Comparing AOMM™ and MATLAB metabolic calculations.

Martina Drecogna · July 2026

University of Padova and University of Pavia, Italy

The objective of this study is to validate the AOMM™ non-OMM indices against an independent mathematical pipeline implemented in MATLAB.

PDF

Dataset and methods

Dataset

The validation framework utilizes a dataset which includes 999 individual records. For each subject, the dataset provides unique identification, demographic and anthropometric characteristics (age at visit, BMI, weight), and plasma glucose, insulin and C-peptide, temporal tracking across a 120-minute OGTT. The physiological kinetic curves include a baseline sample (t = −10 min) and five sampling points at 0, 30, 60, 90, and 120 minutes.

Preprocessing

The preprocessing phase involves unit standardization with AOMM™, which requires plasma glucose expressed in mg/dL, insulin in µU/mL, and C-peptide both in ng/mL and pmol/L for different index calculations. While the raw glucose and insulin profiles were already provided in the required standard units, the C-peptide values were originally reported in ng/mL. Consequently, these values were also converted into pmol/L by multiplying them by 331.

Methods

To systematically verify the mathematical formulas implemented by the AOMM™ software, an independent validation pipeline was developed in MATLAB, evaluating five distinct parameter groups.

1. Static fasting indices

HOMA-IR = \frac{G_{0} \cdot I_{0}}{405}

HOMA-β = \frac{360 \cdot I_{0}}{G_{0} - 63}

QUICKI = \frac{1}{\log_{10} (I_{0}) + \log_{10} (G_{0})}

Where G₀ represents fasting glucose and I₀ denotes fasting insulin, both measured at the baseline of the test (t = 0 min).

2. Dynamic empirical indices

Matsuda ISI = \frac{10,000}{\sqrt{G_{0} \cdot I_{0} \cdot \overline{G} \cdot \overline{I}}}

IGI = \frac{I_{30} - I_{0}}{G_{30} - G_{0}}

Matsuda disposition index = IGI \cdot Matsuda ISI

C-peptide index = \frac{{CP}_{30} - {CP}_{0}}{G_{30} - G_{0}}

ISSI-2 = \frac{{AUC}_{I}}{{AUC}_{G}} \cdot Matsuda ISI

Here, $\overline{G}$ and $\overline{I}$ denote the mean concentration of glucose and insulin across the OGTT, while G_t, I_t and CP_t represent the glucose, insulin, and C-peptide concentrations recorded at time point t (C-peptide in pmol/L). Additionally, the composite IGI/HOMA-IR ratio index was computed.

3. Incremental area under the curve (iAUC)

{iAUC}_{G} = 15 (G_{0} + G_{30}) + 15 (G_{30} + G_{60})

+ 15 (G_{60} + G_{90}) + 15 (G_{90} + G_{120}) - 120 \cdot G_{0}

{iAUC}_{I} = 15 (I_{0} + I_{30}) + 15 (I_{30} + I_{60})

+ 15 (I_{60} + I_{90}) + 15 (I_{90} + I_{120}) - 120 \cdot I_{0}

{iAUC}_{CP} = 15 ({CP}_{0} + {CP}_{30}) + 15 ({CP}_{30} + {CP}_{60})

+ 30 ({CP}_{60} + {CP}_{120}) - 120 \cdot {CP}_{0}

Trapezoidal rule over 0–120 min; C-peptide in pmol/L.

4. Other glycemic indices

G_{60} = G (60)

G_{peak} = \max (G)

t_{peak} = t (\max (G))

G_{range} = \max (G) - \min (G)

G_{120} = G (120)

5. T1D progression indices

{Index}_{60} = 0.3695 \cdot \ln ({CP}_{0}) + 0.0165 \cdot G_{60} - 0.3644 \cdot {CP}_{60}

DPTRS = 1.571 \cdot \ln ({CP}_{0} + 1) + 0.115 \cdot G_{120} - 0.193 \cdot \frac{{CP}_{60}}{{CP}_{0}}

+ 0.0019 \cdot age + 1.5 \cdot \ln (BMI)

C-peptide in ng/mL.

Results

Nineteen scatter plots of AOMM versus MATLAB values, one per parameter, each with a red dashed identity line. All points fall on the line (R = 1.0000) except incremental C-peptide AUC (R = 0.9835), which shows minor scatter at higher values. — Figure. AOMM™ (x-axis) versus MATLAB (y-axis) for each of the 19 parameters (n = 999 records); red dashed line: identity (y = x).

The level of agreement between the two methodologies was evaluated using correlation scatter plots, plotting AOMM™ reference values on the horizontal axis (x-axis) against the MATLAB calculations on the vertical axis (y-axis), with the theoretical identity line (y = x) superimposed in red. In this graphical framework, a perfect linear convergence of the 999 data points along the identity line, paired with a Pearson correlation coefficient of R = 1, mathematically proves that both methods calculated the parameters identically.

18 of the 19 evaluated parameters demonstrated this perfect mathematical agreement (R = 1), with all subject coordinates falling precisely along the red reference line. Only the incremental C-peptide AUC showed small deviations from the identity line (R = 0.9835): while the correlation remains strong, a minor scatter is visible, particularly at higher values. This slight variation is likely due to differences in how the two systems handle the numerical trapezoidal integration across the 60-to-120-minute time gap, or minor variations in baseline subtraction.

Acknowledgements

C Cobelli (University of Padova).

Validation material

The validation material for this study is publicly available on GitHub: github.com/SimonNanoMath/aomm-validation-public.