Many patients came to our clinic with the suspicion that they themselves were depressed, as the website announced that the clinic was recruiting subjects for a biomarker study of PEA levels. The period during which the study was conducted, as specified by the ethics committee, i.e., from 11 June 2011 to 31 March 2020, was divided into three periods, as shown in Figure 1. Period 1, as described below, was devoted to the preparation of the correct measurements of plasma PEA concentrations and the collection of informed consent.
Period 1
The Gyouki-Kai Medical Corporation (GMC) first needed to train the newly assembled staff at the newly established Kawamura Clinic for General Practice to establish a system to obtain informed consent and to prepare/mail plasma samples for accurate measurements of plasma PEA concentrations. The following steps were taken to obtain the best PEA measurements: i) avoid damaging blood vessels by inserting the needle into the same vessel more than once, ii) avoid inserting the needle into vessels that are too thin, iii) ensure that syringe suction is sufficiently low during blood collection, iv) centrifuge samples within 8 hours of collection, v) avoid using faulty ethylenediaminetetraacetic acid (EDTA) blood collection tubes, v) ensure that tubes are sufficiently mixed to prevent the plasma from clotting after centrifugation and to produce plasma with a uniform PEA concentration, and vi) observe the plasma precisely to check for haemolysis and send samples without haemolysis. If haemolysis occurred, the internal PEA from the blood cells dissolved into the plasma, artificially elevating the PEA values in the plasma.
In addition, we found that the time needed to fully observe episodes of MDD was approximately 700 days through observing patients who came to the clinic during Period 1. Therefore, we decided to perform an observational study with the utmost respect for the subjects’ free will during Period 2, with newly recruited patients. No further patients who were recruited in Period 3 were included in this study.
Accuracy of plasma PEA measurements
The methods used to accurately measure plasma PEA concentrations were examined, and potential limitations were considered. In particular, there were numerous cases where the PEA value was incorrect, regardless of how many times the measurement was repeated, because, in rare cases, the ion chromatography–fluorescence detector (IC-FLD) measurement produced another peak that subtly overlapped with the peak in the PEA region; thus, only a reference value could be detected29. Though there were also cases where another peak appeared during Period 2, we did not include these data in this study.
Number of Subjects
The number of study participants is shown in Figure 1. Specifically, in Period 2, 3,617 patients visited the clinic; of these, 2,561 underwent a PEA measurement, and 1,056 did not. A total of 1,320 patients underwent only one PEA measurement.
A total of 1,241 patients indicated that they were willing to provide more than one PEA measurement, but 145 of these patients could not participate in the study because they were unable or unwilling to provide the research staff with information on their medical history regarding mental and physical illness or the chronological changes in the medications they had taken or were taking at the time of their first visit. Forty-six patients were unable to participate in the study because they were unable to complete the structured clinical Interview for the DSM-5 (SCID-5). Additionally, 446 patients were unable to participate in the study because they expressed their willingness to participate in the study but did not provide written, complete informed consent. Twenty-eight patients who did not wish to undergo a PEA measurement at the first visit but did so at numerous subsequent visits were not included in the study.
Thus, there was a total of 576 study subjects. Of these, 264 patients had a disorder or illness other than a depressive disorder (depressive disorders in Chapter 4 of the DSM-5), and 17 patients had only a physical illness.
The above 264 patients were further classified according to their DSM-5 diagnoses: 13 patients had neurodevelopmental disorders, 39 patients had schizophrenia spectrum and other psychotic disorders, 59 patients had bipolar and related disorders, 26 patients had anxiety disorders, 4 patients had obsessive-compulsive and related disorders, 90 patients had trauma and stressor-related disorders, one patient had a dissociative disorder, 4 patients had somatic symptoms and related disorders, one patient had an eating disorder, 14 patients had sleep-wake disorders, one patient had gender dysphoria, 11 patients had alcohol and addictive disorders and one patient had a personality disorder at Table 1, where the mean and standard deviations of plasma PEA concentrations are shown. The PEA values in Schizophrenia, alcoholics were apparently as lower as those of MDD.
