Risk Factors for Cutaneous Squamous Cell Carcinoma Recurrence, Metastasis, and Disease-Specific Death: A Systematic Review and Meta-analysis

Importance  To date, the magnitude of association and the quality of evidence for cutaneous squamous cell carcinoma (cSCC) and risk factors for outcomes have not been reviewed and analyzed systematically.

Objective  To systematically analyze all published data on risk factors for recurrence, metastasis, and disease-specific death (DSD) of cSCC.

Data Sources  Comprehensive search of Ovid MEDLINE In-Process & Other Non-Indexed Citations, MEDLINE, EMBASE, Cochrane Central Register of Controlled Trials, Cochrane Database of Systematic Reviews, and Scopus, from each database’s inception to May 14, 2015.

Study Selection  Inclusion criteria were studies of at least 10 patients, comparative data for at least 1 cSCC risk factor, and an outcome of interest. Exclusion criteria were noncutaneous squamous cell carcinoma (SCC), anogenital SCC, inability to extract cSCC data from other malignancy data, SCC in situ, Marjolin ulcer, and genetic disorders predisposing to cSCC.

Data Extraction and Synthesis  Two reviewers independently abstracted the data. Meta-analysis was performed using the random-effects model. Risk of bias was assessed by the Newcastle-Ottawa Scale.

Main Outcomes and Measures  A priori outcomes were recurrence, metastasis, and DSD.

Results  Thirty-six studies (17 248 patients with 23 421 cSCCs) were included. Significant risk factors for recurrence were the following: Breslow thickness exceeding 2 mm (risk ratio [RR], 9.64; 95% CI, 1.30-71.52), invasion beyond subcutaneous fat (RR, 7.61; 95% CI, 4.17-13.88), Breslow thickness exceeding 6 mm (RR, 7.13; 95% CI, 3.04-16.72), perineural invasion (RR, 4.30; 95% CI, 2.80-6.60), diameter exceeding 20 mm (RR, 3.22; 95% CI, 1.91-5.45), location on the temple (RR, 3.20; 95% CI, 1.12-9.15), and poor differentiation (RR, 2.66; 95% CI, 1.72-4.14). Significant risk factors for metastasis were: invasion beyond subcutaneous fat (RR, 11.21; 95% CI, 3.59-34.97), Breslow thickness exceeding 2 mm (RR, 10.76; 95% CI, 2.55-45.31), Breslow thickness exceeding 6 mm (RR, 6.93; 95% CI, 4.02-11.94), diameter exceeding 20 mm (RR, 6.15; 95% CI, 3.56-10.65), poor differentiation (RR, 4.98; 95% CI, 3.30-7.49), perineural invasion (RR, 2.95; 95% CI, 2.31-3.75), immunosuppression (RR, 1.59; 95% CI, 1.07-2.37), and location on the temple (RR, 2.82; 95% CI, 1.72-4.63), ear (RR, 2.33; 95% CI, 1.67-3.23), or lip (RR, 2.28; 95% CI, 1.54-3.37). Significant risk factors for DSD were: diameter exceeding 20 mm (RR, 19.10; 95% CI, 5.80-62.95), poor differentiation (RR, 5.65; 95% CI, 1.76-18.20), location on the ear (RR, 4.67; 95% CI, 1.28-17.12) or lip (RR, 4.55; 95% CI, 1.41-14.69), invasion beyond subcutaneous fat (RR, 4.49; 95% CI, 2.05-9.82), and perineural invasion (RR, 4.06; 95% CI, 3.10-5.32). Evidence quality was considered low to moderate.

Conclusions and Relevance  Tumor depth is associated with the highest RR of local recurrence and metastasis of cSCC, and tumor diameter exceeding 20 mm is associated with the highest RR of DSD. Unified, consistent collection and reporting of risk factors in a prospective, multicentered effort are needed to further understand the increasing incidence of cSCC.

