Stress hyperglycemia ratio is associated with systemic inflammation and clinical outcomes in diabetic inpatients with pneumonia on admission

1 INTRODUCTION

Pneumonia and diabetes are both among the top 10 causes of death globally according to the World Health Organization.¹ It was reported by the International Diabetes Federation that 537 million adults are currently living with diabetes worldwide, accounting for 10.5% of the total adult population.² With altered immune response and disturbed airway glucose homeostasis, patients with diabetes are proved to have higher morbidity, severity, and mortality of pneumonia.^{3, 4} In diabetic individuals suffered from pneumonia, management of blood glucose is of crucial importance.⁵ Several international guidelines have been developed to guide the management of blood glucose in diabetes to avoid hyperglycemia and hypoglycemia.^5-9 However, stress hyperglycemia is common in pneumonia, acting as a physiological response to acute illnesses or injuries.¹⁰ In case of pneumonia, blood glucose would reflect a composite glucose level of stress blood glucose and background blood glucose. Unlike patients without diabetes, patients with diabetes are of largely varied background glucose levels. Previous studies also indicated that the association between blood glucose, both hyperglycemia and hypoglycemia, and clinical outcomes was altered by former blood glucose level.^11-13

Recently, several studies highlighted the value of stress hyperglycemia, which presents the absolute glycemia change, in the analyses of prognoses in acute physiological events including pneumonia.^13-21 In these studies, new metrics, including stress hyperglycemia ratio (SHR), glycemic gap, or glucose concentration-to-hemoglobin A_1C ratio (GAR), were used to represent stress hyperglycemia. Among previous studies, much attention was paid to the hazard of severe stress hyperglycemia to clinical outcomes but little was paid to the relative deficiency of stress blood glucose. In addition, management of blood glucose in patients with and without diabetes, in patients from different medical settings, and in patients with different severity of illness is not the same.^5-9 Thus far, studies focused on stress hyperglycemia on diabetic patients with pneumonia on admission were limited and of small sample size. Also, systemic inflammatory status in patients with different stress hyperglycemia level needs further study. Here, we carried out a multicenter study with 1631 diabetic inpatients to explore the association of SHR with systemic inflammation and clinical outcomes in patients with pneumonia of different severity by both linear and non-linear analyses.

2 METHODS

3 RESULTS

3.2 Systemic inflammatory factors

As shown in Figure 1, six systemic inflammatory factors, including white blood cell count (WBC), neutrophil-to-lymphocyte ratio (NLR), C-reactive protein (CRP), procalcitonin (PCT), erythrocyte sedimentation rate (ESR), and D-dimer, were chosen to assess inflammatory levels in different SHR quartiles.²⁶ In total, 1631 data on WBC, 1613 on NLR, 1496 on CRP, 1143 on PCT, 434 on ESR, and 1444 on d-Dimer were available for these patients. Generally, compared with patients in other quartiles, patients in Q4 showed significantly elevated systemic inflammatory level with more WBC (9.1 × 10⁹/L in Q4 vs 7.6 × 10⁹/L in Q1, 7.9 × 10⁹/L in Q2, and 8.0 × 10⁹/L in Q3, p < .001), higher NLR (7.0 in Q4 vs 3.6 in Q1, 3.8 in Q2, and 4.0 in Q3, p < .001), higher CRP (52.8 mg/L in Q4 vs 18.9 mg/L in Q1, p < 0.001; 52.8 mg/L in Q4 vs 28.6 mg/L in Q2, p = .002), higher PCT (0.22 ng/mL in Q4 vs 0.10 ng/mL in Q1, 0.09 ng/mL in Q2, and 0.11 ng/mL in Q3, p < .001), and higher D-Dimer (0.67 mg/L in Q4 vs 0.47 mg/L in Q1, 0.50 mg/L in Q2, and 0.47 mg/L in Q3, p < .001). No significant differences in ESR level were found among groups.

3.3 Associations between SHR quartiles and adverse clinical outcomes

Univariable and multivariable logistic regression analyses were done to determine the association between SHR and adverse clinical outcomes in all patients and patients of different severity classified by CURB-65 scores (Figure 2). In the univariable logistic models, in which only SHR was included as a category variable, SHR was shown to be significantly associated with the composite clinical outcome in all patients and patients with different severity (p < .001). In the multivariable logistic regression models, we adjusted ORs of SHR by age, sex, chronic comorbidities (including chronic lung diseases, coronary heart disease, hypertension, stroke, chronic kidney disfunction, cirrhosis, rheumatic disease, and malignancy), current smoking status, and CURB-65 scores. As shown in Figure 2, SHR was still a significant predictor for composite end point in all group sets (p < .001). Taken Q2 as a reference, patients in Q3 and Q4 were of higher risk of developing adverse clinical outcomes (OR: 2.15, 95% CI: 1.18–3.94 for Q3 in all patients, p = .013; OR: 4.77, 95% CI: 2.75–8.27 for Q4 in all patients, p < .001; OR: 3.60, 95% CI: 1.59–8.15 for Q4 in patients with CURB-65 ≤ 1, p = .002; OR: 2.80, 95% CI:1.20–6.56 for Q3 in patients with CURB-65 ≥ 2, p = .018; OR: 5.57, 95% CI: 2.57–12.10 for Q4 in patients with CURB-65 ≥ 2, p < .001). What is more, in patients with CURB-65 ≥ 2, Q1 possessed a slightly but significantly increased risk over Q2 (OR: 2.46, 95% CI: 1.02–5.96, p = .046).

