Poorly differentiated cancer from an unknown primary site

INTRODUCTION — Cancer of unknown primary site (CUP) is a relatively common clinical entity, accounting for approximately 2 percent of all invasive cancer diagnoses [1]. Within this category, tumors from many primary sites with varying biology are represented. Most CUP are adenocarcinomas and can be recognized by routine histologic examination. However, 20 to 25 percent of CUPs are poorly differentiated and cannot be precisely characterized by histologic examination. Approximately 80 percent of these poorly differentiated tumors have features of carcinoma and are termed "poorly differentiated carcinoma" after initial pathologic examination. In the remainder, histologic examination results in the diagnosis of "poorly differentiated neoplasm," signifying the inability to distinguish between carcinoma, melanoma, lymphoma, or sarcoma.

Every attempt should be made to arrive at the most specific diagnosis possible, since the treatments for various cancers in this group can be quite different and may be curative in some patients. The diagnostic approach to poorly differentiated cancers of unknown primary site will be reviewed here, along with the prognostic and therapeutic implications.

Other relevant topics include:

●(See "Overview of the classification and management of cancers of unknown primary site".)

●(See "Adenocarcinoma of unknown primary site".)

●(See "Squamous cell carcinoma of unknown primary site".)

●(See "Head and neck squamous cell carcinoma of unknown primary".)

●(See "Neuroendocrine neoplasms of unknown primary site".)

●(See "Axillary node metastases with occult primary breast cancer".)

CLINICAL EVALUATION

The evaluation of the patient with a cancer of unknown origin usually begins with a biopsy that documents metastatic cancer. Optimum evaluation requires a number of clinical and pathologic studies. (See 'Pathologic evaluation' below.)

●Once a diagnosis of malignancy has been established, the following procedures must fail to identify an anatomic primary site before the diagnosis of CUP can be made: history, physical examination, urinalysis, serum chemistries, complete blood counts, serum PSA (males), mammography (females), and computerized tomography (CT) or magnetic resonance imaging (MRI) of the chest, abdomen, and pelvis. In addition, focused clinical evaluation of signs and symptoms suggestive of metastatic cancer should be performed.

●Serum levels of human chorionic gonadotropin (hCG) and alpha-fetoprotein (AFP) should be obtained in all patients with the initial diagnosis of "poorly differentiated neoplasm" or "poorly differentiated carcinoma," since significant elevations of these tumor markers suggest the diagnosis of extragonadal germ cell tumor. (See "Extragonadal germ cell tumors involving the mediastinum and retroperitoneum".)

●Positron emission tomography (PET) and PET/CT scans can identify a primary site in a minority of patients. However, one prospective study did not show superiority of PET/CT when compared with CT alone [2].

●Other traditional serum tumor markers, such as carcinoembryonic antigen (CEA), CA-125, CA19-9, and CA15-3, are not useful in identifying the primary site or tissue of origin. However, these markers are often elevated, and serial measurement may be useful in subsequently following the response to therapy in an individual patient.

PATHOLOGIC EVALUATION

Initial biopsy review — The first step in the pathologic evaluation of a patient with poorly differentiated CUP is review of the biopsy material.

A nonspecific light microscopic diagnosis can be due to a biopsy specimen that is very small or poorly preserved. In particular, tissue obtained by fine needle aspiration (FNA) provides inadequate diagnostic material in patients with poorly differentiated tumors, because histologic detail is poorly preserved and the ability to perform special studies is limited by the small amount of tissue available.

Examination of poorly differentiated malignancies using routine light microscopy/histology alone is inadequate. There are no light microscopic features that can reliably distinguish between these different tumor groups or between treatment-sensitive and treatment-resistant cancers. Even with careful retrospective review of these cases, some tumors that are highly responsive to specific treatment regimens (germ cell tumor, lymphoma) were not readily distinguished morphologically [3].

Thus, tumor biopsies from these patients should undergo additional pathologic evaluation using immunohistochemistry (IHC). If the IHC results are inconclusive, a molecular cancer classifier assay (MCCA) should be performed to better characterize the cancer and aid in predicting the primary site. In selected cases, electron microscopy (EM) and chromosomal analysis should be considered if other results are inconclusive. Repeat biopsy is indicated if tissue from the initial biopsy is insufficient for these additional diagnostic tests. (See "Adenocarcinoma of unknown primary site", section on 'Molecular cancer classifier assays'.)

