Persistent Inflammation and Recovery after Intensive Care: A Systematic Review
Authors
David M Griffith MD, Matthew E Vale MBBS, Christine Campbell PhD2, Steff Lewis PhD2, Timothy S Walsh MD1
Affiliation
1Department of Critical Care, Centre for Inflammation Research, Queen’s Medical Research Institute, University of Edinburgh, UK
2Centre for Population Health Sciences, Medical School, University of Edinburgh, Teviot Place, EH8 9AG
Corresponding author and institution where work performed
David M Griffith
Department of Critical Care, Room W2.03, Centre for Inflammation Research, Queen’s Medical Research Institute, University of Edinburgh, 47 Little France Crescent, Edinburgh, EH16 4TJ, UK. Tel: +44 131 242 6661, Fax: 0141 242 6578
Email:
Abstract
Purpose
Physical weakness is common after critical illness however it is not clear how best to treat it. Inflammation characterizes critical illness, is associated with loss of muscle mass during critical illness and potentially modifies post-ICU recovery. We sought to identify published reports on the prevalence of systemic inflammation after critical illness and its association with physical recovery.
Methods
Systematic review of the literature. Sources: MEDLINE, EMBASE, CINAHL, CPCISSH, and CPCIC. January 1982-December 2011.
Results
From 7433 references, 207 full text articles were reviewed, 57 were eligible and 22 were included. Inflammation was present in most patients at ICU discharge according to CRP concentration (range 70-100%), pro-calcitonin (range 89-100%), TNFα (100%), and SIRS criteria (range 92-95%). Fewer patients had elevated MPO concentrations (range 0-56%). At hospital discharge, 9/10 COPD patients (90%) had elevated CRP. No studies tested the association between inflammation and physical recovery.
Conclusions
Inflammation is present in most patients at ICU discharge, but little is known or has been investigated about persistent inflammation after this time point. No studies have explored the relationship between persistent inflammation and physical recovery. Further research is proposed.
Keywords
Intensive Care, Critical Care, Critical illness, Rehabilitation, Inflammation, Recovery, Quality of Life
Introduction
Annually, around 10,000 patients are admitted to Scottish Intensive Care Units (ICUs) with a critical illness; numbers are increasing and the aging general population means that numbers of elderly patients are predicted to increase substantially over the next 20 years. Improvements in ICU treatment mean that about 75% of patients survive to hospital discharge [1], but many have persisting physical disability that reduces quality of life, and places high care burden on families and health services. Whilst persistent ICU acquired disability is now recognized, it is not clear how best to prevent or treat it [2].
The most prevalent symptoms for the ICU survivor are fatigue and muscle weakness [3][4]. Muscle biopsy studies reveal skeletal muscle abnormalities in virtually all patients recovering from critical illness [5]. These include axonal neuropathy, denervation, fibre atrophy, non-specific neuropathy and necrotising myopathy. Recovery of muscle function after critical illness is often incomplete [2].
Critical illness is characterised by global activation of the immune system causing a coordinated sequence of events known as the systemic inflammatory response syndrome (SIRS). Inflammatory cytokines have an established role in regulating muscle mass. TNFα, IL-1, IL-6, and endotoxin infusions result in muscle wasting syndromes [6, 7] due to increased protein catabolism [8-12], inhibition of protein synthesis [13], inhibition of muscle cell differentiation [14] and reduced amino acid uptake [15]. Chronic diseases such as cancer, COPD, heart failure and end stage renal disease, as well as normal aging are associated with loss of muscle mass and function. Numerous studies have observed associations between markers of inflammation and muscle function in these groups [16-25].
Inflammation in critical illness has been extensively studied in the acute phase of the illness, but it is unclear how many patients have evidence of ongoing inflammation in the recovery phase. In addition, it is unclear how inflammation and muscle dysfunction are inter-related in the rehabilitation stage of critical illness. The aim of this systematic review is to collate the available data describing the prevalence of persistent, systemic inflammation after critical illness and to establish whether inflammation is linked to markers of physical dysfunction in these patients. We aimed to seek data on persistent inflammation at 3 time points: at the point of ICU discharge, between ICU discharge and hospital discharge, and at any time point after hospital discharge.
Methods
This systematic review has been reported according to the relevant sections of the MOOSE guidelines for Meta-Analyses and Systematic Reviews of Observational Studies [26].
