Search

Evidence-based pathology advocates using a combination of best available data ("evidence") from the literature and personal experience for the diagnosis, estimation of prognosis, and assessment of other variables that impact individual patient care. Evidence-based pathology relies on systematic reviews of the literature, evaluation of the quality of evidence as categorized by evidence levels and statistical tools such as meta-analyses, estimates of probabilities and odds, and others. However, it is well known that previously "statistically significant" information usually does not accurately forecast the future for individual patients. There is great interest in "cognitive computing" in which "data mining" is combined with "predictive analytics" designed to forecast future events and estimate the strength of those predictions. This study demonstrates the use of IBM Watson Analytics software to evaluate and predict the prognosis of 101 patients with typical and atypical pulmonary carcinoid tumors in which Ki-67 indices have been determined. The results obtained with this system are compared with those previously reported using "routine" statistical software and the help of a professional statistician. IBM Watson Analytics interactively provides statistical results that are comparable to those obtained with routine statistical tools but much more rapidly, with considerably less effort and with interactive graphics that are intuitively easy to apply. It also enables analysis of natural language variables and yields detailed survival predictions for patient subgroups selected by the user. Potential applications of this tool and basic concepts of cognitive computing are discussed., (Copyright © 2016 Elsevier Inc. All rights reserved.)

Recent advances in molecular pathology and other technologies such as proteomics present pathologists with the challenge of integrating the new information generated with high-throughput methods with current diagnostic models based mostly on histopathology and clinicopathologic correlations. Parallel developments in the field of medical informatics and bioinformatics provide the technical and mathematical methods to approach these problems in a rational manner. However, it remains unclear whether pathologists or other medical specialists will become primarily responsible for the development and maintenance of these multivariate and multidisciplinary diagnostic and prognostic models that are hoped to provide more accurate, individualized patient-based information. Evidence-based medicine (EBM) and medical decision analysis (MDA) are relatively new disciplines that use quantitative methods to assess the value of information, differentiate fact from myth, and integrate so-called best evidence into multivariate models for the assessment of prognosis, response to therapy, selection of laboratory tests, and other complex problems that influence individual patient care. We review from an epistemological viewpoint the current approach to information in pathology and describe some of the concepts developed by the practitioners of EBM and MDA.

Ovarian tumors of low malignant potential (LMPs) histologically lack invasion and have excellent prognosis in contrast to invasive carcinomas. Integrins are heterodimeric adhesion molecules thought to play a role in cell migration and tumor progression. The alpha(v)beta3 integrin in particular mediates melanoma invasion in vitro and promotes neoplastic angiogenesis, facilitating breast cancer metastasis. In addition, alpha(v)beta3 expression has been detected in ovarian cancer cell lines and in a limited number of human ovarian cancer samples. The distribution pattern of this integrin in LMPs is not known. We examined tissue sections from 19 LMPs and 31 ovarian carcinomas for alpha(v)beta3, in addition to alpha5beta1 and alpha2beta1 integrins, which have been shown to be expressed in ovarian cancer cell lines. Variable immunoreactivities of alpha5beta1 and alpha2beta1 were detected in both LMPs and ovarian carcinomas. Most (74.2%) ovarian carcinomas were alpha(v)beta3 positive, whereas only 36.8% of LMPs were positive, which is statistically significant (P = .009). These results establish the distribution pattern of alpha2beta1, alpha5beta1, and alpha(v)beta3 integrins in LMPs. The biological significance of the less frequent expression of alpha(v)beta3 in LMPs compared with carcinomas of the ovary needs further elucidation.

