Evidence‐based critical care medicine: seeing through a glass darkly

Although evidence-based medicine appears to be a new and novel concept, it does beg the question of what happened before the recent emphasis on the idea. It may come as a surprise to some that medical decisions were generally not taken by inspecting chicken entrails or consulting the temple priestess. The Lancet first appeared in 1823 and volume one of the Journal of Experimental Medicine was published in 1896. Studies involving random allocation are said to date back to 1928 [1]. However, credit must be given to Professor Archie Cochrane for advocating and popularising evidence-based practice, the principles of which were set out in his influential book published in 1972 [2]. The term ‘evidence-based medicine’ is more recent and apparently first appeared in the medical literature in 1992, in a paper by Guyatt et al. [3]. The survey in this issue by Paddle et al. [4] illustrates how evidence can be rapidly incorporated into standard practice, but then, within a short time, be subject to reassessment and a change in routine management. The clinical study by Van den Berghe et al. [5] was based upon good basic science, and showed a significantly lower mortality rate and number of complications in critical care patients given insulin targeted to achieve blood glucose values within a narrow range. This article was rapidly followed by a change in clinical practice in intensive care units in the UK and many other countries. Although further evidence was published to suggest benefit in medical and paediatric patients, other studies showed that tight glycaemic control was of no benefit and may even be harmful [6]. The short history of critical care medicine is littered with apparently well-conducted trials that have shown statistically significant benefit, but where later studies have either proved to be negative or have led to uncertainty about the universality and applicability of the findings. The following is a personal and partial review of some of the evidence, with the intention of illustrating the difficulties and uncertainties in this quest for clinical knowledge. In 1988, Shoemaker et al. demonstrated a stunning decrease in mortality by driving patients with fluid and inotropes to achieve supranormal cardiorespiratory values [7]. This is often referred to as goal directed therapy. Their prospective study was based upon evidence showing that survivors from serious illness had higher median values of cardiac index, oxygen delivery and oxygen consumption. To measure these parameters, it was necessary to insert a pulmonary artery flotation (Swan Ganz) catheter. Many further studies supported the concept of ‘optimisation’ [8], and the use of the pulmonary artery catheter became almost obligatory in intensive care. Despite good evidence and strong advocates, some practitioners continued to express doubt about the widespread use of this technique [9–11]. Not all studies were positive and the practice never became established as routine. Over the years, meta-analysis has suggested that only certain groups will benefit [12]. The waters have been further muddied by evidence showing that the pulmonary artery flotation catheter is not beneficial and may be associated with harm [13]. The role of optimisation is still being explored in sepsis and in the postoperative period, while its place in clinical practice remains uncertain [14]. Other interventions have come and gone much more rapidly. Centoxin, the monoclonal antibody against Gram-negative endotoxin, was rapidly introduced into clinical practice following the positive results of a randomised controlled trial [15], but was withdrawn just as rapidly following a further randomised study [16]. Some treatments refuse to go away. One example is the use of steroids to treat severe sepsis and septic shock. The publication by Annane et al. in 2002 resulted in the widespread adoption of this treatment in intensive care [17]. However, the CORTICUS study in 2008 refuted any benefit [18]. A 2009 review identified 17 randomised trials of steroids with acceptably methodological quality including more than 2000 patients [19]. Meta-analysis found that there had been no clear benefit with respect to mortality. All these reviews and meta-analyses have not dispersed the confusion [20], with the conclusion of one being that ‘further large-scale trials appear mandated’ [21]. Some practices endure for many years until good evidence of efficacy, or its lack, is obtained. The use of naloxone for shock is an excellent example of how research does not necessarily help the clinician reach a firm conclusion. Eighty publications involving humans encompassed only six clinical trials suitable for meta-analysis [22]. These six studies included a total of only 126 patients with shock from a variety of different causes. The main conclusion was that ‘the clinical usefulness of naloxone to treat shock remains to be determined, and additional randomised controlled trials are needed to assess its usefulness’. In contrast, low-dose dopamine to prevent renal