Decreasing Trends in Mean HbA_1c Are Not Associated With Increasing Rates of Severe Hypoglycemia in Children: A Longitudinal Analysis of Two Contemporary Population-Based Pediatric Type 1 Diabetes Registries From Australia and Germany/Austria Between 1995 and 2016

The landmark Diabetes Control and Complications Trial (DCCT) demonstrated the effect of good glycemic control in minimizing the risk for diabetes-related complications (1–3). However, a major adverse finding of the DCCT was an inverse association between the risk of severe hypoglycemia (SH) and lower glycemia, with patients treated intensively experiencing a threefold higher SH rate compared with those receiving conventional treatment (4). Following publication of these findings, individuals with type 1 diabetes and their clinicians have faced the challenge of lowering glycemic targets against the increased risk of SH, which is not only potentially fatal but also associated with longer-term consequences itself (5–9).

Over the past two decades, considerable advances have been made in diabetes therapies and technologies (10). In conjunction with these technological advances and newer insulins, changes in patient education over the past decades include the use of multidisciplinary teams, education of families and groups, the use of carbohydrate counting, and specific education on hypoglycemia awareness. These changes in usual diabetes care have affected both glycemic control and the risk of SH (11–15).

Both population- and single diabetes clinic–based studies in several countries have examined the association between glycemic control and SH rates over recent decades. The majority of these have reported improvements in metabolic control without a concomitant increase in SH rates (16–21). Such longitudinal observations suggest that the inverse association previously reported from the DCCT has weakened in contemporary cohorts of children and adolescents diagnosed with type 1 diabetes and that the use of modern diabetes therapies and technologies may enable lower glycemic targets to be achieved, without increasing the incidence of SH (22). This paradigm shift in the contemporary relationship between glycemic control and risk of hypoglycemia has the potential to remove barriers such as fear of hypoglycemia for parents, but requires a change in education and clinical practice for health care professionals. However, confirming the change in risk of hypoglycemia with lower HbA_1c in more than one cohort using the same study design and longitudinal data is necessary before translation of these new data begins.

Recently, a cross-sectional study of children and adolescents diagnosed with type 1 diabetes aged <15 years attending diabetes clinics during 2011 to 2012, identified from the Diabetes Patient History Documentation (DPV), Western Australian Children’s Diabetes Database (WACDD), and T1D Exchange Clinic (T1DX) registries, found no association between contemporary rates of SH and mean HbA_1c (23). Longitudinal data on glycemic control and SH events are routinely collected at three monthly clinic visits attended by patients in Germany/Austria and Western Australia from the time of diagnosis until the age of 18 years. Therefore, this study aimed to further examine the relationship between glycemic control and SH in these two population-based pediatric cohorts of type 1 diabetes, undertaking a retrospective longitudinal analysis of these data and determining the temporal trends between 1995 and 2016.

The study population was all children and adolescents diagnosed with type 1 diabetes aged <15 years identified from the DPV database in Germany and Austria and the WACDD in Western Australia. The DPV, established in 1995, and the WACDD, established in 1987, are both prospective, longitudinal, population-based registries that have been previously described in detail and have estimated case-ascertainment rates of 80% and >99%, respectively (24,25). Data obtained via clinicians during routine, quarterly diabetes clinic visits are recorded in the DPV and WACDD, including HbA_1c and the number of self-reported episodes of SH since the last visit. As most patients in both of these populations attend diabetes clinics approximately every 3 months, this minimizes the risk of recall bias for self-reported SH events (26).

Data for eligible cases were extracted for all clinic visits between 1 January 1995 and 31 December 2016, inclusive. Records for data analysis were restricted to patients aged <18 years with a duration of diabetes ≥12 months.

SH was defined according to current International Society for Pediatric and Adolescent Diabetes guidelines as a hypoglycemic episode resulting in loss of consciousness and/or seizure (27). The sum of SH events and observation time was calculated for each patient and calendar year.

