The 24-hour urinary cortisol in post-traumatic stress disorder: A meta-analysis

Objective Previous studies found inconsistent results on the relationship between post-traumatic stress disorder (PTSD) and concentrations of 24-hour (24-h) urinary cortisol. This study performed a systematic review and meta-analysis to summarize previous findings on this relationship. Methods We searched in the databases of Web of Science, PubMed, Embase, and Psyc-ARTICLES for articles published before September 2018. We used the random-effects model with restricted maximum-likelihood estimator to synthesize the effect sizes by calculating the standardized mean difference (SMD) and assessing its significance. Results Six hundred and nineteen articles were identified from the preceding databases and 20 of them were included in the meta-analysis. Lower concentrations of 24-h urinary cortisol were observed in patients with PTSD when compared with the controls (SMD = -0.49, 95%CI [-0.91; -0.07], p = 0.02). Subgroup analysis revealed that the concentrations of 24-h urinary cortisol were lower in PTSD patients than in the controls for studies that included female participants or studies that included participants from the United States of America. Conclusions Overall, decreased levels of 24-h urinary cortisol were linked with the pathophysiology of PTSD. Nonetheless, more studies should be conducted to validate the molecular underpinnings of urine cortisol degeneration in PTSD.


Introduction
The search for relevant articles was restricted to articles published in the English language. This was performed in the following four electronic databases: Web of Science, PubMed, Embase, and Psyc-ARTICLES. The search strategy was designed in consultation with experienced librarians. For example, a search strategy in the database of Embase was structured as follows using keywords (search terms): ('urinary free cortisol':ab,ti OR 'urinary cortisol':ab,ti OR 'cortisol in urine':ab,ti OR 'urine cortisol':ab,ti OR 'glucocorticoids in urine':ab,ti OR 'urine glucocorticoid':ab,ti OR 'steroid hormones in urine':ab,ti OR 'urine steroid':ab,ti OR 'urine corticosteroid':ab,ti OR 'urinary cortisol':ab,ti) and ('ptsd':ab,ti OR 'posttraumatic stress disorder':ab,ti). In addition, hand searching was conducted in the lists of references of the retrieved articles by XP and AC. Any inconsistencies between them were resolved by group discussion and consensus with a third party AL.

Eligibility criteria
A study eligible for this meta-analysis had to meet the following inclusion criteria: (1) the study had to be a case control study, which included a control group and PTSD cases, (2) the study reported the mean and standard deviation (SD) of the 24-h urinary cortisol levels, or these could be provided by the authors upon request, (3) the study assessed PTSD cases simultaneously with the assessment of the 24-h urinary cortisol levels, (4) the study was published in English, and (5) the study reported the diagnostic criteria for PTSD. In addition, studies were excluded if they (1) were review articles or case reports; (2) studied PTSD in combination with HPA axis disorder disease, or studied PTSD in patients with other mental illnesses, who used psychotropic medication or other medications which could influence the HPA axis and cortisol concentrations, and (3) studied non-humans or were vitro experiments or animal research.

Data extraction
Two researchers [XW and ZW] independently screened and selected eligible articles [31]. A third party was involved in consultations to make the final decision in the event of disagreements [AL]. Moreover, the grey literatures (non-published literatures) were excluded from our study [32].
In relation to the purpose of this study, the following information were extracted from the eligible studies by two independent investigators [AC and XW]: (1) name of first author, and publication year; (2) geographical area of the study; (3) characteristics of PTSD participants such as trauma type, age (mean, SD), gender distribution, and body-mass index (BMI) (mean, SD); (4) sample characteristics such as sample size, and concentrations of 24-h urinary cortisol (mean, SD); (5) PTSD assessment method; and (6) 24-h urinary cortisol collection and assay methods, intra-assay variation, inter-assay variation, storage temperatures and sensitivity. All the extracted data were organized in EpiData 3.0 and saved in Excel.

Quality evaluation
The Newcastle-Ottawa Quality Assessment Scale (NOS) was used to assess the quality of the eligible studies [33]. Therefore, each eligible study was evaluated based on the three broad perspectives: (1) Selection; (2) Comparability; and (3) Outcome. Two investigators [XP and AC] independently assessed and graded the eligible studies. Any inconsistencies between them were resolved by group discussion with a third party [AL]. According to the pre-specified criteria of this scale, studies scoring 7-9, 4-6, and 0-3 points were graded, respectively, as high, moderate, and low quality.