At a light microscopic level, the separation of C-cell hyperplasia and microscopic medullary carcinoma of the thyroid (MCT) is difficult, and it ultimately rests on the finding of C cells outside of the thyroid follicular basement membranes (FBMs). To date, this has required ultrastructural examination for proper documentation. The assessment of thyroidectomy specimens from patients with multiple endocrine neoplasia, type 2a (MEN2a), a hereditary condition in which there is widespread C-cell hyperplasia (CCH) and multifocal MCT, presented an opportunity to the authors to assess the entire range of C-cell abnormalities. Total thyroidectomy specimens from 17 patients with MEN2a were examined. In addition to hematoxylineosin (H&E) stains, representative tissue sections were labeled for chromogranin A and collagen type IV (CIV), using the avidin-biotin-peroxidase complex (ABC) method. All patients in the study had multifocal C-cell proliferation that was both diffuse and nodular. Fifteen had microscopic MCTs, which were multifocal in eight instances. Three patterns of C-cell proliferation were recognized in CIV immunostains. The first was characterized by complete investment of C-cells by a continuous rim of CIV, corresponding to FBM and confirming an intrafollicular localization; hence, the diagnosis of CCH was made in such cases. The second pattern was distinctive and was typified by defects in the CIV layer; constituent C-cells assumed an extrafollicular location. These images yielded a diagnosis of micro-MCT. The latter findings were also accompanied by focal reduplication of basement membrane that was apparently tumor derived, producing a micronodular or microlobular configuration. The third pattern represented a combination of the first two, with C-cell nodules that were bounded by CIV and clearly situated in an intrafollicular location; however, focal reduplication of basement membranes was also evident in these cases. The biological significance of the third pattern of CIV staining is uncertain, but it may reflect the presence of a preinvasive proliferation of C-cells that is distinct from "usual" CCH in MEN2a.

The bcl-2 gene product (bcl-2 protein, BCLP) prevents apoptotic cell death. Via a 14;18 chromosomal translocation, BCLP is overexpressed in most follicular lymphomas as well as some other non-Hodgkin's lymphomas, and it has also been documented in other nonlymphomatous malignancies. To address the possible prognostic value of this marker in predefined subsets of non-small cell lung carcinoma (NSCLC), the authors studied 126 T1N0M0 cases seen between the years 1986 to 1991 at our institution. Patients were treated by lobectomy (105 cases) or wedge excision (21 cases) with negative margins; neuroendocrine carcinomas of all grades were specifically excluded. The mean follow-up period was 39 months. Immunostaining for BCLP was done using a monoclonal antibody (clone no. 124; DAKO, Carpinteria, CA), and the avidin-biotin-peroxidase complex (ABC) technique. The study cases included 73 adenocarcinomas (ACs) as well as 40 squamous cell (SCC), five adenosquamous (ASC), and eight large cell/poorly differentiated (LCC) carcinomas. As assessed with the Kaplan-Meier method, overall survival was 64% at 5 years (66% AC vs 59% SC). BCLP was detected in 47 of 126 cases (37%) including 32 AC (44%), 10 SCC 925%), two ASC (40%), and three LCC (38%). No significant difference in 5-year survival was noted in a comparison of all cases with BCLP expression (63%) and those without (59%). There was, however, a significant difference in the survival of grade 1 BCLP(+) cases, when compared with grade 2 or 3 BCLP(+) cases (P = .01). A nonstatistically significant trend toward increased survival was observed in BCLP(+) SCC cases (66% 5-year survival in BCLP[+] vs 45% in BCLP[-] [P = .11]). Proportional hazards analysis failed to disclose significant independent risk factors. These data suggest that bcl-2 protein immunoreactivity has limited prognostic value in the pathological evaluation of NSCLC.

Although the capacity for some pulmonary carcinomas to mimic sarcomas is well recognized, their potential resemblance to selected benign lesions of the lung is currently underappreciated. The authors herein report three examples of sarcomatoid bronchogenic carcinoma with a deceptively bland appearance and an investment of reactive inflammation, such that they resembled pseudotumors histologically. These lesions occurred in two men and one woman who were 44, 61, and 63 years old, respectively, at diagnosis. All patients presented with a productive cough, hemoptysis, or chest pain. Their pulmonary masses were irregularly marginated radiographically, and ranged in size from 2.5 to 5.5 cm. Two were treated with lobectomy, and one underwent a wedge excision, followed by radiotherapy to the thorax. Despite these measures, each patient with inflammatory sarcomatoid carcinoma (ISC) died of disease or is likely to do so. Microscopically, ISCs were composed of uniform spindle cell proliferations with only modest nuclear pleomorphism, limited mitotic activity, and an arrangement in fascicles, storiform configurations, or haphazard arrays. Lymphocytes and plasma cells were interspersed throughout each of them, and keloidal stromal collagen was apparent internally in two examples. Two of the neoplasms also invaded pulmonary blood vessels or bronchi. A comparison group of 10 adults with pulmonary inflammatory pseudotumors (IPs) of the fibrous histiocytoma type shared several clinical attributes with ISC and showed closely similar histological features, except that the IPs lacked mitoses and invasiveness, and contained xanthoma cells or multinucleated elements in some cases in this series. Immunohistochemical analyses showed consistent dissimilarities between ISC and IP; keratin and epithelial membrane antigen were present in ISC but not IP, whereas actin was observed only in the proliferating spindle cells of IP. In summary, the potential clinicopathologic overlap between ISC and IP suggests that caution should be exercised in the separation of these two lesions. In particular, it is unwise to attempt to make this distinction in an intraoperative frozen section setting.

