Overweight or obesity, weight variability and the risk of retinopathy in type 1 diabetes

Retinopathy is the most prevalent microvascular complication of type 1 diabetes (T1D), often present during adolescence and eventually affecting the vast majority of subjects with T1D.1 Retinopathy is associated with significant risks of cardiovascular morbidity and preterm mortality.2 Important risk factors for retinopathy include hyperglycaemia, dyslipidaemia, hypertension, obesity and smoking.3 In addition, awareness of the everlasting adverse effects of hyperglycaemia on the risk of retinopathy stresses the importance of tight glycaemic control from the onset of T1D onwards.4 The pathophysiological mechanisms behind this hyperglycaemia-induced 'metabolic memory' include oxidative stress, the formation of advanced glycation end products, inflammation and epigenetic modifications, which may lead to irreversible dysregulation of genes in vascular endothelial and smooth muscle cells.5 If a similar programming effect of non-glycaemic cardiovascular risk factors for microvascular outcomes exists in T1D is unknown. Overweight and obesity are increasingly common in individuals with T1D and are associated with insulin resistance, a major risk factor for micro- and macrovascular complications.6, 7 Obesity is a disease with high recidivism and weight variability; that is, oscillating between overweight/obesity and normal weight seems to increase the risk of cardiovascular events and preterm mortality in individuals with type 2 diabetes.8 In this population-based study from Sweden, we tested the hypothesis that periods of overweight or obesity may induce a similar metabolic memory as periods of hyperglycaemia, with a lasting negative impact on the risk of retinopathy. In addition, we examined whether the transition between overweight/obesity and normal weight impacts the risk of retinopathy. This population-based cohort study included 21 575 individuals with T1D, aged 0-39, identified in the national Swedish diabetes registry between 1 January 1998 and 31 December 2017, and with at least three different registrations. The National Diabetes Register includes information on individuals with any type of diabetes in Sweden. The register includes two separate sections: a paediatric section covering ages 0-18 years (SWEDIABKIDS), and an adult section (National Diabetes Register) with data on individuals ≥18 years. The cohort in the present study is unique as it includes data from two different cohorts, that is, data on all individuals with T1D aged ≥18 years from the National Diabetes Register from 1998 and onwards, and the corresponding data from individuals age <18 years from 2000 and onwards, until 31 December 2017. We included all children, adolescents and adults with a diagnosis of T1D for ≤10 years when first recorded in the registries. T1D was diagnosed according to criteria by the American Diabetes Association and identified in the register with the ICD-10 code. We excluded individuals (n = 836) with implausible values on body mass index (BMI; i.e. >70 or <12 kg/m2), or with <1 year of follow-up in the register. The National Diabetes Register includes prospectively collected data from each clinical visit on glycated haemoglobin (HbA1c), measured weight and height, smoking habits, cholesterol, triglycerides, blood pressure and the presence of microvascular complications. Data are entered continuously from the medical records through local extraction software. For this study, we used data on each individual from the first registration (baseline) and onwards. Accordingly, we had no data on potential retinopathy before baseline. Forward imputation was applied for those variables where a previous registered value was available. If not, the observation was excluded. Data on BMI recorded at the first clinical visit after 1998 was used as the baseline value. We created four different categories of 'weight trajectories' based on the first and subsequent registered BMI for each individual: (a) normal weight (BMI 18.5-24.9 kg/m2) at baseline and at all subsequent visits (reference); (b) normal weight at baseline and overweight/obese (BMI ≥25 kg/m2) at some subsequent visit; (c) overweight/obese at baseline and normal weight at some subsequent visit; and (d) overweight/obese at baseline and all subsequent visits. Categories b [normal weight at baseline and overweight/obese (BMI ≥25 kg/m2) at some subsequent visit] and c (overweight/obese at baseline and normal weight at some subsequent visit) may elucidate risks in individuals with a history of overweight or obesity but who subsequently normalized their weight as compared with the ones who were normal weight at baseline and then developed overweight/obesity later. The risk of retinopathy was assessed at the time of the 'switch' from one category to the other and forward. All categories are defined to not overlap. Once an individual switches from one category to another, he or she remains in that category even if he or she switches back. Categories (a) and (d) capture the stable patients, while categories (c) and (d) capture the volatile patients. For individuals aged <18 years, we used iso-BMI. 'High' weight variability was defined as the ≥75th percentile of the distribution on weight variability of the entire cohort, with the 0-25th percentile as a reference. We calculated incidence rates and hazard ratios (HRs) with 95% confidence intervals (CIs) using Cox regression for any type of retinopathy in relation to weight trajectory and weight variability. Individuals were followed from baseline until the first diagnosis of the outcome, retinopathy or censoring events, emigration, death, or end of follow-up on the 31 December 2017, whichever came first. As retinopathy does not yield any symptoms, it can only be diagnosed at a fundus examination. We only had access to data on potential retinopathy after the first registration in the National Diabetes Register (i.e. baseline). The exact time of onset is unknown, so the outcome is treated as interval censored. First, we fitted a 'crude' model, only adjusted for calendar year. In the multivariable model, we considered variables that have previously been associated with both the exposure and the outcome