OBJECTIVES: Probabilistic sensitivity analysis (PSA) has become mandatory in regular HTA but is often infeasible in early HTA when evidence allowing the definition of parameter distributions is lacking. Our objective is to compare methods for representing uncertainty in early HTA, when only minimum and maximum values are known for all model parameters. METHODS: We investigated 5 methods for estimating outcomes similar to standard PSA outcomes on the incremental cost-effectiveness (ICE) plane. Methods were based on i) uniform/normal distributions per parameter; ii) uniform/normal interpolation, per parameter, on the ICE plane; iii) averaging of points on the ICE plane; iv) minimum distance to points on the ICE plane; and v) smoothed splines fitted to points on the ICE plane. We simulated four scenarios, with true ICE results increasingly deviating from (bivariate) normality, and applied a regular HTA case study on the cost-effectiveness of point-of-care troponin testing to exclude acute coronary syndrome. Method’s inaccuracy was expressed as average absolute difference (in %) between the true CEAC and predicted CEAC, across a willingness-to-pay range of €0-100,000. RESULTS: The best performing methods were averaging of points on the ICE plane (best in 3 scenarios, expected CEAC error across all scenarios: 5.3%) and normal interpolation, per parameter, on the ICE plane (best in 1 scenario where the true uncertainty was crescent-shaped, expected CEAC error across all scenarios: 5.4%). In the troponin case study these methods also performed well, with expected error of only 0.3% (averaging) and 1.4% (normal interpolation per parameter on ICE plane). CONCLUSIONS: Reflecting uncertainty in early HTA may be desirable to indicate the volatile nature of the outcomes. When only minimum and maximum values for parameters are known, the method averaging the corresponding incremental costs and effects is a robust way to represent uncertainty in outcomes, and allows the construction of accurate CEACs.