Cutaneous squamous cell carcinoma (cSCC) is the second most common malignancy of the skin, with an estimated annual incidence of 700 000 cases in the United States.^1,2 Most cases of cSCC portend an excellent prognosis after surgical removal.³ However, 3.7% to 5.2% of patients have nodal metastasis, and 1.5% to 2.1% die of cSCC.^4-8 Although these rates are low compared with many other malignancies, the absolute number of patients with cSCC who have nodal metastasis is estimated at 5604 to 12 572 in the United States alone.¹ Furthermore, the absolute number of cSCC-related deaths is estimated at between 3932 and 8791 annually, the upper limit of which approaches annual melanoma-related deaths.¹ The absence of a national tumor registry for cSCC complicates the analysis of prognostic factors related to outcomes on a broad scale. Therefore, our current understanding of prognostic factors is based primarily on retrospective analyses of single-institution cohorts, with heterogeneously reported data. The objectives of the present study were to perform a systematic review and meta-analysis of all published reports of cSCC risk factors and outcomes and to quantify the magnitude of each risk factor and the quality of the supporting data.

The study was deemed exempt from review by the Mayo Clinic Institutional Review Board. No patient informed consent was required.

A comprehensive and systematic search of Ovid MEDLINE In-Process & Other Non-Indexed Citations, MEDLINE, EMBASE, Cochrane Central Register of Controlled Trials, Cochrane Database of Systematic Reviews, and Scopus was performed from each database’s earliest inception to May 14, 2015, by an experienced librarian (L.J.P.), with input from the study’s principal investigator (C.L.B.) and lead author (A.K.T.). Controlled vocabulary supplemented with keywords was used to search for studies of risk factors in cSCC and association with outcomes. The search strategy is outlined in the eTable in the Supplement. Bibliographies of selected review articles were reviewed for additional relevant studies. Only studies on human data published in or translated into English were included.

Inclusion criteria were studies of at least 10 patients with cSCC that reported comparative data associating at least 1 defined risk factor (ie, depth of invasion, perineural invasion [PNI], diameter, tumor differentiation, immunosuppression, and location on the lip, ear, temple, or cheek) and an outcome of interest (ie, recurrence, nodal metastasis, and disease-specific death [DSD]). Exclusion criteria for the studies included any of the following: squamous cell carcinoma (SCC) in situ, noncutaneous SCC, anogenital SCC, cSCC data that could not be extracted from data on other malignancies (eg, basal cell carcinoma and melanoma), cSCC in patients with genetic disorders that predispose to cSCC (eg, xeroderma pigmentosa), cSCC arising in scar tissue (Marjolin ulcer), and an outcome of interest present at study initiation. No limitations were imposed on the basis of treatment modality.

Two reviewers (A.K.T. and B.F.K.) independently selected studies on the basis of inclusion and exclusion criteria. Disparities in selection were resolved through discussion and ultimately by a third reviewer (C.L.B.). Studies were initially reviewed on the basis of title and abstract, and those deemed relevant were reviewed in full text to establish the final set of studies ultimately included. In cases of study duplication, the more recent and complete studies were selected for inclusion. From these studies, data were abstracted in duplicate (by A.K.T. and B.F.K.) to verify the accuracy.

Risk of bias assessment was analyzed for each article by 2 investigators (A.K.T. and B.F.K.) using the Newcastle-Ottawa Scale.⁹ We considered that the most important factor in determining risk of bias was the ability of a study to adjust for confounders (ie, other risk factors) using multivariable adjustment.

From each study, we extracted or estimated the risk ratio (RR) (presented as the hazard ratio [HR] or odds ratio [OR]) for each risk factor and outcome of interest, with the 95% CI. Various association measures were assumed to be comparable statistically. Multivariable estimates were preferentially extracted when available. Meta-analysis was performed using the random-effects model,¹⁰ reporting the RR (95% CI). The I² statistic¹¹ was also calculated to measure inconsistency, and a value exceeding 50% implied substantial heterogeneity (ie, difference in the estimates derived from each study). Forest plots were constructed for all associations of risk factors and outcomes. Analysis was conducted using statistical software (Comprehensive Meta-Analysis, version 3.0; Biostat).

