Chapter 45: Carcinoma of Unknown Primary

Carcinomas of unknown primary (CUPs), with their heterogeneous presentations, pose a challenge on the diagnostic and therapeutic fronts. Depending on the extent of evaluation, CUP comprises 3% to 5% of all tumors diagnosed (^1,2,3). A working definition for CUP is biopsy-proven metastatic cancer with no identifiable primary source by history; physical examination; chest radiography; complete blood cell count; chemistry; computed tomography (CT) of the chest, abdomen and pelvis; prostate-specific antigen (PSA) in men; and mammography in women (²). The natural history of disease for CUP is diverse and is dependent on multiple variables, such as, age, number of metastatic sites, dominant area of disease, and histology. This considerable heterogeneity presents a challenge to the systematic study of CUPs. In addition, the emergence of sophisticated imaging, robust immunohistochemistry (IHC), and genomic and proteomic tools have challenged the “unknown” designation. Depending on histologic features, sites of disease, and performance status, a small but significant minority of patients will be long-term survivors, and it is important to identify these groups of patients (^4,5).

This chapter discusses the evaluation of patients with CUP and optimal therapeutic strategies in the era of sophisticated diagnostics. The differing natural histories in CUP, depending on both the sites of disease and histology, are also discussed. Studies showed that, in this population, a search for the primary tumor beyond “routine” evaluation is unrewarding in the majority of patients (⁵). This fact has caused much consternation for both patients and physicians. The foundation for cancer treatment traditionally relies on identification of the tumor origin, thereby allowing treatment to be chosen based on the known natural history as well as specific therapies that have been proven effective for the cancer; this is becoming even more important with the rapid emergence of targeted therapies. Without knowledge of the primary site, the oncologist is often hesitant to recommend therapy, especially given the disease heterogeneity. Although most patients with metastatic CUP have tumors that respond poorly to current treatments and will consequently have a poor prognosis, it has become evident over the last two decades that subsets of patients with CUP have a favorable prognosis and respond to chemotherapy or can be successfully treated with regional therapy alone. The current era of sophisticated diagnostics and introduction of targeted therapies has been particularly important in the CUP setting; this cancer entity is the epitome of personalized therapy.

The incidence of CUP cancers as estimated from the cancer registries and databases of “unknown and unspecified cancers” is reported to be approximately 3%-4% of all cancers (⁶). This is probably an overestimation because this group includes a mix of patients: those with true CUP, those with primaries not yet diagnosed at the time of death, and those with difficult-to-diagnose tumors. Further, improved imaging allowing identification of small primary tumors suggests that the incidence of true CUP is decreasing.

A minority of patients (10%) with CUP have a history of an antecedent cancer. In autopsies performed before the advent of CT, the occult primary tumor was identified in 60% to 80% of cases. In one autopsy series, the two most commonly identified primary sites were the pancreas (20%) and lungs (18%) (^7,8). Given the current high-quality CT imaging and positron emission tomographic (PET) scans, it is unclear whether these cancer profiles are still the majority.

The classification of CUP continues to evolve. The last four decades have seen a shift in our understanding of CUP. First, improved imaging increased our confidence for the entity termed occult primary. Later, “favorable” CUP subsets were determined, based primarily on histopathology, the pattern of spread of select CUP cancers, and serum markers. Subsequently, with the advent of novel IHC markers and advances in diagnostic pathology, tissue of origin (ToO) profiles were described that assigned additional putative primary sites to CUP cancers based on IHC patterns. Current research involves the application of proteomic and genomic tools to CUP cancers.

Carcinomas of unknown primary, despite their heterogeneity, are a clinically unique oncologic entity; as such, they share many common features that set them apart from other malignancies. The central unifying clinical feature of CUP is the absence of a detectable primary tumor. Previous studies have shown that, even after an autopsy, the primary site will not be identified in 20% to 40% of cases; that number is likely much lower with significant improvements in imaging. At present, it is not known why primary carcinomas exhibit this unique biological behavior. One current hypothesis is that the acquisition of a “metastatic phenotype” is an early event in CUPs, soon after oncogenesis, thus enabling cells to metastasize early, before the development of a clinically detectable tumor (⁹). It has also been hypothesized that the primary tumors may regress or involute before the metastases become clinically evident, attributed to a host immunologic response. A third hypothesis is that the primary tumor is exposed to antiangiogenic factors locally, whereas the metastases acquire the angiogenic phenotype after a period of dormancy (¹⁰).

Several studies have demonstrated a specific nonrandom pattern of chromosomal aberrations that seems to be unique to CUPs. These data suggest that some of these genetic changes may be the underlying cause of the metastatic phenotype. Carcinoma of unknown primary is characterized by greater genetic instability, with massive GAs, when compared with other distant metastases. In a study by Pantou and colleagues (¹¹), cytogenetic profiling of tumors from 20 patients with CUP was performed, revealing an average of 11 chromosomal changes per case. Of the three histologic subtypes in this study, adenocarcinomas had not only the highest number of cytogenetic changes (16 vs 3) but also involvement of distinct sites (4q31, 6q15, 10q25, and 13q22) when compared with carcinomas or undifferentiated malignancies. The latter group was distinguished by the involvement of changes at 11q22. Overall, the most commonly rearranged chromosomal regions were 1q21, 3p13, 6q21-23, 7q22, 11p15-12, and 11q14-24. The number of cytogenetic alterations was found to be prognostically relevant. Median survival was significantly greater for patients with five or fewer cytogenetic changes compared with those with more than five changes (3 vs 18 months, P = .003). An older study of 12 CUP cell lines also demonstrated a preponderance of chromosome 1 abnormalities. These changes were observed on both the long arm (eg, 1p deletion, isochromosome 1p, and translocations with a 1p breakpoint) and on the short arm (1q21), suggesting the importance of chromosome 1 in the biology of CUPs (¹²). Chromosome 1p aberrations are also commonly associated with advanced malignancies.

