Most follow-up studies of patients with rheumatoid arthritis showed no increased cancer incidence other than a moderately increased risk (about twofold) for lymphoproliferative malignancies [3-6], but none of these studies focused on the Felty syndrome. Other than one patient with non-Hodgkin lymphoma [7], the only type of malignancy consistently linked with the Felty syndrome is T-cell chronic lymphocytic leukemia in a small group of patients with the Felty syndrome who also had the large granular lymphocyte syndrome [8]. We used Veterans Affairs hospital charts and a large computerized database of patient information from Veterans Affairs hospitals to systematically evaluate patients with the Felty syndrome for the subsequent incidence of all cancers, with special attention to lymphoproliferative malignancies.

Methods

Top Methods Results Discussion Author & Article Info References

The study cohort was selected from computerized discharge records (the patient treatment file) for inpatient visits at Veterans Affairs hospitals across the United States. Each patient treatment file record contains up to 10 diagnostic codes, in either the 8th or 9th revision of the International Classification of Diseases. Of 55 000 veterans at least 18 years of age with rheumatoid arthritis, 1135 had the Felty syndrome (code 712.1 in the 8th revision or 714.1 in the 9th revision) listed as a diagnostic code on a discharge record (patient treatment file) between 1 July 1969 and 30 September 1990. The final cohort consisted of 906 white men after the following exclusions: 36 nonwhites, 28 women, 95 men who had a discharge diagnosis of any cancer on or before their first admission for the Felty syndrome (index visit), and 70 men who died during their index visit. Patients who died or had cancer diagnosed during their first visit contributed no person-years and were therefore excluded. Only cancers diagnosed subsequent to the Felty syndrome were considered "incident."

Although the prevalence of rheumatoid arthritis was 0.6% among black veterans and was 1.0% among white veterans in the patient treatment file, the Felty syndrome was diagnosed in only 0.5% of the black men with rheumatoid arthritis compared with 2.2% of the white men with rheumatoid arthritis. Among the black men with the Felty syndrome, only two developed cancer (Hodgkin disease and stomach cancer).

Statistical analyses used standardized incidence ratios (SIR) (ratios of observed-to-expected events) to estimate the risk for specific cancers. The expected numbers of cancers were determined by two methods, one applying internal cancer rates (based on the full Veterans Affairs hospital database), and the other using external cancer rates from the Surveillance, Epidemiology, and End Results (SEER) Program of the National Cancer Institute [9]. The internal rates were obtained by calculating person-years for the four million white men (discharged from Veterans Affairs hospitals between 1 July 1969 and 30 September 1990) from the first discharge to the earliest of the following four dates: first cancer diagnosis (hospital admission), death, 100th birthday, or 30 September 1990 (for censoring). Mortality information from the Social Security Administration was used to truncate the person-years of patients who died outside a Veterans Affairs hospital (we estimate our mortality data to be about 90% complete). Only patients whose cancer was diagnosed in a Veterans Affairs hospital were ascertained. The remaining patients were assumed to be alive and cancer-free. The internal cancer rates were then computed within 5-year age and calendar-time intervals and were computed separately for each type of cancer. Similarly, the external cancer rates were age, calendar-time, cancer-type, race, and sex specific and were calculated for the nine geographically defined SEER Program areas.

Person-years at risk for patients with the Felty syndrome were accrued in the same way, starting from their first date of discharge with a diagnosis of the Felty syndrome and continuing until first cancer admission, death, age 100, or 30 September 1990, whichever date was earliest. The expected number of cancers was calculated by applying internal and external cancer rates to the age and calendar-time distribution of person-years at risk for patients with the Felty syndrome. Standardized incidence ratios based on internal and external rates were computed as the ratio of observed-to-expected events. This procedure is analogous to comparing the rate of cancer in patients with the Felty syndrome with the rate of cancer in patients without the Felty syndrome. Ninety-five percent CIs were computed by the exact procedure of Liddell [10].

Hospital chart review was attempted for all 19 patients with the Felty syndrome who had non-Hodgkin lymphoma and for 19 cancer-free patients with the Felty syndrome. These 19 cancer-free patients were controls from the record abstract study of rheumatoid arthritis who happened to have the Felty syndrome as a hospital diagnosis on computer records (they were matched for race, year of birth, and year of first hospital diagnosis of rheumatoid arthritis to patients with rheumatoid arthritis who had hematopoietic cancer). All charts were reviewed to confirm the relevant diagnoses (the Felty syndrome, cancer, and other immune disorders); to verify dates of diagnoses, duration, and severity of rheumatoid arthritis; and also to determine treatment (alkylating and immunosuppressive agents, methotrexate, gold, phenylbutazone, and radiation therapy). The number of unconfirmed patients with the Felty syndrome or cancer is noted in the text, but these patients were not excluded from the statistical tables.