in the literature. The first adjusted model additionally included variables with potential moderating or confounding effects on the risk of retinopathy: sex, smoking (yes/no), age at onset of T1D (years), hypercholesterolaemia (low-density lipoprotein >2.6 mmol/L), hypertriglyceridaemia (triglyceride >2.0 mmol/L), hypertension (blood pressure >140/90 mmHg at three different occasions) and HbA1c. The variables were entered as categorical (sex, smoking, hypertension, hypercholesterolaemia, hypertriglyceridaemia) or continuous (HbA1c); updated mean and age at onset (age in years) in the multivariable models. All variables except sex and age at onset were treated as time varying, with the most recent observed value in National Diabetes Register. In a second multivariable model, we additionally adjusted for the most recent BMI value. The continuous relationship between weight variability and the risk of retinopathy was assessed using restricted cubic splines. The study cohort included 9406 (44%) females and 12 169 (56%) males with T1D. The median and interquartile range for age at baseline were 21 (19-26) years. The corresponding data for the duration of T1D was 6 (3-9) years, and with a median follow-up time of 4.4 (2.4-73) years. The proportion of individuals with overweight/obesity at baseline was 17%, and at the end of the follow-up it was 33%. Overweight/obesity was more common in males (18%) than in females (16%). One-third developed overweight or obesity with time, whereas only 3% went from overweight/obesity to a normal BMI. In total, 7221 individuals (40%) were diagnosed with retinopathy during follow-up. The crude model, only adjusted for calendar year, showed a significantly increased risk of retinopathy in individuals with overweight/obesity at any time point. The risk was present even in cases of subsequent BMI within the normal range (HR 1.30, 95% CI 1.12-1.51) and comparable with the risk in individuals with persistent overweight/obesity (HR 1.31, 95% CI 1.20-1.43). After adjustment for potential confounders, risk remained increased only in individuals with persistent overweight/obesity (HR 1.14, 95% CI 1.04-1.24) (Table 1). Compared with individuals with stable weight (0-25th percentile), subjects with the highest weight variability (≥75th percentile) had an increased risk of retinopathy in both the crude model (HR 1.13, 95% CI 1.03-1.23) and after adjustment for potential confounders: sex, smoking, hypercholesterolaemia, hypertriglyceridaemia and hypertension (HR 1.14, 95% CI 1.04-1.26). The risk was no longer significantly increased after including BMI in the adjusted model (Table 1 and Figure 1). In this population-based cohort study from Sweden, we tested the hypothesis that periods of overweight or obesity have similar long-lasting adverse effects on the risk of retinopathy as has been shown for hyperglycaemia. Indeed, the crude risk of retinopathy was significantly higher in individuals with a history of overweight or obesity, even if their later BMI was within the normal range. However, the risk was no longer significantly increased after taking HbA1c, obesity-related risk factors and smoking into account. This finding may be interpreted as indicating that the risk increase related to a history of overweight or obesity is not because of overweight/obesity alone but rather by overweight/obesity-associated factors. However, even if attenuated, adjusted risk estimates of retinopathy remained significantly higher in individuals with persistent overweight or obesity. In addition, the difference in weight between overweight/obesity and normal weight was associated with an increased risk of retinopathy. These findings expand on previous knowledge. Weight regain is associated with a larger increment of fat mass than lean mass,9 and weight fluctuation results in a poorer body composition with increased visceral fat mass10 and higher insulin resistance.11 Insulin resistance, in turn, is an independent risk factor for retinopathy.12 It is possible that some form of metabolic memory is established in individuals with a history of overweight/obesity, as well as in those with high weight variability. The common feature could be insulin resistance. Strengths of the current study include the population-based cohort with prospectively collected data on a large group of individuals with T1D. There is no gold standard definition of how to assess weight variability, and there are no reference data. Accordingly, different approaches are possible. In this study, which included a large number of individuals, we defined high and low degrees of weight variability based on the distribution of weight variability estimates for the whole study population. However, there are limitations to consider when evaluating our results. We were not able to differentiate intentional from unintentional weight loss, and we did not have data on important parameters such as body composition and insulin sensitivity. Although the findings of the current study do not allow any conclusion on whether a history of overweight or obesity gives a metabolic imprint, we suggest a prudent approach and the implementation of structured weight management programmes to help individuals with T1D avoid periods of overweight/obesity to limit the risks of retinopathy. All authors affirm that the manuscript is honest, accurate and transparent; that no important aspects of the study have been omitted; and that any discrepancies from the study as planned (and, if relevant, registered) have been explained. MAF and MP are the guarantors 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. This project has received funding from Stockholm City Council (research position for MP) the Swedish Society of Medicine (grant for MP), Karolinska Institute's Research Foundation Grants (for MP 2020-01844). The sponsors were not involved in study design, conduct, reporting or dissemination of our research. The authors have no conflicts of interest to declare. The peer review history for this article is available at https://www.webofscience.com/api/gateway/wos/peer-review/10.1111/dom.15545. Due to Swedish legislation, we cannot provide the data used in this study. However, it is available upon request after Ethics approval.