Incidence of Cutaneous T-Cell Lymphoma in the United States, 1973-2002

Objective  To describe incidence trends for cutaneous T-cell lymphoma (CTCL) in the United States.

Design  Population-based study.

Setting  Data were obtained from 13 population-based cancer registries of the Surveillance, Epidemiology, and End Results (SEER) Program of the National Cancer Institute from 1973 through 2002.

Participants  A total of 4783 cases of CTCL were identified for the period 1973 through 2002.

Main Outcome Measure  Diagnosis of CTCL.

Results  The overall annual age-adjusted incidence of CTCL was 6.4 per million persons. Annual incidence increased by 2.9 × 10⁻⁶ per decade over the study period. Incidence was higher among blacks (9.0 × 10⁻⁶) than among whites (6.1 × 10⁻⁶) and was higher among men (8.7 × 10⁻⁶) than among women (4.6 × 10⁻⁶). The racial differences in incidence decreased with age, while the sex differences increased with age and decreased over time. Substantial geographic variation in incidence was found. Incidence was correlated with high physician density, high family income, high percentage of population with a bachelor's degree or higher, and high home values. Changes in International Classification of Diseases for Oncology (ICD-O) morphologic definitions have resulted in the redistribution of the cases of CTCL among specific subclassifications.

Conclusions  The continued rise in incidence of CTCL is substantial, and the cause of this increase is unknown. The racial, ethnic, sex, and geographic differences in incidence may be of etiologic importance. Changes in ICD-O definitions have made it difficult to evaluate incidence trends for subclassifications of CTCL such as mycosis fungoides. In addition, these changes resulted in the creation of ambiguous histologic codes, which may have caused coding errors. These errors along with the lack of independent verification are limitations of our study. An epidemiological investigation using population-based data is important to better understand this disorder.

Cutaneous T-cell lymphoma (CTCL) is a term that represents a variety of lymphomas with differing clinical presentations, histologic features, and therapeutic considerations. Mycosis fungoides (MF) and Sézary syndrome were the first recognized forms of CTCL, but we now recognize additional disease categories characterized by expansions of malignant T cells within the skin and unique combinations of clinical, histological, and immunophenotypical criteria. The incidence of CTCL was last examined extensively using data for the period 1973 through 1992 and including only cases of MF and Sézary syndrome. We investigated incidence trends for the entire group of diseases classified as CTCL in the Surveillance, Epidemiology, and End Results (SEER) program registries for the 30-year period from 1973 through 2002.

Our incidence data were derived from the 9 original registries of the SEER program of the National Cancer Institute for the period 1973 through 2002 and the 4 additional registries (Los Angeles and San Jose, California; Alaska; and rural Georgia) for the last 11 years of this period only. Together, these 13 registries cover approximately 14% of the US population; each registers patients who reside in their defined geographic area and who are diagnosed as having cancer (other than keratinocyte carcinomas). According to the International Classification of Diseases for Oncology (ICD-O), cases of CTCL were defined for the purposes of this analysis as those with primary site of skin (codes 440-449) and histologic types MF (morphology code 9700), Sézary syndrome (code 9701), mature T-cell lymphoma not otherwise specified (NOS) (code 9702), angioimmunoblastic T-cell lymphoma (code 9705), subcutaneous panniculitis–like T-cell lymphoma (code 9708), CTCL NOS (code 9709), and primary cutaneous CD30⁺ lymphoproliferative disorder (code 9718). Race was classified by the registries as white, black, or “other.” The data were analyzed using Stata SE version 8 (StataCorp, College Station, Texas) and SEER*Stat version 6.1.4 (National Cancer Institute, Bethesda, Maryland) statistical software. Time trends were evaluated by linear regression of year as a predictor of incidence rate. Confidence intervals (CIs) of rates were calculated using 1- and 2-sample proportion tests in Stata SE, and incidence rate ratios (IRRs) were calculated using the IRR calculator function of Stata SE. P values were calculated conventionally and are 2-tailed. Incidence rates were age-adjusted to the 2000 US standard million population unless otherwise indicated.