Chromosome 12 abnormalities have also been shown in CUP. This is of particular interest because one of the observed alterations, isochromosome 12p (i12p), is present in as many as 80% of germ cell tumors. Motzer and colleagues reported that 30% of patient tumors in their series had either i12p or 12q deletions. The presence of either of these two cytogenetic abnormalities was found to be predictive of a complete response to cisplatin-based chemotherapy (75% vs 17%, P = .002) (^13,14). In the current era of sophisticated IHC, we rarely miss a patient with an extragonadal presentation and seldom order cytogenetics to inform therapeutic plans.

The tumor suppressor gene p53 is commonly mutated in human cancers, especially in advanced malignancies. Paradoxically, this does not seem to be true in CUP. Bar-Eli and colleagues found p53 mutations to be less frequent than expected (26%) in CUP after evaluating 15 biopsy specimens and 8 cell lines (¹⁵). However, work by other researchers evaluating IHC studies of p53 in CUP has found this protein to be highly expressed in 70% of the tumors examined (^16,17). Nevertheless, p53 expression has not been found to have prognostic relevance. Molecular studies have also demonstrated the overexpression of other oncogenes, such as c-myc, ras, bcl-2, and Her2/neu, in CUPs, but none have been found to have any correlation with either survival or response to chemotherapy (¹⁸).

More recent data with next-generation sequencing likely has therapeutic implications for patients with CUP, and it is exciting times for CUP researchers to extend the therapeutic envelope. In a recent study, Ross and colleagues presented results from a retrospective study of 200 consecutive CUP tumor specimens that that underwent comprehensive genomic profiling (CGP) using the hybrid-capture–based Foundation-One 10 assay (¹⁹). The DNA extracted from these CUP tumor specimens was analyzed after hybridization capture of 3,769 exons from 236 cancer-related genes and 47 introns of 19 genes commonly rearranged in cancer. There were 125 adenocarcinomas of unknown primary site (ACUPs) and 75 nonadenocarcinomas (non-ACUPs). The authors reported that a large number of CUP samples (85%) harbored at least one clinically relevant genomic alteration (GA) with the potential to influence and personalize therapy. The mean number of GAs was 4.2 GAs per tumor. The ACUP tumors were more frequently driven by GAs in the receptor tyrosine kinase (RTK)/Ras/mitogen-activated protein kinase (MAPK) signaling pathway than non-ACUP tumors. The authors concluded that CGP can identify novel treatment paradigms and suggested that early testing may have utility in CUP management. These data do illustrate some important considerations in the management of CUP cancers, as discussed in this chapter.

The clinical course of patients with CUP varies widely. Median survival in large retrospective studies has ranged from 11 weeks to 11 months. In the University of Texas MD Anderson Cancer Center study, the 5-year overall survival (OS) rate was only 11%. Although survival is poor as a whole, there are certain prognostic variables that correlate with longer survival, including disease limited to one organ site, involvement of lymph nodes only, and histologic diagnoses of squamous or neuroendocrine carcinoma. Variables suggestive of a poor prognosis include male sex, histologic diagnosis of adenocarcinoma, and metastatic involvement of the liver, lungs, bone, pleura, or brain (Table 45-1) (⁴).

By performing multivariate analyses on a consecutive series of 1,000 patients with CUP with classification and regression tree (CART) analysis, Hess and colleagues (⁴) were able to more closely study the interactions between different clinical variables and how this influenced survival.

Patients with CUP present with symptoms and signs similar to those of patients with advanced malignancies of known origin. In one review, the most common symptoms at presentation of CUP were general deterioration (73%), digestive symptoms (58%), liver enlargement (58%), abdominal pain (56%), respiratory symptoms (45%), ascites (26%), and node enlargement (16%) (²⁰). Most patients with CUP present with multiple metastases, with three or more organs involved. In patients with a dominant (or single) site of metastasis, the most common reported sites were liver (25%), bone (22%), lungs (20%), lymph nodes (15%), pleural space (10%), and brain (5%).

In the past, minimalist diagnostic strategies had been advocated, limiting the scope of initial evaluations to differentiate only between treatable and untreatable disease. Others have supported a more aggressive approach, wherein a complete assessment of the extent of the disease and detection of the primary tumor site are attempted. In our experience, a more pragmatic approach is better. Extensive evaluation of all patients presenting with metastases is an expensive and wasteful extreme that does not benefit patients. In one study, the average cost of evaluating a patient with CUP was $17,973 and much higher with recent testing (^21,22). In that study, mean survival was 8.1 months, representative of the natural history of CUP, with only 18% of patients surviving 1 year. However, a strictly minimalist approach may result in the oversight of treatable and potentially curable neoplasms.

An important determinant of the appropriate extent of evaluation for any patient with CUP is whether the data obtained by a diagnostic test will influence treatment decisions. If a treatable or potentially curable cancer is strongly suspected (eg, a germ cell tumor or lymphoma or oligometastatic disease treated with surgery or multimodality therapy), further investigation should proceed until a precise clinical diagnosis can be made, provided that therapy is not unreasonably delayed. The recommended general approach at the present time is thus one of a directed evaluation based on clinical presentation and pathologic findings; predictions of tumor origin or mutations from molecular profiling techniques may also play a role in streamlining the scope of evaluation.

Physical Examination and Laboratory Tests

A thorough medical history should be obtained, and a physical examination, including a digital prostate examination in men and a breast and pelvic examination in women, should be performed. Determination of the patient’s performance status, nutrition, and the presence or absence of concomitant medical illnesses and malignancy-related complications (eg, paraneoplastic syndromes or painful metastases) that may affect patient care is required.

