Fluctuations in glycosylated hemoglobin (HbA1C) as a predictor for the development of diabetic nephropathy in type 1 diabetic patients

1. Introduction 

Diabetes mellitus is reported as the most common single cause of end-stage renal disease (ESRD) in the US and Europe. Patients (20–30%) with type 1 diabetes develop evidence of diabetic nephropathy after a mean of 15years [1], [2]. Thus far, the earliest clinical evidence of diabetic nephropathy is the appearance of low but abnormal levels of albumin in the urine, referred to incipient nephropathy (albuminuria ⩾ 30mg/24h or ⩾20μg/min). These levels of microalbuminuria, when sustained, may lead to overt nephropathy and consequently to ESRD [3].

The incidence of diabetic nephropathy is higher in type 1 diabetic patients with associated risk factors such as genetic factors, presence of hypertension, and poor metabolic control. Essentially, the metabolic control is a modifiable risk factor. The Diabetes Control and Complication Trial (DCCT) has shown that intensive diabetes therapy defined as mean HbA1C of approximately 7% can significantly reduce the risk of the development of microalbuminuria by 34%, as compared to those on conventional treatment where HbA1C was 8.5%. It also showed that intervention to improve metabolic control does reduce the risk of complications independently of previous control [4]. Other studies [5], [6] showed that a between-individual difference in HbA1C of 1–2% increased the risk of developing microvascular complications by at least 25%. Until now, the within individual fluctuations in HbA1C and their consequences on the development of nephropathy were not studied. In fact, DCCT was an interventional study, where patients on intensive treatment were followed up closely (insulin adjusted daily and HbA1C checked monthly) and where HbA1C was sustained close to 7%, unlike the conventional group, whose follow up had been less frequently (every 3months). In this setting, the within individual “fluctuations” in HbA1C would be unexpected in patients with good metabolic control and very probable in those with poor control. Thus, the difference between the two groups in the occurrence of diabetic nephropathy may be explained by the differences of fluctuations within individuals. Although the effect of fluctuations in HbA1C on the development of nephropathy could be suspected indirectly from the findings of DCCT, a study directly addressing this issue is needed.

The purpose of our study was to examine whether fluctuations in HbA1C (as early as the diagnosis up till 5years from diagnosis) are a predictor of the development of diabetic nephropathy, independent of mean HbA1C and of other risk factors such as age at onset of diabetes, body mass index (BMI), the presence of hypertension and the presence of family history of diabetes.

2. Patients and methods 

Type 1 diabetic patients were recruited from a type 1 diabetes center in the Lebanon: the Chronic Care Center for children and young adults with type 1 diabetes mellitus. Type 1 diabetic patients are followed up regularly at least every 3months. At each visit, HbA1C is measured; body weight and the average dose of total insulin required per day are measured. Height is documented once a year. Blood pressure is systematically measured at diagnosis, and recorded at least once a year. Eye examination, for the presence of diabetic retinopathy, is done by an ophthalmologist at the first visit to the center then followed up a yearly basis. Microalbuminuria is tested in each patient 5years after the diagnosis of type 1 diabetes, as recommended by ADA [7], and then yearly.

Patients with type 1 diabetes, as defined by ADA [8], admitted to the chronic care center were studied. The institutional review board (IRB) in the chronic care center approved the protocol. Only patients admitted to the chronic care center within 18months of diagnosis of type 1 diabetes mellitus were included, since structural renal abnormalities due to diabetes usually occurs afterward [9]. Patients were excluded from the study if the duration of diabetes was less than 5years, since diabetic nephropathy is known to occur after at least 5years of the disease [7]. Patients were also excluded from the study if they suffered from wolfram syndrome, or had thalassemia or other hemoglobinopathy. Two hundred and four patients met the inclusion criteria, and 87 patients were excluded. The final sample size was 117 patients.

The following information was obtained: age, gender, date of birth, date of onset of diabetes, date of admission to the center, family history for diabetes, results of microalbuminuria after 5years of diagnosis, the dates and the results of HbA1C at each visit, BMI at baseline and blood pressure. Patients (microalbuminurea vs non-microalbuminurea) were selected based on a cut off point of 24h urine microalbumin of >30mg/24h on more than two occasions.