Statistical analysis
All analyses were conducted in R software (version R i386 3.4.2). Accordingly, meta-analysis was carried out in the package, meta, and meta-regression analysis was performed in the package, metafor. The standardized mean difference (SMD) of the 24-h urinary cortisol levels between the PTSD and the control groups was calculated using Cohen's d [34]. Moreover, by using restricted maximum-likelihood estimator to synthesize the effect sizes reported in the eligible studies, random-effects models were fitted. The Q-test was carried out to examine whether there was heterogeneity in the results from the eligible studies. This heterogeneity was quantified using the I 2 statistic (I 2 = 0% indicates no heterogeneity and I 2 = 100% indicates maximal heterogeneity) [34]. In order to explore sources of heterogeneity between the eligible studies, subgroup analysis was performed with respect to controls type. Specifically, the following subgroups were used for subgroup analyses in this study: study country (USA or not USA); controls type (trauma-exposed controls (TC) or non-trauma-exposed controls (NTC)); and assayed methods (radioimmunoassay (RIA) or other). Besides, the fact that gender-specific 24-h urinary cortisol concentration data were not provided by the eligible studies, an eligible study with over 50% proportion of women in the total sample was defined as examining female subjects. In this way, subgroup analysis in relation to gender was performed. In addition, sensitivity analysis was performed to prove the stability of the results. Also, potential publication bias was assessed using the symmetry of a funnel plot, whose interpretation was verified by the Egger's linear regression test [35].

Literature search and eligible studies
The search of literature yielded 619 relevant articles from Web of Science (373), PubMed (56), Embase (72), and PsycARTICLES (118). After that, 58 duplicates were deleted, leaving a total of 561 relevant articles. The abstracts of these articles were then reviewed to assess their eligibility. Following this assessment, 473 articles did not meet the inclusion criteria, hence they were excluded. Furthermore, full texts of the 88 articles were reviewed and this resulted in the exclusion of 68 articles. In the end, 20 eligible articles met the inclusion criteria and were included in the final meta-analysis (Fig 1). Table 1 presents study characteristics of the 20 eligible studies. Most articles reported controls type, 24-h urinary cortisol collection time, 24-h urinary cortisol collection and assay methods, characteristics of PTSD participants (trauma type, age (mean, SD), BMI, and gender distribution), inter-assay variation, 24-h urinary cortisol intra-assay variation, sensitivity and storage temperature. Three studies were judged to be of high-quality (total score �7), 15 of moderate quality, and 2 of low quality (Table 1).     PTSD, post traumatic stress disorder; SMD, standardised mean difference; DF, degrees of freedom; TC, traumaexposed controls; NTC, non-trauma-exposed controls; RIA, radioimmunoassay; USA, United States of America.

24-h urinary cortisol overall comparison
https://doi.org/10.1371/journal.pone.0227560.t002 Additionally, subgroup analysis with respect to gender indicated that females with PTSD had lower concentrations of 24-h urinary cortisol than female controls (SMD = -0.58 95%CI: [-1.09; -0.07], p = 0.026), but no such difference was observed between males with PTSD and male controls. Besides, gender significantly explained a large amount of the overall heterogeneity, but still with residual heterogeneity (I 2 = 11.3%). Furthermore, subgroup analysis by controls type showed significant lower concentrations of 24-h urinary cortisol in patients with PTSD than in the trauma exposed controls (SMD = -1.10, 95% CI: [-1.77; -0.43], p = 0.001), while there was no significant difference between PTSD patients and the non-trauma-exposed controls. In the subgroup analysis according to assay methods, the concentrations of 24-h urinary cortisol were significantly lower in patients with PTSD than in the controls when radioimmunoassay (RIA) was used (SMD = -0.51, 95%CI [-0.97; -0.04], p = 0.032), while no difference was found when other assay methods were used.

Sensitivity and bias analysis
There was little change in the SMD and corresponding 95% CI when studies were excluded one at a time, indicating low sensitivity of this meta-analysis. Also, no asymmetry was observed in the shape of the Egger's funnel plot, and the p value (0.31) of the Egger's test was not significant (Fig 3), implying small chance of publication bias [35].

