The relative fitness of viral variants has previously been defined as the s
lope of the logarithmic ratio of the genotype or phenotype frequencies in t
ime plots of pairwise competition experiments. Developing mathematical mode
ls for such experiments by employing the conventional coefficient of select
ions, we demonstrate that this logarithmic ratio gives the fitness differen
ce, rather than the relative fitness. This fitness difference remains propo
rtional to the actual replication rate realized in the particular experimen
tal setup and hence cannot be extrapolated to other situations. Conversely,
the conventional relative fitness (1 + s) should be more generic. We devel
op an approach to compute the generic relative fitness in conventional comp
etition experiments. This involves an estimation of the total viral replica
tion during the experiment and requires an estimate of the average lifetime
of productively infected cells. The novel approach is illustrated by estim
ating the relative fitness, i.e., the relative replication rate, of a set o
f zidovudine-resistant human immunodeficiency virus type I variants. A tool
for calculating the relative fitness from observed changes in viral load a
nd genotype (or phenotype) frequencies is publically available on the websi
te at http://www-binf.bio.uu.nl/(similar to)rdb/fitness.html.