Immunohistochemistry — IHC is particularly valuable in identifying the tumor lineage (carcinoma versus lymphoma versus sarcoma, etc) in patients with poorly differentiated neoplasm (table 1). Once a poorly differentiated carcinoma has been identified, IHC can sometimes suggest a specific tumor type and/or primary site (table 2). IHC uses antibodies that are directed at specific cell components or products including enzymes, normal tissue components, hormones, and tumor markers. Such staining can be performed on formalin-fixed, paraffin-embedded tissue. Considerable technical expertise is required to perform these tests, and proper interpretation requires an experienced pathologist.

No IHC staining pattern is completely specific. However, a specific diagnosis is often suggested by the IHC results. A few stains (eg, leukocyte common antigen [LCA] for lymphoma; prostate specific antigen [PSA] for prostate cancer) are quite specific and rarely yield false-positive results. Most IHC stains are less specific and must be interpreted in conjunction with light microscopic and clinical features. The use of panels of IHC stains can also improve the diagnostic specificity, particularly in the differential diagnosis of various adenocarcinomas. To further complicate interpretation, some poorly differentiated neoplasms lose all specific IHC markers, prohibiting further characterization by IHC.

Examples of the potential utility of IHC in the evaluation of poorly differentiated neoplasms include:

●Monoclonal antibodies that react with antigens present on human white blood cells can reliably differentiate hematologic from nonhematologic neoplasms in formalin-fixed sections [4-8]. In one study, immunostaining for LCA (CD45) was applied to 315 lymphomas and 420 nonhematologic tumors [9]. All of the non-Hodgkin lymphomas, 12 of 25 Hodgkin lymphomas, and none of the non-hematologic tumors were LCA positive.

●Positive staining for cytokeratins (EMA, AE1/AE3) provides strong evidence for a carcinoma [10-12], although rarely positive in anaplastic undifferentiated lymphoma.

●Positive staining for vimentin or desmin, in the absence of staining for cytokeratins, is suggestive of sarcoma [10-12].

●Positive staining for S100 protein, HMB-45 antigen, and vimentin supports the diagnosis of amelanotic melanoma [13,14].

●Positive staining for factor VIII-related antigen is a specific but not sensitive marker for angiosarcoma [15,16].

●Positive staining for c-kit (CD117) supports the diagnosis of a gastrointestinal stromal tumor [17].

After a poorly differentiated cancer is recognized to be a carcinoma, IHC staining can sometimes provide further characterization (table 2). Panels of IHC stains are more effective than single stains [18-20].

●A neuroendocrine carcinoma (eg, small cell/large cell lung cancer, carcinoid, islet cell tumor) is suggested by positive immunostaining for chromogranin A or synaptophysin [21].

●Immunostaining for PSA is quite specific for prostate cancer [22].

●IHC staining for octamer binding transcription factor 4 (OCT4, also called POU domain class 5 transcription factor 1) establishes a germ cell origin for a tumor [23].

●Specific combinations of IHC staining (table 2) provide strong evidence for the diagnosis of adenocarcinomas arising in the colon, lung, kidney, or thyroid.

Molecular cancer classifier assays — Molecular cancer classifier assays (MCCAs) are diagnostic gene expression profiling assays that can predict the tissue of tumor origin in patients with CUP. This technique is based on the detection of a site-specific gene expression profile, with some assays demonstrating an accuracy rate of approximately 80 percent [24-27]. MCCAs provide information that is distinct from comprehensive molecular profiling or sequencing, which detects molecular alterations in the cancer and usually cannot identify the site of tumor origin. The role of MCCAs in the diagnosis and management of patients with CUP is evolving, but most evidence suggests the benefit of site-specific therapy, based on the MCCA predictions, in patients with treatment-sensitive tumor types. MCCAs have best been evaluated in adenocarcinoma of unknown primary, and further discussion is detailed elsewhere. (See "Adenocarcinoma of unknown primary site", section on 'Molecular cancer classifier assays'.)

