SEARCH

SEARCH BY CITATION

Summary

  1. Top of page
  2. Summary
  3. Introduction
  4. Methods
  5. Study population
  6. Statistical analysis
  7. Results
  8. Discussion
  9. Acknowledgements
  10. References

Background : There is recent evidence from studies of hospitalized and of undiagnosed patients that the risk of lymphoma for people with coeliac disease may be lower than previously thought. In addition, there have been no precise estimates of small bowel lymphoma risk due to a lack of population data.

Aim : To examine these and other malignant risks in a cohort of patients more typical of those seen in routine clinical practice.

Methods : A prospective cohort study of incident malignancy rates in patients with coeliac disease in southern Derbyshire compared with general population figures.

Results : During 5684 person years of follow-up 31 malignancies (excluding non-melanoma skin cancer) occurred in comparison with 30.30 expected [standardized incidence ratio (SIR) 1.02 (0.69–1.45)]. There were four non-Hodgkin's lymphomas (0.69 expected) SIR 5.81 (1.58–14.86), of which one originated in small bowel (0.02 expected) SIR 40.51 (1.03–225.68). GI malignancy occurred in nine (5.71 expected) SIR 1.58 (0.72–2.99), and breast cancer in three (5.08 expected) SIR 0.59 (0.12–1.73).

Conclusions : There is no increase in the risk of incident malignancy in this population and the risk of non-Hodgkin's lymphoma in general or of the small bowel is lower than previously found from UK coeliac cohorts.


Introduction

  1. Top of page
  2. Summary
  3. Introduction
  4. Methods
  5. Study population
  6. Statistical analysis
  7. Results
  8. Discussion
  9. Acknowledgements
  10. References

It is generally accepted that coeliac disease is associated with an increased risk of non-Hodgkin's lymphoma in general and of small bowel lymphoma in particular. This association was first described in the early 1960s1 and subsequently confirmed2–4 but as yet no consistent picture of its magnitude has emerged. Early studies indicated a relative risk for lymphoma of the order of 50–1003, 5, 6 but were in general from tertiary referral centres with an interest in coeliac disease and almost certainly included patients with malignancy present at the time of diagnosis of coeliac disease. Unsurprisingly, more recent studies not affected by similar selection biases have shown lower relative risks. In the last few years, case–control studies of lymphoma have suggested a relative risk of about 16 for gut lymphomas7, 8 and 3 for all non-Hodgkin's lymphoma7, 9 and a population-based cohort has indicated a relative risk of about 6.10 These studies however have given risk estimates not for coeliac disease treated in gastrointestinal clinics but variously for diagnosed and undiagnosed patients with coeliac disease and for those who had a diagnosis of coeliac disease given at discharge following a hospital stay. Increases in the risk of gastrointestinal carcinoma in general and of carcinomas of the mouth, pharynx, oesophagus and small intestine have also been reported.3, 6, 11 However, previous studies have not been able to estimate the relative risk of small intestinal lymphoma compared with the general population, as relevant population estimates of the incidence of this condition have not been readily available.

In the current study we have therefore set out to establish the risk of lymphoma and other malignancy in a population of people with coeliac disease representative of those diagnosed and treated in routine clinical practice in two district general hospitals and followed up prospectively since 1978. Our study was designed to minimize the hazards of selection bias.

Study population

  1. Top of page
  2. Summary
  3. Introduction
  4. Methods
  5. Study population
  6. Statistical analysis
  7. Results
  8. Discussion
  9. Acknowledgements
  10. References

