Article Text


Study of eight cases of cancer in 426 rheumatoid arthritis patients treated with methotrexate
  1. Christophe Bolognaa,
  2. Marie-Christinne Picotb,
  3. Christian Jorgensena,
  4. Philippe Viua,
  5. Regis Verdierb,
  6. Jacques Sanya
  1. aCHU Lapeyronie, Montpellier, France: Service d’Immuno-Rhumatologie, Fédération de Rhumatologie , bDépartement d’Information Médicale
  1. Professeur Jacques Sany, Service d’Immuno-Rhumatologie, CHU Lapeyronie, 34295 Montpellier Cedex 5, France.


OBJECTIVE To report cancer cases in 426 rheumatoid arthritis patients treated with methotrexate, and determine whether there was an increased incidence of cancer compared with patients never treated with methotrexate (rheumatoid controls) and to the whole regional population.

METHODS The duration of methotrexate treatment was 37.4 (SD 27.9) months. This population was compared with 420 rheumatoid arthritis controls and with a regional population of 812 344 people. Life table analysis was performed to compare the cancer incidence in the two rheumatoid populations. Adjustment for potentially confounding factors was done. The indirect standardisation method was used to compare each rheumatoid population with the regional population.

RESULTS Eight cases of cancer (1.88%; 4.04 cases/1000 person years) were diagnosed in the methotrexate population v six (1.43%; 58.8 cases/1000 person years) in the rheumatoid controls. The life table method showed a higher incidence of cancer in the rheumatoid controls (P = 0.0001). In a multivariate analysis (Cox model), the only significant factor explaining this difference in the cancer incidence was age (P = 0.02). In the regional population there were 6418 new cases of cancer (0.79%; 2.85 cases/1000 person years). By the indirect standardisation method, the ratio of observed cases to expected cases of cancer in each of the rheumatoid populations was not significantly different from 1.

CONCLUSIONS In these eight cases, methotrexate was not found to be responsible for generating cancers. However, because of data regarding lymphomas and methotrexate, and because of the short follow up, especially in the control group, longer prospective studies are warranted.

  • methotrexate
  • rheumatoid arthritis
  • lymphoma
  • cancer.

Statistics from

Weekly low dose methotrexate is increasingly used in the treatment of various conditions, particularly rheumatoid arthritis. However, methotrexate is an antifolate antimetabolite. The oncogenicity of methotrexate is still controversial because of the lack of large series of rheumatoid patients with long term follow up, and because of an increased incidence of lymphomas due to rheumatoid arthritis.1-4 Moreover, other factors must be considered, such as the presence or absence of Sjögren syndrome and the previous or concurrent use of cytotoxic drugs (cyclophosphamide, azathioprine)—thus it was shown in a 20 year follow up study of 404 rheumatoid patients (202 treated and 202 not treated with azathioprine) that the risk of lymphoma was twofold greater in the group which received azathioprine.5

Case studies of cancer diagnosed during methotrexate treatment in rheumatoid patients have already been reported. The cancers are mainly lymphomas and pseudolymphomas. 6-19 The causative role of methotrexate is still difficult to establish as coincidences cannot be ruled out, but regression of tumours after methotrexate withdrawal has been described.9 13 19 20

We have undertaken a long term retrospective study of 426 rheumatoid arthritis patients treated with methotrexate and followed up in our unit. The case notes of patients diagnosed with cancer during the follow up period were reviewed. To determine whether there was an increased incidence of cancer in methotrexate treated rheumatoid patients, we compared the incidence of cancer in this group with that of rheumatoid patients never treated with this drug (rheumatoid arthritis controls) and with the whole regional population.



All patients were recruited in our unit as inpatients or outpatients and fulfilled the American College of Rheumatology criteria for rheumatoid arthritis.21 The case notes of the patients with a diagnosis of cancer were carefully reviewed to determine the diagnosis, particular attention being paid to the histopathological results.

Methotrexate treated patients

The decision to use methotrexate was made by the physician in charge of the patient during hospital admission. Methotrexate was always given in a single weekly dose, by either oral or intramuscular route. Four hundred and sixty nine rheumatoid arthritis patients were treated with methotrexate between 1 January 1985 and 31 March 1994. The date of evaluation was 1 November 1994. For all patients except 11 (2.3%) lost to follow up we had recent medical information from physical examinations (performed less than six months before the date of evaluation). We recorded information on medical and surgical history, current treatment, and the time and reason for methotrexate withdrawal. If methotrexate was still being used, we asked particularly about side effects and the effect of folic acid if prescribed. The analysis was performed on patients for whom we had at least 12 months of follow up data (n = 426), regardless of the duration of methotrexate treatment. The mean follow up was 4.65 (SD 2.34) years (range 1.07 to 9.98), with a total of 1981.15 person-years.

