The Frequency and Determinants of HbA1c Variability in Type 2 Diabetic Patients

Abstract

Aim:

Glycated hemoglobin (HbA1c) is an efficient and easy test to evaluate glycemic control of patients with type 2 diabetes (T2DM). This study aims to evaluate HbA1c variability and associated factors in patients with T2DM.

Methods:

Four hundred four consecutive patients with T2DM who gave consent to participate and who were eligible were included. The inclusion criterion was presence of three or more HbA1c levels in 1 year. A change ≥0.5% in HbA1c was identified as a significant variability in HbA1c in 1 year. Primary endpoint of the study was to identify the factors associated with HbA1c variability. Patients were grouped as (1) without variability, (2) one variability, and (3) more than one variability. Variability frequency and associated factors such as body mass index, smoking, and c-peptide value were assessed.

Results:

There were 404 patients (45.3% male) with mean age 58.91 ± 10.8 years. Thirty-four patients (8.4%) had no variability, 19 patients (4.7%) had one variability, and 351 patients (86.9%) had more than one variability. Patients only on insulin treatment and patients on both oral antidiabetic agents (OAD) and insulin had higher variability than patients only on OAD (P = 0.002; P < 0.01). Patients with variability had higher HbA1c levels than patients without variability (P < 0.01). A 1% increase in HbA1c had a 4.864-fold (95% confidence interval: 2.360–10.023) increased variability risk.

Conclusions:

HbA1c variability is seen in 9 of 10 patients with T2DM and higher HbA1c values and poor glycemic control are associated with a higher risk of HbA1c variability.

Introduction

Optimal glycemic control is of utmost importance for the prevention of complications in diabetes.¹ Glycated hemoglobin (HbA1c) is an efficient and easy method to evaluate glycemic control of patients with type 2 diabetes (T2DM). Various complications of diabetes are attributed to glycemic variability.² An analysis of the Diabetes Control and Complications Trial study has proposed that glycemic variability may be the reason behind increased complication rates compared with patients with similar HbA1c levels and no complications.³ Although it has been used for almost four decades and has been proven to be a reliable measure for both diagnosis and follow-up,⁴ variability in HbA1c results is also important for diabetic complications. Glycemic variability measures daily oscillations in glucose, whereas HbA1c variability targets longer periods of glucose control. There are different methods to measure HbA1c variability. Coefficient of variation (CV) and standard deviation (SD) are mostly used in studies; however, there is no standard measure.⁵

Increased risk of diabetic retinopathy, cardiovascular, and all-cause mortality have been linked with HbA1c variability independent of glycemic control in different studies.^6
–8

Understanding the factors that affect HbA1c variability can be helpful in the management and prevention of glycemic complications. This study aims to evaluate the determinants of HbA1c variability in patients with T2DM.

Methods

This is a cross-sectional study conducted in Turkey. Patients who were referred to Istanbul Medeniyet University Goztepe Training and Research Hospital Diabetes Clinic were screened. Data were collected between May 2018 and May 2019. Four hundred and four adult consecutive patients with T2DM who gave consent to participate and who were eligible were included. The inclusion criterion was the presence of three or more HbA1c levels in 1 year. Exclusion criteria were presence of anemia, hemoglobinopathy, and pregnancy. Patients' demographics, including their height, weight, waist circumference (WC), diabetes duration, smoking status, alcohol use, c-peptide levels, and glomerular filtration rate, were recorded.

HbA1c was assessed using the high performance liquid chromatography (HPLC) method with Primus (Primus, Kansas City, MO). All blood samples of all patients were in the same center and on the same device. Three hemolysate samples representing normal, decision point, and abnormal levels of %HbA1c were 5.76, 7.07, and 10.96, respectively. Estimate of repeatability SD were 0.07, 0.05, and 0.09, respectively. Repeatability %CV were 1.26, 0.72, and 0.85, respectively. Estimate of within-device precision SD were 0.09, 0.09, and 0.16, respectively. Within-device precision %CV were 1.62, 1.28, and 1.50, respectively).⁹

A change ≥0.5% between HbA1c levels in 1 year was identified as the presence of significant variability in HbA1c. HbA1c variability was also measured by intraindividual standard deviation of HbA1c (HbA1c-SD) and coefficient of variation of HbA1c (HbA1c-CV), which is a normalized variability measure as the ratio of HbA1c-SD to intraindividual mean of HbA1c (HbA1c-mean). The primary goal of the study was to identify factors associated with HbA1c variability. Patients were grouped as (1) without significant variability, (2) significant variability between only two HbA1c measurements, (3) significant variability between any of two HbA1c measurements twice or more in 1 year. The association of HbA1c variability with the recorded clinical and laboratory findings was assessed.

