INTRODUCTION Cancer of unknown primary site (CUP) is a common clinical entity, accounting for 2 percent of all cancer diagnoses in the Surveillance, Epidemiology, and End Results (SEER) registries between 1973 and 1987. Within this category, tumors from many primary sites with varying biologies are represented; this heterogeneity has made the design of therapeutic studies difficult.
Substantial improvements have been made in the management and treatment of some patients with CUP. The identification of specific subgroups of treatable patients has been made possible by the development of specialized immunohistologic techniques that can aid in tumor characterization and by the recognition of several clinical syndromes that permit prediction of chemotherapy responsiveness.
The typical patient with CUP presents with symptoms referable to a metastatic site. The initial work-up, including physical examination, laboratory and radiographic study, fails to identify the primary site. Light microscopic evaluation of biopsy material places the tumor into one of five histologic categories, which then guides further evaluation:
The diagnosis and management of patients with adenocarcinoma of unknown primary (AUP) are reviewed here. The diagnosis and management of the other subgroups are discussed separately. (See specific topic reviews).
CLINICAL COURSE The incidence of AUP increases with age. The clinical presentation is determined by the sites of tumor involvement, which are frequently multiple and often include the liver, lungs, lymph nodes, and bones.
The primary site becomes obvious in only 15 to 20 percent of patients during the course of their disease . In autopsy series, the most common identifiable primary sites are the pancreas, hepatobiliary tree, and lung, together accounting for approximately 40 to 50 percent of cases of AUP . Other gastrointestinal sites are also common, and adenocarcinomas from a wide variety of other primary sites account for the remaining cases. Breast and prostate cancers are relatively infrequent as a cause of AUP, despite being the most common cancer types in women and men, respectively .
Even at postmortem examination, 20 to 30 percent of patients with AUP have no detectable primary site. Unfortunately, large autopsy series in patients who have been evaluated with modern radiologic techniques such as computerized tomography (CT) are not available; thus, published autopsy data may not accurately reflect the current patient population with AUP.
Many patients with AUP have widespread metastases and poor performance status at the time of diagnosis. The outlook for most of these patients is poor, with a median survival of four to six months. However, small subsets of patients with a more favorable outlook are contained within this large group. Initial evaluation should be geared toward identification of these subgroups. In addition, empiric chemotherapy incorporating newer agents has produced higher response rates and probably improved survival in patients with good performance status
PATHOLOGIC DIAGNOSIS The diagnosis of adenocarcinoma is usually based on the identification of glandular structures that are formed by the neoplastic cells. Since these features are shared by all adenocarcinomas, the site of the primary adenocarcinoma cannot usually be ascertained by light microscopic examination. Although certain morphologic features can be associated with a particular tumor type (eg, papillary features with ovarian cancer, and signet ring cells with gastric cancer), they may not be sufficiently specific to be considered definitive evidence of the primary site.
Poorly differentiated adenocarcinoma The diagnosis of poorly differentiated adenocarcinoma is usually made when only minimal glandular formation is seen on histologic examination, or in tumors that lack glandular differentiation but stain positively for mucin. It is clear that adenocarcinoma, poorly differentiated adenocarcinoma, and poorly differentiated carcinoma are histologic diagnoses that represent a spectrum of tumor differentiation rather than specific well-demarcated entities. Different pathologists may use slightly different criteria for each of these three diagnoses.
The light microscopic diagnosis of poorly differentiated adenocarcinoma should be interpreted with caution, since a subset of these patients may be distinctive in their tumor biology and responsiveness to systemic chemotherapy. For this reason, evaluation and treatment of patients with poorly differentiated AUP should follow the guidelines outlined for poorly differentiated carcinoma of unknown primary, including the use of immunohistochemical staining and electron microscopy in an attempt to identify a potentially chemotherapy-responsive cancer. (See "Poorly differentiated carcinoma of unknown primary site").
Immunohistochemistry In a few cases, individual immunohistochemical stains may be specific enough to identify the site of origin for some adenocarcinomas
In most other cases, individual immunohistochemical stains are not specific enough for one particular site of origin, but may suggest the most likely primary site. As an example, immunohistochemical staining for the estrogen (ER) or progesterone receptor (PR) suggests metastatic breast cancer in women with AUP. Some breast cancers that do not stain for estrogen and/or progesterone receptors may express androgen receptors that are also detectable by immunohistochemistry . Another example is thyroid transcription factor-1 (TTF-1), a highly specific marker of peripheral bronchogenic adenocarcinomas .
