Primary Cancer Mysteries Persist Even in the Genomic Era

Kanwal P.S. Raghav, MBBS, MD

Kanwal P.S. Raghav, MBBS, MD

Although diagnostic techniques have improved over the past several decades, cancers of unknown primary (CUP) origin are a lingering feature of clinical practice, frustrating attempts to develop new protocols through molecularly driven strategies.

Investigators are using gene expression profiling (GEP) to identify the primary tumor and help guide treatment decisions, and are taking advantage of next-generation sequencing (NGS) to find genomic alterations for which targeted therapies are available.

In a number of studies, GEP tests, also known as molecular cancer classifier assays, have been shown to identify primary cancer types with greater accuracy than immunohistochemistry (IHC) staining. In some cases, these results allow use of site-specific therapies for patients with CUP rather than less efficacious empiric chemotherapy.

However, with evidence lacking for improved survival after GEP, the assays are not recommended or widely used. Investigators say more data are needed from large multicenter trials to delineate CUP subtypes and gauge the impact of molecular profiling and site-specific therapies on patient outcomes.

This means that CUP, at least for now, remains a clinical mystery that clinicians must unravel for each patient.

“At the end of the day, for cancers of unknown primary, you have to just put all the clues together and make a reasonable putative diagnosis,” Kanwal P.S. Raghav, MBBS, MD, said in an interview with OncologyLive^®. Raghav is an assistant professor in the Department of Gastrointestinal (GI) Medical Oncology at The University of Texas MD Anderson Cancer Center in Houston.

“It’s probably never going to be the most accurate diagnosis because there’s no truth to which you can hold that diagnosis, but you have to make a reasonable primary diagnosis and then treat it as close to that primary as possible with a site-specific therapy,” Raghav said. “If, despite your best efforts, you cannot figure out where the cancer is coming from, then, of course, the option is to do empiric chemotherapy.”

Investigators estimate that in the United States, CUP will account for approximately 2% of all cancers, or about 31,500 cases, that will be diagnosed in 2019. Overall, its incidence has been estimated at about 3% to 5% of all tumors, placing it among the 10 most frequently diagnosed tumors in developed countries. Although the percentage of cancers diagnosed as CUP has decreased over the years, these malignancies remain difficult to treat, and prognoses are generally poor.¹

Most patients with a diagnosis of CUP present with metastatic disease, often in the liver, lungs, bones, and lymph nodes.1 Most of the primary tumors found through autopsies are less than 1 cm, and investigators have not been able to determine the mechanisms that caused the cancer to spread from such a small site.²

In a comprehensive review of results from autopsy studies, investigators determined that postmortem examinations of patients whose primary tumor site was never identif ied succeeded in establishing the anatomic origin in 73% of 844 cases.³ A primary site was most frequently identified in the lung (27%), pancreas (24%), hepatobiliary system (8%), and kidneys (8%), with other sites found in the bowel, the genital system and the stomach (Figure).³

Figure. A Picture of CUP From Autopsy Findings³

Notably, however, the same investigators found that a review of results from more than 500 CUP samples using molecular signatures that analyzed 10 to 495 genes reported primary site frequencies that were “strikingly dissimilar” from those identified in the autopsy results. They concluded that the patterns of primary tumors were consistent in the autopsies conducted and that molecular profiling may identify a metastatic genetic signature that “transcends tissue-of-origin distinctions.”³

Pros and Cons of Profiling

In practice, molecular analyses have not yet provided definitive guidance. “Recent suggestions that optimum therapy for CUP can be administered using only the results of comprehensive molecular profiling (ie, knowledge of the primary site is irrelevant) are premature,” investigators from the Sarah Cannon Research Institute in Nashville, Tennessee, noted in an article in the 2018 American Society of Clinical Oncology Educational Book.²

Raghav said his clinic rarely uses GEP in patients with CUP and even then, for the most part, only as part of research. He and his colleagues instead depend on traditional work-ups tailored to each patient, often including IHC, that aim to narrow the range of likely disease origin sites.

“Cancer of unknown primary provides a challenging diagnosis and management conundrum because you have to put a lot of clues together and recognize various patterns of disease presentation as well as pathology and imaging to come to a suitable putative primary according to which you give treatment. It requires you to do a whole lot of searching within the clinical history and presentation of the patient,” Raghav said.

