Sixty-nine female student volunteers were recruited from University College London. Participants were aged between 18 and 25 years, and were screened by structured interview to ensure that they were healthy, not taking any medication and had no previous history of any relevant physical or mental illness. They were instructed not to take antibiotics, ibuprofen or aspirin for 10 days prior to the study and to avoid caffeine, alcohol and excessive exercise during the 12 h prior to testing. On the day of the study, participants were advised to eat a low fat breakfast and were provided with a standardised low fat lunch consisting of vegetable couscous, fruit salad and fresh orange juice. Two participants were excluded from the final analyses since one was extremely obese (BMI 43.3) and another was greatly underweight (BMI 15.6). The study was approved by the joint University College London/University College London Hospital Committee on the Ethics of Human Research and all participants gave their informed consent. We chose to study a healthy population with a wide range of adiposity levels rather than specifically comparing obese versus non-obese individuals, since obese individuals are more likely to have co-morbid conditions such as diabetes or hypertension that may have influenced our measures.

Anthropometric measures were taken by a research nurse at the beginning of the session. Height was measured to the nearest 0.1 cm using the Frankfort plane to standardise the measurement. Body weight was measured to the nearest 0.1 kg. Body mass index (BMI) was calculated as body weight in kilograms divided by height in metres squared. Waist circumference was measured midway between the lowest rib and ileac crest. Body fat mass was estimated using a Bodystat 1500 bioelectrical impedance body composition analyses device (Bodystat, Douglas, Isle of Man). Fat percentage was calculated as fat weight divided by total (fat + lean) body weight.

Cardiovascular data for the study were collected continuously and then averaged over specified 5 min trials. Blood pressure was monitored from the finger using a Portapres 2, a portable version of the Finpres device (TNO-TPD Biomedical Instrumentation, Amsterdam, The Netherlands). Heart rate and heart rate variability (HRV) were assessed by impedance cardiography (ICG; VU-AMS, Amsterdam, The Netherlands) as described previously (^{Willemsen et al., 1996}). HRV was calculated as the root mean square of successive R-R interval differences (RMSSD). A reduction in RMSSD indicates a shift in cardiac sympathovagal balance towards sympathetic control over the rhythm of the heart (^{Malik, 1996}). Cardiac pre-ejection period (PEP) was measured as an index of cardiac sympathetic drive (^{Sherwood et al., 1990}). One minute ensemble averages were derived for the ICG for each minute of tasks and averaged. PEP was defined as the interval between R-wave and B-point plus a fixed Q–R interval of 48 ms. Cortisol was measured in saliva samples obtained throughout the session. Saliva was collected using Salivettes (Sarstedt, Inc. Leicester, UK) and stored at −80 °C prior to analyses. Salivary cortisol was analyzed using a commercially available time-resolved immunoassay with chemiluminescence detection (CLIA; IBL-Hamburg, Hamburg, Germany), at the Technical University of Dresden, Germany. This assay had a detection limit of 0.16 ng/ml and intra-and inter-assay coefficients of variation of <10 and <12% respectively.

For assessment of circulating cytokines, blood samples (10 ml) were drawn using a 21-gauge butterfly needle into Vacutainer tubes containing EDTA as an anti-coagulant, then centrifuged immediately at 1250 × g for 10 min at room temperature. The plasma layer was removed, aliquoted and stored at −80 °C until analyses. Plasma IL-6 levels were measured using a high-sensitivity two-site enzyme-linked immunosorbent assay (ELISA) from R & D Systems (Oxford, UK). This assay had a detection limit of 0.09 pg/ml with intra- and inter-assay coefficients of variation (CVs) of 5.3 and 9.2% respectively. Plasma IL-1Ra concentrations were measured using a commercial ELISA from R & D Systems, with a detection limit of 15 pg/ml and intra-and inter-assay CVs of <10%. Plasma leptin levels were assessed using a commercial ELISA from Ray Biotech at Insight Biotechnology Ltd. (Middlesex, UK). This assay had a detection limit of 6 pg/ml and intra-and interassay CVs of <10% and <12% respectively. Plasma samples were diluted 1:200 prior to leptin analyses.

Mental stress was induced by two 5-min behavioural tasks, administered under time pressure. The first was a computerized Stroop task, involving the successive presentation of target color words (e.g. red) printed in a different color (e.g. green). The task was to press a computer key that corresponded to the position at the bottom of the screen of the name of the color in which the target word was printed. The second task was presented after an interval of 5 min and involved simulated public speaking. Participants were presented with a hypothetical scenario in which they had been wrongly accused of shoplifting, and were instructed to give a speech in their defense by addressing the camera directly in front of them. They were told that their speech would be video recorded and later judged for efficacy and fluency. The experimenter remained in the room with the participant during each of the tasks, and instructed them when to start and stop.

