Op dit moment wordt niet voor alle PTO-patiënten uitgebreide immunohistochemische en moleculaire analyses verricht voor doelgerichte behandeling. De praktijkvariatie is vrij groot: soms worden nauwelijks analyses ingezet voor doelgerichte therapie, terwijl voor andere patiënten grote Next Generation Sequencing (NGS) panels of zelfs Whole Genome Sequencing (WGS) wordt ingezet, vaak in combinatie met enkele immunohistochemische bepalingen. Er is behoefte aan meer duidelijkheid over wanneer welke immunohistochemische en moleculaire testen het beste kunnen worden ingezet met betrekking tot het identificeren van biomarkers/targets voor doelgerichte behandeling en immuunchekpointremmers.

Immunohistochemistry (IHC)

Comprehensive genomic profiling

1. Immunohistochemistry (IHC)

International guidelines

The NCCN Guideline ‘Occult primary’ (2021) describes microsatellite instability (MSI) /mismatch repair (MMR) testing to be indicated for patients with CUP, although the incidence of this feature within this patient group is low (1.8%). Both immunohistochemistry or polymerase chain reaction (PCR) can be used to determine this feature. TRK protein testing by immunohistochemistry is stated to be performed by per physician’s choice. 

Furthermore, the NCCN guideline reports that IHC analyses (such as RAS, HER2 and ALK) can be used to test for tumor biomarkers that might inform treatment decisions.

The ESMO guideline ‘Cancer of unknown primary: ESMO Clinical Practice Guideline for diagnosis, treatment and follow-up’ (2023) states that determination of MSI and PD-L1 is generally recommended when immune checkpoint inhibition (ICI) is being considered. Especially, since MSI is considered to be a predictive biomarker for response to ICI in a tissue-agnostic manner.

Similarly, NTRK-inhibitors have tumor-agnostic approval and therefore testing the presence of NTRK-fusion (most of the time starting with IHC) is recommended.

Additional genetic alterations (potentially found by IHC) that are targeted by compounds licensed in non-CUP entities are described to be an option for patients with CUP harboring these alterations. However, the guideline does not state if or which alterations should be tested.

Description of studies

Mauri (2018) performed a cohort study of Tropomyosin-related kinase A (TRKA) expression, encoded by the NTRK1 gene in various solid tumours. This study was performed at a single cancer center in Italy, within the context of a phase I clinical trial (the AKA-001 trial) evaluating entrectinib. Patients were eligible for this trial if they had: (1) a histologically or cytologically confirmed diagnosis of relapsed or refractory advanced/metastatic solid tumour for which no alternative effective standard therapy was available or for which standard therapy was considered unsuitable or intolerable; (2) a performance status ≤2; (3) a life expectancy of ≥3 months; and (4) adequate organ function. The cohort included a total of 1043 samples, of which 49 samples were from patients with a cancer of unknown primary. No definition of ‘cancer of unknown primary’ was provided.

FFPE samples were tested for TRKA expression using IHC staining. Samples showing weak, moderate or strong IHC staining in at least 10% of cells were analysed by fluorescence in situ hybridization to assess whether NTRK1 gene alterations were present.

Results

In the study by Mauri (2018), out of 49 samples of CUP patients analysed, 1 sample (6%) was characterized by TRKA IHC expression. One sample (2%) displayed NTRK1 copy number gain (3 to 9 copies).

Level of evidence of the literature

The study by Mauri (2018) does not meet the criteria for a GRADE assessment.

2. Comprehensive genomic profiling 

International guidelines

The NCCN Guideline (2021) states that TMB should be determined by a validated and/or FDA approved assay and labels this as a category 2B recommendation. The guideline considers next generation sequencing (NGS) after an initial determination of histology if it would guide therapeutic decision-making. 

The ESMO guideline (2023) states that determination of tumor mutational burden (TMB) is generally recommended, as the presence of high TMB is considered to be predictive for response to ICI in a tissue-agnostic manner. Likewise, BRAF V600E and the RET proto-oncogene are mentioned as cancer type-agnostic targets. However, the guideline does not mention testing these features as mandatory. 

Similarly, NTRK-inhibitors have tumor-agnostic approval and therefore testing the presence of NTRK-fusion is recommended.

As mentioned above, identifying additional genetic alterations could lead to compounds targeting these alterations. However, the guideline does not state if or which alterations should be tested.

The more general ESMO guideline on NGS for metastatic cancer (Mosele 2020) states that while there is no level I evidence multigene sequencing (‘large gene panels’) could be used in patients with CUP. 

Description

Lombardo (2020) performed a systematic review of patients with CUP who underwent next generation sequencing (NGS) to evaluate the frequency and type of genomic alterations for which targeted treatment is available. A systematic search was performed in June 2019 in PubMed, the ASCO meeting library and clincaltrial.gov. English-language publications were eligible if they described a study performed in patients with cancer of unknown primary who underwent NGS. The review also included case reports and abstracts about patients with cancer of unknown primary treated with targeted treatment based on the results of NGS, but these studies are beyond the scope of this literature analysis. Eleven publications (9 full-text publications and two abstracts) were related to genomic alterations. The number of patients included in these studies ranged between 16 and 1806, for a total of 3161 patients. The number of genes included in the panel ranged between 47 genes and 701 genes.

