How publication bias overestimates the risk of atypical femoral fracture and osteonecrosis of the jaw associated with bisphosphonate use: a meta-epidemiological study.

Abstract

Objectives: Bisphosphonates are widely used to treat osteoporosis, bone cancer, Paget's disease, and hypercalcemia to reduce fracture risk. Observational studies have highlighted rare but serious adverse events (AE). Some meta-analyses have found evidence of publication bias, but rarely examined its impact or addressed it in their conclusions. Clinical guidelines are often based on meta-analyses and published literature, which may potentially bias the risk-benefit balance of bisphosphonates. This study assessed the presence and impact of publication bias in bisphosphonate safety evaluations.

Methods: Systematic reviews and meta-analyses of bisphosphonate-related AE were searched. Odds ratios from original clinical studies were collected, and publication bias was assessed using funnel plots, Egger's tests, and robust Bayesian meta-analysis (RoBMA). The effect of bias was quantified by comparing unadjusted and adjusted pooled estimates using trim-and-fill and RoBMA.

Results: The analysis included 42 systematic reviews totalizing 112 clinical studies (58% were observational), providing 148 unique point estimates for 10 AE. Publication bias concerned 2 out of 10 adverse events, high risk of publication bias was detected for atypical femur fractures and osteonecrosis of the jaw. For these, bias inflated effect estimates by 40-45% and 47-67%, respectively, with associations with bisphosphonates use disappearing after adjustment. Publication bias may be due to the observational study design for both, and the non-cancer indication for osteonecrosis of the jaw. No high risk of publication bias was found for eight other AE (atrial fibrillation, myocardial infarction, stroke, kidney dysfunction, oesophageal, breast, gastric, and colorectal cancer). This bias may lead to a distorted assessment of the risk-benefit balance.