SEARCH

SEARCH BY CITATION

Keywords:

  • diabetes;
  • epidemiology;
  • erythrocytes;
  • glucose;
  • HbA1c

Abstract

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conflict of interest statement
  8. Acknowledgements
  9. References

Objective

Hyperglycaemia has multiple effects on the red blood cell (RBC), including glycation of haemoglobin, reduced deformability and reduced lifespan. Red cell distribution width (RDW) is a measure of the heterogeneity of erythrocyte volumes. The aim of this study was to explore the relationships between RDW and glucose, haemoglobin A1c (HbA1c) and incidence of diabetes mellitus (DM).

Design, setting and subjects

RDW and mean corpuscular volume were measured in 26 709 non-diabetic participants (aged 45–73 years) from the population-based Malmö Diet and Cancer cohort. HbA1c and fasting venous blood glucose levels were measured in 4845 subjects.

Main outcome measure

Incidence of DM (= 2944) over 14 years of follow-up was studied by linkage with national and local DM registers.

Results

Individuals with low RDW had significantly higher risk of developing DM [adjusted hazard ratio (HR) 1.48, 95% confidence interval (CI) 1.29–1.70, for 1st vs. 4th quartile], especially in subjects with impaired fasting glucose (= 416) (HR 2.15, 95% CI 1.12–4.14). Low RDW was also associated with significantly higher waist circumference and glucose, insulin and triglyceride concentrations. By contrast, RDW was significantly and positively associated with HbA1c, corresponding an increase in HbA1c of 0.10% per 1 SD increase in RDW.

Conclusion

Low RDW is associated with increased incidence of DM independently of other risk factors. We propose that low RDW could be a surrogate marker of reduced RBC survival, with lower HbA1c due to shorter duration of glucose exposure. RDW is a biomarker that could improve risk assessment for individuals at risk of developing DM.


Introduction

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conflict of interest statement
  8. Acknowledgements
  9. References

The red cell distribution width (RDW) is a measure of the heterogeneity of the volume of red blood cells (RBCs) [1]. Measurement of RDW is used in clinical practice to differentiate between causes of anaemia, and RDW is provided in most routine haematological examinations. Recent studies have demonstrated associations between high RDW and various adverse health outcomes, such as increased mortality [2], increased incidence of atrial fibrillation and heart failure [3, 4], and adverse prognosis in patients with heart failure or coronary heart disease [5, 6]. Substantial correlations between RDW and low heart rate variability, a measure of autonomic dysfunction, were demonstrated in a study of patients with heart failure [7]. In studies of patients with diabetes mellitus (DM) a higher mortality after primary coronary intervention procedures has been reported in individuals with high RDW [8]. In a cross-sectional study of adults with DM from the National Health and Nutrition Examination Study (NHANES), high RDW values were associated with increased odds of cardiovascular disease and nephropathy, leading to the suggestion that RDW may be a useful clinical marker of vascular complications in DM [9]. The mechanism(s) by which RDW predicts mortality and other adverse outcomes is unclear [10]. A study of healthy non-diabetic participants in the NHANES demonstrated significant positive correlations between haemoglobin A1c (HbA1c) and RDW, and the authors of the study suggested the possibility that chronic hyperglycaemia mediates the association between high RDW and cardiovascular disease [11].

It has been shown that hyperglycaemia has multiple effects on the RBC. In addition, it has been reported that RBC counts are increased in pre-diabetic states and decreased in established DM, compared to normal glucose homeostasis [12]. The effects of hyperglycaemia also include glycation of haemoglobin, reduced deformability of RBCs [13, 14] and reduced RBC lifespan [15-17]. To our knowledge, the relationship between RDW and incidence of DM has not been studied previously. The aim of this study was to determine whether RDW is associated with measures of glucose haemostasis and whether RDW predicts incidence of DM in the population-based Malmö Diet and Cancer study (MDC).

