Plasma oxytocin changes and anti-obsessive response during serotonin reuptake inhibitor treatment: a placebo controlled study

We wanted to explore whether pre-treatment plasma oxytocin is related to OCD severity and other clinical features in adult patients with OCD, as compared to previous studies of CSF oxytocin. Furthermore, we planned to investigate in a placebo controlled trial whether SRI treatment in humans are linked to changes of plasma oxytocin and, if so, the direction and magnitude of these changes. Finally, we aimed at testing the hypothesis that oxytocin changes correlate with and possibly predict anti-obsessive response, by following the temporal pattern of plasma oxytocin during the first four weeks of SRI treatment.

In a multicenter drug trial, carried out 1992–1993, comparing paroxetine with clomipramine and placebo for the treatment of OCD [56], our center included some biochemical measurements in addition to the blood samples for safety measures. The blood samples were taken at baseline (before any drug treatment), after 1 week and after 4 weeks of drug treatment. This schedule was chosen in order to detect early changes of biochemical measures that could possibly predict the clinical response after 12 week’s treatment. The blood samples were analyzed in 1995.

Details on the inclusion of 36 patients in this biochemical extension of the drug trial have been published elsewhere [57]. Briefly, patients could be included if they fulfilled DSM-III-R criteria for OCD (the DSM-III-R and DSM–IV criteria for OCD are widely viewed as interchangeable [58]) with at least 6 months’ duration and consented to the blood sampling. Customary exclusion criteria were applied. At the time of inclusion, no patient had been taking antidepressant agents of any kind during the preceding three months, and two thirds (n = 24) were SRI treatment naïve. Females in reproductive age were ascertained not to be pregnant and informed to use effective contraceptive methods, if applicable. For demographic data, see Table 1.

The study was approved by the Research Ethical Committee of the Karolinska Institutet, Stockholm, Sweden, and informed consent was obtained from all participants.

After randomization (2:1:1 ratio), double-blind drug treatment was given for 12 weeks with increasing, flexible doses of either paroxetine 20–60 mg/day (n = 18), clomipramine 50–250 mg/day (n = 9) or placebo (n = 9). Zopiclone 7.5 mg h.s. for insomnia was permitted, if necessary, but cognitive or behavioral psychotherapy was not allowed during the study. For clinical results, see Table 1 and [57].

Clinical data including OCD history, severity and subtype were recorded in a standardized, semi-structured way. The following rating instruments were used: Y-BOCS [59], in order to quantify OCD symptom severity and to evaluate treatment response; National Institute of Mental Health Global Obsessive Compulsive Scale (NIMH-GOCS) [60], a complementary method to measure global severity of OCD; Montgomery Åsberg Depression Rating Scale (MADRS) [61], a clinician rated instrument for measuring depressive symptoms, and the Patients’ Global Evaluation (PGE), a self-rating version of the 7 graded Clinical Global Impression-Improvement scale [62]. Poor insight was evaluated clinically, and autistic traits were assessed with assistance of the High-functioning Autism/Asperger syndrome Global Scale (HAGS) [33]. It covers functional impairment, social and emotional reciprocity, social competence, interests, rigidity, values, self-reflection, speech and language, body posture, gestures, facial expression, and eye contact. The rating consists of four different levels: 1 = an exceptionally empathic and socially competent personality; 2 = more or less normal, “like most people”; 3 = an emotionally blunt and pathological personality with autistic traits, clearly noticeable during the interview; and 4 = an extremely odd personality; the person gives a peculiar, and clearly autistic, impression early in the interview. A HAGS score of 3 or more was considered as sign of autistic traits. Eleven patients were assessed as having autistic traits. To our knowledge, at least five of these patients were later formally diagnosed with ASD.

Treatment response was defined as at least 25% decrease of scores on the Y-BOCS in conjunction with a rating of 1 or 2 (“very much improved” or “much improved”, respectively) on the PGE at endpoint. For premature discontinuers, the last observations were carried forward. PGE, rather than the clinician’s rating of global improvement was chosen in order to also take the patient’s opinion into account. Applying these criteria, 17 of the 36 patients (59% of SRI treated patients) responded to treatment.

