INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Sarcoidosis is a granulomatous inflammatory disease, the cause of which remains largely unknown. A number of features (1) are suggestive of an infectious origin, others underscore the genetic susceptibility (4). In Scandinavian Caucasian patients, there is a strong correlation between specific T-cell receptor gene rearrangements, human leukocyte-associated antigen-DR17 (HLA-DR17), and good prognosis (8) indicating both the presence of a single antigen and the importance of host susceptibility.

Chronic inflammation is associated with an increased risk for malignant lymphomas (9) or cancer in the affected tissue (13). Theoretically this would apply also to sarcoidosis, which most frequently involves intrathoracic organs, the liver, and the skin. Little is known, however, about the cancer incidence following sarcoidosis. Case series have suggested an increased incidence of malignant lymphomas (16, 17). Previous cohort studies (18, 19) have not been able to confirm such an association, partly because of lack of power, but have instead indicated elevated risk for certain other malignancies.

The aim of this study was to assess the incidence of cancer following sarcoidosis, particularly non-Hodgkin's lymphoma and cancers in tissues often involved in sarcoidosis. The Swedish public health care system, high-quality registers, and homogenous population provide a unique setting for cohort identification and nondifferential follow-up. We therefore identified two cohorts of Swedish sarcoidosis patients from 1964 to 1994 which have been followed for cancer occurrence until 1995.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The Uppsala Cohort

This cohort has been described elsewhere (20) and consists of consecutively diagnosed cases of sarcoidosis at Uppsala University Hospital, Sweden, 1966-1980. Cases were identified either because of symptoms or at a regular health screening including chest X-ray and represent all detected cases of pulmonary sarcoidosis in patients above 15 yr of age in a defined catchment area, Uppsala County. For each patient, the National Registration Number (a 10-digit number unique to each citizen [21]), date of birth and diagnosis, mode of detection and diagnosis, corticosteroid treatment, and radiological stage (0-IV) at diagnosis and follow-up were recorded. Maximal radiological stage was defined as most advanced stage during four follow-up visits. A total of 504 patients were identified. Eighteen were excluded as the National Registration Number could not be unambiguously determined; six were lost to follow-up (Tables 1 and 2).

The Inpatient Register Cohort

All Swedish inpatient care is public and population-based. The Inpatient Register contains individual information on inpatient care since 1964. The registration is complete per county, and with increasing number of counties enrolled, the coverage has been nationwide since 1987. For each discharge, the National Registration Number, dates of admission and discharge, main (and up to five contributory) diagnosis, department, and hospital are recorded. In this register all individuals hospitalized for treatment or diagnostic investigations, with a discharge diagnosis of sarcoidosis 1964-1994 were identified in one cohort. In total, 11,698 patients were identified. 701 of these were excluded as data irregularities precluded unambiguous follow-up. Patients with sarcoidosis as a contributory rather than main diagnosis (1,918 patients) were also excluded (Table 1).

Follow-up

The National Registration Number permits linkage to the following population-based registers: The Cancer Register, The Register of Causes of Death, and the Register of Total Population. Through linkage, all incident cancers 1958-1995, deaths, and vital status at the end of the study period were obtained. Start of follow-up was set to date of diagnosis (Uppsala cohort) or date of first discharge (Inpatient cohort). End of follow-up was set to date of death, date of diagnosis of first cancer, or December 31, 1995, whichever occurred first.

Statistics

Standardized incidence ratios (SIR) using population-based rates of cancer incidence were calculated by dividing the observed number of cases with that expected according to sex, 5-yr age groups, and calendar period. Patients with cancer diagnosed prior to entry (six in the Uppsala cohort and 538 in the Inpatient register cohort) or occurring during the first year of follow-up (two and 146, respectively) were excluded. Cancers detected upon autopsy were not analyzed. Assuming a Poisson distribution among the cases, 95% confidence intervals (95% CI) were calculated (22). Trends were analyzed using the Poisson trend statistic (22).

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Uppsala Cohort

Fifty cases of cancer occurred, corresponding to a SIR of 1.2; 95% CI 0.9 to 1.6, higher among men (SIR = 1.5; 95% CI 0.9 to 2.2) than women (SIR = 1.1; 95% CI 0.7 to 1.6) (Tables 3 and 4). The risk estimates were similar in patients older and younger than 60 yr of age at diagnosis of sarcoidosis and among those diagnosed because of symptoms or by chance.

Five cases of lung cancer (three squamous cell cancers and two adenocarcinomas) occurred and although based on small numbers, there was a significantly elevated risk during 5 to 9 yr of follow-up (SIR = 5.7; 95% CI 1.2 to 17) but no increase thereafter. Four of the cases occurred in patients with stage II sarcoidosis, one in a patient with stage I (Table 4). Five cases of malignant melanoma occurred, SIR = 3.3; 95% CI 1.1 to 7.7, all after more than 10 yr of follow-up. There were three cases of nonmelanoma skin cancer (SIR = 2.6; 95% CI 0.5 to 7.5). As for melanoma, the risk was increased in both sexes. Two of the three cases had received corticosteroids. Four cases of liver cancer occurred (SIR = 3.9; 95% CI 1.1 to 9.9), the risk being equal among men and women and persistent over time. Two cases of breast cancer occurred versus 7.6 expected.