The database search identified 1156 publications and meeting abstracts. After initial review of abstracts and titles, 1034 records were excluded (Figure 1). The remainig 122 studies were reviewed in full text, and 36 studies were ultimately included in our analysis.

The 36 studies described 23 421 cSCCs in 17 248 patients. Five were prospective cohort studies, and 31 were retrospective studies. No randomized trials were identified. The number of patients in the included studies ranged from 41 to 6164 (Table 1).

Analyzed risk factors and their respective associations with outcomes of recurrence, metastasis, and DSD are shown in Figure 2, with corresponding RRs, 95% CIs, P values, I² statistics, and forest plots. Risk factors associated with statistically significant increased risk of recurrence were the following: Breslow thickness exceeding 2 mm (RR, 9.64; 95% CI, 1.30-71.52), invasion beyond subcutaneous fat (RR, 7.61; 95% CI, 4.17-13.88), Breslow thickness exceeding 6 mm (RR, 7.13; 95% CI, 3.04-16.72), presence of PNI (RR, 4.30; 95% CI, 2.80-6.60), diameter exceeding 20 mm (RR, 3.22; 95% CI, 1.91-5.45), location on the temple (RR, 3.20; 95% CI, 1.12-9.15), and poor differentiation (RR, 2.66; 95% CI, 1.72-4.14).

Statistically significant risk factors for metastasis were also assessed. They included the following: invasion beyond subcutaneous fat (RR, 11.21; 95% CI, 3.59-34.97), Breslow thickness exceeding 2 mm (RR, 10.76; 95% CI, 2.55-45.31), Breslow thickness exceeding 6 mm (RR, 6.93; 95% CI, 4.02-11.94), diameter exceeding 20 mm (RR, 6.15; 95% CI, 3.56-10.65), poor differentiation (RR, 4.98; 95% CI, 3.30-7.49), presence of PNI (RR, 2.95; 95% CI, 2.31-3.75), immunosuppression (RR, 1.59; 95% CI, 1.07-2.37), and location on the temple (RR, 2.82; 95% CI, 1.72-4.63), ear (RR, 2.33; 95% CI, 1.67-3.23), or lip (RR, 2.28; 95% CI, 1.54-3.37).

Several risk factors showed a statistically significant association with DSD. These included the following: diameter exceeding 20 mm (RR, 19.10; 95% CI, 5.80-62.95), poor differentiation (RR, 5.65; 95% CI, 1.76-18.20), location on the ear (RR, 4.67; 95% CI, 1.28-17.12) or lip (RR, 4.55; 95% CI, 1.41-14.69), invasion beyond subcutaneous fat (RR, 4.49; 95% CI, 2.05-9.82), and presence of PNI (RR, 4.06; 95% CI, 3.10-5.32).

Scores from the Newcastle-Ottawa Scale are listed in Table 2, with risk assessments for each study. Six studies (16.7%) were deemed to have low risk of bias due to the presence of multivariable data, and the other 30 studies (83.3%) were of high or unclear risk owing to a lack of adjustment for confounding variables. The number of studies for each risk factor ranged from 1 to 15. Risk factor and outcome associations are shown in Figure 2, which visually depicts them in a forest plot.

The pooled data for most of the risk factors showed a statistically significant association with the outcomes of interest. However, some previously reported risk factors for each outcome did not have a statistically significant association. Since the 1992 publication of cSCC outcomes by Rowe et al,⁴³ several other studies have contributed to this body of knowledge. Staging systems proposed for the purpose of stratifying patients by outcomes include the American Joint Committee on Cancer (AJCC) seventh edition of the AJCC Cancer Staging Manual (hereafter the AJCC system),⁴⁴ the Union for International Cancer Control (hereafter the UICC system),⁴⁵ and the Brigham and Women’s Hospital (hereafter the BWH system)⁴⁶ staging systems (Table 3). While the BWH system was developed on the basis of a single institution’s experience, the AJCC and UICC systems are based on expert consensus. The present meta-analysis summarizes the entire body of data on these previously reported risk factors. To our knowledge, our data provide the most comprehensive review of previously reported risk factors for cSCC related to the outcomes of local recurrence, metastasis, and DSD.

