Article Text


Non-Hodgkin's lymphoma after immunosuppressive therapy
  1. CRC Cancer Epidemiology Unit, University of Oxford
  2. Radcliffe Infirmary, Oxford OX2 6HE, UK

    Statistics from

    See article on page 514

    The development of renal transplantation in the 1960s, made possible by azathiopine, also allowed the testing of the immunosurveillance proposed by Burnett and Thomas, that cancer has its origins in mutations which would usually be eliminated by a healthy immune system. An increased incidence of non-Hodgkin's lymphoma (NHL) in transplant recipients was soon recognised but not the generalised increases of malignancy that seemed to be predicted by the theory. Instead, increases of skin, liver, and (possibly) cervix cancers, and of Kaposi's sarcoma suggested that immunosurveillance operates primarily against malignancies of infective origin, as animal work had also indicated. It has become clear that this hazard is a feature of other immunosuppressive drugs, including 6-mercaptopurine, cyclophosphamide, cyclosporine, and methotrexate, and is not dependent on the presence of foreign antigens, extending to patients without organ transplants. In fact, an increased incidence of NHL has been found in virtually every type of marked immune impairment (whether or not treatment related) that has been studied in appreciable numbers, included AIDS and rare genetically determined disorders such as the Wiskott Aldrich syndrome.1

    The paper by Farrell and colleagues2 in this issue ofGut on the risk of cancer in 782 patients with inflammatory bowel disease, 30% of whom received immunosuppressive treatment (other than steroids), provides an opportunity for considering certain aspects of such therapy (see page514). There can be no doubt that the excess of lymphomas in their study is related to immunosuppressive therapy, not only because the excess was confined to the group treated in this way, but also because of the totality of the evidence referred to above. It is sometimes implied that an excess of lymphomas needs to be established following such treatment for a particular disorder, as if here it might be free of risk. This is not necessary for no example is known of an iatrogenic carcinogenic risk being restricted in this way; it is thetreatment (the exposure) that is relevant. The size of the risk evident in any given situation is, of course, likely to be influenced by the dosage and the duration of treatment and, as always, the play of chance. The interpretation of an excess of NHL may be complicated by the possibility that it reflects effects of the disorder under treatment, as in rheumatoid arthritis; here, however, the 9.7-fold increase found in seven pooled studies was significantly greater than the 2.2-fold increase in the absence of immunosuppressive treatment (14 studies).1

    The increased incidence of NHL, reported by Farrell and colleagues in their patients is, at over 55-fold, strikingly large and much greater than that usually reported in non-transplant patients treated with immunosuppressive drugs.1 Indeed, it also exceeds the ten-fold increase recently recorded in Scandinavian renal transplant recipients.3 This may partly reflect the doses and duration of treatment but another factor also merits attention. The register of patients with inflammatory bowel disease treated since January 1990, on which the study was based, was only “set up in mid-1996”, by which time three of the four cases of NHL had already been diagnosed. Given that all the patients attended one hospital, the authors were likely to have known of some of these lymphomas and, though not mentioned as a reason, this may have influenced the decision to set up the register for the study. But if so, inclusion of all the lymphomas (and the person years of exposure prior to mid-1996) in the analysis would have introduced bias and rendered the statistical tests invalid.

    At first sight, the logic may not be obvious of analysing data onlyafter the date of diagnosis of the most recent lymphoma to arouse interest. Suppose the “real” increased incidence was only half of that observed, say 25-fold, and that the expected number prior to mid-1996 was 0.04 (not unreasonable given the value of 0.06 for the whole period 1990–99). This would lead on average to one case of NHL (but with a wide range in the confidence interval). Often there would be no lymphoma (as in many transplant centres in the 1970s4) but sometimes two or three. The absence of a case could not prompt a study so this might not be undertaken, whereas the atypically increased occurrence of two or three cases, from the upper part of the confidence range, might well do so. As a general principle, inclusion in the analysis of the cases of disease that led to a study in the first place is likely to upwardly bias the magnitude of the risk estimate. It would seem probable that the 58-fold increase recorded by Farrell and colleagues is indeed inflated by chance.

    Even marked increases of relatively rare diseases such as NHL can go undetected except in large studies and, even more so, can small increases of other malignancies. Transplant recipients have been the main subject of large comprehensive studies of cancer after immunosuppressive therapy, and recently increases of a wider range of sites have been reported than found previously, including colon and lung cancers.3

    The statement of the authors that the incidence of lymphomas in their study is low in absolute terms seems inappropriate and invites misinterpretation. Taken at face value, the four lymphomas caused by treatment (in one in 60 patients) exceeds the total expected number of malignancies of all types in this group (2.15). Even if the incidence of NHL has been exaggerated as indicated above, it is clear that any iatrogenic malignancy, especially among patients with a disorder that is not invariably fatal, must represent grounds for great concern. That these risks are not fixed is indicated by the marked decline in the incidence of post-transplant lymphomas that has occurred since 1970 as doses of steroids and immunosuppressive drugs have been reduced.1 The weighing of benefits and hazards (and explaining these to the patient) is important here as in so much of modern clinical medicine. The study by Farrell and colleagues highlights the risks of immunosuppressive drugs (in any disorder) and the need to keep their use to a reasonable minimum.

    See article on page 514


    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.

    Linked Articles