Interventions for preventing critical illness polyneuropathy and critical illness myopathy (Review)

Interventions for preventing critical illness polyneuropathy and critical illness myopathy (Review) Hermans G, De Jonghe B, Bruyninckx F, Van den Berghe G This is a reprint of a Cochrane review, prepared and maintained by The Cochrane Collaboration and published in The Cochrane Library 2009, Issue 1 http://www.thecochranelibrary.com

T A B L E O F C O N T E N T S HEADER....................................... 1 ABSTRACT...................................... 1 PLAIN LANGUAGE SUMMARY.............................. 2 BACKGROUND.................................... 2 Figure 1...................................... 4 OBJECTIVES..................................... 6 METHODS...................................... 6 RESULTS....................................... 7 DISCUSSION..................................... 11 AUTHORS CONCLUSIONS............................... 12 ACKNOWLEDGEMENTS................................ 12 REFERENCES..................................... 12 CHARACTERISTICS OF STUDIES............................. 16 DATA AND ANALYSES.................................. 23 Analysis 1.1. Comparison 1 Intensive insulin therapy (IIT) versus conventional insulin therapy (CIT), Outcome 1 Occurence of CIP/CIM............................... 25 Analysis 1.2. Comparison 1 Intensive insulin therapy (IIT) versus conventional insulin therapy (CIT), Outcome 2 Duration of mechanical ventilation............................... 26 Analysis 1.3. Comparison 1 Intensive insulin therapy (IIT) versus conventional insulin therapy (CIT), Outcome 3 Duration of ICU stay................................... 27 Analysis 1.4. Comparison 1 Intensive insulin therapy (IIT) versus conventional insulin therapy (CIT), Outcome 4 Death. 28 Analysis 1.5. Comparison 1 Intensive insulin therapy (IIT) versus conventional insulin therapy (CIT), Outcome 5 Severe adverse events.................................. 29 Analysis 2.1. Comparison 2 Corticosteroids versus placebo, Outcome 1 Occurence of CIP/CIM........ 30 Analysis 2.2. Comparison 2 Corticosteroids versus placebo, Outcome 2 Death.............. 31 Analysis 2.3. Comparison 2 Corticosteroids versus placebo, Outcome 3 Severe adverse events (dichotomous data).. 31 Analysis 2.4. Comparison 2 Corticosteroids versus placebo, Outcome 4 Serious adverse events (continuous data).. 32 Analysis 3.1. Comparison 3 Recombinant human growth hormone versus placebo, Outcome 1 Death...... 33 Analysis 4.1. Comparison 4 Immediate postoperative enteral feeding versus no enteral feeding, Outcome 1 Duration of ICU stay.................................... 33 Analysis 5.1. Comparison 5 Inspiratory muscle training (IMT) versus no IMT, Outcome 1 Duration of mechanical ventilation.................................... 34 APPENDICES..................................... 34 WHAT S NEW..................................... 36 HISTORY....................................... 37 CONTRIBUTIONS OF AUTHORS............................. 37 DECLARATIONS OF INTEREST.............................. 37 INDEX TERMS.................................... 37 i

[Intervention Review] Interventions for preventing critical illness polyneuropathy and critical illness myopathy Greet Hermans 2, Bernard De Jonghe 3, Frans Bruyninckx 4, Greet Van den Berghe 1 1 Department of Intensive Care Medicine, Catholic University of Leuven, University Hospitals, Leuven, Belgium. 2 Department of General Internal Medicine, Medical Intensive Care Unit, Catholic University of Leuven, University Hospitals Leuven, Leuven, Belgium. 3 Réanimation Médico-Chirurgicale, Centre Hospitalier de Poissy-Saint-Germain, Poissy, France. 4 Physical Medicine and Rehabilitation, Catholic University of Leuven, University Hospitals Leuven, Leuven, Belgium Contact address: Greet Van den Berghe, Department of Intensive Care Medicine, Catholic University of Leuven, University Hospitals, Herestraat 49,3000, Leuven, Belgium. Greta.vandenberghe@med.kuleuven.be. Editorial group: Cochrane Neuromuscular Disease Group. Publication status and date: New, published in Issue 1, 2009. Review content assessed as up-to-date: 30 October 2007. Citation: Hermans G, De Jonghe B, Bruyninckx F, Van den Berghe G. Interventions for preventing critical illness polyneuropathy and critical illness myopathy. Cochrane Database of Systematic Reviews 2009, Issue 1. Art. No.: CD006832. DOI: 10.1002/14651858.CD006832.pub2. Background A B S T R A C T Critical illness polyneuro-and/or myopathy (CIP/CIM) is an important and frequent complication in the intensive care unit (ICU), causing delayed weaning from mechanical ventilation. It may increase ICU stay and mortality. Objectives To examine the ability of any intervention to prevent the occurrence of CIP/CIM. Search strategy We searched the Cochrane Neuromuscular Disease Group Trials Register (October 2007), MEDLINE (January 1950 to April 2008), EMBASE (January 1980 to October 2007), checked bibliographies and contacted trial authors and experts in the field. Selection criteria All randomised controlled trials (RCTs), examining the effect of any intervention on the incidence of CIP/CIM in adult medical or surgical ICU patients. The primary outcome measure was the incidence of CIP/CIM after at least seven days in ICU, based on electrophysiological or clinical examination. Data collection and analysis Two authors independently extracted the data. Main results Three out of nine identified trials, provided data on our primary outcome measure. Two trials examined the effects of intensive insulin therapy versus conventional insulin therapy. Eight hundred and twenty-five out of 2748 patients randomised, were included in the analysis. The incidence of CIP/CIM was significantly reduced with intensive insulin therapy in the population screened for CIP/CIM (relative risk (RR) 0.65, 95% confidence interval (CI) 0.55 to 0.78) and in the total population randomised (RR 0.60, 95% CI 0.49 1

