Distinct Patterns of Daily Glucose Variability by Pubertal Status in Youth With Type 1 Diabetes

Attaining target glycemic levels remains a challenge in young persons with type 1 diabetes, particularly during adolescence, when A1C rises (1). Physiologic factors related to hormonal changes during puberty have been implicated in suboptimal glycemic control (2). Specifically, a transient period of insulin resistance during puberty has been well characterized in adolescents with type 1 diabetes, with growth hormone and IGF-I emerging as the primary candidate mediators of insulin resistance (2–6). Furthermore, a faster linear growth rate has been associated with higher glucose variability, with possible differences among pubertal stages (7).

Physiologic factors are challenged further by psychosocial factors (1). In early childhood, youth with type 1 diabetes experience substantial glycemic variability (8,9). Wide glycemic excursions may be attributed to modest endogenous insulin production from near-complete β-cell destruction and unpredictable eating and activity patterns (8,10). During adolescence, diminishing treatment adherence can lead to frequent hyperglycemia and suboptimal glycemic control (11). Pubertal status is a marker of physical and psychosocial development and can provide additional insight into these uniquely challenging stages of childhood development and their effects on glycemic control and variability.

A1C remains an important clinical measure of glycemic control but does not provide information on short-term (intraday or interday) fluctuations in glycemia (12). Continuous glucose monitoring (CGM) has emerged as a tool to capture real-time glucose trends and variability in 5-min increments (13–15). Time in and out of the target range and measures of glucose variability derived from CGM data can provide more comprehensive indicators of glycemic control than A1C alone (15–17). One measure of glucose variability is the percentage coefficient of variation (CV), defined by the following equation: (SD of glucose/mean glucose) × 100 (18). Higher glucose variability (i.e., higher CV) in adults with type 1 and adults with type 2 diabetes has been linked to higher rates of hypoglycemia (19,20). A threshold CV of 36% has been proposed to separate stable (CV ≤36%) from unstable (CV >36%) glycemia, which is related to higher glucose variability (15,19).

Recent studies of continuous subcutaneous insulin infusion (CSII) regimens in youth with type 1 diabetes have identified distinct patterns of basal insulin infusion rates associated with age (21,22). As age is a marker of pubertal development, pubertal status probably plays a critical role in the distinct patterns of basal insulin needs. The variability in basal insulin needs, in turn, reflects prescriber efforts to limit glycemic variability and increase time spent in the target glucose range. In this study, we evaluated various metrics of glycemia and glucose variability assessed by CGM according to pubertal status of youth with type 1 diabetes in order to increase understanding of the expected impact of childhood growth and development on glycemic control. We also assessed the relationship between glucose variability defined by CV and CGM-derived time in hypoglycemia.

Study participants were part of a 2-year study assessing CGM utility in youth with type 1 diabetes at a tertiary pediatric diabetes center (23,24). Eligibility criteria included age 8–17 years, duration of diabetes ≥1 year, willingness to use a CGM system, no consistent CGM use in the preceding 6 months, daily insulin dose ≥0.5 units/kg, and A1C between 6.5% and 10.0% (48 and 86 mmol/mol). Participants with significant medical or psychiatric conditions were excluded. All study participants received the Dexcom SEVEN PLUS CGM system or G4 PLATINUM CGM system as part of the study, and they had full access to real-time CGM values. No personal CGM devices or sensor-augmented insulin pumps were used. All study participants/parents provided written informed assent/consent. The study protocol was approved by the Joslin Institutional Review Board, and no study procedures occurred before approval and consent.

Clinical, demographic, and diabetes management data were collected through a review of the electronic medical record and parent/youth interviews. Data on the insulin regimen and daily insulin use were extracted from insulin pump downloads and from participant and clinical reports. Height and weight were obtained by a trained clinical assistant using a calibrated electronic stadiometer and digital scale, respectively. BMI z scores were calculated from reference data from the Centers for Disease Control and Prevention growth charts (25). Pubertal status was assigned as prepubertal, pubertal, or postpubertal according to a previously described methodology that was validated against clinical Tanner staging (26). This methodology is based on a systematic, tiered approach that uses Tanner staging (breasts for girls, testicular size for boys), additional indicators of pubertal development from the electronic medical record, and evaluation of height velocity and growth chart trajectory (26). CGM data were extracted from a 4-week interval before A1C was measured (reference range 4–6% [20–42 mmol/mol]) (Cobas Integra analyzer; Roche Diagnostics, Indianapolis, IN); a minimum of 100 h of CGM data was required for inclusion in analyses.

