A total of 1858 patients with RHD were consecutively screened in the Guangdong General Hospital (Guangzhou, China) between March 2009 and July 2013. All the patients with RHD received at least 1 valve replacement and preoperative coronary angiography. RHD was diagnosed according to previous acute rheumatic fever and/or symptom of precordial abnormalities and presence of heart murmur, and more importantly based on echocardiographic findings. Patients who committed suicide in hospital, or did not have a postoperative glucose data, or have diabetes were excluded from this study. Subjects with HbA1c ≥ 6.5% or fasting plasma glucose (FGB) ≥ 7.0 mmol/L were excluded as well. Finally, 1395 patients were eligible to be included in our analysis (Fig. 1). The primary endpoint was in-hospital all-cause mortality except for suicide during hospitalization, and the secondary endpoint was one-year mortality after operation and in-hospital major adverse clinical events (MACEs), which were defined as composite end points, such as death, renal failure with dialysis, and stroke.

This study was approved by the Ethics Committee of the Guangdong Provincial People’s Hospital, accompanying with a waiver of informed consent due to the retrospective study design. Oral informed consent was obtained from the patients or their close relatives by telephone, and was recorded by trained nurses during the follow-up period.

Baseline clinical characteristics and laboratory data, in-hospital mortality, and the type of surgery were collected. Blood samples for evaluating glucose levels were collected at admission to the intensive care unit (ICU) after operation. The FBG was measured by using an autoanalyzer (Roche AG, Basel, Switzerland). Other venous blood samples for laboratory analysis were collected in the next morning after admission to the ICU before operation. Follow-up data were obtained from interviewing patients, family members, or primary care physicians by telephone or clinical records of hospital readmissions or outpatient interviews for a period of one year after operation.

SPSS software version 13.0 (SPSS, Inc., Chicago, Illinois) was used for the analyses. Continuous data was presented as mean ± SD or medians and interquartile ranges, then compared by the ANOVA or Wilcoxon rank-sum test accordingly. Categorical data was presented as percentage and compared by χ² or fisher test. Variables whose p value was less than 0.05 in univariate logistic regression analysis were included in the multivariable analysis. Kaplan–Meier curve analysis was performed to evaluated cumulative rate of one-year mortality and compared using the log-rank test. A value of p < 0.05 was considered significant.

Here, 1395 patients (female, 67.1%; male, 32.9%; age, 57 ± 6 years) were participated in this study, including 348 in Q1 group (< 9.3 mmol/L), 354 in Q2 group (9.3–10.9 mmol/L), 341 in Q3 group (10.9–13.2 mmol/L), and 352 in Q4 group (≥ 13.2 mmol/L) (Table 1). Patients with a high postoperative glucose level were significantly older (56.3 ± 5.7 vs. 57.2 ± 5.5 vs. 57.5 ± 5.7 vs. 58.0 ± 5.5, P = 0.001) and were more likely female (39.7% vs. 33.1% vs. 31.7% vs. 27.3%, P = 0.006). No significant differences were found among those 4 groups in the percentages of smoking, hypertension, mitral valve replacement, Tricuspid intervention, coronary artery bypass grafting (CABG), the level of serum creatinine, and Glycated hemoglobin. However, the percentages of NYHA > II, and Aortic valve replacement, the level of FBG, and lg C-eactive protein (lgCRP) were significantly higher in patients with a high postoperative glucose. In addition, the rate of left ventricular ejection fraction (LVEF) was lower in those patients. 47 (3.4%) patients died during the hospitalization, among which 4 (1.1%) were in Q1, 8 (2.3%) were in Q2, 6 (1.8%) were in Q3, and 29 (8.2%) were in Q4 (P<0.001). Furthermore, the incidence of in hospital MACEs (2.0% vs. 3.1% vs. 2.6% vs. 9.7%, p < 0.001) was significantly different among the four groups. Receiver operating characteristic (ROC) curve showed that postoperative BG > 13.0 mmol/L was the best threshold for predicting in-hospital death, accompanying with a sensitivity of 61.7% and specificity of 76.9% (area under the curve (AUC) = 0.707, 95% confidence interval (CI): 0.634–0.780, P < 0.001, Fig. 2).

In the univariable logistic regression analysis, the postoperative BG > 13.0 mmol/L was associated with in-hospital death (odds ratio (OR) = 4.666, 95% CI: 2.559–8.507, P < 0.001). The significant variables were age, NYHA III/IV, estimated glomerular filtration rate (eGFR) < 60 ml/min/1.73 m2, lgCRP, LVEF, TVR, and CABG. These variables were entered into a multiple logistic regression analysis, in which the results showed that postoperative BG > 13.0 mmol/L remained an in- dependent predictor of in-hospital mortality (adjusted OR = 3.418, 95% CI: 1.713–6.821; P < 0.001). In addition, it was revealed that the eGFR <60 ml/min/1.73 m2 (OR = 2.952, 95%CI: 1.311–6.645, P = 0.009) and TVR (OR = 3.855, 95%CI: 1.127–3.52, P = 0.032) were independently associated with in-hospital death (Table 2).

A total of 1248 (89.5%) patients completed one-year follow-up after surgery. However, during this period, 58 patients died. Result of Kaplan–Meier curve analysis is displayed in Fig. 3, illustrating the cumulative 1-year mortality among the four groups. The results suggested that the risk of death was increased for a postoperative BG > 13.2 (log-rank 32.762, P < 0.001) (Fig. 3).