to 0.74). Duration of mechanical ventilation, duration of ICU stay and 180-day mortality but not 30-day mortality, were significantly reduced with intensive insulin therapy, in both the total and the screened population. Intensive insulin therapy significantly increased hypoglycaemic events and recurrent hypoglycaemia. Death within 24 hours of the hypoglycaemic event was not different between groups. The third trial examined the effects of corticosteroids versus placebo in 180 patients with prolonged acute respiratory distress syndrome. No significant effect of corticosteroids on CIP/CIM was found (RR 1.09, 95% CI 0.53 to 2.26). No effect on 180-day mortality, new serious infections and glycaemia at day seven was found. A trend towards fewer episodes of pneumonia and reduction of new events of shock was shown. Authors conclusions Substantial evidence shows that intensive insulin therapy reduces the incidence of CIP/CIM, the duration of mechanical ventilation, duration of ICU stay and 180-day mortality. There was a significant associated increase in hypoglycaemia. Further research needs to identify the clinical impact of this and strategies need to be developed to reduce the risk of hypoglycaemia. Limited evidence shows no significant effect of corticosteroids on the incidence of CIP/CIM, or on any of the other secondary outcome measures, except for a significant reduction of new episodes of shock. Strict diagnostic criteria for the purpose of research should be defined. Other interventions should be investigated in randomised controlled trials. P L A I N L A N G U A G E S U M M A R Y Interventions to reduce neuromuscular complications acquired during the acute phase of critical illness Neuromuscular problems are frequent complications in patients with severe disease that require admission to the intensive care unit (ICU). Weakness of limbs and respiratory muscles is most frequently due to critical illness polyneuro-and/or myopathy (CIP/CIM). As a consequence, patients face a delay in weaning from ventilatory support and rehabilitation. Recovery of strength often occurs within weeks to months but can be incomplete or not occur at all. CIP/CIM is associated with increased ICU stay and mortality rates. Prevention and treatment of CIP/CIM is therefore of great importance. We searched for and analysed all randomised controlled trials that examined the effects of any treatment intervention on the incidence of CIP/CIM in adult patients admitted to an ICU. We ultimately found three trials that focused on two different interventions. Two trials with 825 participants examined the effect of intensive insulin therapy, aiming to maintain blood glucose levels within the normal range (80 to 110 mg/dl), versus conventional insulin therapy, aiming to avoid hyperglycaemia (high blood sugar > 215 mg/dl), on the incidence of CIP/CIM in patients staying in ICU for at least one week. Pooling the results of both trials showed that intensive insulin therapy reduces the incidence of CIP/CIM, the duration of mechanical ventilation, the duration of ICU stay and mortality at 180 days. No significant effect on mortality at 30 days was noted. Intensive insulin therapy was associated with a significant increase in hypoglycaemia (low blood sugar), and recurrent hypoglycaemic events. Although no increase in mortality within 24 hours of hypoglycaemia was noted, hypoglycaemia remains an issue of concern when implementing intensive insulin therapy in critically ill patients, as it may cause neurological complications. In both trials, no clinical measurement of weakness of the limbs was reported, nor data on physical rehabilitation. Data were derived from subgroup analysis, which may also limit the conclusions. The third trial compared corticosteroid therapy versus placebo in 180 patients with persisting acute respiratory distress syndrome. Results showed no evidence of an effect of corticosteroids on the incidence of CIP/CIM, no effect on mortality at 180 days, on new serious infections, on glucose levels on day 7 and a trend towards less episodes of suspected or probable pneumonia. The number of new events of shock was reduced. In this trial, only 92 of the 180 patients were prospectively evaluated for CIP/CIM. B A C K G R O U N D Critical illness polyneuropathy (CIP) is an acute and primary axonal motor and sensory polyneuropathy. It occurs in critically ill patients, leading to severe limb weakness and difficulty in weaning from a ventilator. The term CIP was first used in 1986 (Bolton 1986). Later on it became clear that in some patients the disease primarily affected the muscles and the term myopathy in critical illness or critical illness myopathy (CIM), was introduced. Clinical signs of CIP and CIM are basically the same and include 2

flaccid weakness of the limbs. Deep tendon reflexes are usually normal or reduced in pure CIM but may be absent in CIP. In CIP, distal loss of sensitivity to pain, temperature and vibration may also occur. Weakness affects legs more often than arms. Although head, facial, tongue and jaw movement are relatively spared, facial muscles can be involved and ophthalmoplegia may occur. Involvement of the phrenic nerves and diaphragm may cause ventilator weaning problems. Electromyography (EMG) and nerve conduction studies (NCS) are useful to confirm the diagnosis and to exclude other causes of weakness. In contrast to Guillain-Barré syndrome in which demyelination occurs, nerve conduction velocity is normal or nearnormal. Electrophysiological examination can be used early, before clinical assessment is feasible, and typically shows reduced nerve conduction amplitudes and abnormal spontaneous electrical activity. EMG and NCS findings are not specific, however, and occur in CIP as well as in CIM and in other disorders. Therefore, differentiating between both disorders on electrophysiological data is often only possible when patients are fully cooperative, and voluntary motor unit potential recruitment can be obtained. Also, direct muscle stimulation can be helpful, although technically demanding. Co-existence of CIP and CIM can also make interpretation difficult. Ultimately, electrophysiological findings always need to be correlated with clinical findings. Finally, muscle biopsy can confirm muscle involvement and differentiate between the three subtypes of CIM: diffuse non-necrotising myopathy or CIM in the strict sense; thick filament myopathy; and acute necrotising myopathy. Due to the difficulty in differentiating between CIP and CIM, and the frequent association of both, terminology that has been used in the literature is often inconsistent