Yang (2022) performed a retrospective study of 35 patients with cancer of unknown primary from multiple hospitals across China. CUP was defined as follows: ‘a histologically proven metastatic malignant tumour whose primary site cannot be identified during pretreatment evaluation. Patient follow-up has also been performed to confirm that the primary tumours remained unknown until the time of this study’. It is not clear from this definition which initial diagnostic tests were performed. Next generation sequencing was performed with a target gene panel of 416 genes on peripheral blood, FFPE tumour tissue or fresh tumour tissues.

Kang (2021) performed a retrospective study to assess the efficacy of NGS and targeted therapy in 218 patients with CUP from the Asan Medical Center in Seoul, South Korea. These patients had undergone medical history taking, physical examination, baseline blood and biochemistry analyses, imaging studies including chest and abdomen-pelvis computed tomography (CT), and biopsy. Furthermore, 206 patients (94%) had also undergone (PET)/CT as part of the diagnostic work-up. Out of 218 patients, 22 (10%) underwent NGS. Tissues used for NGS includes lymph nodes (59%), bone (18%), and lung (9%). The results reported included level 1, 2,3 and 4 alterations (level of clinical actionability according to OncoKB) and the number of patients that received targeted therapy based on the NGS results.

Hayashi (2020) performed a multicentre phase II trial including 111 patients with cancer of unknown primary from multiple hospitals in Japan. CUP was defined as ‘histologically confirmed metastatic cancer for which identification of the primary site is not possible after an appropriate diagnostic approach’, after a standard evaluation (medical history, physical examination, blood cell counts, chemistry profile, chest-abdomen-pelvis computed tomography scans, positron emission tomography scan, and directed assessment of all symptomatic areas. Next generation sequencing (NGS) was performed using FFPE tumour tissue and included analysis of 257 genes.

Bochtler (2020) performed a cohort (partly prospective, partly retrospective) study of 252 patients with cancer of unknown primary in Germany. CUP was defined according to the ESMO guideline. Targeted next-generation sequencing was performed using FFPE. 104 genes were included in the analysis. Outcomes included genetic alterations and oncogenic activations.

Bochtler (2022) performed a prospective cohort study of 74 patients with CUP in Heidelberg, Germany. The diagnosis of CUP was made according to the European Society of Medical Oncology (ESMO) guidelines. Chromosomal aberrations and chromothripsis were detected by methylation-based copy number variation (CNV) analysis, whereas TMB and MSI were calculated based on large next-generation sequencing (NGS) panels. OS and DFS rates were assessed.

Posner (2023) performed a prospective comparative study of the diagnostic utility of RNA and DNA tests in 183 patients with cancer of unknown primary in Australia. Patients were eligible if: (1) they presented with carcinoma of no confirmed primary site and had undergone a preliminary diagnostic workup including, but not limited to, a detailed clinical assessment, a CT scan of the chest, abdomen and pelvis, pathological review of tumour tissue, and other gender-appropriate diagnostic tests; and (2) they had not started treatment yet or had started treatment within six months before recruitment. Patients were excluded if they were under 18 years of age or had an ECOG performance status ³ 3. The DNA test involved a comprehensive DNA panel sequencing of 386 cancer-related genes. DNA sequencing was performed on matched blood and tumour DNA.

Möhrmann (2022) performed a cohort study of 70 patients with cancer of unknown primary characterized by comprehensive molecular profiling in Germany. Most (61/70; 87%) patients met the criteria defined by the ESMO clinical practice guideline for CUP. To determine the molecular landscape whole exome sequencing or whole genome sequencing was performed for all patients and RNA analysis in 79% of the patients

Es (2021) performed a retrospective study using the AACR Project Genomics Evidence Neoplasia Information Exchange (GENIE) database including 1709 CUP samples to identify druggable targets by genomic mutation and copy number alteration analysis.

Results

In the review by Lombardo (2020), the mean number of patients with a potentially targetable genomic alteration was 43.7%. The most frequent targetable alterations included TP53 (mean 42%; matched drug APR-246), KRAS (19%; matched drugs MEK-i, KRAS G12C), CDKN2A (9%; CDK4/6-I (e.g., abemaciclib), and PIK3CA (9%; matched drugs alpelisib, AKT-i, mTOR-i).