Laboratory tests should include routine biochemical and hematologic surveys. The role of tumor markers in the evaluation of patients with CUP is often not diagnostic. Most tumor markers are nonspecific and are not useful for identifying a primary site or for prognostic purposes. Adenocarcinoma markers (eg, carcinoembryonic antigen [CEA], cancer antigen 125 [CA 125], CA 15-3, and CA 19-9) are often elevated in patients with CUP and cannot be reliably used to identify a specific primary site or to predict either OS or the exact burden of metastatic disease (^23,24,25). Serum tumor markers may play a role in helping to evaluate patients for responses to therapy, although levels are not always predictive of response to chemotherapy.

Their selective use in a directed approach is more helpful than ordering a large battery of tumor markers on all patients who present with CUP. Men who present with metastatic adenocarcinoma and osteoblastic bone metastases CUP should have PSA and prostatic acid phosphatase levels measured. In all men with undifferentiated (or poorly differentiated) midline carcinomas, beta-human chorionic gonadotropin and alpha-fetoprotein levels should be measured, especially if the clinical presentation suggests an extragonadal germ cell tumor. In patients with hepatic tumors, alpha-fetoprotein levels should also be measured if there are risk factors or pathologic characteristics that suggest a possibility of primary hepatocellular carcinoma.

Diagnostic Imaging and Invasive Studies

In the absence of contraindications, a baseline intravenous contrast CT scan of the chest, abdomen, and pelvis is the standard of care, as supported by the National Comprehensive Cancer Network and National Institute for Health and Clinical Excellence CUP radiology guidelines (^26,27). Patients with CUP should then be approached in a “directed” fashion. Endoscopy of the upper and lower gastrointestinal (GI) tract is indicated for patients with abdominal complaints, ascites, liver metastases, or other findings in the initial workup and pathology that are indicative of a possible GI primary tumor.

All women with CUP and adenocarcinoma should undergo mammography. In cases of suspicious findings on a breast examination and negative mammography findings, patients should have breast sonography and a biopsy as indicated. Because both the sensitivity (23%-29%) and specificity (71%-73%) of mammography in detecting an occult carcinoma are low, breast magnetic resonance imaging (MRI) has been evaluated as an alternative in patients with a high suspicion for breast primary cancer. In the setting of isolated axillary adenopathy, MRI is sensitive in detecting occult primary breast cancers (>75%) and should be performed in women with isolated axillary adenopathy and negative mammography findings (^{28,29,30,31,32,33}). Women with adenocarcinoma presenting with metastatic sites other than cervical or axillary adenopathy that are compatible with breast cancer (ie, bone, liver, or lungs) may also undergo breast MRI if the mammography findings are negative (^32,33).

Patients with upper or midcervical adenopathy with a squamous cell carcinoma on pathology should undergo a thorough head and neck evaluation, including panendoscopy (ie, laryngoscopy, bronchoscopy, and esophagoscopy) with random biopsies. Ipsilateral or (more often) bilateral tonsillectomy has also been recommended as part of the staging process because this has been shown to identify an occult primary lesion deep in the tonsillar crypts in up to 30% of patients with this presentation of CUP (^34,35,36). Computed tomography of the head and neck region is routinely done as part of the initial workup. In addition, the utility of 18-fluorodeoxyglucose positron emission tomography (FDG-PET) has been well documented in patients with squamous carcinoma and cervical adenopathy; small prospective and retrospective studies suggest that a primary head and neck tumor is identified in 25% to 30% of these patients (^{36,37,38,39,40,41,42}). A recent retrospective review found that the primary tumor site was identified in 44% of these patients undergoing PET-CT fusion scans, and this modality appears to be emerging as a superior alternative to either PET or CT alone (^38,39).

Outside the indication given previously, the role of PET-CT is unclear. Several small studies have evaluated the utility of PET in CUP patients (^43,44). Moller et al reviewed (¹⁸) FDG-PET as a diagnostic tool for patients with extracervical CUP (⁴⁴). They identified four publications (152 patients); these studies were retrospective and heterogeneous in their inclusion criteria, study design, and diagnostic workup prior to FDG PET-CT. The primary tumor was detected by FDG PET-CT in 39% of patients with extracervical CUP. Lung was the most commonly detected primary tumor site (≈50%). Pooled estimates of sensitivity, specificity, and accuracy of FDG PET-CT in the detection of the primary tumor site were 87%, 88%, and 87.5%, respectively. They concluded that FDG PET-CT may have a role in identification of the primary tumor in extracervical CUP; however, prospective studies with more uniform inclusion criteria are warranted. Although not studied prospectively, PET-CT scans may be useful in selected patients with solitary metastases prior to definitive locoregional therapies and in follow-up of patients with predominant bone disease (²).

Histopathologic Evaluation

All pathologic material obtained at biopsy from a patient with CUP should be evaluated by an experienced pathologist who is familiar with CUP workup. The pathologist should also be informed of the patient’s pertinent history and clinical findings so that he or she can recommend further analysis on the basis of this information. In the CUP world, pathology trumps radiology. Collaboration between the pathologist and treating oncologist is critical. Adequate tissue sampling is essential.

The CUP cancers include adenocarcinoma, poorly differentiated adenocarcinoma (60%); poorly differentiated carcinoma (PDC), undifferentiated carcinoma, or undifferentiated neoplasm (30%); squamous cell carcinoma (5%); and neuroendocrine cancer (2%). Rarely (2%), CUP can present as mixed tumors, including sarcomatoid, basaloid, and adenosquamous carcinomas (Table 45-2).

Adequacy of tissue is essential, especially when the pathologist has to make a diagnosis on deep fine-needle aspirations and there is insufficient tissue for IHC staining. The diagnosis of a poorly differentiated neoplasm implies that the pathologist is unable to classify it into any of the general neoplastic categories (carcinoma, lymphoma, melanoma, or sarcoma). Subsequent evaluation of this group of poorly differentiated lesions by means of special IHC techniques is warranted because some of these patients will have tumors that are potentially curable and very responsive to treatment. Many IHC reagents are at the disposal of the pathologist, making the histologic classification of the tumor easier (Table 45-3).

