Importance
Patients with basal cell carcinoma (BCC) and cutaneous squamous cell carcinoma (SCC) (often termed nonmelanoma skin cancer or keratinocyte carcinoma [KC]) often develop new KCs, but information is limited on the frequency and timing of these subsequent tumors. This information is crucial to guide follow-up care.
Objective
To determine the timing of subsequent new KCs in patients who present with KC.
Design, Setting, and Participants
We enrolled a consecutive cohort of 1426 patients diagnosed as having biopsy-proven KC from January 1, 1999, through December 31, 2000, in a university dermatology practice and its affiliated Department of Veterans Affairs dermatology service. After exclusion of patients with basal cell nevus syndrome and immunocompromise, 1284 patients (90.0%) were followed up prospectively for a mean of 5.7 (range, 0-12.3) years.
Main Outcomes and Measures
We assessed the risks for subsequent KCs over time using single-failure and multiple-failure models. We separately assessed outcomes after first lifetime KCs and after nonfirst lifetime KCs. We also performed secondary analyses of the risk for a subsequent BCC after a prior BCC diagnosis and the risk for a subsequent SCC after a prior SCC diagnosis.
Results
The risk for a subsequent KC was substantially lower after the first lifetime KC diagnosis: 14.5% (95% CI, 11.9%-17.7%) at 1 year, 31.1% (95% CI, 27.3%-35.3%) at 3 years, and 40.7% (95% CI, 36.5%-45.2%) at 5 years, than after a nonfirst KC: 43.9% (95% CI, 42.0%-45.9%) at 1 year, 71.1% (95% CI, 69.1%-73.0%) at 3 years, and 82.0% (95% CI, 80.2%-83.7%) at 5 years. Secondary analyses of the risks for a subsequent BCC after a prior BCC diagnosis and of a subsequent SCC after a prior SCC diagnosis yielded results consistent with the analyses for the pooled KC sample.
Conclusions and Relevance
Although all patients with KC are assumed to be at high risk for subsequent tumors, a subset may not develop another KC after their first tumor. Whether these findings are related to biological or behavioral differences or to differences in health care services should be investigated further to inform and improve care. Ongoing routine screening for subsequent KC may not be indicated for all patients with KC. Skin cancer screening can be improved with a better understanding of the course and frequency of subsequent KC diagnoses.
As the most common malignant neoplasm, keratinocyte carcinoma (KC), which includes basal cell carcinoma (BCC) and cutaneous squamous cell carcinoma (SCC), is a priority for clinical medicine and public health efforts.1 Keratinocyte carcinoma is highly prevalent; in the United States alone, approximately 3.5 million tumors are treated annually1—more than twice as many as all other malignant neoplasms combined.2 Moreover, KC is a problem for patients; for example, the Global Burden of Disease Study reports that patients with KCs have disability-adjusted life-years similar to thyroid cancer and substantially more than Hodgkin lymphoma and testicular cancer.3 Finally, KC is costly, ranking as the fifth most expensive cancer for Medicare in the United States.4,5
Patients with KC are also at substantially increased risk for development of subsequent new KCs. In a meta-analysis of 17 studies,6 persons with a history of KC had approximately 10 times the risk for developing subsequent KC during 3 years of follow-up compared with the general population. A recent meta-analysis of 45 studies reported that 29.2% of patients with BCC were later diagnosed as having a subsequent BCC and 4.3% were diagnosed as having a subsequent SCC; 13.3% of patients with SCC were later diagnosed as having a subsequent SCC and 15.9% were diagnosed as having a subsequent BCC.7 Previous studies have typically focused on only the first subsequent KC after a KC diagnosis, however.6,8,9 Few studies have assessed the development of multiple subsequent KCs, and those available have incorporated multiple subsequent KCs to investigate risk factors only.10,11 Thus, information is limited on the cumulative risks and patterns of all subsequent KCs, especially between patients with first lifetime KCs and those with a history of previous KCs (nonfirst KCs) at baseline.
