Psychoneuroendocrinology (2014) 50, 227—245
Available online at www.sciencedirect.com
ScienceDirect journal homepage: www.elsevier.com/locate/psyneuen
Altered circadian patterns of salivary cortisol in low-functioning children and adolescents with autism Sylvie Tordjman a,b,∗, George M. Anderson c, Solenn Kermarrec a,b, Olivier Bonnot d, Marie-Maude Geoffray e, Sylvie Brailly-Tabard f,g, Amel Chaouch g, Isabelle Colliot g, Severine Trabado f,g, Guillaume Bronsard h,i, Nathalie Coulon b, ¸oise Camus l, Michel Botbol j, Henriette Charbuy k, Franc Yvan Touitou l a
Pôle Hospitalo-Universitaire de Psychiatrie de l’Enfant et de l’Adolescent de Rennes (PHUPEA), CHGR et Université de Rennes 1, Rennes, France b Laboratoire Psychologie de la Perception, Université Paris Descartes, CNRS UMR 8158, Paris, France c Child Study Center, Yale University School of Medicine, New-Haven, CT, USA d Service Universitaire de Psychiatrie de l’Enfant et de l’Adolescent, CHU de Nantes, Nantes, France e Service Universitaire de Psychiatrie de l’Enfant et de l’Adolescent, Hôpital le Vinatier, Bron, France f INSERM U 693, Université Paris-Sud, Faculté de Médecine Paris-Sud, Le Kremlin-Bicêtre, France g AP-HP, CHU Bicêtre, Service de Génétique Moléculaire, Pharmacogénétique et Hormonologie, Le Kremlin-Bicêtre, France h Maison Départementale de l’Adolescent et Centre Médico-Psycho-Pédagogique, Conseil Général des Bouches-du-Rhône, France i Laboratoire de Santé Publique EA3279, Faculté de Médecine de Marseille, France j Service Hospitalo-Universitaire de Psychiatrie de l’Enfant et de l’Adolescent de Brest, EA4686, UBO, Brest, France k Medical Biochemistry and Molecular Biology, Paris 6 School of Medicine, Paris, France l Chronobiology Unit, Rothschild Foundation, Paris, France Received 7 January 2014; received in revised form 14 August 2014; accepted 18 August 2014
∗ Corresponding author at: Chef du Pôle Hospitalo-Universitaire de Psychiatrie de l’Enfant et de l’Adolescent, CHGR et Université de Rennes 1, 154 rue de Châtillon, Rennes 35000, France. Tel.: +33 6 15 38 07 48; fax: +33 2 99 32 46 98. E-mail addresses: [email protected], [email protected] (S. Tordjman).
http://dx.doi.org/10.1016/j.psyneuen.2014.08.010 0306-4530/© 2014 Elsevier Ltd. All rights reserved.
228
KEYWORDS Cortisol; Autism; Circadian rhythm; Basal cortisol levels; Stress response; Repeated stress; Sensitization to stressors; Flattened cortisol patterns; Trait anxiety
S. Tordjman et al.
Summary Background: Reports of higher stress responsivity, altered sleep-wake cycle and a melatonin deficit in autism have stimulated interest in the cortisol circadian rhythm in individuals with autism. Methods: The study was conducted on 55 low-functioning children and adolescents with autism (11.3 ± 4.1 years-old) and 32 typically developing controls (11.7 ± 4.9 years-old) matched for age, sex and puberty. Behavioral assessment was performed using the Autism Diagnostic Observation Schedule (ADOS). Salivary samples for measurement of cortisol were collected during a 24-h period (at least 0800 h-Day1, 1600 h, 0800 h-Day2 for 46 individuals with autism and 27 controls, and 0800 h-Day1, 1100 h, 1600 h, 2400 h, 0800 h-Day2 for 13 individuals with autism and 20 controls). Overnight (2000 h—0800 h) urinary cortisol excretion was also measured. Results: The autism group displayed significantly higher levels of salivary cortisol at all timepoints, flatter daytime and nighttime slopes, higher 0800 h cortisol levels on Day2 compared to Day1, and greater variances of salivary and urinary cortisol. There was a significant relationship between salivary cortisol levels and impairments in social interaction and verbal language. Overnight urinary cortisol excretion was similar in the autism and control groups. Conclusion: Anticipation of the stressful collection procedure appears to contribute to the higher 0800 h-Day2 versus 0800 h-Day1 salivary cortisol levels in autism. This sensitization to stressors might be as, or even more, important clinically than exposure to novelty in autism. The similar group means for overnight urinary cortisol excretion indicate that basal HPA axis functioning is unaltered in low-functioning autism. The elevated salivary cortisol levels observed in autism over the 24-h period in a repeated stressful condition, flattened diurnal cortisol patterns and the apparent effect of anticipation are consistent with prior findings in high trait anxiety. © 2014 Elsevier Ltd. All rights reserved.
1. Introduction Cortisol, the primary hormonal endoproduct of the hypothalamic-pituitary-adrenal (HPA) axis, is a crucial component of the response to environmental stressors followed by the restoration of basal activity via HPA axis feedback inhibition mechanisms. Chronic elevation of cortisol, including hypercortisolemia and elevated nadir has been implicated in the pathogenesis of psychiatric disorders such as depression (for a review, see Keller et al., 2006). Cortisol circadian rhythm has been well documented in typically developing children and its pattern is usually well established by 6 months of age (Onishi et al., 1983; Price et al., 1983). The lowest cortisol levels are observed between 2000 h and 0200 h and levels increase thereafter up to the highest levels observed shortly after awakening regardless of age, followed typically by a decrease throughout day in absence of external stimuli (Weitzman et al., 1971; Touitou et al., 1981; Pruessner et al., 1997). Cortisol is considered as an important marker of the circadian time structure as the pattern of its secretion is highly reproducible from day to day (Selmaoui and Touitou, 2003). Cortisol circadian rhythms depend upon exogenous factors such as light (Levine et al., 1994) and season (Touitou et al., 1983; Matchock et al., 2007; Persson et al., 2008), and might be also influenced by social factors (Aschoff et al., 1971; Moore-Ede et al., 1982). Cortisol circadian rhythms, like any circadian rhythm (such as melatonin circadian rhythm, sleep-wake cycle, body temperature rhythm) allow optimal and anticipatory temporal organization of biological
functions in relation to periodic changes of the environment (Schibler, 2009; Challet, 2010; Pevet and Challet, 2011). Much of the impetus to study cortisol in autism has come from the apparent heightened anxiety and disordered arousal often observed in individuals with autism (Van Steensel et al., 2011 for a meta-analysis of anxiety disorders in Autistic Disorder; Ozsivadjian et al., 2012; Simon and Corbett, 2013). Reports of higher stress responsivity (for a review, see Tordjman et al., 1997), alterations in circadian sleep-wake rhythm (Glickman, 2010; Leu et al., 2011; Kotagal and Broomall, 2012) and altered melatonin secretion (Melke et al., 2008; Mulder et al., 2010; Tordjman et al., 2005, 2012, 2013) have further stimulated interest in the study of cortisol circadian rhythm. The reports of studies on cortisol circadian rhythm in Autistic Disorder are listed in Table 1 Although some of the studies have found abnormalities in cortisol circadian rhythm (see Table 1: Yamazaki et al., 1975; Hill et al., 1977; Hoshino et al., 1987; Aihara and Hashimoto, 1989; Kaneko et al., 1993), the results were not entirely consistent (typical cortisol circadian rhythm was, for example, observed by Richdale and Prior, 1992; Corbett et al., 2006; Marinovi´ c-Curin et al., 2008). Discrepancies in the results might be related to the level of cognitive functioning, study methods (plasma cortisol measures vs. urinary or salivary cortisol measures) and sample sizes. Thus, most of the studies were hampered by relatively small sample sizes due in particular to methodological challenges of repeated blood measures in individuals with Autistic Disorder. With the development of salivary cortisol
Circadian patterns of cortisol in low-functioning autism Table 1
229
Studies of cortisol circadian rhythm in children and adolescents with autism.
