For this discussion, “unknown tumor” encompasses two situations, an undifferentiated tumor (assumed primary – unknown primary cancer (UPC)) that cannot be reliably diagnosed by light microscopy and a metastasis of unknown origin (MUO)

Tumors of unknown primary account for about 3–5% of all cancer patients, which places them in the ten most common malignancies

The goal is to correctly identify the tumor type and/or primary site while minimizing the number of markers (stepwise panels) and/or the testing time; usually the process is a compromise of these two systems

Tumors of unknown primary define a neoplasm that remains undetermined as to site of origin after thorough clinical, laboratory, imaging, and pathology investigation including immunohistochemistry

Molecular gene expression profiling is a promising method for assigning tissue of origin in cancers of undetermined primary site

The ultimate goal is to direct site-specific antitumor therapy whenever possible

Various studies have shown a success rate of 66–90%

The initial discussion is primarily about an undifferentiated primary tumor with the following sections concerning an MUO. However, there is extensive overlap between investigation of a UPC and an MUO

The chapter is structured as several groups of tables of differential diagnoses:

Information can guide and simplify the testing:

The usual approach is that undifferentiated tumors are first screened to identify the type of cancer – carcinoma, lymphoma, melanoma, or sarcoma. The additional “second-tier” testing is performed to identify the specific type or primary site, as clinically required

Differential diagnosis and additional test panels are based on the results of an initial screening panel, tumor primary or metastatic site, and morphology

With appropriate, useful clinicoradiographic information , the process might be more for confirmation than investigation. This can reduce the test panel to a couple of expected positive and pertinent negative (tissue site) markers. For example, adenocarcinoma in the liver with the history of colon cancer and a panel of CK7, CK20, and CA19-9 can usually identify colon cancer (CK7−, CK20+, CA19-9−) and cholangiocarcinoma (CK7+, CK20±, CA19-9+). Hepatocellular carcinoma (HCC) is usually CK7−/CK20−, but the addition of a HepPar1 negative result would help rule out HCC

Ideal tumor markers are highly sensitive (positive in all cases of a particular tumor type) and highly specific (negative in all other tumors). However, less sensitive or specific markers are useful in some situations (e.g., with a limited differential diagnosis). For example, GCDFP-15 is expressed in 35–60% of breast metastases (low sensitivity) but is negative in most other tumors (high specificity). Thus, it is effective when positive but less useful when negative. A probabilistic approach is most appropriate (with an awareness of clinical information, tumor site, morphology, and expected percent positive tumors)

Some markers are positive or negative throughout the tumor tissue (e.g., PSA), but many are typically focal (e.g., CK20, chromogranin). Whether a stain is considered positive depends on the marker, the external and internal controls, the tumor staining intensity and distribution, and the differential diagnosis. For example, a benign schwannoma stains strongly and diffusely for S-100, while a malignant peripheral nerve sheath tumor is weak and focal

Undifferentiated tumor can be a carcinoma, sarcoma, lymphoma, or melanoma. Sarcomas often require IHC confirmation or subtyping

UPC is predominantly adenocarcinoma, most commonly from the lung, pancreas/bile duct, colon/rectum, kidney, breast, stomach, ovary, liver, esophagus, and prostate. Less commonly, poorly differentiated carcinoma, neuroendocrine tumors, melanoma, lymphoma, and sarcoma (3%) are identified

Testing panels and differential diagnosis can be based on morphology including epithelioid (carcinoma, melanoma), glandular (carcinoma), spindle (sarcoma, carcinoma, melanoma), small round blue cell (lymphoma, carcinoma, sarcoma, melanoma), and undifferentiated patterns

Immunoreactivity must be assessed in the critical tumor cells while avoiding background stromal cells such as fibroblasts and vascular endothelium as well as tumor-infiltrating macrophages, mast cells, lymphocytes, and other inflammatory cells

The differential diagnosis of primary tumors is discussed in the immunohistochemistry chapter by organ system and tissue type, with further details included in chapters throughout the book

A few benign and low malignant potential tumors are included when the morphology places them in the differential diagnosis

Tumor genetics are listed in italics. In unknown tumors, molecular testing is most useful for confirming a suspected diagnosis of a sarcoma that has a known and consistent abnormality (e.g., synovial sarcoma). Some genetic results have a prognostic significance. The karyotypic complexity of many tumors limits the usefulness of genetic testing for diagnosis

Molecular testing (cDNA microarrays or RT-PCR) of UPC yields accurate results of 61–89% in various studies. Difficulties occur in test strategy (e.g., microarray design), test performance (e.g., RNA contamination), and tissue availability (e.g., formalin fixed, DNA preservation). Presently, the cost of these studies is high

RT-PCR can be performed on formalin-fixed paraffin-embedded tissue (FFPE) (e.g., c-kit and PDGFRA mutations in GIST)

Using FFPE tissue, FISH probes (e.g., EWS, FKHR, SYT, ALK, CHOP, etc.) can confirm an IHC suspected diagnosis

Refer to the introduction of the immunohistochemistry chapter for a discussion; errors may occur in fixation, tissue sampling, processing, and interpretation

Positive controls should be evaluated for every marker, preferably external and internal. The proper expected staining pattern (nuclear, cytoplasmic, or membranous) must be present

Tumors of a certain cell lineage can lose expression of a typical marker or aberrantly express a marker not usually seen in that cell type. This potentially confounding situation is minimized by performing a panel of several expected positive and negative markers

Necrosis and crush artifact can cause false-positive results

Only very rare markers are nearly 100% sensitive and specific. Markers are listed as positive when a significant majority of tumors express the antigen. Low but significant expression is indicated with a percentage