The pathological findings in the allografts of 14 children who underwent lung transplantation (LT) at St. Louis Children's Hospital, St. Louis, MO, in the period between July 1990 and May 1992 were reviewed. The study is based on histological analysis of 63 transbronchial biopsy (TBB) specimens, eight open lung biopsy specimens, and three pneumonectomy specimens. The mean age at transplantation was 10.5 years (range, 1 to 17 years) and the average follow-up period was 5.7 months. Sufficient tissue for an adequate pathological examination was obtained in 58 (92%) TBB specimens. Each specimen consisted of a mean of 6.12 tissue fragments, but only 4.79 fragments contained actual lung parenchyma for suitable examination. Ten patients (71%) had 23 biopsy-proven episodes of acute rejection with a frequency of 1.64 episodes per patient. The first episode was documented at a mean of 19 days after transplantation. Six patients (42.8%) developed bronchiolitis obliterans (BO). The definitive diagnosis of this condition was made either by open lung biopsy (n = 3) or on allograft pneumonectomy (n = 1), and it was infrequently recognized by TBB. Four of the six patients died less than 9 months after the diagnosis of BO was made, indicating the grave consequences of this complication. Two other deaths were attributed to the development of posttransplantation lymphoproliferative disorders.

Pseudovascular adenoid squamous cell carcinoma (PASCC) is a variant epithelial neoplasm with the ability to simulate the growth pattern of angiosarcoma. It has been documented in the breast, skin, and, recently, lung. We describe three additional examples of pulmonary PASCC occurring in two men and one woman. The patients' ages ranged from 47 to 54 years at diagnosis, and all patients had radiographically and macroscopically typical squamous carcinomas of the lung. Histologically, the tumors showed the presence of interanastomosing pseudoluminal spaces that were lined by obviously atypical epithelioid cells and focally contained erythrocytes. Overtly carcinomatous growth was apparent only very focally. A comparison case of true pleuropulmonary angiosarcoma in a 57-year-old man showed closely similar microscopic features, but it lacked areas that resembled conventional carcinomas. Immunohistologic studies revealed uniform reactivity for keratin in PASCC of the lung. Two cases also stained positively for epithelial membrane antigen and vimentin, but all of them were negative for von Willebrand factor, CD31, CD34, and binding of Ulex europaeus I lectin. Electron microscopic examination of the three cases showed the presence of intercellular desmosomes and cytoplasmic tonofibrils. The example of true pleuropulmonary angiosarcoma demonstrated an endothelial immunophenotype. Two of three patients with pulmonary PASCC survived for at least 20 months, whereas the individual with true angiosarcoma died within 3 months. Together with prior reports on such lesions, these data suggest that angiosarcoma-like carcinomas of the lung differ pathologically and behaviorally from primary pulmonary endothelial malignancies.

Pleomorphic adenomas of the breast (PAB) are uncommon tumors. We studied the clinicopathologic features of 10 cases of PAB, seven of which were assessed immunohistochemically. Nine patients were women, with a median age at diagnosis of 65 years. Eight patients presented with a palpable mass; two had a nipple discharge. Nine of the tumors were periareolar. The latter epidemiologic findings, coupled with histologic observations, appeared to indicate a preferential origin for PAB within large intramammary ducts. These neoplasms histologically resemble their analogues in salivary glands. Moreover, a gradual histologic transition between different morphologic areas in PAB, especially between mesenchymal and epithelial regions, supported the contention that the neoplasm arises from a single cell type capable of divergent differentiation and thus should not be considered a "mixed" tumor. This contention was further substantiated by immunohistochemical findings, in which three intermediate filaments (cytokeratin, vimentin, and glial fibrillary acidic protein) and muscle-specific actin were expressed conjointly in tumor cells with a variety of morphologic appearances. In addition, cells differentiating along mesenchymal lines by conventional microscopy were found to express epithelial membrane antigen or gross cystic disease fluid protein-15 in five cases. The benign nature of PAB was supported by a lack of metastases in this series during a median follow-up period of 4.9 years. However, one lesion recurred locally. Regarding therapy, we believe that PAB can be excised successfully with only a narrow circumferential margin of uninvolved breast.

