OBJECTIVE: For cost-benefit analysis, health technologies with multiple effects should theoretically be valued in a single scenario by a holistic willingness-to-pay (WTP) measure. Recently, decomposed scenarios in which respondents state their WTP for each individual effect were used more often. Evidence can be found that the sum of such decomposed WTPs overestimates the holistic WTP, i.e. the holistic WTP is subadditiv. No theoretically founded and empirically validated aggregation function for decomposed WTPs exists. Our aims were to derive such an aggregation function and to test it empirically. METHODS: Within an expected utility model, we identified risk aversion with respect to money and a positive correlation of better health and marginal utility of money as reasons for subadditivity. Under mutual utility independence of health and money, we derived a theoretically founded aggregation function covering these two impacts. For an empirical application of this function, one has to determine the utilities of the health states, of well-defined states “rich&ill” and “poor&healthy”, and a utility function over money - beside the decomposed WTPs. In a contingent valuation study, we asked 92 patients with diabetes to state their WTP for reductions in the risk of several diabetic complications in decomposed and holistic scenarios. Utilities were measured by standard gamble procedures. 46 patients (group 1) valued two complications (diabetic retinopathy & neuropathy). The remaining 46 patients (group 2) valued also a third complication (diabetic foot). RESULTS: Subadditivity occurred in both groups, but was not statistically significant. The sum of decomposed WTPs overestimated holistic WTP by 8.3% in-group 1 and 16% in-group 2. Theoretically founded aggregation underestimated holistic WTP by 2% in-group 1 and 1.2% in-group 2. CONCLUSIONS: The results of our pilot study suggest that the theoretically founded aggregation function might overcome problems of subadditivity, which can be economically relevant. Further empirical testing of the approach is indicated.