Carcinoma of an Unknown Primary Site

November 1, 2015

By John D. Hainsworth, MD
Lawrence M. Weiss, MD

Article

Carcinoma of an unknown primary site is a common clinical syndrome, accounting for approximately 3% of all oncologic diagnoses. Patients in this group are heterogeneous, having a wide variety of clinical presentations and pathologic findings.

Overview

Carcinoma of an unknown primary site is an uncommon clinical syndrome, accounting for approximately 3% of all oncologic diagnoses. Patients in this group are heterogeneous; they have a wide variety of clinical presentations and pathologic findings. A patient should be considered to have carcinoma of an unknown primary site when metastatic cancer is detected at one or more sites and clinical evaluation fails to define a primary tumor site or pattern.

Although all patients with cancer of an unknown primary site have advanced, metastatic disease, universal pessimism and nihilism regarding treatment are inappropriate. Subsets of patients with specific treatment implications can be defined using clinical and pathologic features. Improved diagnostic methods, including more specific immunohistochemistry (IHC) stains and gene expression profiling, allow accurate prediction of the tissue of origin in most patients with carcinoma of an unknown primary site, so that site-specific treatment can be delivered.

Epidemiology

Gender

Unknown primary cancer occurs with approximately equal frequency in men and women, and the prognosis is the same for both.

Age

As with most epithelial cancers, the incidence of unknown primary cancer increases with advancing age, although a wide age range exists. Some evidence suggests that younger patients are more likely to have poorly differentiated histologies.

Disease Sites

At autopsy, a primary site is identified in 70% to 80% of patients. Above the diaphragm, the lungs are the most common primary disease site, whereas various gastrointestinal (GI) sites of disease (pancreas, colon, stomach, liver) are most common below the diaphragm. Recently, gene expression profiling has also predicted most primary sites to be in the GI tract or lungs, but breast and ovarian cancers have been more commonly seen with this assessment method than in historical autopsy series.

Signs and Symptoms

Sites of Metastatic Involvement

Common sites of metastatic involvement include the lungs, liver, and skeletal system; however, a wide variety of other sites are sometimes involved.

Symptoms and Physical Findings

Patients with unknown primary cancer usually present with signs and symptoms related to the areas of metastatic tumor involvement. In addition, constitutional symptoms, such as anorexia, weight loss, weakness, and fatigue, are common.

Pathologic Evaluation

Optimal pathologic evaluation is critical in the evaluation of patients with carcinoma of an unknown primary site and can aid with the following:

• Distinguishing carcinoma from other cancer types

• Determining histologic type

• Identifying the tissue of origin

• Identifying specific characteristics that may direct specific treatments

Initial Approach

Optimal pathologic evaluation includes specialized studies in all patients with carcinoma of unknown primary site; therefore, a surgical or core needle biopsy is necessary to ensure that sufficient diagnostic material is obtained. Tissue is required for paraffin-section IHC, which can reliably distinguish carcinoma from other neoplasms and can often suggest a specific primary site when interpreted in conjunction with clinical features. Gene expression profiling assays are also valuable in predicting the tissue of origin and should be obtained when IHC stains cannot predict a specific primary site. Electron microscopy, which optimally requires glutaraldehyde fixation, is usually no longer recommended.

Carcinoma vs Other Neoplasms

It is important to rule out the possibility of lymphoma, melanoma, and sarcoma. A battery of antibodies is used in an attempt to distinguish carcinoma from other types of neoplasms, as summarized in Table 1. The staining result obtained with any single marker is unreliable, because exceptions may occur for each antibody. For example, although keratin is a relatively reliable marker of carcinoma, some carcinomas (eg, adrenal cortical carcinoma or undifferentiated carcinoma of the thyroid) may be keratin-negative, whereas some types of sarcoma are characteristically keratin-positive (eg, epithelioid sarcoma).

TABLE 1: Immunohistochemical studies useful in the differential diagnosis of carcinoma vs another neoplasm

Determination of Histologic Type

There may be clues on initial histologic examination. For example, the presence of gland formation or mucin production would indicate an adenocarcinoma, whereas the presence of keratinization would indicate a squamous cell carcinoma. Evidence of neuroendocrine differentiation may be suggested by the presence of a characteristic, relatively fine chromatin pattern. IHC can also be of use, because expression of keratin subtypes 7 and 20 would favor adenocarcinoma, and expression of p40 or 63 or keratin subtypes 5/6 and 14 would favor squamous cell carcinoma. Reliable neuroendocrine markers include chromogranin A and synaptophysin. In situ hybridization studies may be helpful in some circumstances, such as for the identification of Epstein-Barr virus RNA in a cervical lymph node biopsy, which suggests a nasopharyngeal primary site, or the identification of human papillomavirus in an inguinal lymph node biopsy, which suggests a uterine cervical primary site.

