Lymphoma in Rheumatoid Arthritis - The Effect of Methotrexate and Anti-Tumor Necrosis Factor Therapy in 18,572 Patients

 
CONTINUE READING
ARTHRITIS & RHEUMATISM
                                                                                                        Vol. 50, No. 6, June 2004, pp 1740–1751
                                                                                                        DOI 10.1002/art.20311
                                                                                                        © 2004, American College of Rheumatology

                                Lymphoma in Rheumatoid Arthritis

          The Effect of Methotrexate and Anti–Tumor Necrosis Factor Therapy
                                   in 18,572 Patients

                                            Frederick Wolfe1 and Kaleb Michaud2

        Objective. The risk of lymphoma is increased in                         infliximab (with or without etanercept) was 2.6 (95% CI
patients with rheumatoid arthritis (RA), and spontane-                          1.4–4.5). For etanercept, with or without infliximab, the
ous reporting suggests that methotrexate (MTX) and                              SIR was 3.8 (95% CI 1.9–7.5). The SIR for MTX was 1.7
anti–tumor necrosis factor (anti-TNF) therapy might be                          (95% CI 0.9–3.2), and was 1.0 (95% CI 0.4–2.5) for those
associated independently with an increased risk of                              not receiving MTX or biologics. Lymphoma was associ-
lymphoma. However, data from clinical trials and clin-                          ated with increasing age, male sex, and education.
ical practice do not provide sufficient evidence concern-                               Conclusion. Lymphomas are increased in RA.
ing these issues because of small sample sizes and                              Although the SIR is greatest for anti-TNF therapies,
selected study populations. The objective of this study                         differences between therapies are slight, and confidence
was to determine the rate of and standardized incidence                         intervals for treatment groups overlap. The increased
ratio (SIR) for lymphoma in patients with RA and in RA                          lymphoma rates observed with anti-TNF therapy may
patient subsets by treatment group. Additionally, we
                                                                                reflect channeling bias, whereby patients with the high-
sought to determine predictors of lymphoma in RA.
                                                                                est risk of lymphoma preferentially receive anti-TNF
        Methods. We prospectively studied 18,572 pa-
                                                                                therapy. Current data are insufficient to establish a
tients with RA who were enrolled in the National Data
                                                                                causal relationship between RA treatments and the
Bank for Rheumatic Diseases (NDB). Patients were
                                                                                development of lymphoma.
surveyed biannually, and potential lymphoma cases
received detailed followup. The SEER (Survey, Epide-
                                                                                        During the years 1992–1998, the age-adjusted US
miology, and End Results) cancer data resource was
                                                                                incidence rate of lymphoma was 18.3 per 100,000 (15.7
used to derive the expected number of cases of lym-
                                                                                per 100,000 for non-Hodgkin’s lymphoma and 2.6 per
phoma in a cohort that was comparable in age and sex
with the RA cohort.                                                             100,000 for Hodgkin’s lymphoma) (1). These rates in-
        Results. The overall SIR for lymphoma was 1.9                           crease with age. Between the ages of 60 and 64 years, the
(95% confidence interval [95% CI] 1.3–2.7). The SIR for                         modal age group for patients with rheumatoid arthritis
biologic use was 2.9 (95% CI 1.7–4.9) and for the use of                        (RA), the rate of Hodgkin’s lymphoma is 3.1 per 100,000
                                                                                for both sexes, 4.2 per 100,000 for males, and 2.1 per
                                                                                100,000 for females; for patients in the same age group
         The National Data Bank for Rheumatic Diseases has received
grant support from Amgen, Aventis, Bristol-Myers, Centocor, Merck,              with non-Hodgkin’s lymphoma, the respective rates are
Novartis, Pharmacia, Pfizer, Squibb, and Wyeth-Australia. The Inflix-           44.2 per 100,000, 51.5 per 100,000, and 37.7 per 100,000.
imab Safety Registry is supported by grants from Centocor.
         1
           Frederick Wolfe, MD: National Data Bank for Rheumatic
                                                                                        The risk of lymphoma is increased in patients
Diseases, and University of Kansas School of Medicine, Wichita;                 with RA in most studies. Following hospitalization for
2
  Kaleb Michaud, MS: National Data Bank for Rheumatic Diseases,                 RA, at a disease duration of 5–15 years the relative risk
Wichita, Kansas.
         Address correspondence and reprint requests to Frederick               of Hodgkin’s and non-Hodgkin’s lymphoma was 3.9 and
Wolfe, MD, National Data Bank for Rheumatic Diseases, Arthritis                 2.3, respectively among 20,699 patients in a Danish
Research Center Foundation, 1035 North Emporia, Suite 230, Wich-                registry (2). In a Swedish registry of 11,683 RA patients
ita, KS 67214. E-mail: fwolfe@arthritis-research.org.
         Submitted for publication March 15, 2003; accepted in revised          who were seen between 1965 and 1983, the standardized
form February 23, 2004.                                                         incidence ratio (SIR) was 1.98 (95% confidence interval
                                                                         1740
LYMPHOMA IN RA                                                                                                            1741

[95% CI] 1.5–2.6) (3). The risk ratios (determined by          system lacks an adequate denominator, and no associa-
dividing the observed number by the expected number)           tion could be drawn between lymphoma and anti-TNF
for lymphoma incidence and mortality in other large            therapy. In addition, given the absence of age data, it is
studies included a ratio of 2.7 for incidence reported by      not known whether such rates are high or low. Of
Isomaki et al (4), a ratio of 1.9 for mortality reported       interest, however, is the FDA report that in a number of
by Allebeck (5), a ratio of 1.2 for incidence reported by      cases lymphoma developed shortly after the patient
Katusic et al (6), a ratio of 20.0 for incidence reported      started therapy, and that regression of lymphoma oc-
by Prior et al (7), a ratio of 3.5 for mortality reported by   curred in 2 patients following discontinuation of therapy.
Laasko et al (8), ratios of 4.1 and 2.2 for rheumatolo-        Recently, adalimumab, another anti-TNF agent, was
gists’ patients and all patients, respectively, reported by    approved for use in RA (55), and the rates of lymphoma
Tennis et al (9,10), and a ratio of 2.0 for mortality          associated with use of this agent were increased com-
reported by Wolfe et al (11).                                  pared with the rates in the general population (56).
        Increased rates of lymphoma have been attrib-                 Among the issues that remain unresolved is
uted to RA severity, based on detailed epidemiologic           whether the rate of lymphoma is increased by MTX,
studies (12,13) and Epstein-Barr virus studies (14,15),        whether it is increased among patients receiving anti-
although there is some disagreement on this point (16).        TNF agents, and whether such increases are related to
However, there has also been a general concern that            the anti-TNF agents, the worse clinical status of patients
medications, by altering immune function and immuno-           receiving anti-TNF therapy, or a combination of both. In
logic surveillance, might lead to lymphoproliferative          addition, a contrary hypothesis—that lymphoma rates
cancer. This concern has been applied to medications           might be lower in treated patients—has not been inves-
used in RA, including azathioprine (17–21) and metho-          tigated.
trexate (MTX) (22–44). In addition, discontinuation of
MTX has been followed by disappearance of lymphoma
in some patients (reversible lymphoma) (23,25,45,46),                        PATIENTS AND METHODS
thus providing a temporal link between MTX and lym-                     Patients in the study were participants in the National
phoma. Nonetheless, some studies have not found in-            Data Bank for Rheumatic Diseases (NDB) long-term study of
creased rates of lymphoma in MTX-treated patients              the outcomes of RA. Patients were recruited from the prac-
with RA (32,47). There is evidence for a slight increase       tices of 908 US rheumatologists, as previously described (57–
                                                               59), and are systematically followed up by NDB staff, using
in lymphoma rates among RA patients receiving aza-             semiannual questionnaires. The diagnosis of RA was made by
thioprine (18,48), but no epidemiologic studies have           the patient’s rheumatologist. Of the 18,572 patients in this
found an increase in lymphoma in MTX-treated patients          report, 3,244 were enrolled at the time they started receiving
beyond what is expected in RA (32). Studies of cyclo-          leflunomide, and 5,025 were enrolled as part of an infliximab
sporine have also found no increased risk (49). The            safety registry. After the initial enrollment assessment, ⬃8% of
                                                               patients decline to participate each year. The NDB is an open
result of these studies has been to suggest that RA and        cohort, and patients are added continuously. The current study
RA disease activity predispose to lymphoma, but that           concerns 18,572 RA patients without previous lymphoma who
treatment may also play a role. Little consideration has       completed 70,732 biannual questionnaires during consecutive
been given to the idea that effective treatment, by            6-month assessment periods beginning in January 1999 and
reducing disease activity, might also lower lymphoma           ending in June 2002.
                                                                        For each questionnaire assessment, patients report
rates.                                                         demographic and clinical variables (60–70), comorbid condi-
        Recent concern about possible treatment effects        tions, medications, and side effects of treatment. Patients also
and lymphoma has focused on anti-TNF drugs (50–53)             report their cancer status. The identification and validation
because of their profound immunoregulatory effect.             process for lymphoma is shown in Figure 1. At the time of
Reports from MedWatch (the postmarket adverse event            enrollment into the NDB, patients report current (last 6
                                                               months) or past malignancies. They also report current malig-
surveillance system run by the US Food and Drug                nancies in each biannual questionnaire they complete. If a
Administration [FDA]) concerning infliximab and etan-          questionnaire is missed, patients are required to complete the
ercept led the FDA to estimate that the crude incidence        past or “ever” questionnaire again.
rates of lymphoma are 19 per 100,000 patients treated                   For patients reporting a lymphoma or a description
with etanercept and 6.6 per 100,000 patients treated with      suggestive of lymphoma, a validation process is begun (Figure
                                                               1). In general, the first contact with the patient is made by a
infliximab (54). Given the average age of patients with        telephone call from a trained interviewer using a written
RA, these rates are lower than what might be expected          questionnaire. Simple misunderstandings may be identified at
in the general population. However, the MedWatch               this point. For example, patients may confuse lipoma with
1742                                                                                                        WOLFE AND MICHAUD