Blood samples were obtained by cubital venepuncture, between 8 h00 and 9 h00 a.m., when patients had been fasting from midnight and before the morning dose of medication was taken. The sampling was performed by one of two nurses, known to the patients, on each occasion in the same quiet room, at normal room temperature and under comfortable circumstances. Samples were taken at baseline, after 1 week’s double-blind treatment and after 4 weeks of treatment. The samples were collected in tubes containing heparin (10 IU/mL) and Trasylol (500 IU/mL) and centrifuged. Plasma was separated, frozen at -70°C, and blindly analyzed in the same assay in 1995. The concentration of oxytocin was measured with a specific radioimmunoassay (RIA) described by Stock and Uvnäs-Moberg [63]. Briefly, plasma samples were extracted on SEP-PAK C₁₈ cartridges prior to assay. The recovery of this extraction procedure was 95.3 ± 10.1%. For the assay, antiserum K19 (Milab, Malmö, Sweden) was used, which has a cross-reactivity at 70% relative binding (B/BQ) of 0.01% with arginine(A)-vasopressin, <0.01% with lysine(L)-vasopressin and 0.1% with A-vasotocin. The limit of detection is 2 fmol/ml and the intra- and inter-assay coefficients of variation are 11.2 and 13.0% respectively.

Based on previous literature, no prediction of the expected direction of findings was possible. The patients treated with paroxetine and clomipramine were initially analyzed separately, but when they were compared on relevant parameters no meaningful differences between these two groups were found. As the clomipramine cases, when adjoined, did not change the results of the paroxetine cases, these two treatment groups were merged to the SRI group, having the same putative mechanism of action [13, 14]. Since most oxytocin measurements were non-normally distributed, all oxytocin values are reported as medians with 1^st and 3^rd quartiles in parentheses, and nonparametric tests (Mann–Whitney U-test (MW) and Spearman Rank Order Correlation) were utilized. For other group comparisons one-way ANOVA or Chi-2 statistics were used. ANOVA of repeated measures of the oxytocin changes turned out invalid due to the non-normal distributions. This and low sample size prevented multivariate methods to be used. Thus, nonparametric tests of the measures and of the differences between the repeated measures are reported. Also, we calculated the intra-individual range (the difference between the maximal and minimal plasma oxytocin levels among each patient’s three (n = 32) or two (n = 4) samples over the 4 weeks), intended to be a measure of flexibility or responsivity of the oxytocinergic system. The data on age of onset were categorized into tertiles, which corresponded to childhood, adolescence and adult onset, respectively. Statistica 64, version 10, StatSoft Inc. was used. Probabilities < 0.05 were assumed as significant and, when relevant, Bonferroni’s adjustment for multiple comparisons was judiciously implemented. However, due to the explorative nature of this study, also the non-adjusted results are presented.

Plasma oxytocin at baseline was non-normally distributed and had a median of 31.3 (22.7, 39.5) pg/ml and a mean of 33.4 (±13.5). The distribution appeared bi- or trimodal, with a distinct high mode (50–67 pg/ml) and two less distinct lower modes (16–25 and 27–41 pg/ml, respectively); see Table 2 and Figure 1. Oxytocin at baseline was positively related to baseline Y-BOCS scores in the total group (Spearman’s rho = 0.35, n = 36, p = 0.037). The future SRI responders accounted for this correlation (rho = 0.58, n = 16, p = 0.019), that was not retrieved among non-responders (rho = 0.17, n = 19, p = 0.48). On the other hand, baseline oxytocin was unrelated to the future response to treatment, to depression scores (MADRS), sex, age, duration of OCD, history of tics, poor insight, autistic traits and family history of OCD. The age at OCD onset, however, was negatively correlated to baseline oxytocin (rho = -0.36, n = 36, p = 0.030). Median plasma oxytocin of the 12 cases with childhood onset (before 11 years) was higher, 35.4 (31.7, 47.3) pg/ml compared to 21.3 (19.0, 35.6) among the 11 patients with adult onset (from 18 years), MW Z = 2.12, p = 0.034. Baseline plasma oxytocin was also higher among those completing the 12 weeks’ trial, 34.7 (25.8, 40.0) compared to trial discontinuers, 24.2 (20.5, 30.2), MWZ = 2.06, p = 0.039.

The distinct higher mode of oxytocin (50 – 67 pg/ml) contained 6 individuals, that did not differ from the lower modes, separately or merged (16 – 41 pg/ml) pertaining to the above mentioned clinical variables; however 5 of the 6 were cohabiting, a significantly higher rate compared to the lower modes (χ ² = 4.41, d.f. = 1, p = 0.036) (Table 2).

