Fasting Blood Glucose and COVID-19 Severity: Nonlinearity Matters

Novel Communications in Diabetes

¹Department of Environmental Health, Harvard T.H. Chan School of Public Health, Harvard University, Boston, MA
²Dasman Diabetes Institute (DDI), Dasman, Kuwait
³Department of Mathematics, Faculty of Sciences, Kuwait University, Kuwait City, Kuwait
⁴Department of Medical Laboratory Sciences, Faculty of Allied Health Sciences, Health Sciences Center, Kuwait University, Kuwait City, Kuwait

Diabetes Care 2020 Dec; 43(12): 3113-3116. https://doi.org/10.2337/dc20-1941

Article Figures & Tables

Download figure
Open in new tab
Download powerpoint
Figure 1
Exposure-response relationship between FBG levels and the log odds of being admitted to the ICU. The crude relationship (top) and a fully adjusted model (bottom) are shown. The fully adjusted model is adjusted for age, sex, smoking, nationality, diabetes status, hypertension, asthma, cardiovascular disease, malignancy, and renal failure. Tick marks above the horizontal axis indicate individual cases.

Table 1

Odds ratios of different functional forms of FBG and ICU admission

Fasting blood glucose^*	Odds ratio	95% CI	P value
Linear
Per 1 mmol/L increase	1.59	1.3–1.89	<0.001
Per 5 mmol/L increase	10.45	4.8–22.71	<0.001
Dichotomous
<11 mmol/L	Reference
≥11 mmol/L	11.91	4.9–30.81	<0.001
<7 mmol/L	Reference
≥7 mmol/L	14.57	6.87–32.59	<0.001
Categorical
<5.5 mmol/L	Reference
5.5 to 6.9 mmol/L	1.69	0.63–4.05	0.270
≥7 mmol/L	19.21	7.80–51.46	<0.001
Nonlinear
10 mmol/L vs. 5 mmol/L	36.02	23.63–54.91	—
15 mmol/L vs. 10 mmol/L	2.26	1.55–3.30	—
20 mmol/L vs. 15 mmol/L	1.79	0.84–3.79	—

↵* All models were adjusted for age, sex, smoking, nationality, diabetes status, hypertension, asthma, cardiovascular disease, malignancy, and renal failure.