Incidence of pancreatitis in HIV-infected patients, and the association with antiretroviral therapy

We recently reported a low incidence of pancreatitis in a European cohort of HIV-positive individuals followed prospectively from 2001 to 2006, with a rate of 1.27 cases per 1000 person-years ^[1]. Fessel and Hurley ^[2], in an editorial comment in the same issue of AIDS, reported a much higher incidence of approximately five times that seen in our study in the years 1996-2006 in a North American cohort. The authors noted that the rate of pancreatitis remained constant over time. However, we feel that there are important differences in the definition of pancreatitis used in the two studies that may go some way to explaining the disparate incidence rates observed.

Fessel and Hurley use a definition of pancreatitis based on either the presence of plasma lipase greater than four times the upper limit of normal (ULN), amylase greater than six times the ULN, or a pancreatitis diagnosis captured in the electronic medical record. In contrast, the EuroSIDA study used a detailed case definition of pancreatitis, and all events were source verified, reviewed, and classified centrally by the study physicians. Even when considering presumptive pancreatitis, two of the following three events were required: one or more characteristic symptoms or characteristic signs of pancreatitis; raised enzymes; at least one imaging investigation suggesting pancreatitis according to a radiologist or clinician. Furthermore, raised amylases were only considered as a pancreatitis event if other aetiology could be excluded. Only if definitive source documentation could not be obtained was a pancreatitis event assumed without further investigation. Thus, we suggest that the EuroSIDA study group used more stringent criteria to define pancreatitis events that included exclusion of other possible causes of abnormal laboratory values, and required the presence of clinical manifestation of disease in nearly all cases. Thus, a lower incidence of pancreatitis would be expected in our study when compared with that found when using the definition employed by Fessel and Hurley.

Additionally, the authors highlight the lack of an association between pancreatitis and the use of stavudine and didanosine. Although they do not investigate whether this association is present in their cohort, they highlight the fact that a number of other studies have observed such an association ^[3,4]. The authors rightly highlight the fact that use of didanosine and stavudine is less widespread in more recent years. Indeed, much of the stavudine and didanosine use in the EuroSIDA cohort is likely to be historical, rather than current. Awareness of the potential link between these antiretrovirals and pancreatitis may have led to less use of this combination as other nucleosides were developed, and to a reduction in the use in patients most susceptible to pancreatitis. Those susceptible to this complication may have already stopped the antiretroviral(s) prior to the study period, either because of the prior occurrence of pancreatitis, or because of other related issues.

In addition to the helpful suggestions made by Fessel and Hurley, we would also highlight the importance of applying consistent case definitions between studies so that results can be reliably compared. We have already begun further work, investigating the association between pancreatitis and triglycerides ^[5], and strongly agree that further research is needed in this subject area.