Leiomyosarcoma (LMS) of soft tissue is a rare tumor in children. Although LMS may exhibit clinical and histologic features that are typical of smooth muscle neoplasms in adults, they may often be confused with or resemble tumors of presumed fibroblastic, myofibroblastic, or rhabdomyoblastic differentiation. As a result, the diagnosis of LMS in children is often difficult to establish with confidence. To address this problem, we analyzed the immunohistochemical features of six LMS in children (one of deep soft tissue of an extremity, one of paravertebral tissue, one of the retroperitoneum, two of oropharyngeal soft tissue, and one of subcutis); the ultrastructural features of four of these tumors were also studied. Histologically, each of the three tumors of deep soft tissue and one of the retromolar trigone were composed of pleomorphic spindle cells arranged in interweaving fascicles. In contrast, the subcutaneous tumor and the lesion of the hard palate had an epithelioid appearance. Ultrastructural features were typical of adult type LMS. Immunohistochemically, these neoplasms were diffusely reactive for vimentin, while each was negative for cytokeratin, epithelial membrane antigen, and S-100 protein. Desmin was present in all cases, but was expressed only focally in three; a similar pattern of staining was noted for muscle-specific actin, although staining was generally noted in a larger population of cells. alpha-1-Antichymotrypsin was found in five tumors, and cathepsin B reactivity was encountered in four cases. Leu-7 antigen and myelin basic protein were coexpressed by one tumor, but neither was found in the remaining cases. These results indicate that immunohistochemical detection of desmin and muscle-specific actin may be useful in the differential diagnosis of spindle or epithelioid cell proliferations in childhood when ultrastructural analysis in unavailable.

Forty-seven cases of malignant mixed müllerian tumors were reviewed histologically and studied immunohistochemically with three major objectives: to analyze the histogenetic relationship between the carcinomatous and sarcomatous components of these neoplasms, to ascertain the practical role of immunohistochemical studies in diagnosis and classification, and to determine the prognostic significance of immunohistochemically verified rhabdomyoblastic and neuroendocrine differentiation. Epithelial differentiation (cytokeratin and/or epithelial membrane antigen expression) was confirmed in all carcinomatous components; within the sarcomatous areas, it was identified among individual cells (60% of cases) and within poorly formed clusters of cells (57% of cases). There was a statistically significant tendency for concordant expression of alpha-1-antichymotrypsin, Leu-M1, S-100, Leu-7, and neuron-specific enolase between the carcinomatous and sarcomatous components of individual cases. These two findings provide evidence of common origin for the sarcomatous and carcinomatous components of these neoplasms. Histologic review of metastases in 21 cases revealed a biphasic composition in the majority of metastatic lesions (62%), another feature that further supports a common origin for the two components. From a practical standpoint, immunohistochemistry may be helpful in accentuating the biphasic pattern of these neoplasms and in verifying the presence of rhabdomyoblastic differentiation. In most cases, however, careful morphologic examination and thorough sampling will suffice for correct diagnosis and subclassification. The presence of heterologous, rhabdomyoblastic, or neuroendocrine differentiation did not have a statistically significant influence on survival; the last of these was associated with a tendency for a more rapidly fatal course.

