Premium
Assessing the possible association of anti‐TNF use with new malignancies: A neglected methodological consideration
Author(s) -
Ozguler Yesim,
Yazici Yusuf,
Hatemi Gulen,
Tascilar Koray,
Yazici Hasan
Publication year - 2018
Publication title -
pharmacoepidemiology and drug safety
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.023
H-Index - 96
eISSN - 1099-1557
pISSN - 1053-8569
DOI - 10.1002/pds.4579
Subject(s) - medicine , malignancy , rheumatoid arthritis , incidence (geometry) , cancer , medline , population , meta analysis , oncology , environmental health , physics , political science , law , optics
Purpose Whether anti‐TNF agents increase, malignancy risk remains debated. The potential problems with comparing the number of observed malignancies in anti‐TNF treated cohorts to number of expected malignancies in national cancer registries could have been an important issue. National registries give expected number of malignancies per organ involved regardless of if this is a secondary or a higher malignancy while anti‐TNF agents are not or are sparingly used in patients with an existing or history of a malignancy. This results in a biased standardized incidence ratio (SIR). We aimed to formally look at the presence and the magnitude of this bias in published work. Methods A systematic search was performed in PUBMED/MEDLINE about anti‐TNF use in rheumatoid arthritis (RA) up to July 2016. The main outcome sought was malignancy incidence in a RA cohort as compared with that in the general population. Randomized controlled trials, meta‐analyses, and articles published in languages other than English were excluded. Results The potentially biased SIR was discussed in only 1 of 29 studies we analyzed, while in none of these studies, the authors highlighted this as a limitation. We were able to approximate the pooled SIR for overall malignancy in 12 studies before and after correcting for this bias. It was 0.89 (CI: 0.84‐0.95) before and increased to 1.05 (CI: 0.99‐1.11), when corrected. Conclusion Planning and assessment of studies investigating malignancy incidence possibly related to anti‐TNF use in RA, and probably in other diseases, should take into consideration the described bias in SIR calculations.