A total of 4783 cases of CTCL were identified for the period 1973 through 2002 among all 13 SEER registries. Of these cases, 97% were histologically confirmed and most (87%) were reported by hospitals and clinics. Most cases (n = 3430 [72%]) were histologically classified as MF. Other CTCL diagnoses included Sézary syndrome (n = 120 [2.5%]), primary cutaneous CD30⁺ lymphoproliferative disorder (n = 73 [1.5%]), subcutaneous panniculitis-like T-cell lymphoma (n = 10 [< 1%]), angioimmunoblastic T-cell lymphoma (n = 1 [< 1%]), mature T-cell lymphoma NOS (n = 172 [3.6%]), and CTCL NOS (n = 977 [20%]).

The overall annual age-adjusted incidence of CTCL for the original 9 SEER registries from 1973 through 2002 was 6.4 per million persons, representing 0.14% of all cancers other than keratinocyte carcinomas and 3.9% of all non-Hodgkin lymphomas in these registries.

Incidence increased by 2.9 × 10⁻⁶ (95% CI, 2.6-3.1 × 10⁻⁶) per decade (Table 1). This increase occurred at similar rates for all age groups and races. The increase was greater among men (3.5 × 10⁻⁶ [95% CI, 3.1-3.9 × 10⁻⁶] per decade) than among women (2.4 × 10⁻⁶ [95% CI, 2.1-2.7 × 10⁻⁶] per decade) and also varied by geographic location and period (Table 2).

Incidence was higher among blacks than among whites and lower among other racial groups (Table 1). The black-white IRR (1.5 [95% CI, 1.4-1.6]) did not vary significantly by sex, year or registry, but the ratio decreased by age (Table 3). The “other”-white ratio was 0.8 (95% CI, 0.7-0.9). Among the 13 SEER registries from 1992 through 2002, incidence was higher for non-Hispanics (8.3 [95% CI, 8.0-8.6] per million persons) than for Hispanics (5.8 [95% CI, 5.0-6.0] per million persons). This was true among both whites and nonwhites.

Men had significantly higher rates compared with women. The male-female IRR (1.9 [95% CI, 1.8-2.0]) increased with age from a low of 1.1 for individuals diagnosed before age 30 years to 2.1 for those diagnosed after age 60 years (Figure). The male-female IRR was lowest among blacks (1.5) and highest among the other racial groups (2.7). This ratio has decreased over time from 2.5 in the period 1973 through 1982 to 1.7 in 1993 through 2002 (Table 3).

There was a steep increase in incidence with age from 0.1 per million persons (0- to 9-year-olds) to 24.6 per million persons (70- to 79-year-olds) (Table 1). This increase with age was consistent over time.

Of all cases classified as CTCL, 4% (201 cases) were coded as having B-cell lineage on a separate variable that subclassified lymphomas by cell type (an apparent inconsistency in the SEER database), and 86% (172 cases) of those inconsistently classified cases were classified as CTCL NOS. The remainder of these cases were MF (17 cases) or another CTCL subcategory (12 cases). A similar proportion of cases were inconsistently classified for each of the 13 registries. Responses from 3 SEER registries to our specific inquiry about these inconsistencies indicate that many of these inconsistent cases may have been cutaneous B-cell lymphomas that were incorrectly categorized histologically owing to ambiguities in the second edition of the ICD-O (ICD-O-2) (see the “Comment” section). However, it is unclear from our data which of these cases were incorrectly coded by histologic type and which were coded incorrectly as having B-cell lineage.

Considerable variation in incidence exists across registries. Annual incidence was highest in San Francisco, California (9.7 per million persons among whites and 10.8 per million persons among blacks) and lowest in Iowa (3.7 per million persons among whites and 5.8 per million persons among blacks) (Table 2). Incidence in San Francisco was also highest among the 13 SEER registries for the period 1992 through 2002 (14.9 per million persons).