The distinction between malignant epithelioid pleural mesothelioma (MEPM) and peripheral adenocarcinoma of the lung with pleural invasion (PAL) continues to represent a diagnostic challenge in selected cases. In order to provide comparative data on histologic, histochemical, and immunohistochemical features of these neoplasms, we analyzed 51 ultrastructurally categorized MEPMs and 52 PALs with the periodic acid-Schiff-diastase (PAS-D), mucicarmine, and colloidal iron stains, and a panel of immunohistologic reagents. Antibodies to cytokeratin, vimentin, epithelial membrane antigen (EMA), carcinoembryonic antigen (CEA), Leu M1, the B72.3 antigen, blood group isoantigens (BGI), placental alkaline phosphatase, amylase, S100 protein, and Clara cell antigen were used, as applied to paraffin sections with the avidin-biotin-peroxidase complex technique. Ultrastructural studies revealed long, branching microvilli in MEPM cells in all cases, with length-to-diameter ratios (LDR) of 10:1 or more. In contrast, PAL manifested short, nonbranching microvilli with LDR of 8:1 or less. Reactivity with PAS-D and mucicarmine stains was strictly confined to PAL, and hyaluronidase-sensitive colloidal iron-positivity was restricted to MEPM. However, only 63% and 41% of these respective neoplasms demonstrated such histochemical reactivity. Immunohistologic results correlated well with electron microscopic classification. All MEPMs and PALs were reactive for cytokeratin; in addition, the majority of tumors in each group expressed EMA, and a minority were reactive for vimentin. In adenocarcinomas of the lung, Leu M1 was observed in all cases, CEA was apparent in 96%, B72.3 labeled 84%, and BGI were present in 67%; all PALs expressed at least two of these determinants, but none was seen in any mesothelioma. The other markers included in this study also were observed in some PAL cases, but not in MEPM. These findings suggest that immunohistology parallels electron microscopy in efficacy in the diagnostic separation of MEPM and PAL. Using antibodies to Leu M1, CEA, and the B72.3 antigen, reactivity for at least two of these three markers appears to exclude a diagnosis of pleural mesothelioma. The other glycoproteinaceous, oncoplacentofetal, and cytoplasmic antigens we studied can be used to reinforce such a determination, since their distribution is confined to adenocarcinomas.

In an effort to investigate the utility of immunostaining for S-100 protein in the differential diagnosis of chondroblastoma, the expression of S-100 protein in nine chondroblastomas was compared with that in six giant cell tumors, six aneurysmal bone cysts, four giant cell reparative granulomas, six cases of fibrous dysplasia, two cases of osteitis fibrosa cystica, two nonossifying fibromas, and one clear cell chondrosarcoma. Five enchondromas, three typical chondrosarcomas, and one mesenchymal chondrosarcoma were also included as control tumors. The proliferating stromal cells in seven of the nine chondroblastomas stained for S-100 protein, as did the lacunar chondrocytes in all of the enchondromas and chondrosarcomas and rare stromal cells in the clear cell chondrosarcoma. In contrast, none of the other tumefactive bone lesions included in this study demonstrated S-100 protein immunoreactivity. These results suggest that immunohistochemical assessment of S-100 protein may be a method for diagnostically separating chondroblastoma from pathologic entities that could be histologically confused with it in the presence of limited biopsy material. However, clear cell chondrosarcoma would appear to represent an exception to this general statement.

Sixteen cases of congenital mesoblastic nephroma (CMN) were studied. The tumors showed variable patterns of growth, degrees of cellularity, and mitotic activity. Six tumors had the classical pattern of CMN, seven were of the cellular or atypical variant and three showed combined features. The mean ages at presentation were 16 days, 5.3 months, and 2.3 months, respectively. Average size and weight were 5.1 cm and 94 g for classical CMN, 9.1 cm and 620 g for cellular CMN and 10.5 cm and 150 g for combined tumors. Cyst formation, hemorrhage and necrosis were confined to cellular CMNs and to cellular areas of combined CMNs. Mitotic activity ranged from 0 to 1/10 high-power fields (HPFs) in classical tumors to 25 to 30/10 HPFs in cellular tumors. Clear cell sarcoma-like areas were observed in three neoplasms. In ten cases there was invasion of perirenal fat; in one case each, invasion of the psoas muscle, renal vein wall, and renal vein lumen was observed. Ultrastructural and immunohistochemical studies showed features consistent with myofibroblastic differentiation. Flow cytometric analysis revealed euploidy in one classic CMN, one cellular CMN and in classic areas of a combined CMN; cellular areas of the latter tumor were aneuploid. All patients with follow-up were alive without evidence of disease after a mean period of 5 years following nephrectomy alone. No correlation was observed between the pathologic features assessed and the biologic behavior of these neoplasms.

