Evaluation of AJCC tumor staging for cutaneous squamous cell carcinoma and a proposed alternative tumor staging system. JAMA Dermatol. 2013. PMID: 23325457.

SUMMARY: This is the original paper which set out to create a more predictive staging system than 7th edition AJCC staging for cutaneous SCC, specifically with respect to AJCC T2 tumors.  This category (AJCC T2) contains a high rate of events of interest (metastasis and death) and and are still relatively common compared to AJCC T3 and T4 tumors but lacks homogeneity of outcomes (outcomes among T2 tumors vary quite a bit) desired in a staging system.  Of note, an AJCC 8th edition has since been released. 

​

More info...

 

Important notes about the cohort studied are:

• the inclusion of only tumors with at least one risk factor predetermined by the authors (PNI/LVI, poor differentiation, depth beyond fat, diameter of 2cm or more, ear location, vermillion lip location).

• Subsequently, a high rate of tumors with PNI (24%), poor differentiation (20%), depth beyond fat (15%), 2cm or greater (18%), and tumors on the ear and vermillion lip without other risk factors (41%). 

• Not surprisingly, the overall rate of nodal mets (10%), and disease specific death (5%) was also high compared to population rates, estimated at 4% and 1.5%, respectively, in the introduction of the paper (18617440). 

• Low rate of nodal met (0.7%) and disease specific death (0%) in T1 tumors.  Median follow up was 44 months.  

• They did not specify location of tumor beyond lip/ear and non-lip/ear locations making it unclear if these tumors were all head and neck location.  

• Interestingly, in the final T staging system, 134/256 cases were T1 tumors, meaning they had NO risk factors.  The authors only included tumors in the study with at least 1 of the 6 risk factors listed above, and 4 of them were in the final staging system.  Therefore, all 134 of these T1 tumors must have contained one of either LVI or location on lip/ear.  However, in the results they indicate only 104 tumors were of lip/ear location with no other risk factors.  The nature of the remaining 30 tumors is unclear.  

​

Evaluation of American Joint Committee on Cancer, International Union Against Cancer, and Brigham and Women's Hospital tumor staging for cutaneous squamous cell carcinoma. J Clin Oncol. PMID: 24366933.

SUMMARY: Compared to the 7th edition AJCC and UICC, BWH provided better distinction (staging groups correlate with distinct outcomes).  This was accounted for by the highest 2 staging categories, T3 and T4, in which AJCC and UICC categories were indistinct.  It also showed superior monotonicity (increasing stages correlate with worse outcomes), and homogeneity (outcomes were similar within each staging category).  

​

If you refer to table 3 of the paper, the key difference can be illustrated in the predicted rate of nodal mets, in which not only are all 4 T stages for the BWH criteria distinct, but significantly so, meaning the confidence intervals do not even approach each other, and are somewhat evenly distributed: for stages T1, T2a, T2b, and T3, CIs were 0-0.4, 1-5, 13-27, and 41-100, respectively.  For AJCC, stages T1, T2, T3, and T4, CIs were 0-1, 4-9, 37-100, and 34-100.  As you can see, AJCC stages T3 and T3 are not distinct, and there is also a large jump from predicted rate of nodal mets from T2 to T3.

​

More info...

​

Key differences with original BWH staging study (23325457) are:

• SCC of the anogenital and eyelid skin were excluded with the reasoning being these are staged separately in AJCC and UICC systems.  Remember, the initial study did not specify tumor location other than lip/ear or non-lip/ear.

• This study explicitly includes tumors outside the head and neck area and in fact the majority were from these locations (63%).  Of note, the 7th edition of AJCC staging for cutaneous SCC did include non-head and neck tumors but the 8th edition is specifically for head and neck locations only (30003984).

• This cohort contained far fewer cases with certain high risk factors compared to original study: PNI 4% vs 23%, poor dx 13% vs 20%, and depth beyond fat 2% vs 15%.  Tumor size 2cm or greater was closer but still less than the original study, 13% vs 18%.  

• Rate of nodal mets and disease specific death was even lower in T1 tumors in this cohort compared to original: 0.1% vs 0.8% and no disease specific deaths in either.  The overall rate of nodal mets in this study, all T stages, was 1.8%.  

​

Validation of a 40-gene expression profile test to predict metastatic risk in localized high-risk cutaneous squamous cell carcinoma. J Am Acad Dermatol. 2021. Erratum in: J Am Acad Dermatol. 2021. PMID: 32344066. Supplemental at https://doi.org/10.17632/f33w9wmng4.1.

