The Association of Maximum Body Weight on the Development of Type 2 Diabetes and Microvascular Complications: MAXWEL Study

Background Obesity precedes the development of type 2 diabetes (T2D). However, the relationship between the magnitude and rate of weight gain to T2D development and complications, especially in non-White populations, has received less attention. Methods and Findings We determined the association of rate and magnitude of weight gain to age at T2D diagnosis (AgeT2D), HbA1c at T2D diagnosis (HbA1cT2D), microalbuminuria, and diabetic retinopathy after adjusting for sex, BMI at age 20 years, lifestyles, family history of T2D and/or blood pressure and lipids in 2164 Korean subjects aged ≥30 years and newly diagnosed with diabetes. Body weight at age 20 years (Wt20y) was obtained by recall or from participants’ medical, school, or military records. Participants recalled their maximum weight (Wtmax) prior to T2D diagnosis and age at maximum weight (Agemax_wt). The rate of weight gain (Ratemax_wt) was calculated from magnitude of weight gain (ΔWt = Wtmax–Wt20y) divided by ΔTime (Agemax_wt –20 years). The mean Agemax_wt and AgeT2D were 41.5±10.9 years and 50.1±10.5 years, respectively. The Wt20y and Wtmax were 59.9±10.5 kg and 72.9±11.4 kg, respectively. The Ratemax_wt was 0.56±0.50 kg/year. After adjusting for risk factors, greater ΔWt and higher Ratemax_wt were significantly associated with earlier AgeT2D, higher HbA1cT2D after additional adjusting for AgeT2D, and microalbuminuria after further adjusting for HbA1cT2D and lipid profiles. Greater ΔWt and higher Ratemax_wt were also significantly associated with diabetic retinopathy. Conclusions This finding supports public health recommendations to reduce the risk of T2D and its complications by preventing weight gain from early adulthood.


Introduction
The world prevalence of diabetes among adults (aged 20-79 years) was 6.4%, affecting 285 million adults, in 2010, and will increase to 7.7%, and 439 million adults by 2030 [1]. Primary prevention of diabetes and its complications is now an important public health priority worldwide [2].
Obesity is the major risk factor for developing type 2 diabetes mellitus (T2D) [3]. Obesity increases insulin resistance in tissues such as muscle, liver, and adipose tissue. In response to this condition, the pancreatic beta-cells increase insulin production to decrease blood glucose level. Thus, obesity has direct connection with insulin resistance; a condition characterized by increased insulin production and impaired glucose tolerance [4]. Many studies have reported associations between body mass index (BMI) and T2D [5][6][7][8]. These studies have shown that besides obesity per se, an increase in body weight of 3-20 kg is associated with an elevated risk of incidence of T2D. Prevention of weight gain is beneficial for the prevention of T2D in many different ethnicities [9][10][11].
While obesity antedates the development of T2D by some years, quantitative investigation of the relationship between magnitude and rate of weight gain and the development of T2D has been relatively limited, especially in non-White populations. The present study was designed to examine the association of development of T2D and glycemia at diagnosis with weight at age 20 years, maximum lifetime weight before T2D diagnosis, age at maximum weight, and the rate of weight gain, and to identify which of these variables were most predictive of development of T2D, glucose control, and microvascular complications such as microalbuminuria and diabetic nephropathy. We hypothesized that rapid and greater weight gain would increase the risk of T2D diagnosis and its complications.