Rheumatoid arthritis controls

All rheumatoid arthritis patients admitted as inpatients to the unit between 1 January 1990 and 1 November 1994 and never treated with methotrexate were included in the control group (n = 420 rheumatoid arthritis patients). A register of admissions is available but there was no register for patients who attended the unit but who were not admitted. The starting date for the follow up of the control patients was arbitrarily defined as the date of the first hospital admission between 1 January 1990 and 1 November 1994. Four hundred and twenty patients were available for analysis. The mean (SD) number of hospital admissions for these patients during this period was 1.7 (1.3). One hundred and ninety nine patients were seen only once as inpatients. For patients who were admitted more than once during this period (n = 229), the mean follow up duration was 1.0 (0.74) years, with a total of 102 person-years (P = 0.0001, compared with the methotrexate treated group). The present analysis mainly represents a cross sectional study of the incidence of cancers in the rheumatoid population not treated with methotrexate.


Data were obtained from the cancer registry based on the years 1991 to 1992.22 All of these patients lived in the area around Montpellier (Department of Herault, France).


The number of newly diagnosed cancers in the rheumatoid arthritis controls and the regional population was compared with that of the methotrexate treated rheumatoid population using the life table (actuarial) method. We then compared the demographic, clinical, biological, and radiological data at baseline between the two groups (χ2 and Wilcoxon two sample test). Survival estimates were again performed adjusting for any potentially confounding factors found to have a significantly different distribution between the two rheumatoid populations. Comparison of the cancer incidence in the two rheumatoid populations with the incidence in the regional population was performed using an age adjusted incidence method on the regional population (Languedoc-Roussillon, 1991 to 1992). The reference population was the regional population. The ratios of the number of observed cases reported to the number of expected cases were compared to 1 by a χ2 test. P values of 0.05 were considered significant.



The main characteristics of the methotrexate treated rheumatoid patients (n = 426) are presented in table 1. Previous disease modifying treatments (DMARD) were stopped one month before introducing methotrexate, except in 54 patients whose previous DMARD were continued with the methotrexate, or for whom another DMARD was introduced with the methotrexate (mainly hydroxychloroquine and rarely sulphasalazine). Three patients took immunosuppressive drugs, that is, cyclophosphamide (n = 2) and azathioprine (n = 1), before methotrexate; 141 patients took prednisone at the onset of methotrexate [14.1 (SD 5.8) mg/d] and 63 took concomitant folic acid [17.1 (9.1) mg/week].

Table 1

Characteristics of 426 rheumatoid arthritis (RA) patients treated with methotrexate (MTX) and 420 not treated with MTX. Values are means (SD) or range                                                        

The main characteristics of the rheumatoid controls (n = 420) are also presented in table 1. Five control rheumatoid patients (1.2%) previously received an immunosuppressive drug: cyclophosphamide (n = 2) and azathioprine (n = 3). Azathioprine was given to control rheumatoid activity; cyclophosphamide was given in one case for rheumatoid vasculitis associated with glomerulonephritis, and in the other for pulmonary fibrosis. Only one patient with cancer (gastric epithelioma) was previously treated with cyclophosphamide (for vasculitis and glomerulonephritis).


We recorded eight cases of neoplasia out of 426 rheumatoid patients treated with methotrexate (1.88%, or 4.04 cases/1000 person years of observation). The characteristics of these patients are given in tables 2 and 3. None of these eight patients had previously received any immunosuppressive drug. Two patients died. The other patients are still considered to be in remission [follow up of 48.3 (18.1) months].

Cancer-free survival curves (event = occurrence of a cancer) in the two rheumatoid populations: treated with methotrexate (n = 426) and controls never treated with methotrexate (n = 420). The difference between the curves is significant (P = 0.0001, log-rank), indicating a higher occurrence of cancers in the rheumatoid controls.

Table 2

Characteristics of methotrexate (MTX) treated rheumatoid arthritis (RA) patients and of RA controls who developed cancers

Table 3

Characteristics of methotrexate (MTX) treatment in the eight rheumatoid arthritis patients who developed cancers under treatment


Out of the 420 rheumatoid control patients, we detected six new cancer cases (1.43%, or 58.8 cases/1000 person years of observation). The characteristics of these patients are given in table 2. The mean follow up for these six patients after the diagnosis of cancer was 44 (23.8) months. No differences were noted in the demographic data between patients with cancer in the methotrexate treated population and those with cancer in the control rheumatoid population (age at diagnosis of cancer, sex, presence or absence of Sjögren syndrome, duration of rheumatoid arthritis at diagnosis of cancer) (χ2 test and Mann-Whitney U test, data not shown).