3. Statistical analysis 

The outcome of insidious nephropathy was treated as a dichotomous variable. Bivariate analyses were performed between covariates and the outcome, using χ2 or t-test according to whether the variable was continuous or categorical. Correlation of repeated measures (for the differences of two consecutive HbA1C) was explored and showed no evidence of correlation between the two measures. The pattern of missing data was investigated, as recommended by Twisk [10], and was found to be ignorable (neither dependent on the outcome nor on other predictors). Further analyses were performed on a dataset where missing data on HbA1C were inputted using linear interpolation.

The association between HbA1C and the outcome was explored by performing bivariate analysis first between mean HbA1C and nephropathy, and then between “fluctuations” in HbA1C and the outcome.

Furthermore, we studied the association between fluctuations and nephropathy, adjusting for all covariates using logistic regression analysis. Models were compared using Bayesian information criterion (BIC) as a measure of fit, for more details refer to Appendix A, Raftery [11], [12] and Long [13].

To study whether the effect of fluctuations in HbA1C on nephropathy differs with the level of metabolic control, we divided the sample into two groups: those with acceptable metabolic control and those with poor metabolic control. Afterwards, we studied the association between fluctuations in HbA1C and the development of nephropathy separately in each group, using bivariate analysis. Results were considered significant at the 5% critical level (p<0.05). All the calculations were performed using STATA (version 7.0) statistical soft-ware package.

4. Ethics 

The protocol was approved by the institutional review board (IRB) at the Chronic Care Center, Lebanon. All the procedures followed were in accordance with the ethical standards of the responsible committee on human experimentation (institutional or regional) and with the Helsinki Declaration of 1975, as revised in 1983.

5. Results 

The sample analyzed consisted of 117 type 1 diabetic patients, 64 females and 53 males, aged 9–33years at the time of enrollment. Eighteen among 117 (15.38%) developed nephropathy after 5years of onset of diabetes.

As shown in Table 1, there was no association between gender and the development of diabetic nephropathy (p=0.94). There were no significant group differences in the “age at onset of diabetes” or “time period from the onset of diabetes till admission to the chronic care center” (p=0.47 and p=0.80, respectively). BMI at first visit for those who developed nephropathy was not significantly different from the BMI at first visit for those who did not have evidence of nephropathy (p=0.40). There was no significant difference in the outcome between those with and without family history of diabetes. Hypertension was omitted, because of no variability, as patients were not hypertensive.

Table 1. Main characteristics of the study population in relation to the development of diabetic nephropathy.

	Variable	Nephropathy N=18	No nephropathy N=99	p-value
	Gender [n (%)]a
	Male	8 (44)	45 (45)	0.94
	Female	10 (56)	54 (54)
	Age at onset (years)	10.94±4.5	10.12±3.9	0.47
	Time period from onset of diabetes to admission to CCC (years)	3.96±4.2	3.72±4.2	0.80
	BMI at baseline (kg/m2)	19.84±5.2	19.04±3.4	0.40
	Family history of diabetes [n (%)]b
	Positive	8 (47)	58 (59)
	Negative	9 (53)	40 (41)	0.35
	Mean HbA1C (%)	9.4±1.6	8.5±1.1	0.003
	Fluctuations in HbA1C [n (%)]
	Present	15 (83)	54 (54)
	Absent	3 (17)	45 (45)	0.04

Data are means±SD unless otherwise specified.

a [n (%)] indicates the number in each category and (percentage). b Totals do not add up to 117 due to missing data.

5.2. Association between fluctuations in HbA1C and diabetic nephropathy 

The association between fluctuations in HbA1C and diabetic nephropathy is shown in Table 1. Among those who developed nephropathy, 15 of 18 (83%) had fluctuations in HbA1C; compared to those who do not develop nephropathy 54 of 99 (54%) had fluctuations in HbA1C (p=0.04).