3.4 Nonlinear correlations between SHR and adverse clinical outcomes

Given the results of logistic regression analyses using SHR quartiles, we hypothesized that there was a nonlinear correlation between SHR and composite end point. Restricted cubic splines based on univariable and multivariable logistic regressions were explored to determine whether a nonlinear correlation existed or not. As displayed in Figure 3, nonlinear correlations were found in all patients in the univariable analyses (p-nonlinear < .001) and in all groups in the multivariable analyses (p-nonlinear = .002 in all patients; p-nonlinear = .032 in patients with CURB-65 ≤ 1; and p-nonlinear = .029 in patients with CURB-65 ≥ 2). In view of the fact that there was no commonly accepted threshold of low SHR, we roughly divided SHR into two parts based on median 0.889. In the univariable and multivariable logistic regression analyses, we found that when SHR was under median, risk of reaching composite endpoint significantly increased as SHR decreased; and when SHR was over median, risk of reaching composite endpoint significantly increased as SHR increased (Table 2).

TABLE 2. Unadjusted and adjusted odds ratios of SHR for adverse clinical outcomes in diabetic inpatients with pneumonia on admission.

Variable	Univariable logistic regression model		Multivariable logistic regression model
	Odds ratio (95% CI)	p value	Adjusted odds ratio (95% CI)	p value
All patients
SHR ≤ 0.889	0.09 (0.01–0.76)	.026	0.02 (0.00–0.84)	.040
SHR ≥ 0.889	8.31 (4.64–14.86)	.000	4.78 (2.53–9.05)	.000
Patients with CURB-65 ≤ 1
SHR ≤ 0.889	0.10 (0.01–1.86)	.122	0.03 (0.00–0.90)	.043
SHR ≥ 0.889	8.46 (2.98–24.01)	.000	9.53 (3.02–30.07)	.000
Patients with CURB-65 ≥ 2
SHR ≤ 0.889	0.06 (0.00–1.50)	.086	0.02 (0.00–0.84)	.040
SHR ≥ 0.889	4.31 (2.10–8.88)	.000	3.38 (1.58–7.22)	.002

• Note: Statistical analyses were made by univariable and multivariable logistic regression. In the univariable logistic regression model, only SHR was included. In the multivariable logistic regression model, SHR was included along with age, sex, chronic comorbidities (including chronic lung diseases, coronary heart disease, hypertension, stroke, chronic kidney disfunction, cirrhosis, rheumatic disease, and malignancy), current smoking status, and CURB-65 scores.
• Abbreviations: CI, confidence interval; CURB-65, score for pneumonia severity (Confusion, blood Urea nitrogen, Respiratory rate, Blood pressure); SHR, stress hyperglycemia ratio.

4 DISCUSSION

Our study offered a clear and intuitionistic picture of the inflammatory responses and nonlinear association between SHR and adverse clinical outcomes in diabetic inpatients with pneumonia of different severity. Also, our study contained the largest number of patients with diabetes compared with studies of the same topic so far. The findings were as follows. Patients in the highest SHR quartile were of significantly elevated systematic inflammatory biomarkers, indicating inflammatory overreaction in these patients with diabetes. There were J-shaped associations of SHR with adverse clinical outcomes in patients of different severity classified by CURB-65 scores, even when those with absolute hypoglycemia were already excluded from the analyses. The hazard of low SHR was more obvious in patients with CURB-65 ≥ 2 than those with CURB-65 ≤ 1. These results suggested that FBG alone was not sufficient enough for the evaluation of insufficient blood sugar supply, especially in those with severer pneumonia. Predictive value of SHR for adverse clinical outcomes was greater than that of FBG and was further elevated when the nonlinear association was taken into consideration. Our study suggested the need to include stress hyperglycemia in the blood glucose management of diabetic inpatients with pneumonia, especially for the prevention of potential hypoglycemia and recognition of insufficient blood sugar supply in those of severe pneumonia or high HbA_1C.

