Impact of Age of Onset, Puberty, and Glycemic Control Followed From Diagnosis on Incidence of Retinopathy in Type 1 Diabetes: The VISS Study

HbA_1c was introduced in diabetes care in the early 1980s as a measure of average glycemia during the prior 6–10 weeks (1). Many studies have demonstrated that the level of HbA_1c is closely related to development of diabetes complications in type 1 diabetes (2–5). Besides glycemic control, diabetes duration is a strong risk factor for diabetic retinopathy, but there is evidence that the effect of duration is not uniform, being modified by age at diabetes onset (6,7) and puberty (8–10). Duration of diabetes before puberty seems to prolong the time to occurrence of both nephropathy and retinopathy (11). The protective effect of younger age at diagnosis tends to disappear with time (12). Some studies have shown a greater risk for complications in patients with diabetes onset at younger ages (13). Poor glycemic control during the first 5 years after diagnosis has also been suggested to be an independent risk factor for retinopathy (14). A negative influence of puberty on the appearance of diabetic retinopathy may be due to the combination of hormonal changes and the very poor glycemic control often seen in adolescents (15). In a recent report (the VISS Study [Vascular Diabetic Complications in Southeast Sweden]) on development of severe retinopathy (defined as laser-treated proliferative retinopathy) and nephropathy (defined as persistent macroalbuminuria) with HbA_1c followed from diagnosis, we found that after a diabetes duration of 20 years no patient having a long-term mean weighted HbA_1c (wHbA_1c) <7.6% (60 mmol/mol) developed any of these complications (16).

Our aim was to analyze the impact of HbA_1c, followed from diagnosis and during different time periods; relation to puberty; age at diagnosis; and sex as risk factors for developing diabetic simplex and proliferative retinopathy.

An unselected population comprising all 451 patients diagnosed with type 1 diabetes before the age of 35 years from 1 January 1983 to 31 December 1987 in the southeast hospital region in Sweden with just over 900,000 inhabitants was studied. The patients were identified with local registers at each clinic. That all patients were included in this unselected population was validated with help of incidence registers, the Swedish Childhood Diabetes Registry (17) and the Diabetes Incidence Study in Sweden (DISS) registry (18).

Data were retrospectively collected in the patients’ journals or by a questionnaire to their physicians. Most of the patients were followed until 2005–2008. Mean (SD) duration at follow-up was 22.1 (2.0) years. For 17 patients (4%), it was not possible to track data for the whole time period until follow-up. Of these 17 patients, 11 were deceased and 6 had moved abroad. The period of puberty was defined as 10.0–14.9 years of age for girls and 12.0–16.9 years for boys (Swedish growth charts, 2002 [19]).

The Research Ethics Committee of the Faculty of Health Sciences, Linköping University, approved the study.

Retinal screening using color fundus photography was planned every other year for each patient from the onset of diabetes and from 10 years of age. Three standard photographs were taken, after pupil dilation, of each eye: nasal to the optic nerve, the optic nerve and macula, and temporal to the macula. The prevalence of retinopathy was calculated based on reevaluation of fundus photos taken between 2005 and 2008. If no photo was available during this period, the last photo, closest to these dates, was selected. The duration of diabetes at the date of photography was mean (SD) 20.8 (2.9) years. The date of the first laser treatment was collected from clinical records. Photographs or reliable data concerning previous laser therapy for proliferative retinopathy or maculopathy were available for 431 (96%) patients. Eleven patients had died, 6 patients had moved abroad, and 3 patients had not participated in screening. Two ophthalmologists evaluated the photographs independently. The photographs were graded according to the modified Airlie House protocol. For each eye and type of retinopathic lesion, subscales grading the severity of the lesion were applied and then grouped according to the worst eye into four classes: normal, slight simplex, moderate simplex, and proliferative retinopathy or maculopathy or previous laser therapy (20). If the grading was dissimilar, the ophthalmologists reevaluated the photos and then together decided the grading.

The incidence of simplex retinopathy was examined by collecting data from the records from the clinics of ophthalmology in the region. The first recording of any diabetes-related retinal lesion, commonly microaneurysms, was defined as onset of retinopathy. It was not possible to collect these data for patients who had moved out of the region, which is why only 281 (62%) patients participated in this part of the study.

The incidence of proliferative retinopathy, defined as the date of first laser treatment, was also calculated. Here it was possible to get data from 431 of 451 (96%) of the population.