All patients with coeliac disease diagnosed by small intestinal biopsy in southern Derbyshire at the two district general hospitals (Derbyshire Royal Infirmary and Derby City General Hospital) were identified and followed prospectively from 1978 to the end of 2001. The diagnosis of coeliac disease was based on characteristic small bowel biopsy appearances of severe or total villous atrophy. The diagnosis date of patients diagnosed clinically in childhood but who did not have an intestinal biopsy until adult life was considered as the date of the intestinal biopsy. During the whole of this period, regular fortnightly meetings with consultant histopathologists to review small bowel biopsies ensured that all cases of coeliac disease being diagnosed were identified. Information collected for the study included the date of death and date and type of malignancy that occurred. Information on incident cancers was obtained prospectively during clinical review in a dedicated coeliac clinic held weekly, regular review of case notes and comprehensive searching of the histological database of the pathology department of the Derby hospitals. Diagnoses of malignancy were recorded based upon diagnoses made in routine clinical practice. The collection of this information was co-ordinated by a single investigator (GKTH) and stored on a computerized database. The same type of information for all those diagnosed with coeliac disease between 1958, when the first diagnoses of coeliac disease were made in Derby, and 1978 was also collected. Extensive efforts were made to identify all patients in the area and sources of information included the hospital diagnostic index, histopathology records, the dermatitis herpetiformis clinic, the immunology laboratory, and the dietetic department and membership lists of the Coeliac Society. We excluded patients who were referred from other secondary care centres.

Statistical analysis

  1. Top of page
  2. Summary
  3. Introduction
  4. Methods
  5. Study population
  6. Statistical analysis
  7. Results
  8. Discussion
  9. Acknowledgements
  10. References

From the case series, we calculated counts of malignancies diagnosed and patient years of risk time observed by age and sex, and we defined three risk time periods. The first time period included all the prospective follow-up data from 1978 onwards. The second and third periods were calculated with respect to the diagnosis date of coeliac disease for each subject. The ‘peri-diagnosis period’ included 2 years after the diagnosis of coeliac disease. The ‘post-diagnosis period’ included all follow-up time at least 2 years after the diagnosis was made. Prevalent malignancy was defined as cancer being diagnosed prior to, or within 2 years after the diagnosis of coeliac disease. The data for patients without complete follow-up were censored at the date of their last follow-up. Population data against which to compare this were derived from two sources. For all malignancies except small intestinal lymphoma, data were derived from cancer registrations for England and Wales for 1988–92 and the population of England and Wales in the same years. These data were obtained from the Office of National Statistics.12 We estimated small intestinal lymphoma risk in the population on the basis of a recent audit carried out by the West Midlands Cancer Intelligence Unit in which the case notes of registered lymphomas were reviewed (personal communication, West Midlands Cancer Intelligence Unit). Approximate age–sex-specific rates were obtained by standardizing the non-age–sex-specific rates from this audit against cancer registration data, using ICD 0 information to assess whether lymphomas were of small bowel origin. The age–sex-specific rates from cancer registration data were then multiplied by the standardized incidence ratios (SIR) obtained from this in order to correct for the lack of complete ICD 0 data. We excluded basal cell cancers of the skin from the statistical analysis because the expected number of these tumours in the general population is uncertain.

The case series was indirectly standardized against the population to obtain SIRs. The results are presented with 95% confidence intervals (CI), based on the Poisson distribution for observed cases. All statistical analyses were carried out using Stata 7 (Stata Corporation, College Station, TX, USA). Ethical approval for this study was obtained from the South Derbyshire Local Research Ethics Committee.

Results

  1. Top of page
  2. Summary
  3. Introduction
  4. Methods
  5. Study population
  6. Statistical analysis
  7. Results
  8. Discussion
  9. Acknowledgements
  10. References

In total, there were 869 people with coeliac disease with follow-up in Derby to 31 December 2001, excluding one person who was referred from Sheffield. Follow-up was complete for all but three of these individuals. Of these, 772 were under follow-up on 31 December 2001 which based on a catchment population of about 500 000 would give an approximate prevalence of 0.15%. The total follow-up time observed for these patients was 7742 person years and there were a total of 1612 person years in the peri-diagnosis period (Table 1). Of the 869 people with coeliac disease two died prior to the start of prospective follow-up in 1978, these people had therefore no prospective follow-up data available. Thirty-seven of those who did have such data had a malignancy recorded before 2 years after diagnosis, and were therefore excluded and of the remaining 830, 193 had <2 years of follow-up in total and had therefore no uncensored time to contribute (i.e. no follow-up in the ‘post-diagnosis’ period). There were therefore, 637 patients with time to contribute to the analysis of non-prevalent malignancies and of these 546 were diagnosed after the beginning of 1978. In total, they contributed 5684 person years of follow-up in the post-diagnosis period (i.e. after the first 2 years for each case and time before 1978 had been censored). These 637 patients are further summarized in Table 2. The majority of people with coeliac disease were over the age of 40 years at diagnosis. Of those included in the analysis, 444 (70%) were women. The median follow-up time was 6.6 years (interquartile range 2.2–14.5 years) overall and was 24.0 years for prevalent and 5.2 years for incident cases.