Although data are limited, MCCAs appear to be valuable in the diagnosis of patients with poorly differentiated neoplasms of unknown primary site. As an example, in a group of 30 patients who retained this diagnosis after extensive IHC evaluation (median 18 IHC stains), MCCAs predicted the tumor lineage in 83 percent (carcinoma 10, sarcoma 8, melanoma 5, lymphoma 2) [28]. Additional directed testing supported the MCCA diagnoses in 11 of 15 patients. Specific sites of origin were predicted for all 10 carcinomas (including three germ cell tumors and two neuroendocrine carcinomas). Although the study was retrospective, the diagnosis determined by MCCA was used to direct therapy in seven patients (two germ cell, two neuroendocrine, two mesothelioma, one lymphoma), and five remained progression free for 25 to 72+ months.

Gene sequencing techniques (such as comprehensive molecular profiling and assaying specific targetable molecular alterations) may be useful in finding actionable genetic alterations in the tumors of some CUP patients and is discussed elsewhere. (See "Adenocarcinoma of unknown primary site", section on 'Comprehensive molecular profiling'.)

Electron microscopy — Electron microscopy requires special tissue cutting and fixation, and is relatively expensive. For these reasons, electron microscopy should be reserved for neoplasms whose lineage remains unclear after light microscopy, IHC staining, and MCCA.

Electron microscopy is extremely reliable in distinguishing lymphoma from carcinoma and is probably superior to IHC for the identification of poorly differentiated sarcomas. The presence of structures such as neurosecretory granules in neuroendocrine tumors and premelanosomes in melanoma can help identify these neoplasms.

Electron microscopy can also provide evidence for adenocarcinoma or squamous cell carcinoma in cases in which light microscopy is not distinctive. However, it is usually impossible to pinpoint the origin of poorly differentiated adenocarcinomas or squamous cell carcinomas by electron microscopic studies since few ultrastructural features are specific for any site of origin [29].

Undifferentiated tumors often lose specific ultrastructural features as well as typical histology. For this reason, the absence of a particular ultrastructural finding cannot be used to rule out a specific diagnosis.

Chromosome analysis — The identification of tumor-specific chromosome abnormalities has allowed the application of cytogenetic analysis to the evaluation of poorly differentiated neoplasms of unknown primary site in selected cases.

The utility of cytogenetic analysis can be illustrated by the following:

●The majority of B cell lymphomas contain immunoglobulin (Ig) gene rearrangements, and they usually express a unique clonal surface Ig that provides a specific tumor marker [30-32]. In lymphomas that lack detectable surface or cytoplasmic Ig, diagnostic clonal gene rearrangements are often demonstrable. (See "Normal B and T lymphocyte development" and "Tools for genetics and genomics: Cytogenetics and molecular genetics".)

●A specific chromosomal translocation (t[11;22] [q24;q12]) is found in all peripheral primitive neuroectodermal tumors and is also frequent in Ewing sarcoma [33,34]. (See "Pathogenetic factors in soft tissue and bone sarcomas" and "Epidemiology, pathology, and molecular genetics of Ewing sarcoma", section on 'Molecular genetics'.)

●An isochromosome of the short arm of chromosome 12 (i12p) can be detected in a large percentage of testicular and extragonadal germ cell tumors in males [35,36].

●A chromosomal translocation of the NUT gene is associated with aggressive, refractory midline carcinoma in children and young adults [37]. (See 'NUT midline carcinoma' below.)

●Detection of the Epstein-Barr virus genome can be an indicator of nasopharyngeal carcinoma in patients with metastatic disease in cervical lymph nodes and an undetermined primary site. (See "Epidemiology, etiology, and diagnosis of nasopharyngeal carcinoma", section on 'Epstein-Barr virus' and "Epidemiology, etiology, and diagnosis of nasopharyngeal carcinoma", section on 'Introduction'.)

TREATMENT AND PROGNOSIS — In patients with poorly differentiated neoplasms of unknown primary site, the ability to distinguish between different tumor types can have very important prognostic and/or therapeutic implications.

When specialized diagnostic studies identify a treatable neoplasm, therapy should follow guidelines established for that tumor. Examples of treatable tumor types occasionally identified in these patients include neuroendocrine carcinoma, Ewing sarcoma family of neoplasms, and a variety of primitive sarcomas. Other important tumor groups that may be identified are discussed in this section.

Lymphomas — Undifferentiated neoplasms that are identified as lymphoma respond well to the same combination chemotherapy regimens that are used for aggressive non-Hodgkin lymphoma. (See "Initial treatment of advanced stage diffuse large B cell lymphoma" and "Initial treatment of peripheral T cell lymphoma".)