Study Population
The MAXWEL cohort was established in 2006 to investigate the effect of maximum body weight and time interval to maximum body weight on the development of T2D. We consecutively screened all individuals (n = 5,321) aged over 30 years who visited the diabetes clinic first for initial diabetes evaluation at Seoul National University Bundang Hospital (SNUBH), Seongnam, Korea, from January 2007 to December 2009.
After excluding previously diagnosed cases with diabetes, we selected 2977 subjects who had confirmed T2D by glycosylated hemoglobin (HbA1c) $6.5%, based on the American Diabetes Association diagnosis criteria for diabetes [12], and not on antidiabetic medications for more than 1 week before. Of these, those with type 1 diabetes (measured by Glutamic Acid Decarboxylase antibody, n = 32), gestational diabetes (n = 12), or diabetes with secondary causes (n = 16). Patients with malignancy (n = 44), chronic obstructive pulmonary disease (n = 68), depression and/or eating disorder (n = 39), chronic gastrointestinal disorders (n = 39), any medication for weight control for more than 3 months (n = 37), and organ transplantation (n = 4) were excluded. Another 522 subjects were excluded because they were not able to recall their maximum weight or age at maximum weight. They were similar to other participants in anthropometric and biochemical parameters such as age, sex, and glucose control. A total of 2164 newly detected T2D subjects (1220 men and 944 women) men from 2007 to 2009 were included in the current analysis. Medical history and biochemical tests including fasting glucose, HbA1c, and lipid profiles were obtained at the first visit.
The protocol was reviewed and approved by the institutional review board (IRB) of SNUBH (No. B-0909/083-008) and the patient informed consent requirement was waived by the IRB.

Assessment of Weight-related Information
Body weight at age 20 years (Wt 20y ) was obtained in 94.5% of study subjects from the following sources: medical records, military service or college examination records, or personal recording. The remaining 5.5% self-reported their Wt 20y . Maximum weight before T2D diagnosis (Wt max ) and age at maximum weight (Age max_wt ) were also self-reported. Weight around pregnancies was disregarded. In 31.3% subjects who were randomly selected from all participants (n = 678), the recalled Wt max was validated by written document, and the agreement rate was high (r = 0.91). We calculated the rate of weight gain (Rate max_wt ), which was defined as the slope, where weight change (in kilograms) from age 20 years to maximum weight was divided by the time between age 20 years and age at maximum weight (in years). Definition of weightrelated variables and study design were shown in Figure 1.

Assessment of Lifestyle and Characteristics
Interviews were conducted by designated physicians using a standardized survey querying smoking status, alcohol consumption, and exercise habits. Smoking status was divided into three categories: current smokers, ex-smokers and never smokers. Alcohol intake was assessed by frequency and quantity of beer, spirit, sake, and wine intake during the last 12 months. Alcohol intake in grams of alcohol per week was categorized into two categories: # moderate (#199.9 g/week) and heavy intake ($200 g/week). Physical activity was classified into three categories: no, irregular (#2/wk) and regular ($3/wk) exercise. One episode of exercise was defined as exercising for at least 30 min.

Anthropometric and Biochemical Parameters
Height and body weight were measured at the time of T2D diagnosis by standard methods. BMI was calculated as body weight divided by the height squared (kg/m 2 ). Blood pressure measurements were made after subjects had remained seated for 10 min. Measurements were made twice, with a 5-min rest period; the mean value of measurements was used.
We measured HbA1c for diabetes diagnosis along with fasting glucose and insulin and other biochemical parameters in a 12-h fasting state. HbA1c was measured by Bio-Rad Variant II Turbo HPLC analyzer (Bio-Rad, Hercules, CA, USA) in SNUBH, the National Glycohemoglobin Standardization Program (NGSP) level II certified laboratory. The fasting plasma concentrations of glucose, total cholesterol, triglyceride, and high-density lipoprotein (HDL) cholesterol were measured using the Hitachi 747 chemistry analyzer (Hitachi, Tokyo, Japan). Fasting plasma insulin concentrations were measured by radioimmunoassay (Linco, St. Louis, MO, USA).

Diabetic Retinopathy
Complete ophthalmologic examinations including funduscopy on the entire retina after mydriasis were performed on all patients by two ophthalmologists. After the thorough funduscopic examination, patients showing any features of diabetic retinopathy underwent color fundus photography using mydriatic 45u fundus camera (VX-10a, Kowa Inc., Nagoya, Japan). The presence and severity of diabetic retinopathy were graded based on international clinical diabetic retinopathy severity scales proposed by the Global Diabetic Retinopathy Project Group [13]. Non-proliferative diabetic retinopathy (NPDR) was defined as the presence of at least one definite retinal hemorrhage and/or microaneurysm. Subjects were assigned to the PDR group when retinal neovascularization was visible on retinal photographs.

