Postoperative delirium following orthopedic surgery for hip fracture occurs in up to 62% of the elderly patients, many with pre-existing dementia[1]. It is characterized by fluctuating changes in cognition, consciousness and attention and lasts on average, three days[1, 2]. Three clinical subtypes of delirium are known; a hyperactive, a hypoactive, and a mixed subtype [3, 4]. So far, the suggested pathophysiological mechanisms of delirium are mostly hypothetical, but lately there is a lot of interest in the neuroinflammatory system [5–7]. Although patients usually recover after resolution of the underlying cause, delirium appears to be an important risk factor for dementia, even in people without prior cognitive impairment[8].

The high frequency of dementia after delirium possibly reflects irreversible brain damage caused by the detrimental effects of the pathophysiological mechanisms of delirium on the brain. Moreover, it can be hypothesized that postoperative delirium itself might be the consequence of transient cerebral neuronal damage caused by peri-operative hypoxia, micro embolisms[9] or hypotension[10].

S100B-protein (S100B) and Neuron Specific Enolase (NSE) have been used as markers of brain damage in various diseases[11]. S100B belongs to the family of calcium binding proteins, it is expressed mainly by astrocytes and it is found in both intra- and extracellularly in brain tissue[12]. S100B is usually elevated in blood and in CSF due to nervous system damage, via functional disturbance of membrane integrity and/or increased permeability of the blood-brain barrier. NSE is the intracytoplasmic glycolytic enzyme enolase. It is found in neurons and neuro-endocrine tissue and it is elevated in the blood circulation after increased death rate of these cells.

Elevated levels of S100B have been shown to be associated with delirium in patients after abdominal surgery[13], after cardiac surgery[14] and in sepsis-associated delirium[15]. NSE was not elevated in serum in patients with delirium tremens[16] but it has been found to correlate with the degree of cognitive dysfunction at discharge from hospital after cardiac surgery but not after non-cardiac surgery[17, 18]. In these previous studies the delirious state at the day the sample was taken was not taken into account, neither subtype. The aim of this study in elderly hip fracture patients was twofold: (1) to compare changes before and after surgery of S100B and NSE levels in serum in patients with and without postoperative delirium, and to investigate the difference in serum levels before, during and after delirium; (2) to study the serum levels of S100B and NSE in different subtypes of delirium.

During the inclusion period 271 patients aged 65 years and older were admitted. A total of 55 patients were excluded due to lack of informed consent, 63 patients did not consent to extra blood withdrawal and 33 patients were excluded for logistic reasons. In and excluded patients were similar with regard to the male/female ratio, but the excluded patients were younger (mean age of 81.3 years versus 83.9 years, p = 0.005). In total 120 patients were included, 62 (52%) with delirium and 58 without. This resulted in 387 samples in total, 196 samples of delirious patients (21 samples taken before the delirious episode, 120 samples taken during the delirious episode, 55 taken after the delirious episode) and 191 samples of patients without delirium.

To study the course over time of S100B, we included samples from the day before surgery until post-operative day eight because during this period samples from both delirious and non-delirious patients were available. Following a backward approach, type of surgery, type of anesthesia, global cognitive impairment, and age were removed. The final statistical model included delirious state (p < 0.001), day of blood withdrawal (p < 0.001), pre- or postoperative status (p = 0.001) and type of fracture (p = 0.036). To illustrate the effect of delirium on S100B, levels of three different example patients with an intertrochanteric hip fracture were calculated and are presented in figure 1: 1) a patient who was delirious from one day before operation until eight days after operation (continuously delirious), 2) a patient who was never delirious during this period (continuously non-delirious), and 3) a patient who was delirious from the day after operation until postoperative day four (delirious day 1–4). S100B levels were highest 'during' delirium at all time-points. 'Before' and 'after' delirium S100B levels were still higher than those from 'non-delirious' patients. Delirious state was not significant in the same mixed model analysis with NSE in 233 samples. Spearman's rho test showed no significant correlation between S100B and NSE levels in the first sample (p = 0.99).

