Castellsague J, Calingaert B, Rothman KJ, Gutierrez L, van Herk-Sukel MPP, Kuiper JG, Pottegard A, Hallas J, Berglind IA, Sundstrom A, Dedman D, Gallagher A, Kaye JA, Pardo C, Perez-Gutthann S. Probabilistic bias analysis for unmeasured confounders in a study of users of topical tacrolimus, pimecrolimus, and corticosteroids (JOELLE) study. Poster presented at the 32nd ICPE International Conference on Pharmacoepidemiology & Therapeutic Risk Management; August 26, 2016. Dublin, Ireland. [abstract] Pharmacoepidemiol Drug Saf. 2016 Aug; 25(Suppl 3):142-3.

BACKGROUND: Topical tacrolimus (TAC) is indicated for the treatment of moderate to severe atopic dermatitis (AD), and topical pimecrolimus (PIM) for mild to moderate AD. Severity of AD could confound the association between the risk of lymphoma and skin cancer and the use of these medications. We used type of prescriber of first prescription (TPFP) as a proxy for AD severity in a study where information on AD was available in two of the four databases used.

OBJECTIVES: To use probabilistic bias analysis to correct partially adjusted incidence rate ratios (IRR) of lymphoma and skin cancer comparing new users of TAC and PIM with users of moderateto highpotency topical corticosteroids (TCS).

METHODS: Cohort study in the PHARMO Database Network (Netherlands), the Danish and Swedish national registers, and the Clinical Practice Research Datalink (CPRD) (United Kingdom), with RTIHS acting as coordinating/pooled analysis center. We used available information on TPFP in PHARMO and Sweden to estimate input parameters for probabilistic bias analysis to correct IRRs and 95% confidence intervals (CIs) in Denmark and CPRD.

RESULTS: We included 19,948 children and 66,127 adults treated with TAC matched with 79,700 children and 264,482 adults treated with TCS and 23,840 children and 37,417 adults treated with PIM matched with 90,268 children and 149,671 adults treated with TCS. The corrected IRR (95% CI) for lymphoma comparing TAC vs. TCS in children decreased by 28.9% from 5.26 (1.1424.29) to 3.74 (1.0014.06). The corrected IRR for cutaneous Tcell lymphoma (CTCL) comparing TAC vs. TCS in adults decreased by 35.1% from 2.71 (1.355.44) to 1.76 (0.813.79). For PIM, the corrected IRR for lymphoma in children decreased by 40.9% from 1.81 (0.418.02) to 1.07 (0.254.60), and the corrected IRR for CTCL in adults increased by 18.0% from 1.11 (0.284.32) to 1.31 (0.335.14). Smaller IRR reductions were observed for skin cancer for both TAC and PIM.

CONCLUSIONS: Probabilistic bias analysis for unmeasured confounders led to noticeable corrections of the effect of TAC and PIM on lymphoma and skin cancer.