Statistical Analysis
All data are presented as the mean and SD, and were analyzed using SPSS for Windows version 17.0 (SPSS Inc., Chicago, IL, USA). The distributions of triglycerides and U ACR were skewed (Kolmogorov-Smirnov Z = 1.22 and Z = 1.19, both P,0.05). Those values were normalized by logarithmic transformation for all analyses. The variables were compared using student's t or x 2 tests. Correlations between variables were analyzed using Pearson's correlation.
We compared mean values of Age T2D , HbA1c T2D and microalbuminuria, and prevalence of diabetic retinopathy between the highest and lowest quartile of Rate max_wt .
To test independent association of weight variables, we performed three multivariable linear regression models for Age T2D , HbA1c T2D and U ACR , respectively, and one multivariable logistic regression model for diabetic retinopathy.
For Age T2D , DWt and Rate max_wt were included as key independent variables in the multivariable linear regression model with sex, BMI 20y , alcohol consumption, smoking status, exercise habits and family history of diabetes as covariates. In the multivariable linear regression analysis for HbA1c T2D , Age T2D was additionally added as a covariate because glycaemic control might be influenced by age of diagnosis. For log-transformed U ACR , Age T2D , systolic blood pressure (SBP), diastolic blood pressure (DBP), HbA1c T2D , and log-transformed triglycerides/ HDL-cholesterol ratio were additionally added to the multivariable linear regression model because these variables might be able to affect kidney function. To assess multicollinearity of the linear regression models, we checked the variance inflation factor of variables.
For diabetic retinopathy (combined NPDR and PDR), a multivariable logistic regression analysis was performed with the same variables used in the model for U ACR; SBP$140 mmHg or blood pressure medication indicated hypertension to obtain an odds ratio. Since multiple tests were performed in the analysis, we adjusted the number (n = 4) of phenotypes, by multiplying P values by 4. These significance thresholds are conservative given correlation among the phenotype traits themselves. Statistical significance was defined as P,0.05.

Baseline Characteristics of the Participants
The baseline characteristics of the 2164 participants are shown in Table 1. The ranges of Age T2D and BMI T2D were 30-75 years and 15.4-40.1 kg/m 2 , respectively. Almost half of participants had a family history of diabetes. About one fourth of participants (24.3%) had microalbuminuria defined by $30 of U ACR, and one eighth of participants (12.4%) had diabetic retinopathy at the time of T2D diagnosis.

Weight-related Variables
Weight at age 20 years was 59.9 and maximum lifetime weight was 72.9 kg, resulting in 13.0 kg of change in body weight from age 20 years to maximum weight (DWt) ( Table 1). Age at maximum weight (Age max_wt ) was 41.5 years and accordingly it was 21.5 years from age 20 years to Age max_wt before T2D diagnosis (DTime). From these two variables, the Rate max_wt was calculated to be 0.56 kg/year. Seventy four subjects (3.4%) of all participants reported weight loss since age 30 years. In comparison between genders, men showed greater and more rapid weight gain than women. Association among Age T2D , DWt, and DTime Figure 2 shows a three-dimensional graph illustrating association among Age T2D , DWt, and DTime without the participants who lost weight (n = 74). The Age T2D decreased as DWt increased and as DTime decreased (Pearson's correlation coefficients were -0.220 between Age T2D and DWt and 0.495 between Age T2D and DTime, both P,0.01). This illustrates subjects with greater weight gain and shorter duration to maximum weight showed a tendency to be diagnosed with T2D earlier.

Comparison between Rapid and Slow Weight Gainers
After excluding 74 participants who lost weight, we compared Age T2D , HbA1c T2D , microalbuminuria, and diabetic retinopathy (Figure 3) between the highest (4.5169.84, n = 531) and lowest (0.1460.09, n = 534) quartiles of Rate max_wt . The rapid weight gainers showed earlier T2D diagnosis (Age T2D ), higher HbA1c level at diagnosis (HbA1c T2D ), and greater log-transformed U ACR than those of lower weight gainers (42.169.2 years vs. 57.368.6 years, 8.561.7% vs. 7.661.1%, and 3.061.6 vs. 2.361.6, respectively, all P,0.01). The prevalence of diabetic retinopathy was also higher in rapid compared to slow weight gainers.