Optimal management of cSCC is predicated on local tumor control because local recurrence is often the first indicator of aggressive biologic behavior.^3,47 In our analysis, tumor depth (recorded as Breslow thickness in millimeters or anatomic depth) was associated with the highest RR of local recurrence. The highest-quality comparative data on Breslow thickness are by Brantsch et al,⁴ representing one of the few prospective data sets in this analysis. The AJCC staging system for cSCC⁴⁴ includes both Breslow thickness exceeding 2 mm and Clark level of at least 4 as high-risk characteristics. The largest data set describing anatomic depth is a retrospective analysis by Karia et al.⁶ In that study, depth was measured by anatomic depth rather than Breslow thickness, with tumor depth beyond subcutaneous fat considered a high-risk characteristic. Although it is reassuring that the significance of tumor depth is verified through distinct methods of measure, the absence of a uniform measure and reporting contributes to data heterogeneity and ambiguity for the physician.

The theoretical advantage of Breslow thickness measurement is the ability to perform refined analysis of data on a continuous numerical variable. However, from a practical perspective, measurement of Breslow thickness is limited by time and the abundant frequency of transected shave biopsy specimens. Defining depth on the basis of the anatomic depth (eg, invasion beyond subcutaneous fat) simplifies the objective measure assessed in both horizontally and vertically sectioned tissue specimens. If adopted as a standard, the parameters for association with anatomic structures (eg, depth or invasion beyond subcutaneous fat) will require strict unambiguous definition.

After tumor depth, PNI was associated with the next highest RR of local recurrence. As with tumor depth, reporting of PNI was also heterogeneous. Some articles reported PNI as a binary variable,^{8,14,24,33,34,41} whereas other studies^6,48 also analyzed on the basis of the diameter of the involved nerve. For the purpose of the present study, PNI was analyzed as a binary variable, including the above-cited 6 articles,^{8,14,24,33,34,41} and was associated with an RR of local recurrence of 4.30 (95% CI, 2.80-6.60; P < .01). The study by Karia et al⁶ also indicated that the RR increases with the size of the involved nerve, with an RR of regional recurrence for nerves of less than 0.1 mm in diameter of 5.6 (95% CI, 2.0-15.9; P = .001) compared with an RR of regional recurrence for nerves of at least 0.1 mm in diameter of 10.4 (95% CI, 4.4-24.7; P < .001). The AJCC staging system for cSCC⁴⁴ includes PNI as a binary variable for high risk, whereas the BWH staging system includes PNI as a risk factor when the involved nerve is at least 0.1 mm in diameter. As with tumor depth, the uniform definition and reporting of PNI is an area of critical need.

In descending order of RR, tumor diameter, location on the temple, and poor differentiation were associated with local recurrence. A cutoff of 20 mm was used for tumor diameter because it is uniformly included in the AJCC staging system, the UICC staging system, and the BWH staging system. Although 4 studies^6,17,24,35 in this analysis showed statistical significance, the prospective data by Brantsch et al⁴ did not demonstrate an association between tumor diameter and recurrence on multivariable analysis. However, a statistically significant increase in risk was noted on univariable analysis (HR, 3.47; 95% CI, 1.89-6.39; P < .001). Tumor location on the temple, although not part of the AJCC, UICC, or BWH staging systems,⁴⁴ was associated with a higher risk of recurrence (RR, 3.2; 95% CI, 1.1-9.0; P = .03) than tumor location on either the lip or ear, although this finding was based on a single large retrospective analysis by Schmults et al.⁸ Poor differentiation was also found to have a statistically significant association with recurrence (RR, 2.66; 95% CI, 1.72-4.14; P < .01) based on the pooled data from 11 studies.^{4,6,17,19,22-24,34,37,39,41}

Surprisingly, location on the lip or ear and immunosuppression were not associated with a statistically significant RR of local recurrence in the present meta-analysis. The lip⁴³ and ear^6,43 have been reported as risk factors in some studies, but a sufficient body of data suggests otherwise.^4,17,34 Both the lip and the ear are included in the AJCC system but not in the UICC and BWH systems. Immunosuppression is a notable item within the AJCC system but is not included in the staging system, and it is not part of the BWH or UICC staging system.