In a group of 35 patients with histologically unclassified neoplasms that were identified as lymphoma by LCA staining, treatment with standard lymphoma treatment regimens (CHOP and variants) produced an actuarial disease-free survival of 45 percent at 30 months [38]. This outcome was similar to that for a group of concurrently treated patients with aggressive lymphoma and typical histology.

Extragonadal germ cell tumors — A few patients with poorly differentiated neoplasm or poorly differentiated carcinoma of unknown primary site have extragonadal germ cell tumors that are unrecognizable by standard histologic criteria. These patients typically have some of the following characteristics:

●Young age

●Male sex

●Predominant tumor location in the mediastinum or retroperitoneum

●Marked elevation of the serum human chorionic gonadotropin (hCG) or alpha-fetoprotein (AFP) levels

●Presence of 12p chromosomal gain (isochromosome 12p)

●Immunohistochemical staining for OCT4 or placental leukocyte alkaline phosphatase (PLAP) [23]

Most young men with poorly differentiated neoplasm or carcinoma and features of extragonadal germ cell tumor have an excellent response to chemotherapy, and some are cured with cisplatin-based regimens. For this reason, all patients with poorly differentiated neoplasm/carcinoma or poorly differentiated adenocarcinoma of unknown primary site who have these characteristics should be treated according to the guidelines established for poor prognosis germ cell tumors. (See "Initial risk-stratified treatment for advanced testicular germ cell tumors".)

Poorly differentiated carcinoma without characteristics of germ cell tumor — Most patients with poorly differentiated carcinoma of unknown primary site do not have features typical of extragonadal germ cell tumor, and an empiric trial of chemotherapy has been recommended. Several investigators have reported higher overall response rates in this group when compared with patients with adenocarcinoma of unknown primary site, particularly when platinum-based regimens are used [39-42]. In addition, some investigators have reported a minority (15 percent) of patients with complete response and long-term survival [43], while others have not observed such a subset.

While these older results were important in demonstrating the heterogeneity of this patient population, it is likely that currently available diagnostic methods are effective in identifying the highly responsive patients within this large group. Some of these patients have clinical or pathologic features that fit into one of the treatable subgroups described for adenocarcinoma and should be treated accordingly. In other patients, a specific tumor lineage or site of origin can be established. (See "Adenocarcinoma of unknown primary site", section on 'Specific patient subgroups'.)

Although data remain incomplete, support for site-specific therapy directed by the results of IHC stains or MCCA, rather than empiric chemotherapy, is becoming increasingly compelling, with most data in patients with adenocarcinoma of unknown primary site. However, it seems likely that MCCA will also be effective in directing treatment for patients with poorly differentiated carcinoma. For these reasons, the treatment approach for patients with poorly differentiated carcinoma should follow the guidelines used for patients with adenocarcinoma of unknown primary site. (See "Adenocarcinoma of unknown primary site", section on 'Approach to patients not included in specific subgroups'.)

Data are as follows:

●Validation of the available MCCAs was performed in patients with known primary tumors, including both adenocarcinoma and poorly differentiated carcinoma histology. Accuracy was similar regardless of histology.

●In one prospective study, 18 percent of patients had poorly differentiated carcinoma, and an additional 24 percent had poorly differentiated adenocarcinoma [44]. In this study, site-specific treatment directed by MCCA results led to a median survival of 12.5 months in the entire group of 194 patients. Those predicted to have more responsive tumor types had longer median survival than did those with less responsive types (13.4 versus 7.6 months, respectively).

●In two studies, patients predicted by MCCA to have responsive cancer types had longer survival than did those with unresponsive cancer types [45,46].

●In a retrospective series of CUP patients who had the diagnosis of poorly differentiated neoplasm after extensive immunohistochemistry (IHC) evaluation, MCCA predicted the tumor lineage and/or site of origin in 25 of 30 patients (83 percent) [28].

The choice between empiric chemotherapy and site-specific therapy based on IHC or MCCA prediction remains controversial in patients with CUP. Several studies comparing site-specific treatment and empiric chemotherapy in patients with CUP have been reported [47-49]. In randomized studies, the results of the two treatment approaches were similar [47,48]; however, the majority of patients had treatment-resistant tumor types, in which the impact of site-specific treatment is minimal. Numbers of patients with treatment-sensitive tumor types (who would be expected to derive most benefit from site-specific treatment) were too small to be meaningfully compared. These data are discussed in detail separately. (See "Adenocarcinoma of unknown primary site", section on 'Cancers with predicted treatment sensitivity'.)