HbA_1c was measured regularly at the clinical visits, two to four times per year, and analyzed by local hospital laboratories. At the start of the study in January 1983, HbA_1c was analyzed by Isolab minicolumns (Fast Hb Test System; Isolab Inc., Akron, OH) at the four central laboratories. This was replaced during 1984–1987 by high-performance liquid chromatography methods measuring HbA_1c with high precision. The analyzing laboratories calculated intermethod calibrations and conversion factors when the methods were changed. From June 1994, hospital laboratories were participating in an interlaboratory quality program (Equalis, Uppsala, Sweden), where all laboratories analyzed two samples per month. In 1997, a nationwide standardization was introduced, and repeated comparisons were made with National Glycohemoglobin Standardization Program (NGSP) values, which showed the Swedish values to be 1.1% lower than NGSP values (21). The same was demonstrated in a study comparing HbA_1c measured in 1994 at the Linköping Hospital laboratory with results from the Diabetes Control and Complications Trial (DCCT) laboratory (22). All values are converted by formulas to the new International Federation of Clinical Chemistry and Laboratory Medicine (IFCC) reference values (23). The corresponding NGSP values are also stated, making it possible to compare the results with previous studies. The conversion formula is HbA_1c (NGSP) (%) = 0.09153 HbA_1c (IFCC) (mmol/mol) + 2.153. The normal range is 27–42 mmol/mol (IFCC) corresponding to 4.6–6.0% (NGSP). For many of the patients who moved, it was possible to obtain their HbA_1c values from their physicians, and conversions factors to the Equalis reference method was done by the analyzing laboratory. However, 90% of the HbA_1c values come from laboratories in the catchment area. As a measure of long-term glycemic control, all HbA_1c values, from diabetes diagnosis and until the date for first detection of retinopathy or year of laser therapy or time of last follow-up, were used for the calculations. Mean wHbA_1c was calculated using the area under the graph of time versus HbA_1c, divided by the total follow-up time for each patient, thereby compensating for the occasionally irregular intervals between measurements (6,24).

wHbA_1c was calculated for the whole follow-up period as well as for the periods before, during, and after puberty. wHbA_1c for the periods during the first 5 years after diagnosis and >5 years after diagnosis was also calculated.

For statistical analysis, wHbA_1c was divided into five different classes. For some analyses, it was necessary to combine categories because of the absence of readings in the lowest categories.

Differences between groups were tested with t test, ANOVA, or general linear model (GLM) repeated measures with Bonferroni post hoc test.

Life table analysis with Wilcoxon (Gehan) log-rank test was used for analysis of incidence of any retinopathy and laser-treated proliferative retinopathy.

Cox proportional hazards regression model was used to analyze the relative contribution of covariates to the risk of developing simplex retinopathy and laser-treated proliferative retinopathy. Variables included in the model were age at diabetes diagnosis; onset before, during, and after puberty; sex; and mean wHbA_1c during the whole follow-up and during different periods.

Logistic regression was used to analyze the prevalence of retinopathy at the end of the follow-up period. Sex, age at onset, and wHbA_1c during different time periods were included in the models.

Data are shown as mean (SD) unless otherwise stated. The significance level was set as P < 0.05. SPSS, version 21, was used for the analyses.

The number of patients diagnosed before, during, and after puberty and wHbA_1c for the follow-up period of mean (SD) 22.1 (2.0) years are shown in Table 1. There was no difference in wHbA_1c between patients diagnosed before and during puberty, whereas wHbA_1c was lower in those diagnosed after puberty. When separated according to sex, female patients with onset after puberty had lower wHbA_1c in comparison with those with onset before and during puberty, whereas no significant differences were found for males. Both in the whole group and in the subgroup evaluated for incidence of simplex retinopathy, wHbA_1c was higher in females, except in the group diagnosed after puberty (Table 1).

To test the influence of age at onset on development of retinopathy, we categorized age at onset into 5-year intervals (Table 1). In the subgroup evaluated for incidence of simplex, the highest wHbA_1c values were found for females with age at onset 10–14.9 years, and among all patients the highest values were found for females with age at onset 5–9.9 and 10–14.9 years.

wHbA_1c was also calculated for different time periods, i.e., ≤5 years after onset, >5 years after onset, and before, during, and after puberty (Table 2). wHbA_1c was lower during the first 5 years compared with wHbA_1c during the following years. Females had higher wHbA_1c than males after 5 years’ diabetes duration. wHbA_1c before puberty was lower than during and after puberty. Females had higher wHbA_1c than males after puberty (Table 2).