In order to identify the predictors for diabetic nephropathy, we performed a multivariate analysis, by entering all risk covariates into a multiple logistic regression analysis (Table 2). Results from the full model (referred to as Model 1) revealed that mean HbA1C was the only significant predictive factor; all other variables were not significant. Since our hypothesis is to test whether the presence of fluctuations in HbA1C predicts the development of nephropathy adjusting for the mean HbA1C, we further studied three other models one including the two covariates the mean and the “fluctuations” in HbA1C (referred to Model 2), another model including only mean HbA1C as a covariate (referred to as Model 3) and the last model including the fluctuations in HbA1C (Model 4). The Model 2 leads to a smaller BIC than Model 3 (BIC dropped from −101.4 to −104.7), indicating positive evidence for a better fit. We also noticed that the odds ratio of the mean HbA1C decreases from 1.75 to 1.50 when the covariate “fluctuations” is added to the model and becomes closer to 1. Considering Model 4, the odds ratio of the fluctuations in HbA1C is 4.17; however when adjusting for the mean HbA1C (Model 2), the odds ratio dropped to 2.34 and the fluctuations in HbA1c was no more a significant predictor factor.

Table 2. Multivariate analysis for the prediction of diabetic nephropathy (Model 1 includes all covariates, Model 2 includes 2 covariates: the average mean of HbA1C and fluctuations in HbA1C and Model 3 includes only the average mean of HbA1C).

	Variable	Odds ratio (95%CI)
		Model 1	Model 2	Model 3	Model 4
	Average mean of HbA1C	1.66 (1.03; 2.68)	1.55 (1.01; 2.38)⁎	1.75 (1.18; 2.59)⁎
	Fluctuations in HbA1C	1.89 (0.42; 8.41)	2.34 (0.56; 9.77)		4.17 (1.13; 15.31)⁎
	Gender	0.85 (0.27; 2.63)
	Family history	1.32 (0.42; 4.13)
	Age at onset	1.06 (0.88; 1.26)
	Time between onset of diabetes till admission to CCC	0.93 (0.80; 1.08)
	Baseline BMI	0.93 (0.75; 1.14)
	BIC	−123.40	−104.71	−101.40	−104.22

CI denotes the confidence interval. BIC refers to Bayesian information criterion.

⁎ p-value <0.05.

Next, we investigated whether the level of metabolic control affects the association between “fluctuations” in HbA1c and the odds of nephropathy. Forty patients had acceptable metabolic control and 77 had poor metabolic control. Among those with poor control, the incidence of nephropathy was 26.3% for those with fluctuations. This is higher than the 5% incidence for those who did not have “fluctuations” (p=0.056) (Table 3). However, the effect of “fluctuations” did not appear among those with acceptable metabolic control (data not shown).

Table 3. The effect of fluctuations on the incidence of nephropathy among the 77 patients who had a poor metabolic control (HbA1C⩾8%); data are presented as n (%) the number in each category of fluctuation and (percentage).

		Fluctuation present	Fluctuation absent
	Nephropathy present	15 (26%)	1 (5%)
	Nephropathy absent	42 (74%)	19 (95%)

Fisher test: p=0.056.

6. Discussion 

The incidence of insidious nephropathy after 5years of onset of type 1 diabetes, based on our data, was 15.1% similar to the previous findings by others [14], [15], [16]. Lower incidences were reported in the US [17] and Europe [18]. A discrepancy between the different reports may be related to ethnic variability; however it resulted mostly from the use of a different methodology. One main difference is the duration of diabetes, well known as an independent risk factor for the development of microvascular complications [8], [5]. However, in our study, the duration of diabetes was set to 5years after diagnosis. Our finding that the age at onset of diabetes is not associated with the development of diabetic nephropathy is commensurate with recent data [19]; however, as in their study, we did not account for pubertal staging [20].

The results of our study, like many other reports, did not show an association between gender and the development of diabetic nephropathy. This contrasts with a previous finding by Holl et al., showing an impact of female gender on the development of insidious nephropathy [18]. Any association between gender and nephropathy should take into consideration the pubertal stage since the hormonal effects could be at the base of this difference.