and sometimes merely descriptive (for example. ICU acquired paresis and acquired neuromuscular disorders ) and both disorders are frequently grouped together as critical illness polyneuromyopathy or critical illness polyneuropathy and myopathy (CIP/CIM). The incidence rates reported in the literature vary according to patient population, definition and timing of evaluation. For instance, in patients with sepsis or systemic inflammatory response syndrome the incidence is 70% (Witt 1991) and rises up to 100% if complicated with multiple organ failure (MOF) (Tennila 2000). Patients in the intensive care unit (ICU) for at least seven days will acquire CIP/CIM in 49% to 84% of cases (Coakley 1998; Hermans 2006; Van den Berghe 2005). CIP and CIM have important consequences. They cause muscle weakness, paralysis, and impaired rehabilitation. CIP/CIM itself may prolong mechanical ventilation and may increase ICU stay, hospital stay and mortality. Improvement occurs within weeks in mild cases and within months in severe cases. In the more severe cases, recovery may be incomplete or even not occur at all. Hypotheses concerning the pathophysiology of CIP/CIM are outlined in Figure 1. Many risk factors have been related to the incidence of CIP. Both prospective and retrospective trials have found sepsis, systemic inflammatory response syndrome (SIRS) and MOF to play a key part in the development of CIP. Many other factors have been incriminated although results have not always been consistent. Factors identified as independent risk factors by prospective studies are: female sex (De Jonghe 2002), severity of illness (Bednarik 2005; de Letter 2001), duration of organ dysfunction (De Jonghe 2002), renal failure and renal replacement therapy (Garnacho-Montero 2001), hyperglycaemia (Van den Berghe 2005; Witt 1991), hyperosmolality (Garnacho-Montero 2001), parenteral nutrition (Garnacho-Montero 2001), serum albumin ( Witt 1991), duration of ICU stay (Van den Berghe 2005; Witt 1991), vasopressor and catecholamine support (Van den Berghe 2005) and central neurological failure (Garnacho-Montero 2001). 3

Figure 1. Presumed pathophysiological mechanisms and their interactions involved in the development of CIP/CIM. Dark shaded area indicates events taking place in the nerves, light shaded area indicates events taking place in the muscle. (Additional abbreviations: ROS, reactive oxygen species; SR, sarcoplasmic reticulum). Data on the impact of corticosteroids on neuromuscular function have been controversial. Many reports have found CIM to occur in patients treated with a combination of corticosteroids and neuromuscular blocking agents (NMBAs). However, most prospective studies could not identify corticosteroids as an independent risk factor for CIP (Garnacho-Montero 2001) or CIP/CIM (Bednarik 2005; de Letter 2001; Van den Berghe 2005) although two other prospective trials on weakness (De Jonghe 2002; Herridge 2003) did. In one study, corticosteroids were even found to be an independent protective factor for the incidence of CIP/CIM (Hermans 2006). Most prospective trials also could not identify NMBA use as an independent risk factor for CIP or CIP/CIM (Bednarik 2005; De Jonghe 2002;Van den Berghe 2005), although other trials did (Garnacho-Montero 2001; Hermans 2006). Until recently, the only possible way to affect the incidence of CIP and CIM was controlling risk factors. This includes aggressive treatment of sepsis and avoiding or limiting the use of corticosteroids and NMBAs to the lowest dose possible and the shortest time feasible. Interventions studied Nutritional therapy As conflicting data exist on the effect of nutritional status and the route of administration of feeding, we examined the effects of various feeding protocols such as enteral feeding, parenteral feeding and also therapies such as protein and amino acid supplementation. As glutamine has been suggested to be relatively deficient and possibly pathogenically linked to CIM, studies examining glutamine supplementation were evaluated. Arginine is also an indispensable amino acid for maintaining body protein homeostasis and nutrition after burn injury or sepsis. Outcomes in catabolic critical illness have however been variable (Heyland 2003). Parenteral nutrition supplementation with arginine and glutamate improved nitrogen balance and attenuated protein myofibrillar catabolism (Berard 2000). Antioxidant therapy 4

Excess production of free oxygen radicals and diminished endogenous antioxidant mechanisms have been implicated in multiple organ failure and septic shock. Whole blood glutathione (GSH) concentration and absolute synthesis rate were diminished in burn patients (Yu 2002) and muscle biopsies in critically ill patients also have shown decreased levels of GSH (Hammarqvist 1997). Antioxidant therapy may, therefore, improve outcomes due to their ability to scavenge free oxygen radicals and replete GSH stores. Supplementing GSH reduced the oxidative stress indices in critically ill patients. This effect was more pronounced if N-acetylcystein (NAC), a precursor of GSH, was added (Ortolani 2000). Hormonal therapy As the balance between catabolic and anabolic hormones changes in critical illness, hormonal therapy may be beneficial. The use of testosterone to improve anabolism is theoretically attractive. The testosterone derivate oxandrolone improves weight but not muscle strength in patients with AIDS wasting (Berger 1996). In malnourished patients with alcoholic hepatitis, oxandrolone may however improve survival (Mendenhall 1993). Nandrolone decanoate also had marked positive effects on the nitrogen balance in postoperative patients (Michelsen 1982) and improved inspiratory muscle strength in chronic obstructive pulmonary disease (COPD) patients (Schols 1995). Therefore, testosterone derivates may have some functional benefits in critically ill male patients. Basal growth hormone (GH) secretion is increased in acute critical illness. In contrast, circulating levels of insulin-like growth factor I (IGF-I) and II (IGF-II) are low secondary to resistance to GH. The increased protein turnover and negative nitrogen balance in critical illness are partly due to resistance to GH and decreased production and action of IGF-I. Several studies have shown that high doses of recombinant