In general, hematoxylin and eosin staining can be used to determine the diagnosis, but the role of IHC has been increasing in recent years. Most guidelines consider at least a limited IHC study to be necessary for the diagnosis of CUP to exclude a non-epithelial origin of the lesion and to determine the origin of a carcinoma. IHC can accurately predict the origin of ≈35–40% metastatic cancers [9]. We cannot examine all the suggested IHC stains, but screening panels including cytokeratins (CK-7, CK-20), thyroid transcription factor-1 (TTF-1) and breast/ovarian markers are taken into consideration in most guidelines. Some authors have identified a new favorable subset of patients called “adenocarcinoma with colon cancer profile” that have a specific pattern of keratins and tumor-specific antigens (CK20+, CK7., CDX2+). This subset responds to colon cancer-specific combination chemotherapy regimens with a median survival of 20–24 months [10]. Even if experiences on the surgical treatment of this subset have not been published, a role for hepatectomy in selected patients showing this IHC pattern can be foreseen.

Different organ tissues show different patterns of gene expression, with some genes showing greater expression in one organ than another. The specific pattern of gene expression constitutes a “signature” of that tissue. Neoplasms share some of the patterns of the original tissue. Cell type-specific differential gene expression profiles have been used to develop assays that can provide additional information for the classification of tumors, including unknown primary cancers. The accuracy of the classification of these assays is >80% [11]. Multiple gene signatures for the prediction of primary sites have been generated, and there are several commercial tests available using this technology. All of them can be used to identify accurately types and subtypes of cancer on specimens with extremely low numbers of neoplastic cells. The ability to detect tumor signatures on particularly small specimens is extremely relevant for reducing the invasiveness of the diagnostic phase. A molecular profile consistent with six primary sites (lung, breast, colon, ovary, pancreas, and prostate) has been found, evaluating the expression of a 10-gene panel [12], and 10 cancer types could be distinguished by a second panel using 495 predictive genes [13]. Even though validation of this method to predict the origin of a CUP has not been established, its potential role in typing an otherwise “orphan” tumor (and even in determining specific drug response signatures) is enormous.

The widespread availability of complex (but highly effective) technologies such as computed tomography (CT) and magnetic resonance imaging (MRI) means that these imaging modalities are included in most diagnostic protocols for CUPs. The same definition of “putative CUP” cannot be obtained from a chest and abdominal multislice spiral CT, and MRI is often used in the diagnosis of occult tumors of the breast, head or neck.

The need to localize exceptionally small tumors, however, has prompted considerable interest in 18-F-fluorodeoxyglucose-positron emission tomography (18-FDG-PET), especially if combined with CT (PET-CT). CUPs are supposedly too small to be detected in the initial investigation phase, which generally includes CT. The ability of PET-CT to provide functional and metabolic information, along with the imaging, makes this technology particularly useful [14]. A recent meta-analysis demonstrated the excellent sensitivity and specificity of PET-CT in the localization of the primary tumor of an occult metastasis [15]. However, FDG remains a rather non-specific agent that can also accumulate in non-neoplastic tissues. Several other radiotracers are being evaluated for use in PET-CT to enhance its specificity [16,17]. Clinical use of tracers aimed at specific molecular tumor targets and the use of PET-MRI could induce substantial changes to the approach to these tumors.

It could be argued that the role of surgery is extremely limited in a disease that, by definition, is already diffuse at diagnosis and whose biology is probably more aggressive for its early metastatic potential. However, such heterogeneous conditions can be divided into subsets with relatively homogeneous clinicopathological features. Most patients have aggressive disseminated disease that is relatively resistant to chemotherapy and with little curative potential, but 15–20% of subjects belong to subsets with a more favorable prognosis for whom a better outcome can be achieved [18].

Examining this topic under an evidence-based prospective is very difficult. Obtaining a series large enough to achieve a solid level of evidence for each of the significant subsets is not easy in such a heterogeneous group of conditions. Moreover, as research advances and our ability to determine the origin of the tumor increases, more lesions become associated with their putative primary origin, sometimes even if their diagnosis is likely but, in fact, uncertain. These lesions are often studied within a specific group of tumors and excluded from the CUP series (which are left only with unidentified cases). The same technological progress does not allow consideration of series collected over a long timespan because it is extremely likely that the diagnostic methods used in the most recent cases were not available for patients enrolled just few years earlier, making the group inhomogeneous. This is probably the only situation in which the passing of time does not help in accruing larger and more significant series and obtaining a better level of evidence.

Even with the limitations mentioned above, several organizations have recently issued guidelines for the diagnosis and management of CUPs [7, 19– 21] using different methodologies.

A complete and detailed overview of all the diagnostic and multidisciplinary treatments of CUPS lies outside the scope of this chapter. Complete diagnostic and therapeutic algorithms can be found in the guidelines mentioned above. There is no question, however, that identification of potentially treatable anatomoclinical entities, in which it is possible to obtain long-term disease control with appropriate multidisciplinary management, has been one of the most relevant scientific achievements in this matter. Surgery plays a pivotal role for some of these subsets. We will examine the treatment of these subsets but not discuss those with minimal surgical interest.