Determination of Primary Site in Metastatic Adenocarcinoma

IHC staining frequently suggests the primary site in patients with adenocarcinoma of unknown primary site. Panels of IHC stains usually provide more information than single stains. Useful stains are listed in Table 2; however, most of these stains must be interpreted in conjunction with tumor histology and clinical features.

Identification of Specific Treatment Target Characteristics

Even if the primary site is not determined, characteristics of the carcinoma may suggest specific treatment options or impart prognostic information. Examples of the former may include determination of estrogen or progesterone receptors or expression of members of the epidermal growth factor receptor family (eg, human epidermal growth factor receptor type 2 [HER2/neu]). Examples of the latter may include Ki-67, which is a surrogate marker of the proliferation rate of a neoplasm.

TABLE 2: Most useful organ-specific markers

Gene Expression Tumor Profiling

Specific gene expression profiles based on the tissue of origin have been identified for many tumor types. Several assays using either quantitative reverse transcription polymerase chain reaction (qRT-PCR) or gene microarray techniques are now available and can be performed on tumor tissue from formalin-fixed, paraffin-embedded biopsy specimens using very few tissue sections. In tumors of a known primary, these assays can correctly identify the tissue of origin in over 85% of metastases. In a group of 24 patients with unknown primary cancer who had an anatomic primary site clinically identified 2 to 78 months after their initial diagnosis, molecular profiling performed retrospectively on the initial biopsy specimen correctly identified the primary site in 18 (75%) patients. A direct comparison between IHC and gene expression profiling studies in a series of blinded poorly differentiated neoplasms demonstrated the superiority of gene expression profiling (79% vs 69% correct predictions of the tissue of origin). Gene expression profiling can therefore accurately identify the tissue of origin in the majority of patients with carcinoma of unknown primary site, and it should be a part of the diagnostic evaluation if the site of origin is not identified by other studies.

Clinical Evaluation

After a biopsy has established the diagnosis of metastatic carcinoma, a relatively limited clinical evaluation is indicated to search for a primary site. Recommended evaluation includes a complete history; physical examination; chemistry profile; complete blood cell count; computed tomography scans of the chest, abdomen, and pelvis; and specific radiologic and/or endoscopic evaluation of symptomatic areas. If these studies do not identify a primary site, then the patient should be considered to have carcinoma of unknown primary site. However, further evaluation should be performed in patients with specific clinical presentations.

Additional Evaluation Directed by Clinical Features

Women with clinical features suggestive of metastatic breast cancer (eg, metastatic adenocarcinoma involving axillary nodes, bones, or pleura; hormone receptor or HER2/neu-positive tumor) should have mammography, ultrasonography, and breast magnetic resonance imaging. Men with adenocarcinoma of unknown primary site should have serum prostate-specific antigen (PSA) levels measured and/or tumor staining for PSA. In young men with poorly differentiated carcinoma, serum human chorionic gonadotropin (hCG) and alpha-fetoprotein (AFP) levels should always be measured.

Asymptomatic Areas

In general, radiologic or endoscopic evaluation of asymptomatic areas is not productive and should be avoided. However, a couple of exceptions should be mentioned. Colonoscopy should be performed in patients with features suggestive of metastatic colorectal cancer (ie, liver/peritoneal metastases, typical histology and IHC staining), even in the absence of colon-related symptoms. Bronchoscopy should be considered in patients with clinical presentations suggestive of lung cancer (eg, metastatic small cell neuroendocrine carcinoma; metastases involving mediastinal and/or supraclavicular lymph nodes).

Cervical Lymphadenopathy

Metastatic squamous cell carcinoma in cervical lymph nodes usually involves upper- or mid-cervical locations. All patients should undergo a thorough search for a primary site in the head and neck region, including direct endoscopic examination of the oropharynx, hypopharynx, nasopharynx, larynx, and upper esophagus. Any suspicious areas should be biopsied. Fiberoptic bronchoscopy should be considered in patients with involvement of low-cervical nodes. This type of evaluation will identify a primary site, usually in the head and neck, in 85% to 90% of these patients. Further evaluation with PET scanning identifies a primary site in 15% to 30% of the remaining patients. Low-cervical lymph node involvement with adenocarcinoma (level IV, supraclavicular) may represent an upper GI primary site.

Inguinal Lymphadenopathy

Patients with metastatic squamous cell cancer presenting in inguinal lymph nodes almost always have an identifiable primary site in the perineal area. Women should undergo careful examination of the vulva, vagina, and cervix; men should have careful inspection of the penis. Anoscopy should be performed to exclude lesions in the anorectal area.

Treatment

Table 3 summarizes the recommended treatment for various recognized clinicopathologic subsets.

Adenocarcinoma

When evaluating patients with adenocarcinoma of an unknown primary site, several clinical subsets should be identified and treated specifically.