The risk for subsequent KCs is regarded as high enough to justify long-term ongoing care for patients after a KC diagnosis12; typically, patients with KC are seen regularly in the clinic to screen for new tumors. However, whether all patients should be monitored the same way or whether skin cancer screening can be improved with a better understanding of the course and frequency of subsequent KC diagnoses remains unclear.
Our goal was to examine the timing and pattern of subsequent new KCs in patients who present with KC. We tested the hypothesis that patients with a single lifetime KC diagnosis have different risks for subsequent KC than those with greater than 1 lifetime KC.
We conducted and report this study in accordance with the STROBE (Strengthening the Reporting of Observational Studies in Epidemiology) guidelines.13 The study was approved by the institutional review board of the University of California, San Francisco, and was conducted according to the principles of the Declaration of Helsinki. Patients provided written informed consent when required.
Design, Patients, and Measures
We conducted a prospective observational cohort study of all patients with newly diagnosed KC (BCC or SCC, including in situ SCC, based on results of biopsy and histopathologic evaluation) from January 1, 1999, through December 31, 2000, at a university-affiliated dermatology practice or its affiliated Department of Veterans Affairs (VA) dermatology clinic. Details of the study have been described elsewhere.14-16
Overall, 1426 consecutive patients with primary KCs were eligible for inclusion. Because we were interested in typical patients who presented with KCs, we excluded patients with basal cell nevus syndrome (a genetic condition that predisposes to multiple BCCs) and patients with immunocompromise that may increase their risk for KC (defined as those with human immunodeficiency virus infection, history of solid organ transplant, or any other immunocompromise noted in the medical record) (Figure 1).
Baseline characteristics of the patients and tumors were obtained from patient responses to a survey at enrollment and from the medical record. Patients were categorized by whether they had a history of KC before their KCs at enrollment, although the number of previous tumors was not known.
We followed up these patients through medical record review by a trained record abstractor who was also a dermatology nurse practitioner. All subsequent KCs were verified by results of histopathologic evaluation and were identified from pathology records, and subsequent KCs were distinguished from recurrences of enrollment tumors using a formal algorithm.14 Medical record review was complete for 93.8% of eligible patients. Data were right censored on the last date of care in the medical record. To evaluate whether differences existed between patients with shorter and longer follow-up times, we compared baseline characteristics between patients in the lowest and the highest quartile of follow-up using unpaired t tests and χ2 tests as appropriate.
We sought to examine the timing of all subsequent KCs throughout the follow-up period rather than the first subsequent tumor only. Because a greater number of prior KCs has been reported to be a risk factor for subsequent tumors,6 we initially focused on patients who had no history of KC before enrollment and were enrolled with a single KC. Among these patients, the KC diagnosis at enrollment was the patient’s first lifetime KC event and the next subsequent KC diagnosis was the second lifetime KC event, followed by the third, fourth, etc. We separately analyzed each diagnostic interval between these KC events, that is, first lifetime KC to second, second to third, third to fourth, fourth to fifth, and fifth to sixth. For each diagnostic interval, we estimated the probability over time of developing a subsequent KC and created Kaplan-Meier failure curves. We compared the probability over time of developing a subsequent KC after the first lifetime KC event interval with each other interval using a log-rank test and a Cox proportional hazards regression model adjusted for age, sex, history of SCC, history of KC on the head and neck, location of care, and number of visits per patient per year to the dermatology clinic. We then repeated these analyses for patients known to have had KC before enrollment.
Because the timing pattern of subsequent KC development was substantially different for the interval of first to second lifetime KC compared with all the other intervals, we divided diagnostic intervals into 2 groups. The first group consisted of intervals for first to second lifetime KC, and the second group consisted of intervals for all nonfirst to subsequent KC.
