The Danish Cancer Registry, which covers the entire Danish population (approximately 5.1 million persons in 1991), has kept records of all notifications of nonmelanoma skin cancer and other types of cancer for more than half a century. Since 1978, when the International Classification of Diseases for Oncology (ICD-O) was introduced as the coding system of the Cancer Registry [7], it has been possible to distinguish between basal-cell carcinoma and the other, less-common types of nonmelanoma skin cancer.

By using population-based data available in the Danish Cancer Registry, we aim to provide valid and statistically stable estimates of the subsequent risk for cancer in patients who receive a diagnosis of basal-cell carcinoma of the skin. Because of the recently proposed hypothesis that ultraviolet light is involved in the steep increase in incidence of non-Hodgkin lymphoma [4, 8], we devoted special attention to the possible association between basal-cell carcinoma (a disease known to be caused by exposure to ultraviolet light [9]) and the risk for subsequent non-Hodgkin lymphoma.

Methods

Top Methods Results Discussion Author & Article Info References

The procedure for reporting skin cancer to the Danish Cancer Registry is the same as the procedure for reporting other types of cancer. Physicians in hospitals and outpatient settings are required to report all incidents of cancer to the registry. To ensure a high degree of completeness of data in the registry, computerized identity-secure links with other data sources (established by means of a 10-digit personal identity number unique to every Danish citizen) are done at regular intervals to fill in missing data. An annual link is done with the Danish Hospital Discharge Register to ensure that any diagnoses of cancer that have been made in hospitalized patients and that have not been reported to the cancer registry are subsequently included in the cancer files. Similarly, an annual link with the computerized death file of the National Board of Health ensures that the small proportion of cases of cancer known only from death certificates is included [7]. By these means, and because of the thorough scrutiny and centralized coding of all notifications, the accuracy of the Cancer Registry data can be considered to be high. For most types of cancer, the data are believed to be almost 100% complete [10].

We considered a case of cancer to be basal-cell carcinoma if topography, according to the ICD-O [11], was between 1731 and 1739 (skin, excluding lip and external genitalia) and histology was between 80903 and 80933 (variants of basal-cell carcinoma). The files of 37 756 patients with a histologically verified first diagnosis of basal-cell carcinoma were retrieved from the Cancer Registry for 1978 to 1991. The anatomical distribution is shown in Table 1. Some patients had two or more histologically identical basal-cell carcinomas diagnosed either simultaneously or at different times; the interval between these diagnoses differed for each patient. In the registry, such multiple identical cases of skin cancer are coded as only one basal-cell carcinoma (topography code 1738), and patients who have more than one type of cancer retain the date of the first type as the date of diagnosis. In one patient, two basal-cell carcinomas with slightly different histologic characteristics were diagnosed. Such cancers are coded individually, and this patient's date of diagnosis was the date of the first skin cancer. Eighty-two patients (0.2%) were excluded from the analysis: Twenty-eight (15 men and 13 women) were not at risk for subsequent cancer because their skin cancer was diagnosed at autopsy or because they died before the end of the month in which basal-cell carcinoma was diagnosed; 9 (3 men and 6 women) received their diagnosis after they were 100 years of age; and 45 (25 men and 20 women) received a diagnosis of basal-cell carcinoma after they had emigrated from Denmark.

The remaining 37 674 patients (median age, 68 years [range, 8 to 99 years]) were followed for the occurrence of subsequent cancer. We did stratified analyses of the risk for cancer to evaluate the possible importance of various characteristics of the patients, including sex, age (patients were stratified into those <60 and those 60 years of age at the time of diagnosis of basal-cell carcinoma), the anatomical site of the initial skin cancer, and the number of basal-cell carcinomas (patients were stratified into those with only one basal-cell carcinoma and those with two or more). In each analysis, patients were followed from the month after the date of diagnosis of basal-cell carcinoma to 100 years of age, death, emigration, or 31 December 1991, whichever came first. Information about the end of follow-up was obtained through a link with the Civil Registration System, which keeps a continuously updated file on demographic variables (including death and emigration) for all citizens in Denmark. The sex-, age-, and period-specific patient-years of observation in the cohort (in 5-year blocks) were multiplied by the corresponding sex-, age-, and period-specific national incidence rates to determine the expected number of cases of cancer in the cohort. Ratios of actual to expected number of cases of cancer (standardized incidence ratios [SIRs]) measured the relative risk. Assuming a Poisson distribution of the observed cancer events, we calculated 95% CIs for the SIRs using Byar limits [12]. Likelihood ratio tests were done to determine whether SIRs among young and old patients were statistically different at various cancer sites.