Especially useful are the antibodies to common leukocyte antigens present in lymphoma and the antibodies to PSA present in most prostate cancers. Other useful IHC markers include cytokeratin CK7, CK20, and thyroid transcription factor (TTF-1). Thyroid transcription factor is a nuclear transcription factor that is normally expressed in lung and thyroid tissues and in their neoplasms. Staining for TTF-1 is frequently positive in lung cancer, especially in adenocarcinomas (60%-75%) and small cell lung cancers (66%-87%); however, it is inconsistently expressed in squamous cell carcinoma. Among the various monoclonal antibodies against various cytokeratins, CK7 and CK20 can help differentiate between different solid tumors. (For instance, CK7 is more commonly associated with pulmonary or gynecologic malignancies, whereas CK20 is frequently seen in GI adenocarcinomas.) The CK7⁺/CK20⁻ immunophenotype, in conjunction with TTF-1 staining, is suggestive of a lung primary and is a highly sensitive and specific method for differentiating primary pulmonary adenocarcinomas from metastatic extrapulmonary adenocarcinomas (Fig. 45-1) (^{45,46,47,48,49}). In contrast, the CK7⁻/CK20⁺ immunophenotype is suggestive of a colorectal primary site. CK7⁺ and CK20⁺ dual staining suggest a malignancy of urothelial origin. Using light microscopy and IHC, a putative primary tumor may be assigned in up to a third of CUP cases. Immunohistochemistry can also suggest biomarker studies with potential therapeutic impact (eg, Kras, epidermal growth factor receptor [EGFR], Her2, and ALK mutations).

Hep par 1 is an antigen whose expression is confined to benign and malignant hepatocytes and aids in the diagnosis of hepatocellular carcinoma in patients with CUP presenting with liver lesions. In women, depending on the pathology and pattern of metastasis, estrogen receptor and progesterone receptor staining is done to look for a breast primary. Another marker for a breast primary is gross cystic disease fluid protein 15 (GCDFP-15), which is present in 62% to 72% of breast cancers.

Dennis and colleagues (⁵⁰) have identified other novel molecular markers using a bioinformatics approach. All publicly available gene expression data from various adenocarcinomas were pooled together, and four novel proteins not previously recognized as tumor markers were found to be significantly upregulated. This was confirmed by reverse transcription–polymerase chain reaction. One example was lipophilin B, which was found to be restricted to breast, ovarian, and prostate cancers.

The use of cytogenetic analysis in the diagnosis of CUPs is limited. Specific chromosomal abnormalities have been identified in several types of lymphoma (8; 14 translocation in small non–cleaved-cell non-Hodgkin lymphoma), germ cell tumors (i12p), and Ewing sarcoma (t11; 22 or t21; 22). In the cytogenetic study by Pantou et al (¹¹), lymphoma was diagnosed in four patients with CUP on the basis of the presence of immunoglobulin (Ig) H/Alk-1 rearrangement, which was identified by fluorescence in situ hybridization. In addition, one patient was diagnosed with Ewing sarcoma due to detection of characteristic rearrangement of chromosomes 11 and 22. In the era of novel and sophisticated IHC, cytogenetics is rarely ordered to help with therapy decisions. Most useful are patterns and groups of IHC used in an effective algorithm that can prove to be clinically appropriate and cost effective, although the approach to this is not uniform in the community (^51,52).

Molecular Profiling of Carcinomas of Unknown Primary

Tissue of origin molecular profiling using DNA microarray and reverse transcription–polymerase chain reaction (RT-PCR) is a promising technique to establish the putative primary diagnosis in patients with CUP (^53,54) The performance of these assays in known cancers has been validated using independent blinded sets of tumor samples, with an accuracy of approximately 90% (^55,56,57). The feasibility of using formalin-fixed small biopsy or fine-needle aspiration samples makes this practical for use in the clinic setting.

Messenger RNA– or microRNA-based ToO assays have been studied in prospective and retrospective CUP trials. In CUP, these assays suggest a specific primary tumor in about 70% of patients studied (^{58,59,60,61,62,63,64}). Because, by definition, the primary tumor site is not known in CUP, validation of site prediction in CUP remains a challenge. Results from multiple small studies suggest that ToO predictions are likely accurate. Indicators include (1) concordant results of ToO testing and the appearance of a latent primary during the course of the patient’s illness (or many years later); (2) strong correlation between ToO assay results and diagnostic IHC (both suggesting a single putative primary); (3) a prospective study demonstrating that patients treated with ToO-based site-directed therapy had survival similar to those with the corresponding known primary (⁶⁵).

In the United States, there are currently three commercially available Medicare-approved tests. They differ in the platform used, number of potential cancers identified, the size and histologic samples of the training set, and the reporting format. All three assays report on identifying a putative primary in about 75%-85% of patients with CUP. No comparative studies have been conducted, although in our personal experience when IHC is diagnostic, all three tests usually give similar results.

At present, the only outcomes-based study is a prospective single-arm study evaluating the role of the 92-gene assay to predict the ToO and assay-directed site-specific therapy in patients with CUP (⁶⁵). The authors concluded that the median OS of 12.5 months (95% CI, 9.1 to 15.4 months) for patients who received assay-directed site-specific therapy compares favorably with previous studies using empiric therapy. Biliary and urothelial profiles were 33% of the predictions. Unfortunately, firm conclusions of therapeutic impact cannot be drawn from this study given the nonrandomized design; statistical biases; confounding variables, including use of subsequent lines of (empiric) therapy; and the heterogeneity of the CUP cancers.

At this time, we lack randomized controlled trials to gauge the therapeutic impact of ToO molecular profiling assays. Creative trial designs are urgently needed to study CUP subsets and impact of these assays on survival and quality of life of patients with CUP.