There were no cases of Hodgkin's disease but two cases of non-Hodgkin's lymphoma and one case of chronic lymphocytic leukemia, all in patients with stage II-IV sarcoidosis (Table 4).

Inpatient Cohort

In total, 653 cancers occurred corresponding to a SIR of 1.3; 95% CI 1.2 to 1.4 (Table 3). The risk was higher among men (SIR = 1.4; 95% CI 1.2 to 1.5) than women (SIR = 1.2; 95% CI 1.1 to 1.4) and had a tendency to decrease with time since first discharge (p_trend = 0.06) but to increase with age at first discharge (p_trend = 0.002). When restricting the analyses to patients younger than 60 yr of age at entry, the SIR = 1.2; 95% CI 1.1 to 1.3. After exclusion of cancers earlier associated with sarcoidosis (cancers of the oral cavity, liver, lung, malignant melanoma, nonmelanoma skin cancer, and lymphoma) (18, 19), the residual SIR = 1.1; 95% CI 1.0 to 1.3. When the analysis was restricted to those individuals who, on an inpatient basis, underwent bronchoscopy or mediastinoscopy during the year that preceded entry, no difference in overall cancer risk could be found.

The risk for lung cancer was doubled during the first 10 yr of follow-up but decreased to a significant deficit (SIR = 0.5; 95% CI 0.2 to 0.9) thereafter. In a separate analysis, only patients treated at departments of pulmonary medicine (n = 3,459) were included. Their overall relative risk for lung cancer (SIR = 1.6; 95% CI 1.0 to 2.3) was higher than that of the remaining patients of the cohort, but displayed the same temporal variation. Patients undergoing (inpatient) broncho- or mediastinoscopy before the diagnosis did not differ from the remainder of the cohort.

Skin cancer occurred more frequently than expected (SIR melanoma = 1.6; 95% CI 1.1 to 2.3 and SIR nonmelanoma = 2.8; 95% CI 2.0 to 3.8). Elevated risks were seen in both sexes and in all latency intervals. The risk for stomach cancer was elevated (SIR = 1.7; 95% CI 1.1 to 2.3) in all latency intervals. Six cancers of the small intestine (four carcinoids) occurred (SIR = 2.7; 95% CI 1.0 to 5.9) and the risk for colorectal cancer was moderately increased. Initially, no increased risk for liver cancer was evident, but after more than 10 yr of follow-up the SIR = 1.9; 95% CI 0.9 to 3.4. Breast cancer occurred as expected as did cancers of the uterus, ovary, prostate, and testis. Bladder cancer occurred less often than expected, SIR = 0.7; 95% CI 0.4 to 1.2.

For lymphomas, the excess risk for Hodgkin's disease was confined to the first decade following entry into the cohort. The risk for non-Hodgkin's lymphoma (29 cases) displayed the same pattern, i.e., an initially increased risk which after more than 10 yr of follow-up equaled unity. Restricting the analyses to patients treated at departments of pulmonary medicine yielded similar estimates: SIR 1 to 4 yr = 3.6, 95% CI 1.2 to 8.3; SIR 5 to 9 yr = 2.9, 95% CI 0.9 to 6.7; and SIR 10+ yr = 1.2, 95% CI 0.3 to 3.0. Patients undergoing (inpatient) broncho- or mediastinoscopy prior to the diagnosis had a somewhat higher relative risk, which persisted after more than 10 yr of follow-up. Leukemia also occurred more often than expected, particularly myeloid leukemia (SIR = 2.4; 95% CI 1.2 to 4.2), the risk being independent of follow-up time.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Patients with sarcoidosis appear to be at increased risk for cancer, particularly for cancers of the lung, stomach, small intestine, and liver, for melanoma and nonmelanoma skin cancer, non-Hodgkin's lymphoma, and leukemia. The estimated overall relative risk for cancer is equal to or slightly higher than that seen in other chronic diseases such as diabetes (23), inflammatory bowel disease (24), and rheumatoid arthritis (25). The strengths of our study lie in the large number of patients from an ethnically homogenous study population, the long follow-up, the nondifferential ascertainment of outcome through population-based registers, the ability to study incidence rather than mortality, and the possibility to compare findings in the two cohorts.

