INTRODUCTION

The number of people globally with diabetes is increasing, particularly in developing countries⁽¹⁾. While it was estimated in 2019 that there were 463 million people around the world living with diabetes, there are key differences between developed and developing countries⁽²⁾. Population-based surveys have shown that communities experiencing westernization and urbanization associated with lifestyle change are at an even higher risk for diabetes – diabetes is twice as prevalent in urban settings compared to rural settings⁽²⁾. Argentina is one such developing country, where the prevalence of diabetes is 12.7% of the population above 18 years old^(3,4). Between 6 and 10% of people with diabetes (in Argentina) have type 1 diabetes and are using insulin, primarily basal-bolus therapy⁽⁵⁾. Approximately 35% of type 2 diabetes patients in Argentina also use insulin⁽⁶⁾. Due to this population of Argentineans with diabetes, recent literature in Medicina (Buenos Aires) has provided recommendations for how continuous or flash glucose monitoring should be used to improve glucose control amongst people with diabetes⁽⁷⁾.
Effective glycemic control is essential for minimizing microvascular and macrovascular complications associated with diabetes. Hemoglobin A1c (HbA1c or A1c) is currently the gold standard for maintaining glycemic control as it has strong predictive value for diabetes complications⁽⁸⁾. However, A1c does not provide a measure of glycemic variability or hypoglycemic events⁽⁸⁾. Moreover, a range of mean glucose values and glucose profiles can be associated with a single A1c measure⁽⁹⁾. Time in range (TIR) has more recently been identified as a glycemic metric that captures both individual variability in glucose profiles and hypoglycemia and is associated with the risk of microvascular and macrovascular complications-a recent consensus in TIR was published that emphasizes the utility of TIR as a useful and appropriate clinical target^{(10, 11, 12, 13)}. The advent of continuous glucose monitoring (CGM) has provided opportunities to analyze patient data in greater detail (including TIR) and aid in both standardizing and achieving glycemic targets⁽¹⁴⁾.
Regularly monitoring glucose is essential for obtaining and maintaining glycemic targets. Previous studies have shown a strong correlation between higher self-monitored blood glucose (SMBG) frequency and greater glycemic control⁽¹⁵⁾. However, repeated collection of blood glucose measures can be inconvenient and painful, and therefore difficult to maintain long-term^(16,17,18). Flash or continuous glucose monitoring devices enable patients to more conveniently assess their glucose readings, leading to a greater frequency of glucose testing. Recent research demonstrated that flash glucose monitoring in real-world conditions allows more frequent glucose checks associated with better glycemic markers, including greater time in range^(19,20,21).
Given the greater burden of diabetes in such developing countries as Argentina, as well as the ability of the FreeStyle Libre™ flash continuous glucose monitoring system to correlate user behaviors with glycemic outcomes, the goal of this work is to examine and compare usage of flash glucose monitoring in Argentina and worldwide since its introduction (August 2017 and September 2014, respectively), including user behavior surrounding glucose scanning frequency and its association with glycemic markers.