Studies
Measure
Individuals with autism
Controls
Results
Yamazaki et al. (1975) Hill et al. (1977) Aihara and Hashimoto (1989)
Plasma cortisol Plasma cortisol Plasma cortisol
7 LF 6 LF 14 LF
0 3 TD 18 ID
Hoshino et al. (1987)
10 HF 5 LF
7 TD
Kaneko et al. (1993)
Dexamethasone suppression test: DST (salivary cortisol) Salivary cortisol
Altered cortisol rhythm Abnormal cortisol rhythm in patients 11/14 patients showed an atypical 24-h secretion rhythm Atypical cortisol rhythm in patients
21 HF + LF
10 TDa
Marinovi´ c-Curin et al. (2008)
Salivary cortisol
9 NS
7 TD
Corbett et al. (2006)
Salivary cortisol
12 HF
10 TD
Corbett et al. (2008)
Salivary cortisol
22 HF
22 TD
Corbett et al. (2009)
Salivary cortisol
22 HF
22 TD
Kidd et al. (2012)
Salivary cortisol
26 HF + LF
26 TD
Richdale and Prior (1992)
Urinary cortisol
18 HF
19 TD
Atypical rhythm (absence of cortisol variations) in patients Unaltered circadian rhythm in male patients No significant group differences in cortisol circadian rhythm, but the between-child variance in cortisol was greater for the autism group compared to the control one Lack of consistency in day-to-day rhythms (children with autism showed more between- and within-subject variability in circadian rhythms); significantly elevated evening values (30-min before bedtime) in children with autism compared to controls Shallower slope from morning to evening with elevated evening values in high-functioning children with autism compared to controls; significantly greater variability in diurnal cortisol in children with autism than in controls The between-subject variability in cortisol across the day was greater in children with autism than in controls, although the variance estimates were small; higher levels at waking time (but not at midday and bedtime) compared to controls; trend to higher levels when IQ was lower Typical diurnal rhythm in patients; trend to elevated levels throughout the day in children with autism compared to controls
Notes. LF: low functioning; HF: high functioning; TD: typically developing including cognitive development; ID: intellectually disabled; NS: level of cognitive functioning not specified. a Study conducted on healthy adult controls.
assay techniques, research on circadian patterns of cortisol has increased (for a comparison of saliva collection methods in children with high-functioning autism, see Putnam et al., 2012). Saliva samples reflect plasma levels of free cortisol, the unbound physiologically active fraction of cortisol (Kirschbaum and Hellhammer, 1994; Wedekind et al., 2000). Saliva collection is considered to be relatively non-invasive and can be performed in the context of regular activities,
allowing for multiple repeated sampling in larger groups. Given the importance of cortisol as a marker of the circadian time structure and stress responsivity, we have examined the circadian patterns of salivary cortisol in individuals with autism to assess potential circadian rhythm alterations in these patients. In order to better understand our results, basal functioning of the HPA axis was also assessed by measuring overnight urinary free cortisol as it is well established
230 that urinary free cortisol excretion provides a non-invasive index of cortisol secretion over the period of the collection (Findling and Raff, 2001; Stewart and Krone, 2011). The objective of this study was to examine the circadian patterns of salivary cortisol in order to assess diurnal variations and potential circadian rhythm alterations in a large sample of children and adolescents with autism compared to typically developing controls, and to examine also the relationship between the circadian patterns of salivary cortisol and the severity of autism.
2. Methods 2.1. Participants The study was carried out on 55 children and adolescents with autism and 32 typically developing controls matched on age, sex, and Tanner stage of puberty assessed by a pediatrician (Tanner, 1962). Individuals with autism were recruited from outpatient French day-care facilities, and included 36 males and 19 females (mean age 11.3 ± 4.1 years; 28 prepubertal, 13 pubertal, 14 postpubertal). The typically developing controls were recruited over a threemonth period from a preventive medical center where they went for a regular check-up. They were referred by the pediatrician working at the preventive medical center. The control group included 22 males and 10 females (mean age = 11.7 ± 4.9 years; 17 prepubertal, 7 pubertal, 8 postpubertal). The two groups did not differ significantly with respect to age, sex, and pubertal status. All patients and typically developing controls were Caucasian, non-obese, had no history of encephalopathy or neuro-endocrinological disease, and were determined to be physically healthy based on the examination by the pediatrician. Typically developing controls were determined to be free of any significant psychopathological, developmental or neurological disorder, and to have no family history of Autistic Disorder. All subjects were unmedicated for at least 3 months before the saliva collection. All individuals in both subject groups were sleeping in their parents’ house, were going to bed between 2100 h and 2200 h, were waking up between 0700 h and 0745 h (7:45 AM), and were attending school or day hospitals on a daily basis from about 0830 h (8:30 AM) to 1630 h (4:30 PM). Patients and controls had regular set times for eating according to parental reports (breakfast: 7:30—8:30 AM, lunch: 12:30—1:30 PM, and dinner: 7:30—8:30 PM). None were on special diets and no unusual food aversions were noted. The study conformed to standards and ethics for biological rhythm research in humans (Portaluppi et al., 2010). The protocol was approved by the ethics committee of Bicêtre hospital and written informed consent was obtained from parents.
