Response to ‘Non‐Hodgkin's lymphoma and nonsteroidal anti‐inflammatory drugs: A confounding problem’ by Harris
Year: 2004
Type: article
Abstract: Harris raises several concerns about our recently published paper, in which we reported a positive association of the use of non-steroidal anti-inflammatory agents (NSAIDs) with risk of non-Hodgkin lymphoma (NHL) in a prospective cohort of older women.1 The first major concern was that there were no data on NSAIDs exposure by type, strength or duration of intake—a limitation that we acknowledged in our discussion section. However, this is a common limitation to most of the epidemiologic studies conducted to date, in part due to the difficulty of collecting these types of data. For example, the case-control study cited by Harris that reported an inverse association of NSAIDs with risk of NHL was based only on an analysis of ever vs. never use of any NSAIDs for 4 or more consecutive weeks.2 The 2nd concern was the lack of a dose response, which was deemed “not consistent with causality.” We believe that it is premature to evaluate whether this a causal association, but it is accepted that a lack of dose response does not preclude a causal association.3 Third, and most seriously, is the concern that the association of NSAIDs with NHL risk is potentially confounded “with inflammation, treatment and immunosuppression in not just rheumatoid arthritis (RA) but a myriad of other immunosuppressive disorders and medications that have been etiologically linked to NHL development.” It should be noted that statistical adjustment for RA had little influence on the association of NSAID use with the risk of NHL in our data, and our results were similar when stratified by RA history (see Table 3).1 In additional analyses, we adjusted for the history of steroid use, a risk factor for NHL in this cohort,4 based on self-reporting from the 1986 questionnaire. This adjustment did not alter our findings, although a limitation of this analysis is that we did not have the use of steroids from 1986–1992. A recently published case-control study by Kato et al.5 also found a positive association of NSAID use with risk of NHL, and this was independent of corticosteroid use and did not appear to be confounded by indication for the use of NSAIDs. Of the “myriad” of other factors suggested by Harris, those that are clearly etiologically linked to NHL are rare, and can only account for a very small amount of NHL on a population basis.6 Thus, it is not clear that such factors are likely to be a major source of confounding in our study. There is also the potential for confounding by common lifestyle factors that may increase the risk of NHL, which we were able to evaluate but found no evidence for confounding.1 Finally, Harris cites selected mechanistic evidence supportive of a protective role for NSAIDs. However, as we noted in our discussion section, the in vivo effects of NSAIDs on cellular and humoral immunity are currently incompletely understood.7 cyclooxygenase (COX)-independent mechanisms may also be important.8, 9 In conclusion, there is currently no consensus with respect to the role of NSAIDs in the etiology of NHL, and we concur with Harris that future studies should be designed to carefully assess this association, including the role of confounding. Yours sincerely, James R. Cerhan, Kristin E. Anderson, Celine M. Vachon
Source: International Journal of Cancer
DOI: 10.1002/ijc.20087
Institutions: University of Minnesota, Mayo Clinic
Language: English
Subfield: Pharmacology
Field: Medicine
Domain: Health Sciences
Open Access status: bronze
OpenAlex ID: W2073082500