                                         Figure 1. Validation process for reports of lymphoma.

lymphoma or misidentify metastatic cancer with lymph node             report, a total of 1,514 deaths (from all sources) was known to
involvement as lymphoma. The interviewer also identifies              the NDB.
treatments received for the lymphoma, because this may                        All information obtained is reviewed, and the patient’s
provide clues as to the validity of the diagnosis. Patients are       lymphoma is classified as “validated,” “refuted,” or “likely.” As
also asked about the type of lymphoma. A patient reporting            indicated above, most of the difficulty in validation is associ-
“Hodgkin’s disease” or “non-Hodgkin’s lymphoma” provides              ated with older cases. As examples of handling of evidence, if
evidence regarding the validity of the diagnosis for the subse-       we are unable to obtain medical records but the patient states,
quent review process. In all instances, we attempt to contact         “I had Hodgkin’s disease and was treated with radiation,” we
physicians and obtain medical records. In addition, we request        accept such a diagnosis. However, if the patient reports having
all hospitalization and outpatient surgery records, and patients      lymphoma and being treated with antibiotics, we consider the
report all medications, including chemotherapeutic agents. For        diagnosis as refuted.
key records relating to lymphomas, we make multiple, re-                      Among patients who reported lymphoma in their
peated (if necessary) requests to physicians (including oncolo-       biannual questionnaires, those in whom lymphoma is likely
gists) and hospitals to obtain records that include information       might include patients who could not be contacted because of
regarding cell types.                                                 death, those who withdrew consent, those whose rheumatolo-
         As a preliminary contact, we fax a short questionnaire       gist of record was no longer seeing the patient and did not
to the patient’s rheumatologist inquiring about the diagnosis         know about the lymphoma, or patients who had moved and
and validity, and we follow up as necessary by direct contact         had no identifiable forwarding address or telephone number.
with the rheumatologist and by obtaining medical records. This        Patients being followed up as part of the infliximab safety
                                                                      registry were contacted regarding lymphoma immediately
process works very well with patients who report current
                                                                      upon the receipt of such a report. Patients who were not in the
lymphoma. It works less well in terms of lymphomas that may
                                                                      infliximab safety registry were contacted in 2002 regarding
have occurred prior to enrollment in the NDB, because
                                                                      lymphomas reported on their biannual questionnaires for the
previous physicians and records may not be available.                 years 1998–2002 and previously; however, hospital and treat-
         Death records are also reviewed, when available, and         ment records for such patients had already been obtained as a
surviving relatives are contacted for consent to obtain medical       part of NDB surveillance at the time of the report. Based on
records. Death information comes initially from relatives and         validation experience (see Discussion), the probability that a
physicians, often during the process of tracing nonrespondents.       patient who is classified as “likely” lymphoma actually has
We also systematically screen the National Death Index (NDI)          lymphoma is ⬎0.9. Lymphoma cell types were extracted from
(71,72) at yearly intervals for death information and diagnosis.      hospitalization reports, physician reports, and patient reports,
However, because of delays in the processing of death records         in that order of precedence.
by the NDI, the death information from the NDI in this report                 Classification into treatment groups. Patients were
is complete only through the year 2000. At the time of this           grouped into classes according to hierarchical lifetime treat-
LYMPHOMA IN RA                                                                                                                         1743

Table 1. Demographic and lymphoma status of patients in whom lymphoma developed during NDB participation*
                                                                                Treatment category

             Variable                       All           Infliximab†       Etanercept‡       MTX/no biologic        No MTX/no biologic
No. of lymphomas                             29                9                  8                  10                       5
Age, mean ⫾ SD years                    68.4 ⫾ 10.5       69.8 ⫾ 6.4         64.2 ⫾ 7.3          72.2 ⫾ 9.5              62.8 ⫾ 17.9
% male                                      51.7             33.3               62.5                60.0                     40.0
% deceased                                  17.2              0.0               12.5                20.0                     20.0
No. (%) fully validated lymphomas§       82.8 (24)         100.0 (9)          87.5 (7)            70.0 (7)                 80.0 (4)
Lymphoma type
  Non-Hodgkin’s                              22               9                  7                    6                       3
  Hodgkin’s                                   1               0                  1                    0                       0
  Unknown                                     6               0                  0                    4                       2

* Except where indicated otherwise, values are the mean. NDB ⫽ National Data Bank. MTX ⫽ methotrexate.
† Includes 3 patients who also received etanercept.
‡ Includes 3 patients who also received infliximab.
§ See Patients and Methods for validation details.