Discussion
According to the best of our knowledge, this is the first meta-analysis comparing the 24-h urinary cortisol levels data between PTSD patients and the controls. In order to obtain relatively larger sample size, we expanded the scope of article searching in the online electronic databases, which brings 20 eligible studies with a total of 543 participants with PTSD and 738 controls. In the overall study sample, concentrations of 24-h urinary cortisol were lower in patients with PTSD than in controls. Therefore, the conclusions from this study are more comparative and persuasive compared with the previous studies. Besides, different from the preceding studies, this study only concerns with 24-h urinary cortisol concentration levels when studying its connection with PTSD. In addition to data on the differences in the 24-h urinary cortisol levels between PTSD patients and the controls, this study also analyzed almost all factors that may affect 24-hour urinary cortisol levels, including subgroup analyses. In particular, a lower 24-h urinary cortisol levels was found in PTSD patients as compared to the controls. However, the heterogeneity was quite large; suggesting that there may be differences based on some certain subgroups. It is worth noting that our study had significant heterogeneity among the studies, which is not surprising considering the differences between research variations, such as participant characteristics. The previous meta-analyses only included the 24-h urinary cortisol levels data in subgroup analysis, but no significant differences were found between PTSD patients and the controls because of the limited sample size of the studies used [29].
Although previous studies had obtained consistent findings that plasma cortisol levels in PTSD patients was lower than that of the controls [36], they did not find significant difference in the 24-h urinary cortisol levels between PTSD patients and the controls [37]. In addition, these studies did not take into account the sample source when measuring cortisol concentration levels, such as plasma/serum, saliva or urine samples. Yet, it is generally known that the level of cortisol concentration between sufficient samples varies with different sample sources. Therefore, without taking the preceding factors into account, results of a meta-analysis on the critical relationship between cortisol concentration levels and PTSD could be compromised. Furthermore, subgroup analysis indicated that the 24-h urinary cortisol levels were lower in the PTSD patients than in the controls for studies conducted in the United States of America, whereas no difference was found in the studies conducted elsewhere. In general, females are twice as likely as males to develop PTSD; perhaps because females appear to have a more sensitized HPA axis with lower overall plasma cortisol as males, which are especially of interest with respect to cortisol, which interacts with sex hormones [38].
Accordingly, the findings of this study suggested that females with PTSD had lower levels of 24-h urinary cortisol than female controls, whereas there was no significant difference of the same between males with PTSD and male controls. However, up to now, little is known about the biological mechanisms behind the preceding association between the 24-h urinary cortisol levels and PTSD in females. In spite of that, this association could explain why women are more vulnerable than men to the development of post-trauma symptoms, and take longer than men to recover from them [39]. Also, this study found that the 24-h urinary cortisol levels were lower in the PTSD patients than in the controls in studies that used trauma-exposed controls, while no significant group difference of the same was found for studies that used nontrauma-exposed controls.
When 24-h urinary cortisol collection method [40], and storage methods of 24-h urinary cortisol has be unified [41]. The 24-h urinary cortisol could be a quick biomarker assay to assist in screening a possibility for screening a lot of people with PTSD [10,42]. Furthermore, metaregression analysis results suggested that trauma type may be a source of heterogeneity, implying that levels of the 24-h urinary cortisol in PTSD caused by combat, for example, are different from levels of the same in PTSD caused by other trauma types. Nevertheless, the eligible studies did not provide information regarding severity of PTSD; hence this was not taken into account in this study. Therefore, the preceding result should be interpreted with caution. As regards studies using DSM-IV method to diagnose PTSD, the 24-h urinary cortisol levels were significantly different between PTSD patients and controls, while no group difference of the same was found when other diagnostic methods were used. This observation supports the importance of using valid diagnostic tools for PTSD assessment [43].
Also, age in this study explained some of the heterogeneity. This is consistent with results of a previous study, suggesting that the increased risk for developing PTSD is associated with lower concentrations of cortisol in both adults and children [44]. That is, during childhood, trauma exposure may have greater potential of devastating HPA, disrupting brain maturation and affecting the development of the frontal cortex, leading to long-term changes in the HPA response [45]. This may be related to prefrontal cortex development, corpus callosum myelination, and synaptic elimination in the developmental traumatology models. In addition, the low activity of the HPA in adults indicates a greater cumulative lifetime of trauma exposure, adaptation of the HPA axis, and the risk of developing PTSD [44].

Limitations and strengths
However, this study has several limitations. First, due to the limited sample size, the effects of race, age and stressor patterns were not examined. Second, we were unable to use all studies in the subgroup analyses because not all eligible studies reported the subgroups of interest. Third, some studies did not report the use of urine standardization to control variability associated with urine dilution. Lastly, we included only articles published in English and excluded grey literatures. Therefore, this meta-analysis is prone to selection bias.

Conclusions
Despite the preceding limitations, the results of this meta-analysis provide convincing evidence that lower concentrations of 24-h urinary cortisol may be associated with PTSD. However, it is worth noting that numerous often-overlooked factors may have a confounding influence on the concentrations of 24-h urinary cortisol. Therefore, future studies should elucidate whether low 24-h urinary cortisol is related to trauma type, age, or PTSD assessment methods.
Supporting information S1