failure was advocated from at least 1980 [23]. Up until the year 2000, there were many publications and reports associated with its use. However, it took one good randomised, multicentre study of 328 patients to all but eliminate its use as an agent for renal protection [24]. Meta-analyses have their advocates and critics and there is even debate about how they are best carried out [25]. Analysis of the same evidence can result in different interpretations [26]. Some interventions, such as outreach and early warning systems, encompass such diverse interventions investigated with poor quality studies that it is difficult to form any clear conclusions from the evidence [27]. One review by Cochrane collaborators on the use of human albumin in critically ill patients suggested that albumin was associated with increased mortality [28]. The conclusions stated that albumin ‘should not be used outside of the context of rigorously conducted, randomised controlled trials’. However, when the rigorously conducted trial suggested was carried out, mortality was almost identical in albumin and control groups and no significant differences were found in any of the important outcomes [29]. Other treatments fail to be adopted despite good quality randomised controlled trials to support their benefits. An obvious example is selective digestive decontamination [30]. Not everyone would agree that the evidence base is convincing enough to make withholding selective decontamination of the digestive tract from critically ill ethically questionable [31], but it does raise the question of what evidence is required to mandate that a treatment is given. One hopes that the National Institute for Health Research (NIHR) Health Technology Assessment (HTA)-funded SuDDICU study will provide insights into why this treatment has not been widely adopted in the UK [32]. The Surviving Sepsis Campaign is a fascinating example of the evolution of evidence-based practice and has generated much debate [33, 34]. A selected group of experts reviewed the evidence and in 2004 came up with their treatment ‘bundle’ for the management of sepsis [35]. Criticism of the methodology and the rapidly changing evidence mandated a revision of these guidelines published in 2008 [36]. These guidelines received widespread support, being endorsed by 16 societies of intensive care medicine, although two societies chose not to do so [37]. Much advice changed between the two guidelines and the process made it obvious how much of the evidence is weak and inconclusive. However, the guidance was widely adopted and between 2005 and 2008, compliance with the entire management bundle increased from 18% to 36%, although there was a highly variable uptake in the way the individual interventions were applied. Over this time period, hospital mortality decreased significantly and was associated with the length of time a hospital was participating in the campaign [38]. Activated protein C (drotrecogin alfa) is included as one element of the sepsis bundle. The considerable amount of published evidence on its use of is readily available to all. However, concerns over the close relationship between the manufacturer and investigators, as well as uncertainties over the interpretation of the results, have stimulated a further large randomised study that is currently underway [39]. With some interventions, it is unlikely that further research will be forthcoming, at least in the short-to-medium term. One example is the use of extracorporeal membrane oxygenation (ECMO). This was the subject of a randomised controlled trial [40]. The results support those who believe that ECMO is a good treatment, whereas perceived weaknesses in the study design and outcomes provide arguments for those who believe that it is of no benefit [41]. While some may dispute the existence of a coherent condition called adult respiratory distress syndrome (ARDS) [42], many interventions have been tried to improve outcome for patients with this diagnosis. Turning patients prone and the use of inhaled nitric oxide (NO) are two interesting examples. Both of these interventions have been shown to improve oxygenation [43, 44]. However, evidence suggests that this is not associated with a decrease in mortality. In contrast, the study on lung protective ventilation did show a lower mortality in the intervention group [45]. The relevant point is that low tidal volumes were initially associated with worse oxygenation, although ultimately fewer deaths. It is important therefore not to assume that proxy outcomes such as oxygenation are going to tell you what you really want to know. Some studies are clearly fraudulent [46], whereas others may be biased and subject to outside influence [47]. Furthermore, there are many reasons why even well-conducted studies may reach the wrong conclusions, or why later evidence and analysis may lead to a re-interpretation of findings [48]. Caution has been advised in basing clinical practice on the findings of single-centre trials [49]. Such studies commonly lack scientific rigour and may not be generalisable to