Results

Top Methods Results Discussion Author & Article Info References

We followed 18 925 men for 94 162 patient-years (median duration of follow-up, 4.2 years) and 18 749 women for 96 783 patient-years (median duration of follow-up, 4.4 years). After a diagnosis of basal-cell carcinoma, 2234 new cases of cancer were diagnosed in men and 1429 new cases of cancer were diagnosed in women within a maximum of 14 years (Table 2). Compared with the incidence of cancer in the general Danish population, this corresponded to SIRs of 1.18 (95% CI, 1.13 to 1.23) in men and 1.05 (CI, 1.00 to 1.11) in women. When subsequent cutaneous cancer, such as malignant melanoma (n = 134) and other skin cancer (mainly squamous-cell skin cancer, n = 411), were excluded from the analysis, the risk for subsequent cancer in patients with basal-cell carcinoma was almost unchanged (SIR, 1.19 [CI, 1.13 to 1.24] in men and 1.09 [CI, 1.03 to 1.16] in women). Certain types of cancer occurred in statistically significant excess, including cancer of the lip (SIR, 2.07; n = 39), salivary glands (SIR, 2.45; n = 12), larynx (SIR, 1.41; n = 41), lung (SIR, 1.40; n = 589), breast (SIR, 1.13; n = 283), and kidney (SIR, 1.30; n = 114) and malignant melanoma (SIR, 2.64; n = 134), non-Hodgkin lymphoma (SIR, 1.36; n = 87), and leukemia (SIR, 1.34; n = 104) (Table 2).

To evaluate the possible effect of cancer and treatment of cancer that may have occurred before basal-cell carcinoma of the skin, we did a supplementary analysis of 34 146 patients in whom basal-cell carcinoma was their first cancer. The overall risk for all types of subsequent cancer in these patients (SIR, 1.18 for men [observed incidence, 2052] and 1.03 for women [observed incidence, 1264]) did not differ from the pattern seen for the entire cohort. The similarity pertained to all sites at increased risk, including cancer of the lip (SIR, 2.02; n = 35), salivary glands (SIR, 2.46; n = 11), larynx (SIR, 1.49; n = 40), lung (SIR, 1.37; n = 530), breast (SIR, 1.13; n = 254), and kidney (SIR, 1.37; n = 110) and malignant melanoma (SIR, 2.70; n = 126), non-Hodgkin lymphoma (SIR, 1.25; n = 73), and leukemia (SIR, 1.30; n = 92). Consequently, because previous cancer and its treatment had no measurable effect on the risk for cancer after the basal-cell carcinoma, all subsequent analyses were done for the entire cohort of 37 674 patients, regardless of the presence or absence of cancer before the basal-cell carcinoma.

The cohort was split into two groups according to the age at which the patients first received a diagnosis of basal-cell carcinoma. Patients who were younger than 60 years of age at the time of diagnosis of basal-cell carcinoma had a significantly higher relative risk (P < 0.01) for subsequent cancer (SIR, 1.26 [CI, 1.16 to 1.37]; n = 583) than did patients 60 years of age or older (SIR, 1.11 [CI, 1.07 to 1.14]; n = 3080) (Table 3). The moderate but significant excess of breast cancer was entirely caused by an elevated risk in younger patients. Of those who were younger than 60 years of age at the diagnosis of basal-cell carcinoma, 84 patients (of whom 1 was male) developed breast cancer where approximately 61 cases of breast cancer were expected (SIR, 1.37 [CI, 1.09 to 1.70]). In patients 60 years of age or older, the corresponding SIR for breast cancer was 1.05 (CI, 0.91 to 1.24) (P < 0.05). Cancer of the lip also occurred more frequently in younger (SIR, 2.73) than in older patients (SIR, 1.97), as did salivary gland cancer (SIR, 3.75 compared with 2.19) and primary liver cancer (SIR, 2.13 compared with 1.03). No cases of testicular cancer occurred in patients 60 years of age or older, although approximately two cases were anticipated. In contrast, 9 patients younger than 60 years of age developed testicular cancer (SIR, 3.50 [CI, 1.50 to 6.65]) (P < 0.01).