The study was conducted in compliance with the Declaration of Helsinki, and approval was obtained from our hospital's local ethics committee (Decision No: 2018/0073 date: May 17, 2018). Participants gave written informed consent before participating in the study.

Statistical analysis

For statistical analysis Number Cruncher Statistical System 2007 (Kaysville, UT) program was used. Normality of the variables was tested using visual (histogram) and analytic methods (Kolmogorov–Smirnov/Shapiro–Wilk's test), Kruskal–Wallis and Mann–Whitney U test were conducted to compare parameters among groups. Pearson and Spearman correlation analyses were used to test correlations between variables. An overall 5% type-I error level was used to infer statistical significance. For parameters with a normal distribution, if three or more groups were compared, one-way analysis of variance and Bonferroni tests were used. For parameters without a normal distribution, if three or more groups were to be compared, Bonferroni and Dunn's tests were used. Also for comparison of qualitative data Pearson's chi-squared test, Fisher–Freeman–Halton test, and Fisher's exact test were used. The effects of other factors on variability were tested with multivariate logistic regression analysis.

Results

Of 707 patients screened during the study period, 404 were included. The reason 303 patients were not included was lack of data. There were 45.3% males with a mean age of 58.91 ± 10.8 years. Mean body mass index (BMI) was 31.20 ± 5.37 kg/m² and the mean WC was 101.63 ± 11.74 cm. Smoking and alcohol use is reported in Table 1. There were 230 (60.2%) patients with hypertension and 44 patients with coronary heart disease (11.5%). Mean duration of diabetes was 9.06 ± 7.12 years. There were 285 (70.5%) patients who were on only oral antidiabetic agents (OAD), 29 (7.2%) used only insulin, and 90 (22.3%) used both OAD and insulin.

Table 1.

Demographic Characteristics of the Patients

Characteristics	Mean ± SD
Age (years)	58.91 ± 10.80
Male gender (%)	45.3
Height (cm)	162.99 ± 9.15
Weight (kg)	82.65 ± 14.52
BMI (kg/m²)	31.20 ± 5.37
WC (cm)	101.63 ± 11.74
Smoking (%)	18.9
Alcohol (%)	8.7
HTN (%)	60.2
CAD (%)	11.5
Duration of diabetes (years)	9.06 ± 7.12
C-peptide (ng/dL)	2.16 ± 1.14
GFR (mL/min)	88.60 ± 19.62
Glucose (mg/dL)	189.58 ± 79.40
HbA1c (%/mmol/mol)	8.67 ± 2.02 (71)
Urine protein/urine creatinine (mg/dL)	0.40 ± 0.95
Diabetes treatment, n (%)
OAD	285 (70.5)
Insulin	29 (7.2)
OAD+insulin	90 (22.3)

BMI, body mass index; CAD, coronary artery disease; GFR, glomerular filtration rate; HbA1c, glycated hemoglobin; HTN, hypertension; OAD, oral antidiabetic agents; SD, standard deviation; WC, waist circumference.

Thirty-four patients (8.4%) had no HbA1c variability, 19 patients (4.7%) had one variability, and 351 patients (86.9%) had more than one variability. There was no statistically significant difference in demographic features by variability (Table 2).

Table 2.

Presence of Variability According to Demographics

	No variability¹	Presence of variability²	1 variability³	>1 variability⁴	P ^1,2	P ^1,3,4
Age (years)
N	33	360	18	342	0.93^*	0.34^**
Min–Max (median)	22–80 (58)	21–83 (59)	48–73 (63)	21–83 (59)
Mean ± SD	58.76 ± 12.70	58.93 ± 10.63	62.56 ± 6.90	58.73 ± 10.77
Gender, n (%)
N	34	370	19	351	0.82^†	0.94^†
Female	18 (52.9)	203 (54.9)	11 (57.9)	192 (54.7)
Male	16 (47.1)	167 (45.1)	8 (42.1)	159 (45.3)
BMI (kg/m²)
N	33	358	18	340	0.50^*	0.73^**
Min–Max (median)	19.2–43.2 (29.6)	18–52.1 (30.3)	25.1–38.5 (30.2)	18–52.1 (30.3)
Mean ± SD	30.59 ± 5.45	31.25 ± 5.37	30.76 ± 4.28	31.28 ± 5.43
WC (cm)
N	33	358	18	340	0.13^*	0.30^**
Min–Max (median)	65–135 (98)	56–140 (101)	86–120 (100)	56–140 (101)
Mean ± SD	98.70 ± 11.75	101.90 ± 11.71	100.83 ± 8.88	101.96 ± 11.85
Smoking, n (%)
N	33	359	18	341	0.39^†	0.48^§
Yes	4 (12.1)	70 (19.5)	3 (16.7)	67 (19.6)
No	29 (87.9)	282 (78.6)	14 (77.8)	268 (78.6)
Quit	0 (0)	7 (1.9)	1 (5.6)	6 (1.8)
Alcohol, n (%)
N	33	359	18	341	0.33^‡	0.63^§
Yes	1 (3.0)	33 (9.2)	1 (5.6)	32 (9.4)
No	32 (97.0)	326 (90.8)	17 (94.4)	309 (90.6)
HTN, n (%)
N	31	351	18	333	0.60^†	0.72^†
Yes	20 (64.5)	210 (59.8)	12 (66.7)	198 (59.5)
No	11 (35.5)	141 (40.2)	6 (33.3)	135 (40.5)
CHD, n (%)
N	31	350	17	333	1.00^‡	0.65^§
Yes	3 (9.7)	41 (11.7)	3 (17.6)	38 (11.4)
No	28 (90.3)	309 (88.3)	14 (82.4)	295 (88.6)