other units and situations, and there are statistical reasons why these trials may not be as reliable [50]. The management of critically ill patients is complex, often involving multiple interventions and processes, and it is often not realistic or possible to isolate single interventions from the totality of care. The mechanism by which evidence gets translated into clinical practice is also fascinating. Not only does a gap usually exist between empirical evidence and clinical practice, but the empirical evidence is often not directly applicable to the patients we are treating [51]. Most studies will exclude certain groups such as pregnant women and those at the extremes of age [52]. Almost certainly, they will target selected populations and it is only by expanding indications that the clinician can make judgements about whether to use the intervention in many of the patients they will have to treat. Examples include the use of thromboprophylaxis, where much of the research has been conducted in orthopaedic patients, but little evidence exists in other surgical groups or high-risk medical patients [53]. Trials of the use of induced hypothermia following cardiac arrest recruited those whose arrest occurred out of hospital, with very specific exclusions [54]. Should this intervention be used in other groups of patients? Studies examining the effect of an intervention often identify subgroups who get most of the benefit. For example, activated protein C was licensed for use in patients with the most severe sepsis, and subsequent investigation suggested little or no benefit in the less seriously ill. When interventions are applied in routine clinical practice, benefits may not be as great as those seen in the study population, and complications may be more common. The widespread acceptance and speed with which interventions are adopted appear highly variable. Even when there is widespread support for a practice, it may be given in a modified form to a variable percentage of patients. A large number of multicentre randomised clinical trials have been carried out in critically ill patients. A review in 2008, limited to adult studies with more than 50 patients and with mortality as an outcome, found that 72 studies had been published [55]. In ten studies, the intervention was beneficial, in seven it was detrimental and in the remaining 55 there was no effect on mortality. One conclusion was that methodological limitations and flaws may have prevented finding an effect when one really existed. Much energy, time and money are wasted on undertaking poorly designed or underpowered research. The recent history of critical care has been characterised by the formation of clinical trials groups, notably in Canada, the US, Australia and New Zealand, Europe and the UK. Increasingly, there is international cooperation in the design and conduct of studies. Important clinical interventions have been examined in well designed, international investigations. In some cases, a definitive answer to a question is ascertained. More often than not, further questions are raised. Although many areas of uncertainty remain, much knowledge has been gained [56]. Appropriately, those working in critical care have been encouraged to think critically about what they do, and to design good studies to provide evidence to inform clinical practice [20, 48, 57–60]. The ‘care bundle’ approach has become much more common. A group of interventions may improve outcome, although there is little evidence to support each individual component or to prove which elements are most important. Reproducible and protocolised management has the advantage of minimising the risk that vital treatments are omitted or delayed [61]. Information technology has the potential to play an important part in ensuring that treatments are given safely, and in a timely and appropriate manner [62]. Although practitioners may argue about the evidence, and despite the uncertainty surrounding many of our treatments, it is clear that evidence-based interventions can be effective [63]. Over the years, mortality has fallen considerably for critically ill patients in the UK and elsewhere [48, 64, 65]. We can speculate that it is not only the increase in knowledge that has improved outcomes, but the attitudes, training, skills and experience of the intensivist [66], as well as better access to beds and equipment. In 2010, the UK Faculty of Intensive Care Medicine finally came into being, confirming that the speciality is firmly established. As a speciality, we can be proud of our progress and the emphasis that has been placed on obtaining evidence to support our work. Although in many ways it is as though ‘we see through a glass darkly’ (1 Corinthians 13), each piece of evidence adds to the totality, helping us to see that much more clearly. No external funding or competing interests declared.

Récupéré en direct depuis OpenAlex et désinversé. Les résumés ne sont pas conservés dans cette base de données : les index inversés représentent 8,6 Go des 9,3 Go de texte de la base, et le serveur dispose de 13 Go libres.