Golomb's paper [1] on the relation between cholesterol and violence cites our work [2] in a sentence stating "Several studies in humans and nonhuman primates ... suggest a specific connection between low or lowered fats or cholesterol levels and low or lowered serotonin activity." This sentence overstates our findings. Our primary and most statistically powerful analyses noted nonsignificant correlations between cholesterol levels and cerebrospinal fluid 5-hydroxyindolacetic acid (5-HIAA) (the serotonin metabolite) in our samples (n = 84 and 70 in different overlapping samples); correlation coefficients were 0.13 and 0.15, suggesting shared variance less than 3%.

Golomb's Table 7 states that our low-cholesterol participants showed "decreased cerebrospinal fluid metabolite [5-HIAA] (19% reduction)." This statement was derived from secondary subgroup analyses comparing the participants with the highest and lowest cholesterol levels (n = 10 to 14 per group in two different analyses based on different techniques for determining cholesterol levels). We found differences of less than 5 ng/mL between the mean 5-HIAA levels of each group. The within-group SDs ranged from 7 to 15 ng/mL. A calculation of effect sizes from these analyses yields estimates of 0.30 and 0.37, levels that should be considered subtle. Effect-size estimates derived from small samples are subject to outliers. A calculation of the differences between the medians (rather than the means) showed group differences of 0.8 and 5.5 ng/mL. Furthermore, Mann-Whitney tests on between-group comparisons failed to find even trend levels of significance.

In sum, the 5-HIAA levels in our participants with the highest and lowest cholesterol levels were similar (as we previously noted in our paper). The statement in Table 7 should be qualified because this conclusion does not account for a high degree of within-group variance in small subsamples.