2.2. Cognitive and behavioral assessments Cognitive functioning of patients with autism was assessed by two psychologists using the age-appropriate Weschler intelligence scales and the Kaufman K-ABC (Anatasi, 1988). All patients with autism were cognitively impaired (mean full scale IQ ± SD: 42.2 ± 3.2, with a range 40—58; mean
S. Tordjman et al. verbal IQ ± SD: 45.5 ± 2.2, with a range 45—57; mean performance IQ ± SD: 45.6 ± 4.1, with a range 45—80). Based on direct clinical observation of the patients by two independent child psychiatrists, a diagnosis of Autistic Disorder was made according to the criteria of DSM-IVTR (American Psychiatric Association, 2000), ICD-10 and CFTMEA (Misès and Quemada, 1993), and was confirmed by the ADI-R (Autism Diagnostic Interview-Revised) and ADOS-G (Autism Diagnostic Observation Schedule-Generic) scales (Lord, 1997). The ADOS-G (Module 1), an extensive observational scale, was administered by two trained child psychiatrists. The severity of impairments in the behavioral domains of autism was scored using the subset of items included in the ADOS (module 1) algorithm, following a procedure previously described (Tordjman et al., 2001). Because the ADOS items are scored on an ordinal scale (from 1 to 3 according to autism severity; the ‘‘0’’ coding means that the autistic behavior was not present), we took the median value of all items belonging to the same domain of autistic impairment according to the ADOS (module 1) algorithm. This gave a score of central tendency for each of the four main domains: Reciprocal Social Interaction Total (7 items), Communication Total (5 items), Stereotyped Behaviors and Restricted Interests Total (3 items), and Play Total (2 items). The median domain scores of autistic impairments for each individual were also used to dichotomize subjects according to autism severity in each of the domains: individuals were grouped into mild-moderate (median 1—2) and severe (median 3) impairment. Inter-judge reliability with respect to the critical distinction between mild/moderate and severe impairment was excellent, with an inter-judge agreement of 90% observed between the two expert raters (given the large number of severely autistic patients in our sample, it was possible to clearly distinguish the severely autistic subgroup).
2.3. Procedure and determination of salivary cortisol Parents collected from their children in autumn (SeptemberNovember), at home on a Saturday, five saliva samples at fixed times: 0800 h (0800 h-Day1), 1100 h, 1600 h, 2400 h and 0800 h the following morning (0800 h-Day2). The 0800 h time-points were collected 15—60 min after awakening, and to some extent reflect the cortisol awakening response (CAR) that is defined as the rise of cortisol secretory activity in the 30—60 min post-awakening (Wilhelm et al., 2007). Parents were asked to collect at least the 0800 h-Day1, 1600 h and 0800 h-Day2 time-points, and if possible the 1100 h and 2400 h time-points. Most of the parents of children with autism (35 out of 55) chose not to awake their child for the 2400 h time-point. The study was carried out in a single season to limit any effects of circannual variations of cortisol secretion. All parents were asked to limit their child’s intake of caffeine (no coffee or tea) during the day of the collection period and to ensure that their child did not brush their teeth in order to prevent alterations of cortisol measures by gum bleeding. Furthermore, parents were instructed to prevent their children from eating and drinking for at least 1 h prior to sampling. In addition, parents filled out a
Circadian patterns of cortisol in low-functioning autism questionnaire on their child’s sleep pattern, dietary pattern and the occurrence of any stressful events during the collection period. Concerning the parental sleep questionnaire, parents of children with autism and control subjects completed a questionnaire reporting the following information for the night before and for the night of the collection: time to bed, time to sleep, nighttime awakenings other than the midnight awakening necessary for the 2400 h time-point, if their child slept with the light on or got up during the night and turned on the light, and presence of other sleep disturbances. Parents were encouraged to maintain the usual schedule (dietary pattern as well as sleep pattern, including habitual bedtime and wake-up time) for their children during the collection period. On the night before and on the night of saliva collection, all the individuals with Autistic Disorder and the typically developing controls went to bed with lights out between 2100 h and 2200 h. None of them was exposed to light until their morning wake-up between 0700 h and 0745 h (7:45 AM). Based on the brief parental sleep questionnaire, none of the study participants showed any sleep problem for the night before and night of the saliva collection. In addition, no highly stressful or notable events (except for the saliva collection) was reported for any of the subjects during the collection period. Parents were trained on the collection procedure and were provided with dental cotton rolls (dam), syringes, sample tubes and pre-printed labels. Parents were instructed to place the dental cotton roll in their children’s mouth for at least 1 min. Then the wet dam was placed in a syringe and at least 250 l of the saliva was expelled into a pre-labeled 500 l-eppendorf tube. Parents were asked to write for each saliva sample the exact time of the saliva collection on the pre-printed label. They kept the samples in their home freezer and brought them frozen in a cooler to the laboratory within a month of the saliva collection. In the laboratory, the collected samples were stored in a −20 ◦ C freezer until they were all analyzed in a single assay. Blinded analysis of salivary cortisol levels was performed by using the ELISA (Synchron Enzyme Linked Immuno Sorbent Assay) kit from BIO ADVANCE (Emeranville, France). The intra-assay coefficient of variation was 2.3% for a sample with a concentration of 0.53 g/dl.
2.4. Urine collection and determination of urinary free cortisol Nocturnal urine was collected at home and by parents for the autism group during a 12-h period (from 2000 h to 0800 h) in order to reflect overnight basal HPA axis activity. The urine was collected one week after the saliva collection so as to prevent any effect of the saliva collection. Participants were instructed to empty the bladder between 1945 h (7:45 PM) and 2000 h before starting urine collection. Then, all the urine produced from 2000 h to 0800 h was collected, in particular before going to bed and after the morning wake-up for the toilet trained subjects, and using a urine collection bag (Urinocol® ) placed from 2000 h to 0800 h for the enuretic children with autism (18 individuals with autism out of 55 were enuretic). It was not possible to use a diaper for urine sampling in enuretic children given that it was necessary to get the total overnight urine volume collected during the
231 12-h period (from 2000 h to 0800 h). In addition, no bedwetting or accidental loss of urine was reported for any of the subjects. Furthermore, no subjects woke up or were woken up during the night of urine collection to void their bladder. Parents were asked to report also if their child displayed any uncharacteristic behavioral problems during the night of urine collection. The conditions on the night of urine collection were similar to those maintained for the saliva collection and as with the saliva collection no problems were reported. Collected urines were stored in a refrigerator until delivered to the laboratory within 24 h of the urine collection. The volume of the urine collection was measured and a portion frozen until analyzed for creatinine and cortisol. Blinded analysis of urinary free cortisol (UFC) levels was performed by using Fluorescence Polarization Immunoassay (FPIA) technology (TDx/TDxFLx Cortisol assay) from Abbott Laboratories (Abbott Park, IL, USA). The sensitivity was determined to be 0.64 g/dl and the intra-assay coefficient of variation was less than 10%. Cortisol excretion rate was expressed as g/h for the 12-h period collection and also calculated on a g/mg creatinine basis.