To estimate the probability of developing a subsequent KC and to assess whether well-established risk factors differed in the 2 groups, we performed single-failure Cox proportional hazards regressions for the first group (first to second lifetime KC events) and multiple-failure Cox proportional hazards regressions for the second group (nonfirst to subsequent KC events). In the multiple-failure analysis, follow-up time was measured from the previous KC diagnosis, robust SEs were used, and analysis was clustered on patients to account for the same patients included in multiple intervals.17 We performed univariate and multivariate analyses to evaluate the roles of known risk factors for subsequent KC in the 2 groups. Based on previous literature, we adjusted for age, sex, history of SCC, history of KC on the head and neck (the most sun-exposed area of the body), location of care (VA clinic vs university-affiliated clinic), and number of visits per year to the dermatology clinic. Risk factors that could change during the follow-up, such as age, history of SCC, and history of a head and neck tumor, were updated at the beginning of each interval. We assessed risk factors in each of the 2 groups (first to second lifetime KC events and nonfirst to subsequent KC events) separately. For Cox regression models, the proportional hazards assumption was tested graphically and numerically using Schoenfeld residuals.18
In addition, we calculated the new risk for a second KC after a first lifetime KC given that a subsequent KC event had not occurred by 2 or 4 years after the first. We included all patients with a first lifetime KC event at enrollment who had not had a second KC event at 2 or 4 years of follow-up and used single-failure, unadjusted Cox proportional hazards regression to calculate the risk for a subsequent KC in the following 1 and 2 years.
Throughout our primary analyses, we grouped patients who enrolled with BCCs and SCCs and did not differentiate between subsequent BCCs and SCCs diagnosed because our primary interest was in the pattern and timing of new KC diagnoses of all histologic findings. The secondary analyses below describe our separate analyses of both tumor types.
Secondary and Sensitivity Analyses
Because of the biological differences between BCC and SCC tumors, we performed secondary analyses investigating the risks for a subsequent BCC after a prior BCC diagnosis and a subsequent SCC after a prior SCC diagnosis. We also performed a secondary analysis that included Fitzpatrick skin type (a measure of tendency to sunburn) as a covariate in the multivariate analyses. Although we hypothesized that Fitzpatrick skin type could be a risk factor for development of subsequent KCs, we did not use it in the primary analyses owing to missing data. Fitzpatrick skin type was available for 725 patients (56.5% of the study sample).
A final sensitivity analysis excluded KCs diagnosed within 60 days of the previous KC event. The rationale for this analysis was to avoid inclusion of KCs that were already present at the time of the previous KC diagnosis.
All statistical tests were 2 sided. All analyses were performed using commercially available software.19
Patients and Baseline Characteristics
After exclusions, the sample consisted of 1284 patients (90.0%) diagnosed as having 1633 primary KCs at enrollment (Figure 1). Patient and tumor characteristics at enrollment are displayed in Table 1. The mean patient age was 67.0 years, and 71.5% of patients were male. Most patients presented with a single KC, and 49.2% of patients had a history of KC before the enrollment KC. Of the 1633 enrollment KCs, the most common body location was the head and neck (62.2%), followed by the trunk (21.2%).
Patients were followed up for a mean of 5.7 (range, 0-12.3; median, 6.9; interquartile range, 6.7) years. Of our cohort, 75.7% had at least 2 years of follow-up, 58.1% had at least 5 years of follow-up, and 39.7% had 8 years of follow-up. During follow-up, a total of 2932 KCs were diagnosed (mean [SD], 2.3 [4.2] KCs per patient), of which 1895 (64.6%) were BCCs and 1037 (35.4%) were SCCs. Patients had a mean (SD) of 1.9 (2.6) dermatology clinic visits per year during the follow-up. We compared patients in the lowest quartile of follow-up time with patients in the highest quartile of follow-up time and found no statistically significant difference (χ2 test, P > .05) in sex, Fitzpatrick skin type, site of care (VA clinic vs university-affiliated clinic), history of KC at enrollment, or tumor type at enrollment (BCC vs SCC). Patients in the lowest quartile were older than patients in the highest quartile (unpaired t test, P < .01).