ment exposure status. Patients who ever received infliximab or          MTX/no biologics group and begins the observation time at
etanercept were placed in separate groups. Membership in                the time of enrollment in the NDB. This second method, the
these groups was independent of any other treatment, such               “switched” method, is preferred, because it protects against
that current or past use of MTX did not affect membership in            patients who may have discontinued MTX many years in the
the 2 anti-TNF treatment groups. Patients who had ever                  past being counted in the MTX group. The consequences of
received MTX but not infliximab or etanercept were placed in            using different methods for classifying observation times will
the MTX-only group. Patients who did not receive infliximab,            be described in the Discussion section.
etanercept, or MTX were placed in the no MTX/no biologics                       Statistical analysis. The expected number of lym-
group. Patients who received anakinra but were not in any of            phoma cases and the SIRs were calculated based on age- and
the other groups were placed in the anakinra group. Anakinra            sex-matched data from the National Cancer Institute SEER
was received by 233 of the 18,572 patients, but only 58 of these        (Surveillance, Epidemiology, and End Results) resource (1).
patients were not in any of the other groups. In the analyses           The SIR analyses were used to determine the risk of lymphoma
described in Table 4, we identified persons in the no MTX/no            associated with RA and with the various treatments com-
biologics group who received leflunomide, but not MTX,                  pared with the risk in the US population. Incidence rates
etanercept, infliximab, or anakinra, and analyzed them sepa-            represent the number of lymphomas divided by person-years
rately. Treatments received by patients in the no MTX/no                of exposure and are expressed in 100,000 patient-years of
biologics group at a rate of ⬎1% were prednisone (24%),                 exposure. Differences between patients with and without lym-
hydroxychloroquine (26%), injectable gold (4%), auranofin               phoma were assessed by t-tests and logistic regression, on a
(1%), and leflunomide (17%). In 59 persons, lymphoma                    random observation for each patient. For confirmation, assess-
developed before entry into the NDB. These patients were                ments were also performed using Cox regression. Analyses
excluded from analysis and are not further considered.                  were performed with Stata statistical software, version 8.0 (73).
         Definition of observation time and exposure time. The          The 5% significance level was used. All tests were 2-tailed.
observation and analysis time begins at the time the patient
was first exposed to treatment while participating in the NDB.
For example, if a patient enters the NDB at year 0 and starts                                     RESULTS
MTX at year 1, the observation time begins at year 1. However,
if a patient begins a treatment, as defined in the classification               Table 1 shows the age, sex, mortality, validation
method described above, at or before enrolling in the NDB,              status, and lymphoma type for the patients with lym-
the observation time begins at the time of enrollment in the            phoma, according to observation period and treatment
NDB (year 0). The observation time ends at the last observa-            group. Overall, we identified 88 lymphomas, 59 of which
tion in the data bank. The median duration of followup for all          occurred prior to NDB enrollment and 29 of which
study patients was 1.25 years (range 0.1–4.5 years) and was as
follows for the major treatment groups: for MTX, 1.5 years              occurred after NDB enrollment and during the period of
(range 0.1–4.0 years); for etanercept, 1.3 years (range 0.1–4.0         intensive followup. In Table 1, patients are counted as
years); for infliximab, 1.0 years (range 0.1–3.5 years); and for        being in the infliximab or etanercept group regardless of
leflunomide, 1.4 years (range 0.1–4.5 years). For patients who          whether they received MTX (or anakinra) therapy, but
received MTX prior to their enrollment in the NDB and who               patients in the MTX group did not receive infliximab or
were not receiving MTX at the time of enrollment, 2 methods
of treatment classification and observation time were used in           etanercept therapy during the study followup. Patients
the analyses shown in Table 1. The first method is as described         who received both infliximab and etanercept (n ⫽ 3) are
above. The second method places all such patients in the no             counted in both groups. Only 233 patients received
1744                                                                                                  WOLFE AND MICHAUD

Table 2. Cell types in patients with lymphoma*                       ment with those of patients in whom lymphoma did not
 Treatment          Lymphoma                                         develop. Patients who reported lymphoma prior to en-
group/patient          type                      Cell type           rollment were excluded from these analyses. Patients
   1               Non-Hodgkin’s      Marginal zone MALT type        with lymphoma were significantly older (67.3 years ver-
   2               Non-Hodgkin’s      Sclerosing lymphoma, mixed     sus 60.3 years), had more comorbidities (3.1 versus 2.1),
                                        small/large cell             were more likely to be male (51.7% versus 23.3%), had
    3              Non-Hodgkin’s      Large cell
    4              Non-Hodgkin’s      T cell with pancytopenia       more education (14.2 years versus 13.2 years), and were
    5              Hodgkin’s          Hodgkin’s                      more likely to be non-Hispanic whites (100.0% versus
Etanercept/                                                          89.4%; P ⫽ 0.066). Age, sex, disease duration, and
      infliximab
    6              Non-Hodgkin’s      “Lymphocytic lymphoma”         education were entered into a multivariate logistic re-
    7              Non-Hodgkin’s      “Low grade lymphoma”           gression. The odds ratio (OR) for a 10-year increase in
    8              Non-Hodgkin’s      Mixed cell                     age was 1.7 (95% CI 1.2–2.4). Men were 2.6-fold more
Infliximab
    9              Non-Hodgkin’s      Large cell                     likely to have lymphoma than were women (OR 2.6,
  10               Non-Hodgkin’s      Large B cell with underlying   95% CI 1.2–5.5), and each year of education was asso-
                                        small cell                   ciated with an increase in risk (OR 1.2, 95% CI 1.0–1.4).
 11                Non-Hodgkin’s      Unknown
 12                Non-Hodgkin’s      B cell, small cell             The duration of RA remained nonsignificant (P ⫽
 13                Non-Hodgkin’s      Diffuse, large B cell          0.876) in the multivariate analysis. These analyses were
 14                Non-Hodgkin’s      Unknown                        repeated using Cox regression, with similar results. The
Methotrexate
 15                Non-Hodgkin’s      Immunoblastic                  Cox regression hazard ratios and 95% CIs for these
 16                Non-Hodgkin’s      Diffuse large cell, B cell     variables were as follows: for age, 1.58 per 10-year
 17                Non-Hodgkin’s      Large cell T cell              increase (95% CI 1.16–2.18); for male sex, 3.70 (95% CI
 18                Non-Hodgkin’s      “Malignant lymphoma”
 19                Non-Hodgkin’s      Large cell                     1.79–7.68); for education level, 1.16 (95% CI 0.99–1.37);
 20                Non-Hodgkin’s      Unknown                        and for comorbidity, 1.30 (95% CI 1.10–1.54).
 21                Unknown            Unknown                                Table 4 shows the SIRs and lymphoma rates for
 22                Unknown            Unknown
 23                Unknown            Unknown                        patients in the various treatment groups. Because of
 24                Unknown            Unknown                        overlapping therapy for anti-TNF drugs, 3 methods were
No treatment                                                         used to describe these treatments. In the first method,
 25                Non-Hodgkin’s      Diffuse, large cell
 26                Non-Hodgkin’s      Large cell lymphoma            patients who received infliximab or etanercept or both
 27                Non-Hodgkin’s      Large cell                     treatments were analyzed separately. In the second
 28                Unknown            Unknown                        method, patients who received infliximab regardless of
 29                Unknown            Unknown
                                                                     whether they received etanercept were analyzed, and
* MALT ⫽ mucosa-associated lymphoid tissue.                          patients who received etanercept regardless of whether
                                                                     they received infliximab were similarly analyzed. In the
                                                                     third method, all patients who received anti-TNF ther-
anakinra (an interleukin-1 receptor antagonist), and no              apy were studied as a single group.
lymphomas occurred in this group.                                            As shown in Table 4, the overall SIR for lym-
        Of the 29 patients in whom lymphomas devel-                  phoma, regardless of treatment, was 1.9 (95% CI 1.3–
oped after NDB enrollment, 17% were known to be                      2.7). The SIR for lymphoma in patients receiving bio-
dead. Death rates among patients who received MTX                    logics was 2.9 (95% CI 1.7–4.9) and the SIR in those
without biologics (20%), etanercept (13%), infliximab                receiving infliximab, with or without etanercept, was 2.6
(0%), or none of these therapies (20%) reflected pri-                (95% CI 1.4–5.0). For patients receiving etanercept,
marily the shorter duration in the data bank of patients             with or without infliximab, the SIR for lymphoma was
receiving the newer, anti-TNF therapies. Among pa-                   3.8 (95% CI 1.9–7.5). For patients who received lefluno-
tients in whom lymphoma developed after NDB enroll-                  mide only, the SIR was 0 (no cases of lymphoma). In
ment, 22 cases of non-Hodgkin’s lymphoma and 1 case                  addition, the overall SIR for lymphoma in the 6,498
of Hodgkin’s lymphoma were identified. Data on lym-                  patients who had received leflunomide at any time and
phoma type were not available in 6 cases. Table 2 shows              in any combination was 1.0 (95% CI 0.4–2.6). The
the reported cell types. A wide variety of cells types was           analysis was also performed after restricting the sample
reported.                                                            to patients who were not receiving infliximab or etaner-
        Table 3 shows a comparison of the characteristics            cept prior to entry into the data bank. Among 3,024
of patients in whom lymphoma developed after enroll-                 “new starts” on infliximab, the SIR for lymphoma in
LYMPHOMA IN RA                                                                                                                                1745