2.5. Statistical analysis The Kolmogorov—Smirnov test indicated that salivary cortisol levels were not normally distributed; thus all statistical analyses were performed using log-transformed cortisol values. Untransformed means and standard errors for salivary cortisol levels are given to allow comparisons to previous studies. Correlations between continuous variables such as cortisol levels and age were calculated by Pearson correlation analyses. Group comparisons of urine collection volumes, urinary excretion of creatinine and urinary free cortisol levels were performed using two-tailed Student’s t-tests. Group and subgroup comparisons of salivary cortisol levels were performed using repeated measures analyses of variance and two-tailed Student’s t-tests. The observed variability in cortisol at each of the 5 consecutive timepoints was compared across groups, and the within-group variability between the 0800 h-Day1 and 0800 h-Day2 was also compared using a test of homogeneity of variance conducted for those subjects with available measurements at all 5 time-points. In addition, following a procedure previously described (Smyth et al., 1997; Ice et al., 2004), logtransformed salivary cortisol levels were regressed against time to determine the daytime slope of the line from 0800 hDay1 to 1600 h and the nighttime slope of the line from 2400 h to 0800 h-Day2 (i.e., 0800 h the following morning) for each individual. A flat slope was defined by a cycle with a slope between −0.05 and +0.05. Group comparisons of slopes were assessed by 2 test and by t-test. Relationships between salivary cortisol levels and autism severity within the major behavioral domains of impairment (the median domain scores of autistic impairments from 1 to 3 were used for each individual) were studied using repeated measures analyses of variance and Spearman rank-order correlation analyses. In order to balance type I and type II errors in the statistical analysis of behavioral domains, a hierarchical strategy was used (Cohen and Cohen, 1983). First, the total behavioral domains were examined. If there was a significant result for an overall domain, then a further level of analysis
232 occurred on the subscores included in the domain. Group means are reported as ±SEM, unless otherwise indicated; alpha was set at 0.05 for all analyses.
S. Tordjman et al.
A
3. Results 3.1. Initial analyses There was no significant effect of gender, age or pubertal status on salivary and urinary cortisol measures for either the autism or typically developing control groups. In addition, there was no significant effect of IQ on salivary or urinary cortisol measures in individuals with autism.
3.2. Urinary measures Overnight urine was successfully collected by the parents at home for 49 out of 55 children with autism and this group (n = 49) was compared to a large typically developing control group (n = 88) previously described (Tordjman et al., 2005) that included the 32 typically developing controls described in the methods section. Mean (±SEM) urine collection volumes were similar in the autism and control groups (325 ± 77 and 303 ± 15 ml, respectively). Urinary excretion of creatinine also did not differ significantly between the autism and control groups (295 ± 77 and 294 ± 18 mg/collection, respectively). There was no significant difference in urinary cortisol levels between individuals with autism and typically developing controls (mean ± SEM: 1.65 ± 0.27 and 1.44 ± 0.10 g/h, respectively). Similar results were observed when cortisol excretion rate was expressed as g/mg creatinine.
3.3. Relationships between diagnosis and salivary cortisol levels In Fig. 1A, group mean (±SEM) salivary cortisol levels measured at 0800 h (8 AM-Day1), 1600 h and 0800 h the next morning (8 AM-Day2) are shown (27 controls and 46 individuals with autism for whom it was possible to get at least these three saliva collection time-points). In Fig. 1B, the data are plotted for those subjects with measurements at all 5 time-points (20 controls and 13 individuals with autism). Attrition was due to the parents’ choice to not awaken their children at midnight combined with the inherent difficulties of obtaining saliva collection in low-functioning children with autism. Repeated measures analysis of variance conducted on the three time-point collections (0800 h-Day1, 1600 h, 0800 hDay2) indicated significant effects of group (F(1,71 ) = 16.28, P = 0.0001) and time (F(2,142 ) = 78.18, P < 0.0001), and a group-by-time interaction (F(2,142 ) = 9.58, P = 0.0001). Repeated measures analysis of variance conducted on all five consecutive collections (0800 h-Day1, 1100 h, 1600 h, 2400 h, 0800 h-Day2) indicated significant effects of group (F(1,31 ) = 20.74, P < 0.0001), time (F(4,124 ) = 54.07, P < 0.0001), but no group-by-time interaction (F(4,124 ) = 1.59, P = 0.187). For the between-group effect, the ANOVA conducted on all five consecutive collections showed higher mean levels of salivary cortisol in individuals with autism
B
Figure 1 (A) Cortisol circadian cycle: Salivary cortisol levels (mean ± SEM) from 0800 h (8 AM-Day1) to 0800 h the next morning (8 AM-Day2) in individuals with autism and typically developing controls (for individuals with at least three timepoint collections). (B) Cortisol circadian cycle: Salivary cortisol levels (mean ± SEM) from 0800 h (8 AM-Day1) to 0800 h the next morning (8 AM-Day2) in individuals with autism and typically developing controls (for individuals with all five consecutive time-point collections).
compared to typically developing controls at all time-points. In particular, the between-group difference was significant for the 1600 h point, the cortisol nadir (2400 h) and the cortisol peak (0800 h-Day2) (see Table 2). Interestingly, for those subjects for whom all five time-points were obtained, the mean levels of salivary cortisol in the control group (n = 20) at 0800 h-Day1 (0.620 ± 0.070) and 0800 h-Day2 (0.620 ± 0.071) were very similar, whereas the mean level of salivary cortisol in the autism group (n = 13) was significantly higher at 0800 h-Day2 (1.209 ± 0.167) compared to 0800 h-Day1 (0.803 ± 0.134), (t(12 ) = 2.45, P < 0.05). Similar results were observed in the autism group (n = 30) for whom the 2400 h time-point was the only time-point not obtained and who did not experience the midnight waking (0800 h-Day2: 1.149 ± 0.146, 0800 h-Day1: 0.961 ± 0.139, t(29 ) = 1.90, P < 0.05).
3.4. Comparison of cortisol variances The variance of salivary cortisol in the autism group was greater than the variance of salivary cortisol in the control group at all time-points, except at 1100 h, but the difference was not significant; the strongest difference of variance was observed at 2400 h. Interestingly, when analyses were performed using non log-transformed cortisol values, the variance of salivary cortisol in the autism group
Circadian patterns of cortisol in low-functioning autism
233
Table 2 Salivary cortisol levels (mean ± SEM) in individuals with autism compared to typically developing controls for those subjects with available measurements at all 5 time-points. Cortisol level (g/dl)
Group Individuals with Autistic Disorder (n = 13)
8 AM-Day1 11 AM 4 PM Midnight 8 AM-Day2
Table 3 groups.
