Associations between blood glucose level and outcomes of adult in-hospital cardiac arrest: a retrospective cohort study

In this retrospective observational study, we found that post-ROSC BG level is associated with neurological and survival outcomes for IHCA patients. The optimal BG level for IHCA patients might differ according to the presence or absence of DM. Less stringent post-ROSC glycaemic control than that recommended by the ERC (i.e., maintaining the BG level below 180 mg/dL or 10 mmol/L [10] might be appropriate for some patients. Except BG level, other identified significant prognostic factors of IHCA were consistent with those reported from previous studies [18].

In studies of OHCA outcomes, hyperglycaemia has been associated with poor neurological recovery [4‐7]. For OHCA patients with ventricular fibrillation, Müllner et al. [4] and Nurmi et al. [5] indicated that elevated BG level in the early post-ROSC phase was associated with unfavourable neurological outcome. However, comorbidities were not considered in the statistical analyses of either of these studies [4, 5]. Kaukonen et al. [19] demonstrated that hyperglycaemia in critically ill patients may simply be an indicator of illness severity and that the association between hyperglycaemia and mortality was attenuated when lactate levels were considered in the regression analyses. Furthermore, Steingrub et al. [20] indicated that an association between the early post-ROSC BG level and neurologic outcome might not exist because a higher early post-ROSC BG level may only be an epiphenomenon of prolonged arrest duration, which itself would cause poor clinical outcomes. Without considering comorbidities and peri-CPR circumstances [4, 5], the association between hyperglycaemia and neurological outcome might be biased.

For OHCA patients receiving therapeutic hypothermia, Kim et al. [7] noted that elevated BG level in the early post-ROSC phase correlated with poor neurological recovery. However, Losert et al. [21] reported that moderate elevation of BG level at 12 h after ROSC was associated with favourable neurological outcome in non-DM patients. For OHCA patients who survived longer than 48 h after ROSC, Daviaud et al. [6] demonstrated that a high median BG level over this 48-h period was associated with poor neurological outcome. Nevertheless, the mode of death differs between OHCA and IHCA patients, with a higher proportion of patients dying from neurological disability in the former group [22]. The proportion of IHCA patients receiving therapeutic hypothermia is also low [23]. Therefore, it is unclear whether the results of these OHCA studies [6, 7, 21] can be applied to IHCA patients.

Beiser et al. [8] were the first to explore the association between BG level and survival outcome for IHCA by analysing a representative nationwide database. They demonstrated that the patterns of association between BG level and survival outcome were different for DM and non-DM patients. Still, in the study by Beiser et al. [8], maximum and minimum BG levels over the first 24 h after ROSC were analysed separately in the regression analysis, making the results difficult to interpret and apply. Even if BG levels were the same, the survival odds differed depending on the category (i.e. minimum or maximum BG level) to which the value was ascribed, which could only be known retrospectively [8].

We chose to use the mean BG level in our analysis to avoid the issue raised by Beiser et al. [8] Our findings indicate that the patterns of association between BG level and survival were similar to those reported by Beiser et al. [8], demonstrating reverse-U shaped curves (Additional file 6: Figures S3, S4) with lower survival odds for patients with hypoglycaemia or hyperglycaemia. Furthermore, for IHCA patients with DM, a BG level between 183 and 307 mg/dL (10.2–17.1 mmol/L) was significantly associated with favourable neurological outcome; for IHCA patients without DM, the association was not significant, which may be due to inadequate statistical power. Finally, our study demonstrated that the upper bounds of optimal BG levels for both neurological and survival outcomes were higher in DM patients than in non-DM patients. This result corresponds to the findings of Egi et al. [14], who reported that critically ill DM patients were more capable of tolerating the influences of hyperglycaemia than patients without DM, and DM patients experienced significantly lower mortality at all levels of hyperglycaemia. Nevertheless, for both DM and non-DM patients, the upper limits of the optimal BG range were higher than the recommendations of the ERC [10].