In other patients with AUP, multiple immunohistochemical stains, or batteries of stains, have been used to suggest the most likely primary site. These have included cytokeratins (CK) 20 and 7, gross cystic disease fluid protein 15, breast cancer antigen (BCA) 225, carcinoembryonic antigen (CEA), vimentin, CA 19-9. and CA-125 . These stains, although suggestive, may not be specific enough to dictate treatment.
The pattern of CK 20 and 7 may be particularly helpful in suggesting a primary site. CK 20 is a low molecular weight cytokeratin that is normally expressed in the gastrointestinal epithelium, urothelium, and in Merkel cells . CK 7 is expressed by tumors of the lung, ovary, endometrium, and breast, and not in the lower gastrointestinal tract. Thus, the CK phenotype 7-/20+ strongly favors colorectal primary tumors. This was illustrated in a study of 93 autopsy cases of adenocarcinoma involving the liver, in which the CK20+/CK7+ pattern was suggestive of primary localization in the pancreas or biliary tract (11 of 14 cases; 79 percent), while a CK20+/CK7- pattern was characteristic of a colorectal primary (17 of 21 cases; 81 percent). The CK20- group was too heterogeneous to be classified adequately by these two antibodies.
In another second series of 128 metastatic nonmucinous adenocarcinomas in which the primary site was known, the most informative markers were CEA, CA 19-9, CA 125, and BCA225  . With this four-marker panel, the most predictive multiple-marker phenotypes were CEA+, BCA225-, and CA 125- for colon tumors; BCA225+, CEA-, and CA 125- for breast tumors; BCA225+, CEA+, and CA 19-9- for lung tumors; CA 125+ and CEA- for ovarian tumors; and CEA+, CA 19-9+, and CA 125+ for upper gastrointestinal tract tumors. Overall, these phenotype combinations correctly predicted the known primary site in 66 percent of cases.
GENERAL PRINCIPLES OF DIAGNOSTIC TESTING As noted above, the detection of a primary site is unusual in a patient presenting with AUP . For this reason, initial studies are designed to determine the extent of metastatic disease. Routine initial evaluation should include a thorough history and physical examination (including a pelvic examination in women, and a prostate examination in men), standard laboratory screening tests with complete blood count, multichannel biochemistry profile, urinalysis, stool examination for occult blood, and chest radiography  . All men should have a serum PSA determination, and women should undergo mammography if the diagnosis of breast cancer is a possibility. Breast MRI should be strongly considered ion the setting of a negative mammogram in a women with adenocarcinoma involving the axillary lymph nodes
Abdominal CT scan can identify a primary site in 10 to 35 percent of patients and it may identify additional metastatic sites . Positron emission tomography (PET) is also a useful procedure, and results in the identification of a primary site in 20 to 30 percent of patients.
Exhaustive radiographic and endoscopic testing should not be performed since these studies rarely detect the primary site in the asymptomatic patient, and often result in confusing or false-positive information. Instead, the presence of specific signs or symptoms should guide the choice of additional studies.
Commonly measured serum tumor markers (CEA, CA 19-9, CA 15-3, CA 125, human chorionic gonadotropin, and alpha fetoprotein) are not useful as diagnostic or prognostic tests. However, they are commonly elevated in patients with AUP and may be useful in following the response to therapy .
DIAGNOSTIC AND TREATMENT APPROACH FOR SUBSETS OF DISEASE The group of patients with AUP contains several clinically defined subgroups for which specific therapy is available. Patients who do not fit into one of these subgroups should be considered for a trial of empiric chemotherapy
Women with peritoneal carcinomatosis In women, adenocarcinoma causing diffuse peritoneal involvement usually originates in the ovary or in extraovarian tissues with similar histogenesis (see below). Peritoneal carcinomatosis can also occur in women with normal ovaries and no other evident intraabdominal primary site. In one study of 118 women with malignant ascites, the sites of origin were :
Papillary serous carcinoma of the peritoneum Some of these tumors arise from the peritoneal surface, which shares a common histogenesis with ovarian tissues. Many have morphologic features that are typical for epithelial ovarian carcinoma, such as papillary configuration or psammoma bodies. In such cases, this syndrome has been termed papillary serous carcinoma of the peritoneum (PSCP) or multifocal extraovarian serous carcinoma.