We looked at correlations between demographic characteristics of each of the 13 registries (Table 4) and the incidence among whites for the period 1992 through 2002. Incidence was correlated with the following: high physician density (r = 0.6; P = .04), high density of medical specialists (r = 0.7; P = .02), high median family income (r = 0.7; P = .01), high percentage of adults with a bachelor's degree or higher (r = 0.6; P = .02), and high median value of owner occupied housing units (r = 0.6; P = .02). These characteristics were also correlated with each other (Table 4). Incidence was not correlated with population density, percentage of the population who are white, percentage of the population who are foreign born, manufacturing or retail sales, percentage of the population living in the same house in 1995 and 2000, number of housing units, persons per household, and percentage of persons below the poverty level.

The overall incidence of Sézary syndrome was 0.3 per million persons. Men had significantly higher rates than women (0.26 [95% CI, 0.21-0.32] vs 0.09 [95% CI, 0.06-0.12]), and incidence among whites (0.36 [95% CI, 0.31-0.41] per million persons) was substantially higher than among blacks (0.04 [95% CI, 0.00-0.08] per million persons). All but 2 cases occurred in adults 30 years and older.

Coding for cutaneous lymphoma morphology has changed several times throughout the history of the SEER program (Table 5). For cases diagnosed between 1973 and 1985, histologic types are defined in the morphology section of the ICD-O (1976). These definitions changed in 1986, 1992, and 2001 with the introductions of ICD-O Field Trial Edition, ICD-O-2, and ICD-O-3, respectively. Each new ICD-O edition contained either additional histologic classifications or revisions to previous classifications for CTCLs (ICD-O code 970). These new classifications resulted in the redistribution of broadly defined cases of CTCL (code 970) among the more specific subclassifications (codes 9700-9709 and 9718). For example, the incidence of MF in Connecticut dropped from 6.7 per million persons in the period 1981 through 1991 to 1.0 per million persons in 1992 through 2002 after the adoption of ICD-O-2, which included the subclassification cutaneous lymphoma NOS (code 9709). The incidence of cutaneous lymphoma NOS in this registry over the same 2 periods was 0.2 and 8.5 per million persons, respectively.

We used recent and comprehensive data from the SEER program to update the epidemiological features of CTCL and document trends in incidence that differed from those of previous reports.^4,5 We noted a marked rise in incidence and substantial racial and geographic heterogeneity in rates of CTCL.

It is important to examine the incidence of all CTCLs as a group in the SEER program as opposed to the incidence for individual subgroups when analyzing trends over time because the classification changes resulting from the introduction of new ICD-O editions have caused sizeable shifts in the distribution of CTCLs among subcategories. These redistributions may have contributed to the apparent stabilization in the incidence of MF reported previously.⁵

Our study included data from 4 additional registries of the SEER program for the period 1992 through 2002, allowing us to further explore the geographic differences in incidence and to look at other ethnic characteristics such as the incidence among Hispanics. Nearly 3 times as many cases were included in our analysis compared with the most recent report⁵; hence, we were able to examine incidence patterns for various subgroups.

Some limitations of using the SEER program included the lack of independent verification of diagnoses and the lack of detail pertaining to each case. Each SEER registry is responsible for finding every case of cancer in its defined registry area. Nevertheless, because validity of the diagnoses was not determined, cases could have been included that were not actually CTCL, and cases of CTCL in the study registries may have been excluded from the data. Underregistration of cutaneous malignancies in SEER registries has been documented. Missed cases were estimated to constitute 17% of MF in one study,⁵ a figure that is similar to the proportion of invasive melanomas that may be missed.^6-9

Ambiguities in ICD-0 codes may have caused coding errors that also limit our data. These ambiguities in ICD-O-2 morphologic definitions may have resulted in the erroneous classification of several cases of cutaneous B-cell lymphoma as CTCL NOS. The ICD-O-2 (1992-2000) contained the disease classification “cutaneous lymphoma NOS (code 9709)” under the broader disease category “specified cutaneous and peripheral T-cell lymphomas (code 970).”³ The ambiguity created by not indicating T- or B-cell in the subclassification provides an explanation for our finding that approximately 4% of the total cases histologically classified as CTCL were noted as having B-cell lineage under the SEER variable “grade.” This explanation was supported by the review of several cases by individual registries on our request. The Los Angeles Cancer Registry reported that the inconsistency was the result of an error in the way that coders were interpreting the coding manual. The Atlanta registry reported that the grade codes were correct, but the histologic types were coded incorrectly. The Iowa Cancer Registry found that most grade codes were correct and that the inconsistencies were caused by ambiguous ICD-O-2 morphologic definitions.

