Introduction
Of the 1.8 million people in the US diagnosed with cutaneous squamous cell carcinoma (cSCC) each year [
1‐
3], approximately 5–8% develop regional or distant metastasis, with annual mortality estimated to be higher than that for melanoma [
4‐
6]. Consistent with cancer management in general, treatment pathways for cSCC are based upon population-based estimates of risk. Available risk classification systems for cSCC based on clinicopathologic features include National Comprehensive Cancer Network (NCCN) risk groups, Brigham and Women’s Hospital (BWH) staging and American Joint Committee on Cancer Staging Manual, 8th edition (AJCC8) [
7‐
9]. BWH and AJCC8 use tumor characteristics to classify a patient’s risk of disease progression by T-stage, while NCCN includes both patient-specific and tumor factors to stratify risk to offer treatment options for prevention of local recurrence, metastasis or disease specific death. Unfortunately, several of these risk factors are subjective or difficult to measure (i.e., rapidly growing, clinical extent of tumor, neurologic symptoms) or have documented intra- and inter-rater histopathologic variability (grade of differentiation [
10] and depth of invasion [
11,
12]) equating to limited accuracy in determining metastatic risk [
13,
14].
Gene expression profile (GEP) tests have been shown to improve risk stratification accuracy beyond clinicopathologic based staging systems for many types of cancer, including breast and prostate cancer, and cutaneous and uveal melanoma [
15‐
18]. As there is currently no prognostic molecular signature for cSCC, the 40-gene expression profile (40-GEP) test was developed and validated to statistically and independently stratify metastatic risk for cSCC patients to inform treatment pathway decisions. The 40-GEP test uses predictive modeling to categorize patients diagnosed with primary cSCC having one or more high-risk factors (i.e., high-risk cSCC; the intended clinical use group comprised of an estimated 10–15% of all cSCC patients) into three risk categories with increasing metastatic risk, Class 1, Class 2A and Class 2B, with previously reported overall metastatic event rates of < 7%, 20% and > 50%, respectively [
13,
14]. The intended use population is comprised of NCCN high or very high risk patients. Using an extended validation cohort, Ibrahim et al. [
14], demonstrated the ability of the 40-GEP test to predict regional or distant metastasis in a cohort of 420 high-risk cSCC patients. Furthermore, 40-GEP class was an independent and significant predictor of metastasis when included in multivariate models adjusted for individual clinicopathologic risk factors or risk classification systems. Robust validation of the 40-GEP has led to clinical use of the test [
19,
20] to improve precision risk stratification, resulting in reduction of potential overtreatment of biologically low-risk patients and undertreatment of patients with aggressive tumor biology. Algorithms have been developed that detail integration of test results into current treatment approaches [
21].
Multiple studies have shown that clinicians use 40-GEP test results to improve treatment pathway decisions by incorporating tumor biology into risk classification systems so that patients move from population-based treatment pathway decisions to individual risk-based decisions informed by tumor biology [
22‐
24]. Current areas of use include clinical decision-making for nodal assessment (primarily use of imaging), adjuvant radiation therapy (ART) and surgical or therapeutic intervention. These clinical usage management decisions demonstrate a risk-aligned reduction or de-escalation for low-risk, Class 1 patients and increased or escalation in higher-risk Class 2A and 2B patients [
21,
23].
The goal of the current study was to present an independent validation of the 40-GEP test and then, after merging with a previous validation cohort, determine the performance of the 40-GEP test in providing independent prognostic value for the endpoints of regional and distant metastasis. A large cohort of high-risk cSCC further allowed for novel analyses within subgroupings of risk classification systems, individual risk factors and clinically relevant populations, which have not been assessed previously. Evaluation of a large cohort of patients enables a more robust establishment of the significant and independent contribution of the 40-GEP to risk stratification overall and within these novel subpopulations, which could not be achieved with a smaller cohort.
