Nonmelanoma Skin Cancer Mortality (1988-2000): The Rhode Island Follow-Back Study

Objectives  To estimate (1) the magnitude of and the components and factors associated with nonmelanoma skin cancer (NMSC) mortality and (2) the proportion of deathsmisclassified as NMSC.

Design  Population-based follow-back study.

Setting and Patients  All Rhode Island residents whose deaths between 1988 and 2000 were attributed to NMSC.

Main Outcome Measures  Distribution of diagnoses, verification of the causes of death, and characterization of associated factors.

Results  The proportion of misclassified deaths was significantly higher for nongenital NMSC (57%) than for genital NMSC (18%; P<.001).Most of the deaths misclassified as nongenital NMSC were caused by squamous cell carcinoma of mucosal surfaces. The age-adjusted NMSC mortality rate was0.91 (per 100 000 persons per year), of which almost half (0.45) were due to genital carcinoma. Nonmelanoma skin cancer mortality increased sharplywith age. The mortality rate from nongenital NMSC in men was more than twice that in women, but for genital NMSC this ratio was reversed. Skin cancersoriginating on the ear were responsible for more than a quarter of all deaths caused by nongenital NMSC. No cases of NMSC mortality occurred in organ transplantrecipients. Many individuals had comorbid psychiatric conditions or evidence of unreasonable delay in seeking medical care for their lesions.

Conclusions  Misclassifying the cause of death as nongenital NMSC accounts for a large source of error on death certificates in Rhode Island. Overall, nongenitalsquamous cell carcinoma and basal cell carcinoma death rates have declined, and mortality due to genital carcinoma was about half of total NMSC deaths.The dermatology community should emphasize prevention of mortality from genital skin cancer, while continuing to stress the importance of reducing excessiveexposure to UV light and prompt treatment of NMSC.

Nonmelanoma skin cancer (NMSC) is the most commonly diagnosed malignancy in the United States, and recent reports suggest that the incidence is increasing.^1,2 Despite the magnitude of the publichealth burden, investigation of NMSC mortality has been limited.³ Mortality due to NMSC in Rhode Island may be related to several factors including risingincidence, increased awareness leading to earlier detection, aging of the population, the introduction of Mohs surgery in 1989, possible changes inthe accuracy of death certificate coding, and potentially greater scrutiny by the medical community leading to better treatment. In addition, the numberof organ transplant recipients in Rhode Island has increased over the last decade; the risk of NMSC mortality in this subgroup is higher than in thegeneral population. The purpose of this study was to determine the rate of NMSC mortality among Rhode Island residents from 1988 through 2000 and toidentify circumstances that may have contributed to a fatal outcome. We also examined the occurrence of coding errors in cause-of-death certification underthe International Classification of Diseases, Ninth (ICD-9) and 10th (ICD-10) Revision, and compared these results with those from a previous study concerning the period from 1979 through 1987.Whereas mortality from carcinomas arising on genital skin were excluded from the prior investigation, deaths attributed to this cause are included herein.Results of this investigation may be useful in developing strategies to reduce mortality from NMSC.

Medical records (death certificates, Rhode Island Department of Health cancer registry data, and hospital medical records) were sought for all RhodeIsland residents whose deaths from 1988 to 2000 were attributed to NMSC. Death certificates for which the cause of death was attributed to NMSC under ICD-9 and ICD-10 rubrics wereobtained from the Rhode Island Department of Health. Carcinomas of genital skin (ie, vulva, penis, and scrotum) were included, but cutaneous lymphomas(ICD-9 202.1 and 202.2 and ICD-10 C84.0 and C84.1) and cancers of the perianal skin (ICD-9 154.3 and ICD-10 C21.0) were excluded. Institutional review board approval was obtained from the Rhode Island Departmentof Health as well as from Kent, Memorial, Miriam, Newport, Rhode Island, Roger Williams, South County, St Joseph's, and the Veterans Affairs Medical Centerhospitals.

All 239 death certificates meeting these criteria were obtained. Complete medical records were available for 135 cases, incomplete records were availablefor 96 cases, and no records (except for the death certificate) were available for 8 cases. Records were incomplete or absent in some cases in which thediagnosis and treatment of disease occurred out of state or in an outpatient or nursing home setting or in cases in which the hospital records had beendestroyed or could not be located by the medical records department staff.