Table 1.  Available follow-up time by sex to the nearest year. (The complete cohort figures exclude those who contributed no follow-up. Two who died before diagnosis of coeliac disease, and two whose last follow-up was at the date of diagnosis.)
 Number of casesPatient years of follow-up
Complete cohort
 F5935362
 M2722380
 Total8657742
Peri-diagnosis period
 F5771127
 M252485
 Total8291612
Postdiagnosis period
 F4444001
 M1931683
 Total6375684
Table 2.  Characteristics of patients contributing time to the study of incident cancers
 NumberPercentage
Age at diagnosis
 Under 2355
 2–20589
 20–4019330
 40–6021033
 Over 6014122
Sex
 Female44470
 Male19330
Years of analysed follow-up
 0–527343
 5–1014122
 10–157111
 15–20559
 Over 209715
Total637 

Of the whole group of 869 patients with coeliac disease, 83 had developed a total of 92 malignancies, two patients had three cancers, and five had two cancers. Of these 59 occurred either in the peri-diagnosis or postdiagnosis periods and were therefore relevant to the analysis (Table 3). Twelve lymphomas were encountered (all non-Hodgkin's). Five of the lymphomas primarily involved the small bowel of which three were enteropathy-associated T-cell lymphoma, one an anaplastic T-cell and one a B-cell lymphoma. Of these small bowel lymphomas the only one diagnosed more than 2 years after the diagnosis of coeliac disease was the anaplastic T-cell lymphoma, and the only patient to survive more than 1 year was the one with B-cell lymphoma. Five of the lymphomas occurred in the peri-diagnosis period of which three were enteropathy-associated T-cell lymphoma and one a B-cell lymphoma. In the post- diagnosis period, four lymphomas were encountered, of which one was an anaplastic lymphoma arising in the small bowel. Only two patients developed carcinoma of the small intestine, two carcinoma of the oesophagus and 12 had colorectal cancer. A variety of other cancers were encountered in small numbers of patients apart from seven with breast and nine with lung cancer.

Table 3.  Counts of malignancies recorded for patients with coeliac disease subdivided by time period. Malignancies peri-diagnosis occurred between 2 years prior to diagnosis and 2 years after (or death if earlier). Malignancies postdiagnosis exclude those which are censored due to a previous prevalent malignancy in the patient.
 Malignancies peri- diagnosisMalignancies post- diagnosis
Type of cancerFrequencyPercentageFrequencyPercentage
Breast26.6738.82
Lung26.67617.65
Lymphoma (not SB)26.6738.82
SB lymphoma313.3312.94
Colorectal CA516.67411.76
Stomach CA13.3325.88
Oesophageal CA  25.88
Pancreatic CA26.67  
Hepatoma  12.94
SB CA13.33  
Non-melanoma skin CA416.6738.82
Unknown primary  25.88
Melanoma13.33  
TCC bladder  25.88
Myeloma  12.94
Tongue  25.88
Vulva13.33  
Ovary  12.94
Parotid 3.33  
Cervix 3.3312.94
Prostate13.33  
Chronic lymphatic leukaemia 3.33  
Acute myeloid leukaemia    
Seminoma    
Teratoma    
Endometrial CA    
Total2510034100