                    Table 3.   Characteristics of patients with or without lymphoma during NDB participation*
                                                                        No lymphoma                  Lymphoma
                                  Characteristic                        (n ⫽ 18,543)                  (n ⫽ 29)              P
                    Age, years                                            60.3 ⫾ 13.4                67.3 ⫾ 10.6           0.005
                    Male, %                                                   23.3                       51.7              0.001
                    Disease duration, years                               13.6 ⫾ 10.8                15.1 ⫾ 10.9           0.489
                    Education, years                                      13.2 ⫾ 2.4                 14.2 ⫾ 2.1            0.027
                    Non-Hispanic white, %                                     89.4                      100.0              0.066
                    Ever smoker, %                                            53                          59               0.875
                    Total income, US dollars                            44,074 ⫾ 28,605            48,400 ⫾ 35,757         0.451
                    Lifetime comorbidity score (0–11 scale)               2.12 ⫾ 1.6                  3.1 ⫾ 1.8            0.002
                    HAQ (0–3 scale)                                       1.11 ⫾ 0.7                 1.03 ⫾ 0.7            0.576
                    Methotrexate use ever, %                                  77                         79                0.803
                    Biologic agent use ever, %                                47                         48                0.869
                    Infliximab use ever, %                                    35                          31               0.665
                    Etanercept use ever, %                                    19                         28                0.217
                    Leflunomide use ever, %                                   14                         36                0.014

                    * Except where indicated otherwise, values are the mean ⫾ SD. Patients with lymphomas prior to National
                    Data Bank (NDB) enrollment are excluded. Comparisons are between the first visit at which lymphoma
                    was reported and a randomly selected visit for patients without lymphoma. See text for results of Cox
                    regression analyses. HAQ ⫽ Health Assessment Questionnaire.

those receiving infliximab only was 1.0 (95% CI 0.46–                         MTX/no biologics group. Without MTX switching, the
7.4). For new starts on etanercept only, the SIR was 0.95                     SIR in the no MTX/no biologics group was 1.3 (95% CI
(95% CI 0.5–8.3).                                                             0.5–3.1), and with switching it was 1.0 (95% CI 0.4–2.5).
        The results for the MTX and no MTX/no bio-                                   Overall, the lymphoma incidence rate per
logics groups differed according to whether patients who                      100,000 patient-years was 99 (95% CI 69–142). For the
were classified in the MTX group but never received                           various durations of RA, the rates were as follows: for
MTX during their followup in the NDB were “switched”                          0–5 years, 171 (95% CI 82–360), for 5–10 years, 70 (95%
to the no MTX/no biologics group. Without switching,                          CI 29–168), for 10–15 years, 20 (95% CI 3–145), and for
the SIR in those receiving MTX was 1.5 (95% CI                                ⬎15 years, 121 (95% CI 74–198).
0.8–2.7); with switching the rate was 1.7 (95% CI                                    To further explain the potential differences in
0.9–3.2). Switching had a reciprocal effect on the no                         treatment groups, the characteristics of patients in the

Table 4. SIR and incidence rate for lymphoma in RA patients according to exposure group*
                                                   No. of lymphomas
                                                                                                         No. of patients        Time      Incidence
            Treatment group                    Actual        Expected         SIR         95% CI             at risk            at risk      rate
All patients/treatments                            29          15.5           1.9         1.3–2.7            18,572             29,314       98.9
All biologics                                      14           4.9           2.9         1.7–4.9             8,614             10,012      139.8
Infliximab or infliximab ⫹ etanercept               9           3.4           2.6         1.4–4.5             6,465              6,538      137.7
Etanercept or etanercept ⫹ infliximab               8           2.1           3.8         1.9–7.5             3,381              5,099      156.9
Infliximab                                          6           2.7           2.2         1.0–4.9             5,233              4,913      122.1
Etanercept                                          5           1.4           3.5         1.5–8.4             2,149              3,474      143.9
Infliximab ⫹ etanercept                             3           0.7           4.3         1.4–13.2            1,232              1,624      184.7
Anakinra                                            0           0.2           0.0            –                   58                 33        0
Leflunomide only                                    0           0.5           0              –                  758              1,024        0.0
Leflunomide only (switched)                         0           0.9           0              –                1,184              1,850        0.0
No MTX/no biologics                                 5           3.9           1.3         0.5–3.1             3,504              7,122       70.2
No MTX/no biologics (switched)                      5           4.8           1.0         0.4–2.5             4,399              8,938       55.9
MTX only                                           10           6.8           1.5         0.8–2.7             6,396             12,147       82.3
MTX only (switched)                                10           5.8           1.7         0.9–3.2             5,501             10,331       96.8

* Time at risk is expressed in years. Incidence rates are expressed per 100,000 patient-years of exposure. For leflunomide, switched means patients
in the leflunomide-only group who did not receive leflunomide during National Data Bank (NDB) followup were counted as being in the no
methotrexate (MTX)/no biologics group. For MTX, switched means patients in the MTX-only group who did not receive MTX during NDB
followup were counted as being in the no MTX/no biologics group. SIR ⫽ standardized incidence ratio (adjusted for age and sex); RA ⫽ rheumatoid
arthritis; 95% CI ⫽ 95% confidence interval.
1746                                                                                                               WOLFE AND MICHAUD

Table 5. Characteristics of patients with rheumatoid arthritis, by exposure group*
                                               Infliximab              Etanercept                   MTX                  No MTX/no biologics†
             Variable                         (n ⫽ 6,433)              (n ⫽ 2,729)               (n ⫽ 5,593)                 (n ⫽ 4,474)
Age, years                                    60.7 ⫾ 13.4              56.4 ⫾ 12.4               61.2 ⫾ 13.0                   60.4 ⫾ 14.1
% male                                             22.7                     20.7                      24.3                          24.3
Disease duration, years                       13.7 ⫾ 10.7              14.1 ⫾ 10.2               13.5 ⫾ 10.6                   13.5 ⫾ 11.3
Education, years                              13.1 ⫾ 2.5               13.6 ⫾ 2.2                13.2 ⫾ 2.3                    13.3 ⫾ 2.5
Total income, US dollars                    43,057 ⫾ 27,889          49,694 ⫾ 28,792           41,375 ⫾ 26,446               44,094 ⫾ 28,439
Lifetime comorbidities (0–11 scale)             2.2 ⫾ 1.5                2.6 ⫾ 1.8                 2.4 ⫾ 1.8                     2.5 ⫾ 1.9
HAQ (0–3 scale)                                 1.2 ⫾ 0.7                1.2 ⫾ 0.7                 1.1 ⫾ 0.7                     1.0 ⫾ 0.7
Pain (0–10 scale)                               4.2 ⫾ 2.4                4.3 ⫾ 2.4                 3.7 ⫾ 2.3                     3.9 ⫾ 2.6
Global severity (0–10 scale)                    3.9 ⫾ 2.2                3.8 ⫾ 2.1                 3.4 ⫾ 2.1                     3.5 ⫾ 2.4
VAS QOL scale (0–100 scale)                   65.8 ⫾ 17.5              64.3 ⫾ 17.7               66.7 ⫾ 17.5                   65.9 ⫾ 19.5

* Except where indicated otherwise, values are the mean ⫾ SD. Values represent statistics based on the mean for each patient. MTX ⫽
methotrexate; HAQ ⫽ Health Assessment Questionnaire; VAS ⫽ visual analog scale; QoL ⫽ quality of life.
† No exposure to infliximab, etanercept, or MTX while enrolled in the National Data Bank.