There were 295 patients with depressive disorders (Chapter 4 of the DSM-5). Further details are shown in Table 2. Of these patients, 170 had MDD only, and 125 had MDD and comorbidities. Among MDD patients with comorbidities, 29 had attention-deficit/hyperactivity disorder (ADHD), 15 had autism spectrum disorder (ASD) and 4 had both ADHD and ASD. Other than these developmental disorders, among the MDD patients with comorbidities, one had dementia, three had generalised anxiety disorder, 11 had obsessive-compulsive disorder, 12 had panic disorder, 24 had a personality disorder, and five had a physical illness. Moreover, five patients had PTSD (not MDD), and 16 patients had persistent depressive disorder.
SCID-5
We tried to administer the SCID to 1,241 patients who were willing to have their PEA levels measured at the time that informed consent was obtained. Of the patients who provided complete informed consent, SCID data were only formally recorded for patients diagnosed with only MDD. The 125 depressed patients with other comorbidities participated in the SCID, but these data were not included in the analysis due to the difficulties in assessing symptoms. Specifically, in MDD patients with comorbid psychiatric disorders, remission of MDD could not be clearly determined and therefore is not discussed in this paper.
Temporal changes in plasma PEA concentrations and symptoms
Treatment of depression was performed according to Japanese guidelines for the treatment of depression. The overview of the clinical course of the 170 patients with MDD only were pictured in Fig.2. Of the 170 patients with only MDD, 102 were already taking psychiatric medication, 120 had been re-diagnosed with MDD using the SCID at the time of the initial medical interviews at different hospitals or clinics before visiting GMC, 120 had MDD at the time of their first visit to the clinic, and 50 had already experienced partial remission of MDD. More detailed demographic data on the 170 participants with MDD only are presented in terms of changes in PEA values over the course of depression treatment.
Of these 170 participants, three refused pharmacotherapies due to fear of medication. Instead, these participants were provided with psychoeducation on the relationship between MDD and daily lifestyle habits and how to improve MDD; these participants were instructed on how to restore health through engaging in walking exercise. Psychotherapy was also provided by psychotherapists. Surprisingly, all three participants who refused medication achieved remission.
The remaining 167 participants were mainly treated with pharmacotherapy. As a rule, once a decision was made to administer medication, a clinical interview was conducted every eight weeks to assess the drug’s effectiveness, and the drug was changed as needed.
During the treatment, when it was judged that further improvement in depression would be difficult to achieve, the outcome was determined for each patient. Treatment resulted in no change (4 patients), partial remission (75 patients) or complete remission (91 patients).
During treatment, 55 patients left the study due to relocation, family circumstances or personal wishes, while eight patients offered to transfer to another hospital and continue with the study (doctors wrote referral letters), and 47 patients left the clinic without permission from the doctors. Thus, during the period from remission to recovery (the end of the MDD episode present at the first visit), 71 of the 91 patients who achieved complete remission were able to complete drug treatment. This included three patients who achieved remission without pharmacotherapy. Of these 91 patients, 21 did not follow the depression treatment guidelines, and some discontinued treatment on their own or moved to another hospital (thus, their outcome was unknown). Six patients experienced recurrence after recovery. Nine patients experienced relapsed before achieving complete or partial remission.
Patients were classified according to clinical information as 4 patients in no change (NC), 75 patients in partial remission (PR) and 91 patients in complete remission (CR), and the number of days required for the clinical course of each is shown in Table 3. “No change” (NC) refers to people whose diagnostic criteria for depression remained unchanged from their initial diagnosis. “Partial remission” (PR) refers to a state in which more items that do not meet depression diagnostic criteria are no longer clearly depressive and symptoms are reduced, but some items that still meet diagnostic criteria remain. Some left the study/treatment or continued in that state. “Complete remission” (CR) was determined by examining whether remission was present at the time of the interview and at two time points: the start of remission and whether remission persisted one month afterwards. All three clinical outcomes were not significantly influenced by specifiers of MDD suggested by χ2 test (Pearson’s χ2 value=9.90, df=10, p=0.346).