The incidence of CTCL has risen dramatically and consistently since 1973. Changes in classification schemes may have contributed to the rise in incidence, as may improvements in detection or an increase in the underlying etiologic agent(s). Our demographic correlates show that incidence is strongly correlated with the density of physicians. Hence, the rise in incidence may be due, at least in part, to increased efficiency of detection resulting from improvements in medical care over the past few decades.¹⁰ Because of reporting delay, the actual rise in incidence may be greater than the rise found in our data. Reporting delay and reporting error occur when new cases are discovered or erroneous cases are detected in the existing SEER data. Clegg et al¹¹ found that initial incidence case counts accounted for only 88% to 97% of the estimated final counts in the SEER program and that it would take 4 to 17 years for 99% or more of cancer cases to be reported.

The geographic differences in incidence are substantial even after controlling for race. Incidence is correlated with high physician density and several indexes of socioeconomic status such as median family income, percentage of the population with a bachelor's degree or higher, and median home value. We did not confirm the previously reported correlation of incidence with population density.⁵ These geographic differences in incidence may be related, to some degree, by differences in access to medical care.

Several studies have suggested that infectious agents^12-14 or environmental exposures^15-17 may play a role in CTCL. While the unusual geographic differences seen in our data could be explained by an environmental or viral exposure, our data neither directly support nor contradict such hypotheses. The relatively high rates and unique incidence trends observed in the San Francisco registry are similar to incidence trends for both non-Hodgkin lymphoma and Kaposi sarcoma,¹⁸ 2 HIV-related cancers. However, unlike non-Hodgkin lymphoma and Kaposi sarcoma, a previous case-control study failed to find an increased risk for CTCL among never-married men.^19-21

There are significant differences in incidence by race, ethnicity, and sex. While racial differences in incidence decrease with age and have not changed over time, sex differences in incidence increase with age and have decreased over time. Races with low incidence had higher male-female ratios.

Sézary syndrome, a rare type of CTCL, is characterized by erythroderma, generalized lymphadenopathy, and the presence of neoplastic T cells.²² Little data have been reported on the epidemiologic features of Sézary syndrome owing to the rarity of its occurrence. As with CTCL, overall, we found this disease to be relatively more common among men than among women. However, contrary to CTCL, the incidence of this disorder is higher among whites than among blacks.

Our population-based study provides updated trends in the incidence of CTCL in the United States and describes patterns that have not been reported previously. These data may be useful in planning public health strategies, identifying risk factors, and understanding the etiology of this cancer so that it may some day be prevented.

Correspondence: Vincent D. Criscione, AB, Dermatoepidemiology Unit-111D, VA Medical Center, 830 Chalkstone Ave, Providence, RI 02908-4799 (Vincent_Criscione@Brown.edu).

Financial Disclosure: None reported.

Accepted for Publication: January 1, 2007.

Author Contributions:Study concept and design: Criscione and Weinstock. Acquisition of data: Weinstock. Analysis and interpretation of data: Criscione and Weinstock. Drafting of the manuscript: Criscione and Weinstock. Critical revision of the manuscript for important intellectual content: Criscione and Weinstock. Statistical analysis: Criscione and Weinstock. Obtained funding: Weinstock. Administrative, technical, and material support: Weinstock. Study supervision: Weinstock.

Funding/Support: This project was funded by a grant from the Cutaneous Lymphoma Foundation. Dr Weinstock was supported by grants CSP 402 from the Department of Veterans Affairs, Office of Research and Development, Washington, DC, and CA 106592 from the National Cancer Institute, Bethesda, Maryland.