Descriptive statistics for demographic and clinical data are reported as means ± SDs or as percentages. SAS software version 9.4 (SAS Institute, Inc., Cary, NC) was used for all analyses and included Pearson correlations, ANOVA, Student t tests, and Fisher exact tests. A P value <0.05 was considered statistically significant.

Metrics of glycemia, including mean glucose and percentage of time in and out of range 70–180 mg/dL (3.9–10.0 mmol/L), and metrics of glucose variability, including SD and CV, were calculated using CGM data over a 4-week period. Metrics of glycemia and glucose variability were compared by sex, pubertal status (prepubertal, pubertal, and postpubertal), and A1C categories (<7.5% [<58 mmol/mol], ≥7.5% [≥58 mmol/mol]). To assess patterns of glucose variability over 24 h, mean glucose values by hour of day were calculated using the 4 weeks of data, and these values were expressed as the percentage deviation from the 24-h mean glucose. Daily glucose patterns were compared by pubertal status. To assess differences between participants with stable and unstable glycemia, we compared CGM parameters between those with CV ≤36% (stable glycemia) and those with CV >36% (unstable glycemia) within the entire sample and within a subsample of insulin pump users (15,19).

A total of 107 youth with type 1 diabetes participated in the study. Clinical characteristics are summarized in Table 1. Participants (50% male, 95% white) had a mean ± SD age of 13.1 ± 2.6 years and a duration of diabetes of 6.3 ± 3.5 years. Mean ± SD daily insulin dose was 0.9 ± 0.2 units/kg, and 88% of participants received insulin pump therapy (Table 1). For the 4-week interval of CGM data, the corresponding mean ± SD A1C was 7.8 ± 0.8% (62 ± 9 mmol/mol; range 6.1–10.1% [43–87 mmol/mol]), with 34% (n = 36) of participants achieving A1C <7.5% (<58 mmol/mol), while 20% (n = 21) had an A1C ≥8.5% (≥69 mmol/mol). The CGM system was used for a mean ± SD of 115 ± 32 h/week, and this value was similar across pubertal groups (P = 0.6).

Of the 107 participants, 66 (62%) received a pubertal status assignment based on Tanner staging, whereas the remainder received pubertal status assignments based on medical record review according to previously published methods (26). In the sample, 26% (n = 28) were classified as prepubertal, 40% (n = 43) as pubertal, and 34% (n = 36) as postpubertal (Table 1).

Demographic characteristics of each pubertal group were compared (Table 1). Participants in the prepubertal group were younger at diagnosis (P = 0.01) and had a shorter duration of diabetes (P = 0.001) than those in the pubertal and postpubertal groups. The pubertal group had higher insulin requirements (1.0 ± 0.2 units/kg) compared with the prepubertal (0.7 ± 0.1 units/kg) and postpubertal (0.9 ± 0.2 units/kg) groups (P < 0.0001). Postpubertal youth received a higher percentage of the total daily dose of insulin as basal insulin (45 ± 9%) than did prepubertal (40 ± 9%) and pubertal (40 ± 8%) youth (P = 0.01). There were no significant differences in BMI z score or the proportion of males among the three pubertal groups. Mean A1C values were similar across groups.

Metrics of glycemia and glucose variability calculated from the 4-week interval of CGM data are summarized by pubertal status in Table 1. Prepubertal youth had a higher glucose SD (86 ± 12 mg/dL [4.8 ± 0.7 mmol/L]) than did the pubertal (79 ± 13 mg/dL [4.4 ± 0.7 mmol/L]) and postpubertal (77 ± 14 mg/dL [4.3 ± 0.8 mmol/L]) youth (P = 0.01). There was a similar trend for CV, with a higher CV in the prepubertal youth (43 ± 5%) than in the pubertal (41 ± 5%) and postpubertal (40 ± 5%) youth (P = 0.06). No differences were found in mean glucose, percentage of time in range (70–180 mg/dL [3.9–10 mmol/L]), percentage of time in the hypoglycemic range (<70 mg/dL [<3.9 mmol/L]), or percentage of time in the hyperglycemic range (>180 mg/dL [>10 mmol/L] and >250 mg/dL [>13.9 mmol/L]) among the prepubertal, pubertal, and postpubertal groups (all P ≥ 0.4). Metrics of glycemia and glucose variability by pubertal status were compared by sex and assessed in insulin pump users only, and both subanalyses showed patterns similar to those of the entire sample (data not shown).