Results

Top Methods Results Discussion Author & Article Info References

The 906 white male veterans with the Felty syndrome were followed up to 20 years, with an average of 6 years per patient. Their ages at the first Veterans Affairs hospitalization for this condition ranged from 35 to 85 years, with an average age of 61 years.

Table 1 shows the estimated cancer risks for all cancer sites with more than five observed cases. A twofold increase in total cancer was noted, whether compared with internal or with external rates. Rates for all common cancer types were increased except for prostate cancer. The internal and external risk estimates were always in the same direction, although the magnitude varied. For all cancers except lung and buccal cancer, the external risk estimates were lower. Four cancer types were clearly increased when compared with Veterans Affairs and SEER Program rates: non-Hodgkin lymphoma, leukemia, malignant melanoma, and lung cancer. Excess rates of multiple myeloma were also noted (3 patients, SIR = 4.63; 95% CI, 0.9 to 13.5) and Hodgkin disease (1 patient, SIR = 4.04; CI, 0.1 to 22.5), but these were based on small numbers of patients. Of the 13 patients with leukemia, 4 had acute nonlymphocytic (SIR = 8.71; CI, 2.3 to 22.3), 2 had chronic lymphocytic (SIR = 3.38; CI, 0.4 to 12.2), 2 had chronic myeloid (SIR = 8.00; CI, 0.9 to 28.9), 1 had lymphocytic leukemia not otherwise specified (SIR = 8.62; CI, 0.2 to 48.0), and 4 had leukemias not otherwise specified (SIR = 14.3; CI, 3.8 to 36.6).

Table 2 shows the cancer risks stratified by years since first Veterans Affairs hospital diagnosis of the Felty syndrome, a surrogate measure of latency. After the first year, the increased risk for all cancers combined was fairly constant for the remainder of the study period (1.84, 1.94, 1.69). The risk for lung cancer decreased slightly during the 20-year study period (from 2.41 to 1.78), whereas the risk for non-Hodgkin lymphoma remained constant. Even though 3 of the 6 patients with melanoma were diagnosed in the first year, the risk for melanoma remained increased for the entire study period. The increased risk for leukemia was restricted to the first 5 years of follow-up.

No variation in cancer risk was noted with age at the first Veterans Affairs hospital diagnosis of the Felty syndrome or with the number of hospitalizations for the Felty syndrome (not shown). For 60% of the patients, the Felty syndrome was the primary hospital diagnosis on at least one visit. For the other 40%, the primary hospital diagnosis on the first visit with the Felty syndrome was most often a skeletal, circulatory, digestive, or respiratory disorder. Non-Hodgkin lymphoma was the only cancer with a slightly higher risk for patients who ever had a primary diagnosis of the Felty syndrome (data not shown).

Fifteen percent of patients with the Felty syndrome had splenectomy listed as one of the inpatient procedures. The risk for non-Hodgkin lymphoma was somewhat higher among patients who had splenectomy (7 patients, SIR = 27.3; CI, 11.0 to 56.3) than among those who had not received splenectomy (12 patients, SIR = 9.75; CI, 5.0 to 17.0). The splenectomies coincided in time with the hospital diagnosis of the Felty syndrome, not with diagnosis of lymphoma. They preceded the lymphoma by 10 years for 3 of the 7 patients, 8 years for 1 patient, 4 years for 2 patients, and 3 years for 1 patient. None of the 13 patients with leukemia had a splenectomy.

Hospital charts, pathology reports, claims folders, and death certificates for the 19 patients with the Felty syndrome who had non-Hodgkin lymphoma and for the sample of 19 patients with the syndrome who did not have cancer were reviewed for confirmation of the diagnosis of the syndrome. We were unable to obtain the relevant hospital records for 2 of the patients with the Felty syndrome who also had non-Hodgkin lymphoma and for 4 of the cancer-free patients. An additional patient with non-Hodgkin lymphoma failed to meet the diagnostic criteria for the Felty syndrome. Thus, for 16% of the patients with the Felty syndrome who had non-Hodgkin lymphoma and 21% of the patients with the syndrome who did not have cancer, the diagnosis of the syndrome could not be confirmed.

