RT Journal
A1 Rosenberg L
T1 MEta-analysis of alcohol and risk of breast cancer
JF JAMA
JO JAMA
YR 1989
FD January 20
VO 261
IS 3
SP 383
OP 383
DO 10.1001/jama.1989.03420030057027
UL http://dx.doi.org/10.1001/jama.1989.03420030057027
AB To the Editor.— 
    Dr Longnecker and colleagues1 have applied meta-analysis to the question of whether alcohol consumption influences the risk of breast cancer. It makes sense to use this technique for combining data across studies when the studies are randomized trials. Randomization tends to equalize potential confounding factors across the comparison groups; these factors might otherwise distort the results. If each study is valid but too small to give a precise estimate, then combining the results gives a more precise answer, and a valid one.Of course, one cannot randomly assign alcohol consumption, so that all studies of alcohol and breast cancer have been observational. Confounding factors were not equalized across the groups by randomization; instead, they were controlled. For each study, one must decide whether the observed results might be due to some distorting factor. In the case of alcohol and breast cancer, this is a particularly difficult