSK. Vimentin was immunoreactive in all cases, with staining often accentuated in a perinuclear dot-like pattern. This consistent staining served as a control for antigen preservation in each of the blocks studied, making negative results with other immunostains believable even in the absence of internal controls on the slide.

In 13 of 31 cases studied, the cord cells showed weak cytoplasmic staining for muscle specific actin. However, the reactivity was always less intense than that of the pericytic cells surrounding internal capillaries. All 30 cases tested were negative for desmin, including all of the ones that showed weak actin staining.

Tumor cells were not immunoreactive for either cytokeratin AE1/AE3 or CAM5.2 in any of the 41 cases studied. Stains for EMA were similarly negative in 31 cases. Nonetheless, these stains highlighted entrapped native renal tubules that easily could be missed on routine hematoxylin and eosin-stained sections, emphasizing the diagnostically helpful infiltrative nature of these tumors. Of note, the cytokeratin-negative cases included 13 in which distinct epithelioid foci were present: 4 showed trabecular epithelioid patterns, 4 contained microcystic foci, and 5 showed acinar formation simulating true glands. One of these acinar cases was previously reported as a case of CCSK showing epithelial differentiation. 31 The authors illustrated a silver stain which had converted from a single-cell (sarcomatous) pattern to a nested (epithelial) pattern in the acinar foci, although all their immunohistochemical stains had been unsuccessful. By destaining a stained slide and performing antigen retrieval before cytokeratin staining, we found the acinar focus to be cytokeratin-negative with positive staining of adjacent renal pelvic urothelial mucosa serving as an internal control.

Immunohistochemical stains for the MIC2 antigen failed to reveal the characteristic membranous pattern in any of 24 cases studied, including four with cellular foci that raised the differential diagnosis of PNET. Some cases did show blush cytoplasmic staining, a pattern generally considered to be a negative result with this antibody. Other negative stains included S100 protein (23 cases), Factor 13a (18 cases), synaptophysin (20 cases), GFAP (20 cases), LCA (20 cases), NSE (23 cases), c-kit (9 cases), and polyclonal CEA (23 cases). CD34, Factor 8, and Ulex immunostains highlighted thin capillaries within the tumor, but neither antibody-stained tumor cells in 20 cases studied.

To assess proliferative activity, immunohistochemical stains for the MIB-1 antigen were performed on 22 cases. One hundred tumor nuclei were counted from four separate areas on each slide to obtain the percentage of nuclei stained (staining index) for each case. A mean staining index of 21% was obtained (range among cases: 8%–32% of cells staining). No obvious correlation with histologic pattern was noted in the sections available for staining.

A similar scoring system was used to evaluate immunohistochemical stains of p53. Among 29 non-anaplastic CCSKs, 25 showed minimal (less than 5% of nuclei) staining. The four other non-anaplastic CCSKs had intermediate staining indices of 11%, 18%, 25%, and 40%. Of note, the lesion showing 40% staining was a posttreatment metastasis. In contrast, two of three anaplastic CCSKs available for staining showed intense, diffuse (>75% of nuclei) staining for p53. Non-anaplastic foci present on the slides in both of these cases showed the minimal p53 staining pattern usually seen in non-anaplastic CCSK, so that a sharp transition to the anaplastic foci could be appreciated (Fig. 12).

Prognostic Variables

Multivariate analysis revealed four independent significant prognostic factors for survival: treatment with doxorubicin, revised stage, patient age, and the presence of tumor necrosis (Table 3). The significance of tumor necrosis was lost when relapse-free survival was the end point. Variables lacking statistical significance for either survival or relapse-free survival included patient gender, tumor diameter, specimen weight, lymph node status, and any single histologic pattern other than necrosis.

CCSK—multivariate analysis

The addition of doxorubicin to therapeutic regimens significantly improved outcome for children with CCSK (Fig. 2;Tables 4, 5, and 6). This beneficial effect accounts for the progressive improvement in both relapse-free survival and overall survival that was noted from NTWS 1 through 4. Once the progressively increased use of doxorubicin over these trials was taken into account, the survival differences across the trials lost their significance.

