Background. Many studies have noted a strong association between poor social support and premature mortality. A limitation of such studies has been their failure to control adequately for confounders that damage both social supports and physical health.

Methods. A 50-year prospective multivariate study of 223 men was used to examine the possible causal relationships between social supports and health. Alcohol abuse, prior physical health and mental illness prior to age 50 were controlled. Relative social supports were quantified over the period from age 50 to 70.

Results. Adequacy of social supports from age 50 to 70 was powerfully correlated with physical health at age 70 (P<0·001). However, such social supports were also powerfully predicted by alcohol abuse (P<0·001), smoking (P<0·001) and indicators of major depressive disorder (P<0·01) assessed at age 50. When prior smoking, depression and alcohol abuse were controlled, then the association of physical health with social supports was very much attenuated. Some facets of social support like religion and confidantes were unassociated with health even at a univariate level. Surprisingly, in this sample friends seemed more important for sustained physical health than closeness to spouse and to children.

Conclusions. While social supports undoubtedly play a significant role in maintaining physical well-being in late life, much of the association between poor social supports and mortality may be mediated by alcoholism, smoking and pre-morbid psychopathology.