This study among elderly patients admitted for hip fracture showed that delirium was associated with increased level of S100B which could indicate cerebral damage either due to delirium or leading to delirium. The study population represents a typical cross-section of elderly patients admitted for hip fractures. A similar frequency of delirium (52%) has been found in other large studies of hip fracture patients[29]. The expected risk factors for delirium in surgical patients, such as functional and cognitive impairment, were confirmed in our study[30]. Next to delirium, type of fracture, pre- or postoperative status and day of blood withdrawal were all associated with S100B levels. The effect of fracture and surgery and the time course were expected covariables based on the literature[31]. Although the group of delirious patients with known subtype was small, the hyperactive type of delirium was the most common subtype of delirium in our study, similar to what has been described by others[32]. Moreover, selection bias seems improbable since the delirious patients included in this analysis did not differ in associated factors with S100B and levels of NSE and S100B compared to the delirious patients without known subtype. Because this still could have introduced a bias, future studies are needed to confirm the lack of association between subtypes and NSE or S100B.

A limitation of our study is that the NSE and S100B values were measured from peripheral blood and may not necessarily correspond to values in the brain. Under normal conditions serum NSE and NSE in cerebrospinal fluid (CSF) show no significant correlation, whereas serum S100B content is lower than that in CSF[11, 33]. We are not aware of studies that determined NSE and S100B in both CSF and blood among patients with cognitive deficits such as delirium. A second limitation of our study is that only 120 of the 183 (66%) patients consented to multiple blood withdrawals. Although the excluded patients were 2.5 yrs. younger it is unlikely that this small difference would affect the protein levels.

Our results showing elevated S100B levels in delirious patients are in line with former studies [13–15]. Studies looking at postoperative cognitive deficit (POCD) and S100B after surgery show however contradicting results, with no relation between S100B and cognitive performance after cardiac surgery at discharge [17], a questionable relation directly after cardiac surgery [34], but higher levels of S100B in patients with POCD after non-cardiac surgery[18]. The definition of POCD remains controversial, but it seems likely that delirium and short-term POCD are on a continuum of postoperative cognitive brain disorders[35]. Seeing delirium as subset of POCD, our results of NSE correspond to former studies that found no association between NSE and POCD at discharge from hospital after non-cardiac surgery[18]. Maybe NSE is released only in cases of severe brain damage, such as stroke, after resuscitation or neurologically complicated surgery[18].

Importantly, S100B is a marker of cerebral damage and/or reduced integrity of the blood brain barrier. In stroke, global hypoxia and traumatic brain injury a positive correlation between S100B and cognitive outcome has been established [36]. Possibly, the high frequency of dementia after delirium could also be predicted by the level of S100B during delirium. Predictive factors for cognitive outcome after delirium are unknown, largely because of the difficulties of diagnosing pre-existent cognitive functioning in patients with delirium[5]. On the one hand delirium and dementia (or cognitive impairment) could each be caused by the same precipitating factors such as the surgical procedure[7]. On the other hand, it could be hypothesized that dementia is caused by the detrimental effects of delirium itself [5]. The neuroinflammatory system seems to be a rational pathophysiological pathway for delirium [5–7] and this same system has been described in relation to cerebral damage [11]. The correlation between the levels of S100B and the proinflammatory cytokines IL6 and IL8 in the patients in this study can be seen as an indication that the neuroinflammatory system is related to possible cerebral damage due to delirium.

In our study of elderly patients admitted for acute surgical repair of hip fracture we found higher S100B levels in patients with delirium than in patients without delirium with the highest levels of S100B during delirium,. No difference in S100B and NSE levels between the different subtypes of delirium was found. Future studies are needed to elucidate the place of S100B in the pathophysiological pathway leading to delirium (and possibly dementia) and investigate its role as biomarker for delirium.