WBC, NLR, CRP, PCT, ESR, and D-dimer are commonly used inflammatory biomarkers in pneumonia, reflecting the inflammation intensity or degree of lung injury.²⁶ In face of stress and infection, complex interactions exist between stress hyperglycemia and inflammation.²⁷ Briefly, stress hormones regulated by the hypothalamic–pituitary–adrenal (HPA) axis and the sympathoadrenal system profoundly modify the inflammatory response while raising blood glucose.^{27, 28} Glucose in itself is also a proinflammatory molecule, and inflammatory mediators can in turn downregulate glucose transporter-1 and aggravate insulin resistant.^{27, 28} In our study, patients in Q4 showed markedly elevated level in these commonly used inflammatory indicators (Figure 1). These results suggested that patients in Q4 were under strong, or excessive, inflammatory response. Mondal et al found that level of CRP, lactate dehydrogenase, and D-dimer were higher in coronavirus disease 2019 inpatients with SHR≥1.14.¹⁴ Data from Su et al also showed elevated neutrophil counts in patients admitted to intensive care unit (ICU) with higher GAR.¹⁸ Though hypoglycemia was also taken as a kind of stress, no difference was found in the level of inflammatory biomarkers between patients of Q1 (the lowest SHR) and those of Q2 or Q3 (the medium SHR), and so did it in previous studies of similar topics.^{10, 17, 21, 27} This may be explained by the exclusion of patients with absolute hypoglycemia on admission in our analyses and the relatively low severity of pneumonia in Q1 patients.

Associations of SHR with adverse clinical outcomes in diabetic patients with pneumonia were explored by logistic regressions, in which potential confounding factors with complete available data were taken into consideration. SHR was proven to be independently associated with adverse clinical outcomes in patients with different severity on admission in our analyses (as shown in Figures 2 and 3), in line with previous studies on pneumonia or other acute diseases.^13-21 The strong association of excessive stress hyperglycemia and adverse clinical outcomes was quite obvious in our study and previous ones. Stress hyperglycemia is considered as an survival response to ensure an adequate glucose supply for non-insulin dependent tissues.¹⁰ Generally, the HPA axis and the sympathoadrenal system will be activated to certain level according to the degree of stress,¹⁰ and severe stress hyperglycemia can lead to hypertonic state and cause fluid transfer, thus leading to renal impairment and volume depletion.¹⁰ According to the results of our study and previous ones, the harm of excessive stress hyperglycemia could also be partly explained by excessive inflammatory response.^{14, 18} Adequate blood glucose and inflammatory response are indispensable for fighting against pathogens.²⁹ When blood glucose or inflammatory response was out of control, a vicious cycle could be set up.²⁷

Several previous studies displayed the hazard of relative low stress hyperglycemia.^{13, 16, 17, 21} Among these three studies, one owas conducted on community acquired pneumonia (CAP) and two on critical illness. The study on CAP pointed out that patients with the lowest and highest quartiles of glycemic gap both had increased risk of 90-day mortality in all patients after adjusted by status of diabetes, whereas no association was found when admission glycemic gap was taken as a continuous variable.¹⁶ We used restricted cubic splines for nonlinear analyses and made up for its shortcomings in the analysis with stress hyperglycemia as a continuous variable.The former study made the point that diabetic status had no influence on the relation between glycemic gap and 90-day mortality.¹⁶ However, the author of a previous study in ICU pointed out that there appeared to be an elevated mortality in the lowest SHR group among patients with HbA_1C ≥ 6.5% but not among those with HbA_1C < 6.5%.¹³ Because HbA_1C level was quite different in people with and without diabetes, the association of SHR with clinical outcomes might be different in diabetic and nondiabetic people. Another study in ICU revealed a U-shaped association between the glycemic gap and mortality.¹⁷ Here, we clearly displayed the ubiquitous J-shaped relationships between SHR and adverse clinical outcomes in patients with pneumonia of different severity. Tthe hazard of Q1 was more obvious in patients with CUBR-65 ≥ 2. Because CURB-65 scores and inflammatory indicators of patients in Q1 were comparable with those in Q2, the increased risk of adverse clinical outcome might not be resulted from severer illness or stronger inflammatory response in our study, which was supposed as a probable reason in another study conducted among ICU patients.¹⁷ Because we noticed that Q1 patients had higher HbA_1C levels and higher rates of insulin usage for the routine hypoglycemic treatment, we inferred that patients in Q1 were a group of diabetic patients with poorer blood glucose control. Similarly, two previous studies, which found the J-shaped or U-shaped association between stress hyperglycemia and adverse clinical outcomes, both presented significantly higher HbA_1C level in patients with the lowest part of stress hyperglycemia.^{17, 21} It was reported previously that patients with high chronic premorbid hyperglycemia were of increased risk for hypoglycemia in ICU.¹² So we supposed that the possible reason low SHR was associated with worse clinical outcomes could be the increased risk of developing hypoglycemia in the future.⁷ Risk of hypoglycemia is a major concern in the management of patients with diabetes in all guidelines.^5-8 Besides, in face of acute illnesses or injuries, graded responses to the degree of stress are needed and sufficient stress hyperglycemia is essential for immune and cerebral function.^{10, 27} So, another possible reason might be the relatively glucose deficient in these people. Glycemic management in diabetic inpatients should be modified according to previous blood glucose level and disease severity. A recent CONTROLING study reported that there was no survival benefit in ICU patients when blood glucose of patients was controlled to preadmission level compared with <180 mg/dL.³⁰ As suggested in that study, further studies targeting a glycemia over the usual, which matched the stress-induced hyperglycemia, were needed to test the benefit of individualized blood glucose control in inpatients.³⁰ Indexes that can reflect stress-induced hyperglycemia, SHR for instance, can well fit this need. According to restricted cubic splines based on logistic regression analyses, there was a level of SHR with the lowest OR of adverse clinical outcomes in diabetic inpatients with pneumonia. The nadir value might give insight into the individualized blood glucose management in diabetic inpatients with pneumonia.

