Risk of lymphoma in patients with dermatitis herpetiformisBMJ 1994; 308 doi: http://dx.doi.org/10.1136/bmj.308.6920.13 (Published 01 January 1994) Cite this as: BMJ 1994;308:13
- B Sigurgeisson,
- B A Agnarsson,
- B Lindelof
- Department of Dermatology, Karolinska Hospital, Stockholm, Sweden.
- Department of Pathology, University Hospital of Iceland, Reykjavik, Iceland.
- Correspondence to: Dr B Sigurgeirsson, Haaberg 39, 220 Hafnarfjordur, Iceland.
- Accepted 15 October 1993
Objective: To provide accurate estimates of the risk of lymphoma in patients with dermatitis herpetiformis.
Design: Comparison of observed and expected incidence of cancer in patients with dermatitis herpetiformis.
Subjects: 976 patients aged 4 to 97 years with no clinical signs of coeliac disease who were admitted to hospital between 1963 and 1983.
Setting: Data from Swedish Cancer Registry.
Main outcome measures: Incidence and type of cancer.
Results: 106 cancers were diagnosed in 94 patients. The relative risk was 1.4 (95% confidence interval 1.1 to 1.7) in male patients and 1.2 (0.8 to 1.7) in female patients. When the individual cancer sites were analysed a significant risk was found only for malignant, non-Hodgkin's lymphoma in male patients, with a relative risk of 5.4 (2.2 to 11.1).
Conclusions: The risk of lymphoma is greater in male patients with dermatitis herpetiformis, and this calls for increased surveillance in these patients.
Patients with dermatitis herpetiformis have long been thought to have an increased risk of developing cancer
This study found that non-Hodgkin's lymphoma seemed to be the only cancer significantly more common in such patients
Male patients were at greater risk than female patients
Patients with dermatitis herpetiformis should be checked regularly for sign of lymphoma
An increased incidence of gastrointestinal lymphoma has been documented in several studies of coeliac disease. Patients with dermatitis herpetiformis have long been thought to have an increased risk of developing cancer, and there are numerous case reports of patients with dermatitis herpetiformis and lymphoma. To our knowledge, however, no large study has been done to confirm this relation and estimate the risks of lymphoma. In a retrospective study of 109 patients with dermatitis herpetiformis malignant tumours were found in seven, giving a relative risk of 2.4.1 In Sweden another retrospective study of 96 patients, however, failed to show any relation with malignancy.2
It is therefore of interest to determine whether patients with dermatitis herpetiformis without clinical signs of coeliac disease are at risk of developing lymphomas. To provide more accurate estimates of the risk we studied a population based patient cohort of patients with dermatitis herpetiformis who were admitted to hospital in Sweden between 1963 and 1983. Follow up was complete up to 1987 and therefore ranged from five to 25 years after the first evaluation in hospital for dermatitis herpetiformis.
Patients and methods
From 1964 to 1983 the Swedish National Board of Health and Welfare collected nationwide information about people who were admitted to hospital in the country. In 1964, when this registry started, it covered only 20% of the population (8.3 million), but its coverage increased over the following years so that by 1969, 60% of the population was covered. Since 1970 this registry has been virtually nationwide. The information includes a unique individual identification number used in all population statistics in Sweden, the dates of admission and discharge, diagnosis, and nature and dates of operations.
From this registry we selected the diagnosis of dermatitis herpetiformis by using code 70400 from the Swedish adaptation of the International Classification of Disease, seventh revision (ICD-7), for 1964-8 and code 69399 from ICD-8 for 1969-83.
With these methods we identified 1040 patients in whom dermatitis herpetiformis was listed in the registry for the first time from 1964 to 1983. Patients with concomitant coeliac disease were excluded, leaving 976. There were 567 males and 409 females with a mean (range) age of 57.8(4 to 97) years. There were seven (0.7%) children (0 to 15 years of age) (table I). The average follow up was 8.9 years for cancer (to 1987) after correction for deaths during the study period. The number of person years of observation was 8662 after correction for deaths.
Swedish cancer registry
Information from the Swedish Cancer Registry, Stockholm (1958-87) and the 976 patients' records was correlated to identify people with cancer. Nationwide information on incidence of cancer in Sweden has been available since 1958, when compulsory registration began.3 The cancers are classified according to ICD-7 as recommended by the World Health Organisation. The cancer registry collects information on diagnosed cancers from clinicians and pathologists. If a person has more than one cancer, each is registered separately. Basal cell carcinomas are not registered. The completeness of registration is close to 100% for all cancers.4
Original medical records
As the number of patients was so large, and some patients had had numerous stays in different hospitals in Sweden since 1964, we judged it impossible to review all the hospital records, though we did review a sample of 63 records. A dermatologist (BL) reviewed the records and classified them. Six records were not available at the local hospitals, 51 (98%) patients were classified as definitely having dermatitis herpetiformis, and one (2%) patient was classified as not having this diagnosis. Dermatitis herpetiformis was diagnosed on clinical and histopathological grounds and was not always confirmed by immunohistochemistry.