The harm caused by hypoglycaemia in critically ill patients is well recognized [11, 12], but the influence of hyperglycaemia is more controversial. Because of the ease of application in statistical analysis and clinical practice, we selected mean BG level for our analysis. Mean values are subject to influences of extreme values and may, therefore, contribute to the elevated optimal BG level ranges. In previous studies, mean [24], median [6], and maximum/minimum [8] BG levels over certain periods have been used in analyses. Whether a single value of BG could be used to assess glucose control over time is unknown. Eslami et al. [25] documented 30 different published methods to describe the quality of glucose control, and the various indicators of glucose control adopted in different studies [4‐8, 20, 21, 24] might be the reason why the AHA [9] does not recommend a specific target BG level for post-ROSC patients. Future study is warranted to establish the optimal indicator to depict and analyse the dynamics and impacts of BG levels.

Despite the limitation of using a single BG value to represent the dynamics of BG levels, the observed role of moderate hyperglycaemia in our study merits further discussion. Sustained hyperglycaemia could worsen secondary brain injury and lead to poor neurological outcome [26, 27]. However, the definition of hyperglycaemia in terms of BG level has not been well established as it relates to brain injury. The human brain depends almost solely on the availability of systemic glucose supply to maintain normal metabolism. After ischemic insults, the transport of glucose to brain tissues may become inadequate to satisfy cerebral metabolism [28]. When cerebral perfusion is compromised, moderate hyperglycaemia may facilitate glucose transport through the elevated BG diffusion gradient that maximizes cellular glucose uptake [29]. Studies have shown that normalization of BG levels in critically ill patients with brain injury may be associated with greater risk of critical reductions in brain glucose levels and energy crises [30, 31]. Therefore, acute stress hyperglycaemia noted during the early post-ROSC phase might be a physiologic, rather than a pathologic, response and attempts at interfering with this complex adaptive response may be harmful rather than protective [32].

Oksanen et al. [13] indicated that, for OHCA patients, a strict BG control strategy would lead to more severe brain injuries. In other critically ill patients, strict BG control has not been shown to reduce mortality, which was probably caused by an increased incidence of hypoglycaemia [33]. On the basis of our results, relaxing the upper limit of BG levels recommended by the ERC may be justified [10]. A randomized clinical trial targeting two or more BG levels for post-arrest patients would be needed to verify this assumption. Also, as recent studies [34‐36] indicated, the outcomes of DM patients following cardiac arrest might be distinct from those of non-DM patients. Future studies should address the implication of DM in the pathogenesis of mortality and morbidity after cardiac arrest.

This study has several limitations that must be considered. First, this was an observational study, which can only establish an association, rather than a causal relationship, between independent and dependent variables. Second, the effects of unmeasured confounders might bias the results. Third, our study did not involve any specific protocol for BG control. We did not suggest that clinicians actively administer intravenous glucose to maintain the BG level at a certain range or that clinicians should allow permissive hyperglycaemia. Administration of intravenous glucose might be associated with worse neurological outcome for post-ROSC patients [37]. The optimal strategy to maintain BG within a certain range should be examined in a randomized controlled trial. Fourth, about half of patients who achieved sustained ROSC were excluded from our analysis because of lack of measurements of BG level after sustained ROSC. As shown in the Additional file 2: Tables S2 and Additional file 3: Table S3, although there were no significant baseline differences between patients with and without measurements of BG levels, the outcomes of the former were better than the latter. Therefore, patients without any measurement of BG levels may have more unstable haemodynamics, leading to the absence of any BG levels measured after sustained ROSC. This kind of selection bias may only be resolved through a prospective study with a protocol for monitoring post-ROSC BG levels. Finally, this was a single-centre study, which might limit the generalizability of the study conclusion. Since the guideline-based care varied significantly across hospitals [38], a multi-centre study would be needed to confirm our observation.