Individual mean and median HbA_1c were calculated for each calendar year of follow-up using all available HbA_1c values for those individual contributing observations for that year.

Data were also extracted for age, sex, duration of diabetes, and insulin treatment regimen. For each calendar year, patients were categorized into groups according to age at the last visit in the calendar year (<6 years, 6 to <13 years, and 13 to <18 years), duration of diabetes (2 to ≤5 years and >5 years), and treatment regimen (injections or continuous subcutaneous insulin infusion).

For each patient included in the study, the mean annual HbA_1c, total number of SH events, and total observation time were calculated for each calendar year of the study period. To calculate the SH rate per category, the total number of SH events of patients within the category was used as the numerator and their summed total observation time used as the denominator. The resulting rate was then multiplied by 100 to provide the SH rate per 100 patient-years. Similarly, the annual mean HbA_1c for each category was estimated as the mean of individual mean HbA_1c for all patients in that category and calendar year.

Longitudinal trends in mean HbA_1c and SH rates were analyzed using negative binomial regression to account for overdispersion of SH events (28). As individuals could contribute to multiple calendar years in the trend analysis, a random intercept for patient was included in the regression models. Dependent on the category being analyzed, trend analysis included adjustment for sex, age-group, duration of diabetes group, and insulin regimen, as appropriate.

A two-tailed P value <0.05 was considered statistically significant, with the Tukey-Kramer method for correction of P values applied for multiple comparisons. All analyses were performed using SAS version 9.4 (SAS Institute, Cary, NC).

The study was performed according to the principles of the Declaration of Helsinki for medical research and approved by the ethics committees of each registry center, as well as the local institutional review boards of participating centers.

The DPV registry is supported by the German Center for Diabetes Research (grant 82DZ01402). This funding source did not have any involvement in the study design, analysis, interpretation of study findings, or manuscript preparation.

A total of 59,883 and 2,595 cases from the DPV and WACDD registries, respectively, were included in the study, contributing a total observation time of 312,222 patient-years from 1 January 1995 to 31 December 2016 (Table 1). Demographic characteristics were similar for the study cohorts from both registries, with just over 50% of cases being boys (52% in DPV and 51% in WACDD) and a mean age at diagnosis of 8 years (Table 1). The mean duration of study follow-up was 4.3 and 5.4 years, with patients attending an average of 3.7 and 3.3 clinic visits per year, in the DPV and WACDD study cohorts, respectively (Table 1). The overall mean unadjusted HbA_1c for the whole study period was 8.01% (95% CI 7.99, 8.02) for the DPV cohort and 8.35% (95% CI 8.31, 8.40) for the WACDD cohort (P < 0.001). The overall unadjusted SH rate for the whole study period was 4.0 (95% CI 3.9, 4.1) per 100 patient-years for the DPV cohort compared with 10.0 (95% CI 9.1, 10.9) per 100 patient-years in the WACDD cohort (Table 1).

When analyzed over the study period, a significant decreasing trend was observed in both mean HbA_1c and SH rate for both cohorts (Fig. 1). From 1995 to 2016, the mean HbA_1c decreased from 8.3 to 7.8% in the DPV cohort and from 9.2 to 8.3% in the WACDD cohort. In conjunction with the decreasing trend in HbA_1c, a statistically significant decreasing trend was observed in the SH rates in both cohorts. After adjusting for sex, age-group, and diabetes duration group, the mean annual SH rate decreased by an average of 2% (relative risk [RR] 0.983 [95% CI 0.981, 0.986]; P < 0.001) per year over the study period in the DPV cohort and 6% (RR 0.935 [95% CI 0.934, 0.937]; P < 0.001) per year in the WACDD cohort (Fig. 1).

When examined by sex (Fig. 2) or age-group (Fig. 3), similar decreasing trends in both the annual mean HbA_1c and SH rates were observed in all subgroups. Due to the small number of SH events in patients aged <6 years in the WACDD cohort, this age-group was excluded in the analyses stratified by age-group for this cohort (Fig. 3).