The life table analysis showed that there was a significantly higher incidence of cancer in the control group (P = 0.0001, log-rank test) (see the figure, showing the cancer-free survival curves). We looked at potential confounding factors to explain the difference in cancer incidence. As shown in table 1, there are many factors that were differently distributed between the two rheumatoid groups. The lifetest procedure failed to show any influence of sex or rheumatoid factor positivity in the occurrence of cancer when comparing the two rheumatoid populations. The variables found to have a significantly different distribution between the two rheumatoid populations (table 1) were introduced in the multivariate analysis (Cox model). We noted that the only significant factor was age (P = 0.02); methotrexate treatment duration was not significant (P = 0.22). This result means that the difference observed in the survival curves between the two rheumatoid populations is explained by a difference in age distribution.

There was no significant difference in the frequency of immunosuppressive drugs used in the two populations of patients: 3/426 or 0.7% in the methotrexate population versus 5/420 or 1.2% in the control population (P = 0.5).


During the years 1991 to 1992, 6418 cases of cancer were reported out of a mean regional population of 812 344 people (0.79%, or 2.85 cases/1000 person years). Using the indirect standardisation method (reference population = regional population), in the methotrexate group we noted a ratio of eight observed cases for 8.31 expected cases (χ2 = 0.011, NS). Similarly, in the rheumatoid controls the ratio was six observed cases for 3.4 expected (χ2 = 1.98, NS). Therefore, adjusting for age, there was no significant difference in the frequency of cancer in the two rheumatoid populations compared with the regional survey.


We found a total of eight cancers in 426 rheumatoid arthritis patients treated with methotrexate [cumulative frequency of 1.88% during a mean follow up of 4.65 (2.34) years]. As in this study, previously reported cancers have been heterogeneous: thymoma,6 acute and chronic leukaemia,7 23bladder urothelial cancer, mediastinal teratoma, and squamous cell carcinoma of the thoracic skin.8 However, non-Hodgkin lymphoma is the main type of cancer reported9–15 20 23(table 4). These cases must be differentiated from the exceptional case of pseudolymphoma.16 We only observed one case of non-Hodgkin lymphoma in 426 rheumatoid arthritis patients treated with methotrexate; it should be stressed that in this patient—suffering from Sjögren syndrome and Hashimoto thyroiditis—it was a nodal and extranodal lymphoma localised in the thyroid gland. The incidence of non-Hodgkin lymphoma in patients with Sjögren syndrome is 43.8-fold greater than that expected in the general population.25Furthermore, the relative risk of malignant lymphoma of the thyroid gland in chronic lymphocytic thyroiditis is 67.26Consequently, the responsibility of methotrexate treatment in the development of this lymphoma is very difficult to establish.

Table 4

Characteristics of 11 rheumatoid arthritis patients treated with methotrexate (MTX) who developed non-Hodgkin lymphoma

The more frequent association of non-Hodgkin lymphoma and methotrexate treatment in rheumatoid arthritis patients is troublesome. The main argument implicating methotrexate in the appearance of lymphomas in rheumatoid arthritis patients is the complete regression of the tumour after methotrexate withdrawal alone.9 13 19 20 In the latter case,20 when methotrexate was reintroduced because of the evolution of rheumatoid arthritis, there was a recurrence of the tumour. Furthermore, Kamel et al 27 reported 15 cases of lymphoma in rheumatoid arthritis and three in dermatomyositis, of whom 33% (n = 6) presented with an EBV infection. Five of these six patients were treated with methotrexate, and the authors suggested that methotrexate as an immunosuppressor could have contributed to the development of these EBV associated lymphomas.27

An increased incidence of non-Hodgkin lymphomas in a rheumatoid arthritis population has been reported,1-4 the relative risk being 134 and 24.1,2 without any implication of cytotoxic drugs.2 3 28 This increased incidence of lymphomas in rheumatoid arthritis could explain the appearance of lymphomas in rheumatoid patients treated with methotrexate. Another argument against a causative role of methotrexate is that it did not augment the frequency of sister chromatid exchange (a sensitive method of detecting the mutagenicity of a molecule) in eight psoriasis patients treated for two to 10 months.29