In cases for which records were available, the cause of death was ascertained by documenting a histologic diagnosis of NMSC. In addition, the events immediatelyprior to death were reviewed to verify that the NMSC directly contributed to the fatal outcome. If the events leading to death were not well delineated,documentation of advanced disease such as invasion of the skull or orbital bones or unresectable metastases was considered evidence for a direct contributionto death. Cases in which the above information was available were categorized as "confirmed NMSC deaths." Cases were categorized as "probable NMSC deaths"when there was a diagnosis of invasive NMSC or suggestion that it was advanced and contributed to death in the absence of severe comorbid conditions butno specific information on the presence of metastatic disease nor events leading to death. Additional cases in which there was sufficiently detailed languageon the death certificate to suggest that the cause of death was due to NMSC(eg, squamous cell carcinoma [SCC] of the ear) were also categorized as probableNMSC deaths.

Cases in which death was caused by a melanoma, noncutaneous malignancy, or a severe comorbidity were categorized as "misclassified deaths." The sameparameters including documentation of metastasis, bony invasion, and events preceding death were applied to differentiate "confirmed misclassified deaths"from "probable misclassified deaths." Additional cases in which there wassufficiently detailed language on the death certificate to suggest that thecause of death was misclassified (eg, "squamous cell carcinoma" of the head and neck) in the absence of other information from the medical record werealso categorized as probable misclassified deaths. Cases in which death appeared to result from the contribution of both NMSC and more severe comorbid conditionswere categorized as "multifactorial deaths." Cases in which medical records were absent or devoid of useful information regarding NMSC and cause of deathwere categorized as "indeterminate deaths." In addition to the cause of death, data on patient demographics, risk factors for NMSC, the anatomic locationand histologic features of the primary tumor, and the course of disease were abstracted.

For the purpose of examining the issue of misclassification, only confirmed (NMSC and misclassified) deaths were analyzed. For the purpose of calculatingmortality rates, both confirmed and probable NMSC deaths were included in the analysis. Population estimates for Rhode Island were derived from the1990 and 2000 Rhode Island Census. Mortality rates are age-adjusted to the 2000 US standard population (available at: http://seer.cancer.gov/stdpopulations/; accessed September 6, 2003) unless otherwise specified and are expressedas the number of deaths per 100 000 persons per year.

Results are based on 239 Rhode Island residents whose deaths between1988 and 2000 were attributed to NMSC. Of these, 135 deaths (72 confirmedand 63 probable) were caused by NMSC . The remaining 104 deaths were misclassified (56 confirmed and 33 probable), multifactorial (7 deaths), or indeterminate(8 deaths). Hence, 44% of deaths with a confirmed cause (NMSC or misclassified) were, in fact, misclassified (Table 1).Among confirmed cases, misclassified deaths differed from NMSC deaths with respect to age at death (median, 66 vs 78 years; P =.01) and sex (54% vs 43% men; P = .02). An autopsy was performed in only 1 confirmed case. Although not included in the analysisof misclassification, 2 cases categorized as probable NMSC deaths were caused by SCC but were listed as basal cell carcinoma (BCC) on the death certificates.Deaths attributed to malignant neoplasms of the vagina (ICD-9 184.0) were not included in the analysis except for 1 death that was confirmed to be caused by NMSC of the vulva. Of the 72 confirmed NMSCdeaths, 15 (21%) occurred in patients with physical and psychiatric disabilities that may have contributed to a fatal outcome, including anxiety or depression(7 cases), schizophrenia (2 cases), dementia (3 cases), blindness and/or deafness (2 cases), and morbid obesity (1 case).

Of 128 confirmed (NMSC and misclassified) deaths, 84 were attributed to nongenital NMSC; the remaining 44 deaths were attributed to vulvar (32deaths) and penile carcinoma (12 deaths). Of these 84 deaths, 36 (43%) were correctly classified as nongenital NMSC and were caused by SCC (25 deaths),BCC (5 deaths), and other NMSC types (6 deaths). Analysis of misclassified cases revealed that 85% (41/48) were due to SCC arising from mucosal sitesin the head and neck including larynx, pharynx, and oral cavity. These misclassified cases were typically described on death certificates as "squamous cell carcinomaof the head and neck" and were most commonly coded under ICD-9 rubric 173.4 (NMSC of the scalp and neck) (Table 2). The proportion of correctly classified deaths under ICD-9 173 (38% [26/68]) increased under the corresponding ICD-10 C44 (64% [9/14]), but this difference did not reach statistical significance (P = .10).

