Article Text
PDF
Abstract
Objectives The objective of this preliminary study was to explore long-term changes in neurobehavioral parameters, brain morphology and electroencephalography of sepsis patients who received intensive care compared to non-septic intensive care unit (ICU) patients.
Methods Two-centre follow-up study 6–24 months after discharge from hospital using published norms and existing databases of healthy controls for comparison. Patients included 25 septic and 19 non-septic ICU survivors who were recruited from two ICUs of a university and community hospital. Measurements used include brain morphology, standard electroencephalography, cognition and psychiatric health and health-related quality of life.
Results Sepsis survivors showed cognitive deficits in verbal learning and memory and had a significant reduction of left hippocampal volume compared to healthy controls. Moreover, sepsis and to some extent non-septic ICU patients had more low-frequency activity in the EEG indicating unspecific brain dysfunction. No differences were found in health-related quality of life, psychological functioning or depressive symptoms, and depression could be ruled out as a confounding factor.
Conclusions This study demonstrates permanent cognitive impairment in several domains in both septic and non-septic ICU survivors and unspecific brain dysfunction. In the sepsis group, left-sided hippocampal atrophy was found compared to healthy controls. Further study is needed to clarify what contribution sepsis and other factors at the ICU make to these outcomes. Specific neuroprotective therapies are warranted to prevent persisting brain changes in ICU patients.
- Cognitive Neuropsychology
- Anaesthetics
- Clinical Neurology
- Eeg
- Mri
Statistics from Altmetric.com
Introduction
Up to 30% of the patients seen at intensive care units (ICUs) in Europe and the US are diagnosed with sepsis.1 ,2 Septic encephalopathy (SE), defined as acute cerebral dysfunction, is estimated to occur in the majority of such patients and is manifested in slowing of cognitive processing, delirium or coma.3 While most of these symptoms appear transient in nature, there is compelling evidence that SE may actually cause long-term cognitive impairment.4 ,5 Rodent models show long-term spatial memory and fear-based learning deficits even after complete recovery from the initial sepsis syndrome.6–8 Recently global cognitive impairment was found in 17% of a large older cohort of sepsis survivors 8 years post illness, with significant associated functional impairments as well.5 A recent self-report based study of quality of life has pointed to difficulties with sensory processing, emotional functioning, concentration and memory in the 1–4 years following sepsis in a small cohort of eight patients.9 Lowered quality of life in sepsis patients has been verified by this and several other previous studies.10–13 Although much effort has been made to study long-term cognitive consequences of acute respiratory distress syndrome,14 a related but distinct syndrome, and of critical illness in general,15 ,16 to our knowledge Iwashyna et al5 and Lazosky et al9 have provided the only direct examinations of cognitive change specifically as a result of sepsis to date. Specific inferences about which domains are affected could not be made in the former study, and other events occurring in the long period before follow-up (in particular, aging) could have accounted for impairments found. In the latter study, reports indicated domain-level impairments, but with a small sample and only subjective evidence. Hence, an examination of specific cognitive domains after successful recovery from sepsis is warranted.
The present preliminary study extends previous work by exploring brain morphological and electroencephalographic data, as well as psychiatric burden and quality of life in addition to cognition and comparing this data to other, non-septic, ICU survivors. An investigation of the specific mechanisms associated with SE is beyond the scope of this study. Based on previous studies of animal models, it is assumed that factors causing SE include systemic and local generation of inflammatory cytokines, oxidative stress, microglial activation, alterations of the cerebral microcirculation, neurotransmitter imbalances and peripheral organ failure.17 ,18
Patients and methods
Inclusion/exclusion criteria
Inclusion criteria, based on the diagnostic criteria of the American College of Chest Physicians/Society of Critical Care Medicine Consensus Conference, were: history of sepsis (infection plus two systemic inflammatory response syndrome criteria) or severe sepsis (presence of sepsis and concomitant acute organ dysfunction occurring in at least one organ).19 Exclusion criteria were history of neurological disease (stroke, dementia, cerebral trauma) or other diseases that might confound outcome measures (eg, renal failure or hepatic insufficiency) as well as having undergone cardiopulmonary bypass surgery prior to ICU admission.