0.803 0.529 0.365 0.305 1.209
± ± ± ± ±
Test statistic
P value
F(1,31) = 1.49 F(1,31) = 1.60 F(1,31) = 11.78 F(1,31) = 18.81 F(1,31) = 11.95
0.2318 0.2154 0.0017 0.0001 0.0016
Typically Developing Controls (n = 20)
0.134 0.094 0.048 0.077 0.168
0.620 0.429 0.201 0.104 0.620
± ± ± ± ±
0.070 0.089 0.018 0.002 0.071
Measures of variability in urinary cortisol and salivary cortisol at each of the 5 consecutive time-points compared by
Standard deviation
Urinary cortisol Salivary cortisol 8 AM-Day1 11 AM 4 PM Midnight 8 AM-Day2
Group
Test statistic
Individuals with Autistic Disorder (n)
Typically Developing Controls (n)
(homogeneity of variance)
1.88 (n = 49) 0.484 (n = 13) 0.340 (n = 13) 0.172 (n = 13) 0.278 (n = 13) 0.604 (n = 13)
0.94 (n = 88) 0.312 (n = 20) 0.396 (n = 20) 0.079 (n = 20) 0.010 (n = 20) 0.317 (n = 20)
F(48,87) = 4.05 F(12,19) = 2.41 F(19,12) = 1.36 F(12,19) = 4.93 F(12,19) = 788.61 F(12,19) = 3.65
P value
Available online at www.sciencedirect.com
ScienceDirect journal homepage: www.elsevier.com/locate/psyneuen
Altered circadian patterns of salivary cortisol in low-functioning children and adolescents with autism Sylvie Tordjman a,b,∗, George M. Anderson c, Solenn Kermarrec a,b, Olivier Bonnot d, Marie-Maude Geoffray e, Sylvie Brailly-Tabard f,g, Amel Chaouch g, Isabelle Colliot g, Severine Trabado f,g, Guillaume Bronsard h,i, Nathalie Coulon b, ¸oise Camus l, Michel Botbol j, Henriette Charbuy k, Franc Yvan Touitou l a
Pôle Hospitalo-Universitaire de Psychiatrie de l’Enfant et de l’Adolescent de Rennes (PHUPEA), CHGR et Université de Rennes 1, Rennes, France b Laboratoire Psychologie de la Perception, Université Paris Descartes, CNRS UMR 8158, Paris, France c Child Study Center, Yale University School of Medicine, New-Haven, CT, USA d Service Universitaire de Psychiatrie de l’Enfant et de l’Adolescent, CHU de Nantes, Nantes, France e Service Universitaire de Psychiatrie de l’Enfant et de l’Adolescent, Hôpital le Vinatier, Bron, France f INSERM U 693, Université Paris-Sud, Faculté de Médecine Paris-Sud, Le Kremlin-Bicêtre, France g AP-HP, CHU Bicêtre, Service de Génétique Moléculaire, Pharmacogénétique et Hormonologie, Le Kremlin-Bicêtre, France h Maison Départementale de l’Adolescent et Centre Médico-Psycho-Pédagogique, Conseil Général des Bouches-du-Rhône, France i Laboratoire de Santé Publique EA3279, Faculté de Médecine de Marseille, France j Service Hospitalo-Universitaire de Psychiatrie de l’Enfant et de l’Adolescent de Brest, EA4686, UBO, Brest, France k Medical Biochemistry and Molecular Biology, Paris 6 School of Medicine, Paris, France l Chronobiology Unit, Rothschild Foundation, Paris, France Received 7 January 2014; received in revised form 14 August 2014; accepted 18 August 2014
∗ Corresponding author at: Chef du Pôle Hospitalo-Universitaire de Psychiatrie de l’Enfant et de l’Adolescent, CHGR et Université de Rennes 1, 154 rue de Châtillon, Rennes 35000, France. Tel.: +33 6 15 38 07 48; fax: +33 2 99 32 46 98. E-mail addresses: [email protected], [email protected] (S. Tordjman).
http://dx.doi.org/10.1016/j.psyneuen.2014.08.010 0306-4530/© 2014 Elsevier Ltd. All rights reserved.
228
KEYWORDS Cortisol; Autism; Circadian rhythm; Basal cortisol levels; Stress response; Repeated stress; Sensitization to stressors; Flattened cortisol patterns; Trait anxiety
S. Tordjman et al.
Summary Background: Reports of higher stress responsivity, altered sleep-wake cycle and a melatonin deficit in autism have stimulated interest in the cortisol circadian rhythm in individuals with autism. Methods: The study was conducted on 55 low-functioning children and adolescents with autism (11.3 ± 4.1 years-old) and 32 typically developing controls (11.7 ± 4.9 years-old) matched for age, sex and puberty. Behavioral assessment was performed using the Autism Diagnostic Observation Schedule (ADOS). Salivary samples for measurement of cortisol were collected during a 24-h period (at least 0800 h-Day1, 1600 h, 0800 h-Day2 for 46 individuals with autism and 27 controls, and 0800 h-Day1, 1100 h, 1600 h, 2400 h, 0800 h-Day2 for 13 individuals with autism and 20 controls). Overnight (2000 h—0800 h) urinary cortisol excretion was also measured. Results: The autism group displayed significantly higher levels of salivary cortisol at all timepoints, flatter daytime and nighttime slopes, higher 0800 h cortisol levels on Day2 compared to Day1, and greater variances of salivary and urinary cortisol. There was a significant relationship between salivary cortisol levels and impairments in social interaction and verbal language. Overnight urinary cortisol excretion was similar in the autism and control groups. Conclusion: Anticipation of the stressful collection procedure appears to contribute to the higher 0800 h-Day2 versus 0800 h-Day1 salivary cortisol levels in autism. This sensitization to stressors might be as, or even more, important clinically than exposure to novelty in autism. The similar group means for overnight urinary cortisol excretion indicate that basal HPA axis functioning is unaltered in low-functioning autism. The elevated salivary cortisol levels observed in autism over the 24-h period in a repeated stressful condition, flattened diurnal cortisol patterns and the apparent effect of anticipation are consistent with prior findings in high trait anxiety. © 2014 Elsevier Ltd. All rights reserved.
1. Introduction Cortisol, the primary hormonal endoproduct of the hypothalamic-pituitary-adrenal (HPA) axis, is a crucial component of the response to environmental stressors followed by the restoration of basal activity via HPA axis feedback inhibition mechanisms. Chronic elevation of cortisol, including hypercortisolemia and elevated nadir has been implicated in the pathogenesis of psychiatric disorders such as depression (for a review, see Keller et al., 2006). Cortisol circadian rhythm has been well documented in typically developing children and its pattern is usually well established by 6 months of age (Onishi et al., 1983; Price et al., 1983). The lowest cortisol levels are observed between 2000 h and 0200 h and levels increase thereafter up to the highest levels observed shortly after awakening regardless of age, followed typically by a decrease throughout day in absence of external stimuli (Weitzman et al., 1971; Touitou et al., 1981; Pruessner et al., 1997). Cortisol is considered as an important marker of the circadian time structure as the pattern of its secretion is highly reproducible from day to day (Selmaoui and Touitou, 2003). Cortisol circadian rhythms depend upon exogenous factors such as light (Levine et al., 1994) and season (Touitou et al., 1983; Matchock et al., 2007; Persson et al., 2008), and might be also influenced by social factors (Aschoff et al., 1971; Moore-Ede et al., 1982). Cortisol circadian rhythms, like any circadian rhythm (such as melatonin circadian rhythm, sleep-wake cycle, body temperature rhythm) allow optimal and anticipatory temporal organization of biological
functions in relation to periodic changes of the environment (Schibler, 2009; Challet, 2010; Pevet and Challet, 2011). Much of the impetus to study cortisol in autism has come from the apparent heightened anxiety and disordered arousal often observed in individuals with autism (Van Steensel et al., 2011 for a meta-analysis of anxiety disorders in Autistic Disorder; Ozsivadjian et al., 2012; Simon and Corbett, 2013). Reports of higher stress responsivity (for a review, see Tordjman et al., 1997), alterations in circadian sleep-wake rhythm (Glickman, 2010; Leu et al., 2011; Kotagal and Broomall, 2012) and altered melatonin secretion (Melke et al., 2008; Mulder et al., 2010; Tordjman et al., 2005, 2012, 2013) have further stimulated interest in the study of cortisol circadian rhythm. The reports of studies on cortisol circadian rhythm in Autistic Disorder are listed in Table 1 Although some of the studies have found abnormalities in cortisol circadian rhythm (see Table 1: Yamazaki et al., 1975; Hill et al., 1977; Hoshino et al., 1987; Aihara and Hashimoto, 1989; Kaneko et al., 1993), the results were not entirely consistent (typical cortisol circadian rhythm was, for example, observed by Richdale and Prior, 1992; Corbett et al., 2006; Marinovi´ c-Curin et al., 2008). Discrepancies in the results might be related to the level of cognitive functioning, study methods (plasma cortisol measures vs. urinary or salivary cortisol measures) and sample sizes. Thus, most of the studies were hampered by relatively small sample sizes due in particular to methodological challenges of repeated blood measures in individuals with Autistic Disorder. With the development of salivary cortisol
Circadian patterns of cortisol in low-functioning autism Table 1
229
Studies of cortisol circadian rhythm in children and adolescents with autism.