Association with Age T2D
In the multivariable linear regression for Age T2D (Table 2a), greater BMI 20y , heavy alcohol consumption, no exercise, positive family history of diabetes, greater DWt and higher Rate max_wt were significantly associated with earlier Age T2D . When the diagnosis of T2DM was based on fasting glucose concentration ($126 mg/dl), similar result was obtained (data not shown).

Association with HbA1c T2D
In the multivariable linear regression analysis additionally adjusted for Age T2D (Table 2b), the subjects with early diagnosis of T2D diagnosis, ever smoker, no exercise, greater DWt, and higher rate of weight gain showed higher HbA1c level at diagnosis.

Association with U ACR
We conducted another multivariable linear regression analysis for U ACR with weight-related variables (Table 2c). In addition to covariates used in previous model, SBP, log-transformed triglycerides/HDL-cholesterol ratio, and HbA1c T2D were added as covariates.
High BMI at age 20 years, high SBP, high HbA1c T2D , high logtriglyceride/HDL-cholesterol, greater DWt, and higher rate of weight gain were significantly associated with log-transformed U ACR (Table 2c).
Variance inflation factors of all independent factors were less than 1.21, suggesting that there was no significant collinearity among the covariates in the regression models.

Variables Associated with Diabetic Retinopathy
Using a multivariable logistic regression model, we further investigated the independent risk of weight-related variables for the concomitant diabetic retinopathy, where NPDR and PDR were combined. After adjusting for the same variables used in the model for U ACR , high BMI at age 20 years, high SBP or medication, high HbA1c T2D , greater DWt, and high rate of weight gain were found to be significantly associated with presence of diabetic retinopathy at the time of T2D diagnosis ( Table 3).

Gender Difference in the Association of Weight Variables with Diabetic Complications
In gender-specific comparison, similar results were found with slight attenuation in the association of the Rate max_wt with age at T2D diagnosis, HbA1c at T2D diagnosis, urine albumin-tocreatinine ratio at T2D diagnosis ( Table A in File S1 for men and Table B in File S1 for women), and diabetic retinopathy ( Table C in File S1 for men and Table D in File S1 for women), respectively.