Table 4 shows the calculation of SIRs for selected groups of malignancies both in the peri- and post-diagnosis periods. Overall the risk of malignancy showed a statistically significant increase in the peri-diagnosis period (SIR 2.00; 95% CI 1.24–3.06), and was almost exactly the same as for the general population in the post-diagnosis period (SIR 1.02; 95% CI 0.70–1.45). For small bowel lymphomas and for non-Hodgkin's lymphoma there was a statistically significantly increase in risk in both periods. This increase was however, far more marked for the peri-diagnostic period for small bowel lymphoma (SIR 358.80; 95% CI 74.01–1048.34) and for all non-Hodgkin's lymphoma (SIR 20.94; 95% CI 6.80–48.86) than for the post-diagnostic period (SIR 40.51; 95% CI 1.03–225.68 and 1.03–5.80; 95% CI 1.58–14.86 respectively). There was also an increase in gastrointestinal carcinomas (both small bowel and in general) limited to the peri-diagnostic period, and a slight, non-statistically significant reduction in risk of breast cancer (SIR 0.59; 95% CI 0.12–1.73) in the postdiagnosis period. No other SIRs examined were significantly different from 1 at the 5% level.

Table 4.  Observed and expected malignancies and standardized incidence rates for the peri-diagnosis time period (between diagnosis and 2 years after diagnosis) and the postdiagnosis period (at least 2 years after diagnosis)
 ObservedCrude risk/100 000ExpectedSIR95% CI
  1. * All malignancies exclude non-melanoma skin cancers.

Peri-diagnosis period
 All malignancies*21130510.492.001.24–3.06
 Small intestinal lymphoma31870.01358.8074.01–1048.34
 Non-Hodgkin's lymphoma53110.2420.946.80–48.86
 Small intestinal carcinoma1620.0259.971.52–334.12
 GI carcinomas95592.124.241.94–8.05
 Breast cancer21241.591.260.15–4.54
 Lung cancer21241.481.350.16–4.88
Postdiagnosis period
 All malignancies*3154630.301.020.70–1.45
 Small intestinal lymphoma1180.0240.511.03–225.68
 Non-Hodgkin's lymphoma4700.695.801.58–14.86
 Small intestinal carcinoma000.0500.00–78.77
 GI carcinomas91585.711.580.72–2.99
 Breast cancer3535.080.590.12–1.73
 Lung cancer61063.971.510.55–3.29

Discussion

  1. Top of page
  2. Summary
  3. Introduction
  4. Methods
  5. Study population
  6. Statistical analysis
  7. Results
  8. Discussion
  9. Acknowledgements
  10. References

In this large, prospective, population-based study we found no overall increase in the rate of incident malignancy in people with diagnosed coeliac disease compared with the general population. When the site-specific malignancies were examined, we found that patients with diagnosed coeliac disease had a five times greater rate of non-Hodgkin's lymphoma and a 40 times greater rate of small bowel lymphoma compared with the general population. Although these relative rates appear high, the absolute rate of non-Hodgkin's lymphoma and small bowel lymphoma in people with diagnosed coeliac disease are small; one in 1421 person years and one in 5684 person years respectively, indicating the rarity of the occurrence of such malignancies. Of additional interest was the finding of slightly fewer than expected cases of breast cancer.

The data used in this study were collected prospectively from 1978 in a systematic and rigorous manner by a single investigator. This cohort of patients from southern Derbyshire is unique in studies of malignancy in coeliac disease as it represents all cases occurring in a well-defined secondary referral area. Patients come from both primary and secondary care sources and follow-up is over 99% complete. The criteria for the diagnosis of coeliac disease were based on characteristic appearances in small bowel biopsies and remained unchanged throughout the whole of the prospective follow-up period. Referral bias is unlikely to have affected the results because the Derby hospitals are not tertiary referral centres for gastrointestinal diseases and patients in the study all derived from the southern Derbyshire catchment area. The fact that there has been an interest in coeliac disease in Derby for many years may mean that case ascertainment and follow-up are more complete here than elsewhere. If this is the case then it might result in an apparently lower risk of malignancy in this area if the risk is greater where disease is more symptomatic. Such a reduction in risk however could be argued to represent a more accurate assessment of the risk of coeliac disease as it would occur only if previous studies were based on a selected subgroup of patients with higher risk. In addition we excluded the one identified tertiary referral from our analysis. The meticulous surveillance of the cohort is likely to have identified all of the incident malignancies following the diagnosis of coeliac disease and this is probably more accurate than national registration data which have previously shown an approximate 10% underestimation of cancer rates.13 If at all, these factors may have led to a slight overestimation of the relative cancer rates.