Placental-like alkaline phosphatase (PLAP) activity has been reported in various human neoplasms of both somatic and germ cell types. The expression of PLAP was examined with a polyclonal antibody and the immunoperoxidase technique in formalin-fixed, paraffin-embedded sections of 37 germ cell neoplasms and 483 somatic tumors. The expression of keratin and epithelial membrane antigen (EMA) was concurrently assessed to determine whether these stains were helpful in distinguishing germ cell neoplasms from somatic tumors that might mimic them microscopically. All germ cell lesions were reactive for PLAP, but so were 62 somatic carcinomas, usually in female müllerian, intestinal, and lung cancers and less often in carcinomas of the breast and kidney. PLAP-reactive somatic tumors exhibited EMA and keratin positivity in the absence of prior protease digestion, whereas germ cell neoplasms failed to do so. Malignant mesotheliomas were nonreactive for PLAP, as were carcinomas of the nasopharynx, adrenals, liver, pancreas, stomach, prostate, and urinary bladder. PLAP is a highly sensitive but nonspecific immunohistologic marker of germ cell differentiation. However, non-protease-enhanced stains for keratin and EMA allow separation of germ cell and somatic carcinomas, despite their shared capacity for PLAP expression. In somatic neoplasms, PLAP immunoreactivity might be of potential use in predicting possible primary sources for metastatic tumors of unknown origin.

Immunophenotyping of hematopoietic malignancies is usually accomplished in frozen sections or cell suspensions. To determine whether this procedure was also feasible in paraffin sections, we performed a double-blind immunoperoxidase study of 65 hematopoietic tumors whose phenotypes had been determined previously in fresh tissue. A selected antibody panel was used, including anti-LN2, UCHL-1, anti-cathepsin B, anti-Leu M1, anti-MB2, and anti-MT1. A correct phenotype was obtained on paraffin sections in 95% of cases. All 31 B-cell malignancies were properly classified, showing reactivity for LN2 and MB2. In 14 of 15 T-cell hematopoietic malignancies, all cells reacted with anti-MT1 and/or UCHL-1; the 1 case negative for these antigens was misdiagnosed as a B-cell tumor because of misinterpreted LN2 reactivity in benign histiocytes. Four of 5 true histiocytic neoplasms were positive for cathepsin B and LN2 but lacked other antigens; the fifth case was wrongly considered a B-cell proliferation because only bland histiocytes displayed cathepsin B. Only 1 of 7 Hodgkin's lymphomas was misdiagnosed (as a T-cell tumor); in the other 6 cases, Reed-Sternberg cells were reactive for LN2 and LEU M1. Five of 6 extramedullary myeloid leukemias also stained for LN2, MT1, and LEU M1. One showed LN2, MB2, and MT1; this case was classified as a B-cell neoplasm and indeed represented a pre-B-cell transformation of chronic myelogenous leukemia. These results show that the specified panel of antibodies may be useful for immunophenotyping of hematopoietic neoplasms when only paraffin sections are available for analysis. However, it cannot supplant traditional cell-marker studies of hematopoietic tumors because of its lesser accuracy.

The identification of metastatic carcinoma of the breast may be difficult in the absence of a previous history of breast cancer. Various immunophenotypic markers have been introduced to aid in this process. A monoclonal antibody directed at a 15-kilodalton (kd) gross cystic disease fluid protein (GCDFP-15) was applied immunohistochemically to paraffin sections of 105 breast cancers and 585 nonmammary malignancies in order to assess its value in this context. In addition, GCDFP-15 was compared with another putative mammary epithelial marker, alpha-lactalbumin (ALA), with respect to sensitivity and specificity for a diagnosis of breast carcinoma. Overall, the rates of specificity and sensitivity and the predictive value of a positive result for GCDFP-15 were 95%, 74%, and 74%, respectively. Corresponding statistical parameters for ALA were 50%, 50%, and 23%. A consistent congruency between the reactivity patterns of primary and metastatic breast cancers was noted for GCDFP-15 but not for ALA. Besides mammary carcinomas, the major tumor types that expressed GCDFP-15 were carcinomas of the salivary glands, sweat glands, and prostate. Since the latter three types of lesions are unlikely to be diagnosed as metastatic breast cancer, statistical indices were recalculated after exclusion of these three tumor types. Following this exclusion, the adjusted rate of specificity of GCDFP-15 and the predictive value of a positive result for a diagnosis of metastatic carcinoma of the breast were each 99%. In contrast, predictive parameters for ALA were not altered. These results show that GCDFP-15 is a specific marker for breast cancer and is superior to ALA in this respect.