Many cases of malignant melanoma occurred in younger and older patients, but the association was moderately stronger for younger (SIR, 3.22) than for older patients with basal-cell carcinoma (SIR, 2.44). The excess of non-Hodgkin lymphomas occurred primarily in younger patients (SIR, 2.50); no convincing association with non-Hodgkin lymphoma was present in patients older than 60 years (SIR, 1.16 [CI, 0.89 to 1.48]) (P < 0.01) (Table 3). In contrast, older patients with basal-cell carcinoma had higher SIRs for other types of systemic cancers, such as Hodgkin lymphoma and leukemia, than did those in whom basal-cell carcinoma was diagnosed before 60 years of age (Table 3). Age was not associated with cancer of the respiratory organs or kidneys.

The risk for subsequent cancer differed between 7610 patients who had two or more basal-cell carcinomas during the study period and those with only one. Patients with basal-cell carcinoma had an elevated risk for subsequent cancer at sites other than the skin Table 2; those with two or more basal-cell carcinomas appeared to have a reduced risk for subsequent cancer at sites other than the skin (SIR, 0.89 [CI, 0.82 to 0.97]). This deficit appeared because fewer subsequent cancers than expected were seen at most sites, including digestive organs (SIR, 0.73 [CI, 0.62 to 0.86]; n = 143), respiratory organs (SIR, 0.91 [CI, 0.75 to 1.11]; n = 105), male genital organs (SIR, 0.89 [CI, 0.68 to 1.14]; n = 61), the urinary system (SIR, 0.92 [CI, 0.71 to 1.17]; n = 68), and the brain and nervous system (SIR, 0.93 [CI, 0.48 to 1.63]; n = 12). Only lip cancer (SIR, 2.62 [CI, 1.35 to 4.58]; n = 12) and malignant melanoma (SIR, 4.21 [CI, 3.14 to 5.54]; n = 51) occurred more often in patients with several basal-cell carcinomas than in patients with only one such skin cancer.

Dividing the patients into subgroups according to the anatomical site of their basal-cell carcinoma of the skin showed elevated risks for cancer in all subgroups. However, patients with a basal-cell carcinoma on an upper extremity had a very high risk for developing malignant melanoma (SIR, 6.20 [CI, 2.48 to 12.76]; n = 7), as did patients whose basal-cell carcinoma of the skin was located on a lower extremity (SIR, 3.70 [CI, 1.19 to 8.64]; n = 5). Similarly, the highest risks for subsequent non-Hodgkin lymphoma were found in the subgroups that had an initial skin lesion on the arm (SIR, 2.13 [CI, 0.43 to 6.22]; n = 3) or the leg (SIR, 3.66 [CI, 1.34 to 7.96]; n = 6), although these associations were unstable because the number of cases was small. In contrast, the risk for subsequent cancer other than skin cancer and non-Hodgkin lymphoma did not predominate in the subgroups that had basal-cell carcinoma of the skin located on the arm (SIR, 1.16) or the leg (SIR, 1.24).

Discussion

Top Methods Results Discussion Author & Article Info References

Patients with basal-cell carcinoma of the skin are at increased risk for development of new cancer. In addition to having a well-known excess risk for development of new skin cancer, these patients also have an increased risk for subsequent development of various noncutaneous types of cancer.