The incidence of first appearance of simplex retinopathy was studied in the subgroup of patients who had not moved out from our hospital region (Table 1). The level of wHbA_1c was categorized into five levels: ≤6.7% (50 mmol/mol), 6.8–7.6% (51–60 mmol/mol), 7.7–8.6% (61–70 mmol/mol), 8.7–9.5% (71–80 mmol/mol), and >9.5% (80 mmol/mol). As can be seen from Fig. 1A, the first appearance of retinopathy occurred earlier with increasing HbA_1c categories. This difference was most obvious after ∼15 years’ duration. With time, most patients developed simplex retinopathy and the difference between the HbA_1c categories regarding simplex retinopathy decreased. In evaluations at the end of the follow-up period, simplex retinopathy had regressed to no retinopathy in 18 patients. This group had a mean (SD) wHbA_1c value of 7.0% (0.7%) (53 [8] mmol/mol) compared with 8.6% (1.3%) (70 [14] mmol/mol) in the group who did not regress (P < 0.001).

Patients diagnosed with diabetes before puberty had longer time to development of simplex retinopathy than those diagnosed during or after puberty, as shown in Fig. 1B. Figure 1C shows that simplex retinopathy appeared significantly later only in the group with diabetes onset before 5 years of age.

We performed a univariate Cox proportional hazards regression analysis of factors associated with development of first sign of retinopathy using a model (model A) with sex; diabetes diagnosis before, during, and after puberty; and wHbA_1c. Besides wHbA_1c, onset before puberty, but not sex, was significantly associated with later development of first sign of retinopathy (Table 3).

In univariate Cox proportional hazard regression analysis, including sex, age at onset categorized into 5-year intervals, and wHbA_1c, we found that, in addition to wHbA_1c, onset before 5 years of age was associated with delayed onset of simplex retinopathy (P = 0.001) (Table 3 [model B]). If wHbA_1c was divided into wHbA_1c after ≤5 years of diabetes duration and wHbA_1c after >5 years of diabetes duration, both time periods were significantly associated with development of simplex retinopathy (Table 3 [model C]). When wHbA_1c was divided according to the periods before, during, and after puberty, age of onset and wHbA_1c after puberty were associated with simplex retinopathy (Table 3 [model D]). It should be noted that in this model only patients diagnosed before puberty are included.

In univariate Cox proportional hazards regression analysis including wHbA_1c, onset in relation to puberty, and sex (Table 3 [model A]), only wHbA_1c had a major impact (P < 0.001). If puberty was replaced by age at onset in the Cox regression analysis, a similar result was obtained and only wHbA_1c was associated with the risk of proliferative retinopathy (Table 3 [model B]). If wHbA_1c was divided into different time periods, wHbA_1c after 5 years’ diabetes duration and wHbA_1c after puberty were significantly associated with proliferative retinopathy (Table 3 [models C and D]). In model D, age of onset was significantly (P = 0.04) associated with proliferative retinopathy.

At the end of the follow-up only 54 patients (12.5%) had no signs of retinopathy and 145 (33.6%) had slight simplex, 175 (40.5%) moderate simplex, and 57 (13.2%) proliferative retinopathy. There were no cases of maculopathy, and all patients with proliferate retinopathy were laser treated. Twenty of 36 patients with wHbA_1c ≤6.7% (50 mmol/mol) were without any retinopathy but none of 49 with wHbA_1c >9.5% (80 mmol/mol) were without any retinopathy. In logistic regression analysis of risk factors for simplex and proliferative retinopathy, age at onset and onset before, during, and after puberty were not significant, while wHbA_1c for the whole follow-up period was highly significant (P < 0.001) (Supplementary Table 1). In model 3 with wHbA_1c before, during, and after puberty, which only includes patients with onset before puberty, borderline significance was found for the association of proliferative retinopathy with sex and age of onset, while wHbA_1c after puberty was highly significant.

In this study of diabetic retinopathy, HbA_1c was followed continuously for 18–24 years from diagnosis in patients with type 1 diabetes, and to our knowledge this is the only study with HbA_1c followed from diagnosis for such a long time. Long-term mean wHbA_1c was calculated and related to the appearance of simplex and proliferative retinopathy. The time to the appearance of the first signs of simplex retinopathy was found to be markedly influenced by the level of wHbA_1c, as was also reported in the primary prevention cohort of DCCT (2) and in other studies (14). However, with time, the majority, but not all, of the patients developed simplex retinopathy—in agreement with a recent report (12). In the evaluation of the level of HbA_1c and the risk of retinopathy in DCCT, it was concluded that “over the range HbA_1c achieved by DCCT intensive therapy, there does not appear to be a level of glycemia below which the risks of retinopathy progression are eliminated” (25). We can confirm this statement, as even in patients with very well-controlled diabetes, with wHbA_1c of ≤6.7% (50 mmol/mol), 44% had developed simplex retinopathy after 20 years. In this context, it should be pointed out that based on evaluation of epidemiological data, the hyperglycemic threshold for diagnosis of diabetes is set at the level where microangiopathy develops (26). Complete avoidance of retinopathy in patients with type 1 diabetes evidently requires a very tight glycemic control, which is very difficult to achieve with the treatment tools available today and is also dangerous because of the risk of severe hypoglycemia (27). Furthermore, simplex retinopathy does not impair vision, in contrast to proliferative retinopathy, which of course should be avoided. We previously reported that in the patients included in this study, proliferative retinopathy did not occur in any patients with a long-term wHbA_1c <7.6% (60 mmol/mol) (16).