To analyze the purportedly epithelial features of synovial sarcoma, the antigenic profiles of 20 of these neoplasms (including 12 of the monophasic type) were studied by three different immunohistochemical techniques, and the results were correlated with ultrastructural observations in 10 cases. All of the biphasic tumors were immunoreactive for the epithelial markers epithelial membrane antigen (EMA) and cytokeratin (CK) and also had ultrastructural features of epithelial differentiation. In contrast, only one of five monophasic tumors had electron microscopic features suggestive of epithelial differentiation, but eight of the 12 were immunoreactive for CK or EMA by a procedure combining the peroxidase-antiperoxidase and avidin-biotin-peroxidase complex methods. It is concluded that synovial sarcoma, including the monophasic variant, is a mesenchymal tumor with epithelial features. Immunohistochemical studies are more sensitive than ultrastructural analysis for documenting epithelial differentiation.

To better characterize the diagnostic criteria and clinical behavior of malignant histiocytosis, 20 patients treated at the Mayo Clinic during a 25-year period were studied. A wide spectrum of cytologic differentiation was observed, with cells ranging from bland to highly anaplastic; hemophagocytosis was prominent only in conjunction with a bland histologic appearance. In surgical specimens, the diagnosis of malignant histiocytosis necessitated the use of immunoperoxidase methods for lysozyme, immunoglobulin light chain, and alpha-antitrypsin content, as well as cytochemical stains for acid phosphatase and nonspecific esterase. All autopsies showed that organ involvement had varied somewhat from that reported earlier. The correct diagnosis of malignant histiocytosis was made prior to death in only ten of the 20 cases. The mean survival in the 17 fatal cases was 7.6 months; three of the seven patients treated by aggressive chemotherapy achieved complete remission. Relatively longer survival was correlated with initial confinement to the skin and the absence of cytopenia or liver function abnormalities. Three patients with pulmonary involvement from malignant histiocytosis had apparent inappropriate antidiuretic hormone secretion in the absence of central nervous system disease. The accelerated clinical progression of malignant histiocytosis and its response to current chemotherapeutic regimens make rapid diagnosis and familiarity with the pathologic variations seen in this disorder imperative. Routine utilization of special light-microscopic and immunohistochemical stains is mandatory.

The Di George syndrome is an anomaly characterized by the complete or partial absence of derivatives of the third and fourth pharyngeal pouches often associated with defective development of the third, fourth, and sixth aortic arches leading to absence or hypoplasia of the thymus and parathyroid glands and to cardiovascular anomalies. The fifth pharyngeal pouch, often considered a part of the fourth pouch, gives rise to the ultimobranchial body (UB), which becomes incorporated into the thyroid gland and is thought to be the source of thyroid C cells. Robinson suggested that complete or partial absence of the UB should be considered a part of the Di George anomaly. To substantiate this theory, the thyroid glands of 11 patients with the Di George syndrome and 11 age-matched control infants were examined immunohistochemically using the immunoperoxidase technique for presence or absence of thyrocalcitonin (TC)-containing cells. Only three of 11 patients with the Di George syndrome had TC-containing cells in their thyroid glands (27 per cent), and nine of 11 control infants had these cells (82 per cent). It is concluded that thyroid C cell deficiency is present in most patients with Di George anomaly, suggesting a relationship between these cells and development of derivatives of the third through fifth visceral pouches. Furthermore, there is a spectrum of deficiency of thyroid C cells in these individuals comparable with the spectrum of partial to complete absence of third and fourth pharyngeal pouch derivatives regarding thymus and parathyroid glands. Immunostaining for TC of the lungs of all infants with the Di George syndrome and control infants revealed similar numbers of thyrocalcitonin-containing cells in both groups. Asynchronous development of thyroid and lung thyrocalcitonin-containing cells in those with the Di George syndrome favors the theory that the latter develop independently of derivatives of the third through fifth visceral pouches. This study further supports a neural crest origin of the Di George anomaly and strengthens the concept that the Di George anomaly is a neurocristopathy.