Student's t-test.

†

Pearson's chi-squared test.

‡

Fisher's exact test.

Fisher–Freeman–Halton test.

One-way analysis of variance test.

Max, maximum; Min, minimum.

There were statistically significant differences in different treatment modalities by variability. Patients on insulin treatment with or without OAD had more variability than patients only on OAD (P = 0.002; P < 0.01). Although there was no difference in glomerular filtration rate by variability, patients with variability had higher urine protein/creatinine levels P < 0.01, P < 0.05, respectively, than those without (Table 3).

Table 3.

Variability Rate According to Patient Characteristics

	No variability¹	Presence of variability²	1 variability³	>1 variability⁴	P ^1,2	P ^1,3,4	Post hoc
Duration of diabetes (years)
N	30	358	18	340	0.44^*	0.60^**	—
Min–Max (median)	1–30 (5.5)	0–63 (8)	1–28 (6.5)	0–63 (8)
Mean ± SD	8.33 ± 7.01	9.12 ± 7.13	8.61 ± 7.58	9.14 ± 7.12
Diabetes treatment, n (%)
N	34	370	19	351	0.00^†,‡‡	0.00^§,‡‡	4 > 3 = 1
OAD	33 (97.1)	252 (68.1)	17 (89.5)	235 (67.0)
Insulin	0 (0)	29 (7.8)	1 (5.3)	28 (8.0)
OAD+insulin	1 (2.9)	89 (24.1)	1 (5.3)	88 (25.1)
C-peptide (ng/dL)
N	27	330	15	315	0.94^*	0.84^**	—
Min–Max (median)	0–5.9 (2)	0–7.2 (2)	0.5–3 (2.1)	0–7.2 (2)
Mean ± SD	2.07 ± 1.04	2.17 ± 1.14	2.23 ± 1.11	2.16 ± 1.15
HbA1c (%/mmol/mol)
N	34	370	19	351	0.00^a,‡‡	0.00^f,‡‡	4 > 3 = 1
Min–Max (median)	5.4–8.4 (6.5/48)	5.7–16.2 (8.5/69)	5.8–8.8 (7/53)	5.7–16.2 (8.6/70)
Mean ± SD	6.55 ± 0.73	8.87 ± 1.99	7.05 ± 0.87	8.96 ± 1.98
GFR (mL/min)
N	28	331	14	317	0.82^*	0.94^**	—
Min–Max (median)	47.5–125.3 (91.2)	15.9–137.8 (92.5)	21.1–107.3 (95.7)	15.9–137.8 (92.4)
Mean ± SD	88.04 ± 20.82	88.64 ± 19.54	88.94 ± 21.41	88.63 ± 19.49
Glucose (mg/dL)
N	31	347	17	330	0.00^* ^,‡‡	0.00^** ^,‡‡	4 > 3 = 1
Min–Max (median)	80–248 (127)	86–589 (173)	91–187 (128)	86–589 (176)
Mean ± SD	129.39 ± 36.84	194.96 ± 79.99	131.59 ± 25.1	198.23 ± 80.49
Urine protein/urine creatinine (mg/dL)
N	21	251	13	238	0.01^* ^,††	0.03^** ^,††	4 > 1
Min–Max (median)	0.05–0.052 (0.1)	0.05–8.54 (0.14)	0.06–5.48 (0.11)	0.05–8.54 (0.14)
Mean ± SD	0.14 ± 0.13	0.42 ± 0.99	0.57 ± 1.48	0.41 ± 0.96

Mann–Whitney U test.