Of the 128 confirmed (NMSC and misclassified) deaths, 32 were attributed to vulvar carcinoma, of which 26 (81%) were correctly classified. Analysisof misclassified cases revealed that the cause of death was SCC of the vagina; adenocarcinoma of the cervix, uterus, or ovary; or other gynecological cancers.Twelve of the 128 deaths were attributed to penile carcinoma, of which 10 (83%) were correctly classified. The 2 misclassified deaths were caused bymelanoma and ependymoma.

The overall NMSC mortality rate for Rhode Island residents for the period from 1988 through 2000 is estimated at 0.91 (0.99 in men and 0.86 in women)(Table 3).

Nongenital SCC was the cause of 44 (25 confirmed and 19 probable) deaths. The median age at death was 82 years (range, 34-96 years). Twenty-seven deaths(61%) occurred in men, and all deaths occurred in whites, who comprised 91% of the Rhode Island population in 1990. Metastases were documented in 16%of cases at the time the primary tumor was diagnosed and were documented at some point in the course of the malignancy in all but 23% of cases. Mediansurvival from the time of initial diagnosis was 17 months. The Rhode Islandnongenital SCC mortality rate is estimated at 0.29 (0.50 in men and 0.18 inwomen).

Primary lesions occurred most commonly on the skin of the head or neck including the ear (9 cases), face excluding the nose and eyelids (11 cases),lip (5 cases), and scalp (5 cases). Lesions originating on the extremities were documented in 14% (6/44) of cases. Most (86% [38/44]) of the deaths appearedto be related to sun exposure by virtue of their origin on sun-exposed skin (head, neck, back, chest, or extremities) and the lack of evidence for anothercause. Histologically, 20% (9/44) were well differentiated, 32% (14/44) were moderately differentiated, and 34% (15/44) were poorly differentiated; detailedhistological data were not available in 6 cases. We found chronic ulcers in 3 cases, chronic osteomyelitis in 1 case, prior exposure to therapeutic radiationin 2 cases, history of skin cancer in 4 cases, and/or other noncutaneous malignancies in 2 cases, but found no cases in organ transplant recipients. There was documentationof unreasonable delay in seeking medical treatment in 9 cases, of which 3 were found to have metastatic disease at the time of presentation.

Twelve (5 confirmed and 7 probable) deaths were caused by BCC. The median age at death was 78 years (range, 61-103 years); the median ages for men andwomen were 65 and 83 years, respectively. Six deaths (50%) occurred in men; all 12 deaths occurred in whites. Direct invasion into vital structures wasdocumented in 2 cases at the time of presentation and was documented later in the course of malignancy in all but 3 cases. No cases of metastatic BCCwere confirmed, although 1 death in which there was radiographic evidence of metastatic spread of a BCC originating on the leg was categorized as aprobable NMSC death. Median survival from the time of initial diagnosis was61 months. A specific antecedent condition (a chronic leg ulcer) was identifiedin 1 case. Most (83% [10/12]) of deaths appeared to be related to sun exposureby virtue of their origin on sun-exposed skin and the lack of evidence foranother cause. The Rhode Island BCC mortality rate is estimated at 0.08 (0.13 in men and 0.06 in women). The mortality rate ratio of nongenital SCC to BCCis 3.6 (3.8 in men and 3.0 in women). Mortality from nongenital SCC and BCC increased with age and was highest among persons 85 years and older (Table 3).

An additional 13 deaths (6 confirmed and 7 probable) were caused byother histological types of tumors including dermatofibrosarcoma protuberansin 1 case, Merkel cell carcinoma in 4 cases, sweat gland carcinoma in 2 cases,malignant fibrous histiocytoma of the skin in 2 cases, non–AIDS-relatedKaposi sarcoma in 1 case, and 3 additional cases of uncertain histological type. The Rhode Island mortality rate is estimated at 0.09.