Studies
Measure
Individuals with autism
Controls
Results
Yamazaki et al. (1975) Hill et al. (1977) Aihara and Hashimoto (1989)
Plasma cortisol Plasma cortisol Plasma cortisol
7 LF 6 LF 14 LF
0 3 TD 18 ID
Hoshino et al. (1987)
10 HF 5 LF
7 TD
Kaneko et al. (1993)
Dexamethasone suppression test: DST (salivary cortisol) Salivary cortisol
Altered cortisol rhythm Abnormal cortisol rhythm in patients 11/14 patients showed an atypical 24-h secretion rhythm Atypical cortisol rhythm in patients
21 HF + LF
10 TDa
Marinovi´ c-Curin et al. (2008)
Salivary cortisol
9 NS
7 TD
Corbett et al. (2006)
Salivary cortisol
12 HF
10 TD
Corbett et al. (2008)
Salivary cortisol
22 HF
22 TD
Corbett et al. (2009)
Salivary cortisol
22 HF
22 TD
Kidd et al. (2012)
Salivary cortisol
26 HF + LF
26 TD
Richdale and Prior (1992)
Urinary cortisol
18 HF
19 TD
Atypical rhythm (absence of cortisol variations) in patients Unaltered circadian rhythm in male patients No significant group differences in cortisol circadian rhythm, but the between-child variance in cortisol was greater for the autism group compared to the control one Lack of consistency in day-to-day rhythms (children with autism showed more between- and within-subject variability in circadian rhythms); significantly elevated evening values (30-min before bedtime) in children with autism compared to controls Shallower slope from morning to evening with elevated evening values in high-functioning children with autism compared to controls; significantly greater variability in diurnal cortisol in children with autism than in controls The between-subject variability in cortisol across the day was greater in children with autism than in controls, although the variance estimates were small; higher levels at waking time (but not at midday and bedtime) compared to controls; trend to higher levels when IQ was lower Typical diurnal rhythm in patients; trend to elevated levels throughout the day in children with autism compared to controls
Notes. LF: low functioning; HF: high functioning; TD: typically developing including cognitive development; ID: intellectually disabled; NS: level of cognitive functioning not specified. a Study conducted on healthy adult controls.
assay techniques, research on circadian patterns of cortisol has increased (for a comparison of saliva collection methods in children with high-functioning autism, see Putnam et al., 2012). Saliva samples reflect plasma levels of free cortisol, the unbound physiologically active fraction of cortisol (Kirschbaum and Hellhammer, 1994; Wedekind et al., 2000). Saliva collection is considered to be relatively non-invasive and can be performed in the context of regular activities,
allowing for multiple repeated sampling in larger groups. Given the importance of cortisol as a marker of the circadian time structure and stress responsivity, we have examined the circadian patterns of salivary cortisol in individuals with autism to assess potential circadian rhythm alterations in these patients. In order to better understand our results, basal functioning of the HPA axis was also assessed by measuring overnight urinary free cortisol as it is well established
230 that urinary free cortisol excretion provides a non-invasive index of cortisol secretion over the period of the collection (Findling and Raff, 2001; Stewart and Krone, 2011). The objective of this study was to examine the circadian patterns of salivary cortisol in order to assess diurnal variations and potential circadian rhythm alterations in a large sample of children and adolescents with autism compared to typically developing controls, and to examine also the relationship between the circadian patterns of salivary cortisol and the severity of autism.
2. Methods 2.1. Participants The study was carried out on 55 children and adolescents with autism and 32 typically developing controls matched on age, sex, and Tanner stage of puberty assessed by a pediatrician (Tanner, 1962). Individuals with autism were recruited from outpatient French day-care facilities, and included 36 males and 19 females (mean age 11.3 ± 4.1 years; 28 prepubertal, 13 pubertal, 14 postpubertal). The typically developing controls were recruited over a threemonth period from a preventive medical center where they went for a regular check-up. They were referred by the pediatrician working at the preventive medical center. The control group included 22 males and 10 females (mean age = 11.7 ± 4.9 years; 17 prepubertal, 7 pubertal, 8 postpubertal). The two groups did not differ significantly with respect to age, sex, and pubertal status. All patients and typically developing controls were Caucasian, non-obese, had no history of encephalopathy or neuro-endocrinological disease, and were determined to be physically healthy based on the examination by the pediatrician. Typically developing controls were determined to be free of any significant psychopathological, developmental or neurological disorder, and to have no family history of Autistic Disorder. All subjects were unmedicated for at least 3 months before the saliva collection. All individuals in both subject groups were sleeping in their parents’ house, were going to bed between 2100 h and 2200 h, were waking up between 0700 h and 0745 h (7:45 AM), and were attending school or day hospitals on a daily basis from about 0830 h (8:30 AM) to 1630 h (4:30 PM). Patients and controls had regular set times for eating according to parental reports (breakfast: 7:30—8:30 AM, lunch: 12:30—1:30 PM, and dinner: 7:30—8:30 PM). None were on special diets and no unusual food aversions were noted. The study conformed to standards and ethics for biological rhythm research in humans (Portaluppi et al., 2010). The protocol was approved by the ethics committee of Bicêtre hospital and written informed consent was obtained from parents.