Figure 2A depicts Kaplan-Meier plots of the probabilities over time of subsequent KC diagnoses after the first lifetime KC event (ie, for patients with no history of KC at enrollment). Separate plots are given for the intervals of first to second, second to third, third to fourth, fourth to fifth, and fifth to sixth lifetime KC events. The probability curve for the first to second lifetime KC interval was significantly different from each of the 4 other curves at baseline (log-rank test, P < .001) and when adjusted for age, sex, history of SCC, history of KC on the head and neck, location of care, and number of dermatology clinic visits per year (hazard ratio, P < .001). The probability curve for a second lifetime KC event after the first lifetime KC event was also significantly different from the probabilities for all of the diagnosis intervals of patients who had a history of KC at enrollment (eFigure 1 in the Supplement; log-rank test, P < .001; hazard ratio, P < .001). Thus, when separated into intervals, the first to second lifetime KC interval was significantly different from all other nonfirst intervals.
When intervals were grouped into first to second lifetime KC or nonfirst to subsequent KCs, the risk for a new KC was lower after the first lifetime KC (14.5% at 1 year, 31.1% at 3 years, 40.7% at 5 years, and 59.6% at 10 years) than after the nonfirst KC (43.9% at 1 year, 71.1% at 3 years, 82.0% at 5 years, and 91.2% at 10 years). The risk of developing a subsequent KC at 10 years after a first lifetime KC (59.6%) was comparable to the risk of developing a subsequent KC at 2 years after a nonfirst lifetime KC (61.5%) (Table 2 and Figure 2B).
In evaluating the role of established risk factors in the 2 groups, independent predictors of a second KC event after a first lifetime KC were found to be similar to the independent predictors of subsequent KCs after a nonfirst KC event (eTable 1 in the Supplement), with the exception of sex and treatment location, which were statistically significantly associated in the group with a subsequent KC after a nonfirst KC event but not in the group with a second KC event after a first lifetime KC. The proportional hazards assumption was tested graphically and numerically using Schoenfeld residuals. For the interval from the first lifetime to subsequent KCs, the site of care (university-affiliated clinic vs VA clinic) failed to meet the assumption of proportional hazards (P = .045), but stratifying on this predictor resulted in little variation (<10% relative differences) for the hazard ratios of the other predictors. For the interval from the nonfirst lifetime to subsequent KCs, sex failed to meet the proportional hazards assumption (P < .001). In this case, stratifying on this variable again resulted in less than 10% change in the hazard ratios for other predictors, with the exception of history of a head and neck tumor; however, this predictor was not significant before or after stratification. Thus, for simplicity, we report the results from the original Cox proportional hazards regression models in the 2 interval strata (first to second and nonfirst to subsequent).
We also calculated the probabilities of developing a second KC event for patients who enrolled with a first lifetime KC event and had not developed a second KC by 2 years or by 4 years. If a patient had not developed a second KC event 2 years after a first lifetime KC, the probability of developing a second KC in the following year was 10.4% (95% CI, 7.7%-13.9%); in the following 2 years, the probability was 18.0% (95% CI, 14.4%-22.5%). If a patient had not developed a second KC event 4 years after a first lifetime KC, the probability of developing a second KC in the following year was 5.9% (95% CI, 3.6%-9.5%); in the following 2 years, the probability was 10.8% (95% CI, 7.5%-15.4%).
Secondary and Sensitivity Analyses
Secondary analyses of the risk for a subsequent BCC after a prior BCC diagnosis and the risk for a subsequent SCC after a prior SCC diagnosis yielded results consistent with the primary analyses for the pooled KC sample (eFigures 2 and 3 and eTables 2 and 3 in the Supplement). The results of the multivariate analyses for risk factors did not change significantly when adjusted for Fitzpatrick skin type. Fitzpatrick skin type was not an independent predictor of subsequent KC diagnosis and did not significantly improve the model (Wald test, P > .05 for first to second KC intervals and for the nonfirst to subsequent KC intervals). A sensitivity analysis that excluded subsequent KCs diagnosed within 60 days of the last KC also yielded similar results to those of the primary analyses.