groups are presented in Table 5. In general, those in the                 agents was associated with an immediate risk of lym-
anti-TNF groups had slightly more severe disease, as                      phoma, we plotted the time from the start of therapy to
shown by increased scores for the Health Assessment                       the time of lymphoma development (Figure 1). No early
Questionnaire (HAQ) (60), pain, and global severity,                      pattern of lymphoma was detected.
and decreased scores for quality of life. Income and
education levels were higher in etanercept-treated pa-
tients, and patient age was lower in this group than in
other groups. There were no important differences
                                                                          Table 6. Differences in severity measures between the last clinic visit
between groups for age (except for etanercept), sex, or                   before NBD enrollment and the last completed NBD survey*
disease duration.
                                                                                     Patient group/variable              Mean ⫾ SD          P
        To further explore potential disease severity dif-
ferences between groups, we compared baseline and                         Non–registry-enrolled patient (n ⫽ 3,140)
                                                                            Clinic HAQ score                             1.09 ⫾ 0.72       0.763
followup measures of severity (Table 6). Clinic values                      Survey HAQ score                             1.09 ⫾ 0.75
refer to data obtained in a patient’s rheumatologist’s                      Clinic pain score                            4.33 ⫾ 2.62     ⬍0.000
office at the time the patient completed a short ques-                      Survey pain score                            3.78 ⫾ 2.74
                                                                            Clinic patient global score                  4.03 ⫾ 2.54     ⬍0.000
tionnaire and enrolled in the data bank. Survey values                      Survey patient global score                  3.48 ⫾ 2.52
are those data obtained in the NDB questionnaires that                    All infliximab-treated patients (n ⫽ 2,792)
were mailed to patients. Survey values are obtained an                      Clinic HAQ score                             1.32 ⫾ 0.67     ⬍0.001
                                                                            Survey HAQ score                             1.19 ⫾ 0.72
average of 3.5 months after enrollment. For patients just                   Clinic pain score                            4.66 ⫾ 2.71     ⬍0.001
beginning infliximab therapy (new starts; n ⫽ 1,644), the                   Survey pain score                            4.06 ⫾ 2.72
difference between survey and clinic values can be                          Clinic patient global score                  4.02 ⫾ 2.51     ⬍0.001
considered to represent the effect of treatment. As                         Survey patient global score                  3.74 ⫾ 2.41
                                                                          Infliximab new-start patients (n ⫽ 1,644)
shown in Table 6, patients in the infliximab “new start”                    Clinic HAQ score                             1.41 ⫾ 0.64     ⬍0.001
group had much worse clinical status than did the 3,140                     Survey HAQ score                             1.19 ⫾ 0.71
patients enrolling from the practices of US rheumatolo-                     Clinic pain score                            5.24 ⫾ 2.54     ⬍0.001
                                                                            Survey pain score                            4.15 ⫾ 2.73
gists who were not part of treatment registries. In                         Clinic patient global score                  4.55 ⫾ 2.41     ⬍0.001
addition, all infliximab-treated patients (a group that                     Survey patient global score                  3.81 ⫾ 2.41
included both new starts and those continuing to receive                  * Clinic values are those obtained in the patient’s rheumatologist’s
infliximab at the time of the rheumatologist’s evalua-                    office at the time the patient completed a short questionnaire and
tion) had more abnormal scores generally. In addition,                    enrolled in the data bank. Survey values are those obtained in the
                                                                          National Data Bank (NDB) questionnaires that were mailed to
patients in the infliximab new start group had an im-                     patients. Survey values were obtained an average of 3.5 months after
provement in the HAQ score of 0.22 units, while the                       enrollment. For patients just starting infliximab therapy (new start),
HAQ score did not improve in non–registry-enrolled                        the difference between survey and clinic values can be considered to
                                                                          represent the effect of treatment. Health Assessment Questionnaire
patients.                                                                 (HAQ) scores were based on a 0–3 point scale; all other scores were
        To investigate the possibility that use of anti-TNF               based on a 0–10-point scale.
LYMPHOMA IN RA                                                                                                          1747

                       DISCUSSION                                  infliximab-treated patients, the RA SIR is 6.4 (95% CI
                                                                   1.7–16.3), and among adalimumab-treated patients the
        The results of this study show that lymphoma is
                                                                   SIR is 5.4 (95% CI 2.6–10.0) (56). However, do these
increased in RA compared with the general population.
                                                                   increased rates reflect adverse events related to treat-
Previous studies have shown the lymphoma risk in RA
                                                                   ment, RA itself, or a combination of both? It is well
(as measured by ORs and risk ratios) to be 1.98 (3), 2.7
                                                                   known that patients in anti-TNF clinical trials represent
(4), 1.9 (5), 1.2 (6), 20.0 (7), 3.5 (8), 4.1 (9), 2.2 (10), and
                                                                   the most severe 15–20% of RA patients (74), and that
2.0 (11). The highest rates of lymphoma have been
                                                                   such patients are at the greatest risk for lymphoma
found in older studies, and one study did not find an
                                                                   development (12,13). This suggests that disease activity
increased risk (47). In general, in the 5 largest studies
                                                                   might play a very important role.
the risk ratio for lymphoma is ⬃2.0. These data are
                                                                           The article that expressed concern regarding the
consistent with the results of the current study in 18,572
                                                                   potential risk of anti-TNF therapy pointed out that a
RA patients in the NDB, in which the overall SIR for
                                                                   number of lymphoma cases reported to the FDA had
lymphoma was 1.9 (95% CI 1.3–2.7).
                                                                   developed soon after administration of the anti-TNF
        We also found that the SIR for lymphoma was 2.9
                                                                   drugs (54). Such a temporal relationship is worrisome.
(95% CI 1.7–4.9) in patients receiving biologics, 2.6
                                                                   However, problems with the FDA MedWatch system, in
(95% CI 1.4–5.0) in those receiving infliximab, with or
                                                                   which unusual events are preferentially reported, might
without etanercept, and 3.8 (95% CI 1.9–7.5) in patients
                                                                   be an explanation for this finding, because for treat-
treated with etanercept, with or without infliximab.
                                                                   ments that have recently been released, the duration of
Using the “switching method” for MTX, the SIR was 1.7
                                                                   treatment before development of lymphoma cannot be
(95% CI 0.9–3.2), and for patients receiving none of
                                                                   long. By contrast, Figure 2 shows that the times of onset
these therapies the SIR was 1.0 (95% CI 0.4–2.5).
                                                                   of lymphomas in the patients in this study were distrib-
        Although lymphoma following MTX treatment
                                                                   uted evenly over the duration of followup. Given the
has been a concern because of a possible causal rela-
                                                                   remission or recession of lymphoma that has been noted
tionship (22–43), the study results do not show an
                                                                   in some patients following withdrawal of treatment, we
increased SIR in patients receiving MTX compared with
                                                                   suggest the possibility that in some instances treatment
those who were never exposed to MTX. These results
                                                                   might be harmful, while in other instances it might be
are consistent with those in a study by Mariette et al
                                                                   protective. Reversibility (causality) has been noted re-
(32), who found no increase in lymphoma in MTX-
treated patients compared with a population of French              peatedly with MTX therapy (23,25,45,46). However, we
nationals. However, they also found no increase in the             have reported that MTX may protect against mortality
rate of lymphoma in RA, and the incidence rate (⬃30                (75).
per 100,000) was based on estimates of use of MTX in                       Our study has a number of limitations that need
the population.                                                    to be understood. Although the cases presented as
        Despite the differences among treatment groups             lymphoma were fully validated in 83% of patients and
described above, even a sample of ⬎18,500 patients                 were considered highly likely in the other 17% (proba-
could not demonstrate significant differences among the            bility ⬎0.9), it is possible that we overestimated the
studied groups, because of the rarity of lymphoma. In              number of lymphomas. If this were the case, however,
the current study, lymphoma developed in only 29 of the            the rates would be slightly lower, as would the SIRs.
18,572 patients, for an incidence rate of 98.9 per 100,000         However, the converse is also possible—that we failed to
patient-years. By observing the upper 95% confidence               identify some cases of lymphoma. To explore the possi-
intervals, however, a “worse-case” can be obtained,                bility of misclassifying lymphomas, we excluded those
suggesting that the risk ratio for lymphoma does not               cases for which external validation was not possible.
exceed 3.1 for patients receiving no MTX and no                    Sixty-six cases of lymphoma were externally validated,
biologics, 3.2 for those receiving MTX, and 5.0 for those          including those that were reported as occurring prior to
being treated with biologics.                                      study enrollment. Three additional patient reports were
        There is substantial evidence in support of the            found not to be valid (1 case of lipoma and 2 cases of
association of lymphoma and anti-TNF therapy. In                   cancer in lymph nodes), for an overall patient reporting
clinical trials, the SIR for lymphoma in patients receiv-          accuracy of 95.7 %. We also conducted analyses on other
ing etanercept is 2.3 (95% CI 0.9–5.0) or 3.5 (95% CI              cancers, with similar high accuracy rates. The results of
1.6–6.6) if a larger number of cases is used (56). In              these analyses do, in fact, suggest that accepting a
1748                                                                                                          WOLFE AND MICHAUD