The largest study of site-specific therapies for CUP based on GEP used a 92-gene reverse transcriptase-polymerase chain reaction (RT-PCR) cancer classification assay.⁴ Conducted at Sarah Cannon Oncology Research Consortium sites, the prospective, nonrandomized phase II study (NCT00737243) successfully assayed samples from 252 of 289 patients and predicted tissue of origin in 247 (98%).

In the 194 patients who received assay-directed, site-specific treatments, the median overall survival (OS) was 12.5 months (95% CI, 9.1-15.4 months). Median OS was longer in patients predicted to have more responsive tumor types than in those with tumors predicted to be more resistant (13.4 vs 7.6 months; P = .04).⁴

Those findings compare favorably with those of empiric treatment studies that have produced median survival times of less than 10 months. However, the National Comprehensive Cancer Network panel on CUP concluded that the clinical benefit from use of molecular assays remains undetermined.

The panel noted the lack of robust outcomes data and said the results in the Sarah Cannon trial and another smaller study were similar to what might be expected from empiric therapy in patients with good performance status and CUP predominantly below the diaphragm. Neither IHC nor GEP should be used “indiscriminately,” the group concluded.¹

The panel’s skepticism is understandable given the absence of a large randomized trial, said John D. Hainsworth, MD, co-founder and principal investigator for the Sarah Cannon Research Institute and principal investigator for the CUP study. But Hainsworth, who is also a medical oncologist at Tennessee Oncology, argued that GEP is clearly useful for some patients with CUP and said many oncologists are incorporating the assays into their treatments.

“We’ve been doing this routinely in our group because we’ve had interest in this for a long time and we see a lot of these patients as referrals. We have a large number of patients we’ve treated, hundreds, Undetermined neoplasm 10% Squamous ca 23% Undifferentiated Ca Other Histology N = 884 54% Adenocarcinoma 27% Not identified and are completely convinced that this is a useful part of treatment,” Hainsworth said in an interview.

Often GEP assays still predict primary tumor types with poor prognoses, such as biliary tract and pancreatic cancers, but in some cases the predictions open up possibilities for further testing and targeted therapy, Hainsworth said.

“It would be nice, if you were the patient, if you knew you actually had non—small cell lung cancer. It would be even better if you then did molecular profiling and found you had an ALK rearrangement or an EGFR mutation or high TMB [tumor mutational burden]. There are a lot of things that you can do once you accept the diagnosis of lung cancer,” Hainsworth said.

Japanese investigators recently published results from a randomized trial (UMIN000001919) that showed that GEP-directed site-specific therapy did not improve survival, but Hainsworth, Raghav, and others5 have described the study as flawed.

The phase II multicenter trial compared outcomes from site-specific therapy versus empiric therapy with carboplatin and paclitaxel (PC), randomly assigning 130 patients. Molecular profiling by microarray analysis was performed for 101 evaluable previously untreated patients. The study found the 1-year survival rate was lower for site-specific treatment at 44.0% compared with 54.9% (P = .264) for the empiric therapy with PC.⁶

Median OS and progression-free survival (PFS) were 9.8 months and 5.1 months, respectively, for site-specific treatment versus 12.5 months and 4.8 months for empirical PC (P = .896 and .550, respectively). Median OS was better for tumor types predicted to be more responsive compared with those considered less responsive (16.7 vs 10.6 months, respectively; P = .116), as was PFS (5.5 vs 3.9 months; P = .018).

Hainsworth and Raghav said the trial may have relied on faulty pathology because 20% of the initial patient cohort (26 of 130) were predicted to have lymphoma, which they said should never be mistaken for CUP. Hainsworth also noted high rates of patients with hard-to-treat cancers, including 21% each with pancreatic and gastric disease. At the same time, the results showed that GEP can identify patients with more responsive primary tumors who would benefit from site-specific therapies, Hainsworth said.