We found a higher incidence of cancer (P = 0.0001) in the rheumatoid controls, but this difference disappeared when an adjustment for age was made. Nevertheless, there was still no increase in the frequency of cancer in the methotrexate treated group, contrary to what might be feared when using a drug with antifolate properties. The frequency of cancer in the rheumatoid controls could have been underestimated as not all the outpatient data were systematically available and no special attention was paid to these patients, in contrast to the methotrexate treated rheumatoid patients who have always been carefully followed up in our unit. Furthermore, in the analysis we decided to consider patients treated with methotrexate regardless of the treatment duration (even for only one month). Indeed, we wanted to test the worst hypothesis, by considering all cases of cancer diagnosed after the onset of methotrexate treatment as significant. As the data on the rheumatoid controls were only obtained from patients who had been admitted to hospital, the rheumatoid arthritis could be more severe in the control group than in the methotrexate group. It has been suggested that there is an increased risk of lymphoproliferative malignancy when rheumatoid arthritis is more severe. In fact, the rheumatoid arthritis was more active in the methotrexate treated patients, as shown in table1. Rheumatoid activity was not therefore a bias in interpreting the data in this study. Another important point is that, using the age adjusted incidence method, we noted the absence of any difference in the incidence of cancers when comparing each rheumatoid population to the regional population over two years (1991 and 1992). All these data suggest that methotrexate does not augment the incidence of cancer in rheumatoid arthritis. However, as the design of our study is not suited to evaluating the oncogenic potential of methotrexate, and as the follow up duration was short, especially in the control group, one must accept these results with caution. Clearly a longer term follow up is necessary to obtain more person-years at risk and also to answer the question about an increase in longer term risks. A good approximation of the numbers of patients required to show a statistical difference in the cancer occurrence in methotrexate treated and non-treated rheumatoid populations can be obtained using the comparison of two percentages. For the purpose of this calculation the following hypotheses were made: a cancer incidence of 1.88% in the methotrexate treated group as found in this study, a cancer incidence varying between 2% and 5% in the non-treated group (control) as previously published in different cohorts of rheumatoid arthritis patients,1-4 a bilateral test (no prior hypothesis on which type of rheumatoid arthritis population—treated with methotrexate or not—is exposed to a higher risk of cancer), and an α risk of 5%. In this way, the expected number of patients with a β risk of 20% (power of 80%) varies between 424 and 2454 (cancer frequency in the non-treated group of 5% and 3% respectively). These numbers are respectively 679 to 3933 with a β risk of 10% (power of 90%).

Bailin et al 30 previously showed that the incidence of cancer in 205 psoriatic patients treated with methotrexate was not significantly different from that expected in the general population. However, the presence of one case of lymphoma was higher than that theoretically expected (0.2 case). Rustin et al 31 studied the incidence of a second neoplasm in 457 patients treated with high doses of methotrexate for choriocarcinoma. The number of secondary cancers (n = 2) was lower than expected. McKendry and Dale reported on 144 rheumatoid arthritis patients treated with low doses of methotrexate32 in a 13 year retrospective survey. They observed a 12-fold increase in lung cancer (n = 4), but these patients were heavy cigarette smokers. The authors concluded that this increased incidence of lung cancer in the methotrexate treated rheumatoid population was probably a chance occurrence.32In an important retrospective study, 39 out of a total of 16 263 rheumatoid arthritis patients developed a haematological malignancy.33 Twelve had received methotrexate and 27 were never treated with this compound. Based upon the absence of specific histopathology of the haematologic malignancy in the methotrexate treated patients, the authors concluded that the risk from methotrexate is very small and does not appear to be related to cumulative doses of methotrexate or treatment duration.33

The small number of cancers occurring during methotrexate treatment in rheumatoid arthritis, their heterogeneous nature, and the absence of a higher incidence of cancer compared with that of rheumatoid controls and with the general population show that there is little chance that methotrexate is responsible for the generation of these cancers. Despite the difficulties in interpreting this study, the data presented here provide some reassurance for physicians on the use of methotrexate in rheumatoid arthritis. However, it is not possible to conclude that there is no risk with this drug on the basis of our study. Methotrexate could also be only a precipitating factor in the genesis of tumours. Longer prospective surveys of methotrexate treated rheumatoid patients compared to other rheumatoid patients not treated with methotrexate should be conducted to confirm this hypothesis. These studies should specifically target lymphomas, which are the most frequent neoplasms occurring in this situation and for which some intriguing data are reported. As it will be necessary to include a great number of patients in these prospective studies, a meta-analysis of the previously published studies could be another way in the future to answer this important question.


View Abstract

Request permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.