MATERIALS AND METHODS

FreeStyle Libre (Abbott Diabetes Care, Alameda, CA, USA) flash continuous glucose monitor includes a needle-based sensor inserted below the skin that monitors glucose in interstitial fluid for up to 14 days. A smartphone with an app installed or a reader device are used to scan the sensor at any time to collect the current glucose, glucose trend, and up to 8 hours of glucose readings (recorded at 15-minute intervals). The data in this study includes only data collected by readers, not those collected by smartphones.
The reader’s 90-day memory is de-identified and uploaded to a database by the report software, which includes an agreement that this data will be collected by use of the software. All data included in this study was de-identified and anonymous. No demographic or personal data regarding users was available. As such, it was not possible to distinguish between users with T1 and T2 diabetes in this analysis. The analysis required that each sensor was operational for 5 days or longer in order to ensure reliable glucose control measures. Data from all sensors belonging to the same reader were averaged to calculate each user’s glucose metrics.
Glucose measures assessed, which align with the standardized metrics outlined in the international consensus report, included estimated HbA1c (eA1c), glucose standard deviation (SD), glucose coefficient of variation (CV), percent time in range (TIR) (defined as glucose between 70 and 180 mg/dl), percent time in level 2 hyperglycemia (>250 mg/dl), percent time in level 1 and 2 hyperglycemia (>180 mg/dl), percent time in level 1 and 2 hypoglycemia (<70 mg/dl), and percent time in level 2 hypoglycemia (<54 mg/dl)⁽¹³⁾. Mean glucose was converted into eA1c by the method accepted by international professional diabetes societies⁽²²⁾.
Scanning frequency for each reader was assessed by dividing the reader’s total number of scans by the duration of use over all its sensors. Readers were rank-ordered by scan frequency and allocated to 10 equally-sized groups. The glucose control measures were inspected in relation to the 10 scan-frequency groups of readers, and the proportion of each group achieving >70% TIR was also computed.
Statistical comparisons between the lowest and highest scan frequency groups were evaluated using a t-test, and this span of glycemic measures and changes was reported. Database was analyzed by structured query language routines and the Python programming language (www.python.org), and further summarized by KNIME (http://www.knime.org). In view of the limited sample sizes, p<0.05 was considered statistically significant.

RESULTS

From September 2014 to May 2020, the database collected glucose readings from 950.234 readers and 10.672.024 sensors, of which 5.804 readers and 56.268 sensors were from Argentina since the system launched in August 2017. Each of the 10 scan rate groups consisted of 95.023 and 580 readers for the world and Argentina, respectively.
Readers in Argentina had 14.5 scans per day on average. When binning readers into 10 equally sized groups based on daily scans, the lowest decile in Argentina had an average scan rate of 4.1 scans per day, and the highest decile had an average scan rate of 39.4 scans per day (Table I, Figure 1). For all countries, the average scan rate was 13.2 scans per day. The lowest decile of daily scans had an average scan rate of 3.6 scans per day, and the highest decile group had an average scan rate of 36.9 scans per day (Table II, Figure 1).

Table I. Glucose metrics for scan rate deciles in Argentina (N=5,804 total, each group N=580).

Figure 1. Cumulative frequency of daily scans, for both Argentina and world-wide.

Table II. Glucose metrics for scan rate deciles of all countries (N=950,234 total, each group N=95,023).

Comparing the lowest scan rate decile to the highest for Argentina, clinically and statistically significant differences were observed in eA1c (-1.3% from 8.4 % to 7.1 %; p<0.001), time in hyperglycemia above 180 mg/dl (-17.5 % from 49.2 % to 31.6 %; p<0.001), glucose SD (-22.1 mg/dl from 79.5 mg/dl to 57.4 mg/dl; p<0.001), and glucose CV (-4.3 % from 40.8 % to 36.5 %; p<0.001). Similarly, TIR showed a clinically and statistically significant difference (+17.6 % from 44.1 % to 61.7 %; p<0.001) amounting to an additional 4.2 hours per day in range. The small difference in time in hypoglycemia below 70 mg/dl was not significant (-0.04% from 6.69 % to 6.65 %; p=0.92). Figure 2 visualizes the comparison between the lowest and highest scan rate groups for time in level 2 hypoglycemia (<54 mg/dl), level 1 hypoglycemia (54-69 mg/dl), in range (70-180 mg/dl), level 1 hyperglycemia (181-250 mg/dl), and level 2 hyperglycemia (>250 mg/dl).

Figure 2. Differences in time in ranges in Argentina between the lowest and highest deciles of daily scans.
Differences between the lowest and highest decile of daily scans in Argentina for time in level 2 hypoglycemia (<54 mg/dL), level 1 hypoglycemia (54-69 mg/dL), target range (70-180 mg/dL), level 1 hyperglycemia (181-250 mg/dL), and level 2 hyperglycemia (>250 mg/dL).

