Commentary: coeliac disease, mortality and malignancy – authors’ reply

Abstract

We thank Dr Ludvigsson for his thoughtful commentary1 on our study.2 We had decided a priori that it would be acceptable to present combined single estimates if (i) there was no significant heterogeneity and (ii) a sensitivity analysis showed no significant changes. In addition, we performed stratified analyses – such as the serology-only analyses – to further clarify the reported estimates. We agree that a specific undiagnosed vs. diagnosed analysis should have been done in addition to the serology vs. non-serology analysis. As noted by Dr Ludvigsson, the screening serology-only studies find no increased risk of mortality with positive coeliac serology patients.3-5 Our serology-only mortality analysis includes the screening studies but also includes Ludvigsson et al.6 which examined a population of serology-positive, biopsy-negative patients that were not screened. In contrast to the screening studies, Ludvigsson et al. found that serology-positive, biopsy-negative patients had an increased risk of mortality. Removal of Ludvigsson et al. from our serology analysis to form an undiagnosed coeliac disease analysis finds no increased risk of mortality, with an odds ratio (OR) of 1.10 [95% confidence interval (95% CI) 0.94–1.27], and no significant heterogeneity (I2 = 0.00, P = 0.43). One explanation is that as the population in Ludvigsson et al. was not screened, and were indicated to receive a biopsy, it is possible that this population is intrinsically less healthy than a screened population, and thus has an increased risk of mortality. An alternative explanation is that as the risk of death in coeliac patients is only raised by 1.24, it is possible that the screening studies have simply not been large enough to detect this risk. The total number of positive coeliac serology patients in all of the screening studies combined is only 492 in a total cohort of 14 895, whereas Ludvigsson et al. contained 3719 positive coeliac serology patients with negative biopsies in a total cohort of 232 351. Further large-scale cohort studies of screened positive coeliac serology patients will be required to resolve this issue. With regards to diagnosed coeliac disease patients, removal of the serology-only studies to form a diagnosed coeliac disease analysis did not significantly change our original combined estimate, with an OR of 1.26 (95% CI 1.20–1.32), and no significant heterogeneity (I2 = 0.00, P = 0.79). Similarly, the removal of the serology only studies from our non-Hodgkin lymphoma analysis did not significantly change our original combined estimate, with an OR of 2.68 (95% CI 2.09–3.44) and no significant heterogeneity (I2 = 22.97, P = 0.25). Meta-analysis of the two non-Hodgkin lymphoma risk estimates in undiagnosed coeliac disease reported in the only screening serology study by Lohi et al.7 did find an increased risk of non-Hodgkin lymphoma, with an OR 4.04 (95% 1.60–10.21) and no significant heterogeneity (I2 = 0.00, P = 0.41). This result, however, is derived from a population containing only 275 positive coeliac serology patients in a total cohort of 6849, and thus needs replication with larger studies. Dr Ludvigsson is correct that the study by Card et al.8 only shows small differences between estimates using external vs. internal comparisons. However, in addition to using evidence by Card et al. as a rationale to exclude external comparisons, we had also decided to exclude external comparisons due to the fact that the published estimates ranged enormously to the point where we felt it would be unlikely to produce a homogenous result, or even if so, produce a reliable result. A recent meta-analysis on the risk of non-Hodgkin lymphoma and coeliac disease that used external comparisons by Kane et al.9 pooled risk estimates ranging from 300 to 0.28. Even though via stratification Kane et al. ended up producing a homogenous result, we feel that their result of an OR of 4.42 (95% CI 3.72–5.26) is likely to be an overestimate of the actual risk. Declaration of personal and funding interests: None.