Objectives: This study explores how mathematical models can be used to
examine the relationship between the effectiveness and costs of diffe
rent prevention strategies. The analysis, based on a model designed to
simulate the spread of HIV in sub-Saharan Africa, illustrates how the
impact and relative cost-effectiveness of a prevention programme are
sensitive to a number of factors including the rate of spread of HIV p
rior to the introduction of the programme, the measure used to evaluat
e the impact of the programme, and the point when the programme is int
roduced in the epidemic. Results: Assessing the impact and cost-effect
iveness of different HIV prevention strategies is problematic. Not onl
y are there difficulties in collecting the relevant data, but the impa
ct of a prevention programme may be highly non-linear in character bec
ause of the many factors that determine the net rate of viral transmis
sion. The long incubation period of HIV further complicates the analys
is as the benefits from preventing a case of infection extend over a n
umber of years. Conclusions: The results highlight the need to collect
more data on the impact and costs of prevention programmes if we are
to ensure that the available resources for HIV prevention are to be us
ed both efficiently and equitably.