For the analysis including all countries, comparing the lowest scan rate group to the highest, clinically and statistically significant differences were again detected in eA1c (-1.1 % from 8.0 % to 6.9 %; p<0.001), time in hyperglycemia above 180 mg/dl (-15.7 % from 42.8 % to 27.1 %; p<0.001), glucose SD (-15.8 mg/dl from 68.3 mg/dl to 52.5 mg/dl; p<0.001), and glucose CV (-2.6 % from 37.0 % to 34.4 %; p<0.001) (Table II). As it was the case in Argentina, percent time in range showed a clinically and statistically significant difference (+14.8 % from 52.8 % to 67.6 %; p<0.001) corresponding to an additional 3.6 hours per day in range. In contrast to data from Argentina, there was a statistically significant difference in percent time below 70 mg/dl for all countries (0.9 % from 4.4 % to 5.3 %; p<0.001).
For readers both in Argentina and worldwide, figure 3 demonstrates a strong association between higher scanning frequency and less time in hypoglycemia below 54 mg/dl (Figure 3A), greater TIR (Figure 3B), less time in hyperglycemia above 250 mg/dl (Figure 3C), lower eA1c (Figure 3D), and lower glucose CV (Figure 3E). A strong association between higher scan frequency and a higher rate of achieving TIR>70 % was also observed (Figure 4). The lowest scan group in Argentina only had 13.6 % of readers with TIR>70 %, while the highest scan group had 33.6 % of readers achieving this target. These rates were lower than the worldwide cohort, which had 25.7 % and 47.9 % achievement in the lowest and highest scan rate groups, respectively.

Figure 3. Glucose metrics for each scan rate decile, for both Argentina and world-wide.

Figure 4. Achievement of TIR > 70% for each scan rate decile, for both Argentina and world-wide.

DISCUSSION

This is the first known study that investigates the comparative association of sensor-based glucose monitoring to glycemic markers in Argentina, one of the so-called developing countries in Latin America. Given that time in range metrics are still being established, and that both healthcare professionals and users may be inexperienced with CGM devices in Argentina, one might speculate that the treatment strategy in Argentina is focused on lowering average glucose, with less emphasis on managing glucose variability. Such details of diabetes management have become measurable in the data accumulated through the usage of flash CGM. The results show that, like users worldwide, users in Argentina also demonstrate that higher scan frequency is associated with lower estimated A1c, less time in hyperglycemia, greater time in range, and less glucose variability.
Flash glucose monitoring users in Argentina have a mean daily scan rate 1.3 scans per day greater than that of users around the world. Within Argentina, higher scan frequency is associated with better glycemic outcomes, and the difference from the lowest decile of scan frequency to the highest decile is greater in Argentina than it is for all countries. Figure 2 provides a detailed view of how the time in ranges differ between these two scan rate deciles. The time in ranges of the highest scan rate decile illustrate why both eA1c and CV are low for that group-the time at extreme glucose levels (level 2 hyperglycemia and level 2 hypoglycemia) is reduced, particularly for level 2 hyperglycemia. Monitoring time in ranges with CGM complements HbA1c measurements by helping users and healthcare professionals understand how an individual’s glucose variability relates or contributes to A1c.
The mean eA1c for Argentina and all countries in this analysis is 7.7% and 7.5%, respectively. The higher average eA1c in Argentina indicates that the initial group of 5.804 Argentinean users may be earlier in the progression of diabetes management than the typical worldwide user, or similarly, that healthcare professionals in Argentina may not initiate or intensify therapy even when therapeutic goals are not reached. The difficulty of acquiring insurance coverage for CGM in Argentina also means that such technology is generally less accessible, in contrast to more widespread access in many other countries.
Figure 3 shows that users in Argentina have not only higher eA1c, but also higher time in hyperglycemia, higher time in hypoglycemia, and a greater CV compared to the rest of the world. This set of observations is characteristic of brittle diabetes, implying that the current users predominantly have type 1 diabetes^(5,18). It is interesting to note that while all countries show a trend for time in level 2 hypoglycemia that first increases with increased scan frequency before decreasing, in Argentina, time in level 2 hypoglycemia appears to generally decrease as scan frequency increases. This second observation could be partly due to the smaller sample size for Argentina, however, another explanation is that there is a substantial proportion of type 2 diabetes patients in the data for all countries who are not using insulin and are therefore able to achieve low time in hypoglycemia at a low frequency of scanning. Although the data are de-identified, this possible scenario is consistent with the speculation that current CGM users in Argentina have characteristics of type 1 diabetes patients (and are using insulin). The overall greater time in hypoglycemia is also consistent with the previously mentioned speculation that the current treatment strategy in Argentina is focused on lowering average glucose. It is generally not surprising that, compared to the rest of the world, countries like Argentina (which are earlier in the adoption of CGM technology) have yet to demonstrate the same benefit or progression in overall glucose metrics from continued usage of such devices. While the mean scan frequency of Argentinean users is higher than that of users worldwide, figure 1 shows that the cumulative frequency distribution of scans is highly similar between Argentina and all countries, and figure 4 demonstrates that the association between higher scan frequency and greater achievement of TIR > 70% is the same between Argentina and all countries. These combined observations suggest that users in Argentina are in fact engaging with flash CGM similarly to users in other countries and monitoring their glucose more than would be the case with SMBG. The results indicate the acceptance of CGM technology, and also foster the expectation that overall glucose metrics in Argentina can improve, based on previous observations of the longitudinal improvement in glucose control associated with continued usage of flash CGM by higher risk patients⁽²⁰⁾. It will be important to continue analyzing glucose metrics in Argentina as usage of flash glucose monitors expands.