Tumor-specific survival by stage at presentation
Tumor-specific survival by age at diagnosis
Tumor-specific survival by necrosis

The remarkable 98% overall survival rate among revised stage 1 patients was significantly better than those of the more advanced stages (Table 4;Fig. 14). The survival rates for stages 2 and 3 were diminished, but rose from approximately 30% to 75% with the addition of doxorubicin to therapeutic regimens. The few stage 4 patients had a poorer prognosis.

FIG. 14.
FIG. 14.:
Stage-specific survival curves for CCSK patients treated with doxorubicin.

Patients within 1 year of the mean age of diagnosis (ages 2–4 yrs) survived more frequently than those younger (less than 2 years of age) or older (greater than 4 years) than the mean (Table 5). The survival for patients in the latter two categories was improved when doxorubicin was added to therapy.

Necrosis proved to be an adverse prognostic factor for survival. The minority of tumors lacking gross or histologic evidence of necrosis were associated with a high (97%) survival rate (Table 6).

Among 68 patients with tumor relapse, both doxorubicin therapy and tumor necrosis significantly affected tumor-specific survival, whereas age and stage at diagnosis did not. Doxorubicin therapy was associated with a 66% reduction in tumor-related mortality (p = 0.02), whereas necrosis was associated with a 5.8-fold increased risk of death resulting from tumor (p = 0.004).


This study is the largest series of CCSK ever reported. We have confirmed several of the previous concepts of this tumor originating from smaller studies and reached several new conclusions. We confirmed the beneficial effect of doxorubicin therapy by multivariate analysis for both survival and relapse-free survival. Prior studies had demonstrated the trend toward better outcomes with doxorubicin, but the results were not statistically significant. 10 The larger number of patients in the prognostic group in this study and the requirement for central pathologic review likely permitted statistical significance to be attained. With this review, 22 CCSKs were excluded from analysis, leaving a more homogeneous population of tumors. This study also confirmed the striking propensity of this tumor to metastasize to bone, which we documented as its most common site of recurrence. In contrast, Wilms tumor classically spreads to lymph nodes, lung, and liver but rarely metastasizes to bone. We have also documented a large number of unusual sites of metastasis, particularly in the soft tissue, brain, and orbit, that could potentially fool the unwary oncologist. These findings emphasize the need for complete physical examinations and bone imaging 8 in clinical follow up of these patients. Prolonged vigilance is also required, because we found that approximately 20% of documented CCSK metastases occurred 3 years or more after diagnosis and some as long as 10 years later. In contrast, recurrences of Wilms tumor almost always present within 2 years. Doxorubicin therapy tended to prolong the interval to relapse among NWTS cases, but therapy alone cannot account for the propensity to late recurrence that CCSK exhibits. Indeed, the two patients with the longest intervals to relapse (9 and 10 years) did not receive doxorubicin.

An unexpected finding was the high (29%) frequency of lymph node metastases identified at presentation. This figure is extraordinarily high for sarcomas that classically spread through hematogenous routes. This finding is concordant with the high frequency of vascular invasion noted in this study, and together they suggest that CCSK has a propensity to permeate the renal and perirenal lymphovascular system. This concept may explain the high number of local (abdominal) recurrences that we have documented. Complete nodal dissections may be required to remove all tumor in such cases, as is suggested by the cases in which we found positive periaortic lymph nodes despite histologically confirmed benign hilar nodes. Whether such more extensive surgery would decrease the chances of local recurrence or impact on survival remains to be proven.

Another unexpected but likely related finding was the high survival rate (98%) for patients with revised stage 1 disease, results that are at variance with previous studies which have shown a high mortality among all stages. 23,32 Such reports have fostered the view that CCSK disseminates early so that apparent low-stage patients have micrometastases at presentation. Such studies also have shaped NWTSG therapeutic protocols under which CCSKs are treated similarly regardless of stage.

