Causal effects of methotrexate reductions/delays in treatment of high-grade resectable osteosarcoma

Background: We analyzed preoperative data from EURAMOS-1, a large study in resectable osteosarcoma where the regimen is MAP (Methotrexate/Adriamycin/Cisplatin), focusing on the effect on Histological Response (HRe) of reducing Methotrexate (MTX) by one dose. In the analysis of chemotherapy data, toxicity requires special care because it is a time-dependent confounder. Toxicity is at the same time a risk factor for HRe and a predictor for the exposure levels in the next cycle. Indeed, both drug doses and starting date of the next chemotherapy cycle are dynamically allocated to each patient depending on the toxicity levels through the end of the previous cycle.
Methods: We used Marginal Structural Models (MSMs) based on Inverse Probability-of-Treatment Weigthed (IPTW) estimators, which can estimate the causal effect of therapy modifications in presence of time-dependent confounders such as toxicity. MSMs create a pseudo-population by weighting each subject with the inverse of the probability of observing the allocation of a dose delay/reduction. In this pseudo-population toxicity history no longer predicts the next exposure, and MSMs mimic a randomized trial where the reduction of the exposure intensity is no longer confounded by the toxicity. Using the severity of toxic side-effects and both age and gender, we weighted the original population and then fitted a Marginal Structural Logistic Model for HRe with both cumulative reduction and cumulative delay as explanatory variables.
Results: Reducing the courses of MTX has a positive effect on HRe (OR 1.460, p-value 0.216); delaying a cycle by more than a week haa a negative effect (OR 0.564, p-value 0.045).
Conclusions: Although no statistical significance supports the positive effect of reducing MTX, in a clinical scenario where a choice must be taken between (i) giving 2 courses of MTX but delaying the cycle because of high toxicity; (ii) skipping one course of MTX and not delaying the end of the cycle; our findings suggest that the latter decision should probably be taken.