SUMMARY: This is the original paper looking to validated the 40-GEP for SCC, similar to the test for melanoma.  The inclusion criteria were NCCN high risk or greater, or AJCC 8 or BWH T stage of 2 or higher tumors, but 62.6% of the tumors were AJCC8 T1 and 57.9% BWH stage T1, presumably because of the loose NCCN criteria.  The overall metastatic rate of 16% in the cohort, up to 8 times the 1.9-2.6% population rate estimated from a cohort of nearly 9000 patients (22592943), is one of the major critiques.  Regarding accuracy of the tests (accuracy being how often is the test correct), vs the staging systems, utilizing the formula accuracy = (sensitivity) x (prevalence) + (specificity) x (1-prevalence), only when comparing 40-GEP class 2B vs 1/2A is there an advantage over BWH or AJCC8 staging.  Comparing class 2 vs 1 decreases accuracy by approximately 10% compared to BWH and AJCC8.

​

More info...

​

DETAILED REVIEW:

​

Introduction: Authors state that "early detection of metastasis is correlated with better outcomes", but do not provide a citation for this important claim.  The final statement of the introduction is also a conclusion of the paper: "the 40-GEP test is an independent predictor of outcomes and improves on risk prediction with staging systems, supporting its potential clinical use in conjunction with standard staging and patient management criteria".  If this statement is based on prior literature it should be cited appropriately.  Otherwise, it should be saved for the discussion.  

 

Methods: Authors state that the tissue was prospectively designed on archival tissue.  This contradictory statement should be clarified in order to clearly represent the timing of events pertaining to collection and analysis of tissue. 

 

They also do not clearly specify what tissue was archived: was it initial biopsy specimens, was it Mohs surgery debulks, etc.?  They do indicate that all data from the definitive surgery (presumably including Mohs) was included, suggesting analysis could have included Mohs debulk specimens. 

 

The inclusion criteria were interesting: NCCN high risk or greater, or AJCC 8 or BWH T stage of 2 or higher.  For NCCN criteria this excluded recurrent tumors, but presumably could have included any "rapidly growing" or "poorly defined" tumor, fairly subjective variables, although there is no indication that a tumor was included solely on one of these factors.  Similarly, any acantholytic or other high risk histologic subtype would have qualified, regardless of size or location.  Regarding AJCC 8 and BWH criteria, they intended to include T stages higher than T1, but 62.6% of the tumors were AJCC8 T1 and 57.9% BWH stage T1.  The larger proportion of T1 tumors in AJCC8 is likely from the size criteria for which a 2cm tumor without other risk factors is BWH T2 but AJCC8 T1, and the poorly dx criteria which is at least BWH T2 but does not factor into AJCC8 and would result in a T1 staging if no other factors are present.  If this is the case, then the 186 BWH T1 tumors included would have been so based on the NCCN criteria.  Factors that would result in a high risk NCCN designation but remain BWH T1 include: head, neck, hands, feet, pretibia, and anogenital locations, tumors with poorly defined borders, tumors in immunosuppressed patients, tumors in a radiation field or chronic inflammatory process, rapidly growing tumors, tumors with histology of acantholytic (adenoid), adenosquamous (showing mucin production), or metaplastic (carcinosarcomatous), tumors with small caliber perineural involvement, and possibly those with neurologic symptoms, although these would like present with other high risk factors.  It would be helpful to know which of these criteria were met to better understand the cohort.  Finally, they did not specify location other than head and neck vs other which is significant because 33% were non head and neck location and therefor technically cannot be staged with AJCC8 criteria.  

 

Results: The first major critique is the overall metastatic rate of 16%, up to 8 times the 1.9-2.6% population rate estimated from a cohort of nearly 9000 patients (22592943).  The most notable impact of this would be to significantly overestimate the PPV of this test.  Furthermore, if calculating the accuracy of the tests vs the staging systems (accuracy being how often is the test correct), utilizing the formula, accuracy = (sensitivity) x (prevalence) + (specificity) x (1-prevalence), only when comparing 40-GEP class 2B vs 1/2A provides an advantage over BWH or AJCC8 staging.  Comparing class 2 vs 1 decreases accuracy by approximately 10% compared to BWH and AJCC8.

​

Enhanced metastatic risk assessment in cutaneous squamous cell carcinoma with the 40-gene expression profile test. Future Oncol. 2022. PMID: 34821148.  Supplemental at https://www.futuremedicine.com/doi/suppl/10.2217/fon-2021-1277.

