Sir: The process of converting scientific evidence into clinical guidelines involves making judgments. Few judgments are absolute – there is usually a penumbra of uncertainty around the decision, and the decision itself will be influenced by factors which may not be answered by current evidence. We used a voting process to identify the point of consensus and the extent of uncertainty for those elements of the guidelines for which there was no unanimity. These included selective digestive tract decontamination (SDD), steroids, glucose control, and activated protein C, among others.

Drs. Spoelstra and Girbes correctly draw attention to the divergence between the scientific literature in favor of SDD and the absence of a recommendation by the guidelines group. However, they describe this as bias, and we would not use this term. The voting process demonstrated the presence either of maximal uncertainty (from a group perspective), or of polarized views, one weakly opposed and the other weakly in favor of using this intervention. Intensivists, microbiologists, and infectious disease physicians are equally represented on both sides of this divide. Those committee members who did not support SDD were influenced by fear of future harms (antibiotic resistance) and predominance of trial data in preventing infection in non-infected patients not preventing a second infection in septic patients. This makes it very difficult to design a research study which would ‘convince the sceptics’, particularly as a recent unpublished large study suggests that SDD has a less beneficial impact on survival than suggested by earlier research (personal communication, Marc Bonten, Utrecht, Netherlands). Given current enthusiasm for the aggregated non-antimicrobial interventions referred to as ‘ventilator bundles’, it seems reasonable to propose that future SDD research should be undertaken in conjunction with an evaluation of these bundles, for example using a factorial design, since both are directed at preventing ventilator-associated pneumonia (VAP).

Concerning glucose control, we gave a strong recommendation to control blood glucose and a weak recommendation to target a blood glucose level of less than 8.3 mmol/l, and recently elaborated on our rationale for that target [1]. Would Spoelstra and Girbes not control blood sugar at all? If they would do so, at what level would they prescribe an insulin infusion? The van den Berghe study in medical (non-operative) patients suggested benefit for those patients randomized to normoglycemia as well as other clinical non-mortality benefits [2]. Two other studies demonstrated significant hypoglycemia using a normoglycemia target [3, 4]. The higher target level we propose is pragmatic and likely safe, and our recommendations are consistent with the data from two high-quality studies as judged by our group.

Concerning stress ulcer prophylaxis (SUP), the meta-analysis by Messori et al. was considered, but its influence was limited by the fact that it dealt with only one specific H2 blocker (ranitidine) [5]. In addition, for the purposes of this letter we performed an additional meta-analysis (Fig. 1) of six older studies included in Cook's meta-analysis mentioned by Spoelstra and Girbes, using mostly cimetidine, and five newer studies using ranitidine included in the Mes sori study [5, 6]. The results, obtained using the same method as in the Messori work (Peto odds ratio with fixed-effect model), showed a statistically significant effect on reduction in GI bleeding with the use of H2 blockers: OR 0.43, 95% CI 0.24–0.76. A proper meta-analysis would obviously require a more involved process, but we still judge the evidence in favor of beneficial effect on bleeding.

Fig. 1
figure 1

Meta-analysis of six previous studies

Full size image

A separateissueistheriskofVAP with SUP. The GRADE approachdictates a decision on what constitutes the critical outcome which is included in the clinical question. We made it explicit that the recommendation is based on GI bleeding risk reduction, and that it should be balanced by the potential for increase in VAP.