Two seminal reviews (IARC, 2002; CDHS, 2002) of possible health effects from power-frequency EMFs reached partly different conclusions from similar epidemiological evidence. These differences have an impact on precautionary policy. We examine the statistical aggregation of results from individual disparate studies. Without consistent exposure metrics, the advantage of meta-analysis to estimate magnitude of effect is lost. However, counting positive and statistically significant results yields important information. This is not a substitute for meta-analysis, but a fall-back when meaningful meta-analysis is not available. Representative results from 33 independent adult leukemia studies tabled by IARC yielded 23.5 positives (p approximately 0.01) and 9 significant-positives (p<10(-7)). From 43 representative results from CDHS, there were 32 positive (p<0.001) and 14 significant-positives (p<10(-12)). There were no significant-negative results in either list. Results for adult brain cancer gave a similar, but less clear, message. Childhood leukemia EMF studies have been sufficiently comparable to allow selective pooled analysis, which was important in classifying carcinogenicity. Aggregating all the studies suggests that results for childhood leukemia are not stronger, numerically, than those for adult leukemia. CDHS did not note the number of significant-positives, but noted the meta-analytic summary and the number of positives, forming a view about the strength of these findings. IARC shows no evidence of considering the aggregation of results other than subjectively. It considered individual studies but this led to a tendency to fragment and dismiss evidence that is intrinsically highly significant. We make recommendations for future reviews.