Patient selection
Medical records of all patients admitted to the operative (anesthesiological, surgical and cardiothoracic surgery) ICUs at the University of Bonn, the Department for Anesthesiology and surgical ICU of the Helios Clinic in Siegburg, Germany, between January 2004 and August 2006 were studied. Those who met the inclusion criteria (sepsis: n=184; non-septic ICU: n=194) were contacted by mailed letter. Sepsis survivors (n=25, 12f, 13m) comprised patients diagnosed with either sepsis (n=8) or severe sepsis/septic shock (n=17). All severe sepsis patients required vasopressor therapy, thereby also fulfilling criteria for septic shock. Non-septic ICU survivors were treated at the ICU within the same time period but without a diagnosis of sepsis (n=19, 8f, 11m). The clinical and demographic characteristics of both patient groups are presented in table 1.
Demographic and clinical characteristics of sepsis and ICU participants
Control samples
No healthy control group was actively recruited for this study. Published historical norms were used for analysis of the cognitive, psychiatric and quality of life and are described in detail below. Sex-matched healthy controls selected from two different established databases at our clinic were used to evaluate the resting-state EEG data and MRI data. MRI healthy controls (HC) consisted of 31 (16f, 15m) subjects with no personal history of neurological disorders (including epilepsy) or major medical illness. All MRI images were reviewed and confirmed to be normal on visual inspection. EEG HC consisted of 20 subjects (7f, 13m) who received an EEG in order to exclude the presence of epilepsy after having had vasovagal syncopes.
Study design
This is a two-centre follow-up study of long-term outcomes in a small sample of septic and non-septic ICU survivors using published norms and existing databases of healthy controls for comparison (detailed descriptions and references below). Outcome measures included cognitive performance, brain MRI, resting-state EEG, and psychiatric health and quality of life collected within a time period of 6–24 months after discharge from the ICU. Diagnostic laboratory data was collected during ICU stay. Brain imaging was done within 1 week of neuropsychological assessment post ICU stay. This study protocol was approved by the local Ethics Committee (05/04) and written informed consent was obtained from all participants prior testing.
Laboratory work-up
All patients underwent routine diagnostic laboratory tests while at the ICU comprising blood cell count, liver enzyme count, creatinine levels, blood coagulation parameters, C reactive protein and electrolyte levels. Medical data collected included are shown in table 1. The treatment of all patients was carried out according to standard clinical procedures.
Neuropsychological assessment
Since all patients underwent intubation and analgosedation during their time on the ICU, it was not possible to perform cognitive testing during ICU.
Estimated premorbid verbal ability
Determining the level of premorbid intelligence is crucial to detecting presence of neurocognitive deficits prior to the ICU experience. Due to the methodological difficulty of obtaining a baseline cognitive performance score for sepsis patients, premorbid intelligence was estimated based on a vocabulary test. This widely-used method presumes that verbal knowledge (vocabulary) is not or only minimally affected by brain damage and that estimated premorbid verbal ability is highly correlated with premorbid intelligence. Although this not a full-scale premorbid intelligence estimate, we used a German vocabulary test (Multiple Choice Word Test-B, German: Mehrfach-Wortschatz-Intelligenztest) that is valid (r=0.94) and reliable (rtt=0.87–0.95) and has been shown to accurately measure premorbid intelligence in clinical samples.20 ,21 For the sake of clarity, however, we denote this measure estimated premorbid verbal ability.
Cognition
Cognition was assessed using Neuro Cognitive Effects (NeuroCogFx), a validated computerised assessment battery that includes eight cognitive subtests: short term memory, working memory, alertness, selective attention, interference, verbal memory, figural memory and phonetic verbal fluency.22 Psychomotor speed and divided attention were assessed using the Trail Making Test A and B.23
Auxiliary testing
A subset of patients (n=14) received further memory testing after a preliminary analysis indicated memory deficits: the German version of the Auditory Verbal Learning Test and the Rey Complex Figure Test (RCFT). The Auditory Verbal Learning Test includes 15 aurally presented words with five immediate recall trials, an interference word list, short and long recall trials and a word recognition task.24 The RCFT includes a Copy task in which a complex figure is directly copied from an original presentation, and a recall task in which the patient redraws the figure from memory after a short delay (3 min) and long delay (30 min).25
Comparison to estimated premorbid verbal ability
Neuropsychological profile analysis was used to determine cognitive change compared to own estimated premorbid verbal ability levels. This method is widely used in neuropsychological assessment to reveal changes in specific domains which may be hidden when comparing performance to an historical norm alone.26
First, a composite score for cognition (Ccomp) was calculated using the unweighted average of cognitive performance across the NeuroCogFX and the TMT scores only. Cognitive performance and estimated premorbid verbal ability were then compared by subtracting the premorbid z-score from each individual cognitive test z-scores: ((Cognitive Test z-score)—(Multiple Choice Word Test-B z-score)) to determine the extent of deficit score (Zdiff).27