human GH improved nitrogen balance (Gore 1991; Jeevanandam 1995; Pape 1991; Ponting 1988; Voerman 1992; Voerman 1995; Ziegler 1990) and resulted in improved peripheral muscle strength (Jiang 1989), may have facilitated weaning from mechanical ventilation (Knox 1996) and improved inspiratory muscle force (Pape 1991). However, other studies did not show these beneficial effects on muscle strength or ventilator weaning despite improved nitrogen balance (Pichard 1996; Suchner 1990; Takala 1999). Also various side effects such as hyperglycaemia (Takala 1999) and increased splanchnic oxygen consumption have been reported (Dahn 1998) and even increased mortality (Takala 1999). Studies evaluating the effects of exogenous IGF-1 in critical illness have not demonstrated beneficial effects on the nitrogen balance (Yarwood 1997). Conflicting evidence exists concerning the role of hyperglycaemia. Various mechanisms have been proposed to explain its possible detrimental effects such as impairment of the microcirculation to the peripheral nerve, caused by hyperglycaemia (Bolton 2005; Witt 1991) and passive uptake of glucose with increased generation/deficient scavenging system for reactive oxygen species (Van den Berghe 2004). It was hypothesised that avoiding hyperglycaemia might exert a protective effect on the neural mitochondrial function (Van den Berghe 2005) and also reduce organ damage by endothelial protection through diminished release of nitric oxide (Langouche 2005). Two randomised controlled trials showed that maintaining normoglycaemia with intensive insulin therapy reduced the incidence of CIP/CIM (Hermans 2006; Van den Berghe 2005). In the first trial glycaemic control was an independent protective factor for the development of CIP/CIM (Van den Berghe 2005), although it was also hypothesised that some of the beneficial effects of this therapy may be due to insulin itself ( Van den Berghe 2004). Some rationale also exists for treatment with corticosteroids (CS) as these drugs may not only improve survival in some ICU subpopulations (septic shock, acute respiratory distress syndrome (ARDS), acute asthma) but could also reduce duration and severity of multiple organ failure, a major determinant of CIP/CIM ( Bollaert 1998; Boyer 2006; Briegel 1999; Chalwa 1999; Meduri 1998; Rowe 2001; Steinberg 2006). Furthermore, CS showed a preventive effect on CIP/CIM in one prospective trial (Hermans 2006). Intravenous immunoglobulins As immune mechanisms may be implicated in the pathogenesis, immunoglobulins were administered intravenously in three patients, without beneficial effects (Wijdicks 1994). However, in a retrospective study early treatment of gram negative sepsis with immunoglobulins may have prevented CIP (Mohr 1997). Physiotherapy and rehabilitation programs Most authors agree that physiotherapy is useful, probably by obviating disuse atrophy that undoubtedly dominates this disorder ( Pandit 2006). Physiotherapy is also considered to be important in preventing joint contractures and pressure sores (Bolton 1986). This review is important for several reasons. Two previous systematic reviews on CIP/CIM have been published (De Jonghe 1998; Stevens 2007). The first review considered prospective cohort studies and focused on assessment and outcomes. The second review aimed to determine prevalence, risk factors and outcome. In the current systematic review, we aimed to specifically assess the effect of any intervention, studied in a RCT, on the incidence of CIP/CIM. In ICU patients, interventions to reduce mortality are obviously of paramount importance. However, in acute phase survivors, reduction of morbidity is also imperative. CIP/CIM is now recognised as a major complication of severe critical illness and its highly sophisticated management in the ICU. As both locomotor and respiratory muscles can be involved, CIP/CIM significantly influences weaning from mechanical ventilation and recovery of physical autonomy. Therefore, interventions demonstrated to reduce the incidence of CIP/CIM are expected to have a 5

beneficial effect on weaning and mechanical ventilation duration, muscle function and overall locomotor autonomy in the recovery phase. Although less clearly investigated to date, other potential adverse consequences of CIP/CIM include negative psychological impact, higher residual mortality after the acute phase, and increased costs. No systematic review of interventions to prevent CIP/CIM is known to exist. O B J E C T I V E S The objective was to systematically review the evidence from randomised controlled trials concerning the ability of any intervention to reduce the incidence of CIP or CIM in critically ill patients. M E T H O D S one week after ICU admission. As no international criteria exist for this diagnosis, we defined CIP/CIM for the purpose of this review as weakness of the limbs or respiratory muscles, or EMG documented peripheral polyneuropathy or myopathy for which other causes than CIP/CIM have been excluded. Secondary outcomes Secondary outcome measures included clinically relevant consequences from a possible reduction in the incidence of CIP/CIM and also any possible side effects of the therapy. (a) Weaning from mechanical ventilation. This was defined as the time to actual and final liberation from the ventilator. (b) Duration of ICU stay (c) Death at 30 days (after ICU admission) (d) Death at 180 days (after ICU admission) (e) Serious adverse events from the treatment regimens, which were fatal, life-threatening or required prolonged hospital stay (eg hypoglycaemia, hyperglycaemia, organ failure). Criteria for considering studies for this review Types of studies We included all RCTs in humans in which the efficacy of any treatment used to prevent or reduce the incidence of CIP or CIM was compared to placebo, no treatment or a different treatment. Types of participants Adult participants ( >18 years) of either sex, admitted to a medical, surgical or mixed ICU. Types of interventions We included in the review any form of intervention which has been related to a decreased risk of CIP or CIM or both in the literature, such as nutritional interventions: enteral versus parenteral feeding and supplemental therapies such as protein and amino acid supplementation (glutamine and arginine). Antioxidant