Assessing cardiotoxicity in randomised trials of anthracycline chemotherapy

Chemotherapy has improved survival in many types of cancer. However, it can lead to a range of toxic side effects that can occur while the patient is on treatment but may also appear months or years later. One of the more serious harmful side-effects of chemotherapy is heart disease and anthracycline chemotherapy is particularly implicated in this. Observational studies have shown that the risk of congestive heart failure in anthracycline treated patients rises rapidly after cumulative doses of about 400 mg/m2, but that there is also some risk below this dose. This thesis explored several methodological issues that affect meta-analyses of harms. First meta-analyses of harms often include terms for those harms in their search strategies. Evaluation of a trial search including cardiac terms against the trial list maintained by the Early Breast Cancer Trialists' Collaborative Group showed that only 48% of their trials were found. After cardiac terms were removed from the search strategy, 87% of the EBCTCG trials were found. Second, anthracycline cardiotoxicity is often not reported in a manner that enables data to be extracted for meta-analyses. To address this, risk of outcome reporting bias in trials included in this thesis was assessed. There was a high risk of outcome reporting bias for cardiac morbidity in 50% of trials, and 35% of trials had high risk of outcome reporting bias for cardiac mortality. Third, previous meta-analyses of anthracycline cardiotoxicity included only a limited number of trials. In order to increase the amount of information about anthracycline cardiotoxicity available for meta-analysis, a method was devised for including confounded trials in the meta-analysis, taking into account prior knowledge as to the likely effect of other drugs on cardiotoxicity. Finally, published and individual patient data meta-analyses of cardiac death in randomised trials of anthracycline versus not were undertaken. The published data meta-analysis included 32 trials, 9 of which were unconfounded, and found a 36% (95% CI 1-83%) increased risk of cardiac death in anthracycline-allocated patients compared to those not allocated to anthracycline. No conclusions could be drawn about the effect of confounder drugs on the risk of cardiac death. A meta-analysis grouped by lower and higher dose showed that there was a statistically significant increase in risk of cardiac death in anthracycline allocated patients compared to those not allocated anthracyclines in the higher dose group only (OR=1.74, 95%CI 1.15 – 2.62). In conclusion, this thesis has generated new insights into the methods used for meta-analyses of harms. It has also produced a new risk estimate of cardiac death in anthracycline-treated patients that clinicians can use to inform their treatment decisions.