All sessions were run in the afternoon in a light and temperature-controlled laboratory, and participants were tested individually. They were provided with water but were not allowed to eat during testing. Anthropometric measures were obtained and the impedance cardiogram was fitted for continuous assessment of heart rate, HRV and cardiac PEP. Participants were then seated comfortably and fitted with finger cuffs so that blood pressure (BP) could be monitored using the Portapres-2. A venous cannula was inserted into the lower arm for blood sampling, and they were left to rest for 30 min. Cardiovascular measures (BP, heart rate, HRV and cardiac PEP) were recorded for the last 5 min of the baseline rest period, then a saliva sample was obtained and a baseline blood sample was drawn. At this time, participants were asked to rate their subjective feelings of stress on a 7-point Likert scale ranging from 1 = low to 7 = high. Next, the two tasks were administered in a fixed order. Five-minute recordings of cardiovascular activity were made during each of the tasks and cortisol samples and subjective stress ratings were obtained after each task. Following the tasks, a second blood sample was taken and participants rested for a further 45 min. Additional recordings of cardiovascular activity were obtained at 10–15, 25–30 and 40–45 min post-tasks, and a third blood sample was drawn at 45 min. Further measures of subjective stress and cortisol were obtained at 15, 30 and 45 min post-task.

First, we tested whether the tasks used in our study were indeed stressful, by assessing participants’ subjective and physiological responses to these tasks using repeated measures analyses of variance (ANOVA). The repeated measures analyses of subjective stress and salivary cortisol involved six trials (baseline, post-Stroop, post-speech, 15, 30 and 45 min post-tasks). Similarly, analyses of heart rate, HRV, cardiac PEP and blood pressure involved six trials (baseline, Stroop, speech, 10–15, 25–30 and 40–45 min post-tasks), while the analyses of plasma cytokines involved three trials (baseline, immediately post-tasks and 45 min post-task). The Greenhouse-Geisser correction of degrees of freedom was applied when sphericity assumptions were violated. The distribution of plasma leptin was skewed so data were square root transformed before analyses. Heart rate variability (RMSSD) data were also skewed and were log transformed, however raw values are presented for comparability with other studies. Post hoc tests were conducted using Tukey's least significant difference (LSD) test.

We then investigated the association between adiposity measures and physiological stress responses using a multiple linear regression approach. Three aspects of physiological response were analyzed: baseline levels, stress reactivity (computed as the change in levels between baseline and stress) and stress recovery (computed as the change between baseline and 45 min post-task). Age, smoking status and ethnicity were included in all models as covariates. In regressions involving reactivity and recovery measures, baseline levels of the relevant dependent measure (HR, HRV, etc.) were included as additional covariates in the model. The same regression approach was used to assess the relationship between basal plasma leptin levels and physiological stress responses. This time adiposity measures (BMI, waist circumference and percentage body fat) were also included as covariates, since adiposity is a major determinant of leptin levels and is associated with stress reactivity. Results are presented as standardised regression coefficients (β) with standard errors (s.e.). Associations between adiposity (or leptin) measures and reactivity are illustrated by displaying the mean cardiovascular and inflammatory responses of individuals in the lower and higher tertiles of waist circumference or plasma leptin, adjusted for covariates.

Sample characteristics at baseline are presented in Table 1. Participants were relatively young with a mean age of 21. The majority were White non-smokers. All were normotensive and had glycated haemoglobin (HbA1c) levels in the normal range. Although they were not overweight on average, there were large individual differences in adiposity measures; BMI ranged from 18.4 to 34.0 kg m⁻², waist circumference ranged from 57.0 to 95.0 cm and percentage body fat ranged from 10.1 to 40.5%. Participants also varied widely in basal plasma leptin levels, ranging from 5.7 to 105.5 ng/ml, with a mean concentration of 35.7 ng/ml, s.d. 22.0. These levels are in the expected physiological range for non-fasting women (^{Rosenbaum et al., 1996}).