At least 2 other phase III randomized trials using GEP assays for patients with CUP are under way internationally. The European GEFCAPI04 trial (NCT01540058), which seeks to enroll 223 patients, is comparing PFS and OS following test-guided therapy versus empiric strategy of gemcitabine and cisplatin. The study is using an RT-PCR assay to predict primary tumor type. A Chinese trial (NCT03278600) is using tissue-of-origin profiling via a 90-gene assay and aims to enroll 172 participants.⁶

Numerous studies have sought to establish the accuracy of GEP assays in identifying primary tumors. In tests using tumors of known origin, GEP assays identify the site of origin in 85% or more of cases, a higher rate than with IHC staining, particularly for poorly differentiated carcinomas and when a first round of IHC testing is inconclusive.^7-9

A meta-analysis of 7 studies with 549 patients found that molecular tumor profiling tests have a pooled sensitivity of 89% (95% CI, 0.85-0.92) and a pooled specificity of 74% (95% CI, 0.67-0.80).¹⁰ A recent study that used NGS data and a machine learning approach developed by Caris Life Sciences to construct a tumor lineage classification system was able to identify a tumor type in 93.7% of nearly 6000 CUP cases analyzed.¹¹

Those findings come with caveats, however. Although molecular profiling may reach definitive conclusions more often than IHC, confirming the assays’ accuracy when no primary tumor can be found remains difficult. “Gene expression profiling is an objective way of getting another clue, but I’m not quite sure I can say this is the word of God. We found many times the gene expression profiling comes back with results that are completely contrary to even the pathology,” Raghav said.

Another challenge is heterogeneity within known tumor types. Raghav said the established molecular profiles to which specimens are matched reflect the most common types of known cancers; in the case of colon cancer, for example, that type is consensus molecular subtype (CMS) 2.

However, it is possible that CUP tumors that are metastases from colon cancer derive more often from another type. “If cancers of unknown primary reflect the biology of CMS4, every time you get cancer of unknown primary, which could have been CMS4, you’re not going to call it colon cancer because your training set is made of the most common kind of colon cancer,” Raghav said.

Questions About Tumor Biology

One of the unanswered questions about CUP malignancies is whether they are all versions of known tumor types or whether some have changed so substantively that they are effectively a different cancer requiring different therapies.

Another possibility is that some CUPs are derived from unknown types of cancer, such as tumors that metastasize when very small and then disappear. With GEP, “the big assumption you’re making is that all cancers of unknown primaries have the same gene expression signature as metastases from a known primary; that is, their biology is exactly the same. I’m not quite sure research has even definitively concluded that,” Raghav said.

Rather than relying on identification of the primary tumor, a different experimental approach seeks to use comprehensive genomic profiling of CUP tumors to find clinically actionable mutations in the metastases.

At Memorial Sloan Kettering Cancer Center, 333 patients with CUP were evaluated, and archival tissue from 150 was sequenced using the MSK-IMPACT assay.¹² Ninety-one percent (137 of 150) had at least 1 alteration with a known or likely oncogenic function, with an average of 3.1 oncogenic alterations per tumor (range, 0—9).

Thirty percent (45 of 150) had potentially targetable alterations. Investigators identified 27 level 2 alterations in 23 tumors, the most common of which were ERBB2 amplification and BRAF V600E mutations; 27 level 3 alterations were identified in 25 tumors, the most common of which were PIK3CA mutations.

Fifteen of the patients received targeted therapies. Time to treatment failure ranged from less than 1 month to 14 months, and several patients remained on targeted therapy at the time of data cut-off.

Smaller studies have found similarly high rates of genomic alterations. MD Anderson Cancer Center investigators tested 17 patients and found alterations in 15 (88%) (median, 2 aberrations; range, 0—8; total, 59 alterations).¹³ Nine patients had altered cell signaling, 8 had impaired epigenetic regulation and DNA methylation, 8 had aberrant cell cycle regulation, and 7 had alterations in tumor suppressor genes. Ten (59%) patients had alterations in transcriptional regulators. Eleven patients participated in a phase I clinical trial and were treated with a targeted therapy, including 7 who received a therapy matching their mutational profile. Four of the 11 had stable disease lasting 4 or more months, including 3 of those patients whose treatments matched their profile.

As part of the ongoing University of Michigan Oncology Sequencing Program, 32 patients with CUP underwent biopsies for comprehensive NGS analysis, including tumor and normal whole-exome, whole-genome, and transcriptome analysis.¹⁴

Investigators also identified pathogenic germline variants (PGVs) conferring increased cancer risk. Of those patients, 28 (87.5%) had at least 1 clinically actionable alteration (average, 1.8 alterations; range, 0-5); treatment changed for 12 of those (43.0%) as a result.

