Introducing the Continuous Glucose Data Analysis (CGDA) R Package: An Intuitive Package to Analyze Continuous Glucose Monitoring Data

Introduction

The first continuous glucose monitors (CGM) have been introduced more than 20 years ago and today multiple options exist. ¹ These options consist of real-time CGM (rt-CGM) and intermittent CGM (iCGM) or Flash monitors. The rt-CGM measure the glucose levels continuously and can alert if a patient is moving toward a hypo- or a hyperglycaemic event. The iCGM, however, only show the glucose levels when the patient actively scans the device.

Multiple studies have shown that implementing CGM in routine clinical care has positive effects, regardless of the type of CGM. ² The main benefit is that CGM show real-time glucose levels and fluctuation, thereby overcoming limitations of “old” metrics such as HbA1c and fasting glucose levels. However, despite recommendations, routine use of CGM in clinical practice remains scare. ³ A possible explanation for this is the fact that using and interpreting results from CGM remains difficult for health care practitioners. Furthermore, analyses of the data in clinical trials differ widely and are therefore difficult to generalize and use in clinical practice. ²

Recently, the American Diabetes Association (ADA) published a guideline on how to use CGM in daily practice. ³ To date, no tool exists that can easily and intuitively provide these data from CGM. Several statistical packages do exist that can assist in extracting data from the CGM (e.g., HbA1c, glycemic excursions).^4,5 However, a major drawback of these packages is that the user needs to be familiar with programming and the output is often in mg/dL or %, making implementation and interpretation for clinicians that use other metrics difficult. Moreover, analyses packages of manufacturers are not insightful in how analyses are done and which algorithms are used.

Through this letter we wish to introduce the Continuous Glucose Data Analysis (CGDA) Package in the programming language R. This is an easy-to-use, free package that incorporates the ADA recommendations, has graphical output, and provides output in all commonly used metrics (i.e., glucose mmol/mol, mg/dL, and HbA1c % as well as mg/dL).

Methods

The CGDA package has been forked from the open source package “cgmanalysis,” ⁴ and supports glucose data from all CGM. The complete R package can be found on GitHub (https://GitHub.com/EvdVossen/CGDA), making it modifiable for users that want to change certain threshold values. GitHub also contains a step-by-step guide on how to install and run the package, with an example data set. Moreover, this information can also be provided directly by the corresponding author.

Results

The output of the CGDA package is a CSV/Excel file with summary statistics (Table 1). The files can be presented as individual subjects, or as groups. The package will also provide an output using the ADA Time-In Range criteria, which are particularly useful when studying a diabetic cohort.

Table 1.

Output of Summary Measures by the CGDA Package.

subject_id	ID number of the subject
date_cgm_placement	Date of placement of the CGM device
percent_time_wear	Percentage of time that the CGM device was worn
num_days_good_data	Number of “good days” reading, based on at least 80% data availability
total_sensor_readings	Total measurements of the CGM device
average_sensor_glucose*	Average glucose value of the CGM device in mmol/L
estimated_a1c_percent	Estimated A1c in percentage
estimated_a1c*	Estimated A1c in mmol/mol
estimated_gmi_percent	Estimated gmi in percentage
estimated_gmi*	Estimated gmi in mmol/mol
q1_sensor_glucose*	First quartile of the glucose level
median_sensor_glucose*	Median of the glucose level
q3_sensor_glucose*	Third quartile of the glucose level
standard_deviation_glucose*	Standard deviation
cv_glucose*	cv
min_sensor_glucose*	Lowest glucose value measured
max_sensor_glucose*	Highest glucose value measured
excursions_over_13.9	excursions_over_13.9 mmol/L
min_spent_over_13.9	min_spent_over_13.9 mmol/L
percent_time_over_13.9	percent_time_over_13.9 mmol/L
excursions_over_13	Number of excursion over 13mmol/L
excursions_over_10_under_13.9	excursions_over_10_under_13.9 mmol/L
min_spent_over_10_under_13.9	min_spent_over_10_under_13.9 mmol/L
percent_time_over_10_under_13.9	percent_time_over_10_under_13.9 mmol/L
min_spent_over_3.9_under_10	min_spent_over_3.9_under_10 mmol/L
percent_time_over_3.9_under_10	percent_time_over_3.9_under_10 mmol/L
min_spent_over_3_under_3.9	min_spent_over_3_under_3.9 mmol/L
percent_time_over_3_under_3.9	percent_time_over_3_under_3.9 mmol/L
excursions_under_3.9	Number of excursion under 3.9 mmol/L
min_spent_under_3.9	min_spent_under_3.9 mmol/L
percent_time_under_3.9	percent_time_under_3.9 mmol/L
excursions_under_3	Number of excursion under 3mmol/L
min_spent_under_3	min_spent_under_3 mmol/L
percent_time_under_3	percent_time_under_3 mmol/L
total_auc*	total_auc
average_auc_per_day	average_auc_per_day
auc_over_2sd	auc_over of more than 2 SD
average_auc_2sd	average_auc_ of more than 2 SD
auc_over_10	auc_over_10 mmol/L
average_auc_10	average_auc_10 mmol/L
auc_over_13	auc_over_13 mmol/L
average_auc_13	average_auc_13 mmol/L
auc_over_13.9	auc_over_13.9 mmol/L
average_auc_13.9	average_auc_13.9 mmol/L
conga_1*	CONGA_1
Modd*	MODD
r_mage*	r_MAGE

Abbreviation: CGM, continuous glucose monitors.

Metrics with a * are derived from the cgmanalysis package. ⁴

Conclusion

The CGDA package offers a simple and intuitive function to analyze data from CGM, regardless of manufacturer. The CGDA provides graphical output and summary statistics making classical parametric and nonparametric testing feasible. The CGDA package can provide a basis for clinicians and researchers in standardizing the methods in which CGM data are analyzed.