Type 2 diabetes mellitus and sepsis: state of the art, certainties and missing evidence

Sepsis is defined as a “life-threatening organ dysfunction caused by dysregulated host response to an infection”, and represents a leading cause of death worldwide, with a mortality rate > 10% [1]. In 2017, almost 50 million incident cases of sepsis were estimated worldwide and 11 million sepsis-related deaths were reported, representing nearly 20% of all global deaths [2]. Septic shock is highly prevalent in the general population, occurring in the 8–10% of Intensive Care Unit (ICU) patients, with a high mortality rate (almost 40%) [3]. The expanding elderly population suffering from extensive comorbidity burden, physiological frailty and immune senescence [4] leads to predict an increased mortality rate for sepsis over the next couple of decades [5].

With the rising globalization of Western diet and lifestyle, the incidence of Type 2 Diabetes Mellitus (T2D) is increasing and its prevalence is expected to exceed 700 million worldwide in the near future, reaching pandemic proportions [6]. T2D and diabetes-related complications are also a leading cause of hospitalization, disability and mortality [7, 8].

Although still under debate [9‐11], several lines of evidence indicate that diabetic patients have an increased risk of infection [9, 12‐17], and a 2 to 6 times higher risk of sepsis compared to the age-matched non-diabetic people [12, 17]), and higher sepsis-related morbidity and mortality compared to non-diabetic individuals [12, 15, 18, 19]. Diabetic patients are also likely to have higher rates of colonization by resistant pathogens, including methicillin-resistant Staphylococcus aureus, than non-diabetics [20]. These considerations support the finding that diabetes is an increasingly common comorbidity among septic patients [21, 22]. As a matter of fact, during a 25-year study period (1979–2003), sepsis occurred in 12.5 million of 930 million acute-care hospitalizations, and diabetes was reported in 17% of cases [21]. Moreover, diabetic patients account for the largest population experiencing post-sepsis complications and rising mortality [15].

Despite current improvements in diagnosis and treatment options, diabetes and sepsis remain common, costly and lethal worldwide [3, 11, 14]. This work aims at reviewing the current state of knowledge about: (1) the impact of diabetes and sepsis on the immune system, (2) the influence of diabetes on the risk of sepsis and its outcomes, and (3) the optimal target for blood glucose control during sepsis in patients with diabetes.

A MEDLINE/PubMed search was conducted from inception to December 31, 2020, using the MeSH terms Diabetes mellitus AND Sepsis AND the following: Immune system processes, Glycated hemoglobin, Insulin, Hypoglycemic agents, Metformin, Sulphonylurea compounds, Thiazolidinediones, Incretin, Multiple-organ dysfunction syndrome, Lung injury, Acute respiratory distress syndrome, Acute kidney failure, Blood glucose, Mortality.

All types of publications and articles related to human studies were initially included. Out of 583 records retrieved through the initial database search, 425 remained after removing duplicates. After manual assessment based on title/abstract, 150 remained for full-text assessment for eligibility. Articles without full text or not written in English, case reports and studies involving generically critically ill patients or patients with specific infective focus were excluded. Based on these exclusion criteria, 92 records were excluded, while 58 articles remained. An additional 46 records were identified by manual search among the references cited in these records and further assessed for eligibility according to the above-mentioned criteria, leading to exclude 27 and include 19.

Finally, 77 studies were included in the qualitative analysis (Fig. 1).

T2D is a complex clinical syndrome characterized by persistent hyperglycemia, associated with decreased insulin secretion and sensitivity [13]. Several metabolic abnormalities, including inflammation and insulin resistance driven by both chronic and stress-induced hyperglycemia, and T2D-related obesity and dyslipidemia, additionally worsen the host response against infections.

Also, sepsis exerts a global impact on the immune system, impairing the lifespan, generation and function of innate and adaptive immune cells and leading to perturbation of the immune homeostasis [23].

