Defining prognosis in Cancer of Unknown Primary (CUP)


Nicholas Pavlidis


Karim Fizazi

Date of publication: 14 May 2024
Last update: 14 May 2024


A patient is considered to have cancer of unknown primary when a tumor is detected at one or more metastatic sites and routine evaluation fails to identify the primary site. CUP is divided into favourable and unfavourable types. The unfavourable CUPs accounts for the 80% of CUP tumors and have the worse prognosis. Conventional chemotherapy did not yield the anticipated survival improvement. During the last 10 years prospective randomized trials between traditional cytotoxic agents and targeting treatment have been evaluated.


Cancer of unknown primary (CUP) are heterogenous, undifferentiated metastatic tumours in which the primary lesion responsible for metastatic spread cannot be identified, despite a thorough diagnostic work up. Diagnosing CUP, therefore, is especially challenging, and fewer than 20% of patients survive beyond 1 year following initial diagnosis. We explore the historical contributions made by European clinical researchers on the diagnostic, treatment and clinical trial progress of CUP, and reveal why overall survival and/or quality of life gains remain an unmet medical need.

##History and Epidemiology

CUP was first described in 1907 by Wiliam Halsted in three patients with hidden breast carcinoma (Halsted, 1907). CUP accounts for 2% of all human cancers and by definition is fraught with diagnostic and treatment challenges because the primary tumour is undefined in over 80–85% of cases (JD Hainsworth et al. J Clin. Oncol. 2013). These tumours have early dissemination, are aggressive with an unpredictable metastatic growth pattern. Around 85% of CUPs are adenocarcinomas, 10% squamous cell carcinomas and 5% undifferentiated neoplasms.

Diagnostic work up

Providing a complete diagnostic work up and tailoring treatment to this diverse, undefined tumour category remains highly complex. There is no standard staging system for CUP. The typical diagnostic work up for CUP includes a thorough physical examination, analyses of blood samples for classic tumour markers, as well as evaluation of tumour biomarkers and CT scan of the thorax, abdomen and pelvis (E Mnatsakanyan et al. Cancer Causes Control, 2014). PET-CT scan remains optional (A. Kramer et al Annals of Oncology 2022). Ultimately, histopathology, imaging and endoscopy techniques are employed to diagnose CUP. The involvement of lymph nodes and patient symptoms are important poor prognostic indicators. Tests can be done to find where the primary cancer began and to get information about where the cancer has spread. When tests are able to find the primary cancer, the cancer is no longer considered a CUP, and treatment is based on the type of primary cancer. More recently, whole-genome sequencing can provide the correct diagnosis in 60–70% of cases (L J Schipper et al. ESMO Open 2022).

Prognostic subgroups: favourable and unfavourable

Patients with CUP are categorised into two prognostic subgroups according to clinicopathologic features: those with a “favourable” or an “unfavourable” prognosis. Around 15–20% of patients can be given a putative primary tumour diagnosis on the basis of the histological and pathological findings, and such patients typically have a more favourable prognosis, with survival beyond 1 year. Most patients with CUP, however, (80–85%) have an unfavourable prognosis, and are treated with empiric chemotherapy, with a poor survival expectation of 4 to 10 months (Pavlidis & Pentheroudakis, Lancet, 2012). The unfavourable patient subgroups tend to be resistant to chemotherapy and have poor performance status (PS) and elevated lactate dehydrogenase (LDH) levels, criteria associated more generally with metastatic disease.

In 2002, the French Study Group on Carcinomas of Unknown Primary, GEFCAPI, identified a prognostic model for patients within these subsets, based on performance status and serum lactate dehydrogenase (LDH) levels, later referred to as the GEFCAPI Index (S Culine et al. J. Clin. Oncol, 2003). The study further investigated which treatments are advantageous for patients with CUP entities for which no treatment was suggested by the clinicopathologic features. Patients with good PS status (<1) and normal lactate dehydrogenase level have a better outcome (life expectancy of 1 year) than those with poor PS (>2) and raised LDH level (median overall survival of 4 months). As Fizazi explains “we still had difficulties to identify where the primary was located and thus how best to treat patients. There was no standard of care for chemotherapy. The biology was not very well known and there were no targeted agents or immunotherapies”. Responsive patient subsets

In 2003, Nicholas Pavlidis (at the University of Ioannina Hospital in Greece), together with Evangelos Briasoulis (also at Ioannina) and the US researchers John Hainsworth and Anthony Greco, attempted to classify CUP patients into clinicopathological prognostic groups to ask whether it is possible to identify responsive and unresponsive patient subgroups. Using factors such as histological subsets, lymph node involvement, number of metastatic sites and serum markers, among others, the authors proposed “favourable” and “unfavourable” subsets of CUP (Pavlidis N et al. Eur J. Cancer, 2003).