Sarcoidosis is sometimes diagnosed during examination for other reasons. In the Inpatient cohort, increased cancer occurrence resulting from underlying morbidity could therefore have led to spurious associations. Therefore, the Inpatient cohort was restricted to patients discharged with sarcoidosis as a main (rather than contributory) diagnosis. This had little influence on the age-specific risk estimates but lowered the proportion of patients older than 60 yr of age at entry. Further restrictions (e.g., to patients treated at internal or pulmonary medicine departments or who underwent broncho- or mediastinoscopy on an inpatient basis) did not affect the results. Because the relative risks in the Uppsala cohort, with a high proportion of screening-detected cases, were similar to that of the Inpatient cohort, both in magnitude and in distribution in different cancer sites, latency intervals, and sex, substantial impact of confounding from underlying morbidity remains unlikely.

Increased medical surveillance of sarcoidosis patients could also have biased our results. Scandinavian sarcoidosis is, however, often self-healing, which explains why most patients are only followed for a few years. Furthermore, most of the studied cancers are, if untreated, lethal and to diminish the risk for surveillance bias, cancers detected during autopsy were excluded from the analyses.

Sarcoidosis following malignancy could theoretically also affect the results. Therefore, patients with a cancer diagnosis antedating entry into the cohorts were excluded and follow-up terminated at first cancer. To minimize the risk for misclassification of sarcoid reactions or prevalent cancers as sarcoidosis, cancers occurring during the first year of follow-up were excluded. In terms of diagnostic accuracy, inclusion of the Uppsala cohort provided a cohort with high exposure validity.

In lung cancer, misdiagnosis of sarcoidosis could be part of the explanation for the increased risk during the first years of follow-up. Considering the poor prognosis in lung cancer, such misclassification is, however, an unlikely explanation for the increased risk, for instance, after 5 to 9 yr of follow-up. Even though lung cancer may have a long preclinical period, this period is unlikely to exceed, for example, 5 yr counted from the presentation with (bilateral) hilar enlargement initially misclassified as sarcoidosis. We cannot, however, completely exclude this possibility. Because sarcoidosis appears to be inversely related to smoking (26), the 50% reduction in risk thereafter most likely reflects a lower proportion of smokers in the cohort than in the general population, which is further supported by the, albeit nonsignificant, reduced risk for bladder cancer. The consistency in magnitude and time between our two cohorts, and one of the existing cohort studies (19), argues against a chance finding. It appears therefore as if sarcoidosis is associated with a marked but transiently increased risk for lung cancer, especially in patients with parenchymal involvement. This is in line with the increased cancer risk seen in other inflammatory respiratory disorders (29) and it may be the associated regeneration of cells rather than sarcoidosis per se that mediates the risk.

Also for malignant lymphomas, misdiagnosis is a likely explanation for the observed initial increase in risk. Bearing in mind the poor prognosis following diagnosis of malignant lymphoma, such misclassification is likely to be confined to the first years of follow-up. Although the true risk during the first years after entry may be uncertain, the nearly doubled risk during 5 to 9 yr after diagnosis is, however, unlikely to be explained by misclassification. The large proportion of self-healing sarcoidosis precludes more detailed comparisons with other, chronic, inflammatory conditions (10, 11, 25) but our results do not contradict an association with inflammatory activity. An alternative explanation is that the agent causing sarcoidosis also confers a transiently increased risk for malignant lymphomas.

The increased risk for melanoma and nonmelanoma skin cancer was present in virtually all subsets of patients studied and corroborates earlier findings (17) but is higher than the risk found in other surveyed cohorts (23, 25). Nonmelanoma skin cancer is associated with immunosuppressive therapy but malignant melanomas are not (30). Skin sarcoidosis is often associated with progressive disease (31) and may heal with moderate scarring. The skin malignancies could have arisen from such lesions. Both types of skin cancer are, however, associated with ultraviolet light and sarcoidosis could somehow be associated with exposure to sunlight. In comparison with lung cancer, the persistency of the risk increase argues against common carcinogenic pathways.

In sarcoidosis, liver granulomas can often be found (32). The increased risk for liver cancer is therefore interesting, and supported by one of the two earlier cohort studies of sarcoidosis (18). Liver inflammation is a well-known risk factor for hepatocellular cancer (33) and could be the underlying mechanism also in sarcoidosis. Misclassification of primary biliary cirrhosis is unlikely because none of the patients who developed liver cancer in the Uppsala cohort had undergone liver biopsy. The incidence of stomach cancer in Sweden displays a geographic gradient and the moderately increased risk in the Inpatient cohort is probably explained by the higher than expected proportion of cases diagnosed in such high-incidence areas.

Although based on small numbers, the increased occurrence of cancer of the eye, nasal sinuses, and buccal cavity merits mentioning because these sites are often affected in sarcoidosis. Both earlier cohort studies of sarcoidosis also found elevated risks for lip and pharynx cancer (18, 19).

In conclusion, patients with sarcoidosis appear to be at significantly increased risk for cancer, in particular lung cancer, malignant lymphomas, and cancer in other organs known to be affected in sarcoidosis. Therefore, despite the often favorable prognosis in sarcoidosis, these patients deserve increased medical attention. Although the underlying mechanism is unclear, chronic inflammation may be a putative mediator.