Results
In total 230 pregnant women were included. The age and gestation were 34 (30-37) years and 28 (20-35) weeks respectively. Our hospital serves a large ethnically diverse catchment area. The reported ethnicities were; Black 41% (n=94), White 28% (n=63), Asian 8% (n=19), other/Mixed 9% (n=22) and undisclosed 14% (n=31).
Paired fasting and 2-hour glucose data were available for all women. The glucose concentrations in the entire cohort at fasting were 4.7 (4.3-5.0) mmol/L using POC-StatStrip® and 4.5 (4.2-4.8) mmol/L with Lab-VPG. Glucose concentrations at 2-hour were 7.6 (6.6-8.7) mmol/L using POC- StatStrip® and 7.0 (5.8-8.3) mmol/L with Lab-VPG.
The correlation (R) of all glucose concentration from both methods was 0.905. For glucose concentrations at fasting r = 0.871 and for concentrations at 2-hour r = 0.919. However, the correlation between 2-hour POC-StatStrip® concentrations in the range of 7.5-9.5 mmol/L with the corresponding Lab-VPG concentrations was weaker (r=0.72) (Figure 1a, 1b, 1c and 1d).
Bland Altman Plot showed good agreement between the two methods. However, it demonstrated a positive bias of 0.35 mmol/L for POC- StatStrip® with most values scattered within a tight limit of agreement (LoA) (95% LoA from-0.74 mmol/L to +1.44 mmol/L). Bland Altman plot was applied to fasting and 2-hour glucose concentrations separately and showed a positive bias of 0.14 mmol/L at fasting (95% LoA -0.44 to +0.73) and 0.55 mmo/L (95% LoA -0.76 and +1.82) at 2-hours (Figure 2a, 2b and 2c).
Passing Bablok regression analysis showed good agreement between the two methods. Regression equation:POC-StatStrip® (mmol/l) = -0.2182 + 1.091 Lab-VPG, (intercept=- 0.22 and 95% Confidence Interval (CI) -0.44 to -0.04, slope -1.01 and 95% CI 1.06 to 1.13) (figure 3). However, the agreement was less pronounced when the POC-StatStrip® glucose concentration within the range of 7.5 to 9.5 mmol/L was considered separately: POC-StatStrip® mmol/l = 3.829 + 0.5714 Lab-VPG.
NICE, 2015 diagnostic thresholds for GDM were used to compare diagnostic performance between POC-StatStrip® and VPG-Lab. The number (n) and the percentage of women with glucose concentration above the GDM diagnostic thresholds using POC-StatStrip® versus Lab-VPG for the whole test, at fasting and 2-hour OGTT, sensitivity, specificity, FP and FN rates are listed in table (1).
The specificity and NPV for fating POC-StatStrip® glucose ≤ 5.0 mmol/L and 2-hour glucose < 7.5 were 100% while the sensitivity and PPV for fating POC-StatStrip® glucose > 9.5 mmol/L were 100%.
The Youden Index for POC-CBG thresholds were 0.86 and 0.76 for fasting and 2-hour respectively.
ROC curve analysis for fasting and 2-hour POC-StatStrip® glucose are shown in figure 4a and 4b. For fasting glucose, the area under the curve (AUC) was 0.97 (95%: CI 0.96 to 0.99) and for 2-hour, AUC was 0.97 (95% CI: 0.99 to 0.99). The diagnostic accuracy for POC-StatStrip® in our cohort for GDM diagnosis was 84%.