Incremental costs and effectiveness may be correlated if patients receiving the most health benefit from an intervention tend to be those who incur greater (or lesser) incremental cost. Such correlation has the potential to bias the synthesis of cost-effectiveness evidence, yet is infrequently reported. The objective of this study was to assess the sensitivity of an incremental cost-effectiveness ratio (ICER) for interventions A vs C, estimated indirectly from studies of A vs B and B vs C, to correlations between incremental cost and effect. Incremental costs and effects of interventions A vs B and B vs C were simulated from lognormal and Normal distributions respectively of given mean, standard deviation (SD) and cost-effectiveness correlation. The distribution of the ICER for interventions A vs C was calculated. Results from various parameter choices were compared. One parameter set choice was that the mean (SD) incremental cost of A vs B was $100,000 ($40,000) for a health benefit of ten(2) and that the incremental cost of B vs C was $200,000 ($50,000) for a health benefit of 12(2). For the parameter set described, the mean (SD) ICER of intervention A vs C assuming no correlation between incremental costs and health effects was $13,760 ($2637). Varying the two correlation parameters between -1 and +1 altered the mean ICER by up to $200 and the ICER's SD by up to $1679. The ICER's mean was fairly insensitive to the two correlation parameters but its SD was more sensitive. Correlations between cost and effectiveness can therefore be expected to be more important in questions of decision uncertainty rather than in the estimation with best existing knowledge of relative cost-effectiveness. More generally, sensitivity was greater when distributional variance and skewness of incremental cost and health benefit were greater.