To further examine the relationship between HbA_1c and SH over time, records were categorized into four time periods (1995–2001, 2002–2006, 2007–2011, and 2012–2016) and the mean adjusted SH rate for each time period and HbA_1c group was estimated using negative binomial regression as described previously. For both the DPV and WACDD cohorts, a decrease in the estimated marginal mean SH rate was observed in each HbA_1c group for each successive time period (Fig. 4).

This study reports decreasing trends in mean HbA_1c and SH rates having occurred simultaneously over the past 20 years in two independent, population-based pediatric cohorts with type 1 diabetes receiving usual care in different parts of the world. This observation is in contrast to findings reported by the DCCT in the late 1990s, in which decreasing HbA_1c levels were associated with a significant increased risk of SH (4). More recently, a longitudinal follow-up study of the DCCT/Epidemiology of Diabetes Interventions and Complications (EDIC) cohort reported a continued 13–15% increased risk of SH for every 10% decrease in HbA_1c during EDIC (12).

However, several other studies, notably in pediatric patients with type 1 diabetes, have reported changes in the relationship between HbA_1c and SH rates over the past few decades. In 2004, a Western Australian study reported improvements in HbA_1c having occurred over the previous decade without a concurrent increase in SH (16). In 2007, a U.S.-based study analyzing two small type 1 diabetes cohorts identified from the same single pediatric diabetes clinic 5 years apart reported that intensification of diabetes management during the 5 years between the two cohorts studied had resulted not only in lowered HbA_1c, but also in a reduction of SH rates (17). More recently, in Denmark, lowering of HbA_1c at the population level was reported to have occurred without a corresponding increase or change in SH rates (19,20). Consistent with this, analysis of DPV showed that compared with the period 1995–2003, in 2004–2012, low HbA_1c was a minor risk factor for SH in this cohort (21). A cross-sectional analysis of pediatric patients attending diabetes clinics between 2011 and 2012 identified from the U.S. T1D Exchange Clinic, DPV, and WACDD registries found no association between HbA_1c and SH rates in these pediatric cohorts of type 1 diabetes (23).

Concurrent decreasing trends in HbA_1c and SH rates have been reported in independent studies of the WACDD (29,30) and DPV (18,31) pediatric type 1 diabetes cohorts over the past decades. The DPV and WACDD cohorts represent pediatric type 1 diabetes cohorts receiving usual care in developed nations on separate continents, with different models of diabetes care and national government funding for diabetes-related technologies. For example, Western Australia has a centralized model of care with pediatric patients diagnosed with diabetes managed at a single center in one tertiary pediatric hospital. All patients are therefore managed according to the same clinical protocol by the same multidisciplinary team, which includes diabetes educators. Data from the DPV relate to patients treated at multiple centers, by different diabetes specialists and multidisciplinary teams, with potential variation in their models of clinical care. In addition, unlike Germany/Austria, although Australia has a national public health service, access to medical devices such as insulin pumps and continuous glucose monitoring is not necessarily available to patients free of charge, meaning there may be some disparity in access to latest technologies by patient socioeconomic status. Therefore, this international collaborative study was undertaken to analyze the longitudinal data available from these two separate, independent diabetes registries together.

This study reports that over the past several decades, there has been an overall decrease in mean HbA_1c in both of these independent pediatric type 1 diabetes cohorts, without a corresponding increase in SH rates. Of interest, the mean HbA_1c and SH rates observed in the DPV cohort were lower than those observed in the WACDD across the study period. In addition, the pattern of decline in mean HbA_1c in the WACDD cohort appears steeper in the earlier part of the study period followed by a more gradual and stable decline thereafter. Importantly, this is not explained by any differences in data collection or completeness during the study period time, as the WACDD was established in 1987 and was well established by 1995 when this study period commenced. The observational data analyzed in this study were not set up to infer any causal relationships between specific changes in diabetes management/therapies and ensuing changes in HbA_1c and/or severe hypoglycemic risk over time, including any associations that could account for the change in slope of decline observed in the WACDD.

