Prediction models for treatment-induced cardiac toxicity in patients with non-small-cell lung cancer: A systematic review and meta-analysis

Background: To maximize the likelihood of positive outcome in non-small-cell lung cancer (NSCLC) survivors, potential benefits of treatment modalities have to be weighed against the possibilities of damage to normal tissues, such as the heart. High-quality data-driven evidence regarding appropriate risk stratification strategies is still scarce. The aim of this review is to summarize and appraise available prediction models for treatment-induced cardiac events in patients with NSCLC.

Methods: A systematic search of MEDLINE was performed using a Boolean combination of appropriate truncation and indexing terms related to "NSCLC", "prediction models", "cardiac toxicity", and "treatment modalities". The following exclusion criteria were applied: sample-size of less than 100, no significant predictors in multivariate analysis, lack of model specifications, and case-mix studies. The generic inverse variance method was used to pool the summary effect estimate for each predictor. The quality of the papers was assessed using the Prediction model Risk Of Bias Assessment Tool.

Results: Of the 3,056 papers retrieved, 28 prediction models were identified, including seven for (chemo-)radiotherapy, one for immunotherapy, and 20 for surgical resection. Forty-one distinct predictors were entered in the prediction models. The pooled effect estimate of the mean heart dose (HR = 1.06, 95%CI:1.04-1.08) and history of cardiovascular diseases (HR = 3.1, 95%CI:1.8-5.36) were shown to significantly increase the risk of developing late cardiac toxicity after (chemo-)radiotherapy. Summary estimates of age (OR = 1.17, 95%CI:1.06-1.29), male gender (OR = 1.61, 95%CI:1.4-1.85), and advanced stage (OR = 1.34, 95%CI:1.06-1.69) were significantly associated with higher risk of acute cardiac events after surgery. Risk of bias varied across studies, but analysis was the most concerning domain where none of the studies were judged to be low risk.

Conclusion: This review highlights the need for a robust prediction model which can inform patients and clinicians about expected treatment-induced heart damage. Identified clues suggest incorporation of detailed cardiac metrics (substructures' volumes and doses).