This study has two key strengths: 1) A large number of readers were analyzed with consistent methodology making it possible to compare individual country numbers with worldwide data. 2) The results capture real-world data for CGM users obtained without any experimental intervention. Over 950.234 readers and 10.672.024 sensors from around the world and over 5.804 readers from Argentina are included in this analysis. However, several limitations are also acknowledged in this analysis-despite the large volume of data, little information is available regarding specific characteristics of the users, including type of diabetes, therapy type, and age. Therefore, the speculation that current users in Argentina have predominantly type 1 diabetes cannot be confirmed. Moreover, because continuous glucose monitoring is reimbursed in some countries, but not in others, the constitution and motivations of the user population may differ from one country to another. In Argentina for example, people with diabetes are only prescribed CGM when they surpass very high HbA1c and glucose variability levels, and otherwise would pay for the device ‘out-of-pocket’. Such challenges in cost and insurance may result in a selection bias in the Argentina data towards users for whom glucose control is the most problematic, and users with the greatest motivation to actively manage their diabetes. Nonetheless, this sensor-based data uniquely enables detailed measurement of glycemic control via time-in-range metrics.

CONCLUSIONS

This study suggests that flash CGM users in the developing country of Argentina have a high frequency of glucose scans, comparable to that of their counterparts worldwide, and the data from both Argentina and around the world unequivocally demonstrate that higher scan frequency is associated with greater glycemic control. In addition, the recently introduced TIR metrics reveal critical components of glycemic control not captured by traditional HbA1c measurement. Observing and understanding TIR metrics through the use of CGM is thus an important complement to HbA1c that can lead to better treatment decisions and glucose control. At present, however, insurance for CGM is still difficult to acquire for some people with type 1 diabetes, and especially people with type 2 diabetes, and further studies are therefore recommended as access to CGM in Argentina grows. More analyses like this, with different populations, will also continue to enrich the characterization of diabetes management in the real-world setting.

Conflict of interest: Kao K, Barbieri DE, Brandner L, and Dunn T are employees of Abbott Diabetes Care. Litwak L and Ré M has been members of advisory panels/received speakers fees from Abbott Diabetes Care. This study was funded by Abbott Diabetes Care.