A meta-analysis of salivary cortisol responses in the Trier Social Stress Test to evaluate the effects of speech topics, sex, and sample size

BACKGROUND: The Trier Social Stress Test (TSST) is one of the most widely used laboratory-based psychological stress paradigms. Previous studies have shown that males have a more robust cortisol response than females in the TSST. However, the effects of sample size, speech topic, and interaction between sex and speech topic on cortisol responses in TSST remain elusive. Our goal was to evaluate these influencing factors in the TSST using salivary cortisol reactivity as an objective measure. METHODS: We collected TSST research articles in Web of Science, PubMed, PsycNet, and CNKI. We only included TSST studies that had measures of salivary cortisol both before and after task completion. A total of 65 articles involving 76 sub-studies met our inclusion criteria, with a total of 5171 participants (2040 females and 3131 males). The effects of sample sizes were assessed to determine if results of studies with various sample sizes were stable. We performed multivariate meta-regression to determine the effects of speech topic, sex, and the interaction between sex and speech topic after controlling their confounding effects. Subgroup analysis of sex was conducted to detect inter-group differences. We further evaluated the baseline and peak salivary cortisol concentrations for males and females independently to detect the sources of sex differences. RESULTS: The average effect size (i.e., Cohen's d) of salivary cortisol reactivity was 0.93, 95% CI = 0.82 to 1.04, p < .001. The small studies produced larger variations in the reported effect sizes than the large-sample studies (r = -0.24, p = .041). A sample size of 40 was necessary to provide sufficient statistical power to detect significant changes of salivary cortisol in TSST. Speech topics, sex, and sex-speech topic interaction could predict salivary cortisol responses (F(df1 = 3, df2 = 72) = 11.98, p < .001) and explained 42.68% of the total experimental variation. Sex was the only significant contributing factor (p < .00025) in the regression model. Salivary cortisol responses in males were significantly higher than in females (Q(B) = 42.89, df = 1, p < .001). Further, significant differences between males and females were detected at baseline (t = -2.03, df = 74, p = .046) and peak (t = -4.96, df = 74, p < .001). CONCLUSIONS: The TSST effectively induces stress response as measured by salivary cortisol change. Forty samples is the minimum sample size for detecting the robust salivary cortisol responses. We confirmed that males have more robust salivary cortisol reactivity than females in TSST. Speech topics that we tested did not significantly contribute to differences in salivary cortisol responses. No significant interaction between sex and speech topic on salivary cortisol responses was detected.