Out of the initial 36 patients, oxytocin samples were obtained from 35 patients after 1 week and from 33 after 4 weeks. Plasma oxytocin medians in the total group after 1 and 4 weeks were 31.2 (22.8, 43.8) and 36.7 (27.0, 43.6), respectively. Median oxytocin plasma levels at the three time points for SRI treated versus placebo treated patients, respectively, were: baseline 32.1 (22.7, 39.5) vs 31.2 (22.6, 39.5), after 1 week 30.4 (22.5, 40.5) vs 36.1 (25.1, 47.3), after 4 weeks 36.7 (27.3, 43.8) vs 37.0 (26.1, 43.6), and intra-individual range 11.6 (5.1, 16.8) vs 7.4 (5.8, 12.4). None of these measures or the differences between time point measures differed significantly between treatment groups. Also, when the clomipramine group and the paroxetine group were analyzed separately, no significant differences between them or between them and the placebo group were detected on the different oxytocin measures. According to plasma drug levels at week 4, all SRI patients but one non-responder seemed to comply with treatment.

While the baseline and week 1 samples did not differ between responders and non-responders, the 17 treatment responders (including 1 placebo responder) had higher oxytocin at week 4, than the 16 non-responders (MW Z = 2.31, p = 0.021, missing data = 3). This difference remained if only SRI-treated subjects were included in the analysis (16 responders and 8 non responders) (MW Z = 2.14, p = 0.032, missing data = 3).

The individuals’ changes of plasma oxytocin between the three time points were analyzed and compared between response groups (Table 3). A significant association appeared between final treatment response and the difference of plasma levels between week 1 and week 4, showing an increase of oxytocin among treatment responders and a decrease among non-responders (8.5 (-2.9, 17.5) and -3.1 (-7.5, 2.8), respectively, MW Z = 2.24, p = 0.025). The sole placebo responder had an increase of oxytocin within the highest quartile between baseline and week 4. The intra-individual plasma oxytocin range significantly differentiated responders from non-responders in the total group (median 24.2 (15.7, 37.5) vs 8.9 (4.9, 12.7) pg/ml (MW Z = 3.61, p = 0.0003). Furthermore, this oxytocin range also differed between those with autistic traits (n = 11) and those without (5.1 (3.5, 9.7) vs 15.6 (7.6, 27.3) MW Z = 2.76, p =0.006, missing data = 2). With Bonferroni adjustment, the two latter findings remained significant.

There was a strong negative correlation between both Y-BOCS and NIMH-GOCS% decrease at endpoint and plasma oxytocin change between baseline and first week of SRI treatment; i.e. those with the largest initial increase of plasma oxytocin had the poorest outcome. Reversely, a strong positive correlation was found between oxytocin change between week 1 and 4 and overall improvement in OCD symptoms. None of these measures correlated with improvement of depressive symptoms. The oxytocin range correlated strongly with improvement in OCD and to a lesser extent also with amelioration of depression (Table 4 and Figure 2).

There were no significant correlations between the previously reported [57] measures of serotonin in whole blood from the same patients and the presently reported oxytocin measures. Specifically, neither the baseline serotonin level nor the decrease of serotonin between baseline and week 1 (that correlated with clinical improvement) showed any connection with the oxytocin measures (all Spearman rho values below 0.10), indicating that these measures are independent.

In the present study it was not possible to identify intra-cerebral events; thus the nature of the implied central nervous counterparts giving rise to and mediating our plasma findings is an unresolved issue. Almost all plasma oxytocin is released from the magnocellular neurons in the SON and PVN via the posterior pituitary. The functional links between the neurohypophyseal release and the intra-cerebral circuits operating with oxytocin are as yet poorly understood; but even if they could be regulated separately [20, 23, 26], they are most likely to interact. In a study were both CSF and plasma oxytocin levels were analyzed in the same individuals [64], central and peripheral oxytocin measures did not correlate with each other, while both correlated with some measures of suicidality. In various experiments, elevated oxytocin has been linked to relaxed, affiliative situations, implying anxiolytic and antidepressant effects [47, 65], but in other experiments oxytocin is increased in relation to stress [26, 66]. These disparate findings indicate that different segments of the central oxytocin system may act in different directions. The difficulty of predicting effects within the oxytocinergic system is further underscored by recent studies were intranasal oxytocin induced lowered mood in women with postnatal depression [67], and increased agonistic behaviors with dysregulated HPA axis in piglets [68], respectively, in both cases contrary to expectation.