Questionnaires were completed after remission had been reached and within one month of blood collection. Blood collection for PEA level measurement and the SCID were performed free of charge, but the questionnaires involved a fee, so a few questionnaires were missing because busy patients went home without completing the questionnaire on the day that they were asked to do so. Scores on the State-Trait Anxiety Inventory (STAI) were distributed between 20 and 80 points, and scores on the Center for Epidemiological Studies Depression scale (CESD) were distributed between 0 and 60 points.
PEA measurement was performed at the judgement day; the time of these finalised interviews for judgement of MDD status. The main objective of this paper is to obtain statistics on changes in psychological indicators and PEA values between the initial visit and the date of this judgment.
Blood PEA levels were assessed at four time points: the first time, the second time, the date the judgement was implemented, and when the maximum value was taken. Each measurement interval was tried usually three months. Some patients hesitated drawing blood and postponed it. The maximum value could be obtained before, after or on the date of judgement implementation, depending on the individual.
PEA values were not normally distributed according to the Shapiro‒Wilk and Kolmogorov‒Smirnov tests either in all participants (both p>0.2) or in the 170 participants with MDD only (both p>0.2), so rank standardisation was performed to normalise the overall distribution by Blom’s rank standardisation and compare changes in PEA values among time points30. The results of rank standardisation were successfully significant (Both p<0.001).
Changes in PEA values and questionnaire scores are shown in Figure 3. It is more intuitive to report the real PEA values, but the statistics were performed using paired-sample t tests on the variables after rank standardisation.
Patients with no change in depression (n=4)
The four persons without a change in depression symptoms were not included in the statistical analysis, but for the purpose of clarity, we show them in the Figure 3a. All the values were represented mean (SD).
The time between the first visit and the judgement of no change was 137(63) days (4 individuals). No change was determined according to SCID and the following psychological test values: at the first visit, CESD scores of 32.5(13.3), STAI state anxiety subscale (STAI-S) scores of 52.8(8.9) and STAI trait anxiety subscale (STAI-T) scores of 66.3(8.0); and at the time of categorization, scores of 14.2(10.7), 41.8(12.5) and 44.2(12.9), respectively. In contrast to our predictions, plasma PEA levels were 1.22(0.2) µM at the first visit and apparently increased to 1.34(0.2) µM at the time of judgement, but this difference was not significant (Fig 3a). These four patients were transferred to other hospitals as they experienced no improvement in depression.
Patients in partial remission (n=71)
Regarding patients who achieved partial remission, the number of days from their first visit to the time of the final judgement was 358(345) days. Although the number of depression diagnostic criteria met according to the SCID decreased, at least one diagnostic criterion remained met at the time of judgement, and the patients reported various physical symptoms and anxiety symptoms.
Regarding the psychological test values, scores on the CESD, STAI-S, and STAI-T at the first visit were 31.9(11.5), 57.0(11.0) and 61.2(10.1), respectively; and those at the time of judgement were 37.0(9.5), 51.3(6.8) and 68.3(3.2).
Plasma PEA levels were 1.20(0.19) µM at the first visit and significantly increased to 1.50(0.29) µM at the time of judgement (Fig 3b).
Patients who achieved complete remission (n=91)
The number of days from the first visit to the time of final decision was 287(212) days. Scores on the CESD, STAI-S and STAI-T were 30.2 (11.0), 54.6(11.5) and 59.1(10.9), respectively, at the time of the first visit; these scores were 14.2(10.7), 41.8(12.5) and 44.2(12.9) at the time of the final judgement.
The PEA concentration increased significantly from 1.23(0.18) µM at the first visit to 1.75(0.28) µM at the time of judgement (Fig 3c).
Plasma PEA concentrations in MDD patients changed over the clinical course according to their mental status and increased with clinical improvement.
Significant differences in plasma PEA concentrations between patients with complete and partial remission
The PEA values of patients who later achieved complete and partial remission were not significantly different at the first visit (t=.1.14, df=164, p=0.257) However, the values at the second measurement (t=6.00, df=157, p<0.001), at the time of judgement (t=2.79, df=156, p<0.001) and at the maximum value (t=7.376, df= 143.7, p<0.001) throughout the course of treatment were all significantly higher in the complete remission group. The remission group was highly responsive, but this does not necessarily mean that they were quicker to respond to treatment.
