Meta-analysis of side effects (SEs) of aromatase inhibitors (AI) and of tamoxifen (TAM) in large randomized clinical trials (RCT)

Abstract

11015 Background: The published data from large RCTs for the endocrine adjuvant treatment of postmenopausal breast cancer patients (ATAC and BIG 1–98) have shown us the difference in SEs between an AI vs TAM. Observed SEs in these trials are the result of an effect of both drugs: e.g. AIs induce bone loss and lead to an increase in fracture risk whereas TAM reduces fracture risk. We did a meta- analysis of comparable SEs in ATAC and BIG and combined this information with a second meta-analysis comparing an AI vs nil and a third meta-analysis comparing TAM vs nil. Methods: A relative risk (RR) meta-analysis was done on published data of SEs comparing AI vs TAM (ATAC Lancet 2002; BIG 1–98 NEJM 2005). SEs studied were: (1) fracture risk; (2) ischemic cerebrovascular events; (3) hypercholesterolemia; (4) hot flushes; (5) venous thromboembolic events and (6) any musculoskeletal events. Data for the comparison of AI vs nil were taken from MA.17 (NEJM 2003) and the combined arm of ATAC (Lancet 2002). Data for the comparison of TAM vs nil were taken from the P1 (JNCI 1998) and IBIS 1 (Lancet 2002). The assumption of a fixed effects model was used unless significant heterogeneity (Cochran Q test) was found. Results: In the comparison of fractures (ATAC and BIG combined) there was a 50% increase in RR for fractures in the AI arm. Our second meta-analysis showed that half of this increase was due to an effect of the AI and (according to the third meta- analysis) about half was due to a decrease of the RR for fractures thanks to TAM. Significant differences in cerebrovascular, hypercholesterolemia and hot flushes related events were due to a TAM effect. Concerning musculoskeletal events, significant heterogeneity was found when comparing the effect of an AI in the presence or absence of TAM. When patients get TAM together with an AI (ATAC combined arm) there is no increase in musculoskeletal events compared to the TAM alone arm. When an AI was compared to nil (MA.17) a significant increase in events was observed. We hypothesize that this heterogeneity is due to an estrogenic effect of TAM. Conclusion: This RR analysis allowed us to dissect the SEs profile of an AI and of TAM. The results for musculoskeletal events generate a new biological hypothesis. No significant financial relationships to disclose.