therapy (GSH or NAC), hormone therapy (testosterone, oxandrolone, GH and IGF-1, intensive insulin therapy, glucocorticoids), intravenous immunoglobulin and physiotherapy, electrical stimulation and rehabilitation programs were also included. Types of outcome measures Primary outcomes The primary outcome measure was the incidence of CIP and/or CIM during ICU stay. We extracted this outcome measure at least Search methods for identification of studies Electronic searches We searched the Cochrane Neuromuscular Disease Group Trials Register (searched October 2007) for randomised trials using the following search terms: polyneuropathies, muscle weakness, rhabdomyolysis, quadriplegia, paresis, neuromuscular manifestations, myopathy, neuromuscular disorder, neuromuscular disease, neuromyopath, motor syndrome, muscle function, critical illness, intensive care, therapeutics, enteral nutrition, parenteral nutrition, glucocorticoids, exercise therapy, diet therapy, prevention and control, rehabilitation, therapies, investigational, amino acids, arginine, glutamine, antioxidants, acetylcysteine, growth hormone, androgens, insulin-like growth factor I, immunoglobulins, insulin, amino acids, arginine, glutamine, anti oxidant, N- acetylcysteine, glutathione, growth hormone, androgen, testosterone, oxandrolone, Immunoglobulin, glucocorticoids, physiotherapy, electrostimulation. We adapted this strategy to search MEDLINE (January 1950 to April 2008) and EMBASE (January 1980 to October 2007). We used a combination of MeSH and keyword searching in these databases and we used the RCT strategy from the Cochrane Handbook Appendix 5b. The search strategies are given in Appendix 1 and Appendix 2. Searching other resources We reviewed the bibliographies of the randomised trials identified and contacted authors and known experts in the field to identify additional unpublished data. 6

Data collection and analysis Selection of studies Two of the four authors independently checked titles and abstracts identified from the Trials Register. The full text of all potentially relevant studies was obtained and the authors decided which trials fit the inclusion criteria. If the two authors did not agree, consensus was achieved by discussion. outcome data are presented as relative risks (RRs) and risk differences (RD) with 95% confidence intervals (CIs). Continuous outcome data are presented as weighed mean differences (WMDs) with 95% CIs. The fixed-effect model was used. Assessment of reporting biases Sensitivity analysis was not performed and a funnel plot was not drawn as these were not appropriate. Data extraction and management Two authors independently extracted data. We obtained data on methodological quality criteria, baseline characteristics and relevant data for the primary and secondary outcome measures if available. In case of missing data we contacted the investigator whenever possible. Information about randomised but excluded patients was also sought and, if available, incorporated into the analysis. Assessment of methodological quality of included studies Assessment of methodological quality took into account security of randomisation, allocation concealment, observer and patient blinding, completeness of follow-up, intention-to-treat analysis, explicit diagnostic criteria and adherence to the regimen. Allocation concealment relates to the randomisation procedures, whereby investigators involved in participant allocation should not be able to influence how the groups are assembled. It involves concealment of the next allocation in the randomisation sequence, such that neither the investigator nor the participants can be aware of the next group assignment until after the decision about whether an individual is eligible for the trial has been finalised. We graded these items A: adequate; B: moderate risk for bias; C: inadequate or not done. If the information was not available, the item was graded as inadequate. For some items graded as not done or unclear, we attempted to contact the authors to obtain additional information about the trial design. In the case of disagreement between the two authors, we reached agreement by consensus. Data were entered into the Cochrane statistical package, Review Manager, by one review author, and checked by a second author. Measures of treatment effect We analysed and presented results according to the statistical guidelines of The Cochrane Collaboration as described in Chapter 8 of the Cochrane Handbook for Systematic Reviews of Interventions (Deeks 2005). The reliability of the various pieces of evidence obtained form part of the discussion. Trials of each intervention were analysed and presented separately. Dichotomous R E S U L T S Description of studies See: Characteristics of included studies; Characteristics of excluded studies. The MEDLINE search revealed 18 possibly relevant references, out of the 2397 that were retrieved. We excluded 11 of these. One trial was not randomised (Hsieh 2006) and one trial was a retrospective analysis (Mohr 1997). Seven trials did not provide any clinical nor electrophysiological outcome data on the neuromuscular system (Berard 2000; Bouletreau 1985; Fläring 2003; Gamrin 2000; Kuhls 2007; Sevette 2005; Tjader 2004). Finally, another two trials were not performed in the ICU (Nava 1998; Porta 2005). In the latter, although the patient population is described as a respiratory intensive care unit population, patients were randomised to two rehabilitation programs, that only started to differ from each other as soon as patients were able to walk and, therefore, can not be considered ICU patients any longer. Seven trials remained that fulfilled the selection criteria (Caruso 2005; Hermans 2007; Johnson 1993; Pichard 1996; Steinberg 2006; Van den Berghe 2005; Watters 1997). Our search of EMBASE up to October 2007 revealed another three out of 191 retrieved trials, which could possibly be included. Two of these, however, were not randomised trials (Knox 1996; Ziegler 1990) and the third did not concern ICU patients (Paddon-Jones 2005). Searching the Cochrane Neuromuscular Disease Group Trials Register resulted in three references, which had been identified by previous search strategies. Personal contact with experts in the field revealed two additional trials that could be included (Takala 1999; Takala 1999 1). A total of nine trials fulfilling the inclusion criteria remained. Two of these trials compared treatment with intensive insulin therapy versus conventional insulin therapy and included a total of 825 participants (see Characteristics of included studies). The first trial (Van den Berghe 2005) evaluated the incidence of CIP/CIM in surgical ICU patients, staying in ICU for at least seven days. The second trial (Hermans 2007) was performed in a medical ICU. Both trials were subanalyses of large randomised controlled trials, 7