We found that the risks for a second KC after a first lifetime KC are significantly lower over time than the risks for a third KC after a second, a fourth after a third, a fifth after a fourth, and a sixth after a fifth. Our analysis shows that at 10 years, approximately 40% of patients diagnosed as having a first lifetime KC will not have developed another tumor. When patients who enrolled with a first lifetime KC developed at least 1 subsequent KC, however, they had substantially higher risks for subsequent tumors, similar to the risks seen in patients with a history of KC at enrollment. The risk of developing a subsequent KC at 10 years after a first lifetime KC (59.6%) was similar to the risk of developing a subsequent KC at only 2 years after a nonfirst lifetime KC (61.5%). These findings remained after adjustments for potential risk factors and confounders and were also consistent in analyses of each tumor type (subsequent BCC after a prior BCC diagnosis and subsequent SCC after a prior SCC diagnosis). Although the literature reports that a history of KC increases the risk for subsequent KC diagnosis, our study demonstrates that the risk over time for a second KC is substantially lower than the risk over time for a subsequent KC after at least 2 KCs have been diagnosed, and that patients diagnosed as having a single lifetime KC have a good chance of remaining free of subsequent KCs, even at 10 years after the first tumor.
The studies in the literature on this topic have limitations. A meta-analysis published in 20006 that included 17 studies and more than 64 000 patients with KC reported that having a personal history of KC carried an approximately 10-fold increase in the risk for subsequent KC at 3 years of follow-up. However, the studies sourced for this meta-analysis and subsequent studies8-11,20 have limitations. Many of the studies are retrospective record-linkage analyses, and prospective cohort studies, when available, tend to be small. In addition, some studies are secondary analyses of data from randomized clinical trials of skin cancer prevention strategies,20,21 and selective participant criteria may limit their generalizability. Our study, which prospectively followed up a large consecutive sample of patients with KC for more than 10 years, is well suited to address both limitations and to determine in a novel way the timing and patterns of multiple subsequent KC diagnoses in patients with first KCs or nonfirst KCs. We observed similar associations with established risk factors for the development of subsequent KCs for first to second KCs and for nonfirst to subsequent KCs. Further investigation is needed to understand the patterns observed in these 2 groups. Certain individuals might be genetically less predisposed to develop multiple KCs over time. In addition, some patients may modify their behavior after the diagnosis of a first lifetime KC enough to protect themselves from subsequent KCs, whereas others do not. Other nonbiological and nonbehavioral factors might apply, and further research should focus in particular on patients who have only 1 lifetime KC event and do not develop subsequent tumors.
These results have implications for clinical care. Patients presenting with their first lifetime KC may be the group most likely to benefit from preventive counseling, whereas patients presenting with a history of KC may benefit from more aggressive or more frequent screening for subsequent tumors. Also, we found that patients who did not develop a subsequent tumor by 2 or 4 years after a first lifetime KC diagnosis were at low risk for subsequent tumors. That is, a patient seen at a follow-up visit 2 years after a first lifetime KC had an 82% likelihood of remaining free of subsequent KCs 2 years later, and a patient seen at 4 years after a first lifetime KC had an 89% likelihood of remaining free of subsequent KCs 2 years later. Clinicians should consider whether these patients could be seen in the clinic less frequently.
Although the demographic and histologic characteristics of the patients with KC in this cohort are typical of patients with KC,22-24 this cohort was from a single city and from a VA or university-affiliated dermatology clinic, both of which potentially limit the generalizability of our findings. Some patients may have received dermatologic care at sites other than the study sites, so some subsequent KCs might have been diagnosed in other clinics and our estimates may be underestimates. However, more than 80% of the patients in our cohort had a skin examination at one of the study sites in the year before their final follow-up date, and approximately 70% of patients had at least 1 dermatology visit every 2 years, indicating that most had continued to receive dermatologic care that would be captured by this study. Data on the known behavioral risk factor of sun exposure were not available and could not be included in our investigation of risk factors, although the focus of this work was the pattern and timing of subsequent tumors rather than risk factors. Finally, detailed review of medical records was used to ensure that subsequent KC diagnoses were primary KCs and not recurrences of the enrollment KC, but we cannot be sure that every subsequent KC was not a recurrence of a different subsequent KC that had been diagnosed during follow-up. Overall recurrence rates after treatment of KC are very low, however.14
The risk for subsequent new KC over time is substantially lower after a first lifetime KC diagnosis than after a nonfirst KC diagnosis. Although all patients with KC are assumed to be at high risk for subsequent tumors, a subset may not develop another KC. Whether these findings are related to biological or behavioral differences or to differences in health care services should be investigated further to inform and improve care.