Methods

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conflict of interest statement
  8. Acknowledgements
  9. References

Study population

The MDC is a prospective cohort study from the city of Malmö in southern Sweden [3, 18]. Briefly, citizens from Malmö were invited to the screening examination by postal invitation and through newspaper advertisements. A total of 28 449 subjects (11 246 men, born 1923–1945; 17 203 women, born 1923–1950) attended a baseline examination between 18 March 1991 and 25 September 1996. Participants underwent peripheral venous blood sampling and anthropometric measurements, and filled out a self-administered questionnaire. Subjects with a history of DM at the baseline examination were excluded (= 1190). In addition, 550 subjects were excluded due to missing information on waist circumference, smoking habits, mean corpuscular volume (MCV), haemoglobin and RDW. Thus, the final study population in the analysis consisted of 26 709 subjects (10 395 men and 16 314 women), aged 45–73 years.

Between 7 October 1991 and 25 February 1994, a randomly selected subgroup was invited to take part in a cardiovascular substudy which included measurements of fasting blood glucose, insulin, triglycerides and HbA1c. Of the 6103 invited subjects, 5533 agreed to participate in a second visit at the screening centre and donated blood after fasting overnight [18]. Complete information about RDW, HbA1c, waist circumference and smoking was available for 4845 non-diabetic subjects (aged 46–68 years; 61% women).

All participants provided written informed consent. The study was approved by the ethics committee at Lund University (LU 51/90).

Measurements and definitions

Body weight and height were measured at the screening centre by two trained nurses. Standing height was measured with a fixed stadiometer calibrated (in cm). Weight was measured to the nearest 0.1 kg using a balance beam scale with subjects wearing light clothing and no shoes. Body mass index (BMI) was calculated as weight divided by height squared (in kg m−2). Waist circumference (in cm) was measured midway between the lowest rib margin and the iliac crest. Information on smoking habits and country of birth was obtained from a self-administered questionnaire. Subjects were categorized into current smokers (i.e. those who smoked regularly or occasionally) or non-smokers (i.e. former smokers and never smokers).

History of cardiovascular disease (i.e. myocardial infarction or stroke) and heart failure was retrieved by linkage with the hospital discharge diagnoses from the Swedish inpatient register [3, 19].

The diet assessment methods in the MDC study have been described elsewhere [20]. Dietary intakes of iron, vitamin B12 and folate were log normalized and adjusted for total energy intake before being entered as covariates in the multivariate models, as described previously [21]. High alcohol intake was defined as ≥40 g alcohol per day for men and ≥30 g per day for women.

The methods of measuring triglycerides, insulin and the homeostatic model assessment (HOMA) index have previously been described [18]. Insulin resistance was calculated according to the HOMA index using the formula: plasma insulin*glucose/22.5. HbA1c was determined by ion exchange chromatography, using the Swedish Mono-S standardization system; reference values were 3.9–5.3% in non-diabetic individuals.

Diabetes mellitus at baseline was defined as self-reported DM (according to the questionnaire), use of anti-diabetic medication or any recording in registers (see below) of DM prior to the baseline examination. In the subgroup, fasting venous whole blood glucose ≥6.1 mmol L−1 was also used as a criterion of pre-existing DM [22].

Red cell distribution width, haemoglobin, MCV, and RBC and leukocyte counts were analysed consecutively in fresh blood. Erythrocyte diameter was measured using a fully automated assay (SYSMEX K1000 haematology analyser; TOA Medical Electronics, Kobe, Japan). RDW was calculated as the width of the erythrocyte distribution curve at a relative height of 20% above the baseline; reference values were 35.1–43.9 fL in men and 36.4–46.3 fL in women.

Incidence of DM

All subjects were followed from the baseline examination until first diagnosis of DM, death, emigration from Sweden or 31 December 2009, whichever came first. New-onset cases of DM were identified in the Malmö HbA1c Register (MHR) (56% of all cases), the Swedish National Diabetes Register (NDR) (14%), the Swedish Hospital Discharge Register (40%), the Swedish Outpatient Register (38%), the nationwide Swedish Drug Prescription register (65%) and the regional Diabetes 2000 register of the Scania region (22%) [19, 23]. In addition, 44% of the cases were identified at re-examinations of the cohort [24]. At least two independent sources confirmed the diagnosis for 72% of cases, and 53% were identified in three independent data sources. NDR and the Diabetes 2000 register required a physician diagnosis according to established diagnostic criteria (fasting plasma glucose concentration of ≥7.0 mmol L−1, which corresponds to a fasting whole blood glucose of ≥6.1 mmol L−1 [22], measured on two different occasions). All HbA1c samples collected from individuals in institutional and non-institutional care in the greater Malmö area from 1988 onwards were analysed and recorded by the MHR at the Department of Clinical Chemistry, Malmö University Hospital. Individuals who had at least two HbA1c values ≥6.0% recorded in the MHR with the Swedish Mono-S standardization system (corresponding to 7.0% with the US National Glycohemoglobin Standardization Program) after the baseline examination were defined as having incident DM. The relationship between RDW and incidence of DM was tested for each of the data sources used for endpoint retrieval, with essentially the same results (data not shown). A total of 2944 subjects had new-onset DM during the mean (±SD) follow-up period of 14.3 (±3.9) years.