Discussion
In the MAXWEL cohort, greater and rapid weight gain were significant predictors of early diagnosis of T2D, high HbA1c level at diagnosis, and microalbuminuria independent of other important clinical variables. The magnitude and the rate of weight gain were also independently associated with diabetic retinopathy. These results quantitate the increased risk associated with magnitude and rate of weight gain, which are associated with earlier diagnosis of diabetes, poor glycemic control, and microvascular complications, independent of other common risk factors.
Previous studies mainly focused on amount of weight gain during a certain period. In the US First National Health and Nutrition Examination Survey, weight gain for 10 years was associated with substantially increased risk of diabetes among overweight adults [14]. Another study from US showed that there was a progressive rise in weight before development of diabetes [15]. More specifically, in a previous study, gain of .10% of body weight was associated with a significant increase in risk of T2D compared with stable weight after adjustment for multiple risk factors including initial BMI [7]. In another study, weight gain dose-dependently increased risk of T2D even among non-obese men with a low initial BMI ,21 kg/m 2 [16].
In contrast with previous studies, we considered time and magnitude of weight gain together. In our study, the beta coefficient of rate of weight gain for age at T2D diagnosis was 20.166, corresponding to 1 year earlier T2D diagnosis with 6 kg/ year of rate in weight gain (20.996 year = 20.166 year/kg 6 6 kg). The impact of obesity or weight gain on T2D incidence may differ depending on when obesity is assessed [16][17][18]. A previous study from Pima Indians showed that weight in childhood and adolescence was one of the most significant predictors of T2D [19]. Ford et al. found that participants who gained more than 5 kg over the previous 10 years had a higher chance of diabetes compared with participants whose weights remained relatively stable, even at overweight or obese levels, in a US national cohort [5]. A study from the UK showed that .10% weight gain was associated with a significant increase in risk of T2D compared to stable weight after adjustment for initial BMI [7]. These results show that the time and magnitude of weight gain should be taken into account when the impact of body weight on T2D incidence is assessed.
Pancreatic b-cell function starts to deteriorate from early age [20]. A study with healthy, glucose tolerant Caucasians showed that b-cell function is greatest around age 20 years and declines with age at a rate of about 1% per year [21], providing the rationale for choosing age 20 years as the baseline in our study. In contrast, insulin sensitivity was not affected by aging within the time frame studied [21].
A study demonstrated that the risk of diabetes increases with early weight gain and decreases with later weight loss [22]. Another study showed that BMI in childhood was a negative and independent predictor of insulin secretion at adulthood after adjusting for age, sex, and fat percent, indicating that pancreatic bcell capacity may be set early in life [23]. Conceivably, rapid increase of weight could be more damaging to pancreatic b-cell function than slow increase, given the briefer period of time available to adapt to weight increase [24]. Taken together, these data suggest that rapid weight gain is more harmful to pancreatic function than slow weight gain, particularly in younger age.
The current study extends prior work by providing the detriments of weight gain on concomitant microvascular complications of T2D. In a study from the Atherosclerosis Risk in Communities, weight gainers had significantly less favorable glucose and lipid levels when compared with weight maintainers [25]. Another study showed that greater weight gain was associated with glycaemic progression in non-diabetic subjects [26]. In the present study, rapid weight gainers showed earlier diagnosis of T2D, higher level of HbA1c, and higher prevalence of microalbuminuria and diabetic retinopathy compared to relatively slow weight gainers. Although rapid weight gain may indicate other comorbidities compared to slow weight gain, these findings highlight the importance of the dynamics of weight change associated with development of T2D, glycemic control, and diabetic microvascular complications.
Several mechanisms for the weight gain and development of T2D and its complication can be postulated. Weight gain, particularly rapid increase in adiposity, leads to the alteration in gene expression of growth factors and cytokines such as transforming growth factor-b that are important in the development of diabetes and obesity-associated glomerular injury [27]. Hyperlipidemia, commonly accompanied by obesity, is a risk factor for the development of albuminuria by promoting glomerular injury through renal upregulation of sterol-regulatory Table 2. Variables associated with age at T2D diagnosis, HbA1c at T2D diagnosis, and urine albumin-to-creatinine ratio (U ACR ) at T2D diagnosis.  element-binding proteins, which in turn induces mesangial cell proliferation and cytokine synthesis [28]. The prevalence of diabetic retinopathy was also associated with rapid and greater weight gain in our study. Obese people were 6.5 times more likely to have PDR than were those with normal weight, and the degree of obesity was positively associated with increasing severity of diabetic retinopathy [29]. These findings suggest that diabetic retinopathy is a multifactorial microvascular complication, which is associated with obesity, hyperglycemia, and blood pressure.
The MAXWEL cohort has several novel strengths. First, weight information at age 20 years was accurately obtained from official written documents in 94.5% of participants. Second, identification of diabetes was based on laboratory results, not based on selfreport. Third, only newly detected subjects with diabetes were included, which enabled us to assess glycaemia and status of diabetic complications at the time of diagnosis.
The primary limitation of this study is its cross-sectional design with retrospective components: the identification of maximum weight and age at maximum weight were based on self-report. When prevalence estimates for obesity were compared, it was found that bias in self-reported weight was smaller in-person interviews than in telephone interviews [30]. In the setting of rigorous in-person interviews by physicians, it has been shown that relationships between self-reported and measured weight are strong [31]. In our sample, self-reported weight was highly accurate in randomly selected subjects. In addition, we did not assess weight fluctuation, which may affect pancreatic b-cell function [32,33]. However, effect of weight fluctuation has not been significant after adjustment for overall weight status or attained BMI in previous studies [32,33].
In conclusion, we found that both rapid and great weight gain were associated with not only early development of T2D and glycemic status but also microalbuminuria and diabetic retinopathy. These results support public health recommendations to reduce the risk of T2D and its microvascular complications by preventing weight gain from adolescent or early adulthood. Healthcare providers may also consider reviewing patients' weight histories when assessing their T2D risk.

Supporting Information
File S1 Supporting Tables: Table A. Variables associated with age at T2D diagnosis, HbA1c at T2D diagnosis, and urine albumin-to-creatinine ratio (U ACR ) at T2D diagnosis in men. Table B. Variables associated with age at T2D diagnosis, HbA1c at T2D diagnosis, and urine albumin-to-creatinine ratio (U ACR ) at T2D diagnosis in women.