The role of prepubertal diabetes duration in development of microangiopathy has been a matter of debate, with reports suggesting a range from no or minimal effect (9) to being an additional independent risk factor for proliferative diabetic retinopathy (13). When onset before, during, and after puberty, defined as 10.0–14.9 years for girls and 12.0–16.9 years for boys, was tested in Cox regression adjusted for sex and wHbA_1c, onset before puberty was associated with later onset for development of simplex retinopathy. When age of onset was introduced in the model, only onset before 5 years of age was associated with later development of simplex retinopathy, in agreement with other reports (11,14). Porta et al. (12) defined prepubertal age as 0–12 years in males and 0–11 years in females and found that the prevalence of any retinopathy by duration was lower in patients with onset before puberty, but after 20 years’ duration the difference disappeared, as in our study. Thus, there is strong evidence that age at onset of diabetes affects the time to develop simplex retinopathy, while it is unclear whether puberty by itself has an effect beside age at onset.

wHbA_1c was lower during the first 5 years after diabetes onset than later. This is probably due to persistent endogenous insulin secretion. Most patients go into partial remission after starting insulin treatment (28,29). In Cox regression analysis, wHbA_1c both during the first 5 years after diabetes onset and later were associated with the incidence of simplex retinopathy, whereas only wHbA_1c in the period after >5 years’ duration was associated with proliferative retinopathy. This suggests that wHbA_1c during the first years after diabetes onset has less impact on development of proliferative retinopathy. We can only speculate from our study that the explanation may be that the total glycemic load is of greater importance than HbA_1c during shorter time periods.

In Cox regression, we found no association between proliferative retinopathy and age at onset or onset in relation to puberty. This contrasts with the EURODIAB Prospective Complications Study, which found an increased risk for patients with onset before puberty (13). In that study, HbA_1c was not followed from diagnosis, and age at diagnosis <12 years was defined as before puberty. Beside HbA_1c, baseline retinopathy was a strong risk factor and with adjustment for this and urinary albumin excretion, age at diagnosis <12 years was no longer a significant risk factor (P < 0.08). In a sample of 1,117 patients with type 1 diabetes, consecutively recruited in the Finnish Diabetic Nephropathy (FinnDiane) Study and divided into age-at-onset groups 0–4, 5–14, and 15–40 years, the age-group 0–4 years had the longest mean duration, 24.3 years (95% CI 22.7–25.9), to development of proliferative retinopathy (30). In a Cox regression model, the highest risk of proliferative retinopathy was observed in the 5–14 years age-group. Also, in that study HbA_1c was not followed from diagnosis. As in the FinnDiane Study, in our study there was a positive association of age at onset with proliferative retinopathy in the Cox regression model D with wHbA_1c before, during, and after puberty—in which only patients with onset before puberty are included. These data are not quite conclusive but may suggest that age at onset/puberty somehow modifies the long-term risk of proliferative retinopathy. For the ability to draw firm conclusions about the role of age of onset and wHbA_1c after ≤5 years in development of proliferative retinopathy, a diabetes duration longer than 18–24 years, as in our study, is needed.

The strength of our study is that we have followed HbA_1c in an unselected population of patients with type 1 diabetes from diabetes onset and managed to follow a large proportion of the whole cohort for >20 years. The study started in the early 1980s, when HbA_1c was introduced in diabetes care, which has given us the opportunity to follow HbA_1c since diabetes diagnosis. In the calculations, we could adjust for long-term HbA_1c during the whole follow-up and during different time periods, which could perhaps explain the difference of our results from those of many other studies, where HbA_1c is only analyzed during limited periods. Still, there is a risk of type II errors where we have found no significant associations. The age of the patients at inclusion varied from 0 to 35 years, which makes the study suitable for reviewing the effect of age of onset and puberty. A weakness is that like most other studies, we defined puberty by age range (12,14) and not by clinical investigation. Most of the patients developed simplex retinopathy during the duration of the study, which is favorable for assessment of risk factors. The mean duration for development of proliferative retinopathy is >20 years, which is why the observation time in our study may be too short, or the numbers of cases of proliferative retinopathy too few, to fully assess risk factors for proliferative retinopathy (30).