The difference in the staging system we used compared with previous studies likely accounts for the differences in our results. Under updated NWTS 5 criteria, invasion of renal sinus vasculature, not gross protrusion beyond the hilar plane, is a basis of upstaging from stage 1 to stage 2. This criterion makes more biologic sense, because the sinus vessels represent the primary route of lymphatic and venous outflow and hence tumor exit from the kidney. Twenty-three of the NWTS cases previously classified as stage 1 became stage 2 tumors on our review, usually on the basis of renal sinus vascular invasion. These tumors, having access to the systemic circulation, would be expected to have a diminished prognosis. Once this aggressive subset was excluded, a more pristine population of stage 1 tumors was obtained which proved to have an excellent prognosis. The long-held concept that stage 1 CCSKs disseminate early seems at variance with their slow growth rate, as reflected by the relatively low MIB-1 staining indices we obtained, the documented long intervals to recurrence, and the low percentage of patients who initially presented with stage 4 disease. Furthermore, the four revised stage 1 patients in this study who did not receive the benefit of doxorubicin therapy all survived, suggesting that these tumors were not disseminated and may have been cured by surgical excision. Instead, it is likely that stage 2 tumors disseminate microscopically in this fashion, as is reflected in their diminished prognosis compared with stage 1, and account for the previously cited results with “low-stage” tumors.

The only histologic variable that independently correlated with survival was the presence of necrosis. Because necrosis is typically a feature of aggressive high-grade sarcomas, a correlation of the absence of necrosis with favorable outcome is not unexpected. This concept is supported by the magnified effect of tumor necrosis on the outcome of relapsed tumors in which surgery is unlikely to be curative but instead response to chemotherapy is required. As the effects of stage and age are both mediated through relapse, these variables may reflect the probability of complete excision. Regardless of the potential explanations, it is somewhat comforting that we can identify subsets of CCSK (that is, stage 1, ages 2–4 yrs, no tumor necrosis) with a superior prognosis; such patients may not need the intensive therapy given for patients with advanced disease. The need to minimize unnecessary therapy is underscored by the fact that treatment-related deaths (2) outnumbered tumor-related deaths (1) among our revised stage 1 patients.

Whereas the variant histopathologic patterns identified in CCSK do not appear to carry prognostic import, they do highlight differential diagnostic dilemmas. Of particular difficulty is the distinction of cellular CCSK from blastemal Wilms tumor and PNET, both of which have a prominent vascular pattern. Useful differential diagnostic criteria are presented in Table 7. Both blastemal Wilms tumor and PNET are more aggressively invasive than CCSK, entrapping whole islands of native renal parenchyma as opposed to the single tubules entrapped by CCSK. Both feature coarser chromatin than CCSK. Of particular use is our finding of consistent MIC2 negativity in CCSK. Hence, in a small biopsy or other challenging case, true membranous MIC2 staining strongly argues against CCSK. An equally challenging diagnostic problem is the distinction of epithelioid CCSK patterns, particularly acinar types, from the true tubular differentiation of a Wilms tumor. We show that stains for cytokeratin can reliably make this distinction; no CCSK in our study expressed cytokeratin regardless of how epithelioid it appeared. Finally, predominantly spindled CCSKs can be difficult to distinguish from the plump cell variant of cellular congenital mesoblastic nephroma, metanephric stromal tumor, and other sarcomas. This is particularly problematic because the characteristic fine chromatin of CCSK is more difficult to appreciate in a narrow spindled nucleus. Besides its chromatin, the best clue to CCSK is the presence of foci of its classic pattern, which may be absent in small biopsy samples of large tumors. Given that all CCSKs in our study stained negatively for desmin and S100 protein, positivity for either of these markers can be used as strong evidence against CCSK. In contrast, a significant percentage of congenital mesoblastic nephromas stain with desmin. 24

Renal cellular blue cell lesions: differential diagnosis

This study also clearly defines the existence of an anaplastic subset of CCSKs. This subtype is slightly less common (3%) among CCSKs than anaplasia among Wilms tumors (5%). Our results refute our previous assumption that anaplasia in CCSK was usually related to treatment effect. First, anaplasia was not overwhelmingly more frequent in post-therapy specimens examined (3 of 53 [6%]). Second, in only one case did we document that a non-anaplastic tumor became anaplastic after therapy. Hence, anaplasia in CCSK, as with Wilms tumor, appears to arise de novo. Of particular interest was the dramatic p53 overexpression in the anaplastic foci of two tumors in which the non-anaplastic regions did not overexpress p53. The low frequency of p53 positivity in non-anaplastic CCSK likely represents native p53 expressed by proliferating cells; it is unlikely to represent mutated p53, which usually accumulates enough to be detected in a higher percentage of cells. These results are consistent with a model set by other pediatric tumors, such as Wilms tumor and rhabdomyosarcoma, in which only anaplastic tumors show p53 protein overexpression that results from gene mutation. 1,26 The absence of genetic losses at 17p in a recently completed comparative genomic hybridization study of CCSK further argues against p53 alterations being a primary genetic event in CCSK. 28

This study more sharply defines the clinical behavior and pathologic features of CCSK but the histogenesis of this tumor remains an enigma. While its nearly exclusive occurrence in the kidney suggests an origin from the developing renal mesenchyme, no viable candidate cell of origin is apparent. The presence of verified extrarenal tumors in this and other studies 14,33 that are indistinguishable from CCSK perhaps favors origin from a non-organ-specific mesenchymal cell. The eventual delineation of this cell of origin will no doubt further our understanding of this tumor.


