This is to our best knowledge the first observational prospective cohort study to determine whether delirious critically ill patients who are given drugs with serotonergic properties show physical signs suggesting central serotonergic toxicity. Since a few serotonergic drugs are frequently and often for a longer period of time administered, we hypothesize that some of these delirious patients have serotonin toxicity without physicians suspecting this. In our relatively small sample, 16 % of the delirious patients who were given drugs with serotonergic properties demonstrated physical signs of serotonin toxicity and met the Hunter serotonin toxicity criteria compatible with serotonergic toxicity. Of all neurological symptoms, an inducible clonus, with or without hyperreflexia, and especially in the lower extremities, seems to be most important indication that serotonergic toxicity may be present, along with fluctuating blood pressure in the absence of vasopressor or inotropic medication. Even though half of the patients were still delirious after 1 week and serotonergic drugs were still administered, no physician suspected serotonin toxicity as a possible cause. This may suggest that physicians are not sufficiently aware of the serotonergic effects of some frequently used drugs in the ICU. Nor do they apparently expect serotonin toxicity to occur in patients who are not on antidepressants.
We used the CAM-ICU as an instrument for detecting delirium. Although it is the most widely used instrument in critically ill patients, it can lead to false positives in case of heavily sedated patients, i.e., RASS-2 or 3. A study of Heangi et al. [13, 14] showed that excluding patients with RASS-2 and 3 leads to a reduction in positives CAM-ICU of 22 %. In our cohort, the CAM-ICU assessment has only been performed in case of a RASS of -2 of higher, i.e., when a patient is lightly sedated, thus expecting few false positives.
Very little research has been done on serotonergic toxicity in the ICU. Pedavally et al.  performed a retrospective analysis of the clinical presentation of patients with serotonin toxicity in the ICU.
They included patients who were already diagnosed with serotonin toxicity before admission to the ICU, as well as patients who developed serotonin toxicity after admission. Most patients, who developed serotonergic toxicity during ICU admission, were on antidepressant therapy and developed symptoms indicative of serotonergic toxicity after administering additional serotonergic drugs in the ICU. Patients in our cohort did not receive antidepressants, but drugs with less serotonergic potency.
Fentanyl is a synthetic piperidine opioid with serotonergic properties. Systemic injection causes efflux of 5-HT in the dorsal raphe nucleus, either through direct 5-HT1A postsynaptic stimulation or through indirect mu-opioid stimulation, which causes 5-HT reuptake inhibition [16–18]. The cytochrome P450 iso-enzyme family, especially CYP3A4, is responsible for the elimination of fentanyl. Concomitantly subscribing fluconazole or voriconazole, both CYP3A4 inhibitors, has shown to decrease the elimination of fentanyl significantly . In addition, human cytochrome P450 genes are highly polymorphic, leading to inter-individual differences in drug metabolism .
Metoclopramide, though it is primarily a dopamine antagonist, also seems to have moderate–low affinity to 5-HT2A, 5-HT3 and 5-HT4 receptors . Combined with fentanyl, it might enhance the risk of inducing serotonergic toxicity.
Although tramadol is not an analgesic frequently prescribed in the ICU, a recent case series by Ibister et al.  suggested that tramadol does not seem to provoke serotonin toxicity clinically and a study by Barann et al.  showed a concentration-dependent decrease in 5-HT reuptake, leading to an increase in serotonin.
Since this study contained a small cohort, no definite conclusions can be drawn on differences in drug use between the two groups. It is of note that the large majority of all our patients receive fentanyl at some stage. All patients with suspicion of serotonin toxicity had received fentanyl, started after admission to the ICU, and half of the patients had received two or more serotonergic drugs combined, mostly with metoclopramide. Yet no differences in the amount of days or total received dosage of fentanyl between the groups were present. However, four out of seven patients received a CYP3A4 inhibitor while receiving fentanyl, leading possibly to higher dosages. In addition, three out of four received besides a CYP3A4 inhibitor also a CYP2D6 inhibitor. Metoclopramide depends on CYP2D6 for inactivation. Combining CYP2D6 inhibitors (like haloperidol) with metoclopramide could therefore potentially be an extra risk factor for developing serotonergic toxicity, because of its moderate–low affinity to some postsynaptic 5HT receptors, when a patient has already been given other serotonergic drugs.