Some patients may present with a poorly differentiated adenocarcinoma that does not exhibit a papillary configuration (analogous to poorly differentiated epithelial ovarian carcinomas); they should be approached similarly to those with typical papillary serous histology.
Tumors arising within the mesothelial cells may share a common biology with ovarian carcinoma, a concept that is supported by the following observations:
Treatment Patients with PSCP often respond well to chemotherapy regimens that are effective in the treatment of advanced epithelial ovarian cancer. Several studies have documented high initial response rates, with long-term remissions in 15 to 20 percent of cases . In most early reports, patients were treated with platinum/cyclophosphamide. However, taxane/platinum regimens have proven superior for advanced ovarian cancer, and should be the treatment of choice for these patients.
Surgical cytoreduction should be considered in patients with bulky disease. In patients with epithelial ovarian cancer, debulking may provide the best chance for long-term remission, although the optimal timing is controversial .
Women with axillary lymph node metastases Breast cancer should be suspected in women who have AUP and axillary lymphadenopathy. Measurement ER, PR, and Her2 overexpression must be performed on the initial lymph node biopsy in such patients. Mammography may provide useful information, but is often misleading. Negative mammograms have been found in patients with breast cancer, and abnormal mammograms have been found in patients without breast cancer
Mammographically occult primary breast cancers may be identifiable with breast magnetic resonance imaging (MRI) . Bilateral breast MRI is now considered a standard approach to evaluation of the breasts in such patients. If a focal lesion is identified, further diagnostic evaluation should follow standard guidelines for suspected breast cancer.
Women who have no evident primary breast lesion and whose metastases are isolated to axillary lymph nodes after completion of routine staging evaluation are potentially curable, and should be managed according to standard guidelines for stage II breast cancer. Primary therapy should include either modified radical mastectomy or axillary lymph node dissection followed by radiation therapy to the breast . An occult breast cancer will be identified in 33 to 82 percent of patients when mastectomy is performed even when physical examination, mammograms, and MRI are normal . The primary tumor is usually less than 2 cm in diameter; in occasional patients, only carcinoma in situ is identified .
Observation of the breast without definitive local therapy is not recommended since a high percentage of these patients will develop a clinically manifest breast tumor. Selection of adjuvant therapy should be based on hormone receptor status, Her-2 status, and pertinent clinical features, following standard guidelines for patients with node-positive breast cancer.
Women with metastatic sites in addition to axillary lymph nodes may also have metastatic breast cancer. These women should receive a trial of systemic therapy using the guidelines for the treatment of metastatic breast cancer. Patients with positive hormone receptors may derive major palliative benefit from hormonal therapy. Patients whose tumors overexpress Her-2 by immunohistochemistry (ie, 3+) or by fluorescence in situ hybridization (FISH) should be treated with trastuzumab in combination with other cytotoxic agents.
Men with skeletal metastases When bone metastases are the first manifestation of metastatic adenocarcinoma, the most common primary tumor sites are the lung, prostate, and, less often, liver, kidney, thyroid, and colon . Metastatic prostate cancer should be suspected in men with AUP predominantly involving bone, particularly if the metastases are blastic or sclerotic. Elevated serum levels of PSA or tumor staining with PSA provides confirmatory evidence of prostate cancer. Occasional patients have a significantly elevated serum PSA or tumor staining for PSA, but a clinical presentation that is atypical for prostate cancer (eg, metastases to the lung or mediastinal or upper abdominal lymph nodes, without concomitant involvement of bone or pelvic lymph nodes) . All of these patients should be treated according to guidelines for advanced prostate cancer.
Adenocarcinoma presenting as a single metastatic focus In occasional patients, only a single metastatic lesion is identified after a complete staging evaluation. Such single lesions have been described in a variety of sites including lymph nodes, brain, lung, adrenal gland, liver, and bone. The possibility of an unusual primary site (eg, apocrine, eccrine, or sebaceous carcinoma) mimicking a metastatic lesion should be considered, but can usually be excluded on the basis of clinical or pathologic features.