Correlations between mean glucose and A1C were compared in the entire sample and by pubertal status. In the entire sample, mean glucose and A1C were highly correlated (r = 0.71; P < 0.001), and these correlations remained strong in each pubertal group (Supplementary Fig. 1). In youth achieving a target A1C <7.5% (<58 mmol/mol), the prepubertal group had a significantly higher mean glucose and spent a larger percentage of time with glucose >250 mg/dL (>13.9 mmol/L) than did the pubertal and postpubertal groups, despite similar A1C levels (Supplementary Table 1).

To characterize patterns of glucose variability among pubertal groups, deviations from the 24-h mean glucose were calculated by hour of day (Fig. 1A). In all three pubertal groups, the mean glucose declined overnight; this reduction was most pronounced in the prepubertal group. In addition, the greatest excursions from the mean glucose occurred in the prepubertal youth, with the largest nadir overnight and the largest peaks in the morning and early evening. These excursions from the mean glucose support the observations that glucose variability, as defined by SD and CV, was greatest in the prepubertal youth. To further assess glucose variability patterns over 24 h, mean CV was calculated by hour of day (Fig. 1B). Notably, CV was highest in the prepubertal youth across most hours of the day and night. In the pubertal youth, CV was the lowest in the early morning, likely reflecting the absence of a significant overnight drop in mean glucose, as shown in Fig. 1A.

The greater variability observed in the prepubertal group was especially evident in a subanalysis of 36 youth with A1C <7.5% (<58 mmol/mol) (Supplementary Table 1). Among those with A1C <7.5%, prepubertal youth had a higher glucose SD (81 ± 7 mg/dL [4.5 ± 0.4 mmol/L]) than did the pubertal (69 ± 11 mg/dL [3.8 ± 0.6 mmol/L]) and postpubertal (67 ± 11 mg/dL [3.7 ± 0.6 mmol/L]) youth (P = 0.006). The trend was similar but nonsignificant for CV, with a higher CV in the prepubertal youth (43 ± 4%) than in the pubertal (40 ± 4%) and postpubertal (40 ± 5%) youth (P = 0.2). There were no differences in mean glucose or other glucose parameters in those with A1C ≥7.5% (≥58 mmol/mol) (Supplementary Table 1).

We compared CGM parameters between youth with low glycemic variability (defined as CV ≤36%) and those with high variability (defined by a CV >36%) (Table 2). Among the sample, only 20% (n = 21) had low glycemic variability (CV ≤36%). Glucose SD was significantly higher in those with a CV >36% than in those with a CV ≤36% (P < 0.0001). Notably, the group with a CV >36% spent 4% of the day, or ∼1 h, in the hypoglycemic range (<70 mg/dL [<3.9 mmol/L]), whereas the group with a CV ≤36% spent 1% of the day, or ∼15 min, in the hypoglycemic range (P < 0.0001). Indeed, CV was strongly correlated with the percentage of time <70 mg/dL (<3.9 mmol/L) across all youth (r = 0.79; P < 0.0001) (Fig. 2). There was a modest suggestion that a higher CV may be correlated with a smaller percentage of time with glucose >180 mg/dL (>10.0 mmol/L; r = −0.25); however, this was significant only in the postpubertal group (P = 0.02). No correlation existed between CV and the percentage of time with glucose at 70–180 mg/dL (3.9–10.0 mmol/L) or >250 mg/dL (>13.9 mmol/L) (Supplementary Fig. 2).