In clinical practice, it is of great importance to find the balance between the risk of potentially dangerous hypoglycemic events and quality of life and the risk of severe microvascular complications to be able to recommend an evidence-based optimal level of HbA_1c both in the short-term and in the long-term. The observation that wHbA_1c before and during puberty did not influence the prevalence of proliferative retinopathy at 20 years’ diabetes duration is of clinical importance in the setting of targets for glycemic control in young children for whom severe hypoglycemia might be especially dangerous.

Simplex retinopathy is not sight threatening, even if advanced simplex retinopathy is a risk factor for proliferative retinopathy (13). However, simplex retinopathy may regress, and in our study simplex retinopathy regressed in a group of patients with mean wHbA_1c 7.0% (SD 0.7%) (53 [8] mmol/mol). Proliferative retinopathy is clinically more relevant and should be avoided. We previously showed that the threshold for proliferative retinopathy is higher than for simplex retinopathy (28). Proliferative retinopathy did not occur in this material in patients with wHbA_1c <7.6% (60 mmol/mol), which indicates what should be an important goal for glycemic control. This is in close agreement with the position statement for type 1 diabetes in children and adolescents recently issued by the American Diabetes Association recommending an HbA_1c target of <7.5% (58 mmol/mol) (31).

In summary, after 20 years of diabetes duration, there is a strong positive association between long-term mean wHbA_1c followed from diagnosis and appearance of both simplex and proliferative retinopathy. Diabetes onset at <5 years of age and lower wHbA_1c the first 5 years after diagnosis are associated with longer duration before development of simplex retinopathy but not proliferative retinopathy. Proliferative retinopathy does not appear in patients with wHbA_1c <7.6% (60 mmol/mol).

Acknowledgments. The authors are grateful for colleagues all over Sweden who helped with medical data and hospital records. The ophthalmologists Mariann Abrahamsson and Meryl Dihr, Department of Ophthalmology, Vrinnevi Hospital, performed the examinations of all the fundus photos. Christina Ekenberg (Endocrine Unit, Linköping University Hospital, Linköping, Östergötland, Sweden) typed data from the records into the computer and did a fantastic job finding data from patients who had moved from the region.

Funding. This study was supported by Barndiabetesfonden (the Swedish Child Diabetes Foundation) and Forskningsrådet i Sydöstra Sverige (FORSS) (the Research Council of Southeast Sweden).

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. M.N. and H.J.A. designed the study, performed the literature research, performed the statistical analysis, interpreted data, and wrote the first draft of the manuscript. M.F. performed the statistical analysis and interpreted data. J.L. designed the study, performed the literature research, and interpreted data. M.N., M.F., J.L., and H.J.A. reviewed and approved the final version of the manuscript. H.J.A. 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 in abstract form at the 53rd Annual Meeting of the European Association for the Study of Diabetes, Lisbon, Portugal, 11–15 September 2017, and at the 43rd Annual Meeting of the International Society for Pediatric and Adolescent Diabetes, Innsbruck, Austria, 18–21 October 2017.

Clinics participating in the VISS Study Group. Region Jönköpings län, Johan Blomgren (Department of Internal Medicine, Höglandsjukhuset, Eksjö, Sweden), Oskar Lindholm (Department of Internal Medicine, Ryhovs sjukhus, Jönköping, Sweden), Karen Wahlin (Department of Internal Medicine, Värnamo sjukhus, Värnamo, Sweden), and Karin Åkesson (Department of Paediatrics, Ryhovs sjukhus, Jönköping, Sweden); Landstinget i Kalmar län, Marianne Fagerberg (Department of Internal Medicine, Kalmar sjukhus, Kalmar, Sweden), Ulrik Mathiesen (Department of Internal Medicine, Oskarshamns sjukhus, Oskarshamn, Sweden), Marika Berg (Department of Internal Medicine, Västerviks sjukhus, Västervik, Sweden), Göran Lundström (Department of Paediatrics, Kalmar sjukhus, Kalmar, Sweden), and Jan Åke Hammersjö (Department of Paediatrics, Västerviks sjukhus, Västervik, Sweden); Region Östergötland, Ulf Rosenqvist (Department of Internal Medicine, Motala Lasarett, Motala, Sweden) and Anna-Maria Ottoson (Department of Internal Medicine, Vrinnevisjukhuset, Norrköping, Sweden).