Statistic analysis

The RDW values were categorized into sex-specific quartiles, with similar proportions of men and women in each quartile. RDW was also tested as a continuous variable (per 1 SD). One-way anova and logistic regression were used to compare the distribution of risk factors across quartiles of RDW and between subjects with and without DM during follow-up. Due to skewed distributions, insulin, HOMA index and triglycerides were log transformed before analysis and presented as geometric means. Unless otherwise stated, results are presented as mean ± SD.

Multiple linear regression was used to assess the relationship between RDW and HbA1c (dependent variable). Cox proportional hazards regression was used to study incidence of DM. Age, sex, waist circumference, BMI, smoking, leukocytes, MCV and high alcohol intake, i.e. factors that are known to be associated with RDW, were used as covariates in the Cox model. In the subgroup with available glucose measurements, we also adjusted for glucose, HbA1c, triglycerides and insulin. The P-value for trend was calculated by fitting the quartiles of RDW as an ordinal variable. The fit of the proportional hazards model was confirmed by plotting the incidence of DM over time for the different risk factors. A subgroup analysis was performed for subjects with impaired fasting glucose at the baseline examination (i.e. fasting blood glucose 5.6–6.1 mmol L−1). Potential interactions between exposure variables were tested by interaction terms in the Cox models. Because high RDW is associated with mortality, a competing risk analysis was performed [25], in which deaths were considered as competing events for risk of developing DM. spss (v20) (www.spss.com) and stata (v12) software (www.stata.com) were used in all statistical calculations.

Results

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conflict of interest statement
  8. Acknowledgements
  9. References

Baseline characteristics

A total of 26 709 participants (10 395 men and 16 314 women) were included in the study. Mean age was 58.1 ± 7.6 years (59.1 ± 7.0 years in men and 57.4 ± 7.9 years in women). Mean waist circumference was 93.4 ± 9.9 cm in men and 77.5 ± 10.2 cm in women.

Fasting glucose and HbA1c measurements were available in a subgroup of 4845 subjects. In this subgroup, mean glucose was 5.0 ± 0.43 mmol L−1 in men and 4.8 ± 0.44 in women. Mean HbA1c was 4.75 ± 0.42% and 4.80 ± 0.42% in men and women, respectively. A total of 723 subjects (2.7%) had a history of myocardial infarction or stroke, and 57 (0.2%) subjects had previously been hospitalized due to heart failure. Blood pressure-lowering medication was used by 4521 subjects (16.9%), and 743 (2.8%) used lipid-lowering drugs. The distribution of risk factors across quartiles of RDW is presented in Table 1.

Table 1. Distribution of risk factors in relation to sex-specific quartiles (Q1–Q4) of red cell distribution width (RDW)
 Q1Q2Q3Q4P for trend
  1. Values are mean ± SD unless otherwise stated. Values of triglycerides, insulin and HOMA are presented as geometric means.

  2. WC, waist circumference; RBC, red blood cell; MCV, mean corpuscular volume; BMI, body mass index; HOMA, homeostatic model assessment.