The 19 patients with lymphoma included the 2 with inaccessible hospital charts (above), 1 patient with lymphoma who had antecedent prostate cancer, 1 patient diagnosed before detection of the Felty syndrome, and another within 20 days. The histologic results for the remaining patients with lymphoma varied: Seven patients had malignant lymphoma not otherwise specified; 2 had immunoblastic lymphoma; and 2 had large cell, 1 had diffuse lymphocytic, 1 had follicular, and 1 had hairy cell leukemia. B-cell origin was noted for 2 patients (1 had lymphoma not otherwise specified and 1 had immunoblastic lymphoma), whereas 1 patient had the acquired immunodeficiency syndrome and another had the Sjogren syndrome. Two of the cancer-free patients also had the Sjogren syndrome. Only one of the lymphoma charts mentioned treatment with a cytotoxic or immunosuppressive drug (azathioprine), and no mention was made of methotrexate treatment for any of the patients with lymphoma.

Discussion

Top Methods Results Discussion Author & Article Info References

This cohort study of 906 white men with a hospital diagnosis of the Felty syndrome showed a twofold excess of total cancer (SIR = 2.09), a greatly increased risk for non-Hodgkin lymphoma (SIR = 12.8), and a moderately increased rate of lung cancer (SIR = 1.98). The risk for leukemia (SIR = 7.67) was increased but only within 5 years of the first hospitalization for the Felty syndrome. The risk for melanoma was also increased (SIR = 6.95), with half the cases occurring in the first year after diagnosis of the Felty syndrome.

The twofold increased risk for all cancers combined among patients with the Felty syndrome distinguishes this study from previous studies of rheumatoid arthritis, which have shown no overall increased risk for cancer [3]. Perhaps the susceptibility to various forms of cancer relates to features of the Felty syndrome that are not found in other patients with rheumatoid arthritis, such as the low neutrophil count, splenic dysfunction, increased susceptibility to infections, or more aggressive rheumatoid disease. Cancer rates were increased throughout the 20-year study period, as indicated by a SIR of 1.69 ten or more years after the diagnosis of the Felty syndrome. Factors considered indicative of closer surveillance or of increased severity of disease, such as more hospital visits or having the Felty syndrome ever listed as their primary hospital diagnosis, were not associated with a greater risk for total cancer.

The 12-fold increased risk for non-Hodgkin lymphoma among patients with the Felty syndrome exceeds the twofold risk with non-Hodgkin lymphoma found in most surveys of rheumatoid arthritis [3], but it is similar to the risk for the Sjogren syndrome [11], a syndrome that has been associated with the Felty syndrome [12]. In our study, only one of the patients with non-Hodgkin lymphoma also had the Sjogren syndrome. Because most of the lymphomas started 5 or more years after the first hospital visit for the Felty syndrome, it is unlikely that diagnoses of the Felty syndrome were merely patients who had lymphoma in a prodromal stage. Further, non-Hodgkin lymphoma was the only neoplasm with an increased risk for patients who had ever had a primary hospital diagnosis of the Felty syndrome. Cytotoxic or immunosuppressive drugs [13] or methotrexate [14] could have contributed to the increased rate of lymphoma; however, only one of the hospital charts mentioned the use of these agents (azathioprine). Although the risk for non-Hodgkin lymphoma was increased among patients who had received splenectomy, which has been hypothesized as a risk factor for lymphoma [15], it could not explain the 10-fold increased risk for the patients with the Felty syndrome who had not had splenectomy. The susceptibility to lymphoma may be related to chronic immune stimulation, decreased lymphokines, and decreased natural killer cell activity, or to increased susceptibility to Epstein-Barr virus infection that may accompany rheumatoid arthritis, especially in severe or systemic cases [3, 16]. Patients with the Felty syndrome have an increased frequency of a particular human leukocyte antigen DQ haplotype [12] that has been linked to control of T-cell suppressor function [17].

