Stern RS. Prevalence of a History of Skin Cancer in 2007Results of an Incidence-Based Model. Arch Dermatol. 2010;146(3):279-282. doi:10.1001/archdermatol.2010.4
Copyright 2010 American Medical Association. All Rights Reserved. Applicable FARS/DFARS Restrictions Apply to Government Use.2010
To estimate the 2007 person prevalence of common types of nonmelanoma skin cancer (NMSC), basal cell carcinoma (BCC) and squamous cell carcinoma (SCC), or both, in the United States using an incidence-based mathematical model; and to compare the prevalence of skin cancer with that of other common cancers.
I developed a mathematical model to estimate the prevalence of NMSC in the United States in 2007. This model used age-specific incidence data adjusted to reflect changes in incidence from 1957 to 2006, the age distribution of the population from 1957 to 2006, and the likelihood that an incident tumor was the first ever for that person. I performed sensitivity analyses that varied my assumption about change in incidence over time and proportion of incident tumors that were a first-ever NMSC for an individual. I used standard methods for analysis of survey data to calculate the number of persons who report a history of the selected cancers and published Surveillance, Epidemiology, and End Results (SEER) estimates for incidence-based estimates for prevalence of cancers other than NMSC.
National Health Interview Survey (NHIS) 2007 data, National Cancer Institute Skin Cancer Incidence data (1977-1978), and SEER data.
Main Outcome Measure
Incidence-based estimate of prevalence of NMSC and melanoma and patient reports of a history of skin and selected other cancers.
Approximately 13 million white non-Hispanics living in the United States at the beginning of 2007 have had at least 1 NMSC. About 1 in 5 seventy-year-olds have had NMSCs, and most of those affected have had multiple NMSCs. In the 2007 NHIS estimates, only 5 million persons report a history of skin cancer, less than half the number estimated based on incidence and survival data.
My incidence-based model indicates that the prevalence of a skin cancer history is about 5 times higher than that of breast or prostate cancer and greater than the 31-year prevalence of all other cancers combined. Despite their high frequency, population-based incidence and burden data for BSC and SCC are largely lacking.
Modeling of earlier surveillance incidence data to reflect the increase of incidence over time estimated the 1994 incidence of nonmelanoma skin cancers (NMSCs) in the United States, which are primarily basal cell cancer (BCC) and squamous cell cancer (SCC), to be between 0.9 and 1.2 million.1,2 If the temporal trends assumed by Miller and Weinstock1 from 1978 to 1994 were projected to today, NMSC incidence would approach 2 million persons affected each year, which is substantially higher than recent estimates of annual NMSC incidence.3,4 The 1994 study estimate of lifetime risk of NMSC did not account for the likelihood that an individual might develop NMSCs more than a year apart.1 Incorporating the risk of an individual developing multiple NMSCs over a lifetime should improve the accuracy of the population prevalence and lifetime risk of NMSCs. The 2007 National Health Interview Study (NHIS)5 provides interview-based data that estimates the prevalence of a variety of cancers including NMSC, melanoma, and skin cancer of unknown type.
I present the results of my model and a sensitivity analysis that estimates the cumulative incidence of NMSCs that have occurred in the lifetimes of all white residents in 2007 of the United States and the prevalence of NMSC. I compare incidence-based estimates of prevalence with the NHIS estimate of NMSC and skin cancer prevalence. I also compare incidence- and survey-based estimates of the prevalence of melanoma and 2 other common cancers. The discrepancies between incidence- and interview-based estimates of prevalence of NMSC and melanoma suggest that patients with melanoma may not accurately recall their diagnosis, and that NMSC is not a meaningful term to many patients who have had a BCC or SCC of the skin. The high prevalence of NMSC suggests that better information of their incidence, burden, and prevention is needed.
To estimate the total number of persons living in the United States in 2007 (population prevalence) who had had at least 1 NMSC, I varied age-specific incidence rates from a 1978 federal study to account for the change in the incidence that has been observed over time.2 I then applied these age- and year-specific incidence rates to the number of persons of each age in 2007 who were of that age in prior years to calculate the total number of persons who had 1 or more NMSCs in each year from 1957 to 2006. To calculate the number of unique individuals alive in 2007 who had 1 or more NMSCs in their lifetime, I adjusted annual person incidence to account for the proportion of tumors in each decade for each age group that were the first ever for an individual. I performed sensitivity analyses that varied the assumed change in annual incidence between 1957 and 2006 and the probability that an incident NMSC was the first-ever NMSC for persons of that age in that decade.