Only a few recent studies [1-3] have reported on the subsequent risk for cancer at sites other than skin in patients with basal-cell carcinoma. In their follow-up study of 657 patients, Chute and colleagues [1] found 43 cases of cancer where the expected number was 44.5 (SIR, 1.0 [CI, 0.7 to 1.3]). Unfortunately, the authors did not present detailed site-specific analyses. A study of 468 Swedish patients [2] also failed to find a statistically elevated risk for additional cancer after basal-cell carcinoma of the skin: Fifty-four cases of cancer were seen where 51.0 had been expected (SIR, 1.1 [CI, 0.8 to 1.4]). A larger study from Sweden [3] followed 1973 patients for 12 867 patient-years and found 221 subsequent noncutaneous cancers where 177.9 had been expected. This corresponds to a statistically significantly increased SIR of 1.24 (CI, 1.08 to 1.42), which is close to the estimate at which we arrived. Unfortunately, anatomical details about the cancers were not presented. The limited power of these previous investigations to detect small increases in the risk for various types of cancer should be recognized. Together, the patient-time of these three studies equals only 10% of the patient-time of our study.

Basal-cell carcinoma is increasing in incidence and has become the most commonly diagnosed cancer in industrialized parts of the world [13]. The public health problem associated with diagnosing, treating, and following cases of basal-cell carcinoma has become considerable [9]. Nevertheless, most population-based cancer registries do not include basal-cell carcinomas in their files. To our knowledge, our study is the first to describe the pattern of incidence of cancer in a large and truly population-based cohort of patients followed for as long as 14 years after a diagnosis of basal-cell carcinoma.

Nevertheless, certain inherent limitations must be considered in this kind of register-based research. Because our investigation involved no direct contact with patients, the data available for investigation included only those that were collected as part of the notification routines. Records of some exposures and lifestyle variables (such as smoking habit, concurrent diagnoses, socioeconomic variables, and reproductive factors) that may be important to the cancers studied are not contained in the Cancer Registry. Consequently, we could not address the degree to which factors that are not registered could explain the excess risk for cancer among patients with basal-cell carcinoma. This inability to adjust for potential confounding factors should be recognized so that unjustified conclusions are avoided.

Despite the high quality of the data for most types of cancer in the Danish Cancer Registry, the potential problem of underreporting of nonmelanoma skin cancer should be considered. Unlike most internal cancer, for which patients are notified primarily from hospital departments through well-functioning notification routines, many reports of nonmelanoma skin cancer come from primary-care physicians [14], whose notification procedures may not be as thorough as those of hospitals. Thus, an unknown degree of underreporting of nonmelanoma skin cancer is likely to occur.

Surveillance bias is another concern that should be considered as a possible explanation for the associations. Theoretically, the modest yet statistically significant increase in the risk for subsequent cancer could be at least partly explained by an increased clinical alertness devoted to patients treated for skin cancer. However, in a large follow-up study of all 379 941 Danish patients in whom cancer other than nonmelanoma skin cancer was diagnosed and reported to the Danish Cancer Registry between 1943 and 1980, even fewer new cases of cancer occurred than were expected (SIR, 0.91 [CI, 0.90 to 0.92]) [15]. It is not plausible that patients treated for basal-cell carcinoma lesions that clinically resemble benign tumors should be subject to more thorough follow-up for internal cancers than are patients with diagnoses of more severe cancer. Consequently, it is unlikely that the excess risk seen among patients with basal-cell carcinoma should be the result of increased surveillance.

The counterintuitive observation that patients who had more than one basal-cell carcinoma were at a reduced risk for subsequent cancer compared with patients who had only one such lesion is probably the result of an overestimation of the patient-time at risk for those patients with two or more skin cancers. As described previously, follow-up for patients with two or more identical basal-cell carcinomas (those with ICD-O topography code 1738) began when their first skin cancer was diagnosed. Therefore, most patients who had several basal-cell carcinomas of the skin and had originally had only one such cancer (initially coded as ICD-O 173X, with the X depending on the anatomical location) survived for a variable period before the topography code was changed to 1738 as the result of a new, histologically identical basal-cell carcinoma. By definition, these patients had an artificially reduced risk for developing lethal cancer in the time between development of the first and development of the second basal-cell carcinoma of the skin.