Quality of life/psychiatric health
Perceived health-related quality of life was assessed by the Short Form-36 Health Survey (SF36).28 The Global Severity Index was used as a comparative measure in the data analysis. The summed scores for physical function, role physical, bodily pain and general health served as a composite physical health score (SF36 Physical). The summed scores for vitality, social function, role emotional and mental health served as a composite score for mental health score (SF36 Mental). A second questionnaire Symptom Check List-90-R (SCL-90-R) was used to assess the severity of common medical and psychiatric symptoms and complaints. It includes nine dimension scores with summary measures for physical health and mental health.29 Symptoms of clinical depression were assessed by the Beck Depression Inventory (BDI).30
Magnetic resonance imaging and brain volumetry
MR scanning of the patient groups was performed on three scanners with two magnetic field strengths (Tesla 1.5 and Tesla 3). Scanning of healthy controls in the database was done with a Philips 1.5T Achieva whole body system. For all groups, a 3D FFE sequence (TE/TR/FLIP: 15/3.6 mesc/30°) was made with 140 slices and a resolution of 1×1×1 mm3. Data were converted to the analyse-format and brain volume measures was performed with the Analyse 7.0 software, according to a previously published protocol.31 The intrarater reliability was assessed by blindly measuring 10 independent test MR volumes. Intracranial volumes were obtained by automated tissue segmentation with SPM5 (Wellcome Department of Cognitive Neurology, London) using tissue probability maps. The volumes of hippocampi were divided by the total intracranial volume to adjust for differences in head size.
EEG recording and spectral analysis
Ten to 20 min of EEG were recorded using the international 10–20 system with additional T1 and T2 electrodes. Linked ear-lobe electrodes were used as reference (A1 and A2). Data was sampled at a rate of 256 Hz using an anti-aliasing low pass filter. Segments containing artefacts like eye movements, eye-blinks, other movement artefacts or drowsiness were identified and excluded from further analyses by consensual evaluation of two board-certified EEG experts (GW and FM), blinded for the group membership of the cases. The first 5 min of awake, resting, eyes closed and artefact-free EEG segments were selected for further analysis. Power spectra were calculated for consecutive 4-s windows (1024 samples each) for each electrode contact, and absolute spectral band power for conventional EEG frequency bands (δ: 0.5–4 Hz; θ: 4–8 Hz; α: 8–13 Hz; β: 13–20 Hz; γ: 20–40 Hz) were averaged across different windows. Since global changes were expected, the band power values were finally averaged over all electrode contacts.
Statistics
Data were analysed using the Statistical Package for Social Sciences (V.17.0; SPSS Inc, Chicago, Illinois, USA). Homogeneity of variance and normality of data distributions were evaluated for the study and control groups. Missing values for duration of hospitalisation and Acute Physiology and Chronic Health Evaluation (APACHE) II scores were imputed using nearest neighbour imputation. The problem of unequal cell counts was resolved using the default unweighted means method in SPSS V.18.0. Student t-tests, Pearson's correlation coefficients, one-way analysis of variance, and Multivariate Analysis of Covariance (MANCOVA) and were used for data analysis. For all comparisons, the α-level was set at 0.05, with Bonferroni corrections as appropriate. Due to very unequal sample sizes, sepsis and severe sepsis were not analysed separately, although information on demographics, cognition and quality of life are provided in online supplemental tables S1–S3.
Results
Demographic and clinical characteristics
There was a significant difference between length of stay at the ICU, APACHE II and sequential organ failure assessment (SOFA) scores between the two groups, the first two of which were used as covariates in later analyses. There was no difference in age or estimated premorbid verbal ability (table 1). There were no differences in gender distribution between the patient groups and MRI or EEG healthy controls, however the control groups were younger (MRI HC: n=31, Age=40.3±1.852; EEG HC: n=20, Age=35.6±3.043). Therefore age was also used as a covariate.
Cognitive performance
Sepsis survivors underperformed on eight scores, while the non-septic ICU survivors showed lowered scores on six subtests compared to historical norms (table 2).
Results of neuropsychological testing (in standard values) for sepsis and ICU participants compared to historical norms (HN)
Both patient groups performed better than average on the copy task of the RCFT. Since both patient groups had above average estimated premorbid verbal ability, a direct comparison of post-ICU performance to estimated premorbid verbal ability is a better indicator of actual impairment. Figure 1 reveals long-term mild cognitive impairments in attention, verbal fluency, executive function and verbal memory in sepsis survivors, and impairments in executive function and recall in non-septic ICU survivors when compared to own estimated previous intelligence level. Correlation and regression analysis revealed that these deficits are not due to length of ICU stay, time since hospital release, ventilator days, nor to severity of disease (APACHE II and SOFA; data not shown).