The concurrent decreasing trends in HbA_1c and SH rates observed in this longitudinal study of two independent cohorts provide strong evidence that the relationship between HbA_1c and SH rates has changed in developed countries (19,21,23) where modern diabetes therapies and technology are being used to treat pediatric patients with type 1 diabetes by health professionals in well-developed pediatric diabetes care programs. Advances in diabetes therapies and technologies over the past decades including insulin analogs, pump therapy, continuous glucose monitoring, and patient/carer education, together with increasing clinical experience in the use of these diabetes management tools, are likely to have resulted in reduction in HbA_1c without an increase in SH rates (10,13,32,33). Importantly, reduced SH rates over time were observed in all HbA_1c groups, including the lowest HbA_1c group, in both of the cohorts analyzed in this study.

The strengths of this study are its use of data from prospective, population-based diabetes registries with high case ascertainment levels of >90% for the DPV and >99.9% for the WACDD with consistent data collection from routine clinical visits attended throughout the study period for all participants. The completeness of these population-based registries minimizes the risk of selection bias, enabling an accurate analysis to be undertaken with results that are generalizable to all children and adolescents diagnosed with type 1 diabetes in these populations. Furthermore, both the DPV and WACDD are well-established longitudinal registries with data on SH events systematically and routinely collected at diabetes clinic visits for >20 years. Most patients in both of these cohorts attended clinic visits every 3 months, which minimizes the risk of recall bias for self-reported SH events in this study. An important strength of this report is that the data provide information on outcomes over an extended 22-year time period in population-based samples from the real world, in contrast to data from clinical trials in which limitations to interpreting changes in hypoglycemia rates include potential biases from participant selection and trial effects.

It is important to note that our study findings relate to population-level HbA_1c and SH rates and cannot account for individual-level variation in significant risk factors (e.g., history of previous severe hypoglycemic event, impaired hypoglycemic awareness, and residual β-cell function), which help determine the most appropriate clinical management goals for each patient. Although the requirement for individualized glycemic targets remains unquestioned, this study provides further strong support for the revision of the historical dogma that lower glycemic targets are associated with an increased risk of SH. This will require a shift in the clinical approach of health care professionals and education aimed at reducing the fear of hypoglycemia for patients and their carers. Until this happens, the commonly held belief that increased rates of SH occur with lowering HbA_1c is likely to impede efforts to increase the number of children and adolescents with type 1 diabetes reaching optimal glycemic targets and diabetes-related outcomes (7,9,33–35).

Funding. The DPV registry is supported by the German Center for Diabetes Research (grant 82DZ01402).

Duality of Interest. T.W.J. has received honoraria for speaking and advisory board membership from Eli Lilly and Company, Novo Nordisk, Sanofi, and Medtronic. No other potential conflicts of interest relevant to this article were reported.

There was no involvement by the funding sources acknowledged in this study in any aspect of the study design, data collection, data analysis and interpretation, or writing of or decision to publish this manuscript.

Author Contributions. A.H. conducted the literature search, interpretation of data analysis, synthesis of results, and compilation of tables and figures and wrote the manuscript, incorporating feedback and comments from all coauthors. J.M.H. was responsible for data aggregation and data management, conducted the data analysis, assisted with interpretation of the results, and reviewed the tables, figures, and manuscript. H.C., S.E.H., B.K., and T.W.J. reviewed the results, tables, and figures and contributed to the manuscript. E.A.D. and R.W.H. were joint senior authors for this manuscript and contributed to study design, reviewed the literature search, results, tables, and figures, and contributed to the manuscript. E.A.D. and R.W.H. are the guarantors of this work and, as such, had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.