Attaining target glycemic control is challenging for youth with type 1 diabetes throughout childhood and especially during adolescence (1,27). Contributing factors, particularly during adolescence, include the physiologic insulin resistance of puberty and the “behavioral resistance” of attending to diabetes self-care during the teen years (2,11). As A1C does not capture short-term fluctuations in glycemia, CGM data in addition to A1C can provide a more complete picture of glycemia and its variability in efforts to understand glycemic patterns and ultimately optimize control (13–15).

Using 4 weeks of CGM data, we identified distinct 24-h glucose patterns according to pubertal status. The greatest drop in mean glucose overnight occurred in prepubertal youth. The greatest excursions from the mean, with the most pronounced peaks and nadirs, also occurred in the prepubertal group. Consistent with these findings, prepubertal youth had higher glucose variability as defined by SD and a trend toward higher CV compared with pubertal and postpubertal youth, despite similar glycemic control.

Greater excursions from mean glucose in prepubertal youth could be due to multiple causes. One cause might be their insulin sensitivity, which is higher than that in pubertal and postpubertal youth. Although early hormonal changes of puberty can be first detected in prepubertal youth along with rising serum IGF-I levels, pubertal insulin resistance is not yet manifest (28,29). During mid- and late puberty, growth hormone and IGF-I levels rise significantly and seem to mediate increasing insulin resistance (2–6,29,30). Thus, the potential greater glucose variability that occurs in prepubertal youth may reflect variations in insulin sensitivity throughout the day and night. In contrast, the insulin resistance resulting from pubertal hormonal changes in the older, pubertal youth might lead to more consistently elevated glucose, with less glucose variability throughout the day and night, albeit with lower mean glucose than in the prepubertal youth.

Another cause of greater excursions from mean glucose in the prepubertal youth in our sample might relate to their younger age at diagnosis, which is associated with more aggressive β-cell loss and less residual β-cell function (10). Younger age at diagnosis was in fact correlated with higher glucose variability, as defined by CV, and confirmed in a multivariable model (data not shown). Diminished endogenous insulin production and near-complete β-cell loss may lead to greater glucose variability in response to diet, exercise, and other lifestyle factors.

Psychosocial factors related to diabetes management could also contribute to greater glycemic excursions in prepubertal youth. Parents of young children may be less aggressive with insulin dosing and/or may overtreat low blood glucose levels because of fear of hypoglycemia (31,32). In addition, a high-carbohydrate diet could contribute to wide fluctuations in glucose following meals (33,34). On the other hand, diminished treatment adherence during adolescence can lead to frequent hyperglycemia, and thus decreased glucose variability, in both pubertal and postpubertal youth (11). In addition, parental education was notably the highest among prepubertal youth, which might affect the management of type 1 diabetes and thus glycemic variability.

Prepubertal youth appear to have higher mean glucose levels than are reflected by their A1C levels, particularly those who achieve target A1C levels (<7.5% [<58 mmol/mol]). Discrepancies between mean glucose and A1C could be attributed to differences in red blood cell turnover rates and/or differences in glycation rates (15). A1C should be interpreted in association with CGM data, particularly in prepubertal children, in efforts to optimize glycemia. Future studies might provide further insight into the relationship between A1C and CGM-derived mean glucose in youth of all ages with type 1 diabetes. Furthermore, daily glucose and CV patterns could be stratified by A1C in order to assess patterns across different levels of glycemic control. Stratification by A1C in our study, however, resulted in very small numbers of participants in each pubertal group (i.e., n = 4), which limited the feasibility of such analyses.

In addition to glycemic excursions by time of day, we assessed glucose variability by CV and its association with hypoglycemia. Higher glucose variability as defined by higher CV has been associated with hypoglycemia in adults (19,20). Similarly, we found that a higher CV was associated with more time in the hypoglycemic range (<70 mg/dL [3.9 mmol/L]) across all three pubertal groups in our sample. The majority of youth (80%) had a CV >36%, which was associated with ∼1 h/day spent in the hypoglycemic range (<70 mg/dL [<3.9 mmol/L]). These findings suggest that reducing glycemic variability and CV might decrease the risk of hypoglycemia in youth with type 1 diabetes, independent of pubertal status.