Vulvar carcinoma was the cause of 53 deaths (32 confirmed and 21 probable). Primary lesions were commonly identified on the labia majora, labia minora,or clitoris. We found documentation of specific antecedent conditions including human papillomavirus infection in 1 case, Bowen disease in 1 case, leukoplakiain 2 cases, lichen sclerosis in 1 case, and another noncutaneous malignancy in 1 case. In 6 cases, documentation of unreasonable delay in seeking medicaltreatment was identified. Metastatic disease was diagnosed at the time ofpresentation in 11 cases, of which 3 also had evidence of unreasonable delay.The median age at the time of diagnosis was 78 years (range, 34-103 years); the corresponding age at death was 80 years (range, 40-93 years). The mediantime to death was 1.2 years after diagnosis. Of the 53 tumors, 7 (13%) were well differentiated, 6 (11%) were moderately differentiated, and 10 (19%)were poorly differentiated; detailed histological data were not availablein 30 cases. The Rhode Island mortality rate for vulvar carcinoma is estimatedat 0.56. Mortality increased with age and was highest among persons 75 years and older.

Penile carcinoma was the cause of 12 deaths (10 confirmed and 2 probable). An additional case of SCC of the scrotum was identified. Primary lesions occurredmost commonly on the glans and shaft. Information on circumcision was poorly documented. We found documentation of specific antecedent conditions includinghuman papillomavirus infection (2 cases), Bowen disease (1 case), and erythroplasia de Queyrat (1 case). In 3 cases, unreasonable delay in seeking medical treatmentwas documented; the same number of cases presented with evidence of metastaticdisease. The median age at the time of diagnosis was 67 years (range, 49-86years) and the median age at death was 71 years (range, 53-87 years), although there was no significant difference between these age groups (P = .57). The median time to death was 1.4 years after diagnosis. Ofthe 10 confirmed tumors arising on the penis, 2 (20%) were well differentiated,40% (4/10) were moderately differentiated, and 40% (4/10) were poorly differentiated.The Rhode Island mortality rate for penile carcinoma is estimated at 0.24. Mortality increased with age and was highest among individuals 85 years andolder (Table 3).

The adjusted (to the 2000 US standard) Rhode Island mortality rate forNMSC is estimated at 0.91. Half of these deaths were caused by genital carcinoma.Greater than half of all deaths attributed to nongenital NMSC were misclassified. The proportion of misclassified deaths was highest for ICD-9 code 173.4 (92%) and written cause "squamous cell carcinoma ofthe head and neck" (92%). There was considerably less misclassification into other diagnostic categories. In many cases there was evidence of significantpsychiatric morbidity or delay in seeking treatment. No confirmed deaths occurred in organ transplant recipients. Skin cancers originating on the ear contributedto greater than 25% of deaths caused by nongenital NMSC.

The present study is limited by several factors. The foremost is the limited detail available from retrievable medical records. An autopsy wasperformed in only 1 case, although detailed histopathological information was available in most cases. Strict criteria were used to categorize deathsas confirmed, probable, multifactorial, or indeterminate. Only confirmed deaths were included in the analysis of misclassified cases, whereas both confirmedand probable cases were included in mortality rate calculations. Owing to the nature of this study, it was not possible to ascertain the proportionof deaths caused by NMSC that were misclassified under other causes of death.In addition, deaths attributed to NMSC of the perianal skin (154.3) as wellas deaths attributed to cutaneous lymphomas (ICD-9 202.1 and 202.2) were not included.

Based on results from the present study, the Rhode Island nongenital SCC mortality rate (adjusted to the US 1970 population) is estimated at 0.21for the current period compared with 0.26 for the previous period (1979-1987).³ The estimated rates for men and women for the currentperiod are 0.35 and 0.12, respectively, compared with 0.49 and 0.13 for theearlier period. The BCC mortality rate for the current period is estimated at 0.05 compared with 0.10 for the earlier period. Hence, the BCC and nongenitalSCC mortality rates appear to be declining over time in Rhode Island. Similar findings have been reported in Finland and in Australia.^4,5

The proportion of deaths misclassified as nongenital NMSC for the currentperiod (57%) was similar to that of the earlier period (54%).⁶ Mostof the misclassified deaths were caused by SCC arising from mucosal surfaces in the head and neck (73%). These misclassified deaths were most commonlycoded under the ICD-9 173.4 rubric (92%), but rubrics 140 to 149 would have been correct in most of these cases. In the presentstudy, the proportion of misclassified cases was significantly lower for deaths attributed to genital carcinoma in both men (17%) and women (19%) comparedwith nongenital SCC (P<.001). Previous studies examining the proportion of deaths falsely attributed to NMSC suggest thatthe problem of misclassification is not confined to the United States.^7,8