Our use of national cancer registration data to estimate the general population rate of malignancy is routine in studies of this type. However, we were particularly interested in the rate of small bowel lymphoma, which has not had a specific ICD 8 or 9 code but for which some data are available using the ICD 0 data. As ICD 0 data are far from complete, estimates of the incidence of this condition based upon national statistics are likely to be inaccurate. We therefore used estimates of the incidence of small intestinal lymphoma generated in a study performed in the West Midlands. These estimates are likely to be robust because of the validation of each registered case of lymphoma by reference to original case notes. Our study is the first to attempt to estimate the risk of small intestinal lymphoma compared with the general population, presumably because no previous studies have had relevant population data available.

In order not to inflate the risks, cancers occurring within 2 years of the diagnosis of coeliac disease in our study were excluded from consideration because investigation of these might have led to the diagnosis of coeliac disease. Other studies have used a shorter 1-year exclusion period or none at all and may have allowed inclusion of patients with already established malignancy.6, 10 It is interesting that in a cohort of over a 1000 patients with coeliac disease gathered from centres in Italy the risk of death from non-Hodgkin's lymphoma was 70 times that expected.14 The excess deaths however, occurred from non-Hodgkin's lymphoma, ulcerative jejunitis or other cancers within 3 years of the diagnosis of coeliac disease. After this time the coeliac patients had a similar death rate to that of the general population. This suggests that those dying within the first 3 years did so of a malignancy that precipitated the diagnosis of coeliac disease. If we had used a cut-off of 3 years after the date of diagnosis of coeliac disease we would have excluded the single small bowel lymphoma included in our study. This would have dramatically reduced the relative rate of small bowel lymphoma to zero.

For some specific malignancies, our results are in agreement with other recent studies of cancer incidence in coeliac disease. The overall observation of no increased rate of malignancy in the postdiagnosis period is similar to the largest reported population-based study, from Sweden.10 Askling et al. found that with increasing length of follow-up the SIR for people with coeliac disease was only slightly and non-significantly elevated. Our point estimate of the risk of all malignancy in the peri-diagnosis period (1.24) was similar to theirs but not significant at the 5% level. Their finding of an SIR of 1.3 was limited to adults, whereas we have included all ages. In addition, they had considerably more statistical precision for their estimate but included only hospitalized cases of coeliac disease. Other studies finding a higher level of risk are in general older3, 4 (and might therefore be describing a more symptomatic group of patients), but have also in many cases not censored early malignancies and may therefore be susceptible to ascertainment bias. The absolute number of lymphomas occurring in patients with coeliac disease is small. In our series, only 12 lymphomas were identified and of these only five involved the small bowel of which only three were enteropathy associated T-cell lymphoma. Others have similarly found very few cases of enteropathy-associated T-cell lymphoma.7, 10 We also found no incident small bowel carcinoma; this likewise is a very rare tumour.15 Interestingly, we found no increase in the risk of gastrointestinal carcinomas in general. The observation of fewer cases of breast cancer than expected, although not statistically significant, is of interest particularly in the light of similar observations in previous reports.10, 16 The reasons for this decrease could be related to the lower body weight of people with coeliac disease17 or their lower rates of smoking.18

In conclusion, in this large, prospective, population-based cohort study we found no overall increase in the risk of incident malignancy in clinically diagnosed coeliac disease. Patients with coeliac disease are at greater risk of non-Hodgkin's lymphoma and small bowel lymphoma than the general population but the absolute risk (1 in 1421 person years and 1 in 5684 person years respectively) is very small. In addition we have demonstrated no increase in risk of gastrointestinal carcinoma and found fewer than expected cases of breast cancer.