Two prospectively defined blinded studies of difficult-to-diagnose primary cancers (several poorly differentiated) have reported on the cost effectiveness of ToO molecular profiling over IHC. Samples were evaluated by IHC/morphology analysis or the ToO molecular profiling test (⁶⁶). Accuracy was defined based on comparison with pathology of known primary. In one, the assay demonstrated overall accuracy of 79% for tumor classification versus 69% for IHC/morphology analysis (P = .019). Mean IHC use was 7.9 stains per case (range 2 to 15). The other study also confirmed similar findings (⁶⁷); the assay accurately identified 89% of specimens, compared with 83% accuracy using IHC (P = .013). In the subset of poorly differentiated and undifferentiated carcinomas, assay accuracy exceeded that of IHC (91% to 71%, P = .023). These results have important implications for management of CUP cancers and warrant a study of an integrated algorithm evaluating ToO molecular profiling complementing IHC in select patients. At this time, we lack randomized controlled trials to gauge the therapeutic impact of ToO molecular profiling assays. Creative trial designs are urgently needed to study CUP subsets and impact of these assays on survival and quality of life of patients with CUP.

Carcinomas of Unknown Primary Presenting as Isolated Brain Metastases

In up to 15% of patients presenting with brain metastases, the primary site remains unknown (^68,69). The important factor in treating patients with brain lesions is to distinguish patients with metastatic disease from those with primary brain tumors. Once this distinction has been made, patients with single metastatic lesions should be considered for surgery, and those with multiple lesions should receive radiotherapy. In a recent small prospective study, patients with CUP who had single brain metastases treated with gross total resection and subsequent whole-brain radiotherapy (WBRT) had a median survival of 13 months. Patients with CUP who had multiple brain metastases and who underwent either WBRT alone or gross resection of symptomatic lesions followed by adjuvant WBRT had a median survival of only 6 to 8 months (⁶⁹). Stereotactic brain radiation is often used in the CUP setting using the same principles as for brain metastases from known primary cancers.

Carcinomas of Unknown Primary Presenting as Metastatic Cervical Adenopathy

In this subgroup, patients present with high-to-midcervical or -supraclavicular adenopathy; on histopathology, these tumors are squamous cell or PDCs. For squamous cell carcinoma, the primary site is eventually identified during follow-up in approximately 20% of patients, with the tonsil being the most common site, followed by the pyriform sinus and base of the tongue.

Adenocarcinoma is much less common and is generally from either metastatic nonpapillary thyroid carcinoma or advanced malignant disease from a distant site (GI, lung, or breast carcinoma presenting as a metastatic supraclavicular node). Of all malignancies of the head and neck, only 5% to 10% are classified as unknown primary after imaging and panendoscopy. The prognosis for patients with cervical CUP overall is better than that for other CUP clinical subgroups, but even within this group, significant heterogeneity exists. Yalin et al (⁷⁰), in a retrospective study of 107 patients with cervical CUP (62% PDC, 24% squamous carcinoma, and 14% adenocarcinoma), reported a 5-year OS rate of 35.5%. In another retrospective study by Issing et al (⁷¹), 5- and 10-year OS rates were 42.7% and 30.6%, respectively. The prognosis is significantly worse in the presence of any of the following: adenocarcinoma, level III/IV lymphadenopathy, multiple lymph nodes, and bulky disease.

Patients with supraclavicular adenopathy have a far worse prognosis than those with adenopathy in other lymph node–bearing areas. Carcinoma affecting supraclavicular lymph nodes on the right most commonly arises from occult primary tumors of the lungs and breast. When disease affects the lymph nodes on the left side, spread from intra-abdominal malignancies by way of the thoracic duct (Virchow node) is an additional possibility.

The management of patients with cervical CUP has become increasingly controversial, primarily because of the question of postoperative radiation therapy. The notion of adjuvant irradiation of all potential mucosal sites has been questioned because of the absence of any demonstrated survival benefit in randomized studies. To date, postoperative radiation therapy in cervical CUP significantly improves locoregional control, but this does not translate into improved OS. This being said, combined-modality therapy (surgery and radiation therapy) is better than either modality alone (⁷²). Most patients with only cervical or supraclavicular involvement should have regional therapy consisting of surgery, postoperative radiation therapy, and close follow-up. Patients who undergo an excisional biopsy for diagnosis usually do not need additional surgery if no gross disease is left behind, only a single lymph node measuring less than 6 cm is involved, and no extracapsular extension is noted on pathologic review. If any of these features is present, a neck dissection is indicated. In addition, for patients with squamous cell carcinoma, unilateral tonsillectomy ipsilateral to the presenting neck mass is commonly advocated as part of the surgical treatment because occult tonsillar carcinomas are usually found in 18% to 39% of patients who undergo tonsillectomy (^71,72). Identification of the primary site would thereby reduce morbidity by limiting the field of radiation and would improve surveillance.

In patients with N1 or N2a disease (squamous cell), it is unclear whether postoperative radiation improves local control because studies have been contradictory. In this case, close surveillance would also be an acceptable option after surgery. All other patients should receive postoperative radiation to the bilateral neck covering all potential occult primary sites (ie, nasopharynx, oropharynx, and hypopharynx). The 3-year survival rate after radical neck dissection or radical neck irradiation ranges from 35% to 60%. Within this group, patients with N1 disease have a better prognosis; patients with N3 disease, regardless of the local treatment modality used (surgery, radiotherapy, or both), fail to achieve complete remission in 65% of cases.

Although the role of chemotherapy in patients with cervical CUP remains poorly defined, extrapolation of phase II/III data in head and neck cancer indicates a role in patients with advanced nodal disease (N3). A recent large meta-analysis of more than 10,000 patients in 63 trials with head and neck squamous cell carcinoma demonstrated a small but significant absolute survival benefit of 4% at 5 years for chemotherapy (⁷³). Intensive concurrent chemoradiotherapy in unresectable squamous cell head and neck cancers with cisplatin/5-fluorouracil–based and cetuximab-based regimens has resulted in improved complete response rates, locoregional control, and preservation of organ function, albeit at the cost of significant toxicities.