2.2. Cognitive and behavioral assessments Cognitive functioning of patients with autism was assessed by two psychologists using the age-appropriate Weschler intelligence scales and the Kaufman K-ABC (Anatasi, 1988). All patients with autism were cognitively impaired (mean full scale IQ ± SD: 42.2 ± 3.2, with a range 40—58; mean
S. Tordjman et al. verbal IQ ± SD: 45.5 ± 2.2, with a range 45—57; mean performance IQ ± SD: 45.6 ± 4.1, with a range 45—80). Based on direct clinical observation of the patients by two independent child psychiatrists, a diagnosis of Autistic Disorder was made according to the criteria of DSM-IVTR (American Psychiatric Association, 2000), ICD-10 and CFTMEA (Misès and Quemada, 1993), and was confirmed by the ADI-R (Autism Diagnostic Interview-Revised) and ADOS-G (Autism Diagnostic Observation Schedule-Generic) scales (Lord, 1997). The ADOS-G (Module 1), an extensive observational scale, was administered by two trained child psychiatrists. The severity of impairments in the behavioral domains of autism was scored using the subset of items included in the ADOS (module 1) algorithm, following a procedure previously described (Tordjman et al., 2001). Because the ADOS items are scored on an ordinal scale (from 1 to 3 according to autism severity; the ‘‘0’’ coding means that the autistic behavior was not present), we took the median value of all items belonging to the same domain of autistic impairment according to the ADOS (module 1) algorithm. This gave a score of central tendency for each of the four main domains: Reciprocal Social Interaction Total (7 items), Communication Total (5 items), Stereotyped Behaviors and Restricted Interests Total (3 items), and Play Total (2 items). The median domain scores of autistic impairments for each individual were also used to dichotomize subjects according to autism severity in each of the domains: individuals were grouped into mild-moderate (median 1—2) and severe (median 3) impairment. Inter-judge reliability with respect to the critical distinction between mild/moderate and severe impairment was excellent, with an inter-judge agreement of 90% observed between the two expert raters (given the large number of severely autistic patients in our sample, it was possible to clearly distinguish the severely autistic subgroup).
2.3. Procedure and determination of salivary cortisol Parents collected from their children in autumn (SeptemberNovember), at home on a Saturday, five saliva samples at fixed times: 0800 h (0800 h-Day1), 1100 h, 1600 h, 2400 h and 0800 h the following morning (0800 h-Day2). The 0800 h time-points were collected 15—60 min after awakening, and to some extent reflect the cortisol awakening response (CAR) that is defined as the rise of cortisol secretory activity in the 30—60 min post-awakening (Wilhelm et al., 2007). Parents were asked to collect at least the 0800 h-Day1, 1600 h and 0800 h-Day2 time-points, and if possible the 1100 h and 2400 h time-points. Most of the parents of children with autism (35 out of 55) chose not to awake their child for the 2400 h time-point. The study was carried out in a single season to limit any effects of circannual variations of cortisol secretion. All parents were asked to limit their child’s intake of caffeine (no coffee or tea) during the day of the collection period and to ensure that their child did not brush their teeth in order to prevent alterations of cortisol measures by gum bleeding. Furthermore, parents were instructed to prevent their children from eating and drinking for at least 1 h prior to sampling. In addition, parents filled out a
Circadian patterns of cortisol in low-functioning autism questionnaire on their child’s sleep pattern, dietary pattern and the occurrence of any stressful events during the collection period. Concerning the parental sleep questionnaire, parents of children with autism and control subjects completed a questionnaire reporting the following information for the night before and for the night of the collection: time to bed, time to sleep, nighttime awakenings other than the midnight awakening necessary for the 2400 h time-point, if their child slept with the light on or got up during the night and turned on the light, and presence of other sleep disturbances. Parents were encouraged to maintain the usual schedule (dietary pattern as well as sleep pattern, including habitual bedtime and wake-up time) for their children during the collection period. On the night before and on the night of saliva collection, all the individuals with Autistic Disorder and the typically developing controls went to bed with lights out between 2100 h and 2200 h. None of them was exposed to light until their morning wake-up between 0700 h and 0745 h (7:45 AM). Based on the brief parental sleep questionnaire, none of the study participants showed any sleep problem for the night before and night of the saliva collection. In addition, no highly stressful or notable events (except for the saliva collection) was reported for any of the subjects during the collection period. Parents were trained on the collection procedure and were provided with dental cotton rolls (dam), syringes, sample tubes and pre-printed labels. Parents were instructed to place the dental cotton roll in their children’s mouth for at least 1 min. Then the wet dam was placed in a syringe and at least 250 l of the saliva was expelled into a pre-labeled 500 l-eppendorf tube. Parents were asked to write for each saliva sample the exact time of the saliva collection on the pre-printed label. They kept the samples in their home freezer and brought them frozen in a cooler to the laboratory within a month of the saliva collection. In the laboratory, the collected samples were stored in a −20 ◦ C freezer until they were all analyzed in a single assay. Blinded analysis of salivary cortisol levels was performed by using the ELISA (Synchron Enzyme Linked Immuno Sorbent Assay) kit from BIO ADVANCE (Emeranville, France). The intra-assay coefficient of variation was 2.3% for a sample with a concentration of 0.53 g/dl.
2.4. Urine collection and determination of urinary free cortisol Nocturnal urine was collected at home and by parents for the autism group during a 12-h period (from 2000 h to 0800 h) in order to reflect overnight basal HPA axis activity. The urine was collected one week after the saliva collection so as to prevent any effect of the saliva collection. Participants were instructed to empty the bladder between 1945 h (7:45 PM) and 2000 h before starting urine collection. Then, all the urine produced from 2000 h to 0800 h was collected, in particular before going to bed and after the morning wake-up for the toilet trained subjects, and using a urine collection bag (Urinocol® ) placed from 2000 h to 0800 h for the enuretic children with autism (18 individuals with autism out of 55 were enuretic). It was not possible to use a diaper for urine sampling in enuretic children given that it was necessary to get the total overnight urine volume collected during the
231 12-h period (from 2000 h to 0800 h). In addition, no bedwetting or accidental loss of urine was reported for any of the subjects. Furthermore, no subjects woke up or were woken up during the night of urine collection to void their bladder. Parents were asked to report also if their child displayed any uncharacteristic behavioral problems during the night of urine collection. The conditions on the night of urine collection were similar to those maintained for the saliva collection and as with the saliva collection no problems were reported. Collected urines were stored in a refrigerator until delivered to the laboratory within 24 h of the urine collection. The volume of the urine collection was measured and a portion frozen until analyzed for creatinine and cortisol. Blinded analysis of urinary free cortisol (UFC) levels was performed by using Fluorescence Polarization Immunoassay (FPIA) technology (TDx/TDxFLx Cortisol assay) from Abbott Laboratories (Abbott Park, IL, USA). The sensitivity was determined to be 0.64 g/dl and the intra-assay coefficient of variation was less than 10%. Cortisol excretion rate was expressed as g/h for the 12-h period collection and also calculated on a g/mg creatinine basis.
2.5. Statistical analysis The Kolmogorov—Smirnov test indicated that salivary cortisol levels were not normally distributed; thus all statistical analyses were performed using log-transformed cortisol values. Untransformed means and standard errors for salivary cortisol levels are given to allow comparisons to previous studies. Correlations between continuous variables such as cortisol levels and age were calculated by Pearson correlation analyses. Group comparisons of urine collection volumes, urinary excretion of creatinine and urinary free cortisol levels were performed using two-tailed Student’s t-tests. Group and subgroup comparisons of salivary cortisol levels were performed using repeated measures analyses of variance and two-tailed Student’s t-tests. The observed variability in cortisol at each of the 5 consecutive timepoints was compared across groups, and the within-group variability between the 0800 h-Day1 and 0800 h-Day2 was also compared using a test of homogeneity of variance conducted for those subjects with available measurements at all 5 time-points. In addition, following a procedure previously described (Smyth et al., 1997; Ice et al., 2004), logtransformed salivary cortisol levels were regressed against time to determine the daytime slope of the line from 0800 hDay1 to 1600 h and the nighttime slope of the line from 2400 h to 0800 h-Day2 (i.e., 0800 h the following morning) for each individual. A flat slope was defined by a cycle with a slope between −0.05 and +0.05. Group comparisons of slopes were assessed by 2 test and by t-test. Relationships between salivary cortisol levels and autism severity within the major behavioral domains of impairment (the median domain scores of autistic impairments from 1 to 3 were used for each individual) were studied using repeated measures analyses of variance and Spearman rank-order correlation analyses. In order to balance type I and type II errors in the statistical analysis of behavioral domains, a hierarchical strategy was used (Cohen and Cohen, 1983). First, the total behavioral domains were examined. If there was a significant result for an overall domain, then a further level of analysis
232 occurred on the subscores included in the domain. Group means are reported as ±SEM, unless otherwise indicated; alpha was set at 0.05 for all analyses.