The National Wilms Tumor Study Group is supported primarily by United States Public Health Service Grant CA-42386. Additional funding for this study was provided by a generous award from the Children's Cancer Foundation of Maryland. The authors thank the many institutional pathologists who have over the years provided the case material on which this study is based. They also thank those pathologists who sent blocks specifically for use in this study: Dr. David Kelly (Birmingham, AL) and Dr. Philip Faught (Indianapolis, IN); the staff of the NWTSG Data and Statistical Center, particularly Dr. Weiva Sieh, Tracy Bergemann, and Janice Takashima, for their tireless work; Mary North, Josephine Geh, and George Pettis for performing the immunohistochemical stains; Lisa Madden for typing the manuscript and creating Figure 1; and Pete Lund for photographic assistance.


1. Bardeesy N, Falkoff D, Petruzzi MJ, et al. Anaplastic Wilms' tumour, a subtype displaying poor prognosis, harbours p53 gene mutations. Nat Genet 1994; 7:91–7.
2. Beckwith JB. Metanephric stromal tumor (MST): a new renal neoplasm resembling mesoblastic nephroma (MN) but related to metanephric adenofibroma (MAF) [Abstract]. Mod Pathol 1998; 11:1P.
3. Beckwith JB. Renal tumors. In: Stocker JT, Askin FB, eds. Pathology of Solid Tumors in Children. New York, NY: Chapman & Hall Medical, 1998: 1–23.
4. Beckwith JB, Larson E. Case 7. Clear cell sarcoma of kidney. Pediatr Pathol Lab Med 1989; 9:211–8.
5. Beckwith JB, Palmer NF. Histopathology and prognosis of Wilms tumors: results from the First National Wilms' Tumor Study. Cancer 1978; 41:1937–48.
6. Cheah PL, Looi LM. Implications of p53 protein expression in clear cell sarcoma of the kidney. Pathology 1996; 28:229–31.
7. Faria P, Argani P, Epstein JI, Reuter VE, Beckwith JB, Ladanyi M. Primary synovial sarcoma of the kidney: a molecular reappraisal of a subset of so-called embryonal renal sarcoma [Abstract]. Mod Pathol 1999; 12:94A.
8. Feusner JH, Beckwith JB, D'Angio GJ. Clear cell sarcoma of the kidney: accuracy of imaging methods for detecting bone metastases. Report from the National Wilms' Tumor Study. Med Pediatr Oncol 1990; 18:225–7.
9. Florine BL, Simonton SC, Sane SM, Stickel FR, Singher LJ, Dehner LP. Clear cell sarcoma of the kidney: report of a case with mandibular metastasis simulating a benign myxomatous tumor. Oral Surg Oral Med Oral Pathol 1988; 65:567–74.
10. Green DM, Breslow NE, Beckwith JB, Moksness J, Finklestein JZ, D'Angio GJ. Treatment of children with clear-cell sarcoma of the kidney: a report from the National Wilms' Tumor Study Group. J Clin Oncol 1994; 12:2132–7.
11. Haas JE, Bonadio JF, Beckwith JB. Clear cell sarcoma of the kidney with emphasis on ultrastructural studies. Cancer 1984; 54:2978–87.
12. Hennigar RA, Beckwith JB. Nephrogenic adenofibroma. A novel kidney tumor of young people. Am J Surg Pathol 1992; 16:325–34.
13. Ishii E, Fujimoto J, Hara S, Tanaka S, Hata J. Human sarcomatous Wilms' tumor lines: evidence for epithelial differentiation in clear cell sarcoma of the kidney. Cancer Res 1989; 49:5392–9.
14. Kataoka Y, Shimada H, Sugimoto T, et al. Congenital sarcoma in the terminal ileum histologically resembling clear cell sarcoma of the kidney: a case report with an immunohistochemical study. Hum Pathol 1993; 24:1026–30.