Five of the 32 (16%) had a PGV identified. Of the patients whose treatment was informed by NGS, 2 had “exceptional” responses and remained on treatment for over 1 year. One of them, with metastatic cholangiocarcinoma and an FGFR2—CCDC6 gene fusion, received an FGFR inhibitor in a phase I trial, whereas the other had a PGV in MSH2 and received pembrolizumab.

Laurence H. Baker, DO, who founded the CUP clinic at the University of Michigan Medical School in Ann Arbor, noted the importance of testing for PGVs, which he said is rarely done in comprehensive genomic profiling assays. “If you don’t do that test, then you’re going to miss 16% of the diagnoses,” said Baker, a professor of internal medicine and pharmacology at Michigan Medicine.

Another general benefit of comprehensive genomic testing is that it provides new information to patients with CUP, who may be distressed by their diagnosis or may not understand why it is incomplete. “Patients deserve to have some understanding of where their cancer arose, even if that doesn’t necessarily seem logical to some. Most patients get psychological relief by knowing where it started,” Baker said.

Future Considerations

The findings from genomic testing studies have raised questions about the use of empiric regimens for CUP. Some investigators suggest that NGS testing should be studied as a frontline alternative to time-consuming laboratory and imaging studies that often do not lead to improved outcomes, given the greater benefit from targeted therapies compared with standard cytotoxic chemotherapy.¹⁵ Yet others note that most patients do not have useful targets; Hainsworth estimated the percentage of patients with CUP with druggable oncogenes at no more than 30%.

Tumor-agnostic therapies may also prove useful for treating CUP, although more research is needed. Significant numbers of patients with different types of CUP have been found to have high TMB, including 7.6% of patients with adenocarcinoma of unknown primary, 10.8% of those with carcinoma of unknown primary, and 20.6% of patients with squamous cell carcinoma of unknown primary.¹⁶ Investigators are studying high TMB as a possible predictor of positive response to immunotherapy in some cancers.

A specimen review of tumors of unknown primary type using FoundationOne sequencing found a 2.4% rate of high microsatellite instability,17 which can be treated with the PD-1 inhibitor pembrolizumab (Keytruda) regardless of tumor type. Two British phase II trials are studying the efficacy of molecularly guided therapy for CUP. The CUPISCO trial is comparing targeted therapy or immunotherapy with chemotherapy and aims to enroll 790 patients (NCT03498521). The CUPem trial is testing pembrolizumab in CUP and intends to enroll 77 patients (NCT03752333).

Experts said that understanding which approaches work best for which patients will require accumulating much more data and conducting large randomized trials. Obstacles include the rarity of patients with CUP at individual clinics, which makes trial accrual difficult, and only moderate interest from pharmaceutical manufacturers who might fund research, they said. Raghav and Baker said nationwide, or international multicenter trials are needed, as are collaborative data collection efforts.

“We need large-scale studies to define the true subsets of CUP. For example, our group worked on defining a CDX2-positive, CK20-positive CUP subset, which we described as a bowel-profile subset. The treatment is a 5FU [fluorouracil]—based regimen,” said Raghav. “In a retrospective series we have shown the outcomes are actually pretty good, similar to what you would expect for colon or small-bowel cancers.

“We need to actually do studies that can establish whether that form of therapy is better than giving somebody empiric carboplatin/Taxol or gemcitabine/cisplatin. It’s important, especially in an era [when] treatments are improving and targeted treatments and immunotherapy are coming into the picture,” Raghav said. “We need to be able to do dedicated studies accounting for the heterogeneity that we find in CUP.”