In 1992, when the patients in this study were enrolled, OCD was regarded as a rare disorder and many psychiatrists in Sweden were not aware that they had treated any OCD patients in their practice. Therefore, most of the patients were recruited through advertisement in the local paper, and only a handful was clinically referred. The patients were nevertheless ill, and although they were mostly in their early forties, they had a mean duration of OCD of 26 years and one third was defined as depressed. Just about half the sample was able to work full time and the majority lived as singles. Most of these patients had not conveyed their symptoms prior to this study, possibly due to shame. Now, they were assigned a well-informed and devoted psychiatrist that fully understood their symptoms and provided hope for improvement. Such circumstances are often put forward in discussions of problematic placebo response in randomized controlled trials. However, in our study, the rate of placebo response was very low, in accordance with the clomipramine studies of the early days [69]. In spite of this, the social bonding and affiliative aspects of the clinical setting could hypothetically have enhanced oxytocin release and thus influenced the results, resulting in the increase of plasma oxytocin that we saw the first week in both placebo treated patients and SRI non-responders. However, in the SRI responders, the decrease of plasma oxytocin at week 1 may correspond to a response-related serotonin-oxytocin interaction that surmounted this affiliative effect. In this perspective, it may also be of interest that the patients that completed 12 weeks of double-blind treatment had higher baseline oxytocin than those that did not. Nonetheless, the substantial increase of plasma oxytocin in our only placebo responder suggests that oxytocin may be relevant for future studies on placebo response. Incidentally, the relevance of our oxytocin measurements for affiliative aspects was corroborated by the higher number of married or cohabiting individuals in the group with highest baseline oxytocin.

According to El Mansari and Blier [14], the neurophysiological change, responsible for improvement in SRI treatment of OCD, is that orbitofrontal pre-synaptic 5-HT_1D receptors are down-regulated by long term (8 weeks) treatment with SRIs. This will lead to an increased transmission over 5-HT₂ receptors, eventually leading to decreased activity in the “OCD loop”, consisting of orbitofrontal cortex, the head of the caudate nucleus, a direct and an indirect pathway through the basal ganglia, the thalamus, and back to orbitofrontal cortex. It remains a possibility that our changes of oxytocin plasma levels only represent peripheral reflections of such serotonergic events without any functional importance. However, the baseline correlation with OCD severity in our and a previous study [41] supports the view that the oxytocinergic system is not merely a bystander.

An alternative hypothesis would be that down-regulation of post-synaptic 5-HT_1A receptors, which is known to take place in the hypothalamus during SRI treatment [51], will result in decreased oxytocinergic transmission in relevant parts of the forebrain. The credibility of this hypothesis, though, depends on conceivable links between the oxytocinergic system and the above mentioned, well documented OCD neural circuit. Even if no evidence was found for frontal cortex oxytocin receptors in an autoradiography study of humans [70], more recently and with more advanced technology oxytocinergic fibers of medium density were identified in the medial orbital and the frontal association cortices of rats [23]. Moreover, in a recent study, intranasal oxytocin challenge in humans caused an increased activity of the caudate nucleus (an essential part of the OCD neural circuit), and a significant correlation between plasma oxytocin and caudate activity was reported [71]. Also, nucleus accumbens has been implicated in OCD [72] and receives oxytocinergic innervation from the PVN. It has even been shown that oxytocin and serotonin interact closely in the nucleus accumbens, related to social reward [73]. Accordingly, it is not inconceivable that overactive oxytocinergic neurons of hypothalamic origin contribute to OCD severity by increasing striatal and orbitofrontal engagement. In such case, a SRI-induced down-regulation of hypothalamic 5HT_1A receptors [51] may modulate this oxytocinergic over-activity, thereby eventually contributing to an anti-obsessive response. Since the magnocellular neurons in the PVN transmit oxytocin both by axons projecting to amygdala and nucleus accumbens and by hormonal release into the peripheral circulation [22, 23], changed regulation of their central activity may well be reflected in plasma oxytocin levels. However, there is a lack of detailed knowledge on the regulation of these separate activities, especially concerning the effects of SRIs. Admittedly, it is reasonable that the entire oxytocinergic system change synchronized under the influence of SRI treatment, but the possibility remains that different parts of the system react differently to serotonergic changes. In such case, the already discussed, seemingly contradictory findings of OCD-oxytocin relationships in the present and a previous study [53] may find an explanation. However, until future studies have shed more light on the regional serotonergic regulation of oxytocin transmission, and the effects of psychopharmacological manipulations, this remains conjecture.

On the other hand, increased oxytocin activity has been linked to anxiolytic effects, exerted e.g. in the amygdala [23] or the median raphe nucleus [45], effects that may also be involved in the reduction of OCD symptoms. Then, the oxytocin decrease in responders after 1 week’s treatment could hypothetically be linked to an increase in anxiety. Interestingly, when starting SRI treatment in panic disorder, an initial paradoxical increase of anxiety is commonly observed [74, 75], however, this is less commonly reported in OCD treatment. Since specific anxiety ratings were not included in the present study, we do not know whether the initial decreases of oxytocin correspond to increases of anxiety. If this were the case, however, oxytocin deficit may contribute to an explanation of this intriguing phenomenon.