RDW range (fL) (men/women)<38.2/<38.638.3–40.1/38.7–40.640.2–42.5/40.7–42.8>42.5/>42.9 
n (% women)6689 (61)6576 (62)6838 (61)6606 (61) 
Age (years)56.7 ± 6.957.8 ± 7.558.6 ± 7.859.2 ± 8.0<0.001
WC (cm) (men/women)94 ± 9.5/79 ± 1094 ± 9.4/78 ± 1093 ± 9.8/77 ± 1093 ± 11/76 ± 10<0.001
BMI (kg m−2)26.1 ± 3.925.8 ± 3.925.6 ± 3.925.1 ± 3.0<0.001
Haemoglobin (g L−1)142 ± 12142 ± 12142 ± 12141 ± 120.02
RBC count (1012 L−1)4.8 ± 0.394.7 ± 0.464.6 ± 0.464.5 ± 0.64<0.001
MCV (fL)86 ± 2.988 ± 2.790 ± 3.093 ± 3.7<0.001
Leukocyte count (109 L−1)6.0 ± 1.56.3 ± 2.06.4 ± 1.76.8 ± 3.7<0.001
Smokers (%)14.321.730.347.4<0.001
High alcohol intake (%)3.33.54.16.4<0.001
Born in Sweden (%)85.988.089.389.5<0.001
Subgroup with measured blood glucose (= 4845)
n (% women)1231 (59)1168 (60)1208 (60)1238 (61) 
Age (years)56.4 ± 5.857.5 ± 6.057.8 ± 6.058.1 ± 5.8<0.001
WC (cm) (men/women)93 ± 8.5/77 ± 9.593 ± 9.1/77 ± 1092 ± 9.5/77 ± 1091 ± 10/75 ± 9.4<0.001
BMI (kg m−2)25.8 ± 3.625.7 ± 3.825.6 ± 3.825.0 ± 3.8<0.001
Smokers (%)8.314.723.943.2<0.001
Glucose (mmol L−1)4.91 ± 0.454.89 ± 0.434.88 ± 0.454.86 ± 0.460.014
HbA1c (%)4.6 ± 0.404.8 ± 0.414.8 ± 0.404.9 ± 0.42<0.001
Triglycerides (mmol L−1)1.241.191.141.13<0.001
HOMA1.421.391.331.25<0.001
Insulin (mIU L−1)6.536.426.175.80<0.001

RDW, HbA1c and glucose

In the subgroup with available glucose and HbA1c data, HbA1c was positively associated with RDW (4.6% vs. 4.9%, for 1st vs. 4th quartile, < 0.001) (Fig. 1). By contrast, concentrations of insulin (< 0.001) and glucose (= 0.014), HOMA (< 0.001), waist circumference (< 0.001) and triglyceride levels (< 0.001) were inversely associated with RDW (Table 1 and Fig. 1). A 1-SD increment in RDW (3.57 fL) corresponded to an increase in HbA1c of 0.11% (SE 0.006) in unadjusted regression analysis. This figure was 0.10% (SE 0.007) after adjustments for age, sex, MCV, glucose, waist circumference, BMI and smoking in a multiple linear regression. The relationship was similar in men (0.10%, SE 0.01) and women (0.10%, SE 0.008).

image

Figure 1. Mean glucose (mmol L−1) and HbA1c (%) levels in relation to quartiles of red cell distribution width (RDW) in 4845 non-diabetic subjects. Error bars represent ±2 SEM.

Download figure to PowerPoint

Incidence of DM

A total of 2944 (11.0%) subjects were identified as new DM patients during the follow-up. Mean RDW at baseline was 40.1 ± 3.3 and 40.9 ± 3.4 fL in those who did and did not develop DM, respectively (< 0.001). As expected, those who developed DM had higher fasting levels of glucose (5.3 ± 0.43 vs. 4.8 ± 0.44 mmol L−1), insulin (geometric mean 8.7 vs. 6.0 mIU L−1) and HbA1c (5.00 ± 0.44% vs. 4.75 ± 0.41%) at baseline (all P-values <0.001) (Table 2).

Table 2. Baseline characteristics of initially non-diabetic participants who did and did not develop diabetes mellitus during the follow-up
 Diabetes during follow-up P
YesNo
  1. Values of triglycerides, insulin and HOMA are presented as geometric means.