SUMMARY: This is the Ibrahim et al paper.  Of note, overall metastatic rate was 15%; BWH T1 tumors had a 9.5% metastatic rate, T2b a 15.2% rate; AJCC8 T1 12.6% and T2 9.3%.

​

More info...

 

Aim: this study aimed to couple the 40-GEP test with clinical factors.  In the introduction they state that 30% of metastatic outcomes come from cases classified as low risk (BWH T1/T2a) and that 25% of high risk SCCs (BWH T2b/T3) do not develop mets.  Because of this, they sought to combine gene expression profiling with clinical factors (which are used for BWH staging) to better risk stratify cutaneous SCCs, similar to the integrated 31-GEP for melanoma.  

​

​Methods: tissue was obtained from FFPE biopsies or WLE specimens.  They included cases that were treated with Mohs surgery but it does not appear as if debulks from MMS was used; possibly, because these are often discarded or processed as frozen specimens and not submitted for FFPE.  

​

For statistics they don't really provide much detail, at least in the methods section.  They reference a prior paper, 32344066, and say it was previously described, but this referenced paper also does not describe the statistical methods and refers you to "previously described" without providing a citation.  Additionally, this was stated to be a new model combining 40-GEP and clinical factors, so presumably the statistical analysis would be different and the methods would focus on this unique aspect.  Oddly enough, they do go on to perform multiple statistical calculations that they discuss in the results section as they present the results.  

​

They indicate 7 clinical risk factors they are looking at: tumor size and location (as one variable, in the same manner as it is used in the NCCN risk stratification), tumor depth, immunosuppression, PNI, LVI, degree of differentiation, and histologic subtype.  They then assign equivalent value to each risk factor and assign each case 1 point for each risk factor present.  They use this later on for their combined 40-GEP and clinical risk factor assessment where they categorize them in a binary fashion into 1 risk factor or 2 or more risk factor groups.  This is a little strange, and smells a little of a retrofitted pseudo-staging system that they found produced a publishable effect when combined with the 40-GEP.  It’s unclear why they didn’t use primary staging systems like BWH or AJCC8 in conjunction with the 40-GEP (stratifying each T stage, not creating binary low and high risk groupings as they later do), or create a new model altogether including the above risk factors along with the 40-GEP class all as independent variables akin to the melanoma i31-GEP (they provide a multivariable risk assessment later in the paper, but not a risk model). 

​

Results: this is where some statistical gymnastics begins.  The meatiest part is the discussion of the combined 40-GEP and clinical risk factor assessment, and it gets hard to follow quickly.  The gist is this: a 40-GEP Class 1 result predicts a lower metastasis rate than this new binary risk factor classification system they came up with would have predicted (8.2% for 1 risk factor and 19.7% for 2-7 risk factors), and a Class 2A or Class 2B result predicts a higher and much higher (respectively) metastasis rate.  Notably, a Class 2B result combined with 2 or more risk factors leads to a lower predicted metastatic rate than a 2B result with 1 risk factor, demonstrating inferior monotonicity to current clinical factor staging alone.  

​

They try their luck again, this time risk stratifying tumors with the conventional NCCN method (high and very high risk) and then stratifying them by 40-GEP class.  Monotonicity was preserved, but apart from a Class 2B result which predicted noticeably higher rates of metastasis, Class 1 and 2A results showed more modest risk stratification on baseline predicted risk based on NCCN risk stratification.  

​

They then try to make the case for the accuracy of the test compared to BWH and AJCC8, and while accuracy is better when comparing 40-GEP Class 2B to BWH T2b/T3 and AJCC8 T3/T4 by roughly 5% and 8%, respectively, its significantly worse when comparing 40-GEP Class 2A/B by roughly 25% and 22%, respectively.

​

Later on they provide a univariate analysis which demonstrated poor differentiation, PNI, deep invasion, and tumor diameter as significant; by multivariate analysis, poor differentiation, deep invasion, and 40-GEP result were all predictors of metastasis.  A risk model with these factors would be interesting if not useful.  

​

The supplemental also has some nice data, specifically supplemental figure 2 which allows one to stage a patient clinically with BWH or AJCC8 and then modify risk with the 40-GEP.  The only drawback to this table, and its a big one, is that it’s likely not generalizeable to most patient populations as the metastatic rate in T1 tumors was 9.5% and 12.6%, respectively, for BWH and AJCC8 staged tumors.