For the purpose of the present analysis, immunosuppression was included as a general category, without further stratification. Although immunosuppression was associated with recurrence in the prospective data by Brantsch et al⁴ and in certain subsets of well-defined patients (eg, those with chronic lymphocytic leukemia²³), the results of these 2 studies were likely offset by other studies^6,34,39 that did not stratify on the basis of the type of immunosuppression. Taken together, the data suggest that clarification of the nature of the immunosuppression is critical for understanding the factors that influence local recurrence. Notably, the I² statistics for location on the lip or ear and for immunosuppression were increased at 52%, 59%, and 39%, respectively, indicating substantial data inconsistency, which also could contribute to a lack of statistical significance.

The correlation between cSCC risk factors and metastasis was the most well-studied and well-represented finding in our meta-analysis. The implications of metastasis are obvious because they generally are associated with poor outcomes.^8,49,50 Therefore, identifying risk factors associated with metastasis in cSCC is crucial for proper staging and for early identification of high-risk patients. As with local recurrence, our study showed that tumor depth was associated with the highest RR of metastasis, regardless of whether depth is measured as Breslow thickness or anatomic depth. These results are based on 7 studies,^{4,6,25,34,38,40,42} one of which presents data on both Breslow thickness and anatomic depth.³⁸ The RR for metastasis in our analysis was greater for Breslow thickness exceeding 2 mm than for Breslow thickness exceeding 6 mm. This finding may reflect the absolute number of patients in the studies who had nodal metastases with primary tumor depths of 2 to 6 mm vs the smaller population of patients with the same outcome and tumor depth exceeding 6 mm. No patients in the prospective data by Brantsch et al⁴ had nodal metastasis with Breslow thickness of less than 2 mm. The BWH group found that the anatomic depth associated with metastasis was the depth beyond subcutaneous fat (RR, 7.0; 95% CI, 2.4-20.3; P < .001).⁶ The practice gaps related to reporting of depth hold true for association with metastasis and for association with local recurrence.

Tumor diameter exceeding 20 mm and poor differentiation were associated with similar RRs of 6.15 (95% CI, 3.56-10.65) and 4.98 (95% CI, 3.30-7.49), respectively, for the development of metastasis. Eight included studies^{4,6,19,26,30,38-40} were based on tumor diameter and metastasis. Poorly differentiated cSCC and metastasis represented the most well-studied association in this analysis, with a total of 18 publications that provided comparative data.^{4-7,13,15,16,19,25-28,30,34,38-40,42}

The remaining risk factors associated with increased RR of metastasis in this analysis were PNI (RR, 2.95; 95% CI, 2.31-3.75), location on the temple (RR, 2.82; 95% CI, 1.72-4.63), location on the ear (RR, 2.33; 95% CI, 1.67-3.23), location on the lip (RR, 2.28; 95% CI, 1.54-3.37), and immunosuppression (RR, 1.59; 95% CI, 1.07-2.37). The relative absence of heterogeneity for these risk factors is reflected in the I² statistics of 0%, 6%, 27%, and 0% for location on the temple, ear, and lip and immunosuppression, respectively. This absence of heterogeneity further underscores associations between these risk factors and metastasis. Tumor localization on the temple is not included as a risk factor in the AJCC, UICC, and BWH staging systems. Yet, these results suggest that it might be advisable to categorize this location with the other high-risk anatomic locations, such as the ear and lip. Furthermore, a more stratified approach to the specific type of immunosuppression (eg, human immunodeficiency virus infection, solid-organ transplant, or chronic lymphocytic leukemia) would afford greater precision in our risk analysis for metastasis in these patients.