Cognitive performance compared to estimated premorbid verbal ability level. Cognitive performance and estimated premorbid verbal ability were compared by subtracting premorbid ability score from each individual cognitive test z-score. Overall extent of cognitive deficit indicated by Ccomp. Means and standard error of measurements are shown since only a subgroup received auditory verbal learning test and Rey Figure tests. TMT, Trail Making Test.
MRI
Visual inspection found no presence of lacunae and white matter lesions in either patient group.
Volumetry
The Pearson correlation coefficient between volumetry ratings was r=0.98 for the rater of the present study. Age-related changes such as lesions or microbleeds did not occur in either patient group. A MANCOVA was carried out for volumetric MRI data using age and intracranial brain volume as covariates (figure 2A). Univariate results show volume differences for the left hippocampus and total hippocampus across the three groups. There were no differences in grey matter, white matter, intracranial or cerebral spinal fluid volume. Pairwise comparisons revealed a reduction of the left hippocampus and total hippocampus compared to healthy controls (figure 2A). The hippocampal volume in non-septic ICU survivors was between that of sepsis patients and healthy controls.
Hippocampal volume and resting state EEG for patients and healthy controls. (A) Univariate results of MANCOVA showed a significant difference for left and total hippocampal volume. Pairwise comparisons revealed a significant difference between sepsis and healthy controls for the left hippocampus and total hippocampal volumes. *p<0.05, **p<01. (B) Pairwise comparisons revealed pathologically increased θ and δ power in sepsis survivors compared to healthy controls. Non-septic intensive care unit survivors also had increased δ power. *p<0.05, **p<0.01. MANCOVA, multivariate analysis of covariance.
EEG data
Patients who had survived sepsis showed more θ and δ power than healthy controls (one-sided t-test, p<0.021 and p<0.037, respectively) (figure 2B). Non-septic ICU survivors had more δ power compared to healthy controls (p<0.025). More spectral power in high-frequency bands (α, β, γ) was also found for the patient groups compared to healthy controls. These findings indicate unspecific brain dysfunction in both patient groups, with greater dysfunction in the sepsis survivors than in the non-septic ICU survivors.
Psychological and physical quality of life
Both groups showed somewhat lowered psychological and physical quality of life on all assessed measures from the SCL-90-R, SF-36, and BDI (table 3) compared to historical norms. For the composite scores physical and mental health multivariate testing revealed no group differences. A group difference on Role Emotional from SF-36 indicated reduced work ability due to psychological problems in sepsis participants compared to ICU. The participants in both groups did not score high enough on the BDI to be classified as clinically depressed.
Results of quality of life inventories (in standard values) for sepsis and ICU patients compared to historical norms
Discussion
We offer a specific profile of long-term cognitive deficits in sepsis in attention, verbal fluency, executive function and verbal memory. Visual memory and visuoconstructive ability is preserved in these patients. In addition, sepsis survivors showed reduced cognitive functioning in the absence of psychiatric disturbance, and only marginally reduced physical quality of life. The somewhat lowered health-related quality of life (HRQOL) reflects findings from earlier studies of ICU patients.9 ,10 ,13 ,32
Since cognitive reserve in our patient samples was high, as evidenced by high estimated premorbid verbal ability and visuoconstructive ability, expected cognitive performance level should have been much higher. Thus, the actual cognitive performance of sepsis and non-septic ICU patients was around 1.5 SDs below estimated premorbid verbal ability levels.
Intriguingly, sepsis, which is a ‘generalised’ factor, predominantly affected the left hippocampus in our study population. This may be explained by asymmetrical distribution of neurotransmitters in the brain, including dopamine, noradrenaline, GABA and choline acetyltransferase. Of note, higher levels of noradrenaline, which exerts portent anti-inflammatory mechanisms besides its classical action as a neurotransmitter,33 have been attributed to the right hemisphere. Consequently, this biochemical asymmetry could finally predispose the left hemisphere to a higher degree of vulnerability in response to inflammatory insults.34 ,35 Additionally, several neurodegenerative diseases show asymmetrical disease progression including semantic dementia and Alzheimer's disease.35 ,36 Positron emission tomography studies also reveal left-greater-than-right metabolic dysfunction in early dementia.37 These asymmetries indicate that the left hemisphere might be more susceptible to neurodegeneration in general, and this phenomenon may also occur in response to sepsis. Asymmetrical reduction in the left hippocampus of the sepsis patients compared to healthy controls is in line with previous results from animal models that demonstrated sepsis-induced long-term behavioural impairment accompanied by hippocampal loss of neurons.6 ,7
The fact that sepsis and ICU patients had more low-frequency activity in the EEG could indicate unspecific brain dysfunction. For animals exposed to sepsis, a similar finding has been reported,7 albeit these changes were present during sepsis and not evaluated thereafter. These changes seem to persist in humans and future studies are required to monitor the development of individual EEG changes longitudinally.