Discussion
We have previously shown that the 40-GEP test provides clinically and statistically significant stratification of metastatic risk in cSCC patients with one or more clinicopathologic high-risk factors in both development and validation cohorts [
13,
14]. Importantly, these studies also demonstrated that the 40-GEP provides independent prognostic value in multivariate models with clinicopathologic risk classification systems as well as individual risk factors. The analysis of a larger high-risk cSCC cohort enabled the current study to demonstrate a stronger validation of the performance of the 40-GEP. This cohort allowed for evaluation of test performance in novel, clinically meaningful subpopulations and enabled detailed analysis of the assay in combination with clinicopathologic risk stratification systems, an accomplishment not possible with smaller cohorts of patients.
The NCCN guidelines provide treatment pathway recommendations using traditional clinicopathologic risk factors. Recommendations include guidance for clinical follow-up and referrals, nodal assessment decisions (clinical exam, imaging using ultrasound or advanced imaging modalities, sentinel lymph node FNA or biopsy), ART and immunotherapy. The 40-GEP test improves the accuracy of risk stratification and enables personalized treatment and surveillance decisions. These refinements can result in escalation or de-escalation of treatment intensity.
Patients categorized as NCCN high or very high risk are eligible for consideration of ART; yet, as a group, most of these patients will not experience metastasis and will not benefit from radiation therapy. As demonstrated within this study, patients categorized as NCCN high risk with a Class 2B test result had an event rate of 30.7%, which is 19% higher than the NCCN very-high-risk group without 40-GEP. By comparison, a NCCN high-risk patient with a Class 1 test result had an event rate of 3.0%, which is 6% lower than the NCCN high-risk group without 40-GEP. Different treatment pathways may be considered as the study results suggest that the Class 2B patients would benefit from ART [
27] and the Class 1 patients have a low likelihood of developing metastatic disease [
14]. These results are consistent with previous studies and have led to the development of physician-derived NCCN-aligned patient management pathways that integrate the 40-GEP test with clinicopathologic features [
11,
20,
21].
BWH T1 and T2a tumors lack the clinical factors, or a number of clinical factors, that are considered very high risk within the framework of this staging system but can occur in the setting of other high-risk features, including those identified as high risk or very high risk by NCCN guidelines (e.g., location on the head or neck, immunosuppression and LVI, among others) that are clinically concerning and confer eligibility for 40-GEP testing. Risk assessment in these lower staged subsets is important because as many as one-third of all metastatic events in cSCC patients have been observed in T1-staged patients alone because of the high denominator of patients diagnosed with T1 tumors [
28]. Within this study, patients with BWH T1 or T2a tumors and Class 1 results experienced low event rates compared to patients with Class 2A or Class 2B tumors. In these patient subsets, the 40-GEP also identified patients who are considered low risk by current staging but harbor aggressive tumor biology with a substantial increase in metastatic risk associated with a Class 2B result. The Class 2A result adds a similar degree of metastatic risk as if a patient presented with an additional clinicopathologic risk factor. Consistent with clinical utility studies, the Class 1, 2A and Class 2B results each lead to risk-aligned changes in patient management (including ART, imaging and follow-up visit frequencies), demonstrating that clinical adoption of risk-aligned treatment plans informed by clinical, pathologic and 40-GEP data together results in comprehensively risk-aligned treatment plans for the benefit of patients [
23,
24].
Test performance in the advanced age populations is particularly significant, given that patients generally have multiple age-associated comorbidities, increasing the need for improved risk stratification to identify patients who can safely forgo more invasive treatments given the increased potential for complications and substantial additions to healthcare costs, including Medicare. Removing younger patients did not impact the significant stratification of risk by the 40-GEP test, demonstrating strong clinical utility for patients diagnosed at ≥ 65 years of age, when comorbidity concerns can start to pose a particular challenge to clinicians. This also held true for patients diagnosed at ≥ 80 years of age, who more often present with additional age-related comorbidities. Thus, use of the 40-GEP can help clinicians identify patients in the population of patients of advanced age who can safely waive treatments and those who are at substantial risk of disease progression and therefore can benefit from more intensive treatment despite the potential for complications and increased cost. Risk-aligned management decisions about treatment options, such as ART [
29‐
31] and immunotherapy [
32‐
34], become increasingly important in the Medicare-eligible population as determining the risk-to-benefit ratio for a given patient can become increasingly complex. This highlights the need for additional risk stratification tools, such as the 40-GEP, to identify patients who can forgo treatments with high morbidity.