Of all the outcomes, DSD had the least amount of comparative data available for analysis. The risk factor with the highest RR for DSD was diameter exceeding 20 mm (RR, 19.10; 95% CI, 5.80-62.95). However, this conclusion is based solely on 1 study by Karia et al.⁶ Poor differentiation, location on the ear or lip, invasion beyond subcutaneous fat, and PNI were associated with a statistically significant increase, ranging from 4-fold to 6-fold, in the RR of DSD. Although significant, tumor depth was not as highly associated with DSD as with local recurrence and metastasis, likely a reflection of a paucity of data and the heterogeneity of available studies, reflected in the I² statistic of 76%. As a risk factor for DSI, PNI had an I² statistic of 0%, reinforcing this association. In the single study⁶ with the largest data set related to DSD, risk factors in their multivariable analysis with the greatest association with DSD were diameter exceeding 20 mm (HR, 19.1; 95% CI, 5.8-63.0; P < .001), invasion beyond subcutaneous fat (HR, 11.1; 95% CI, 3.4-35.8; P < .001), location on the ear (HR, 10.1; 95% CI, 1.8-57.0; P = .008), and poor differentiation (HR, 10.0; 95% CI, 3.4-28.9; P < .001). Only 1 study⁶ (among 141 patients) that reported immunosuppression as a risk factor for DSD was included in our analysis, and the authors of that report acknowledged that their study was underpowered to assess prognostic significance. Given the increasing burden of cSCC and the estimated magnitude of DSD, the results of our analysis further emphasize the need for comprehensive reporting of these results.

This analysis was limited by several factors. First, it is possible that there were pertinent studies that were not identified because of inherent limitations in database literature searches. Second, the quality of evidence is limited by data largely derived from single-center experiences and retrospective analyses, with heterogeneous data reporting and study design. Third, some included studies had variable and at times limited follow-up data. However, no eligibility criteria were imposed on the basis of follow-up. Rather, such limitations were reflected in the Newcastle-Ottawa Scale scoring (Table 2). Fourth, a large number of the studies provided unadjusted estimates, and many risk factors can plausibly be codependent. Fifth, we included studies in which a small proportion of patients had SCC in the setting of scarring, a genetic disorder, or the anogenital region.

To our knowledge, this systematic review and meta-analysis is the largest and most comprehensive study of risk factors related to outcomes for cSCC. These results verify the significance of many previously reported factors, while providing a novel, robust quantitative risk for each risk factor and the associated outcomes of local recurrence, metastasis, and death. In the short term, these results may help guide physicians in their risk assessment of patients, particularly those with only 1 identified risk factor, while keeping in mind the inherent limitations of the data. In the long term, these results may be used to refine the evolving work on staging systems for cSCC, while providing a renewed call to action for data collection. Not only are unambiguous definitions for each risk factor needed, but also comprehensive, uniform reporting of risk factors and outcomes is necessary to provide optimal care for the increasing number of patients with cSCC in the United States and globally.

Accepted for Publication: October 19, 2015.

Corresponding Author: Christian L. Baum, MD, Department of Dermatology, Mayo Clinic, 200 First St SW, Rochester, MN 55905 (baum.christian@mayo.edu).

Published Online: January 13, 2016. doi:10.1001/jamadermatol.2015.4994.

Author Contributions: Drs Thompson and Baum had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

Study concept and design: Thompson, Murad, Baum.

Acquisition, analysis, or interpretation of data: Thompson, Kelley, Murad, Baum.

Drafting of the manuscript: Thompson, Baum.

Critical revision of the manuscript for important intellectual content: Thompson, Kelley, Murad, Baum.

Statistical analysis: Murad.

Obtained funding: Murad.

Administrative, technical, or material support: Prokop.

Study supervision: Baum.

Conflict of Interest Disclosures: None reported.

Funding/Support: This study was supported in part by Center for Clinical and Translational Science grant UL1 TR000135 from the National Center for Advancing Translational Sciences.

Role of the Funder/Sponsor: The funding source had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.

Disclaimer: The contents herein are solely the responsibility of the authors and do not necessarily represent the official views of the National Institutes of Health.

Previous Presentation: This study was presented in part at the 47th Annual Meeting of the American College of Mohs Surgery; April 30, 2015; San Antonio, Texas.