Participants rated tasks as stressful with a mean score of 4.64 ± 1.1 and 4.42 ± 1.4 during the Stroop and speech task respectively. Subjective stress ratings returned to baseline levels during recovery, falling below baseline at 45 min post-task (F (3, 196) = 174.4, p < 0.001). Participants’ blood pressure and heart rate increased significantly during the tasks, with mean increases of 13.3 mmHg (F (3, 167) = 38.2, p < 0.001) and 10.3 mmHg (F (4, 221) = 44.5, p < 0.001) in systolic BP and diastolic BP respectively, and an average rise of 11.5 bpm in heart rate (F (2, 149) = 116.3, p < 0.001). At the same time, tasks induced a substantial decrease in participants’ HRV and cardiac PEP; HRV decreased by 18.6 ms on average during tasks (F (2, 149) = 42.4, p < 0.001), and cardiac PEP decreased by 6.0 ms on average (F (2, 131) = 44.7, p < 0.001). There were large individual differences in all cardiovascular responses. For example, changes in heart rate ranged from −1.8 bpm to +30.3 bpm, changes in HRV ranged from −99.4 ms to +4.8 ms, and changes in cardiac PEP ranged from −28.0 ms to +13.0 ms, during tasks. In addition, there was a small but significant (14.5%) rise in salivary cortisol in response to tasks, returning to baseline levels during the rest period (F (2, 122) = 9.96, p < 0.001).

Tasks induced a small but significant increase in participants’ plasma levels of IL-6 (F (1, 78) = 21.1, p < 0.001) and leptin (F (2, 105) = 26.3, p < 0.001), with maximum levels detected at 45 min post-tasks. Plasma levels of IL-6 increased by 37% on average at 45 min, with responses ranging from −0.45 pg/ml to +1.72 pg/ml. Similarly, plasma leptin levels increased by 14% on average at 45 min, with responses ranging from −6.7 ng/ml to +20.2 ng/ml. IL-1Ra levels were not altered by stress in the sample as a whole. However, there were large individual differences in this response, with changes in plasma IL-1Ra ranging from −148.1 pg/ml to +124.2 pg/ml at 45 min post-task. Plasma concentrations of IL-1Ra and leptin were positively correlated immediately post-stress (r = 0.30, p = 0.023) and at 45 min post-stress (r = 0.32, p = 0.020).

There was a significant relationship between adiposity and diastolic BP. Baseline diastolic BP was associated with BMI, waist circumference and percentage body fat (β = 0.29–0.37, s.e. = 0.13–0.14, all p < 0.05), independent of age, ethnicity and smoking status. Similarly, stress-induced increases in diastolic BP at 45 min were related to percentage body fat (β = 0.33, s.e. = 0.13, p = 0.014) and waist (β = 0.26, s.e. = 0.14, p = 0.059 ns trend), independent of age, ethnicity, smoking and baseline diastolic BP. There were no associations between adiposity and any other cardiovascular or neuroendocrine measures.

Basal plasma cytokine levels were related to adiposity, independent of age, ethnicity and smoking. Baseline leptin was associated with BMI, waist and percentage body fat (β = 0.49–0.62, s.e. = 0.11–0.13, all p < 0.001). Similarly, IL-6 levels were related to BMI and percentage body fat (β = 0.28–0.29, s.e. = 0.13–0.14, p < 0.05), and basal plasma IL-1Ra was associated with waist (β = 0.41, s.e. = 0.13, p = 0.003). There was also a significant relationship between central adiposity and cytokine stress responses. Waist circumference was associated with individual differences in stress-induced increases in plasma leptin immediately post-task (β = 0.34, s.e. = 0.15, p = 0.022) and stress-induced increases in plasma IL-1Ra at 45 min post-task (β = 0.32, s.e. = 0.15, p = 0.040), independent of age, ethnicity, smoking and baseline levels of the respective cytokine. The relationship between waist and IL-1Ra responses is illustrated in Fig. 1, showing mean plasma IL-1Ra concentrations in individuals in the lowest and highest tertiles of waist circumference. People with greater central adiposity had 15% higher IL-1Ra levels at 45 min post-task.

Plasma levels of leptin at baseline were significantly related to participants’ cardiovascular and inflammatory stress responses, independent of age, smoking, adiposity and baseline measures of the respective dependent variable. Specifically, basal leptin was associated with stress-induced increases in heart rate (β = 0.53, s.e. = 0.18, p = 0.006) and with decreases in HRV (β = −0.44, s.e. = 0.18, p = 0.015) and cardiac PEP (β = −0.51, s.e. = 0.17, p = 0.004) during tasks, independent of covariates. There was also a positive association between basal leptin levels and stress-induced increases in IL-6 at 45 min (β = 0.35, s.e. = 0.17, p = 0.042). These effects are illustrated in Fig. 2 (A–D). Women in the highest versus lowest tertile of basal leptin had a 2.2-fold greater IL-6 response to stress at 45 min. Similarly, during tasks, women in the highest leptin tertile had 187% greater increases in heart rate, 86% greater reductions in HRV and >11-fold larger decreases in cardiac PEP. All associations remained significant when controlling for ethnicity (p < 0.05). There was no relationship between basal leptin and measures of diastolic or systolic BP, salivary cortisol or plasma IL-1Ra.