Using standard methods for the analysis of survey data,6 I analyzed data from the 2007 NHIS survey to calculate the number of persons reporting a history of skin cancer.5 The 2007 NHIS included more than 20 000 adults.5 For internal controls, I examined comparable data for breast and prostate cancer. In 2007, participants were asked if they ever had been told if they had cancer of specific types, including breast and prostate cancer, melanoma and NMSC, or skin cancer of unknown type. I used published Surveillance, Epidemiology, and End Results (SEER) cancer surveillance system estimates of the 31-year prevalence of melanoma, breast and prostate cancer, and cancer of all types.7,8 All statistical computation was performed using Stata statistical software (version 9; StataCorp, College Station, Texas). I then compared the population prevalence of a history of cancer calculated from NHIS interview data with that available from SEER for breast and prostate cancer and melanoma in 2006 and from my model for NMSC.
My model estimates that 13 million white non-Hispanic persons living in the United States in 2007 have had a total of 22 million BCCs and SCCs during their lifetimes. Under a wide range of assumptions concerning the proportion of incident tumors that were a patient's first ever in my sensitivity analysis of reoccurrence risk, 11 to 15 million persons in the United States in 2007 had a history of NMSC (Table 1). In my base case, the annual rate of change in NMSC incidence since 1978 was 2%. Under this assumption, my model estimates a 2007 NMSC annual incidence of about 1.3 million persons.
I performed a number of additional sensitivity analyses. First, I assessed the effect of change in incidence rates over time. Assuming incidence rates changed 3% per year from 1977 rather than the approximately 2% I used in the base case for my model, the 2007 incidence of NMSC would be about 1.8 million persons, which is substantially higher than most recent estimates of annual NMSC incidence.3,4 An increase in incidence of 1% per year would predict a 2007 NMSC incidence of less than 1 million, a figure substantially below current estimates of NMSC incidence. For most scenarios, with a fairly broad range of assumptions about the risk of multiple tumors over time and with age as well as the annual rate of change in NMSC incidence per year from 1977 (ie, assuming a 3% vs 2% or 1% vs 2% change in annual incidence rate), my estimate of total population prevalence varies by 12% to 25% from my base case estimate of 13 million persons with a history of BSC and/or SCC. Incidence- and survey-based estimates of melanoma prevalence range from 0.76 to 1.1 million persons.7 After accounting for the substantial proportion of patients with melanoma who also develop NMSC and including lower risk populations, the prevalence of skin cancer history is likely to be about 14 million persons, nearly 5% of the US population.
By age 70 years, nearly 1 in 5 white US residents has had at least 1 BCC or SCC. About 1 000 000 current US residents aged 50 years already have had at least 1 NMSC. About 30% of current white US residents who have survived to age 90 years have had at least 1 NMSC. On average, each affected patient of this age had 1 or more NMSCs in 2 different years.
Nearly 50% more persons surveyed gave a history of melanoma compared with SEER estimates based on melanoma incidence and survival. Less than half as many white persons indicated that they had a skin cancer of any type than is likely based on incidence and survival data (Table 2). In contrast, the NHIS interview survey and SEER prevalence estimates for breast and prostate cancer are nearly identical (Table 2). Overall, more persons are likely to have had skin cancer than have had cancers of all other types in the past 31 years.
More than 30 years have passed since the last systematic national population-based study of NMSC incidence in the United States. In 1994, Miller and Weinstock1 updated these incidence estimates to reflect changes in population and a range of assumptions about the increase in incidence of these tumors from 1977 to 1994. Based on about a 4% annual increase in incidence of NMSC, they estimated that there were about 0.9 to 1.1 million persons with NMSC in 1994. Current estimates of NMSC incidence indicate annual incidence range from 1.1 to 1.3 million persons, a figure consistent with the 2% per annum change in incidence since 1978 that I used in my model.3,4
Based on the age distribution of white non-Hispanics in 2007 and the age- and year-specific incidence rates for NMSC, I calculated both the lifetime number and annual number of NMSCs that have occurred in this population. As of 2007, about 13 million white non-Hispanic US residents (6%) have had more than 22 million NMSCs. My model accounted for an individual developing NMSCs in different calendar years. The probability that an NMSC will be a person's first ever decreases with age and with increasing risk in the population. Therefore, my model varied the proportion of persons developing NMSCs in a year whose tumors were second or subsequent tumors for persons of a given age in a given calendar year. Under a wide range of assumptions concerning the risk of multiple tumors in an individual, I estimate that 12 to 15 million white non-Hispanics living in the United States in 2007 have had at least 1 NMSC in their lifetime.
The incidence data I used included only the first NMSC occurring in an individual in that year. The limited available population-based data on multiple NMSCs occurring in the same year suggests that the tumor incidence of BCC and SCC per year is likely to be 10% to 15% greater than the number of persons affected by these tumors each year. Therefore, in 2007 I estimate that 1.2 to 1.5 million BCCs and SCCs were diagnosed in the United States in 2007.