Women With Carcinomas of Unknown Primary and Isolated Axillary Adenopathy

Women who present with adenocarcinoma in the axillary lymph nodes compose another subset with a more favorable prognosis. These patients are often managed as women with stage II breast cancer. Isolated axillary adenopathy is an uncommon presentation of breast cancer, accounting for only 1 to 3 of every 1,000 diagnosed breast cancers. Mammography and ultrasound should be performed, and biopsies should be performed on any identified lesions. If mammography findings are normal, additional imaging of the breast with MRI is indicated because of its greater ability to detect small primary breast tumors (70%-95% sensitivity). Magnetic resonance imaging has a very low false-negative rate. Of approximately 40 women reported in the literature with isolated axillary adenocarcinoma and negative breast MRI findings, only 4 were found to have breast cancer at surgery or during follow-up (⁷⁴).

The present recommended management of women with CUP of the axilla includes axillary dissection, axillary radiotherapy for those at high risk of local recurrence (eg, extracapsular invasion or more than four positive lymph nodes), and appropriate systemic therapy for breast cancer, depending on age and menopausal status. If breast MRI findings are negative, neither mastectomy nor breast irradiation is recommended (^75,76,77). If the breast MRI is positive or suspicious, radiation to the breast is usually recommended. The prognosis is not as favorable in men who present with axillary adenopathy only (⁷⁵).

This management paradigm is changing as molecular profiling complements pathology as a diagnostic tool in this subset of patients. All women with axillary adenopathy do not have occult breast cancer. Profiling for ToO can help with treatment decisions especially if the IHC does not correlate with breast cancer and ER, PR, and Her-2 status is negative.

Carcinomas of Unknown Primary Presenting as Isolated Inguinal Adenopathy

A few patients with CUP present with inguinal adenopathy. Undifferentiated (anaplastic) carcinoma is identified in at least half of these cases. Some of these anaplastic “carcinomas” appear to be melanomas with no obvious primary skin lesion. The remaining patients have squamous cell carcinomas arising from the skin, genitourinary tract, anus, or pelvis. A detailed investigation for primary lesions in these areas is important because curative therapy is available for carcinomas of the anus, vulva, vagina, and cervix even with spread to regional lymph nodes. In patients with carcinomas and PDCs confined to the groin nodes, where no primary site was identified, a superficial groin dissection should be performed with or without radiation therapy. Bimodality therapy with surgery and radiation may increase the risk of significant lymph edema and requires careful planning. Chemotherapy, before definitive therapy and in the context of a clinical trial, may be offered to patients with bulky locoregional adenopathy and is not an uncommon practice in the clinic.

Carcinomas of Unknown Primary and Isolated Pleural Effusions

Most patients with isolated pleural effusions have adenocarcinomas, which may sometimes be difficult to differentiate from mesotheliomas. Newer IHC markers (eg, calretinin, CK 5/6, and WT1 [Wilms Tumor-1]) that are more sensitive in differentiating epithelioid malignant mesothelioma from pulmonary adenocarcinoma can assist in the diagnosis (⁷⁸). Additional IHC markers, including TTF-1, CK 7/20, and breast markers, should routinely be done as first- and second-tier diagnostics to aid in treatment. If the effusion reaccumulates quickly, pleurodesis may be attempted to slow the rate of fluid reaccumulation, or as done more often currently, a pleural catheter with daily aspirations is preferred (this can be removed after chemotherapy response is noted and the flow decreases). Chemotherapy is initiated in most patients based on their IHC profile and taxane plus carboplatin versus gemcitabine plus cisplatin are commonly used doublets.

Carcinomas of Unknown Primary Presenting as Malignant Ascites

Patients with malignant ascites usually belong to one of two subsets, each with a very different natural history of disease. The first group consists of patients with mucin-producing adenocarcinoma, who may present with ascitic fluid that contains signet-ring cells. These patients often have multiple peritoneal implants, with the primary site most likely being the GI tract (ie, stomach, small bowel, appendix, colon, or pancreaticobiliary). Given the current armamentarium of drugs available for treatment of metastatic colon cancer and the improved survival, it is important to consider those combinations for patients with IHC suggestive of colon profile (CK20⁺, CK 7⁻, and CDX-2⁺). The second subset is composed of women patients with primary serous papillary peritoneal carcinomatosis. This disease is often also associated with pelvic adenopathy or masses. These patients may have elevated CA 125 levels but do not have detectable ovarian cancer. Some investigators consider these patients to have true unknown primary ovarian tumors or primary serous carcinomas of the peritoneum (^79,80). Disease management should be the same as for women with ovarian carcinoma. A prolonged median survival of 13 months, with 25% of patients having a progression-free survival lasting more than 2 years, was reported for paclitaxel/carboplatin-based chemotherapy in patients with peritoneal carcinomatosis. In this study, a high overall response rate (ORR) and number of complete responses were reported for this subgroup of patients with CUP (68.4% and 20%, respectively).

Carcinomas of Unknown Primary Presenting as Isolated Bony Metastases

When bone metastases are detected, men should be evaluated for prostate cancer and women for breast cancer given that they may be candidates for hormonal therapy, which is relatively easier compared to cytotoxic therapies. Other cancer profiles include lung, cholangiocarcinoma, renal, and rarely melanoma. Patients with a single bony metastasis may be candidates for surgery or radiation and then monitored. Patients with disease at multiple sites and good performance status and whose tumors progress after radiation therapy should be offered a trial of chemotherapy. Many experimental agents are currently available in ongoing clinical trials. Therapy with bone-seeking radioisotopes (eg, strontium 89) may be useful in the treatment of disseminated painful bone metastases in a few patients. Bisphosphonates are routinely used as in other malignancies, such as multiple myeloma, breast cancer, and prostate cancer. Often, PET-CT is the imaging modality of choice to follow response to therapy for disseminated osseous metastatic disease.