In the retrospective study by Yang (2022), 37 patients were enrolled and two patients were excluded because of lack of evaluable tumour tissue and plasma, and due to use of a different sequencing panel. Targetable mutations (according to the OncoKB database) were found in 8 out of 35 patients (23%) and included KRAS G12C (n=3; 9%), PIK3CA E545K (n=2; 6%), PIK3CA E542K (n=1; 3%), BRAF V600L (n=1; 3%), EML4-ALK (n=1; 3%), and ALK L1196M (n=1; 3%).The median tumor mutational burden (TMB) was 12.5 mutations/Mb. For all 26 patients, microsatellite instability (MSI) analysis showed that all tissue specimens were microsatellite stable.

In the retrospective study by Kang (2021), 10 out of 22 patients showed level 1, 2 or 3 alterations, which is the level of clinical actionability according to OncoKB (table 2). Targeted therapy was offered to two out of 22 patients (9%) who underwent NGS. This included one patient with an NTRK fusion who received entrectinib and one patient who had an AKT2 activating mutation and received ipatasertib.

In the nonrandomized phase II study by Hayashi (2020), of the 111 enrolled patients, fourteen patients (12.6%) did not receive the trial therapy, therefore the efficacy analysis was performed for 97 patients. d. Genetic alterations were evaluated by targeted sequencing in all 97 patients in the efficacy analysis. The most common genetic alterations were in TP53 (45 [46.4%]), KRAS (19 [19.6%]), and CDKN2A (18 [18.6%]). Several genetic driver mutations with implications for treatment selection were also detected, including those affecting EGFR and KRAS.

In the cohort study by Bochtler (2020) 28/198 (14.1%) patients displayed at least one targetable mutation, including BRAF (n = 7), BRCA1/2 (n = 7), ATM (n = 7), IDH1 (n = 2) and PALB2 (n = 1), ALK-, RET and NUTM1 translocations (n = 1 each) as well as MET

amplification (n = 1).

In the prospective cohort study by Bochtler (2022), 11/74 (14.9%) patients were found to fall into the TMB-high category (>12 mutations/Mb). Two CUP tumors were found to be MSI (2/74)..

In the study by Posner (2023), the median tumour mutation burden (TMB) was 4.4 mutations/Mb (range 0.5-149 mutations/Mb). The most frequently mutated genes were TP53 (55%), LRP1B (20%), PIK3CA (17%), KMT2D (15%), KRAS (12%), ARID1A (11%), and SMARCA4 (11%). Out of 215 patients, 86 patients with cancer of unknown primary (40%) had actionable mutations, as defined by the CUPISCO trial criteria.

In the study by Möhrmann (2022), 3 out of 70 samples showed a high TMB (>10 mutations per megabase), all samples that could be tested for microsatellite instability were microsattelite stable. With RNA sequencing in 17 patients relevant fusions were found: FGFR2 (6 cases), EML4-ALK (3 cases), BRAF (3 cases) and EWSR1-WT1 (2 cases). Genomics-guided treatment could be provided in 20 out of 70 patients (29%).

Es (2021) reported that in 1709 CUP samples, 52 significant mutated genes were identified, of which 13 (25%) were potentially targetable with approved drugs or drugs evaluated in clinical trials..

Level of evidence

Non of above-mentioned studies meet the GRADE criteria.

A systematic review of the literature was performed to answer the following questions:

Question 1: immunohistochemical analysis

What is the diagnostic yield, in terms of finding a target for treatment, of immunohistochemical analysis (MMR, NTRK, HER2) compared with no immunohistochemical analysis in patients with provisional or confirmed cancer of unknown primary?

Question 2: comprehensive genomic profiling

What is the diagnostic yield, in terms of findings a target for treatment, of comprehensive genomic profiling compared with less comprehensive genomic profiling or no genomic profiling?

Relevant outcome measures

The guideline development group considered identification of targets for treatment as a critical outcome measure for decision making. A priori, the working group did not define the outcome measures listed above but used the definitions used in the studies. The working group defined no minimal cut-of as clinically (patient) important difference since finding a treatment target is of such importance for this patient group in whom promising treatment options are missing.

Search and select (Methods)

The databases Medline (via OVID and Embase (via Embase.com) were searched with relevant search terms until 1 February 2023. The detailed search strategy is depicted under the tab Methods. The systematic literature search resulted in 1180 hits. Studies were selected based on the following criteria: 1) full text publication in English or Dutch, published between 2018-2023; (2) population and intervention according to PICO 1 and/or 2.

26 studies were initially selected based on title and abstract screening. After reading the full text, 15 studies were excluded (see the table with reasons for exclusion under the tab Methods), and 11 studies were included. For question 1, about the diagnostic yield of immunohistochemical analysis, one relevant study (Mauri, 2018) was included.  For question 2, about the diagnostic yield of comprehensive genomic profiling, a systematic review by Lombardo (2020) was found, as well as nine original studies that were published after the search by Lombardo (2020).

Results

Eleven studies were included in the analysis of the literature. For question 1, the study of Mauri (2018) was summarized. For question 2, the review of Lombardo (2020) was included as well as nine original studies published after the search by Lombardo). Important study characteristics and results are summarized in the evidence tables. The assessment of the risk of bias is summarized in the risk of bias tables.