15. Kidd JM. Exclusion of certain renal neoplasms from the category of Wilms tumor [Abstract]. Am J Pathol 1970; 58:16a.
16. Knisely AS, Magid MS, Alonso D, Hilgartner M. Clear-cell sarcoma of the kidney can coexpress cytokeratin and vimentin [Abstract]. Lab Invest 1985; 56:4P.
17. Kodet R, Stejskal J, Malis J, Horak J. Bone metastasizing renal tumor of childhood. A clinicopathological study of eleven cases from the Prague Pediatric Tumor Registry. Pathol Res Pract 1994; 190:750–8.
18. Looi LM, Cheah PL. An immunohistochemical study comparing clear cell sarcoma of the kidney and Wilms' tumor. Pathology 1993; 25:106–9.
19. Marsden HB, Lawler W. Bone-metastasizing renal tumour of childhood. Br J Cancer 1978; 38:437–41.
20. Marsden HB, Lawler W. Bone metastasizing renal tumour of childhood. Histopathological and clinical review of 38 cases. Virchows Arch [Pathol Anat] 1980; 387:341–51.
21. Marsden HB, Lawler W, Kumar PM. Bone metastasizing renal tumor of childhood: morphological and clinical features, and differences from Wilms' tumor. Cancer 1978; 42:1922–8.
22. Morgan E, Kidd JM. Undifferentiated sarcoma of the kidney: a tumor of childhood with histopathologic and clinical characteristics distinct from Wilms' tumor. Cancer 1978; 42:1916–21.
23. Murphy WM, Beckwith JB, Farrow GM. Tumors of the Kidney, Bladder, and Related Urinary Structures (Third Series, Fascicle 11). Washington, DC: Armed Forces Institute of Pathology, 1993:67–81.
24. Nadasdy T, Roth J, Johnson DL, et al. Congenital mesoblastic nephroma: an immunohistochemical and lectin study. Hum Pathol 1993; 24:413–9.
25. Ogawa K, Nakashima Y, Yamabe H, Hamashima Y. Clear cell sarcoma of the kidney. An immunohistochemical study. Acta Pathologica Japanica 1986; 36:681–9.
26. Pascasio J, Triche T, Sorensen P, Barr F, Qualman S, Intergroup Rhabdomyosarcoma Study. P53 expression correlates with anaplasia in rhabdomyosarcoma [Abstract]. Mod Pathol 1998;11:4P.
27. Pawade J, Soosay GN, Delprado W, Parkinson MC, Rode J. Cystic hamartoma of the renal pelvis. Am J Surg Pathol 1993; 17:1169–75.
28. Perlman EJ, Schuster AE, Argani P, et al. Genetic analysis of clear cell sarcoma of kidney (CCSK) by comparative genomic hybridization (CGH) [Abstract]. Mod Pathol 1999; 12:4P.
29. Rodriguez-Galindo C, Marina NM, Fletcher BD, Parham DM, Bodner SM, Meyer WH. Is primitive neuroectodermal tumor of the kidney a distinct entity? Cancer 1997; 79:2243–50.
30. Sandstedt BE, Delemarre JF, Harms D, Tournade MF. Sarcomatous Wilms' tumour with clear cells and hyalinization. A study of 38 tumours in children from the SIOP nephroblastoma file. Histopathology 1987; 11:273–85.
31. Schmidt D, Harms D, Evers KG, Bliesener JA, Beckwith JB. Bone metastasizing renal tumor (clear cell sarcoma) of childhood with epithelioid elements. Cancer 1985; 56:609–13.
32. Sotelo-Avila C, Gonzalez-Crussi F, Sadowinski S, Gooch WM, Pena R. Clear cell sarcoma of the kidney: a clinicopathologic study of 21 patients with long-term follow-up evaluation. Hum Pathol 1985; 16:1219–30.
33. Weeks DA, Malott RL, Zuppan C, Mierau GW, Beckwith JB. Primitive pelvic sarcoma resembling clear cell sarcoma of kidney. Ultrastruct Pathol 1991; 15:403–8.

Clear cell sarcoma; Kidney; Childhood cancer

© 2000 Lippincott Williams & Wilkins, Inc.