       Figure 2. Number of months from the start of therapy to the development of lymphoma. Bars show the 5th and 95th percentiles.

patient’s report of lymphoma when external validation is                observation time for the anti-TNF agents, given their
not possible is an acceptable method.                                   recent introduction.
        It is possible that death records that are not yet                     In addition, there is no really good way in which
available will show additional lymphoma cases. It is also               to model MTX exposure in the current data set. Some
possible that patients in whom lymphoma develops                        patients may have had long exposures to MTX years
might drop out of the study and not report their                        prior to entry into the NDB, but no exposure while in
lymphoma. If such were the case, we would underesti-                    the NDB. The alternative (switched versus nonswitched)
mate the actual SIR. To understand whether such events                  methods for MTX indicate the effect of differing as-
occurred in the current study, we analyzed available data               sumptions on SIRs for the MTX and no MTX/no
to determine whether we learned of any lymphomas only                   biologics groups. In spite of these differences in method,
by death records. Although the absence of cases is weak                 the actual SIRs do not differ substantially. Finally, it
supporting evidence of case ascertainment, we found no                  should be noted that by definition the no MTX/no
such cases. Although such misclassification might alter                 biologics group has to have the longest observation
the SIR, there is no evidence that differential misclassi-              (exposure) time, because the start of drug therapy begins
fication could have occurred. Therefore, we can be                      at the time of cohort enrollment. Thus, we may slightly
confident as to the relative relationships in SIR among                 underestimate the risk associated with no MTX/no
the treatment groups.                                                   biologic therapy.
        Incidence rates and SIRs can be affected by the                        Figure 2 is included to address the issue of how
method we chose to calculate observation time (expo-                    soon after the start of therapy lymphomas may occur.
sure). We assumed that once exposure to treatment                       We did not notice a clustering of very early cases. If
occurred, the observation time should continue to the                   lymphoma occurs soon after the start of treatment, this
end of followup. This assumption is reasonable because,                 may provide some evidence of a causal association, but
assuming there is a risk associated with anti-TNF ther-                 the lack of a temporal association provides no informa-
apy, 1) we do not know how long after exposure the risk                 tion on the likelihood that the medication causes lym-
continues, 2) lymphoma that might be caused by drug                     phoma. It also possible that inhibition of TNF␣ is not
treatment might develop after the drug is discontinued,                 causally related to lymphoma development but facili-
and 3) actual exposure time is a large fraction of                      tates the growth of tumors that are already present.
LYMPHOMA IN RA                                                                                                                 1749

Because of possibilities such as this, we did not set a       imab, with or without etanercept, was 2.6 (95% CI
minimum treatment exposure time in this study.                1.4–5.0). For patients being treated with etanercept, with
        Several methods are available to adjust for treat-    or without infliximab, the SIR was 3.8 (95% CI 1.9–7.5).
ment assignment, including propensity scores and mar-         The SIR for patients receiving MTX was 1.7 (95% CI
ginal structural models. We did not use these methods in      0.9–3.2), and that for patients in the no MTX/no bio-
this study, because baseline data that were crucial to        logics group was 1.0 (95% CI 0.4–2.5). Lymphoma was
understanding treatment assignment were often unavail-        associated with increasing age, male sex, and level of
able. In addition, potentially key variables such as the      education. It seems possible that the apparently in-
C-reactive protein level or the ESR and the swollen joint     creased rates of lymphoma are, in fact, reduced by
count were also unavailable.                                  therapy, and that the “increase” may reflect channeling
        Although we did not observe a relationship be-        bias whereby patients with the highest risk of lymphoma
tween average disease severity among patients in whom         preferentially receive anti-TNF therapy. It appears that
lymphoma developed and those in whom lymphoma did             neither clinical trial data nor data from the current study
not develop, as has been shown previously when labora-        are sufficient to establish a causal relationship between
tory and physical examination data were available, it is      RA treatments and the development of lymphoma.
possible that the average values of the HAQ, pain, and
global severity scores that were measured during the
surveys (Table 5) masked the severity of disease. We                                   REFERENCES
investigated this possibility by examining the severity        1. National Cancer Institute. Surveillance, Epidemiology and End
scores of patients who enrolled in the NDB in their               Results page. URL: http://seer.cancer.gov/.
                                                               2. Mellemkjaer L, Linet MS, Gridley G, Frisch M, Moller H, Olsen
rheumatologists’ clinics. As shown in Table 6, patients
                                                                  JH. Rheumatoid arthritis and cancer risk. Eur J Cancer 1996;32A:
receiving infliximab had much worse clinical status than          1753–7.
did non–registry-enrolled patients. This suggests the          3. Gridley G, McLaughlin JK, Ekbom A, Klareskog L, Adami HO,
possibility that patients receiving infliximab had more           Hacker DG, et al. Incidence of cancer among patients with
                                                                  rheumatoid arthritis. J Natl Cancer Inst 1993;85:307–11.
severe disease generally. If disease severity was a long-      4. Isomaki HA, Hakulinen T, Joutsenlahti U. Excess risk of lympho-
term process that was interrupted by infliximab, then it is       mas, leukemia and myeloma in patients with rheumatoid arthritis.
possible that the infliximab group was at a high risk for         J Chronic Dis 1978;31:691–6.
                                                               5. Allebeck P. Increased mortality in rheumatoid arthritis. Scand
lymphoma regardless of treatment. We do not have data             J Rheumatol 1982;11:81–6.
(HAQ, pain, etc.) from the time of etanercept initiation,      6. Katusic S, Beard CM, Kurland LT, Weis JW, Bergstralh E.
but we expect that such data would not be different from          Occurrence of malignant neoplasms in the Rochester, Minnesota,
those for patients treated with infliximab. We wish to            rheumatoid arthritis cohort. Am J Med 1985;78:50–5.
                                                               7. Prior P, Symmons DP, Hawkins CF, Scott DL, Brown R. Cancer
make clear, however, that the above discussion is con-            morbidity in rheumatoid arthritis. Ann Rheum Dis 1984;43:
sistent with data from the study but is, of course, not           128–31.
proof.                                                         8. Laakso M, Mutru O, Isomaki HA, Koota K. Cancer mortality in
                                                                  patients with rheumatoid arthritis. J Rheumatol 1986;13:522–6.
        In contrast to the nonassociation with HAQ             9. Tennis P, Andrews E, Bombardier C, Wang Y, Strand L, West R,
scores noted in this report, we previously observed a             et al. Record linkage to conduct an epidemiologic study on the
strong association of lymphoma with ESRs (13), with an            association of rheumatoid arthritis and lymphoma in the Province
                                                                  of Saskatchewan, Canada. J Clin Epidemiol 1993;46:685–95.
ESR of ⬎40 mm/hour having a hazard ratio of 9.2 (95%          10. Tennis P, Bombardier C, Malcolm E, Downey W. Validity of
CI 2.0–42.7). In that study, however, there was no                rheumatoid arthritis diagnoses listed in the Saskatchewan Hospital
association with the HAQ score. Similarly, Baecklund et           Separations Database. J Clin Epidemiol 1993;46:675–83.
                                                              11. Wolfe F, Mitchell DM, Sibley JT, Fries JF, Bloch DA, Williams
al, who also considered clinical and laboratory data,             CA, et al. The mortality of rheumatoid arthritis. Arthritis Rheum
found an OR of 25.8 (95% CI 3.1–213.0) for patients               1994;37:481–94.
with high inflammatory activity compared with those           12. Baecklund E, Ekbom A, Sparen P, Feltelius N, Klareskog L.
                                                                  Disease activity and risk of lymphoma in patients with rheumatoid
with low inflammatory activity (12). Thus, the results of         arthritis: nested case-control study. Br Med J 1998;317:180–1.
the current study, which did not include laboratory or        13. Wolfe F. Inflammatory activity, but not methotrexate or pred-
swollen joint count data, should not be used as evidence          nisone use predicts non-Hodgkin’s lymphoma in rheumatoid ar-
                                                                  thritis: a 25-year study of 1,767 RA patients. Arthritis Rheum
against the association of inflammation and lymphoma.             1998;41 Suppl 9:S188.
        In summary, for 18,572 patients with RA in the        14. Newkirk MM, Shiroky JB, Johnson N, Danoff D, Isenberg DA,
NDB, the overall SIR for lymphoma was 1.9 (95% CI                 Shustik C, et al. Rheumatic disease patients, prone to Sjögren’s
                                                                  syndrome and/or lymphoma, mount an antibody response to
1.3–2.7). The SIR for those receiving biologics was 2.9           BHRF1, the Epstein-Barr viral homologue of BCL-2. Br J Rheu-
(95% CI 1.7–4.9) and that for patients receiving inflix-          matol 1996;35:1075–81.
1750                                                                                                              WOLFE AND MICHAUD