Acknowledgements

  1. Top of page
  2. Summary
  3. Introduction
  4. Methods
  5. Study population
  6. Statistical analysis
  7. Results
  8. Discussion
  9. Acknowledgements
  10. References

T. R. Card and J. West are both supported by Wellcome Training Fellowships in Clinical Epidemiology (numbers 060529 and 063800). We are grateful to the West Midlands Cancer Intelligence Unit for the use of their audit data.

References

  1. Top of page
  2. Summary
  3. Introduction
  4. Methods
  5. Study population
  6. Statistical analysis
  7. Results
  8. Discussion
  9. Acknowledgements
  10. References
  • 1
    Gough K, Read A, Naish J. Intestinal reticulosis as a complication of idiopathic steatorrhoea. Gut 1962; 3: 2329.
  • 2
    Austad WI, Cornes JS, Gough KR, McCarthy CF, Read AE. Steatorrhea and malignant lymphoma. The relationship of malignant tumors of lymphoid tissue and celiac disease. Am J Dig Dis 1967; 12: 47590.
  • 3
    Harris OD, Cooke WT, Thompson H, Waterhouse JA. Malignancy in adult coeliac disease and idiopathic steatorrhoea. Am J Med 1967; 42: 899912.
  • 4
    Holmes GKT, Stokes PL, Sorahan TM, Prior P, Waterhouse JA, Cooke WT. Coeliac disease, gluten-free diet, and malignancy. Gut 1976; 17: 6129.
  • 5
    Selby WS, Gallagher ND. Malignancy in a 19-year experience of adult celiac disease. Dig Dis Sci 1979; 24: 6848.
  • 6
    Holmes GKT, Prior P, Lane MR, Pope D, Allan RN. Malignancy in coeliac disease – effect of a gluten-free diet. Gut 1989; 30: 3338.
  • 7
    Catassi C, Fabiani E, Corrao G, et al. Risk of non-Hodgkin lymphoma in celiac disease. JAMA 2002; 287: 14139.
  • 8
    Johnston SD, Watson RG. Small bowel lymphoma in unrecognized coeliac disease: a cause for concern? Eur J Gastroenterol Hepatol 2000; 12: 6458.
  • 9
    Mearin ML. Coeliac disease and non-Hodgkin's lymphoma. In: 10th International Symposium on Coeliac Disease. Paris, 2002.
  • 10
    Askling J, Linet M, Gridley G, Halstensen TS, Ekstrom K, Ekbom A. Cancer incidence in a population-based cohort of individuals hospitalized with celiac disease or dermatitis herpetiformis. Gastroenterology 2002; 123: 142835.
  • 11
    Swinson CM, Slavin G, Coles EC, Booth CC. Coeliac disease and malignancy. Lancet 1983; 1: 1115.
  • 12
    Cancer 1971–1997. CD ROM. London: Office for National Statistics, 1998.
  • 13
    Dickinson HO, Salotti JA, Birch PJ, Reid MM, Malcolm A, Parker L. How complete and accurate are cancer registrations notified by the National Health Service Central Register for England and Wales? J Epidemiol Community Health 2001; 55: 41422.
  • 14
    Corrao G, Corazza GR, Bagnardi V, et al. Mortality in patients with coeliac disease and their relatives: a cohort study. Lancet 2001; 358: 35661.
  • 15
    Howdle PD, Jalal PK, Holmes GKT, Houlston RS. Primary small-bowel malignancy in the UK and its association with coeliac disease. QJM 2003; 96: 34553.
  • 16
    Logan RF, Rifkind EA, Turner ID, Ferguson A. Mortality in celiac disease. Gastroenterology 1989; 97: 26571.
  • 17
    West J, Logan RF, Hill PG, et al. Seroprevalence, correlates, and characteristics of undetected coeliac disease in England. Gut 2003; 52: 9605.
  • 18
    Snook JA, Dwyer L, Lee-Elliott C, Khan S, Wheeler DW, Nicholas DS. Adult coeliac disease and cigarette smoking. Gut 1996; 39: 602.