Search Strategy
Electronic databases EMBASE, MEDLINE, and CINAHL were systematically searched using the OVID user interface. In addition, grey literature sources were searched for conference citations (CPCISSH and CPCIC) using the Web of Science interface. An example search strategy for the MEDLINE database is given in Table 1. We searched for studies published between January 1982 and December 2011 of human intensive care unit patients who had a clinical or biochemical marker of systemic inflammation measured.
Study characteristics
Inclusion and exclusion criteria are summarized in table 2. Studies carried out in medical, surgical, or mixed intensive care units were considered. Studies including children, neonates, neurosurgical, or post-operative cardiothoracic patients were not considered.
A study was deemed to include a measure of systemic inflammation if it recorded all of the systemic inflammatory response syndrome (SIRS) criteria (i.e. white cell count, respiratory rate, body temperature, heart rate), C-reactive protein (CRP), or any established pro-inflammatory mediator (e.g. IL-1, IL-6, or TNF-alpha).
For a study to be considered, the marker of systemic inflammation had to be measured at one of 3 pre-specified time points: within 24 hours of ICU discharge, between ICU discharge and hospital discharge, and after hospital discharge.
If a study reported a measurement of systemic inflammation whilst the patient was in ICU, it was included if sampling continued until ICU discharge. If there was no reference to ICU discharge, the study was only considered if the last sample taken was at a time point >14 days after ICU admission. This considers that there was reasonable probability that the majority of patients being sampled at this time point would have been discharged from ICU. In such studies, the authors were contacted for further information.
No language restrictions were placed on the search. Where an English abstract was available, the study remained in the review provided there was sufficient information in the abstract. Where no English abstract was available, foreign language publications were excluded.
Selection of studies
De-duplication was carried out automatically using the OVID user interface (Ovid Technologies, New York), then manually using Endnote X4 software (Thompson Reuters, New York). Following this, the title list was searched to remove clearly irrelevant studies (e.g. studies of paediatric, neonatal, cardiothoracic, or neurosurgical patients, review articles, editorials, case reports and commentaries). The abstracts of the remaining studies were screened independently by 2 authors, and those not meeting the inclusion criteria were excluded. Disagreements about eligibility were resolved by discussion between the 2 screening authors. An inclusive approach was adopted. Where it was not clear from the abstract whether a study should be included, it remained in the review list.
Full text versions of the remaining articles were obtained whenever possible using the resources of the NHS, University of Edinburgh, and the British Library. Where an article could not be retrieved in full text, and there was insufficient information in the abstract to determine eligibility, it was excluded from the review (4 articles).
The full text articles were reviewed independently by 2 authors against inclusion and exclusion criteria. This resulted in a final short list for further evaluation and data extraction.
Data extraction
Each short-listed article was reviewed by 1 author looking specifically for an estimate of prevalence of systemic inflammation. Where a prevalence estimate was not provided in the text, attempts were made to contact authors for raw data to allow calculation of prevalence estimates. Acknowledging that raw data may not be available in older studies, authors were asked if they could provide summary measures (central tendency and sample variability). Authors were contacted by email and traditional mail on 2 occasions, 1 month apart, thus allowing 2 months in total to respond after the initial contact.
Data was extracted using a standard form. Parameters included were: author, publication title, publication journal, publication year, number of patients at start of study, number of ICU survivors, number of patients in whom inflammatory marker was available, inflammatory mediator including units of measurement, time point, prevalence estimate and/or summary estimate. Articles were also screened for any statistic that related persistent inflammation and physical recovery after critical illness.
Data synthesis
For each circulating biomarker, the upper limit of normal was defined as the 97.5th centile (or suitable alternative) from a previously published study of healthy volunteers. Prevalence estimates were calculated as the proportion of included patients exceeding this limit. In addition summary estimates (a measure of central tendency and distribution) were quoted.
Meta-analysis
Meta-analysis was not considered to be methodologically appropriate due to the considerable heterogeneity of the study populations under study and high risk of selection bias. For example, some studies focused on single diseases, certain ICU complications, or specific settings.
Risk of Bias Assessment
A bespoke ‘risk of bias’ instrument was developed by the authors to allow assessment of bias in prevalence estimates across a variety of study designs. This instrument was a modification of the instrument produced by Hoy & Colleagues [27] taking into account the major sources of bias affecting prevalence estimates, and the criteria identified previously by consensus [28]. External validity was assessed according to 4 criteria (target population, sampling frame, selection method, and risk of non-response bias). Internal validity was assessed according to 3 criteria (case definition, measurement instrument, and data collection method). For each of the 7 criteria, studies were assessed as high risk, low risk or not reported (NR). Within each domain (internal or external validity), an overall assessment of risk of bias was given according to the following rules: 0 criteria at high risk – low risk; 1 criterion at high risk – moderate risk; 2 or more at high risk – high. In the case of missing information, risk of bias was deemed to be ‘unclear’.