As Pavlidis states: “since we were fascinated with the entity of CUP in general; we wanted to ask why this group does well and why the other group does not do well”. This study categorised favourable and unfavourable prognostic patient subsets. While the study was a joint European and US collaboration, the idea was driven by the group at Ioannina, who had the idea for developing a classification by clinicopathological prognosis. This work in 2003 led to the first recognised classification of CUP patients into clinicopathological prognosis groups, which was a collaborative Europe-US effort. This prognostic model entered the ESMO Clinical Practice Guidelines for CUP (2015), and provided the first major stride forward in diagnostic and prognostic understanding and clinical management of patients with CUP.

Favourable subgroups include women with breast adenocarcinoma with only involved axillary lymph nodes, or papillary adenocarcinoma of the peritoneum, squamous cell carcinomas involving cervical lymph nodes, poorly differentiated neuroendocrine cancers, or men with blastic bone metastases and raised PSA levels. Patients with unfavourable features included metastatic cancers of the liver, those with multiple cerebral metastases, and patients with multiple pleural or lung metastases, or with multiple metastatic bone disease.

A Canadian open registry study of patient samples collected from 2000–2005 looked at survival outcome differences based on treatments using knowledge of the primary tumour site for patients with CUP and known primary site. Patients with a known primary had considerably better outcomes (median overall survival of 11.9 months) than those with an unknown site (1.9 months). “For good prognosis groups and known primary sites, you have exactly what you want. You don't need more guidelines. We need more for the poor prognosis group to find new treatments,” explains Pavlidis (JD Hainsworth et al. J Clin. Oncol. 2013).

Prognosis is not always helpful

As Fizazi notes, “the problem with prognostic factors is that they're not super helpful. Performance status is a very easy way to assess patients, and it is easy and cheap to measure LDH levels.” He continues, “You know if your patient has a poor PS and or an elevated LDH he or she is going to die within 4 months. Now, if the patient is not one of those, his or her life expectancy is more than 1 year.” The issue is how can we improve on these classic clinicopathologic features to predict prognosis and treatment outcome?

Setting the chemotherapy standard through trial approaches

There was no standard of chemotherapy care for patients with CUP, so in 1998, a randomised study was launched (GEFCAPI) to generate some guidelines based on what was already known regarding the management of patients with CUP. GEFCAPI 01 compared chemotherapy with cisplatin and irinotecan and chemotherapy with cisplatin and gemcitabine. GEFCAPI 01 showed that the combination of cisplatin and gemcitabine had promising antitumour activity and a favourable pattern of tolerance (S Culine et al. J. Clin. Oncol, 2003). As Fizazi highlights, “what we found was that cisplatin and gemcitabine was probably at least as good in terms of balance between efficacy and safety. So, cisplatin and gemcitabine became the standard”.

Notably, the study made use of the prognostic model previously published by GEFCAPI to classify patients into “good risk” and “poor risk” prognostic groups. The study only included patients for whom no treatment implications were established based on clinical or pathological features. In 2003, GEFCAPI decided to launch two trials, based on the data collected in GEFCAPI 01. GEFCAPI 02 tested cisplatin with or without gemcitabine in patients with a favourable prognosis, and GEFCAPI 03 tested the role of chemotherapy in patients with an unfavourable prognosis. Because of a low accrual rate, the trial was closed before reaching the planned participant number.

Molecular gene-expression profiling

Molecular gene-expression profiling of tissue samples has improved the ability to detect the primary site, which can help approaches to tailoring treatment according to epigenomic and transcriptomic classifiers of tumour samples. Liquid biopsy sampling of the blood can also provide a new diagnostic avenue that might lead to improvements in the management of patients with CUP. While prospective, non-randomised studies provided promising results, this was not replicated in randomised studies. Most evidence supporting the use of biomarker-guided therapies is based on case reports. In the precision medicine era, some advances in understanding the biology of CUP have led to improved diagnosis and classification to enable a more-tailored therapeutic approach.

##Role of molecular profiling to predict outcomes

In a 2013 trial to assess the predictive value of molecular profiling to ascertain site of tumour origin and use of site-directed therapy in patients with CUP, the median survival was 12.5 months, which compared favourably with previous results using empiric systemic therapies. The results of this study concluded that molecular profiling contributes to the management of patients with CUP and should be part of their standard evaluation. One study that used a molecular profiling algorithm approach in identifying primary tumours of unknown origin for the diagnostic work up showed that this approach correctly identified the tumour type in 68% of cases (L J Schipper et al. ESMO Open 2022). (Rassy E et al Nat Rev Clin Oncol 2020, 17 (9): 541-554)

Molecular-based trials less encouraging

After the failure of GEFCAPI 02/03 (M Gross-Goupil et al. Eur. J. Cancer 2012) in accrual, in 2019 GEFCAPI started a European trial on CUP, to study precision medicine as an option for patients with CUP. A phase III randomised controlled trial (GEFCAPI 04) of patients with unfavourable CUP compared the 1-year overall survival outcomes for site-specific tumour guided treatment on the basis of gene-expression profiling or empirical carboplatin and cisplatin chemotherapy (K Fizazi et al. Ann. Oncol. 2019). As Fizazi explains: “the idea in this 04 trial was to compare standard of care with cisplatin plus gemcitabine versus to a microarray test and treat your patient according to the results. So, if the result indicates a likely lung cancer, you would treat according to guidelines for lung cancer, and so on.”