We estimated the expected number of cancers by using incidence data from the cancer registry and a specially developed computer program, CANEST (CANcer ESTimates).5 The programme calculated the risk of cancer for each patient individually. The calculation was based on incidence data for the years between the first diagnosis of dermatitis herpetiformis and 1987 or death if the patient died before 1987. Information about deaths in the cohort was obtained from the Swedish Cause of Death Registry. Total risk was calculated from the sum of the individual risks. The relative risk of cancer, which is the ratio between observed and expected numbers of malignancies, was used to assess the risk of cancer. The observed numbers of cancers in the cohort were obtained from the cancer registry. We assumed that the distribution was normal and calculated 95% confidence intervals accordingly.6 This method has been described elsewhere in detail.5 A result was considered significant when the lower limit of the 95% confidence interval exceeded 1.0.
All statistics that were not carried out with the CANEST program and tabulations were processed with the SPSS statistical package.
Among the 976 patients with dermatitis herpetiformis, 106 cancers were diagnosed in 94 patients at the same time or after dermatitis herpetiformis was first diagnosed. The relative risk was 1.4 (95% confidence interval 1.1 to 1.7) in male patients and 1.2 (0.8 to 1.7) in female patients. This increase in risk was not significant when we excluded lymphomas. When we analysed individual cancer sites we found a significant increase only for malignant, non-Hodgkin's lymphoma in males (relative risk 6.4; 2.7 to 12.5). Table II summarises these results. The total number of lymphomas diagnosed in this group of patients (before and after diagnosis of dermatitis herpetiformis) was 14. Original paraffin blocks (in one case cytological material) of 12 of the 14 tumours diagnosed as lymphoma were made available to us. We reviewed this material and carried out immunoperoxidase studies with antibodies of leucocyte common antigen (LCA), B cell antigen CD20 (L26, Dako), and T cell antigen CD45RO (UCHL1, Dako). One of the 14 cases was on review found to be a carcinoma, leaving 13 cases of lymphoma. The mean time lag between the first admission to hospital and the diagnosis of lymphoma was 4.6 years, and the mean age at diagnosis of lymphoma was 79.6 years. Table III summarises the characteristics of these 13 cases.
The association between dermatitis herpetiformis and lymphoma was first proposed in 1970, when Gjone and Nordoy described a patient with dermatitis herpetiformis and lymphoma.7 Since then several cases of dermatitis herpetiformis and malignancy, particularly lymphomas, have been reported.*RF 8-10* At least six series have reported on the incidence of lymphomas among patients with dermatitis herpetiformis. Although not confirmed in all studies, there seems to be an increased incidence of lymphomas in patients with dermatitis herpetiformis (table IV). In the only study in which relative risks have been calculated a risk of 100 for developing lymphoma was found and a risk for cancer on the whole was 2.4.1
Our study confirms the increased risk of lymphoma, although the relative risk seems much lower (6 V 100) than in the study of Leonard et al.1 Interestingly, the ratio of patients with lymphoma is not so different (1.3% V 2.8%). They estimated the incidence of lymphoma by using figures from the 1976 cancer registry for the whole study period, but we calculated the incidence individually for each patient by using data on incidence specific for age and date for the corresponding years. Obviously the quality of the incidence data used to calculate the expected number of malignancies is as important as the observed number of cancers. In our study the estimated number of cancers is reliable as since 1958 the cancer registry has provided yearly incidence figures which allow for accurate estimations of cancers.
The increased risk of lymphoma in the gastro-intestinal tract of patients with dermatitis herpetiformis is probably the result of a polyclonal stimulation of lymphocytes by gluten that causes transformation into a malignant clone. If this is true patients on a gluten free diet probably have a lower risk of developing a lymphoma. Interestingly, however, in our study most lymphomas were found outside the gastrointestinal tract and most were of B cell phenotype. Only one case could thus be classified as T cell lymphoma associated with enteropathy as defined by O'Farrelly et al.11 This must be clarified in future studies. We found an average time lag of 4.5 years between the first admission to hospital and diagnosis of dermatitis herpetiformis and the subsequent diagnosis of lymphoma. If the time is calculated from the first appearance of symptoms it is probably longer, and this is supported by published case reports in which the mean time between diagnosis of dermatitis herpetiformis and lymphoma is much longer, as long as 14 years. 12 Many of the patients in our study may have had dermatitis herpetiformis for some time before the first admission for the condition. The same probably applies to the mean age of our patients, which is much higher than that previously reported (65 V 54). The mean age at diagnosis of lymphoma is high, and there is a high mortality after diagnosis.
Our findings show that patients with dermatitis herpetiformis have a moderately increased risk of cancer compared with the general population. If lymphomas are excluded this risk does not exist. On the other hand, patients with dermatitis herpetiformis have a definite increase in risk of lymphoma. In these patients an extensive search for lymphoma at regular intervals is therefore warranted.
This study was supported with grants from the Icelandic research foundation and the Swedish Welander foundation.