S. Tordjman et al.
A
3. Results 3.1. Initial analyses There was no significant effect of gender, age or pubertal status on salivary and urinary cortisol measures for either the autism or typically developing control groups. In addition, there was no significant effect of IQ on salivary or urinary cortisol measures in individuals with autism.
3.2. Urinary measures Overnight urine was successfully collected by the parents at home for 49 out of 55 children with autism and this group (n = 49) was compared to a large typically developing control group (n = 88) previously described (Tordjman et al., 2005) that included the 32 typically developing controls described in the methods section. Mean (±SEM) urine collection volumes were similar in the autism and control groups (325 ± 77 and 303 ± 15 ml, respectively). Urinary excretion of creatinine also did not differ significantly between the autism and control groups (295 ± 77 and 294 ± 18 mg/collection, respectively). There was no significant difference in urinary cortisol levels between individuals with autism and typically developing controls (mean ± SEM: 1.65 ± 0.27 and 1.44 ± 0.10 g/h, respectively). Similar results were observed when cortisol excretion rate was expressed as g/mg creatinine.
3.3. Relationships between diagnosis and salivary cortisol levels In Fig. 1A, group mean (±SEM) salivary cortisol levels measured at 0800 h (8 AM-Day1), 1600 h and 0800 h the next morning (8 AM-Day2) are shown (27 controls and 46 individuals with autism for whom it was possible to get at least these three saliva collection time-points). In Fig. 1B, the data are plotted for those subjects with measurements at all 5 time-points (20 controls and 13 individuals with autism). Attrition was due to the parents’ choice to not awaken their children at midnight combined with the inherent difficulties of obtaining saliva collection in low-functioning children with autism. Repeated measures analysis of variance conducted on the three time-point collections (0800 h-Day1, 1600 h, 0800 hDay2) indicated significant effects of group (F(1,71 ) = 16.28, P = 0.0001) and time (F(2,142 ) = 78.18, P < 0.0001), and a group-by-time interaction (F(2,142 ) = 9.58, P = 0.0001). Repeated measures analysis of variance conducted on all five consecutive collections (0800 h-Day1, 1100 h, 1600 h, 2400 h, 0800 h-Day2) indicated significant effects of group (F(1,31 ) = 20.74, P < 0.0001), time (F(4,124 ) = 54.07, P < 0.0001), but no group-by-time interaction (F(4,124 ) = 1.59, P = 0.187). For the between-group effect, the ANOVA conducted on all five consecutive collections showed higher mean levels of salivary cortisol in individuals with autism
B
Figure 1 (A) Cortisol circadian cycle: Salivary cortisol levels (mean ± SEM) from 0800 h (8 AM-Day1) to 0800 h the next morning (8 AM-Day2) in individuals with autism and typically developing controls (for individuals with at least three timepoint collections). (B) Cortisol circadian cycle: Salivary cortisol levels (mean ± SEM) from 0800 h (8 AM-Day1) to 0800 h the next morning (8 AM-Day2) in individuals with autism and typically developing controls (for individuals with all five consecutive time-point collections).
compared to typically developing controls at all time-points. In particular, the between-group difference was significant for the 1600 h point, the cortisol nadir (2400 h) and the cortisol peak (0800 h-Day2) (see Table 2). Interestingly, for those subjects for whom all five time-points were obtained, the mean levels of salivary cortisol in the control group (n = 20) at 0800 h-Day1 (0.620 ± 0.070) and 0800 h-Day2 (0.620 ± 0.071) were very similar, whereas the mean level of salivary cortisol in the autism group (n = 13) was significantly higher at 0800 h-Day2 (1.209 ± 0.167) compared to 0800 h-Day1 (0.803 ± 0.134), (t(12 ) = 2.45, P < 0.05). Similar results were observed in the autism group (n = 30) for whom the 2400 h time-point was the only time-point not obtained and who did not experience the midnight waking (0800 h-Day2: 1.149 ± 0.146, 0800 h-Day1: 0.961 ± 0.139, t(29 ) = 1.90, P < 0.05).
3.4. Comparison of cortisol variances The variance of salivary cortisol in the autism group was greater than the variance of salivary cortisol in the control group at all time-points, except at 1100 h, but the difference was not significant; the strongest difference of variance was observed at 2400 h. Interestingly, when analyses were performed using non log-transformed cortisol values, the variance of salivary cortisol in the autism group
Circadian patterns of cortisol in low-functioning autism
233
Table 2 Salivary cortisol levels (mean ± SEM) in individuals with autism compared to typically developing controls for those subjects with available measurements at all 5 time-points. Cortisol level (g/dl)
Group Individuals with Autistic Disorder (n = 13)
8 AM-Day1 11 AM 4 PM Midnight 8 AM-Day2
Table 3 groups.
0.803 0.529 0.365 0.305 1.209
± ± ± ± ±
Test statistic
P value
F(1,31) = 1.49 F(1,31) = 1.60 F(1,31) = 11.78 F(1,31) = 18.81 F(1,31) = 11.95
0.2318 0.2154 0.0017 0.0001 0.0016
Typically Developing Controls (n = 20)
0.134 0.094 0.048 0.077 0.168
0.620 0.429 0.201 0.104 0.620
± ± ± ± ±
0.070 0.089 0.018 0.002 0.071
Measures of variability in urinary cortisol and salivary cortisol at each of the 5 consecutive time-points compared by
Standard deviation
Urinary cortisol Salivary cortisol 8 AM-Day1 11 AM 4 PM Midnight 8 AM-Day2
Group
Test statistic
Individuals with Autistic Disorder (n)
Typically Developing Controls (n)
(homogeneity of variance)
1.88 (n = 49) 0.484 (n = 13) 0.340 (n = 13) 0.172 (n = 13) 0.278 (n = 13) 0.604 (n = 13)
0.94 (n = 88) 0.312 (n = 20) 0.396 (n = 20) 0.079 (n = 20) 0.010 (n = 20) 0.317 (n = 20)
F(48,87) = 4.05 F(12,19) = 2.41 F(19,12) = 1.36 F(12,19) = 4.93 F(12,19) = 788.61 F(12,19) = 3.65
P value