Data on the association between BMI, an index of metabolic state, and the development of diabetic nephropathy, is scarce [18]. In our study, BMI was measured at the first visit to the center, when most of the patients had poor metabolic control (reflected by the elevated mean HbA1C at that visit) that might have negatively affected the weight. Although the baseline BMI was found to be associated with the development of microvascular complications [5], the impact of BMI was apparent only at higher values. Following BMI longitudinally and accounting for pubertal changes would help in establishing the associations between BMI and diabetic nephropathy.

Metabolic control was the only established and consistent predictor for the development of diabetic nephropathy. In reviewing the literature, different measures have been used in order to study the association between metabolic control and diabetic nephropathy. The mean HbA1C is repeatedly used [4], [20]; the median has also been used as a summary measure [14]. Based on the results of our study, the mean HbA1C remains the only significant predictor for the development of diabetic nephropathy in type 1 diabetic patients, even after adjusting for “fluctuations”. However, the magnitude of this association decreases when accounting for “fluctuations”. While the odds of developing nephropathy increased by at least 18%, with an increase in mean HbA1C of 1%, accounting for the “fluctuations” in HbA1C showed that the odds ratio of developing nephropathy, at least, did not change when the mean HbA1C increased by 1%.

The use of “fluctuations” in HbA1C as a longitudinal measure for the change in the metabolic state is original. It may better reflect the changes in ambient glycemia within one individual. This latter was found to be the culprit in the development of diabetic nephropathy through activation of the proteinase C [21], upregulating the heparanase expression [22], enhancing sensitivity to TGF beta 1 [23] and increasing VEGF (vascular endothelial growth factor) expression [24]. Our data showed that “fluctuations” in HbA1C predicted the incidence of nephropathy, based on the positive evidence that the model including fluctuations fits the data better. In addition, an increase of greater than 2% between two consecutive measurements seems to increase the risk of developing nephropathy by 21% among those with poor metabolic control. This may have many implications: first, these findings may help to achieve a better understanding of the pathophysiology of diabetic nephropathy, since they suggest that, although this latter is accelerated by the chronic hyperglycemia (manifested as mean HbA1C), it is much worse during acute increases in glycemia which is reflected by fluctuations in HbA1C. Second, our data highlight the issue that a single jump in HbA1C have a durable effect, this agrees with the hypothesis of “long time- glycemic memory” and supported by findings from DCCT on microvascular complications. Third, as diabetic nephropathy has an insidious onset, one large increment in HbA1C during the first 5years, would be an indicator of a development of nephropathy well before the appearance of microalbuminuria.

Nevertheless, our data were unable to establish the association between fluctuations in HbA1C and the development of nephropathy in diabetics with acceptable control. The sample size was small to permit the comparison between the different groups; this was well seen by the wide confidence intervals. Interestingly, taking the whole model, the mean HbA1C explains 10% the prediction for the development of diabetic nephropathy. Other factors, such as genetic predisposition, have been known to be associated with the development of nephropathy. Family history of hypertension [25], kidney disease and other cardiovascular risk factors [26], were used as a measure for genetic predisposition. However, our study lacks this information. Puberty data were not obtained; this is a limitation for our results. since puberty onset and its related physiologic changes had been known as risk factor for the development of diabetic nephropathy [27].

Our data should be interpreted with caution, since the measure used to define “fluctuations”, which is at the cornerstone of the results, had been based on our assumptions. This measure is not yet validated as a measure for “fluctuations”, and needs to be replicated in order to add more confidence to the results.

Finally, we conclude that type 1 diabetic patients who have a similar mean HbA1C, in the long run, may behave differently in terms of developing nephropathy, depending on the fluctuations in HbA1C and more precisely, depending on the frequency of the acute “jumps” in the HbA1C. However, this effect seems to be more pronounced among those who have higher values of HbA1C considered as poorly controlled.

Declaration of competing interests 

None to declare.