#728 Predicting individual patient response to corticosteroids in IgA nephropathy: a secondary analysis from the TESTING cohort

Résumé fourni par la source

Abstract Background and Aims Corticosteroids are an effective treatment for IgA nephropathy but are associated with considerable adverse events. The TESTING clinical trial showed an average treatment effect of a 47% relative risk reduction in the primary composite outcome for methylprednisolone versus placebo (hazard ratio 0.53, 95% CI 0.39–0.72). This is an aggregate result that cannot be applied to individual patients to make personalized treatment decisions, making it challenging to identify appropriate patients for corticosteroid treatment. To address this problem, we conducted a secondary analysis of the TESTING cohort to generate a model that can predict, for an individual patient, the probability that they will respond to methylprednisolone resulting in a lower risk of kidney disease progression. Method Time to the primary outcome (40% reduction in eGFR, kidney failure or death due to kidney disease) was first evaluated in a Cox proportional hazards model in which all potential treatment effect modifiers including demographic, clinical and MEST-C variables were evaluated as main effects using backwards elimination. The selected variables were then forced into a multivariable model along with treatment exposure and interaction terms between treatment and each other variable. This model was used to generate the predicted 4-year absolute risk of the primary outcome for each patient under separate counterfactual scenarios of being treated with methylprednisolone or placebo. The difference in risk between the two scenarios was the predicted individual treatment effect on absolute risk reduction (ARR). Model performance was assessed using discrimination plots, restricted mean survival time (RMST, an estimate of the additional time methylprednisolone provides without experiencing the primary outcome) and the C-statistic for benefit (ability of the model to discriminate between patients who got more versus less benefit from methylprednisolone). Results A total of 483 patients were included (median age 36 years, proteinuria 2.0 g/day, eGFR 57 mL/min). During 43 (median) months of follow-up, 176 participants experienced the primary outcome. Compared to the average ARR associated with methylprednisolone (16.1%, 95% CI 15.5–16.8), the predicted individual-level ARR was highly variable ranging from zero (for patients who experience minimal or no benefit) to more than 30% (for patients who experience considerable benefit) (Fig. 1, left panel). Patients with predicted ARR >10% had a substantially greater observed benefit from methylprednisolone (ARR 24%) compared to those with predicted ARR ≤10% (ARR −5%) (Fig. 1, right panel). A policy of treating patients with higher predicted benefit (ARR >10%) and not treating patients with low predicted benefit (ARR ≤10%) had a longer RMST than using random treatment allocation as was done in the main trial (1,194 v 1,028 days). The C-statistic for benefit was 0.63 (95% CI 0.56–0.70). Calibration plots showed considerable agreement between predicted and observed ARR. Findings were consistent in both the high-dose and reduced-dose methylprednisolone cohorts. The pattern of treatment effect modifiers was similar when the outcome was changed to annualized eGFR slope. Conclusion We have generated a model that can predict individual patient response to methylprednisolone and inform personalized treatment decisions in IgAN so that corticosteroid therapy can be targeted to those most likely to benefit.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.