Accepted for Publication: August 23, 2014.
Corresponding Author: Mary-Margaret Chren, MD, Department of Dermatology, University of California, San Francisco, 2340 Sutter St, Room N412, Box 0808, San Francisco, CA 94143 (chrenm@derm.ucsf.edu).
Published Online: January 14, 2015. doi:10.1001/jamadermatol.2014.3307.
Author Contributions: Dr Chren had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: Wehner, Linos, Chren.
Acquisition, analysis, or interpretation of data: All authors.
Drafting of the manuscript: Wehner, Chren.
Critical revision of the manuscript for important intellectual content: All authors.
Statistical analysis: Wehner, Parvataneni, Boscardin, Chren.
Obtained funding: Wehner, Linos, Chren.
Administrative, technical, or material support: Linos, Stuart, Chren.
Study supervision: Linos, Chren.
Conflict of Interest Disclosures: Dr Chren serves as a consultant for Genentech. No other disclosures were reported.
Funding/Support: This study was supported in part by grants from the Doris Duke Charitable Foundation (Dr Wehner), by the Dermatology Foundation (Dr Linos), by award KL2RR024130 from the National Center for Research Resources of the National Institutes of Health (NIH) (Dr Linos), and by grants R01 AR 054983 and K24 AR052667 from the National Institute of Arthritis and Musculoskeletal and Skin Diseases, NIH (Dr Chren).
Role of the Funder/Sponsor: The funding sources 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.
1.Rogers
HW, Weinstock
MA, Harris
AR,
et al. Incidence estimate of nonmelanoma skin cancer in the United States, 2006.
Arch Dermatol. 2010;146(3):283-287.
PubMedGoogle ScholarCrossref 2.American Cancer Society. Cancer Facts & Figures 2013. Atlanta, GA: American Cancer Society; 2013.
4.Housman
TS, Feldman
SR, Williford
PM,
et al. Skin cancer is among the most costly of all cancers to treat for the Medicare population.
J Am Acad Dermatol. 2003;48(3):425-429.
PubMedGoogle ScholarCrossref 5.Mudigonda
T, Pearce
DJ, Yentzer
BA, Williford
P, Feldman
SR. The economic impact of non-melanoma skin cancer: a review.
J Natl Compr Canc Netw. 2010;8(8):888-896.
PubMedGoogle Scholar 6.Marcil
I, Stern
RS. Risk of developing a subsequent nonmelanoma skin cancer in patients with a history of nonmelanoma skin cancer: a critical review of the literature and meta-analysis.
Arch Dermatol. 2000;136(12):1524-1530.
PubMedGoogle ScholarCrossref 7.Flohil
SC, van der Leest
RJ, Arends
LR, de Vries
E, Nijsten
T. Risk of subsequent cutaneous malignancy in patients with prior keratinocyte carcinoma: a systematic review and meta-analysis.
Eur J Cancer. 2013;49(10):2365-2375.
PubMedGoogle ScholarCrossref 8.Ramachandran
S, Rajaratnam
R, Smith
AG, Lear
JT, Strange
RC. Patients with both basal and squamous cell carcinomas are at a lower risk of further basal cell carcinomas than patients with only a basal cell carcinoma.
J Am Acad Dermatol. 2009;61(2):247-251.
PubMedGoogle ScholarCrossref 9.Graells
J. The risk and risk factors of a second non-melanoma skin cancer: a study in a Mediterranean population.
J Eur Acad Dermatol Venereol. 2004;18(2):142-147.