Women with isolated axillary adenopathy

Treatment appropriate for stage II breast cancer should be administered. Mastectomy reveals an occult primary cancer in 50% to 60% of these patients, even when findings from the physical examination and mammography are normal. Axillary dissection with breast irradiation is also a reasonable treatment, although there are no definitive comparisons of this approach vs mastectomy. Adjuvant systemic therapy, following standard guidelines for stage II breast cancer, is also indicated.

TABLE 3: Recommended treatment for recognized clinicopathologic subsets

Women with peritoneal carcinomatosis

Often, the histopathology in these patients suggests ovarian cancer (ie, serous cystadenocarcinoma or papillary adenocarcinoma). However, all women with this syndrome should be treated as if they had stage III ovarian cancer. Initial cytoreductive surgery should be followed by chemotherapy with a taxane/platinum combination, as recommended for advanced ovarian cancer. In these patients, serum CA-125 can often be used as a tumor marker.

Men with bone metastasis

Metastatic prostate cancer should be suspected and usually can be diagnosed in patients with either an elevated serum PSA level or positive tumor staining for PSA. In such patients, treatment should follow guidelines for metastatic prostate cancer; androgen deprivation therapy is often of palliative benefit.

Patients with a colon cancer “profile”

Patients with a metastatic pattern (liver and/or peritoneal involvement) and pathologic findings (adenocarcinoma, CK20-positive/CK7-negative and/or CDX2-positive) strongly suggestive of metastatic colorectal cancer should be treated with modern regimens for advanced colon cancer. The prediction of a colorectal tissue of origin by molecular gene profiling can also be used to guide therapy.

Patients with a single metastatic site

Surgical resection or radiation therapy should be administered to patients who present with clinical evidence of a single metastasis. Before proceeding with local therapy in these patients, positron emission tomography (PET) scanning should be considered to rule out additional metastatic sites. Some of these patients have prolonged survival after local therapy, particularly those who present with a sole metastasis in an isolated peripheral lymph node group. The role of “adjuvant” systemic therapy is undetermined in these patients.

Squamous Cell Carcinoma

Squamous cell cancer accounts for only 10% of light microscopic diagnoses in patients with unknown primary cancer. Isolated cervical adenopathy is the most common presentation for squamous cell carcinoma of an unknown primary site; other patients have isolated inguinal adenopathy at presentation. Specific management is essential for both of these subgroups, since both have the potential for long-term survival following treatment.

Patients with cervical lymphadenopathy

Patients with cervical lymphadenopathy in whom no primary site is identified should be treated as if they had a primary site in the head and neck. Concurrent treatment with chemotherapy and radiation therapy is the treatment of choice for locally advanced head and neck cancer. Radiation therapy doses and techniques should be identical to those used in treating patients with known head and neck primary sites. In addition to the involved neck, the nasopharynx, oropharynx, and hypopharynx should be included in the radiation field. Radical neck dissection should be considered in patients who have any evidence of residual cancer following combined-modality therapy.

Five-year survival rates with combined-modality therapy are 60% to 70% and appear superior to results with local modalities alone (30% to 50%). The extent of cervical lymph node involvement is the most important prognostic factor. (For additional discussion, see “Unknown Head and Neck Primary Site” in the “Head and Neck Tumors” chapter.)

Patients with inguinal lymphadenopathy

Identification of a primary site in the perineal area is important in patients with inguinal lymphadenopathy, because curative therapy is available for some patients, even after metastasis to inguinal lymph nodes. In the uncommon patient for whom no primary site is identified, inguinal node dissection, with or without radiation therapy, can result in long-term survival. Although limited data exist on this uncommon subgroup, the demonstrated superiority of combined-modality therapy vs local treatment alone for primary squamous cell cancers in the perineal area (eg, cervix, anus) has led to a suggestion that the addition of platinum-based chemotherapy may improve treatment results.

Poorly Differentiated Carcinoma

This heterogeneous group includes a minority of patients with highly responsive neoplasms and therefore requires special attention in initial clinical and pathologic evaluations. All patients with poorly differentiated carcinoma of unknown primary site require evaluation with specialized pathologic studies, including IHC and gene expression profiling. Several treatable subsets can be recognized.

Men with extragonadal germ cell cancer syndrome

Young men with a predominant tumor location in the mediastinum and retroperitoneum and/or high levels of serum hCG or AFP should be treated as if they had a poor-prognosis germ cell tumor (ie, four courses of chemotherapy with cisplatin/etoposide/bleomycin, followed by surgical resection of residual radiographic abnormalities).

Molecular genetic analysis can identify an i(12p) chromosomal abnormality diagnostic of a germ cell tumor in some of these patients, even when the diagnosis cannot be made by any other pathologic evaluation. Patients with germ cell tumors diagnosed in this manner have been shown to be as responsive to treatment as patients with extragonadal germ cell tumors of typical histology.