15. Van de Rijn M, Cleary ML, Variakojis D, Warnke RA, Chang PP,        35. Kono H, Inokuma S, Matsuzaki Y, Nakayama H, Yamazaki J,
    Kamel OW. Epstein-Barr virus clonality in lymphomas occurring           Hishima T, et al. Two cases of methotrexate induced lymphomas
    in patients with rheumatoid arthritis. Arthritis Rheum 1996;39:         in rheumatoid arthritis: an association with increased serum IgE.
    638–42.                                                                 J Rheumatol 1999;26:2249–53.
16. Kamel OW, Holly EA, van de Rijn M, Lele C, Sah A. A                 36. Georgescu L, Quinn GC, Schwartzman S, Paget SA. Lymphoma in
    population based, case control study of non-Hodgkin’s lymphoma          patients with rheumatoid arthritis: association with the disease
    in patients with rheumatoid arthritis. J Rheumatol 1999;26:             state or methotrexate treatment. Semin Arthritis Rheum 1997;26:
    1676–80.                                                                794–804.
17. Urowitz MB, Lee P. The risks of antimalarial retinopathy, aza-      37. Sukenik S, Ariad S, Flusser D. Malignancy in patients with
    thioprine lymphoma and methotrexate hepatotoxicity during the           rheumatoid arthritis treated with methotrexate [letter]. J Rheu-
    treatment of rheumatoid arthritis. Baillieres Clin Rheumatol            matol 1997;24:806.
    1991;4:193–206.                                                     38. Usman AR, Yunus MB. Non-Hodgkin’s lymphoma in patients
18. Silman AJ, Petrie J, Hazleman B, Evans SJ. Lymphoproliferative          with rheumatoid arthritis treated with low dose methotrexate.
    cancer and other malignancy in patients with rheumatoid arthritis       J Rheumatol 1996;23:1095–7.
    treated with azathioprine: a 20 year follow up study. Ann Rheum     39. Viraben R, Brousse P, Lamant L. Reversible cutaneous lymphoma
    Dis 1988;47:988–92.                                                     occurring during methotrexate therapy. Br J Dermatol 1996;135:
19. Pitt PI, Sultan AH, Malone M, Andrews V, Hamilton EB.                   116–8.
    Association between azathioprine therapy and lymphoma in rheu-      40. Kerr LD, Troy K, Isola L. Temporal association between the use
    matoid disease. J R Soc Med 1987;80:428–9.                              of methotrexate and development of leukemia in 2 patients with
20. Hazleman BL. The comparative incidence of malignant disease in          rheumatoid arthritis. J Rheumatol 1995;22:2356–8.
    rheumatoid arthritics exposed to different treatment regimens.      41. Marlier S, Chagnon A, Brocq O, de Jaureguiberry JP, Jaubert D,
    Ann Rheum Dis 1982;41 Suppl 1:12–7.                                     Paris JF, et al. Lymphoma induced by low-dose methotrexate in
21. Van Wanghe P, Dequeker J. Compliance and long-term effect of            rheumatoid arthritis with severe lymphopenia. Ann Med Interne
    azathioprine in 65 rheumatoid arthritis cases. Ann Rheum Dis            (Paris) 1995;146:206–8.
    1982;41 Suppl 1:40–3.                                               42. Brackertz D. Occurrence of a highly malignant non-Hodgkin
22. Cobeta-Garcia JC, Ruiz-Jimeno MT, Fontova-Garrofe R. Non-               lymphoma in connection with azathioprine therapy and metho-
    Hodgkin’s lymphoma, rheumatoid arthritis and methotrexate.              trexate therapy in a patient with rheumatoid arthritis. Z Rheuma-
    J Rheumatol 1993;20:200–2.                                              tol 1992;51:257–9.
23. Kamel OW, Vanderijn M, Weiss LM, Delzoppo GJ, Hench PK,             43. Shiroky JB, Frost A, Skelton JD, Haegert DG, Newkirk MM,
    Robbins BA, et al. Reversible lymphomas associated with Epstein-        Neville C. Complications of immunosuppression associated with
    Barr virus occurring during methotrexate therapy for rheumatoid         weekly low dose methotrexate. J Rheumatol 1991;18:1172–5.
    arthritis and dermatomyositis. N Engl J Med 1993;328:1317–21.       44. Bachman TR, Sawitzke AD, Perkins SL, Ward JH, Cannon GW.
24. Morris CR, Morris AJ. Localized lymphoma in a patient with              Methotrexate-associated lymphoma in patients with rheumatoid
    rheumatoid arthritis treated with parenteral methotrexate.              arthritis: report of two cases. Arthritis Rheum 1996;39:325–9.
    J Rheumatol 1993;20:2172–3.                                         45. Salloum E, Cooper DL, Howe G, Lacy J, Tallini G, Crouch J, et al.
25. Shiroky JB, Newkirk MM. Reversible lymphomas. N Engl J Med              Spontaneous regression of lymphoproliferative disorders in pa-
    1993;329:1657–8.                                                        tients treated with methotrexate for rheumatoid arthritis and other
26. Taillan B, Garnier G, Castanet J, Ferrari E, Pesce A, Dujardin P.       rheumatic diseases. J Clin Oncol 1996;14:1943–9.
    Lymphoma developing in a patient with rheumatoid arthritis          46. Liote F, Pertuiset E, Cochand-Priollet B, D’Agay MF, Dombret H,
    taking methotrexate. Clin Rheumatol 1993;12:93–4.                       Numeric P, et al. Methotrexate related B lymphoproliferative
27. Buskila D, Thomson GT, Klein M, Keystone EC. Avascular                  disease in a patient with rheumatoid arthritis: role of Epstein-Barr
    necrosis of bone mimicking symmetric polyarthritis in a patient         virus infection. J Rheumatol 1995;22:1174–8.
    with malignant lymphoma treated with high-dose steroids. Isr        47. Moder KG, Tefferi A, Cohen MD, Menke DM, Luthra HS.
    J Med Sci 1992;28:804–5.                                                Hematologic malignancies and the use of methotrexate in rheu-
28. Kingsmore SF, Hall BD, Allen NB, Rice JR, Caldwell DS.                  matoid arthritis: a retrospective study. Am J Med 1995;99:276–81.
    Association of methotrexate, rheumatoid arthritis and lymphoma:     48. Jones M, Symmons D, Finn J, Wolfe F. Does exposure to
    report of 2 cases and literature review. J Rheumatol 1992;19:           immunosuppressive therapy increase the 10 year malignancy and
    1462–5.                                                                 mortality risks in rheumatoid arthritis? A matched cohort study.
29. Klein HP. Occurrence of highly malignant non-Hodgkin’s lym-             Br J Rheumatol 1996;35:738–45.
    phoma in correlation with Imurek therapy and later MTX therapy      49. Van den Borne BE, Landewé RB, Houkes I, Schild F, van der
    of chronic polyarthritis. Z Rheumatol 1992;51:256–9.                    Heyden PC, Hazes JM, et al. No increased risk of malignancies
30. Ellman MH, Hurwitz H, Thomas C, Kozloff M. Lymphoma                     and mortality in cyclosporin A–treated patients with rheumatoid
    developing in a patient with rheumatoid arthritis taking low dose       arthritis. Arthritis Rheum 1998;41:1930–7.
    weekly methotrexate. J Rheumatol 1991;18:1741–3.                    50. Bathon JM, Martin RW, Fleischmann RM, Tesser JR, Schiff MH,
31. Thomson GT, Keystone EC, Sturgeon JF, Fornasier V. Erythema             Keystone EC, et al. A comparison of etanercept and methotrexate
    nodosum and non-Hodgkin’s lymphoma. J Rheumatol 1990;17:                in patients with early rheumatoid arthritis [published erratum
    383–5.                                                                  appears in N Engl J Med 2001;344:240]. N Engl J Med 2000;343:
32. Mariette X, Cazals-Hatem D, Warszawki J, Liote F, Balandraud            1586–93.
    N, Sibilia J. Lymphomas in rheumatoid arthritis patients treated    51. Targan SR, Hanauer SB, van Deventer SJ, Mayer L, Present DH,
    with methotrexate: a 3-year prospective study in France. Blood          Braakman T, et al, and the Crohn’s Disease cA2 Study Group. A
    2002;99:3909–15.                                                        short-term study of chimeric monoclonal antibody cA2 to tumor
33. Starkebaum G. Rheumatoid arthritis, methotrexate, and lym-              necrosis factor ␣ for Crohn’s disease. N Engl J Med 1997;337:
    phoma: risk substitution, or cat and mouse with Epstein-Barr            1029–35.
    virus? J Rheumatol 2001;28:2573–5.                                  52. Sandborn WJ, Hanauer SB. Antitumor necrosis factor therapy for
34. Georgescu L, Paget SA. Lymphoma in patients with rheumatoid             inflammatory bowel disease: a review of agents, pharmacology,
    arthritis: what is the evidence of a link with methotrexate? Drug       clinical results, and safety. Inflamm Bowel Dis 1999;5:119–33.
    Saf 1999;20:475–87.                                                 53. Maini R, St Clair EW, Breedveld F, Furst D, Kalden J, Weisman
LYMPHOMA IN RA                                                                                                                                    1751