Strengths of our study include the use of 4 weeks of CGM data and a large sample of youth of various ages and pubertal stages. Although 14 days of CGM data have been shown to be sufficient to reflect glycemic control for a 3-month interval (35), our analyses included 4 weeks of CGM data and thus should allow a substantial assessment of the glycemic patterns in our sample. Another strength is the use of a standardized and validated method of assigning pubertal status based on Tanner staging and linear growth evaluations, which allowed us to evaluate glycemic variability in prepubertal, pubertal, and postpubertal youth without relying solely on age.

Limitations of our study include the lack of biochemical evidence of insulin resistance and pubertal development. Although in our sample Tanner staging and linear growth evaluations served as biomarkers of pubertal development, we acknowledge that not all participants received a pubertal status assignment based on physical examination. However, our method of assessing pubertal status on the basis of medical record review and linear growth has a sensitivity of 87% compared to physical Tanner staging (26). Our study used older CGM devices, namely the Dexcom SEVEN PLUS and G4 PLATINUM, both of which perform less well than currently available CGM systems (36). However, the G4 PLATINUM has demonstrated adequate performance in pediatric patients with diabetes (37). Further, our use of 4 weeks of CGM data should help to mitigate any performance limitations of the older systems that we used.

We also acknowledge that many additional factors, including dietary practices and timing of meals and snacks, exercise patterns, sleep habits, and psychosocial stressors, can affect glucose variability, but these were not assessed in our study. Experience with and use of the CGM system might have also affected glycemic parameters, including glucose variability. Although none of the participants were using CGM in the 6 months preceding the study, a few (∼10%) had past exposure to real-time CGM use, but all had discontinued use, likely because of the poorer performance of early CGM devices. In addition, 88% of our sample population used insulin pump therapy, which might limit the generalizability of our findings to youth on injection insulin regimens. Last, our sample reflected modest racial and ethnic diversity. Future studies should include a more diverse sample of youth, as well as youth with varying insulin regimens and a wider range of A1C levels, in order to investigate additional psychosocial, treatment-related, and biochemical factors associated with glycemic patterns and glucose variability.

In summary, prepubertal youth had greater glucose variability that was independent of A1C than did pubertal and postpubertal youth. Across all youth, greater glucose variability was associated with significantly more time in the hypoglycemic range. Current clinical practice guidelines for youth with type 1 diabetes recommend glycemic control targets based on A1C alone, although a recent consensus conference recommended the use of CGM metrics (e.g., CGM-derived time in range) as part of routine clinical care (38–40). Increasing use of CGM systems in the pediatric population offers a unique opportunity to assess fluctuations in glycemic control and attainment of glycemic goals once time in range becomes an accepted target. Our findings show that A1C alone does not capture the full range of glycemia in youth with type 1 diabetes and highlight the added insight provided by CGM data in the management of type 1 diabetes (16,17). In particular, assessment of glucose variability through the use of CV can be considered in clinical practice in an effort to reduce the risk of hypoglycemia in youth with type 1 diabetes. Evaluation of additional biochemical, behavioral, and psychosocial variables in future studies could further advance our understanding of glucose variability and help to tailor individualized approaches to optimize glycemic control. Finally, longitudinal studies are needed that incorporate findings from CGM data with A1C in clinical interventions aimed at improving glycemic control in youth with type 1 diabetes.

Funding. This research was supported by funding from the National Institutes of Health (grants R01DK089349, K12DK094721, P30DK036836), the Eleanor Chesterman Beatson Fund, the Maria Griffin Drury Pediatric Fund, and the Katherine Adler Astrove Youth Education Fund.

The content is solely the responsibility of the authors and does not necessarily represent the official views of these organizations.

Duality of Interest. L.M.L. provides consulting services for Johnson & Johnson, Eli Lilly, Sanofi, Novo Nordisk, Roche Diagnostics, Dexcom, AstraZeneca, and Boehringer Ingelheim. No other conflicts of interest relevant to this article were reported.

Author Contributions. J.Z. conceived the research question, analyzed and interpreted the data, and wrote the first draft of the manuscript. L.K.V. designed and implemented the study, interpreted the data, and reviewed and edited the manuscript. L.M.L. conceptualized, designed, and implemented the study; interpreted the data; and reviewed and edited the manuscript. L.M.L. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Prior Presentation. Parts of this study were presented at the 78th Scientific Sessions of the American Diabetes Association, Orlando, FL, 22–26 June 2018.