examining the effects of intensive insulin therapy versus conventional insulin therapy in a surgical (Van den Berghe 2001) and medical ICU (Van den Berghe 2006 a). The patients screened for CIP/CIM were those who were still in ICU on day seven. Four hundred and five long-stay patients of a total of 1548 patients randomised were included in the analysis in the surgical trial (Van den Berghe 2005) and 420 out of 1200 patients randomised in the medical trial (Hermans 2007). CIP/CIM was diagnosed by weekly electrophysiological examination. Significantly fewer patients in the intensive insulin therapy group developed CIP/CIM compared to the conventional insulin therapy group (reduction surgical trial:109/224 (49%) to 46/181 (25%), P < 0.0001; medical trial: 107/212 (50%) to 81/208 (39%) P = 0.02).The need for prolonged mechanical ventilation, defined as mechanical ventilation for at least 14 days, was significantly reduced in the intensive insulin therapy group compared to the conventional insulin therapy group (reduction surgical trial 93/224 (42%) to 57/181 (31%), P = 0.04; medical trial: 99/212 (47%) to 72/208 (35%), P = 0.01). Both trials were single centre trials. Three other trials compared treatment with recombinant human growth hormone (rhgh) versus placebo (Pichard 1996; Takala 1999; Takala 1999 1), and included a total of 552 patients. A mean dose of 0.3 IU/kg (Takala 1999; Takala 1999 1) and 0.43 IU/kg (Pichard 1996) rhgh was given for 12 days (Pichard 1996) to up to 21 days (Takala 1999; Takala 1999 1). None of these trials provided data on our primary outcome measure. The two large trials (Takala 1999; Takala 1999 1) reported an excess of in hospital mortality (Finish trial: 467/119 (39%) versus 25/123 (20%), P < 0.001, multinational trial: 61/139 (44%) versus 26/141 (18%), P < 0.001) and, therefore, analyses of other outcome parameters were considered inappropriate. One trial compared therapy with corticosteroids versus placebo, in 180 patients with unresolving ARDS (Steinberg 2006). Intravenous methylprednisolone was given at a starting dose of 2 mg/kg/day then gradually tapered over more than three weeks. In this trial, the primary outcome data were only prospectively available in 92 patients and no difference was noted between both treatment groups (incidence CIP/CIM placebo 11/48 (23%), intervention 11/44 (25%), P = 0.67). One trial compared inspiratory muscle training, consisting of twice-daily exercise sessions, in which resistance was adapted according to the individual patient s capacity and gradually increased, with no inspiratory muscle training in 40 patients (Caruso 2005). No data were provided on our primary outcome measure, no difference was found in the duration of mechanical ventilation. One trial compared early postoperative enteral feeding, started at 20 ml/h six hours postoperatively, increased to half of the target dose on day two and full dose at day three, versus no enteral feeding in the first six days (Watters 1997). No data were provided on our primary outcome measure, nor on any of our secondary outcome measures. Finally, one trial compared administration of magnesium 6 g/16 hours versus placebo in a cross-over design in 20 patients (Johnson 1993). No data were provided on our primary outcome measure, nor on any of our secondary outcome measures. Risk of bias in included studies Table 1 gives the quality scores for each trial. In seven trials randomisation and allocation concealment was adequate. In two trials however, no information is given with regards to the randomisation procedure and these were, therefore, graded C (Johnson 1993; Pichard 1996). Patients were unaware of the regimen they were allocated to in all trials with two exceptions, in which this was not possible due to the nature of the intervention (Caruso 2005; Watters 1997). In three trials (Hermans 2007; Van den Berghe 2005; Watters 1997) the treating physician could not be blinded, also due to the nature of the intervention, and therefore observer blinding was scored C. In one trial, due to inconsistency on this subject in the paper, both observer blinding and outcome assessor blinding were graded C (Pichard 1996) and in one trial physician and outcome assessor were not blinded, and graded C for these items (Caruso 2005). For the other trials, observer blinding was adequate. In one other trial, due to lack of information, outcome assessor blinding was graded C (Johnson 1993). One trial was an unblinded trial and therefore also outcome assessor blinding was graded C (Watters 1997). In the remaining trials this was adequate. Completeness of follow-up was adequate in all trials, except for the two (Takala 1999; Takala 1999 1) in which grip strength was only measured in survivors, and therefore graded B. In one trial, in which 40 patients were randomised but data only presented on 25, completion of follow-up was graded C (Caruso 2005). Intention-to-treat analysis was only adequate in one trial ( Johnson 1993). Two trials (Hermans 2007; Van den Berghe 2005) reported on electrophysiological examination, only in those patients staying in ICU for at least seven days. Concerning intention-to-treat analysis, therefore, no data are available on presence or absence of CIP/CIM in patients dead or discharged within one week. We have graded this item as a moderate risk for bias (see Discussion). In one trial measurements were not available in 2/20 patients, and therefore it was also graded B (Pichard 1996). In two trials (Takala 1999; Takala 1999 1) intention-to-treat analysis was performed on all outcome measures except for strength, which was only measured in survivors. Again, no data were available for dead patients and this item was graded B. In one trial (Steinberg 2006), prospective evaluation of muscle force was performed in only 92/180 patients and intention- to- treat analysis was graded C. In one trial 25 out of 40 patients randomised, were accounted for, and this was graded C. Finally, in 1one trial 3/31 patients were not accounted for in the analysis and therefore the trial was graded B (Watters 1997). The major problem concerned the lack of adequate diagnostic criteria in seven trials (Caruso 2005; Johnson 1993; Pichard 1996; Steinberg 2006; Takala 1999; Takala 1999 1; Watters 1997), of which only one trial provided data on 8