The period from 1988 through 2000 includes the year in which public health departments began using the ICD-10 for causeof death certification rather than the ICD-9. Deaths attributed to NMSC during 1988 to 1998 were coded under ICD-9 rubrics; the years 1999 to 2000 correspond to ICD-10. The proportion of correctly classified deaths under the ICD-9 173 rubric (38%) increased under the corresponding ICD-10 C44 rubric (64%), although this difference did notreach statistical significance. Of note, 2 deaths attributed to BCC were caused by SCC of the skin. Coding rubrics for ICD-9 and ICD-10 do not permit differentiation of BCC from SCC. Inaddition, 1 death that was attributed to 184.0 (carcinoma of the vagina) was caused by NMSC of the vulva. The correct ICD-9 codewould have been 184.4. It is assumed that the number deaths caused by NMSCthat are misclassified under non-NMSC rubrics is low. However, with the exceptionof deaths attributed to code 184.0, we did not investigate this potential source of misclassification. Nevertheless, estimated mortality rates mustbe interpreted with appropriate caution.

The incidence of SCC and BCC in Rhode Island has not been directly measured. However, incidence data for the nearby state of New Hampshire are availablefrom a recent population-based study, although genital carcinomas were not included in this investigation.² The incidencerate ratios of nongenital SCC to BCC were 0.3 in men and 0.2 in women. By contrast, the corresponding mortality rate ratios in Rhode Island were 4.1in men and 2.6 in women. Hence, these data indicate that case fatality for nongenital SCC is substantially higher than that for BCC.

We are unaware of previous population-based estimates of mortality rates for genital carcinomas. The Rhode Island mortality rates for vulvar and penilecarcinomas suggest that mortality due to genital NMSC represents nearly half of all deaths caused by NMSC in Rhode Island. In addition, while men are twiceas likely to die from nongenital SCC than from genital SCC, women are more than 3 times as likely to die from vulvar carcinoma than from nongenital SCC.Nevertheless, efforts by the dermatology community to promote primary prevention of human papillomavirus infection in the United States pale in comparisonto those promoting reduction of exposure to UV light. Leaders in the dermatology community need to recognize the risk of mortality from genital carcinoma andplace greater emphasis on strategies to prevent it.

The challenges to conducting population-based mortality studies are readily apparent. Although Rhode Island is the smallest state in the UnitedStates, considerable difficulty was encountered with respect to medical record accrual. Because of space limitations, several medical institutions have anongoing schedule for the destruction of medical records. More medical care is being delivered in an outpatient setting, which often promotes a decentralizedsystem of record keeping. Some European countries continue to maintain detailed national cancer registries that can be cross-referenced with cause of deathregistries, yielding a powerful tool for conducting epidemiology research. Obvious concern over privacy of medical records in the United States has ledto increased regulatory burdens that inhibit the creation and use of centralized health information databases for approved research purposes.

Mortality is one critical measure of the public health impact of disease. For NMSC, the accuracy of cause of death certification must be improved toascertain trends in mortality more reliably. This study also underscores the importance of genital skin cancer and suggests that the dermatology communitycould be a forceful advocate for its prevention and early detection.

Correspondence: Martin A. Weinstock, MD, PhD, Dermatoepidemiology Unit, VA Medical Center—111D, 830 Chalkstone Ave, Providence, RI 02908(maw@brown.edu).

Accepted for publication January 5, 2004.

This study was supported by the Medical Student Fellowship from the American Dermatological Association, Millwood, NY (Mr Lewis), and grant CSP402from the Department of Veterans Affairs and grant 78800 from the National Cancer Institute, Bethesda, Md (Dr Weinstock).

We would like to acknowledge the contribution of Suleka Neelagaru, BS, in acquiring death certificates. John Fulton, PhD, extracted data from theRhode Island cancer registry and facilitated access to death certificates. Numerous people in medical records departments at hospitals across Rhode Islandprovided an invaluable contribution through the acquisition of patient medical charts.