Table 1 Characteristics of included studies

Righettoni [17]

Healthy subjects sampled after overnight fast and after lunch.

8

22-55 years

7 male, 1 female

PTR-TOF-MS: Acetone, ethanol, methanol, isoprene Nano sensing films: Breath acetone

Finger prick measurement with Bayer Contour Blood Glucose Meter

After overnight fast a high correlation between sensors and glucose, and acetone, ethanol, methanol and glucose was found. These high correlations were not found after lunch.

Morning: PTR-TOF-MS: Acetone:0.98 Ethanol:0.9 Methanol: .93 Isoprene:0.00 Nano sensing films: 0.96 Afternoon: PTR-TOF-MS: Acetone:-0.08 Ethanol:0.11 Methanol:-0.16 Isoprene:-0.40 Nano sensing films: −0.02

Storer [18]

T2DM subjects not asked to fast but to refrain from eating. Cross-sectional study

38, T2DM

32-76 years, median age 62

13 male, 25 female

SIFT-MS: Acetone

Finger prick measurement with Abbot Optium Xceed

No strong correlation found between blood glucose and breath acetone. Breath acetone was found to be significantly higher in men.

r = 0.003

Minh [22]

Clamp study. Overnight fast. T1DM subjects were asked not to take long acting insulin.

25 (17 healthy, 8 T1DM)

Healthy: 28 ± 1 years T1DM: 25,8 ± 1,7 years

11 male, 14 female

GCMS: Group A (Ethanol, acetone, methyl nitrate, ethyl-benzene) Group B (2-pentyl nitrate, propane, methanol, ethanol) Room samples collected.

IV catheters in antecubital veins; Beckman Glucose analyzer II

Group A: healthy, mean r of 0.836, T1DM, mean r of 0.950. B: healthy, mean r of 0.829, T1DM, mean r of 0.920.

Healthy: r = 0.8325 T1DM: r = 0.935

Turner [21]

Clamp study. Overnight fast. T1DM subjects

8, T1DM

28 ± 3 years

SIFT-MS: Acetone

IV distal catheter in hand. Hand warmed to arterialize the sample. YSI.

No strong correlation at baseline. Linear correlation between acetone and blood glucose values. Breath acetone decreased when blood glucose decreased. In healthy volunteers the opposite was seen: Low blood glucose values yield high acetone values.

r = 0.816(0.598-0.940)

Lee [20]

Clamp study. Healthy subjects admitted to lab after overnight fast.

10

26 ± 4 years

5 male, 5 female

GCMS: Ethanol, Acetone, Methyl nitrate, ethylbenzene, o-oxylene, m/p-xylene. Room samples collected.

IV catheters in antecubital veins; Beckman Glucose analyzer II

Best 4 gas model: Ethanol, acetone, methyl nitrate, ethyl benzene (mean r of 0.913(0.698-0.977)) 9 samples per patient

r = 0.913 (0.698-0.977)

Fritsch [19]

OGTT. Healthy volunteers admitted after 10 hours fast.

6

24-32 years

5 male, 1 female

Electrochemical analyzer, laser spectrometer, and breath hydrogen: Carbon monoxide measured with Micro smokerlyzer.

Finger prick measurement, Accu check Aviva.

No strong correlation between glucose and carbon monoxide

None

Novak [39]

Clamp study. T1DM subjects admitted after eating light breakfast. Patients on insulin followed normal regimen.

10, T1DM

13,8 ± 0,5 years

7 male, 3 female

GCMS: Methyl nitrate Room samples collected.

IV lines in arms, Blood samples every 30 min. Beckman glucose analyzer II

Methyl nitrate had strongest correlation with blood glucose levels. Correlation increased with 30-minute lag time. Ethanol and Acetone DID NOT correlate with glucose

One subject mentioned, r = 0.99

Galassetti [32]

OGTT. Healthy subjects admitted to research center in morning after overnight fast.

10

27,4 ± 3,1

5 male, 5 female

GCMS: Ethanol and acetone. Room samples collected.

IV catheter. Determined with a quantitative enzymatic measurement.

Multiple linear regression analysis with ethanol and acetone gave an average r of 0.70.

r = 0.700

Paredi [27]

OGTT in 5 patients, CO and glucose measured. Only CO measured in larger cohort

5

33 ± 4 years

3 male, 2 female

Micro smokerlyzer: Carbon monoxide

Finger prick measurement, Reflolux S.

The maximal glucose increase was associated with a significant increase in exhaled CO concentration. Both parameters returned to the baseline at 40 min after glucose administration.

Unknown