This paper describes the mixed-integer nonlinear programming (MINLP) synthesis of reactor networks combined by a successive analysis in the attainable region (AR). The first prerequisite in the development of the combined approach was to develop a suitable superstructure of the reactor network which would be compact and simple enough to be used for simultaneous reactor network and overall process synthesis. The second prerequisite was to modify the theory of the attainable region (AR) so that economical criteria (gross profit, profit per unit of a product, etc.) are used rather than the conventional performance criteria (conversion, selectivity, etc.). Thus, the optimal MINLP results can now be directly verified in the modified AR and all the promising suggestions obtained by the analysis step can be brought back to modify the superstructure and perform another MINLP step until the analysis step suggests no improvement. The superstructure-base approach is thus enhanced by the rigorous AR technique. The nontrivial example illustrates how the results obtained by the MINLP approach can be further improved by the use of the combined approach. Furthermore: it has been shown that in special cases the recycle reactor (RR) may construct the border of the AR, which indicates the necessity to include the RRs in the superstructure.