Further relevance for connections between serotonin and oxytocin in the human brain has been demonstrated by mechanistic studies of 3,4-methylenedioxymethamphetamine (MDMA or “Ecstasy”). It has been shown in humans that one of the acute effects of MDMA intake is elevation of plasma oxytocin together with pro-social effects [e.g. [76]. Since MDMA purportedly exerts its effects through the serotonin transporter and SRI pretreatment blocked the oxytocin elevation, the authors suggest a primary role for serotonin in the effects of MDMA on oxytocin release. According to Hunt et al. [77], MDMA-induced increase of oxytocin depends on 5-HT_1A transmission and takes place in the SON and PVN. Interestingly, two cases where MDMA use was related to de novo onset of OCD have been reported [78], seemingly consistent with our hypothesis.

One further link between OCD psychopharmacology and oxytocin is provided by the effects of antipsychotic drugs. In OCD, resistant to SRI treatment, the best documented treatment option is to add an antipsychotic; both haloperidol and risperidone have strong short-term data, while those of olanzapine and quetiapine are mixed [17]. Conversely, in patients with schizophrenia, obsessive-compulsive symptoms may emerge related to antipsychotic use, the risk seemingly higher with clozapine and olanzapine than with haloperidol and risperidone [79, 80]. The effects of these antipsychotics on the oxytocin system have been investigated [81, 82], showing most markedly increased release of oxytocin and activation of oxytocinergic neurons by clozapine, closely followed by olanzapine, while the effects of risperidone and haloperidol were much less pronounced or absent. Accordingly, if oxytocin contributes to obsession and compulsion severity, this may explain the differential effects of antipsychotics as SRI augmentation in OCD treatment as well as de novo OCD provocation among patients with schizophrenia.

Thus, we suggest that the present study supports the idea that oxytocin is involved in OCD, but based on our data we cannot conclude on the preferred direction of oxytocin changes during OCD treatment. However, gleanings from other research shift the balance in the direction of an increased activity of some part of the central oxytocin system in OCD, as previously proposed [18, 19, 39], and that this activity is moderated by SRI treatment.

This study includes a small sample of patients and it was carried out in the early nineties. On the other hand, considering how widely spread SSRI medication is today, it would be a challenge to obtain a group of mainly drug naïve, chronically ill patients such as those included in the present study. The small sample size is to some extent compensated by the low placebo response, as is validated by the significantly higher response rate with SRI compared to placebo treatment. Also, since the two active treatments did not differ on any relevant measures, they were merged in order to increase statistical power. Because the placebo group remained problematically small, the comparisons with placebo should be seen as tentative; however, the response categories within the SRI group have a more reasonable size.

Oxytocin in plasma was measured at three time points; however, in four patients it was only measured twice due to patient related factors. All samples were obtained during the first 4 weeks of SRI treatment; it would have been of interest to measure oxytocin also after 12 weeks in order to follow further changes. However, our measurement schedule was based on the presumption that the biochemical changes appearing during the first phase of treatment are decisive for treatment result; furthermore the risk of drop-outs increases with the length of the study. Oxytocin was only measured in plasma; it would have been preferable to also include CSF levels, but most patients with OCD are likely to refuse spinal tap due to extensive worries regarding its consequences. Oxytocin is released in a pulsatile manner and may also vary in relation to the menstrual cycle and use of oral contraceptives. These factors were not accounted for in the present study, but it seems unlikely that the correlations to response would have appeared as a spurious result of this omission. On the other hand, the RIA method we used for oxytocin analysis, including plasma extraction, belongs to the most reliable types of oxytocin analysis [83]. The importance of plasma extraction for the validity of these analyses has recently been further emphasized [84]. Furthermore, a link between our oxytocin measurements and mental functions is to some extent validated by the association between oxytocin levels and cohabitation status, in line with previous research; see e.g. [85, 86].

Recent findings suggest specific interactions between genetic polymorphisms of the serotonin transporter and oxytocin receptor genes [46], which would have been interesting to explore in our patient group. However, at the time of our study such genotyping was not available.

The lack of a relationship between depressive symptoms and oxytocin measures may be due to our patients’ low but variable levels of depressive symptoms, thus representing a type II error. Finally, we did not use specific rating scales for tics, and autistic traits were only measured with a global scale, HAGS. Again, at the time of the study, other instruments for assessing ASD in adults with normal intellectual ability were not developed.