In most of these patients, other metastatic sites become evident within a relatively short time. A PET scan should be considered to rule out additional unrecognized sites of metastatic disease prior to definitive local therapy .
If no evidence of additional disease can be found, resection of the solitary lesion should be considered. If resection is not feasible based on the location of the metastatic lesion, definitive local radiation therapy should be administered. Local treatment sometimes results in long disease-free intervals. In some instances (eg, after resection of a solitary brain metastasis), local radiation therapy may also be appropriate to maximize the chance of local control
The role of adjuvant chemotherapy in this setting is undefined. Patients with poorly differentiated carcinoma may benefit from platinum-based therapy.
Empiric chemotherapy Most patients with AUP do not fit into any of the clinical subgroups outlined above. In such cases, empiric chemotherapy may be considered. In the past, 5-fluorouracil- and doxorubicin-based regimens, initially developed for the treatment of gastrointestinal primary tumors, produced relatively low response rates and few complete responses . Various cisplatin-based regimens have also been evaluated, but were more toxic and did not show greater efficacy than non-cisplatin-containing combinations .
The introduction of newer antineoplastic agents with broad clinical activity has changed the standard treatment of several common epithelial cancers. These drugs include the taxanes, topoisomerase I inhibitors, gemcitabine, and vinorelbine. Although none of these agents has been completely evaluated for the treatment of AUP, results with various taxane and platinum containing regimens sugest improved response rates and survival when compared retrospectively with earlier regimens . Combinations containing other new agents (gemcitabine/cisplatin, irinotecan/cisplatin, docetaxel/gemcitabine, gemcitabine/carboplatin) have produced similar results . Although response rates have varied from 20 to 55 percent, median survivals with several regimens have been similar at 8 to 11 months. When reported, two year survivals are in the 20 to 25 percent range. Addition of a third drug to a taxane and platinum regimen does not appear to improve efficacy . Unfortunately, the absence of comparative trials and the marked heterogeneity of this patient population prevents confident determination of the most efficacious regimen.
Limited experience now exists with second line therapy. Single agent gemcitabine (1000 mg/m2 weekly three of four weeks) has modest activity . The combination of gemcitabine and irinotecan produced a 10 percent overall response rate, with an additional 18 percent of patients stable more than six months, in a group of 31 patients previously treated with taxane/platinum combinations .
Antiangiogenic therapy Antiangiogenic therapy may also be active in patients with metastatic adenocarcinoma or poorly differentiated carcinoma of unknown primary. In a phase II trial, 49 evaluable patients, 75 percent of whom had received prior chemotherapy, were treated with the combination of bevacizumab (10 mg/kg IV every two weeks) plus erlotinib (150 mg/day) . Objective responses were seen in 4 patients (8 percent) and another 30 patients (59 percent) had stable disease. Median progression-free and overall survival were 6.2 and 8.9 months respectively.
Prognostic factors Retrospective analyses have identified clinical and pathologic features associated with a favorable response to treatment with empiric chemotherapy . These include:
In one prognostic factor analysis that included 150 patients with CUP who were seen at a single institution over a ten year period, the performance status and serum LDH could be used to separate patients into good and poor risk categories. The median survival durations were 11.7 and 3.9 months, and one year survival rates were 45 and 11 percent, for good and poor risk patients, respectively
A multivariate analysis of prognostic factors based upon a series of 317 consecutive patients found that a normal serum albumin and the absence of liver metastases identified a favorable subset of patients (median survival 371 versus 103 days, in patients with a low serum albumin and/or liver metastases) . In the same report, the favorable prognosis associated with the combination of a normal serum albumin and the absence of liver metastases was validated in a second patient cohort of 124 patients with CUP.
Recommendation All patients with good performance status should be considered for a trial of empiric chemotherapy. Several regimens containing newer drugs appear to have more antitumor activity than older regimens. However, the best regimen has not been defined in prospective randomized trials. At present, the combination of paclitaxel and carboplatin is a reasonable choice for first-line therapy, based on the relatively large experience with this combination in AUP. The value of adding a third agent (either etoposide or gemcitabine) is unclear, based upon data from existing phase II trials. Patients with poor performance status are much less likely to benefit from chemotherapy, and optimal management may include supportive measures only.