  2. WC, waist circumference; BMI, body mass index; RDW, red cell distribution width; HOMA, homeostatic model assessment.

  3. a

    Based on 478 cases with diabetes and 4367 without diabetes during follow-up.

n 294423765 
Age (years)59.0 ± 7.057.9 ± 7.7<0.001
Women (%)48.262.7<0.001
WC (cm) (men/women)99 ± 11/87 ± 1292 ± 9.4/77 ± 10<0.001
BMI (kg m−2)28.4 ± 4.325.3 ± 3.7<0.001
Leukocyte count (109 L−1)6.7 ± 3.66.4 ± 2.2<0.001
Smokers (%)28.528.40.90
High alcohol intake (%)5.14.20.04
RDW (fL)40.1 ± 3.440.9 ± 3.4<0.001
HbA1c (%)a5.0 ± 0.444.8 ± 0.41<0.001
Glucose (mmol L−1)a5.3 ± 0.434.8 ± 0.44<0.001
Triglycerides (mmol L−1)a1.441.15<0.001
Insulin (mIU L−1)a8.76.0<0.001
HOMAa2.01.3<0.001

The incidence of DM was significantly higher in subjects with low RDW at baseline (Fig. 2). The hazard ratio (HR) adjusted for age and sex was 1.56 [95% confidence interval (CI) 1.41–1.74] for the 1st versus the 4th quartile of RDW. The HR remained significant after further adjustments for waist circumference, BMI, smoking, leukocyte count, MCV, country of birth and alcohol intake (HR 1.48, 95% CI 1.29–1.70, Table 3; HR per 1 SD 1.24, 95% CI 1.19–1.30). The relationship between RDW and incidence of DM remained significant after further adjustments for glucose, HbA1c, insulin and triglycerides (Table 3). As expected, waist circumference (HR per 1 cm 1.05, 95% CI 1.03–1.07), HbA1c (HR per 1% 2.7, 95% CI 2.1–3.6), glucose (HR per 1 mmol L−1 5.1, 95% CI 4.0–6.5) and insulin (HR per 1 log unit 1.4, 95% CI 1.2–1.8) were all major risk factors for DM in the final multivariate model (all < 0.001).

Table 3. Incidence of diabetes mellitus (DM) in relation to sex-specific quartiles (Q1–Q4) of red cell distribution width (RDW)
 Q1Q2Q3Q4P for trend
  1. aHazard ratio adjusted for age and sex; badjusted for age, sex, body mass index (BMI), waist circumference, country of birth, smoking, leukocyte count, mean corpuscular volume (MCV) and high alcohol intake. cHazard ratio adjusted for age, sex, glucose and HbA1c; dadjusted for age, sex, country of birth, glucose, HbA1c, BMI, waist circumference, log insulin, log triglycerides, leukocyte count, MCV, smoking and high alcohol intake. Data are based on 4699 individuals and 458 events.

  2. PY, person-years; HR, hazard ratio; CI, confidence interval.

Full cohort (= 26709)
n 6689657668386606 
Incident DM (n)905820649570 
Incident DM (per 1000 PY)9.58.86.86.3 
HR (95% CI)1.56 (1.41–1.74)1.43 (1.29–1.60)1.07 (0.95–1.19)1.00<0.001
HR (95% CI)1.48 (1.29–1.70)1.39 (1.23–1.57)1.05 (0.93–1.18)1.00<0.001
Subgroup with measured blood glucose (= 4845)
n 1231116812081238 
Incident DM (n)13512712690 
HR (95% CI)1.43 (1.09–1.86)1.40 (1.07–1.84)1.37 (1.05–1.80)1.000.02
HR (95% CI)1.73 (1.31–2.27)1.58 (1.20–2.07)1.49 (1.14–1.96)1.00<0.001
HR (95% CI)1.61 (1.11–2.33)1.47 (1.07–2.03)1.33 (0.99–1.77)1.000.02
image

Figure 2. Incidence of diabetes mellitus in relation to quartiles of red cell distribution width.

Download figure to PowerPoint

There was a significant interaction between RDW and age at baseline, with respect to incidence of DM. The HRs (per 1 SD reduction in RDW, adjusted for risk factors) were 1.07 (95% CI 0.96–1.18), 1.22 (95% CI 1.11–1.34) and 1.46 (95% CI 1.28–1.65) for subjects aged 45–54, 55–64 and 65–73 years, respectively. A significant interaction was also observed between RDW and MCV (see below). No interactions were seen between RDW and smoking, waist circumference or alcohol intake. The HR (per 1 SD reduction in RDW, adjusted for risk factors) was stronger for women than for men (1.32, 95% CI 1.21–1.43, < 0.001 vs. 1.11, 95% CI 1.03–1.20, = 0.01). However, there was no significant interaction between RDW and sex with respect to incidence of DM.

