One of the main goals of an individual participant data meta-analysis (IPD-MA) of intervention studies is to investigate whether treatment effect differences are present, and how they are associated with patient characteristics. Examining treatment heterogeneity due to a continuous covariable (e.g., BMI or age) may be challenging, since there is often no prior knowledge on functional form of the conditional association between the outcome and the continuous variable. Modelling treatment effect differences whilst accounting for non-linear functional shapes may provide the opportunity to accurately make inferences whether a patient should be treated or not. To account for non-linearities the authors may estimate the functional shape of the associations and investigate potential treatment effect differences. So far, a variety of methods that account for non-linear functional shapes has been proposed. In this manuscript, the authors focus on the use of splines since they can capture both non-linear main effects and non-linear treatment-covariable interaction effects without the need to pre-specify their functional form. The goal is to explain and illustrate how to predict a conditional absolute treatment effect, as this measure is most relevant for clinical decision-making. The authors describe the various spline approaches and their application in IPD-MA using pointwise meta-analysis, multivariate meta-analysis, and generalised additive mixed effects models, and the authors provide the corresponding R-code. They also describe the results of the aforementioned spline and pooling methods using an empirical individual participant data-set, investigating the effect of antibiotics in children with acute otitis media (AOM).