Currently, the molecular network that cooperates to worsen clinical outcomes in patients with T2D and sepsis remains uncertain [15]. Figure 2 summarizes the current knowledge on the mechanisms of sepsis and the effects of chronic hyperglycemia, both impacting on the immune system and translating into poor patient outcome [13, 15].

Progression of sepsis is associated with changes in insulin and cortisol circulating levels, resulting in significant glucose perturbations, organ damage and activation of the immune system [38]. Besides the well-known stress-induced hyperglycemia, hypoglycemia may also reflects a pathological acute stress response. Indeed, hypoglycemia is commonly associated with sepsis and considered an epiphenomenon of severe organ dysfunction preceding death. Although the mechanisms and relationships between hypoglycemia and the severity of the disease in septic patients are still debated, the role of inflammatory cytokines has been proposed [39].

In critical settings, derangement of glycemic control is associated with more severe disease and poorer prognosis [39‐41]. However, diabetes may modulate the relationship between dysglycemia and mortality in sepsis [40]. Indeed, the risk of mortality associated with hyperglycemia is lower in diabetic than non-diabetic patients [42] and is not influenced by hypoglycemia [39] or glycemic variability [43].

Despite strong recommendations for early insulin administration, how to monitor and treat stress-induced hyperglycemia remains under debate [41, 44].

Several large-size trials have investigated the optimal acute blood glucose control in critically ill patients, including septic ones [22, 45‐47]. However, only a few small studies were restricted to septic patients [48, 49], and none specifically targeted diabetic patients. Table 1 reports the main clinical trials evaluating the impact of different targets of acute glycemic control in critically ill and septic patients. Van den Berghe et al. [45] first evaluated patients admitted to surgical Intensive Care Units (ICU) who were randomly assigned to receive intensive insulin therapy (blood glucose target 80–110 mg/dl) or conventional therapy (target 180–200 mg/dl). Although the number of septic patients was not reported at baseline, intensive insulin therapy reduced episodes of nosocomial septicaemia of about 46% and the proportion of patients requiring prolonged antibiotic therapy. Specifically, a tight glucose control (TGC, i.e., blood glucose levels < 110 mg/dl) was associated with lower morbidity and mortality rates (with a 43% relative risk reduction of ICU mortality). However, this result relies of the benefit obtained in the subgroup of those patients staying in ICU for more than 5 days and in cardiac surgical patients (accounting for the majority of the study population) who previously received intravenous glucose load for nutritional purpose. In a subsequent study in medical ICU patients, the same group [46] failed to confirm a benefit on mortality in the overall population, since demonstrated that TGC prevents morbidity in all patients, but reduces mortality only in those staying in the ICU for at least 3 days. Moreover, concerns raised on the high rates of hypoglycemic events (more than sixfold higher than the previous study) in this subgroup of patients. A post-hoc analysis [50] of pooled data from the two Leuven studies [45, 46] further confirmed that TGC carried a significantly higher risk of hypoglycemia (which occurred in 11.3% of patients on TGC vs. 1.8% of those on conventional insulin therapy, p < 0.0001). However, even if hypoglycemia was not associated with early deaths and/or neurological sequelae, a higher risk of death was reported. Such pooled data finally revealed that TGC significantly reduced morbidity and mortality in mixed medical/surgical ICU (particularly in patients staying in ICU at least 3 days). In addition it was reported that all patient subgroups, including those admitted for sepsis, benefit from TGC. Only for diabetic patients, no survival benefit was reported. A rapid normalization of blood glucose levels rather than hypoglycemic events has been proposed to explain the lack of TGC benefit in diabetic patients.