Patients with poorly differentiated neuroendocrine carcinoma

With the improved immunoperoxidase stains now available, neuroendocrine features are recognized more frequently in patients with poorly differentiated carcinoma. These tumors are distinct in biology and therapeutic implications from well-differentiated neuroendocrine tumors (eg, carcinoid tumors, islet cell tumors) of an unknown primary site, which almost always present with multiple liver metastases. In contrast to typical carcinoid tumors, poorly differentiated neuroendocrine tumors are difficult to recognize by light microscopic examination alone, although some of the latter tumors have neuroendocrine or “small-cell” features.

Patients with poorly differentiated neuroendocrine carcinoma of an unknown primary site should receive a trial of chemotherapy as recommended for the treatment of small-cell lung cancer. In a group of 51 such patients, the complete response rate was 28% following treatment with cisplatin/etoposide-based regimens; the overall response rate was 71%. Eight patients (16%) had durable complete remissions.

The combination of paclitaxel, carboplatin, and etoposide is also highly active in the treatment of poorly differentiated neuroendocrine tumors (response rate, 53%; 2-year survival, 33%). However, the addition of paclitaxel increases myelosuppression and is not clearly superior to platinum/etoposide therapy.

Although the identity of most poorly differentiated neuroendocrine tumors remains unknown, this group of chemotherapy-responsive patients can be reliably identified using specialized, but widely available, pathologic evaluation.

Other patients with poorly differentiated carcinoma

A small percentage of patients with poorly differentiated carcinoma of unknown primary site have highly responsive tumors and derive major benefit from combination chemotherapy. In a group of 220 such patients treated at a single institution with cisplatin-based regimens effective for germ cell tumor, the overall response rate was 64%, with 27% complete responses. Median survival in this group was 20 months, and 13% of patients remained disease-free for more than 8 years. The young median age of 39 years indicates that this was a select patient group; improved diagnostic methods now available allow these chemotherapy-sensitive patients to be identified and included in one of the previously described treatable subsets, or to have a specific treatable tumor recognized. The remaining patients with poorly differentiated carcinoma have a prognosis similar to that of the large majority of patients in the adenocarcinoma group, and they should be managed according to the same guidelines.

Site-Specific Therapy Based on the Predicted Tissue of Origin

Approximately 80% of patients with cancer of unknown primary site do not fit into any of the favorable treatment subsets previously discussed. For many years, the standard treatment for this group of patients has been a trial of empiric chemotherapy. Although some patients derive substantial benefit, the results of treatment for most patients in this group are poor, and the median survival with empiric chemotherapy is only 8 to 11 months.

In recent years, the treatment for many types of cancer has improved and has also become more site-specific, so that adequate “broad-spectrum” coverage with an empiric chemotherapy regimen is not feasible. Improved diagnostic techniques, including panels of IHC stains and gene expression profiling, enable accurate prediction of the tissue of origin in the majority of patients with carcinoma of unknown primary site. Increasing clinical data support the use of these predictions to direct treatment for these patients.

Three studies support the efficacy of treatment for metastatic colorectal cancer in patients with carcinoma of unknown primary site predicted to have a colorectal site of origin. In one of these studies, a colorectal site of origin was predicted by IHC staining (CDX2-positive with or without CK20-positive/CK7-negative staining), while in two studies, predictions were made by gene expression profiling. Patients in all three studies underwent colonoscopy, without identification of an anatomic primary site. With treatment for metastatic colorectal cancer, patients in these studies had median survivals of 28, 21, and 27 months, respectively, similar to the expected survival of patients with known metastatic colorectal cancer.

One prospective study has been performed, in which patients with newly diagnosed carcinoma of unknown primary site had gene expression profiling; when a site of origin was predicted, patients received site-specific therapy. Gene expression profiling predicted a primary site in 98% of patients; in 194 patients who received site-specific therapy, the median survival was 12.5 months. In patients predicted to have treatment-responsive tumor types, median survival was 13.4 months (vs a median survival of 7.6 months in patients predicted to have less responsive tumors). The median survival of patients with individual tumor types mirrored survival of similar tumors of known primary site (eg, pancreas, 8 months; ovary and breast, > 30 months).

Although clinical data are still emerging, all studies to date support the use of assay-directed, site-specific therapy as a new treatment paradigm. In some cases, tissue-of-origin predictions can lead to other molecular testing (eg, assays for epidermal growth factor receptor–activating mutation and ALK rearrangement in patients predicted to have non–small-cell lung cancer), which provide further treatment options for individual patients. A trial of empiric chemotherapy is still reasonable for patients with carcinoma of unknown primary site in whom new diagnostic techniques fail to predict a site of origin, but widespread use of this approach is no longer recommended.