We observed an increased risk for leukemia, which we anticipated for two reasons: 1) a rare form of leukemia, T-cell chronic lymphocytic leukemia, occurred in a group of patients with the Felty syndrome who had large granular lymphocyte syndrome [1, 8]; and 2) cytotoxic drugs used to treat certain patients with severe rheumatoid arthritis can increase the risk for acute nonlymphocytic leukemia [18]. We consider patients with the large granular lymphocyte syndrome to be a subset of patients with Felty syndrome; this syndrome was first recognized in the 1980s and is distinguished from the Felty syndrome by the presence of large granular lymphocytes on blood smear [19-21]. Among the 13 patients with leukemia in our cohort were 2 with chronic lymphocytic leukemia and 4 with acute nonlymphocytic leukemia; all patients with leukemia were diagnosed within the first 5 years of follow-up. The early onset of diverse forms of leukemia suggests that prodromal symptoms of leukemia may have been misdiagnosed as the Felty syndrome [22]. However, the role of immune alterations, such as those responsible for bone marrow depression in the Felty syndrome [23], and the use of alkylating and immunosuppressive agents should be studied as possible risk factors for leukemia as well as lymphoma. Despite reports [24, 25] that splenectomy enhances the risk for acute nonlymphocytic leukemia after treatment of Hodgkin disease, none of the 13 patients with leukemia in our study had splenectomy.

A slight increase in lung cancer has been observed previously in patients with rheumatoid arthritis [4-6, 26, 27] and may be related to rheumatoid interstitial lung disease [28]. We cannot exclude the influence of tobacco smoking, because data on smoking were not available. Immunologic factors may play a role in lung cancer, as suggested by the increased risk among renal transplant recipients [13] and by the familial association of lymphoproliferative disease with lung cancer [29]. Malignant melanoma occurs excessively in certain states of immune suppression [30, 31]; however, three of the six diagnoses of melanoma in patients with the Felty syndrome were found in the first year of follow-up, suggesting the influence of diagnostic surveillance.

The advantages of our study include the large size, the ability to link records and follow patients for up to 20 years on a nationwide scale, the estimation of cancer incidence rather than prevalence or mortality, comparison with both internal and external cancer rates, and access to medical records. The disadvantages include incomplete follow-up and ascertainment of cancers, patients who use Veterans Affairs hospitals may not be representative of the U.S. population, hospital patients are presumably more severely afflicted with rheumatoid arthritis or other diseases, the Felty syndrome and cancer diagnoses could not all be confirmed, and treatment information for the whole cohort is not readily available.

We did not adjust the risk estimates for the findings resulting from chart review, because our preliminary analyses of the reliability of chart retrieval suggest that similar proportions of charts have been misplaced or lost in any of the relevant categories. If we had decreased our "observed" by the number of patients with non-Hodgkin lymphoma that we were unable to confirm, we would also have had to decrease our "expected" by the proportion of patients with the Felty syndrome without cancer that we were unable to confirm. Because we compared the rate of lymphoma in patients with the Felty syndrome with the rate in patients without the syndrome, our estimates could be inflated by ignoring the record review findings only in the following manner. Our estimates could be inflated if patients with the Felty syndrome were more likely to have an unconfirmed diagnosis of lymphoma than were patients without the syndrome or if patients with the syndrome who had lymphoma were more likely to have a previously unconfirmed diagnosis of the syndrome than were patients with the syndrome who did not have lymphoma.

We also did not exclude the six patients with the Felty syndrome who had hematopoietic cancer and who prompted this study, because it would have been difficult to compute "expected" values. If we had removed some of the "observed," we would have to remove a comparable and equitable fraction of the cohort. In fact, the 12-fold association that prompted our study was confirmed when the number of patients tripled (from 6 patients to 19).

Having both internal and external rate comparisons gives us a better estimate of risk, because advantages and disadvantages exist for using either method alone. The internal rate comparisons are probably more appropriate for two reasons: 1) Because cancers are not completely ascertained in the Veterans Affairs hospital database, the comparison with external rates probably underestimates risk; and 2) veterans who use Veterans Affairs hospitals may be different from other U.S. white men with regard to cancer risk factors such as tobacco use, alcohol use, socioeconomic status, and other medical conditions [32], so that differences in risk may reflect differences in lifestyle.

We initially observed that the Felty syndrome was over-represented among a sample of patients with rheumatoid arthritis who had hematopoietic cancer. We then studied total cancer incidence after patients were diagnosed with the Felty syndrome in Veterans Affairs hospitals. Cancer incidence was increased 2-fold, and the risk was 12-fold for non-Hodgkin lymphoma, 7-fold for leukemia and melanoma, and 2-fold for lung cancer. The large risk for non-Hodgkin lymphoma was not explained by treatment or concomitant diseases such as the Sjogren syndrome or the large granular lymphocyte syndrome. Further research is needed about the carcinogenic risks associated with the Felty syndrome as well as other immunologic disorders.

Abbreviations

SEER: Surveillance, Epidemiology and End Results

SIR: standardized incidence ratios