The authors strongly prefer site-specific treatment for patients with treatment-sensitive tumor types. Either empiric chemotherapy or site-specific treatment is acceptable in treatment-resistant tumor types. If empiric chemotherapy is selected in these patients, a platinum-based regimen is probably preferable [40-42,50-53]. (See "Adenocarcinoma of unknown primary site", section on 'Approach to patients not included in specific subgroups'.)

NUT midline carcinoma — NUT midline carcinomas are defined by the presence of a chromosomal rearrangement of the NUT gene [37,54]. The molecular translocation involves the BRD4 or BRD3 gene in most cases, but translocation of other genes has also been described. NUT midline carcinomas are aggressive, poorly differentiated tumors that include variable degrees of squamous differentiation in approximately one-half of cases.

In a retrospective analysis that included 141 cases of NUT midline carcinoma from the NUT Midline Carcinoma Registry compiled over a 24-year period, the median age at diagnosis was 23.6 years; 51 percent of patients had predominant tumor location in the thorax, and 49 percent had other sites (head/neck, soft tissue, bone, others) [37]. More than 60 percent of patients had one or more sites of metastases. The median overall survival was 6.5 months for all patients regardless of therapy, but three risk groups were defined by anatomic site and NUT fusion type: thoracic tumor (n = 67), median survival 4.4 months; nonthoracic tumor with BDR4-NUT alteration (n = 45), median survival 10 months; and nonthoracic tumor with BRD3- or NSD3-NUT alteration (n = 12), median survival 36.5 months.

More cases of NUT midline carcinoma are being diagnosed, presumably due to increased awareness of this entity and better diagnostic capabilities. The commercially available C52 monoclonal antibody can be used for IHC testing of poorly differentiated carcinomas without glandular differentiation that arise in the chest or head and neck [53].

SUMMARY AND RECOMMENDATIONS

●Pathologic evaluation – Light microscopy generally cannot establish a diagnosis in patients with poorly differentiated cancers of unknown primary site (CUPs). Specialized studies, including immunohistochemistry, molecular cancer classifier assays, and in selected cases, electron microscopy and/or chromosomal analysis, can usually distinguish between a diagnosis of carcinoma, germ cell tumor, lymphoma, melanoma, or sarcoma. (See 'Pathologic evaluation' above.)

●Treatment and prognosis – When specialized pathologic studies are consistent with a diagnosis of poorly differentiated lymphoma, extragonadal germ cell tumor, melanoma, or sarcoma, treatment should be based upon that diagnostic category. (See 'Treatment and prognosis' above.)

●Management of patients in treatable subgroups – Several clinical subsets requiring specific treatment have been defined for patients with adenocarcinoma of unknown primary site. Patients with poorly differentiated carcinoma who fit into one of these clinical subsets should receive treatment as described for adenocarcinoma. (See "Adenocarcinoma of unknown primary site", section on 'Specific patient subgroups'.)

●Management of patients not included in specific subgroups – For other patients with poorly differentiated carcinoma who do not fit into any treatable subgroup, the information derived from immunohistochemistry and molecular cancer classifier assay often suggests the site of origin. Optimal therapy for these patients is controversial, and the treatment approach is similar to patients with adenocarcinoma of unknown primary. (See "Adenocarcinoma of unknown primary site", section on 'Approach to patients not included in specific subgroups'.)

•For patients predicted to have treatment-sensitive tumor types, we favor site-specific therapy. (See "Adenocarcinoma of unknown primary site", section on 'Cancers with predicted treatment sensitivity'.)

•For patients with treatment-resistant tumor types, empiric chemotherapy with a platinum-based regimen or site-specific treatment produces similar results; either approach is acceptable. (See "Adenocarcinoma of unknown primary site", section on 'Cancers with predicted treatment resistance or no primary site identified'.)

•Comprehensive molecular profiling of biopsy specimens or liquid biopsies may identify important therapeutic targets for some patients. (See "Adenocarcinoma of unknown primary site", section on 'Comprehensive molecular profiling'.)