our primary outcome measure (Steinberg 2006). In this trial however, even in the prospectively evaluated population no diagnostic criteria were stated. In the other trials various volitional and non-volitional measurements of peripheral and respiratory muscle force were performed. However, no cut-off values for these measurements have been described to discriminate between normal and weak patients. In the remaining two trials reporting data on our primary outcome measure (Hermans 2007; Van den Berghe 2005), the diagnostic criteria used were electrophysiological criteria. As no internationally accepted diagnostic criteria exist for CIP/CIM, we graded this item B (see Discussion). Adherence to the protocol was adequate except for four trials graded C. In three (Johnson 1993; Pichard 1996; Steinberg 2006) no information is given on this item, in one trial preset aims appear inadequately reached (Watters 1997). One trial was graded B for this item as 144 out of 167 training sessions planned, were completed (Caruso 2005). The most common problem was that the primary outcome criterion was not stated in the methods section in six out of nine papers. Unpublished information derived from the authors of five trials ( Caruso 2005; Hermans 2007; Takala 1999; Takala 1999 1; Van den Berghe 2005) was used to grade certain items. Table 1. Methodological quality scores Study ID Patient blinding Observer blinding Outcome assessor blinding Secure randomisation Allocation concealment Completeness of follow-up Intentionto-treat analysis Diagnostic criteria Adherence Van den Berghe 2005 Hermans 2007 Pichard 1996 Johnson 1993 Takala- Finish trial 1999 Takalamultinational trial 1999 A A A C A A B B A A A A C A A B B A C C A C C A B C C C C A A C A A C C A A A A A B B C A A A A A A B B C A 9

Table 1. Methodological quality scores (Continued) Steinberg 2006 Watters 1997 Caruso 2005 A A A A A A C C C A A C C C A B C C A A C C C C C C B Effects of interventions Primary outcome measure: occurence of CIP/CIM Data on the electrophysiological incidence of CIP/CIM were only available in three trials, of which two trials compared intensive insulin therapy versus conventional insulin therapy (Van den Berghe 2005; Hermans 2007), and one compared corticosteroids versus placebo (Steinberg 2006). Data on the incidence of CIP/CIM were available for patients in the ICU for at least seven days in 389 patients treated with intensive insulin therapy, and 436 patients treated with conventional insulin therapy. The meta-analysis showed a significant benefit of intensive insulin therapy with a RR 0.65, 95% CI 0.55 to 0.78 ( Analysis 1.1). We chose the fixed-effect model, although statistical tests indicated possible heterogeneity as only two studies were available, and therefore a random-effects analysis would provide poor estimates of the distribution of intervention effects. Based on the current data however, a difference in the effect size between medical and surgical patients can not be excluded. As from a clinical point of view patients can not be identified as needing ICU for at least one week at the time of admission, we also analysed data on the primary outcome measure as well as on the secondary outcome measures in the total randomised population, consisting of 1360 intensive insulin therapy and 1388 conventional insulin therapy patients. As no electrophysiological examination was performed in patients in ICU for less then one week, we used imputation of negative results in these patients. This showed that intensive insulin therapy had a significantly beneficial effect on CIP/CIM with RR 0.60, 95% CI 0.49 to 0.74 (Analysis 1.1). Primary outcome data were also available in terms of clinical weakness in 93 patients, of which 48 received placebo and 44 received corticosteroids. No significant difference was observed between both treatment groups (RR 1.09, 95% CI 0.53 to 2.26) (Analysis 2.1). We also performed an intention-to-treat analysis, by imputing data for the first 87 patients that were not prospectively evaluated. For this purpose we used the retrospective data available for these patients. Results show no significant effect with a RR of 1.27, 95% CI 0.77 to 2.08. Secondary outcome measures a. Duration of mechanical ventilation Data on this outcome measure were available in the insulin trials and in the trial of inspiratory muscle training (IMT). For the insulin trials, in the total population randomised, as well as in the population screened for CIP/CIM, there was a significant beneficial effect of IIT on the duration of mechanical ventilation (WMD -2.00, 95% CI -2.93 to -1.07 for total population randomised and WMD -2.55, 95% CI -4.60 to -0.51 for the population screened) (Analysis 1.2). A fixed-effect model was used. As in the end only the patients surviving ICU stay will benefit from an intervention that reduced CIP/CIM, we also analysed data for ICU survivors only. This showed that the beneficial effect on duration of mechanical ventilation was also present considering only ICU survivors of the total population randomised, which consisted of 1181 IIT patients an 1163 CIT patients (WMD -1.00, 95% CI -1.86 to - 0.14) (Analysis 1.2). When considering only the screened patients that survived ICU stay, the effect was not significant (WMD - 1.59, 95% CI -3.98 to 0.79). In the IMT trial, no effect was found on the duration of mechanical ventilation (WMD -1.00, 95% CI -5.90 to 3.90) (Analysis 5.1). b. Duration of ICU stay Data on this outcome measure were only available in the insulin trials. Applying the fixed-effect model, in the total population that was randomised, as well as in the population that was screened for CIP/CIM, the duration of ICU stay was significantly reduced with IIT (WMD -1.48, 95% CI -2.43 to -0.54 for the total population randomised and WMD -3.59, 95% CI -5.70 to -1.48 in the 10