Carcinomas of Unknown Primary Presenting as Hepatic Metastases

Patients with hepatic metastases constitute 30% to 40% of people with CUPs; they compose a clinical subgroup with a relatively poor prognosis, with reported median OS between 49 days and 7 months. The most important diagnostic considerations in this class are to distinguish primary liver and biliary tumors (hepatocellular carcinoma and cholangiocarcinoma) from cancers that have metastasized to the liver and to identify patients with neoplasms of a more indolent nature (eg, neuroendocrine tumors). A careful pathologic review with IHC of liver biopsy specimens is therefore essential. The two most common histologies in primary CUP of the liver are adenocarcinoma (55%) and poorly differentiated/undifferentiated carcinoma (30%). The recommended initial therapy for unresectable disease is systemic chemotherapy, and surgery may be considered an option for those with resectable disease.

Neuroendocrine Tumors of Unknown Primary Site

Neuroendocrine tumors compose about 4% of all CUPs and commonly present with diffuse liver or bone metastases. Histologically, neuroendocrine tumors can be well differentiated or low grade, with features that are typical of carcinoid or islet cell tumors exhibiting a more indolent behavior. Management of these tumors should be similar to established guidelines for metastatic low-grade neuroendocrine tumors from a known primary site. In patients with limited disease, surgical resection or chemoembolization may be appropriate. If not amenable to local therapy, then targeted therapy is considered with (anti–vascular endothelial growth factor [anti-VEGF] agents, including sunitinib, or mammalian target of rapamycin [mTOR] inhibitors, including everolimus.

A second group involves high-grade neuroendocrine tumors that may present as PDC by light microscopy but have strong neuroendocrine features revealed by IHC (ie, neuron-specific enolase, chromogranin A, and synaptophysin positive). These high-grade neuroendocrine tumors are treated like small cell lung carcinoma with etoposide plus platinum or irinotecan plus platinum combinations, with high reported response rates.

Carcinomas of Unknown Primary and Extragonadal Germ Cell Syndrome

As a group, patients who have undifferentiated carcinoma or PDC are younger than 50 years and present with rapidly growing midline tumors involving the lymph nodes, mediastinum, or retroperitoneum; their tumors have been found to be very responsive to chemotherapy, particularly to platinum-containing regimens. It is believed that these patients have poorly differentiated extragonadal germ cell tumors. They have response rates to chemotherapy of 35% to 50%, and those who achieve a complete response often enjoy a durable remission. In a prospective study by Hainsworth and colleagues of 220 patients with PDC or poorly differentiated adenocarcinoma (PDAC) treated between 1978 and 1989 with cisplatin-based chemotherapy regimens, approximately half of the patients had a predominant tumor location in the mediastinum, retroperitoneum, or peripheral lymph nodes. The ORR was 63%, with 26% complete responses and an actuarial 10-year disease-free survival rate of 16%.

However, this was not found to be true by Lenzi and colleagues (⁴), who retrospectively reviewed the clinical outcomes of 337 patients with PDC/PDAC. No prolonged survival was observed in this cohort of patients, and no significant survival advantage resulted from cisplatin-based chemotherapy. Moreover, elevated serum levels of alpha-fetoprotein or beta-human chorionic gonadotropin, contrary to other reports in the literature, were not found to be predictive of an improved median OS. This discrepancy may have resulted from several confounding factors.

First, older studies of extragonadal germ cell syndrome included patients with PDCs who in actuality did not have CUP but had other highly treatable malignancies (⁴). In a study by Hainsworth et al (⁸¹), of the 36 long-term survivors, 20% were subsequently found to have either lymphoma (⁵), testicular cancer (¹), or leiomyosarcoma (¹). Conversely, in the study by Lenzi, patients in whom the primary site was identified were excluded from the analysis. Most of these patients were found to have highly treatable malignancies, such as lymphoma (6%), breast cancer (8%), ovarian cancer (3%), germ cell tumors (2%), and prostate cancer (1%). Exclusion of these patients would significantly reduce response and median survival rates.

Second, even among patients with PDC/PDAC of unknown primary, significant heterogeneity exists. In the study by Lenzi, CART analysis of 337 patients revealed different groups with widely discrepant survival times. The group with the longest median OS (40 months) included patients with PDC, lymph node involvement, and only one or two metastatic sites. By contrast, patients with non–lymph node metastases had a very poor prognosis, with a median OS of only 7 months (⁴).

Carcinomas of Unknown Primary and Single Sites Discovered Incidentally on Resection

Carcinomas of unknown primary are notorious for unusual, isolated presentations. Such lesions may appear on the skin, in single isolated lymph nodes removed during surgery for benign unrelated conditions, and at other, even more unusual sites. Patients should be examined for primary tumors and other sites of metastasis, as described previously. If no primary tumor and no additional sites of metastasis are found, complete removal of the lesion must be ensured; this often requires additional excision with wider margins (if skin or subcutaneous). The patient may then be monitored without therapy and in selected cases are candidates for radiation. Many such patients may enjoy prolonged survival. Patients with isolated skin lesions may have an undifferentiated primary integumentary tumor with a potential for cure after adequate local surgical treatment.

Combination regimens using newer chemotherapeutic agents have demonstrated greater benefit than did older single-agent therapies for CUP cancers. Several difficulties arise when survival and response rates reported in different chemotherapy trials are compared. For example, histologic criteria for patient selection often varied from study to study. Moreover, in older studies, IHC methods were not used to evaluate pathologic specimens. Despite these difficulties, no study has firmly established any chemotherapy regimen as the “gold standard” in CUP. The median survival in most studies, regardless of regimen, has ranged between 5 and 13 months, with response rates of less than 30% and without a significant improvement in survival (Table 45-4).