PubMedGoogle ScholarCrossref 10.Kiiski
V, de Vries
E, Flohil
SC,
et al. Risk factors for single and multiple basal cell carcinomas.
Arch Dermatol. 2010;146(8):848-855.
PubMedGoogle ScholarCrossref 11.Flohil
SC, Koljenović
S, de Haas
ER, Overbeek
LI, de Vries
E, Nijsten
T. Cumulative risks and rates of subsequent basal cell carcinomas in the Netherlands.
Br J Dermatol. 2011;165(4):874-881.
PubMedGoogle ScholarCrossref 12.Zanna
P, Lucchese
A, Sevilla
MC,
et al. Molecular and ultrastructural analysis of a multiphasic oral malignant melanoma.
Ultrastruct Pathol. 2011;35(1):37-41.
PubMedGoogle ScholarCrossref 13.von Elm
E, Altman
DG, Egger
M, Pocock
SJ, Gøtzsche
PC, Vandenbroucke
JP; STROBE Initiative. The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement: guidelines for reporting observational studies.
PLoS Med. 2007;4(10):e296. doi:10.1371/journal.pmed.0040296.
PubMedGoogle ScholarCrossref 14.Chren
MM, Linos
E, Torres
JS, Stuart
SE, Parvataneni
R, Boscardin
WJ. Tumor recurrence 5 years after treatment of cutaneous basal cell carcinoma and squamous cell carcinoma.
J Invest Dermatol. 2013;133(5):1188-1196.
PubMedGoogle ScholarCrossref 15.Chren
MM, Torres
JS, Stuart
SE, Bertenthal
D, Labrador
RJ, Boscardin
WJ. Recurrence after treatment of nonmelanoma skin cancer: a prospective cohort study.
Arch Dermatol. 2011;147(5):540-546.
PubMedGoogle ScholarCrossref 16.Chren
MM, Sahay
AP, Sands
LP,
et al. Variation in care for nonmelanoma skin cancer in a private practice and a Veterans Affairs clinic.
Med Care. 2004;42(10):1019-1026.
PubMedGoogle ScholarCrossref 17.Prentice
RL, Williams
BJ, Peterson
AV. On the regression analysis of multivariate failure time data.
Biometrika. 1981;68(2):373-379.
Google ScholarCrossref 18.Schoenfeld
D. Partial residuals for the proportional hazards regression model.
Biometrika. 1982;69(1):239-241.
Google ScholarCrossref 19.StataCorp. Stata Statistical Software: Release 12. College Station, TX: StataCorp LP; 2011.
20.Pandeya
N, Purdie
DM, Green
A, Williams
G. Repeated occurrence of basal cell carcinoma of the skin and multifailure survival analysis: follow-up data from the Nambour Skin Cancer Prevention Trial.
Am J Epidemiol. 2005;161(8):748-754.
PubMedGoogle ScholarCrossref 21.Karagas
MR, Stukel
TA, Greenberg
ER, Baron
JA, Mott
LA, Stern
RS; Skin Cancer Prevention Study Group. Risk of subsequent basal cell carcinoma and squamous cell carcinoma of the skin among patients with prior skin cancer.
JAMA. 1992;267(24):3305-3310.
PubMedGoogle ScholarCrossref 22.Gallagher
RP, Hill
GB, Bajdik
CD,
et al. Sunlight exposure, pigmentation factors, and risk of nonmelanocytic skin cancer, II: squamous cell carcinoma.
Arch Dermatol. 1995;131(2):164-169.
PubMedGoogle ScholarCrossref 23.Gallagher
RP, Hill
GB, Bajdik
CD,
et al. Sunlight exposure, pigmentary factors, and risk of nonmelanocytic skin cancer, I: basal cell carcinoma.
Arch Dermatol. 1995;131(2):157-163.
PubMedGoogle ScholarCrossref 24.Vitaliano
PP, Urbach
F. The relative importance of risk factors in nonmelanoma carcinoma.
Arch Dermatol. 1980;116(4):454-456.
PubMedGoogle ScholarCrossref