†

Pearson's chi-squared test.

‡

Fisher's exact test.

Fisher–Freeman–Halton test.

Kruskal–Wallis test.

††

P < 0.05.

‡‡

P < 0.01.

Student's t-test.

One-way analysis of variance test.

With logistic regression analysis, insulin use and urine protein/creatinine had no statistically significant association with variability (P > 0.05); whereas HbA1c level did (P < 0.01). A 1% increase in HbA1c increased the variability risk 4.864-fold (95% confidence interval [CI]: 2.360–10.023) (Table 4).

Table 4.

Logistic Regression Analysis of Risk Factors Affecting the Variability

	P	Odds ratio	95% CI
	P	Odds ratio	Lower	Upper
Insulin usage	0.445	2.319	0.268	20.082
HbA1c (%/mmol/mol)	0.001^**	4.864	2.360	10.023
Urine protein/urine creatinine (mg/dL)	0.433	4.239	0.115	156.378

^**P < 0.01.

The distribution of HbA1C measurements and the relationship of HbA1c median value with CV and SD were examined (Table 5). A positive statistically significant weak correlation between the HbA1c median and HbA1c-SD (r = 0.407; 95% CI: 0.302–0.502), and a very weak statistically significant positive correlation between the HbA1c median and HbA1c-CV were found (r = 0.240; 95% CI: 0.125–0.350).

Table 5.

The Distribution of Glycated Hemoglobin Measurements

	N	Min	Max	Median	Mean	SD
HbA1c-SD	271	0.05	5.73	0.78	1.01	0.86
HbA1c-Median	271	5.55	14.00	7.90	8.02	1.43
HbA1c-Mean	271	5.58	13.80	8.03	8.15	1.42
HbA1c-CV	271	0.79	58.26	9.06	12.02	9.47

When we evaluated the effect of insulin use, HbA1c median measurement, and urine protein/creatinine ratio on variability (HbA1c-CV ≥9.06) with Enter Logistic regression analysis, it was seen that the model was not statistically significant (P = 0.081; P > 0.05), but the explanatory coefficient of the model was 62.0%.

When we evaluated the effect of insulin use, HbA1c Median measurement, and urine protein/creatinine ratio on variability (HbA1c-SD ≥0.78) with Enter Logistic regression analysis, the model was found to be significant (P = 0.001; P < 0.01) and the explanatory coefficient of the model was 63.5%.

Although the effect of insulin use and urine protein/creatinine ratio on variability was not statistically significant in the model (P > 0.05); the effect of HbA1c median measurement was found to be statistically significant (P = 0.001; P < 0.01). One unit of increase in HbA1c measurement increases the risk of variability by 1.588 times (95% CI: 1.226–2.057). Accordingly, the HbA1c median measurement is an independent risk factor.

Discussion

In this study, HbA1c variability was associated with insulin use, high HbA1c levels, and proteinuria. Index HbA1c levels were higher in patients with variability. There was an almost fivefold increase in variability risk with a 1% increase in HbA1c. Treatment modalities can have a direct effect on these findings. Patients on insulin are often patients with poor glycemic control; therefore, their HbA1c variability can be attributed to their glycemic status instead of insulin. Regression analysis also demonstrated no direct relation between insulin and HbA1c variability. Our study showed high HbA1c variability rate. More than 90% of our patients had one variability and 86.9% had two or more. This rate was 78% in a study from England.⁸ The difference between this study and ours may be because our study was conducted in a tertiary level university hospital.

Sociodemographic factors associated with HbA1c variability were examined in a prospective study by Mellergård et al. Conducted in 2020 on 158 patients. HbA1c variability was found to be higher in males and those with a BMI >30 kg/m².¹⁰ In a study in which the results of ∼10,000 patients were evaluated in 2017, it was concluded that the most important causes of variability were the intensive treatment regimen and the young male gender.¹¹ It is thought that sex hormones and each of the factors that cause obesity may have an effect on this difference between genders.¹² In our study, no effect of gender and BMI on variability was found.

Another study from Taiwan showed that HbA1c variability is a reliable and consistent indicator of the development of microalbuminuria.¹³ Sugawara et al. concluded that HbA1c variability has an effect on the development of microalbuminuria independent of HbA1c levels.¹⁴ This hypothesis is also supported by our findings. However, it can also be interpreted as the expected results of poor glycemic control with high HbA1c levels that can cause variability in HbA1c.