Receiver operating characteristic (ROC) analysis
ROC analysis was performed to test the diagnostic power of PEA values at the first visit among 295 patients with depressive disorders, 264 patients with other psychiatric disorders and 17 patients with physical illnesses who came to the hospital because they suspected that they themselves had MDD, for a total of 281 patients. The results showed that the area under the curve (AUC) was 0.838 (95% CI: 0.803-0.872), with a standard error of 0.018 and an asymptotic significance probability of p<0.001, which is significant. In other words, a diagnosis of MDD or depressive disorders was highly probable when PEA levels were low.
Lifestyle habits, questionnaire scores, and clinical information correlated with plasma PEA concentrations
No significant correlation was observed between age and normalised plasma PEA concentrations (Pearson’s correlation coefficient=-0.055, p<0.185). Table 4 lists each variable correlated with normalised plasma PEA concentrations. Lifestyle habits that influenced PEA levels were duration of sleep the night before the measurement, smoking and drinking alcohol. PEA levels were significantly negatively correlated with CESD scores.
Feelings related with fear, self-injurious behaviours, suicidal thoughts, symptoms seen in obsessive-compulsive disorder, and symptoms seen in panic attack were significantly and negatively correlated with PEA values. Psychomotor retardation and agitation (symptoms seen in depressive disorders) were negatively correlated with PEA values (both p<0.1)
Interestingly, the number of depressive episodes and the length of time since the last depressive episode were unrelated, and the PEA value at the first visit was also unrelated to the number of days of subsequent visits.
Univariate analysis of variance
To analyse factors and variables that influence the plasma PEA level, according to the results of the correlation study above, with the omission of patients belonging to diagnostic categories to which fewer than four patients belong, a model was constructed that had an significant adjusted R2 of 0.398 (Type III square sum=256.0, df=66, average square=3.9, F=6.5 and p<0.001) , which includes age (Type III square sum =1.33, df=1, average square=1.33, F=5.3 and p=0.14), duration of sleep just before the first visit to GMC (Type III square sum=3.2, df=1, average square=3.2, F=5.3 and p=0.022), thought about hurting oneself (Type III square sum=1.01, df=1, average square=1.01, F=1.7 and p=0.193), fear for committing suicide (Type III square sum=2.3, df=1, average square=2.3, F=3.9 and p=0.050), smoking (Type III square sum=0.21, df=1, average square=0.21, F=0.35 and p=0.555), alcohol intake (Type III square sum=0.880, df=1, average square=0.01, F=0.023 and p=0.880), gender (Type III square sum=4.30, df=1, average square=4.30, F=7.2 and p=0.008) and the diagnosis according DSM-5 (Type III square sum=83.94, df=10, average square=8.39, F=14.0 and p<0.001). Significant statistical interactions were observed between smoking, gender and DSM-5 diagnosis (Type III square sum=10.56, df=5, average square=2.11, F=3.5 and p=0.004) and between smoking, alcohol consumption, gender and DSM-5 diagnosis (Type III square sum=3.12, df=1, average square=3.12, F=5.22 and p=0.023).
Instead, the model with only psychiatric diagnosis had a significant adjusted R2 of 0.379 (Type III square sum=224.3, df=13, average square=17.3, F=27.9 and p<0.001).
Analysis of variance
ANOVA was conducted to elucidate the disorders or illnesses that do not differ their PEA concentrations statistically at the first visit of GMC using diagnoses by DSM-5 criteria with the omission of patients belonging to diagnostic categories to which less than 10 patients belong. Significant differences were detected between diagnostic categories (between group: type III square sum=214.0, df=8 average square=26.8 F=43.4 p=0.001, within group: type III square sum=341.08, df=555 average square=0.62) and, according to a post-hoc test by Scheffe, the PEA values of patients diagnosed as depressive disorder were not significantly lower than those diagnosed as disorder with schizophrenia (p=0.137) and disorders of alcohol related and addiction (p=1.00), but significantly different from those diagnosed as the other categories (for all other comparison: p<0.003).