The ultrastructural and immunohistochemical features of 19 hepatoblastomas were examined to evaluate the phenotypic expressivity of this solid embryonic neoplasm of childhood. Electron microscopy confirmed the embryonal and fetal characteristics of the neoplastic hepatocytes, but in addition, cells with features intermediate between these two cell types were identified. Dense bundles of collagen corresponding to the osteoid-like material by light microscopy surrounded nests of cells; the cells within this matrix stained for epithelial membrane antigen and vimentin and focally for cytokeratin, and they showed ultrastructural features of epithelial cells. The two cases of small cell hepatoblastoma reacted positively for vimentin and cytokeratin; the remaining 17 cases were immunoreactive for cytokeratin and alpha-fetoprotein, and some also for alpha 1-antitrypsin, ferritin, and vimentin. A histogenetic scheme based on our findings is proposed to explain the divergent morphologic features of this neoplasm.

In an effort to provide immunocytochemical data that would be useful in distinguishing between small cell epithelial tumors of the anorectal region, 10 cases of neuroendocrine small cell colorectal carcinoma (NSCCC) and five cases of cloacogenic carcinoma (CC) were studied with antibodies to cytokeratin, epithelial membrane antigen (EMA), chromogranin, blood group isoantigens (BGI), carcinoembryonic antigen (CEA), Leu-M1, Leu-7, leukocyte common antigen (LCA), S-100 protein, neurofilaments (NF), neuron-specific enolase (NSE), serotonin, and 14 neuropeptides. The diagnoses for all 15 tumors were verified ultrastructurally. Among the antigenic determinants considered, reactivity for low- and medium-molecular-weight cytokeratin, EMA, NSE, and NF was seen in the majority of NSCCCs, whereas the CCs were positive for all cytokeratin classes, BGI, EMA, and CEA. In addition, Leu-M1, Leu-7, and chromogranin were each expressed in three cases of NSCCC. None of the other antisera yielded positive results in tumors of either type. All 10 patients with NSCCC died of their tumors within 11 months of clinical presentation, while four of the five CCs proved fatal, with an average survival of 28 months. One of the patients with CC was free of disease 31 months after diagnosis. These data suggest that an immunocytochemical panel, consisting of antibodies to high-molecular-weight cytokeratin, BGI, CEA, NSE, and NF (and possibly Leu-7 and chromogranin as well), is capable of distinguishing between NSCCC and CC in problematic cases. Although tumors of both types are aggressive, it is possible that the survival statistics for both may be improved through more accurate diagnostic classification.

Two cases of endovascular papillary angioendothelioma (EPA) of the skin were studied immunohistochemically to assess the expression of endothelial, leukocytic, and epithelial determinants by these lesions. The tumor cells in both cases were labeled by antibodies to Factor VIII-related antigen, vimentin, and blood group isoantigens; they also bound Ulex europaeus I agglutinin. On the other hand, immunoreactivity for epithelial membrane antigen, Leu-M1, HLA-DR (Ia-like antigen), and leukocyte common antigen (LCA) was not observed. Stromal and intravascular lymphocytes were labeled intensely with antibodies to LCA and HLA-DR and were intimately associated with papillary configurations of proliferating intravascular tumor cells. These results confirm the endothelial nature of EPA and suggest that the cells of this lesion may differentiate toward "high" endothelial cells, which have been shown to interact functionally with circulating lymphocytes.

A panel of paraffin effective antibodies recognizing B cells and T cells (LN-2, MB1, L26, MT1, UCHL1, kappa, lambda) was used to characterize the immunophenotypes of 26 sinonasal non-Hodgkin's lymphomas. Seventeen tumors were stage I, five were stage II, one was stage III, and three were stage IV. Nine lymphomas were classified morphologically as large cell, six were large cell immunoblastic, six were small cleaved cell, two were mixed small and large cell, two were small noncleaved cell, and one was lymphoblastic. None were follicular. Twenty-two lymphomas had a B cell immunophenotype, three were T cell neoplasms, and one was immunoreactive only for MT1. This predominance of sinonasal lymphomas with a B cell immunophenotype in patients residing in the United States contrasts with the almost exclusive occurrence of T cell sinonasal lymphomas in Chinese patients living in Hong Kong and Japanese patients residing in regions of Japan that are nonendemic for human T cell leukemia virus-1.