The risk for non-Hodgkin lymphoma was markedly elevated in patients younger than 60 years of age but not in those 60 years of age or older when basal-cell carcinoma was diagnosed. In contrast, the other types of systemic cancers (Hodgkin lymphoma and leukemia) occurred less frequently in younger patients. Persons who are genetically predisposed to basal-cell carcinoma of the skin or who are most heavily exposed, at an early age, to exogenous factors causing this tumor may also be predisposed to non-Hodgkin lymphoma. Although the hypothesis remains controversial [16], it has recently been proposed that ultraviolet light is a common causal factor that might explain the synchronous increases in the incidence rates of non-Hodgkin lymphoma and skin cancer [4, 6, 8]. One indication that such common risk factors may exist was our finding of a parallel occurrence of subsequent non-Hodgkin lymphoma and malignant melanoma subsequent to basal-cell carcinoma of the skin. These subsequent cancers occurred more often in young patients than in old patients, and the risk for either cancer was higher in patients whose initial skin cancer was located on the extremities than in patients with basal-cell carcinoma located elsewhere.

Our study confirms previous findings of a close association between nonmelanoma skin cancer and cancer of the salivary glands [5, 6, 17, 18]. The mechanism behind this association is not known, but because the salivary glands and the skin emerge from the ectodermal layer of the fetus, a common susceptibility toward external risk factors has been proposed [19].

As do younger (<60 years of age) patients with squamous-cell skin cancer [6] or anal cancer [20], younger patients with basal-cell carcinoma of the skin had an elevated risk for developing breast cancer in our study. In addition, as with malignant melanoma and non-Hodgkin lymphoma, higher SIRs were present for lip cancer, salivary gland cancer, primary liver cancer, and testicular cancer in younger patients with basal-cell carcinoma. The explanation for this age relation is not clear, but some genetic predisposition might exist in patients who are younger when they receive a diagnosis of skin neoplasms that usually occur later in life.

The finding of an elevation of approximately 40% in risks for laryngeal cancer and lung cancer merits consideration. The prevailing evidence suggests that smoking is not a risk factor for basal-cell carcinoma. One large cohort study [21] found no association between heavy cigarette smoking ( 25 cigarettes per day) and the risk for this type of skin cancer. Another study [22] suggested that the risk for new identical lesions in patients with one or more previous basal-cell carcinoma of the skin was even lower in heavy smokers than in nonsmokers. Therefore, rather than being a genuine risk factor for basal-cell carcinoma of the skin, smoking may be related to a causal factor in a way that generates confounding. If persons exposed to ultraviolet light tend to smoke more heavily than those who are not exposed, a spurious association between smoking-related phenomena and basal-cell carcinoma may occur. This might explain the association between basal-cell carcinoma of the skin and both laryngeal cancer and lung cancer.

Theoretically, confounding could also have contributed to the excess risk for other types of cancer in these patients if exposures related causally to the skin cancer and to cancers that subsequently occurred in excess had been present before the date of diagnosis. However, in comparing the subsequent incidence of cancer in the entire cohort with that of patients for whom basal-cell carcinoma was their first cancer, we found no indication that cancer or treatment of cancer before the basal-cell carcinoma of the skin developed had any effect on the subsequent risk for cancer. Consequently, we believe that the pattern of subsequent incidence of cancer is not confounded by such treatment-related carcinogenic exposures as alkylating chemotherapy or radiography before the appearance of basal-cell carcinoma.

Although the statistical power of our study may appear impressive, additional large-scale cohort studies must be done in other settings to assess, with the least risk for bias, the subsequent risk for noncutaneous cancer in patients with basal-cell carcinoma of the skin. Our study supports the well-documented notion that patients with this type of skin cancer should be followed at regular intervals for the occurrence of new cutaneous cancer. However, these patients appear to constitute a group at high risk not only for new skin tumors but also for cancer at other sites. Nevertheless, the increase in absolute risk for noncutaneous cancer does not appear to be large enough to justify general screening procedures for internal cancer in patients with basal-cell carcinoma. Rather, although breast and scrotal symptoms should always be carefully examined, it seems prudent to advocate that such symptoms be taken particularly seriously in younger patients who have previously had basal-cell carcinoma in order to diagnose breast cancer or testicular cancer at an early stage. In addition, when a younger patient with a history of basal-cell carcinoma of the skin presents with noninfectious lymph node enlargement, thorough examination should be done to exclude the possibility of non-Hodgkin lymphoma.

Dr. Olsen: Danish Cancer Society, Division for Cancer Epidemiology, 49 Strandboulevarden, DK-2100 Copenhagen O, Denmark.