Results
Included Studies
Following electronic database searching and de-duplication, 7433 unique references were retrieved. In total, 3327 abstracts were scrutinised and from these, 207 articles fulfilling or potentially fulfilling eligibility criteria were retrieved for full text review. 57 papers appeared to fulfill eligibility criteria for the review. A flow diagram detailing exclusions at various stages of the review are detailed in Figure 1. Details of the included studies can be found in table 3.
Data Completeness
Of the 57 papers considered to be eligible after full text review, none had prevalence estimates for systemic inflammation and only 3 studies had summary estimates. Therefore the authors of all these studies were contacted to provide further data. The authors for 34 (65%) of the articles responded [29-60]. Seven of these did not measure inflammation at an appropriate time point [34, 49, 51-53, 60, 61] and were excluded. Two studies [55, 59] used the same data as other included studies [42, 48] and were excluded. Raw data to allow calculation of prevalence estimates was provided for 13 studies (23%) [29-31, 35, 39, 41, 42, 45-47, 56, 62, 63]. These studies were included in the analysis. In the 12 studies where prevalence data was not provided, summary estimates of biomarker concentrations were available for 5 studies and these were also included in the analysis [32, 33, 38, 40, 43, 64-66]. The remaining 7 studies were excluded. None of the studies measured physical function after ICU discharge. One investigator had measured health-related quality of life but was unable to provide data to allow calculation of association with inflammatory markers [33]. Finally, 1 author volunteered data from another published study [67]. This study was missed from the initial search because inflammation was not the main focus of the paper. The summary data from this study was included.
Study Design
Of 22 included papers, 19 (86% were observational, 3 (14%) were interventional. Of the observational studies, 3 (16%) were case control studies, 1 (5%) was cross-sectional, and 15 (79%) were cohort studies.
Biochemical measures of inflammation
C-reactive protein (CRP) was measured in 20 (91%) of studies. Pro-calcitonin (PCT) was measured in 3 (14%) studies. IL-6 was measured in 3 (14%) studies. TNF α was measured in 1 (5%) study. SIRS criteria were measured in 1 (5%) study. Myeloperoxidase (MPO) was measured in 1 study (5%). The cut-off values derived from healthy populations are given in the Electronic Supplement (eTable 1).
Validity
A summary of the risk of bias assessment is provided in the final columns of table 3. The detailed scoring can be found in the electronic data supplement (eTable2). In external validity terms, 13 papers were at high risk, 5 papers were at moderate risk, and 4 papers lacked enough information to make an assessment. In internal validity terms, 14 papers were at low risk. The remaining 8 papers lacked enough information to make an assessment.
CRP concentration at ICU discharge
Of the 22 included studies, 18 (82%) measured CRP at the point of ICU discharge. CRP concentration was elevated (>10mg/L) in the majority of patients ranging from 70% in a large study of mixed medical and surgical ICU patients [62] to 100% in patients with severe sepsis [47] and a cohort of patients who subsequently were readmitted to ICU [39].
The CRP concentration varied according to the population studied. The mean of the median concentrations of CRP at ICU discharge in the mixed medical / surgical cohorts was 60mg/L. Lower mean CRP concentration was observed in trauma ICU patients (23mg/L), patients with VAP (46mg/L), prolonged length of stay (45mg/L), and medical ICU patients (36mg/L). Higher mean CRP concentrations were noted in sepsis survivors (107mg/L) and surgical ICU patients (99mg/L). Unsurprisingly, the patients selected as cases for the observational studies of ICU readmission [39, 65] and unexpected death after ICU discharge [45] had high concentrations of CRP in their blood at ICU discharge (131 and 218mg/L respectively).
IL-6 concentration at ICU discharge
Three studies (14%) measured IL-6 at ICU discharge [41, 42, 56]. These included one study of mixed medical and surgical ICU survivors [56], one study of ICU patients with a length of stay longer than 6 days [41], and one study of ICU patients with sepsis [42]. The percentage of patients in each of these samples with IL-6 concentration above 3.5 pg/mL was 99%, 63% and 100% respectively. Median (IQR) IL-6 concentration at ICU discharge in these samples were 80 (42-183) pg/mL, 76 (2-100) pg/mL, and 20 (15-39) pg/mL. There is therefore evidence of significant elevations in IL-6 concentration in the 3 studies at ICU discharge.