What is more, SHR was considered a better predictive factor for adverse clinical outcomes in acute illness or injuries compared with blood glucose.^{14, 15, 18} Without the commonly accepted threshold of low SHR, we used SHR as a spline term, as quartiles, and a continuous variable to analyze its predictive value (Table 3). We found that the predictive value of SHR was improved when taking the J-shaped association into consideration, especially in patients with more severe pneumonia. These results suggested that there seemed to be a range, with both lower and upper limits, of SHR that was associated with favorable clinical outcomes in diabetic patients with pneumonia on admission.

It was indicated in a previous study that the distribution of a risk factor in the study population greatly influenced its nadir value in a U- or J-shaped relationship between a particular risk factor and a future health outcome, as well as the exact hazard ratio or OR.³¹ Consequently, considering the sample size of this study, we did not point out the nadir value or make further investigation on the exact range. Further large-scale studies are needed to find this nadir value and favorable range of SHR and make them better serve in the prediction of adverse clinical outcomes in diabetic inpatients with pneumonia.

There are certain limitations in the present study. First, data were retrospectively collected from medical records and missing data of certain degree existed, especially on HbA_1C and CURB-65 scores. The sample size and the data incompleteness made it hard to address other potential confounding factors in this study, such as body mass index, vaccination status, insulin treatment, diabetes-related complications, etc. Further prospective researches are needed to remedy this shortcoming. Second, CURB-65 score is a simple but rough evaluation method of pneumonia severity. Third, we presented only the J-shaped association of SHR with adverse clinical outcomes but did not provide the exact cutoff points of SHR and other evaluation indicators in the prediction model except AUC. Fourth, restricted by the low positive rate of pathogen detection, limited number of participants with type 1 diabetes, and small number of patients with adverse clinical outcomes, we needed to use a composite end point in the prognostic analysis, as in a previous study³² and did not verify the applicability of our findings with regard to cause of pneumonia, type of diabetes, and separate adverse clinical outcome. Further studies with larger sample sizes or clinical trials are needed to verify the benefits of SHR, or other forms of stress-induced glycemia, in the individualized management in diabetic patients with pneumonia or other acute illnesses.

In conclusion, we found that high SHR was associated with elevated systematic inflammation in diabetic inpatients with pneumonia on admission. J-shaped associations of SHR with adverse clinical outcomes commonly existed in diabetic inpatients with pneumonia of different severity. The inclusion of SHR in the blood glucose management of patients with diabetes may be beneficial, especially for the prevention of potential hypoglycemia or the recognition of glucose insufficient in those of severe pneumonia or high HbA_1C.

AUTHOR CONTRIBUTIONS

Min Zhou conceived and designed the study. Bing Liu, Yu Chen, and Liping Yu collected and analyzed the data. Bing Liu drafted the manuscript. Min Zhou revised the manuscript. All authors read and gave final approval of the version to be submitted.

FUNDING INFORMATION

This work was supported by the National Key R&D Program of China (Grant Nos. 2017YFC1309700 and 2017YFC1309701), Shanghai Municipal Key Clinical Specialty (shslczdzk02202), Shanghai Top-Priority Clinical Key Disciplines Construction Project (2017ZZ02014), Shanghai Key Laboratory of Emergency Prevention, Diagnosis and Treatment of Respiratory Infectious Diseases (20dz2261100), and Cultivation Project of Shanghai Major Infectious Disease Research Base (20dz2210500).

CONFLICT OF INTEREST STATEMENT

The authors declare no conflicts of interest in this work.