However, further studies failed to confirm these benefits from TGC [22, 46‐48], although differences in study design, selection of patients, nutritional support, targeted glucose range and blood glucose measurements make the comparison challenging [41]. As a matter of fact, a further trial specifically involving patients with severe sepsis [48] not only failed to demonstrate a benefit on mortality from TGC, in both diabetic and non-diabetic patients, but was early stopped for safety reasons (e.g., a significantly increased rate of severe hypoglycemic events). Two further large-scale trials including mixed populations of medical and surgical patients, the Glucontrol study [47] and the Normoglycemia in Intensive Care Evaluation—Survival Using Glucose Algorithm Regulation (NICE-SUGAR) Trial [22], reported higher rates of hypoglycemia in the TGC group. The former, prematurely stopped for the high rate of unintended protocol violations, did not find differences in mortality from TGC, while the second one revealed that a less stringent glycemic control translates into lower mortality rate, regardless of diabetes. Finally, a recent meta-analysis by Yamada et al. [51] confirmed the absence of clinical benefits of a stringent glycemic control in term of mortality, while reporting an increased rate of hypoglycemia in both diabetic and non-diabetic patients on TGC compared to patients on mild (140–180 mg/dl) and very mild control (180–220 mg).

The U-shaped curve describing the relationship between glycemic control and mortality (patients with low and high glucose levels have worse outcomes than those in the normal/moderate range) suggests that moderately elevated glycemic level may represent the ideal target in diabetic patients [9, 15, 52]. However, whether this effect was actually due to such glucose levels or to confounding variables driving hypoglycemia and poor outcome is still a matter of debate [52]. Additionally, the relationship between longer time in blood glucose range 70 to 140 mg/dl and lower mortality rate, clearly described in non-diabetic patients, is missing in diabetics [53].

Current guidelines recommend to treat hyperglycemia in critical patients to a target of 140–180 mg/dL, regardless of the presence of previously known diabetes [54, 55]. The need for specific targets of glycemic control in diabetic patients has been postulated [44, 56, 57], and some studies have suggested that less stringent glycemic control (e.g., targeting blood glucose levels at 180–250 mg/dL) may be beneficial in critical patients with premorbid chronic hyperglycemia (e.g., HbA1c level > 7% or > 53 mmol/mmol). However, concerns (including increased risk of infection, glycosuria and polyneuropathy) have raised against such permissive glucose levels in critically ill diabetic patients [58]. Based on these observations, Egi et al. [58] proposed to adopt a uniform blood glucose target for patients with and without diabetes (≤ 180 mg/dL), at least until the randomized control LUCID trial (Liberal GlUcose Control in Critically Ill Patients with Preexisting Type 2 Diabetes trial) [59] will inform on the risks and benefits of more liberal glucose control strategies.

Finally, a role was suggested for closed-loop glucose control systems and immunomodulatory treatment options, to avoid hypoglycemia during insulin therapy and to control the rise in circulating cytokine levels in diabetic patients with severe sepsis and septic shock [60].

The occurrence of organ dysfunction in diabetic patients with sepsis was first evaluated in a cohort of 12.5 million people admitted to hospital for sepsis between 1979 and 2003, among whom over 2 million had diabetes [21]. The study revealed that diabetic patients were less likely to develop acute respiratory failure (9% versus 14%, p < 0.05), regardless of the source of infection, but more likely to develop acute renal failure (13% versus 7%, p < 0.05) than non-diabetic ones. No differences were found in dysfunction of other organs (cardiovascular failure occurred in the 4% of the overall population, while hepatic, hematological, metabolic and central nervous system dysfunction globally occurred in the 6%).

In both sepsis and chronic hyperglycemia, injury of the glycocalyx, due to generation of reactive oxygen species and inflammatory mediators, impacts the microcirculation and leads to organ damage [61]. The coexistence of diabetes and severe sepsis additionally compromises red blood cell deformability, worsens the microcirculation and hastens the progression of organ dysfunction [62].

A recent retrospective observational study [61] supports the link between premorbid chronic hyperglycemia and progression to organ dysfunction in septic patients. The authors demonstrated that, in septic patients admitted to the ICU, HbA1c values ≥ 6.5% (≥ 47.5 mmol/mmol) were independently associated with the progression of liver and kidney failure within 72 h, and with ICU mortality. Only, a positive trend for the progression of lung and cardiac dysfunction and clotting disorders was reported.