screened population) (Analysis 1.3). Again when only analysing the ICU survivors, the benefit remained present in the total population randomised, but did not reach statistical significance in the screened population (WMD total population randomised -1.00, 95% CI -1.90 to -0.10, screened population WMD -2.18 95% CI -4.66 to 0.30) c and d: death at 30 and 180 days In the insulin trials, there was no statistical effect on mortality at 30 days in the total population or in the screened population (total population: RR 0.94, 95% CI 0.80 to 1.10, screened population 0.91, 95% CI 0.72 to 1.14) (Analysis 1.4). A fixed-effect model was applied. The meta analysis however did show a significant beneficial effect on 180-day mortality (total population RR 0.87, 95%CI 0.76 to 1.00, screened population 0.78, 95% CI 0.66 to 0.93). Again for 180-day mortality, statistical tests indicate possible heterogeneity. As mentioned above, we chose the fixed-effect model and a different effect size for medical and surgical patients can not be excluded from the current data. In the steroid trial, no effect on mortality was shown at 180 days (RR 0.99, 95%CI 0.64- to 1.52)(Analysis 2.2). No data were available at 30 days, as this trial s primary outcome measure was death at 60 days, which also showed no difference between treatment groups. Subanalyses however revealed increased mortality in the steroid-treated patients in the subgroup enrolled at least 14 days after the onset of ARDS at 60 days as well as at 180 days. Pooling of data from two growth hormone therapy trials showed a significant increase in mortality at 30 days (RR 2.27, 95% CI 1.64 to 3.15)(Analysis 3.1) as well as at 180 days (RR 1.98, 95% CI 1.54 to 2.54) in patients treated with growth hormone. This analysis was performed again using the fixed- effect model, despite statistics suggesting heterogeneity, for the same reasons as mentioned above. e. Severe adverse events In the insulin trials, the incidence of hypoglycaemia, defined as blood glucose below 40 mg/dl and the occurrence of at least two hypoglycemic events were significantly higher in the intensive insulin therapy group versus conventional insulin therapy. This occurred in the total patient population randomised, as well as in the screened patients (Analysis 1.5). There was no increase in death 24 hours after the last hypoglycaemic event in the total population randomised (RR 1.60, 95% CI 0.42 to 6.10), nor in the screened patients (RR 0.77, 95% CI 0.15 to 3.96). Treatment with corticosteroids appeared to have a protective effect on the occurrence of new episodes of shock (RR 0.41, 95% CI 0.17 to 1.01) (Analysis 2.3), no significant effect on new serious infections (RR 0.68, 95% CI 0.42 to 1.11) and a trend towards fewer episodes of suspected or probable pneumonia (RR 0.44, 95% CI 0.18 to 1.09). Blood glucose levels reported in this trial on day seven were not significantly different (WMD 15.00, 95% CI -3.41 to 33.41) (Analysis 2.4) D I S C U S S I O N Nine studies, examining the effects of six interventions were included but only three trials on two interventions provided data for our primary outcome data, although none of these specifically evaluated the prevention of critical illness polyneuropathy/myopathy as a primary outcome. This is a very small number of trials. One of the main reasons for this is that no clear accepted diagnostic criteria for CIP/CIM exist. In several trials some measurements of peripheral or respiratory muscle force were made, such as maximal inspiratory pressure, hand grip force and thumb muscle force without the existence of a cut-off value that allows differentiation between normal and weak patients. To date, concerning clinical evaluation, an arbitrarily cut-off value is only described for the MRC sum score. Also, several trials have evaluated N-metabolism and protein balance of several interventions without clinical or electrophysiological correlation. Two included trials evaluated the effect of intensive insulin therapy on CIP/CIM and studied a total of 825 participants. One trial was in a surgical ICU and one trial in a medical ICU at a single centre. Our primary outcome measure, incidence of CIP/CIM after at least one week in ICU, was significantly reduced by intensive insulin therapy (RR 0.65 95% CI 0.55 to 0.78). As ICU stay of at least seven days can not be predicted accurately on admission, we also evaluated all outcome measures for the total population randomised. As patients discharged or dead within one week were not screened for CIP/CIM, we used imputation of negative results for these patients. Meta-analysis then showed a significant beneficial effect in the total population randomised (RR 0.60 95% CI 0.49 to 0.74). Imputation of data does have limitations. In patients discharged within one week, it is likely that no clinically relevant CIP/CIM was present. In patients dead within the first week, some diagnoses of CIP/CIM may have been missed, as electrophysiological signs may occur early. The diagnosis of CIP/CIM was made using only the presence of abundant spontaneous electrical activity on electrophysiological examination, which are observed either in axonopathy or muscle necrosis, two important components of the neuromuscular involvement in patients with CIP/CIM. On the other hand, some myopathies with muscle membrane inexcitability may therefore have been missed. Concerning the secondary outcome measures, a beneficial effect on duration of mechanical ventilation, ICU stay and 180-day mortality was noted, in the total population randomised, as well as in the population screened for CIP/CIM, but not for 30-day mortality, which is likely to be too early to see the benefit of this intervention. Whether the reduction in duration of mechanical ventilation is due to improved respiratory muscle force or other beneficial effects of intensive insulin therapy, such as reduced infections, organ failure remains unclear. Hypoglycaemia and at least two hypoglycaemic events occurred more frequently in the intensive insulin therapy group than in the conventional insulin therapy group in the total population randomised as well as in the screened patients. No increase in mor- 11