References

1. NCCN Clinical Practice Guidelines in Oncology. Occult Primary, version 2.2019. National Comprehensive Cancer Network website. nccn.org/professionals/physician_gls/pdf/occult.pdf. Published January 23, 2019. Accessed June 10, 2019.
2. Hainsworth JD, Greco FA. Cancer of unknown primary site: new treatment paradigms in the era of precision medicine. Am Soc Clin Oncol Educ Book. 2018;38:20-25. doi: 10.1200/EDBK_100014.
3. Pentheroudakis G, Golfinopoulos V, Pavlidis N. Switching benchmarks in cancer of unknown primary: from autopsy to microarray. Eur J Cancer. 2007;43(14):2026-2036. doi: 10.1016/j.ejca.2007.06.023.
4. Hainsworth JD, Rubin MS, Spigel DR, 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;31(2):217-223. doi: 10.1200/JCO.2012.43.3755.
5. Wapner J. Phase 2 trial examines gene-expression profiling for cancer of unknown primary site. Cancer Therapy Advisor website. cancertherapyadvisor.com/home/cancer-topics/general-oncology/cancer-unknown-primary-site-gene-expression-phase-2-trial/. Published April 5, 2019. Accessed June 10, 2019.
6. Hayashi H, Kurata T, Takiguchi Y, et al. Randomized phase II trial comparing site-specific treatment based on gene expression profiling with carboplatin and paclitaxel for patients with cancer of unknown primary site. J Clin Oncol. 2019;37(7):570-579. doi: 10.1200/JCO.18.00771.
7. Hainsworth JD, Greco FA. Gene expression profiling in patients with carcinoma of unknown primary site: from translational research to standard of care. Virchows Arch. 2014;464(4):393-402. doi: 10.1007/s00428-014-1545-2.
8. Handorf CR, Kulkarni A, Grenert JP, et al. A multicenter study directly comparing the diagnostic accuracy of gene expression profiling and immunohistochemistry for primary site identification in metastatic tumors. Am J Surg Pathol. 2013;37(7):1067-1075. doi: 10.1097/PAS.0b013e31828309c4.
9. Meiri E, Mueller WC, Rosenwald S, et al. A second-generation microRNA-based assay for diagnosing tumor tissue origin. Oncologist. 2012;17(6):801-812. doi: 10.1634/theoncologist.2011-0466.
10. Inocencio RT, Kesari K, Smith S. Diagnostic accuracy of molecular tumor profiling in determining the tissue of origin in adult patients with cancer of unknown primary: a meta-analysis. J Clin Oncol. 2017;35(suppl 7; abstr e23092). meetinglibrary.asco.org/record/146419/abstract.
11. Abraham J, Heimberger AB, Marshall J, et al. Machine learning algorithm analysis using 55,780 cases from a commercial 592-gene NGS panel to accurately predict tumor type for carcinoma of unknown primary (CUP). Poster presented at: American Society of Clinical Oncology Annual Meeting; May 31-June 4, 2019; Chicago, IL. Abstract 3083. meetinglibrary.asco.org/record/175034/abstract.
12. Varghese AM, Arora A, Capanu M, et al. Clinical and molecular characterization of patients with cancer of unknown primary in the modern era. Ann Oncol. 2017;28(12):3015-3021. doi: 10.1093/annonc/mdx545.
13. Subbiah IM, Tsimberidou A, Subbiah V, Janku F, Roy-Chowdhuri S, Hong DS. Next generation sequencing of carcinoma of unknown primary reveals novel combinatorial strategies in a heterogeneous mutational landscape. Oncoscience. 2017;4(5-6):47-56. doi: 10.18632/oncoscience.352.
14. Yentz S, Bhave M, Cobain E, Baker L. Cancer of unknown primary. In: DeVita, Hellman, and Rosenberg’s Cancer Principles & Practice of Oncology. 11th ed. Philadelphia, PA: Wolters Kluwer; 2019:chap 108.
15. Ross JS, Wang K, Gay L, et al. Comprehensive genomic profiling of carcinoma of unknown primary site: new routes to targeted therapies. JAMA Oncol. 2015;1(1):40-49. doi: 10.1001/jamaoncol.2014.216.
16. Gay LM, Fabrizio D, Frampton GM, et al. Mutational burden of tumors with primary site unknown. J Clin Oncol. 2017;35(suppl; abstr 3039). meetinglibrary.asco.org/record/145086/abstract. Poster presented at: American Society of Clinical Oncology Annual Meeting; May 31-June 4, 2019; Chicago, IL. Abstract 3039. meetinglibrary.asco.org/record/145086/abstract.
17. Agarwala V, Gossai A, Singal G, et al. Use of cancer immunotherapies in the real-world in the setting of microsatellite instability. J Clin Oncol. 2018;36(suppl 5S; abstr 30). meetinglibrary.asco.org/record/156619/abstract.