Unlike lower risk of acute respiratory dysfunction [14], higher risk of acute kidney injury in diabetic patients was confirmed in a large nationwide retrospective study [63] and in a recent meta-analysis [14].

Increased susceptibility to infection and sepsis in diabetes is extensively documented [9, 12‐17], but equivocal results from epidemiological studies pose doubts on the association between diabetes and increased risk of infection-related morbidity and mortality [9‐11]. Different study populations (including lack of stratification into Type 1 and Type 2 Diabetes, different adjustments for comorbidities, sepsis etiology and disease severity), drug administration regimens to control blood glucose and methods to measure outcomes have been proposed to explain this heterogeneity [10, 11].

Table 2 reports recent clinical studies investigating the association between diabetes and mortality for sepsis. Among them, a large-size observational study [72] demonstrated that diabetic patients experienced an increased mortality from infectious diseases (persisting even after adjustment for comorbidities) and a twofold increased risk of mortality for sepsis compared to the general population. Additionally, two large-size retrospective cohort studies [12, 19] found higher mortality rate related to infections in diabetic compared to non-diabetic patients, whereas others [33, 34, 63, 66, 70, 73, 74] failed to demonstrate such association, and Esper et al. [21] even reported improved survival in diabetic patients.

Few observational studies have investigated the link between premorbid glycemic state and sepsis outcome, showing that HbA1c levels at admission are in direct correlation with hospital mortality in diabetic patients with sepsis [37, 61].

The results by Tayek et al. [75], firstly reporting a global benefit on mortality from diabetic status, were confirmed in the meta-analysis by Wang et al. [14], which demonstrated that diabetes is not associated with adverse outcomes in patients with sepsis, while beneficial. As a matter of fact, some studies notably demonstrated an association between hyperglycemia and increased mortality in non-diabetic patients, unlike in diabetic patients, suggesting that acute hyperglycemia may drive different pathophysiologic effects in diabetic and non-diabetic patients. Nevertheless, whether the link between hyperglycemia and mortality in non-diabetics relies on hyperglycemia-induced toxic effects or is the hallmark of disease severity still remains to be clarified [9]. Although the mechanisms for such protective effect driven by diabetes remain uncertain, previous exposure to high glucose has been proposed to enhance immune adaptation and to induce benefits [9, 14, 75]. The role of inflammation has been also investigated in this context. In particular, Stegenga et al. [73] reported comparable cytokine levels and procoagulant responses in critical septic patients with and without preexisting diabetes, while a different study unveiled the presence of elevated levels of markers of endothelial cell activation in patients with diabetes and septic shock, compared to patients without diabetes [76]. Beneficial effects of insulin administration, prevention of acute lung injury, adaptation to previous oxidant stress and nutritional intake in obese patients with diabetes were also proposed as protective against sepsis [9].

Sepsis represents a rising cause of mortality worldwide and diabetes is a common and increasing comorbidity in septic patients. Although the higher risk of infection in diabetic patients is well documented, the impact of diabetes on the outcome of sepsis and the mechanisms underlying their interactions are still debated.

Critical issues that need clarifying include the impact of diabetes and sepsis on the immune system, the role of glycemic control and the potential protective role of anti-diabetic treatments, on the occurrence of sepsis and its outcome, including the risk of renal failure and acute respiratory dysfunction. Also, recommendations for glycemic targets during sepsis do not stand on firm grounds.

Further large-size prospective studies, randomized controlled trials whenever possible, specifically including diabetic patients with sepsis instead of generically critically ill or patients with specific infective focus, could clear some of these unsolved questions, including the risk/benefit balance of more liberal acute glycemic control.

Finally, interesting and challenging therapeutic options, including immunomodulatory approaches targeting the pathways activated in T2D and sepsis, are under investigation and may result in clinical benefits.