Most studies find that associations between mortality and either body mass
index or weight change are U-shaped. Previous studies, however, have not co
nsidered independent effects of weight level and weight change while also c
ontrolling for confounding by diseases leading to weight change. We used fo
llow-up data on average 10-year total mortality from a Danish population of
15,113 men and women, who had their weight measured at about 5 year interv
als. We obtained information on preexisting disease at surveys and by linka
ge to hospital discharge registers. We defined subclinical disease as incid
ent disease or death during the first 4 years of follow-up. We decomposed t
he mortality risk associated with weight change into a static effect, corre
sponding to the difference in mortality at stable weight at the initial and
attained weight, and a dynamic effect, estimated as the difference between
mortality after weight change and mortality at stable weight at the attain
ed weight. Both weight level and weight change had independent effects on t
otal mortality, with both these associations being U-shaped. Adjustment for
smoking and the exclusion of subjects with preexisting and subclinical dis
ease did not alter the associations.