      M, et al, and the ATTRACT Study Group. Infliximab (chimeric                  rheumatoid arthritis: a prospective study of 400 patients. J Rheu-
      anti-tumour necrosis factor ␣ monoclonal antibody) versus pla-               matol 1988;15:932–41.
      cebo in rheumatoid arthritis patients receiving concomitant meth-      64.   Ware JE, Sherbourne CD. The MOS 36-item short-form health
      otrexate: a randomised phase III trial. Lancet 1999;354:1932–9.              survey (SF-36). I. Conceptual framework and item selection. Med
54.   Brown SL, Greene MH, Gershon SK, Edwards ET, Braun MM.                       Care 1992;30:473–83.
      Tumor necrosis factor antagonist therapy and lymphoma develop-         65.   Hurst NP, Kind P, Ruta D, Hunter M, Stubbings A. Measuring
      ment: twenty-six cases reported to the Food and Drug Adminis-                health-related quality of life in rheumatoid arthritis: validity,
      tration. Arthritis Rheum 2002;46:3151–8.                                     responsiveness and reliability of EuroQol (EQ-5D). Br J Rheu-
55.   Weinblatt ME, Keystone EC, Furst DE, Moreland LW, Weisman                    matol 1997;36:551–9.
      MH, Birbara CA, et al. Adalimumab, a fully human anti–tumor            66.   Wolfe F, Hawley DJ. Measurement of the quality of life in
      necrosis factor ␣ monoclonal antibody, for the treatment of                  rheumatic disorders using the EuroQol. Br J Rheumatol 1997;36:
      rheumatoid arthritis in patients taking concomitant methotrexate:            786–93.
      the ARMADA trial. Arthritis Rheum 2003;48:35–45.                       67.   Choi HK, Michaud K, Wolfe F. Utility measures in rheumatic
56.   FDA. Briefing Document. US Food and Drug Administration:                     disorders [abstract]. Arthritis Rheum 2002;46 Suppl 9;S76.
      2003.                                                                  68.   Brazier J, Roberts J, Deverill M. The estimation of a preference-
57.   Wolfe F, Anderson J, Burke TA, Arguelles LM, Pettitt D.                      based measure of health from the SF-36. J Health Econ 2002;21:
      Gastroprotective therapy and risk of gastrointestinal ulcers: risk           271–92.
      reduction by COX-2 therapy. J Rheumatol 2002;29:467–73.                69.   Brazier J, Usherwood T, Harper R, Thomas K. Deriving a
58.   Wolfe F, Flowers N, Burke TA, Arguelles LM, Pettitt D. Increase              preference-based single index from the UK SF-36 Health Survey.
      in lifetime adverse drug reactions, service utilization, and disease         J Clin Epidemiol 1998;51:1115–28.
      severity among patients who will start COX-2 specific inhibitors:      70.   Wolfe F, Hawley DJ, Wilson K. The prevalence and meaning of
      quantitative assessment of channeling bias and confounding by                fatigue in rheumatic disease. J Rheumatol 1996;23:1407–17.
      indication in 6689 patients with rheumatoid arthritis and osteoar-     71.   Doody MM, Hayes HM, Bilgrad R. Comparability of national
      thritis. J Rheumatol 2002;29:1015–22.                                        death index plus and standard procedures for determining causes
59.   Wolfe F, Flowers N, Anderson J. The National Rheumatic Disease               of death in epidemiologic studies. Ann Epidemiol 2001;11:46–50.
      Data Bank: case mix and severity characteristics of patients in        72.   Edlavitch SA, Baxter J. Comparability of mortality followup
      rheumatological practice. Arthritis Rheum 1998;41 Suppl 9:S132.              before and after the National Death Index. Am J Epidemiol
60.   Fries JF, Spitz PW, Young DY. The dimensions of health out-                  1988;127:1164–78.
      comes: the health assessment questionnaire, disability and pain        73.   Stata statistical software: release 8.0. College Station (TX): Stata
      scales. J Rheumatol 1982;9:789–93.                                           Corporation; 2003.
61.   Fries JF, Spitz P, Kraines RG, Holman HR. Measurement of               74.   Wolfe F, Pincus T, O’Dell J. Evaluation and documentation of
      patient outcome in arthritis. Arthritis Rheum 1980;23:137–45.                rheumatoid arthritis disease status in the clinic: which variables
62.   Hawley DJ, Wolfe F. Depression is not more common in rheuma-                 best predict change in therapy. J Rheumatol 2001;28:1712–7.
      toid arthritis: a 10-year longitudinal study of 6,153 patients with    75.   Choi HK, Hernan MA, Seeger JD, Robins JM, Wolfe F. Metho-
      rheumatic disease. J Rheumatol 1993;20:2025–31.                              trexate therapy and mortality in patients with rheumatoid arthritis:
63.   Hawley DJ, Wolfe F. Anxiety and depression in patients with                  a prospective study. Lancet 2002;359:1173–7.
You can also read