Importantly, “we did the test in both arms. In arm A we were blinded, so we used chemotherapy, we didn't know what we were treating. In arm B, based on suspected tumour type we treat accordingly.” But the good thing is that because the test was run in both arms, it was possible to compare at the end, a posteriori, how patients performed; for example, for suspected lung cancer, suspected kidney cancer, suspected colorectal cancer, etc.”

The primary end point of 1-year overall survival was 9.8 months in the site-specific guided therapy arm and 12.5 months in the empirical arm, a difference that was not statistically significant. Ultimately, molecular testing did not enable patients to do better as compared to just starting on standard chemotherapy. Subgroup analysis showed if you identify kidney cancer and treat the patients with therapy according to that cancer, they do better than using empirical chemotherapy (i.e. control arm treatment), although caution is needed in subgroup analysis as the suboptimal statistical power does not enable definitive conclusions.

Future opportunities

Ultimately it is still unclear whether molecular profiling increases the accuracy of identifying the primary site, and it does not seem to drastically improve targeted treatment decisions.

No positive phase III trial has been completed for patients with CUP. However, some of the results coming out of European studies have entered guidelines and treatment. These include the combination of cisplatin and gemcitabine for chemotherapy, the division into favourable and unfavourable prognostic groups, and prognostic factors such as serum LDH levels and PS. Further trials and studies are ongoing to assess how molecular profiling could better treat patients with poor-prognosis CUP.


Halsted W. Cancerous axillary glands with non-demonstrable cancer of the mama = Report on three cases. Annals of Surgery Vol XLVI, No 1, July 1907

Culine S et al. Cisplatin in combination with either gemcitabine or irinotecan in carcinomas of unknown primary site: results of a randomized phase II study--trial for the French Study Group on Carcinomas of Unknown Primary (GEFCAPI 01). J. Clin. Oncol. (2003) 15;21 (18): 3479–3482. doi: 10.1200/JCO.2003.12.104.

Fizazi K et al. A phase III trial of empiric chemotherapy with cisplatin and gemcitabine or systemic treatment tailored by molecular gene expression analysis in patients with carcinomas of an unknown primary (CUP) site (GEFCAPI 04). Ann. Oncol. (2019) 30 (Suppl. 5): v851–v934 doi:10.1093/annonc/mdz394

Gross-Goupil M et al. Cisplatin alone or combined with gemcitabine in carcinomas of unknown primary: results of the randomised GEFCAPI 02 trial. Eur. J. Cancer (2012) 48(5):721–727. doi: 10.1016/j.ejca.2012.01.011

Hainsworth JD et al. Molecular gene expression profiling to predict the tissue of origin and direct site-specific therapy in patients with carcinoma of unknown primary site: a prospective trial of the Sarah Cannon Research Institute. J. Clin. Oncol. (2013) 10:31(2):217–223. doi: 10.1200/JCO.2012.43.3755

Rassy E et al. Progress in refining the clinical management of cancer of unknown primary in the molecular era. Nat Rev Clin Oncol (2020), 17 (9): 541-554

Mnatsakanyan E et al. Cancer of unknown primary: time trends in incidence, United States. Cancer Causes Control (2014) 25: 747–757 doi: 10.1007/s10552-014-0378-2

Pavlidis N et al. Diagnostic and therapeutic management of cancer of an unknown primary. Eur. J. Cancer (2003) 39(14):1990–2005. doi: 10.1016/s0959-8049(03)00547-1

Pavlidis N et al. Cancer of unknown primary site. Lancet (2012) 379(9824):1428–1435. doi: 10.1016/S0140-6736( 11)61178-1.

Schipper L J. Complete genomic characterization in patients with cancer of unknown primary origin in routine diagnostics. ESMO Open (2022) 7(6):100611. doi: 10.1016/j.esmoop.2022.100611


First report on 3 patients with axillary lymph nodal metastases (adenocarcinoma) without detectable primary site (by W Halsted)


Combination chemotherapy regimens failed to improve overall survival in the majority of CUP patients


CUP is divided into unfavourable (80%) group with 2-10 months survival and favourable (20%) group with a median survival of more than 2 years


During the last 10 years clinical trials with site specific treatment (targeted therapy) are under investigation