My incidence-based estimate differs greatly from the results of the 2007 NHIS survey that asked about a history of NMSC and a skin cancer of “unknown type,” and melanoma but not specifically about BCC or SCC. The NHIS estimate of skin cancer prevalence is less than half the prevalence I estimate for BCC, SCC, and melanoma. This difference may reflect patients recalling their cancer as a BCC or SCC and a lack of patients' awareness that these common tumors are NMSCs. The findings suggest that a more accurate history of skin cancer might be elicited if patients are asked about each common type of skin cancer by name (ie, BCC, SCC, and melanoma).
Nearly 50% more persons reported a history of melanoma compared with SEER estimates. This finding is consistent with the clinical observation that many patients equate the diagnosis of “skin cancer” with melanoma. As a result, hundreds of thousands of patients may be unnecessarily burdened with the belief that they have had a potentially lethal cancer (melanoma) when in fact they have had a skin tumor that is very unlikely to be lethal. Overall, NHIS estimates of the prevalence of skin cancers overall, and of NMSC specifically, are far lower than their prevalence. This finding further emphasizes the benefit of clearly informing patients about the specific type of skin cancer they have had and its clinical implications. Although their prevalence is far lower than that of skin cancer, the NHIS interview and SEER incidence- and survival-based estimates of prevalence are remarkably consistent for breast and prostate cancer.
My study has both strengths and weaknesses. It provides a contemporary estimate of the incidence of NMSC and estimates the prevalence of persons with a history of skin cancer in the United States. I performed sensitivity analyses that vary the estimates of the change in NMSC incidence over time and the proportion of incident NMSCs that are a person's first ever. My sensitivity analysis incorporates the medically and statistically reasonable assumption that the risk of subsequent tumors increased with age and was higher in high-risk populations.9- 17 My sensitivity analyses included a wide range of estimates of the age- and calendar year–specific proportion of individuals for whom an NMSC was their first ever.
My findings suggest that for clinical history, separate consideration of BCC and SCC is likely to be preferable. However, I chose to estimate the person prevalence of these 2 types of NMSCs together for a number of reasons. There is greater variability in estimates of the incidence of BCC and SCC than of NMSC overall. Many persons who have had an SCC have also had a BCC and vice versa. Treatment and treatment cost for BCC and SCC are often similar. As a result, my approach to provide person prevalence estimates for NMSC is likely to be more precise than separate estimates of the person prevalence of persons with BCC only, SCC only, and with both.
A July 2009 PubMed search revealed only a few studies in English concerning the accuracy of recall of diagnosis and skin cancer. My findings are consistent with a study of California teachers performed nearly 15 years ago.18 It found a low sensitivity (54%) for self-reports of NMSC.18 In contrast, Ming et al19 studied patients enrolled in a case-control study, among whom nearly three-fourths had skin cancer. They found patients' self-reports of skin cancer to be highly accurate. Their findings may reflect the specialized nature of the population that they19 studied. In an Australian study, even those with a documented melanoma frequently indicated that family members had melanoma when they had not—further support of my observation that persons often incorrectly believe a skin tumor was a melanoma when it was, in fact, a less serious lesion.20
As with all findings based on mathematical modeling, the precision of my estimate of NMSC prevalence is not known. However, my sensitivity analyses indicate that under a wide range of assumptions concerning current incidence, change in incidence over time, and an individual's risk of multiple tumors, in 2007 at least 10 million but probably less than 15 million US residents have had at least 1 skin cancer. Most of those patients have had NMSC, and most persons who develop NMSC will have multiple skin cancers during their lifetime. The prevalence of a history of skin cancer is far higher than that of any other cancer and exceeds that of all other cancers diagnosed since 1975.8 Yet, recent population-based data concerning skin cancer incidence, morbidity, and cost of care are lacking for the most common types of skin cancer, BCC and SCC.
Correspondence: Robert S. Stern, MD, Department of Dermatology, Beth Israel Deaconess Medical Center, 330 Brookline Ave, Boston, MA 02215 (email@example.com).
Accepted for Publication: November 11, 2009.
Financial Disclosure: Dr Stern has been a consultant for Johnson & Johnson, Vertex, and Takeda, and an expert witness for Alphapharm, Mutual Pharm, and Johnson & Johnson.
Funding/Support: This study was supported in part by the National Institutes of Health contract NO1-AR-0-2446 and the Department of Dermatology, Beth Israel Deaconess Medical Center.
Role of the Sponsors: The sponsors had no role in the design and conduct of the study; in the collection, analysis, and interpretation of data; or in the preparation, review, or approval of the manuscript.