Nevertheless, patients with certain clinical subtypes (eg, peritoneal carcinomatosis and lymph node–predominant disease) do benefit from chemotherapy. Historically, cisplatin-based combination chemotherapy regimens were frequently used to treat patients with CUP. Response rates in the literature range from 12% to 26% and median survival from 5 to 7 months. Combining paclitaxel with carboplatin has modestly improved both survival and response rates. In patients with widespread metastases and poor performance status, however, systemic chemotherapy is unlikely to be beneficial, and only supportive therapy is usually indicated.

In a phase II study by Hainsworth and colleagues, patients with CUP (n = 55) received paclitaxel (200 mg/m² day 1), carboplatin (AUC = 6 day 1), and oral etoposide (50 mg alternating with 100 mg days 1-10) every 21 days (⁸²). Most were previously untreated, with only four having received prior chemotherapy. Most patients had moderately to well-differentiated adenocarcinoma (55%) or PDC/PDAC (38%), with squamous (2%) and neuroendocrine (5%) histologies less prevalent. The dominant sites of disease were lymph nodes (25%), liver (16%), and lungs (16%). Approximately 24% of patients in the study had multiple sites of disease, with 42% of patients having more than two metastatic sites. Response rates were equivalent in all histologic subgroups, with a reported ORR of 47% and a median OS of 13.4 months. This regimen was well tolerated, with myelosuppression the most common grade 3/4 toxicity. No treatment-related deaths were reported.

Briasoulis and colleagues (⁸³) found equivalent response rates and median OS in CUP with carboplatin (AUC = 6) and paclitaxel (200 mg/m²) without oral etoposide. In this phase II trial, patients (n = 77) were given a maximum of eight cycles of chemotherapy. In addition, granulocyte colony-stimulating factor was administered on days 5 to 12. The proportions of differing histologic subtypes were comparable to those in the Hainsworth study: adenocarcinoma (61%), undifferentiated (35%), and squamous (4%). Three distinct clinical subsets were present in this study: peritoneal carcinomatosis (25%, mostly women); visceral or bony metastases (43%); and predominant nodal or pleural disease (30%). The reported ORR, median response duration, and median OS were 38.7%, 6 months, and 13 months, respectively. Although response rates were equivalent for adenocarcinoma and undifferentiated carcinoma, significant differences were seen among the three clinical subsets: liver/bone or disseminated metastases (ORR, 15.1%; median OS, 10 months); nodal/pleural disease (ORR, 47.8%; median OS, 13 months); and peritoneal (ORR, 68.4% [75% for women]; median OS, 15 months), P = .01. Three patients with nodal-predominant disease had durable responses lasting longer than 2 years. Grade 3/4 neutropenia was only 4%, with two reported septic deaths.

The results of docetaxel in combination with carboplatin in one small phase II study appear to be inferior to those of paclitaxel/carboplatin in the trials mentioned. The ORR was 22%, the median OS was 8 months, and the 1-year OS rate was 29%. Differences in sites of disease and histology among these three studies may account for the discrepancy. Severe grade 3/4 myelosuppression was more frequent with docetaxel (50%) than with paclitaxel, with two reported septic deaths (⁸⁴).

Patients with undifferentiated or PDCs not fitting into the extragonadal germ cell or neuroendocrine clinical subgroups have traditionally been given a trial of a cisplatin-based regimen. Patients with squamous cell carcinomas who require chemotherapy are also often treated effectively using a cisplatin-based regimen.

The role of salvage chemotherapy in CUP is poorly defined. Gemcitabine has been published as second-line therapy in patients with previously treated CUP. In a phase II study by Hainsworth and colleagues (⁸⁵), gemcitabine was administered weekly at 1,000 mg/m² (on days 1, 8, and 15 of a 28-day cycle). All patients (n = 39) received two cycles and were then evaluated for response. Chemotherapy was continued for a maximum of six cycles for either an objective response or stable disease. Approximately 90% of patients had failed a prior regimen containing platinum and a taxane. Most patients had either adenocarcinoma (59%) or PDC/PDAC (31%). Median time to progression was 5 months. Gemcitabine was well tolerated, with 92% of patients receiving two or more cycles. The most common grades 3 to 4 toxicities were fatigue/weakness and mucositis/esophagitis.

Hainsworth et al (⁸⁶) reported on a combination-targeted therapy trial of bevacizumab and erlotinib in 51 patients; 25% were chemotherapy naïve with advanced bone or liver metastases, and 75% had been treated with one or two chemotherapy regimens. Responses were noted in 4 patients (8%), and 30 patients (59%) experienced stable disease or a minor response. The median OS duration was 8.9 months, with 42% of patients alive at 1 year.

These combination therapy trials have been a significant contribution in the post 5-flurouracil and cisplatin era of second-generation chemotherapeutic agents. They have certainly served their function in allowing the access to several broad-spectrum chemotherapies in patients with CUP and helped us understand the responses to these therapies. Although evaluation of empiric regimens was the preferred approach in the past, with the emergence of modern molecular diagnostic trials that help define CUP subtypes, our focus has shifted from empiric combinations to more tailored regimens, especially as directed by IHC and where helpful ToO or mutational profiles. Further, as therapies for known cancers improve and become more selective based on evolving predictive markers, the newer therapeutic approaches should be evaluated in the appropriate CUP subtypes as well.

All patients with CUP should undergo a directed diagnostic evaluation for the primary tumor and a detailed pathologic evaluation of the metastatic specimen. A subset of patients defined by clinicopathologic criteria and considered to have favorable prognosis benefit significantly from selective or aggressive treatments. For most patients who present with advanced disseminated CUP, the prognosis remains poor, and no unique empiric combination therapy of established efficacy is available. We have moved away from the paradigm of one treatment fits all to a more focused approach that integrates clinical presentation, pathologic evaluation, and the evolving diagnostic tools. Our current focus is to study the impact of molecular profiling and next-generation sequencing-based studies that help individualize CUP treatments. More broadly, there is an extensive push toward personalizing cancer care through the use of genomic tools to identify driver mutations in an individual tumor. For this approach to be successful, it will require both additional molecular insights and novel drugs that are effective against specific mutations.