In our study, HbA1c variability was defined as ≥0.5% change in HbA1c. Although the current literature finds this definition practical, a rate instead of a constant number may be more accurate. In the study published by Forbes et al. it is stated that when HbA1c is between 6% and 10%, it would be reasonable to take the variability with a difference of 0.5%.⁸ The average of our cases was ∼8%; therefore, we used 0.5%, acknowledging the fact that there are small number of patients with HbA1c <6% and >10%.

There are some limitations in this study. Although 707 patients screened during the study period, only 404 patients who had at least three HbA1c measurements were included. The 404 patients had also some missing data for other parameters (age 11 years, BMI 13, diabetes mellitus duration 16, glucose 26, and urinary protein 32 of 404 patients), but when evaluated separately, they did not statistically affect the results. Also the retrospective nature of the study is another limitation.

Conclusions

HbA1c variability is seen in 9 of 10 patients with T2DM and higher HbA1c values and poor glycemic control are associated with a higher risk of HbA1c variability.

Footnotes

Authors' Contributions

Concept and materials by M.V.K.; design and supervision by M.V.K. and A.O.; data by M.V.K. and M.T.; analysis by O.T.C.; literature search by M.V.K. and O.T.C.; writing by M.V.K., M.T., and O.T.C.; critical revision by A.O. Informed written consent was obtained from all volunteers.

Ethics Committee Approval

This study was approved by the ethics and research committee of a University Clinical Research Ethics Committee (Approval number: 2018, Approval date: May 17, 2018) and verbal consent was obtained from each participant.

Author Disclosure Statement

No conflicting financial interests exist.

Funding Information

No funding was received for this article.

References

UK Prospective Diabetes Study (UKPDS) Group. Effect of intensive blood-glucose control with metformin on complications in overweight patients with type 2 diabetes (UKPDS 34). Lancet, 1998; 352:854–865.

American Diabetes Association. 6. Glycemic targets Standards of medical care in diabetes. Diabetes Care, 2019; 42:61–70.

Hirsch

, Brownlee

. Should minimal blood glucose variability become the gold standard of glycemic control?. J Diabetes Compl, 2005; 19:178–181.

The Diabetes Control and Complications Trial Research Group. The relationship of glycemic exposure (HbA1c) to the risk of development and progression of retinopathy in the Diabetes Control and Complications Trial. Diabetes, 1995; 44:968–983.

Gorst

, Kwok

, Aslam

, et al. Long-term glycemic variability and risk of adverse outcomes: A systematic review and meta-analysis. Diabetes Care, 2015; 38:2354–2369.

Hermann

, Hammes

, Rami-Merhar

, et al. HbA1c variability as an independent risk factor for diabetic retinopathy in type 1 diabetes: A German/Austrian multicenter analysis on 35,891 patients. PLoS One, 2014; 9:911–937.

Wan

, Fung

, Fong

, Lam

. Association of variability in hemoglobin A1c with cardiovascular diseases and mortality in Chinese patients with type 2 diabetes mellitus—A retrospective population-based cohort study. J Diabetes Compl, 2016; 30:1240–1247.

Forbes

, Murrells

, Mulnier

, Sinclair

. Mean HbA1c, HbA1c variability, and mortality in people with diabetes aged 70 years and older: A retrospective cohort study. Lancet Diabetes Endocrinol, 2018; 6:476–486.

Hoelzel

, Weykamp

, Jeppsson

, et al. IFCC Working Group on HbA1c Standardization. IFCC reference system for measurement of hemoglobin A1c in human blood and the national standardization schemes in the United States, Japan, and Sweden: A method-comparison study. Clin Chem, 2004; 50:166–174.

10.

Mellergård

, Johnsson

, Eek

. Sociodemographic factors associated with HbA1c variability in type 2 diabetes: A prospective exploratory cohort study. BMC Endocr Disord, 2020; 20:102.

11.

Noyes

, Soto-Pedre

, Donnelly

, Pearson

. Characteristics of people with high visit-to-visit glycaemic variability in Type 2 diabetes. Diabet Med, 2018; 35:262–269.

12.

Kautzky-Willer

, Harreiter

, Pacini

. Sex and gender differences in risk, pathophysiology and complications of type 2 diabetes mellitus. Endocr Rev, 2016; 37:278–316.

13.

Hsu

, Chang

, Huang

, et al. HbA 1c variability is associated with microalbuminuria development in type 2 diabetes: A 7-year prospective cohort study. Diabetologia, 2012; 55:3163–3172.

14.

Sugawara

, Kawai

, Motohashi

, et al. HbA 1c variability and the development of microalbuminuria in type 2 diabetes: Tsukuba Kawai Diabetes Registry 2. Diabetologia, 2012; 55:2128–2131.