In conclusion, there is indeed evidence of post-ICU cognitive dysfunction, left-sided hippocampal atrophy and brain dysfunction in sepsis survivors. However, we do not know if this is attributable to sepsis or perhaps other factors during ICU stay (disease severity, dosage of vasopressors, sedatives and analgesics or episodes of hemodynamic instability) or thereafter (physiotherapy, cognitive rehabilitation, psychotherapy, support with reintegration into work and family).
In addition, small sample sizes and the use of multiple historical norms and healthy control groups limit interpretation. The time-dependent nature of cognitive and brain volume changes could not be thoroughly examined with a one time-point assessment. Future studies should include several assessments at even shorter periods following ICU release. Finally, it would be highly valuable to examine sepsis effects in an older cohort, as the ability to recover from neuroinflammation may potentially weaken with age as well as with cumulative brain insults.
Nonetheless, comparisons to patients’ own estimated premorbid verbal ability objectify a large extent of cognitive decline in sepsis survivors, general brain dysfunction and left-sided hippocampal atrophy, all of which appear to be irreversible. For acute as well as long-term outcome, future treatment strategies should therefore consider specific neuroprotective therapies. Furthermore, therapeutic interventions to ameliorate cognitive deficits and to promote psychological health in sepsis survivors are desirable and should be tested in future studies.
Acknowledgments
We wish to thank Dr Lukas Scheef, MD, of the Radiology Department at the University Hospital of Bonn for discussion of the radiological data and to Dr Malte Bewersdorff, MD, of the Neurology Department at the University Hospital of Bonn for study and technical assistance with MRI and EEG data. Thank you to Dr Rolf Fimmers, of the Biometry Department at the University Hospital of Bonn for statistical consultation. We also thank Alexandra Lindlau, medical student at the University Hospital of Bonn, for acquisition of clinical data from patient files.
References
Supplementary materials
Supplementary Data
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
Files in this Data Supplement:
- Data supplement 1 - Online supplement
Footnotes
AS and CNW contributed equally
-
Contributors AS: study concept, acquisition of data, drafting/revising manuscript; CNW: drafting/revising manuscript, statistical analysis and interpretation of demographic, clinical, cognitive, psychiatric, life quality and MRI data; TO: study concept, acquisition of data; HU: study concept, acquisition and interpretation of MRI data; MK: study concept, patient selection, patient contact/follow-up, acquisition of data and revision of the manuscript for intellectual content; GW: analysis and interpretation of EEG data; FM: study concept, acquisition and analysis of EEG data; JW: acquisition and interpretation of cognitive data; KF: study concept, interpretation of the data, revising manuscript for intellectual content; AH: acquisition and interpretation of data; FJ: acquisition of MRI data, analysis and interpretation of MRI data, revising manuscript for intellectual content; CP: acquisition of data, revising manuscript for intellectual content; MTH: Guarantor, study concept, study supervision, drafting/revising manuscript for intellectual content.
-
Funding This work was supported by a grant from the German Research Council (DFG, Clinical Research Group 177) to MTH.
-
Competing interests HU serves on the Editorial Boards of Neuroradiology and Clinical Neuroradiology. KF is supported by the German Research Council (Deutsche Forschungsgemeinschaft, Grant FL715-1). With kind permission of the University of Bonn he receives honoraria for the published version of ‘NeuroCogFX’. FJ is a member of the advisory boards of AC Immune, UCB, Lilly, GE, Jansen-Cilag and receives honoraria from Pfizer, Eisai, and Novartis. MTH serves on the editorial boards of the Journal of Chemical Neuroanatomy and the Journal of Neurochemistry.
-
Ethics approval The study was approved by the local Ethics Commission at the Medical School of the University of Bonn, Bonn, Germany (Reference 05/04).
-
Provenance and peer review Not commissioned; externally peer reviewed.
-
Data sharing statement Professor MT Heneka has access to all of the data and the right to share any and all data.