Licorice Extract Consumption and Salivary Testosterone Concentrations

Robert A Josephs, Jennifer S Guinn, Michelle L Harper, Frederick Askari

Licorice consumption has been shown to substantially reduce serum testosterone concentration. An explanation for this result was that the active component in licorice (glycyrrhizic acid) interfered with 17ß- hydroxysteroid dehydrogenase, which has been shown in vitro to catalyse the conversion of androstenedione to testosterone. We twice attempted to replicate this effect of licorice but could not. We identified differences between our methods and those of the previous study and possible statistical anomalies (including inappropriate use of statistical tests) in the earlier report. Lancet 2001; 358: 1613-14

Armanini and colleagues (1) reported that serum testosterone concentration was reduced by 35% after 4 days of licorice consumption in seven men. These findings have important implications for licorice consumers and those who study testosterone. We were interested in these effects of licorice and attempted and failed to find an effect of licorice on spatial cognition (measures of spatial cognition and changes in testosterone concentration are correlated). (2) Our failure led us to attempt a replication of Armanini and colleagues' results. We followed Armanini and colleagues' methods. (1) We gave licorice to men daily for 4 days and measured their testosterone concentrations in saliva. 20 college-aged men were given 5·6 g of licorice (Baschetti Licorice Extract) daily for 4 days. This preparation contained the same amount of glycyrrhizic acid used by Armani and colleagues (0·5 g). Potential participants were excluded if they indicated current or former use of illicit or controlled substances or had a history of hypertension. Testosterone concentration was measured by RIA before the study and on the day after the study ended. We noted a statistically insignificant decrease in testosterone concentration of 9·5% (table).

Hormone concentrations in college-aged people before and after licorice consumption

Day 0 Day 4 Effect size Power Study Armanini and colleagues
Serum testosterone 25·7 (7·5) 14·4 (1·5)† 2·51 >0·995 (mmol/L)
Our study (1) Saliva testosterone 0·59 (0·30) 0·53 (0·24) 0·20 0·15 (mmol/L)
Our study (2) Male saliva 0·57 (0·19) 0·53 (0·19) 0·21 0·12 testosterone (mmol/L) Female saliva 0·27 (0·13) 0·24 (0·18) 0·22 0·13 testosterone (mmol/L) Male saliva cortisol 60·7 (17·9) 91·0 (21·5)* 1·35 0·90 (mmol/L) Female saliva cortisol 49·7 (17·9) 69·0 (21·5)† 0·99 0·70 (mmol/L) Data are mean (SD). *p=0·001 for day 0 comparison. †p=0·05 for day 0 comparison.

We repeated the study. We changed licorice sources (Frutarom; Haifa, Israel), and recruited a new group of participants (11 college-aged men and ten college-aged women), and obtained independent validation of glycyrrhizic acid content (Independent Laboratories; Denver, CO, USA). A further test of the licorice potency and validity of the experimental procedure was provided by measuring changes in cortisol concentrations in response to licorice. Cortisol concentrations increase in response to licorice consumption because of inhibition of 11ß-hydroxysteroid dehydrogenase, which catalyses the conversion of cortisol to cortisone. (3) Concentrations of testosterone and cortisol in saliva were measured before the study and for the next 4 days (only initial and final values are presented, table). Testosterone concentration did not significantly decrease. We compared the results of our two studies by 2 test. The effects of licorice on testosterone concentration were the same in both studies (table). We noted statistically significant increases in cortisol concentrations (table).

We did a meta-analysis to combine and compare effect sizes: we combined the effect of licorice on testosterone across our two studies, and compared the size of this effect with that reported by Armanini and colleagues. The ten-fold difference in effect size between our results and those of Armanini and colleagues was statistically significant (p=0·001). We did not replicate the results of Armanini and colleagues despite substantially higher statistical power in our studies. The licorice source (Saila) used by Armanini and colleagues was different to ours. However, three things suggest that this difference is unimportant. First, in our second study, we replicated the well-known effect of licorice on cortisol, and thus we validated the potency of the licorice and the effectiveness of our experimental procedure. Second, verification of glycyrrhizic acid content confirmed that our participants ingested the same quantity of metabolically active component as the participants in Armanini and colleagues' study. Third, the decrease in testosterone concentration was the same in both our investigations, despite differences in licorice products and sample composition. Armanini and colleagues measured changes in serum testosterone concentrations, whereas we used concentrations in saliva. Although serum concentration is much higher than saliva concentration, changes in testosterone concentration in saliva are equivalent to changes in serum concentration. Saliva and serum testosterone concentrations are very strongly correlated. (4) Therefore, this procedural difference does not explain the large discrepancy between our results and those of Armanini and colleagues. Without examination of Armanini and colleagues' raw data, we can only speculate as to why our results are so different from theirs. However, Armanini and colleagues' statistical analysis could provide an explanation for their effect. They reported a drop in SD of 80% from the start to the end of their study (table). This reduction in error variance violates assumptions that underly the statistical tests that they used and renders them inappropriate. Furthermore, Armanini and colleagues' large pre-administration error variance indicates that at least one of their participants probably began the study with a testosterone concentration that was either substantially greater or lower than the mean. The effect sizes suggest that this participant had a higher concentration of testosterone than the mean. If this assumption is correct, then the remaining participants began the study with testosterone concentrations that were below the mean. For these participants, the actual licorice-induced decrease in testosterone concentration was more modest than the mean. Therefore, removal of statistical outliers (standard procedure is to remove participants whose testosterone concentrations are greater than 2 SD than the mean) would lower Armanini and colleagues' effect to be closer to ours. Armanini and colleagues concluded that men with low libido or hypertension should be questioned about licorice ingestion. We confirm that licorice increases cortisol, suggesting a contraindication for its use in people with hypertension. Because we detected only a small reduction in testosterone, we cannot confirm that people with low libido should avoid licorice consumption, nor can we strongly endorse licorice for use in behavioural or medical research on testosterone.

1. Armanini D, Bonanni G, Palermo M. Reduction of serum testosterone in men by liquorice. N Engl J Med 1999; 341: 1158.

2. Kimura D, Hampson E. Cognitive pattern in men and women is influenced by fluctuations in sex hormones. Curr Dir Psych Sci 1994; 3: 57-61. [PubMed]

3. Heilmann P, Heide J, Hundertmark S, Schoneshofer M. Administration of glycyrrhetinic acid: significant correlation between serum levels and the cortisol/cortisone-ratio in serum and urine. Exp Clin Endocrinol Diabetes 1999; 107: 370-78. [PubMed]

4. Vittek J, L' Hommedieu D, Gordon G, Rappaport S, Southren A. Direct radioimmunoassay (RIA) of salivary testosterone: correlations with free and total serum testosterone. Life Sci 1985; 37: 711-16. [PubMed]

Department of Psychology, University of Texas at Austin, TX, 78712, USA (R A Josephs PhD, J S Guinn BS, M L Harper MS); and Department of Internal Medicine, University of Michigan, Ann Arbor, MI (F Askari PhD)