Some patients showed no signs of serotonin toxicity, despite having received multiple serotonergic agents. The most frequent combination of drugs in this group was fentanyl combined with ondansetron. Ondansetron is a 5-HT3 receptor antagonist. Turkel et al.  hypothesize that by blocking the 5-HT3 subtype receptor and simultaneously increasing the level of serotonin through serotonergic drugs, the excessive serotonin selectively agonizes the other serotonin subtypes, such as 5-HT1A and 5-HT2A. However, for a relevant toxicity, 5-HT levels need to be increased 40-fold over baseline . Ondansetron-like drugs might, thus, not play such a big role in provoking serotonergic toxicity. Another possible explanation, for why some patients develop serotonergic toxicity and others do not, might be a combination of factors, such as unmeasured differences in severity of illness, different intra-individual susceptibility to develop serotonergic toxicity in terms of central serotonergic neurotransmission, blood–brain barrier disturbances or cytochrome P450 polymorphisms.
Serotonin toxicity is not an easy diagnosis to make. Little research has been conducted, and most is known through case reports. No gold standard diagnostic test exists, and serotonergic toxicity is essentially a phenomenological diagnosis. Dunkley et al.  defined the Hunter serotonergic toxicity criteria, which compared to a diagnosis made by clinical toxicologist have a sensitivity and specificity of 84 and 97 %, respectively. Although the Hunter criteria have not been developed for ICU patients and their specificity in the ICU has not been tested, we relied on the Hunter serotonin toxicity criteria.
We acknowledge that ICU patients are a difficult population to study. Many exhibit autonomic symptoms due to the critically ill condition they are in, and especially critically ill patients might have various reasons for being delirious. We have to emphasize, however, that our patients were diagnosed with delirium at the moment when they were stabilized in terms of hemodynamics and respiratory failure and sedation was minimal. Yet all seven patients showed autonomic instability and hyperactivity without active infection, support of inotropic or vasopressors, or drug/alcohol intoxication. Furthermore, the patients showed neurological symptoms that otherwise rarely occur in patients. Relying on the Hunter criteria makes misclassification unlikely.
Clinically serotonin toxicity might be difficult to distinguish from malignant neuroleptic syndrome (MNS). Excessive dopaminergic blockage can lead to MNS, characterized by fever, muscular rigidity with elevated serum CK, altered mental status and autonomic dysfunction. It is a rare idiosyncratic drug reaction to (therapeutic) doses of neuroleptic agents, and although it could in theory develop anytime, symptoms usually develop within the first 2 weeks after starting a neuroleptic agent. Serotonin toxicity, to the contrary, is a spectrum of symptoms and a toxic reaction due to an excess of serotonin in the CNS, and develops quickly after administering a serotonergic drug .
MNS is characterized by generalized lead pipe rigidity and bradykinesie and normal reflexia, whereas serotonin toxicity is characterized by neuromuscular hyperactivity (e.g., clonus, hyperreflexia) and predominantly in the lower extremities.
It is unlikely that the patients with suspected serotonin toxicity, in our cohort, were suffering from MNS. Haloperidol was administered to only two out of seven patients, metoclopramide to only four patients and both for a relatively short period of time. Haloperidol was one of the inclusion criteria and physical examination followed within a maximum of 48 h. Furthermore, all seven patients showed neuromuscular hyperactivity, a characteristic of serotonin toxicity and no significant elevation of serum CK. Physical symptoms started after administration of serotonergic agents, and delirium was diagnosed on average on day five of admission.
Further studies might involve a bigger cohort, possibly in comparison with non-delirious patients, to determine risk factors in terms of drug use and precipitating and predisposing factors for developing serotonin toxicity. Furthermore, it would be of use to know whether the administration of a 5-HT antagonist, such as cyproheptadine, in patients with suspicion of serotonin toxicity, would benefit in terms of symptoms that subside. Furthermore, most importantly future studies should aim to investigate whether treatment of serotonin toxicity leads to earlier recovery and better patient outcome.
In summary, we have demonstrated that a significant proportion of patients in the ICU who have been diagnosed with delirium might in fact be classified as suffering from serotonin toxicity. We furthermore showed that awareness of potential serotonin toxicity is low among physicians in our tertiary university ICU. However, lack of existing studies strongly suggests that this is not unique for our ICU. We urge for further studies on the incidence and prevalence of serotonin toxicity in ICU patients and the effect of possible early therapeutic interventions.