The multivariable analyses reported in Tables
2 and
3 and modeling reported in Table
4 show that using the 40-GEP test in combination with clinicopathologic risk assessment leads to more accurate determination of metastatic risk. Additionally, evaluation of accuracy metrics within the entire cohort of 897 cases demonstrated the improved PPV and NPV when including 40-GEP results with clinicopathologic factors and clinically actionable metastatic risk prediction within in stage or subset (Table
5). Based on the reported results, a Class 1 result in the context of a BWH stage T1, T2a or T2b diagnosis provides a level of metastatic risk that is lower than the risk that is expected based on staging alone, given the increased NPV for the combined result compared to staging alone. Conversely, when patients receive a Class 2B result, the observed increase in PPV compared to staging alone increases decision-making confidence to intensify management plans for the patient. Similar improvements in predictive values are also seen for both AJCC8 staging and NCCN risk groups.
A limitation of the study, and unfortunately, for the cSCC population as a whole, is that standard clinical and pathologic factors were not consistently recorded on pathology reports during the time period of this study. To address this limitation, histologic review was conducted on 100% of cases by an independent, board-certified dermatopathologist (blinded to patient outcome and 40-GEP result). Additionally, 100% monitoring of source documentation, including pathology reports, surgical reports and other medical records, was conducted.
Acknowledgements
The authors thank the following individuals and centers for their contributions to this project: David Beynet (VA-Los Angeles), Philip Scumpia (UCLA), Kimberly Brady (Roswell Park), David Pariser (Virginia Clinical Research, Inc), Ramona Behshad (Saint Louis University), Laura Ferris (University of Pittsburgh Medical Center), Mark Nestor (Skin and Cancer Associates), Thomas Knackstedt (MetroHealth-Cleveland), Augusti Toll (Hospital Clinic De Barcelona), Nima Gharavi (Cedars Sinai), Hugh Greenway (Scripps Health), Shawn Kwatra (Johns Hopkins), Kenneth Reed (Derm ASAP), Chrysalyne Schmults (Brigham and Women's Hospital), Ian Maher (University of Minnesota), Yang Xia (Brooke Army Medical Center), Leila Tolaymat (Mayo Clinic Florida), Charles Love, (Radiant Complexions-Iowa Dermatology), Joseph Curry (Thomas Jefferson University Hospital), Catherine Chung (Ohio State University), Diamondis Papadopoulos (Metro Derm- ACCR), Gene Kim (University of Southern California), Robert Bednarek (Parkview Research Center), Rogerio Neves (Penn State, Hershey), Christine Weinberger (University of Vermont), Hooman Khorasani (Mt Sinai), Martin Fleming (University of Tennessee Health Science Center), Simon Yoo (Northwestern University), James Lewis (University of Tennessee Medical Center-Knoxville), Keith Duffy (University of Utah), Evans Bailey (Naaman Clinic), Tim Hansen (McFarland Clinic), John Lyons (Mary Bird Perkins Cancer Center), Nathan Cleaver (Cleaver Dermatology), Manish Gharia (Ascension/Columbia St. Mary's), John Campana (Porter Adventist), Bruce Brockstein (Northshore University Health System), Emily Smith (University of Missouri). J Cañueto is partially supported by the (Gerencia Regional de Salud de Castilla y León (grants GRS2338/A/21 and GRS2549/A/22) and by the Instituto de Salud Carlos III through the project PI21/01207, co-funded by European Union. The authors also thank Sarah Kurley, PhD, for her contribution to the project. This work was conducted with support from Castle Biosciences, Inc. (CBI)