The results were virtually identical after further adjustment for dietary intake of vitamin B12, folate and iron in a sensitivity analysis (HR 1.48, 95% CI 1.29–1.70 for 1st vs. 4th quartile of RDW).

A competing risk analysis was performed considering death as a competing event. After adjustment for risk factors, the subhazard ratios were 1.64 (95% CI 1.4–1.9), 1.50 (95% CI 1.3–1.7), 1.10 (95% CI 0.98–1.3) and 1.00 (reference) for 1st, 2nd, 3rd and 4th quartiles, respectively. This analysis indicates that differences in mortality between quartiles of RDW do not explain the relationship with incidence of DM.

Impaired fasting glucose

A total of 416 individuals (57.5% men) had impaired fasting glucose (IFG; i.e. fasting blood glucose concentration of 5.6–6.1 mmol L−1) at the baseline examination. RDW was significantly and positively associated with HbA1c in subjects with IFG. A 1-SD increase in RDW corresponded to an increase in HbA1c of 0.095% (SE 0.02) in an unadjusted regression analysis and of 0.16% (SE 0.03) after adjustments in a multiple linear regression for age, sex, MCV, glucose, BMI, waist circumference and smoking.

In the group with IFG, 149 (35.8%) individuals developed DM during follow-up. After adjustment for age, sex, HbA1c, glucose, MCV, leukocyte count, BMI, waist circumference, country of birth, alcohol intake and smoking, the HR of incident DM was 2.15 (95% CI 1.12–4.14), comparing the 1st versus the 4th quartiles of RDW.

RDW, HbA1c and incidence of DM by quartiles of MCV

As high RDW could be present both in subjects with macrocytosis and in those with normocytosis, the relationship between RDW, HbA1c and DM was investigated by sex-specific quartiles of MCV (Table 4). RDW was positively associated with HbA1c in subjects with a low or medium MCV (i.e. quartiles 1–3). After adjustment for age and sex, a 1-SD change in RDW corresponded to a change in HbA1c of 0.19–0.23% in subjects with MCV in quartiles 1–3 (Table 4). This relationship was weaker in subjects with a high MCV (i.e. quartile 4).

Table 4. HbA1c and incidence of diabetes mellitus (DM) in relation to 1 SD reduction in red cell distribution width (RDW), by quartiles of mean corpuscular volume (MCV)
 HbA1cIncidence of DM
  1. Data are presented as age- and sex-adjusted beta-coefficients (SE) for HbA1c and HR (95% CI) for incident DM, per 1 SD reduction in RDW (3.57 fL).

  2. Cut-off points for MCV (median, interquartile range): men, 89.1 (86.6–91.9); women, 89.4 (87.0–92.1).

  3. HR, hazard ratio.

Quartile of MCV n %-units change in HbA1c per 1 SD reduction in RDWN (n events)HR for DM per 1 SD reduction in RDW
Q1 (low)1208−0.19 (0.02)6722 (919)1.20 (1.08–1.34)
Q21239−0.23 (0.02)6612 (749)1.24 (1.11–1.38)
Q31169−0.21 (0.02)6733 (679)1.15 (1.03–1.29)
Q4 (high)1229−0.10 (0.01)6942 (597)1.08 (0.98–1.18)

A low RDW was significantly associated with an increased incidence of DM in subjects with a low or medium MCV (quartiles 1–3, Table 4). The HRs per 1 SD reduction in RDW were 1.20 (95% CI 1.08–1.34), 1.24 (95% CI 1.11–1.38) and 1.15 (95% CI 1.03–1.29) for quartiles 1–3, respectively. A low RDW was not significantly associated with incidence of DM in subjects with a high MCV (Table 4).

Discussion

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conflict of interest statement
  8. Acknowledgements
  9. References

Recent studies from clinical and population-based cohorts, including the MDC study, have demonstrated associations between high RDW and increased risk of various adverse health outcomes, such as increased mortality [2], increased incidence of heart failure [3] and adverse prognosis in patients with heart failure or coronary heart disease [5, 6]. The results of the present study of initially non-diabetic subjects from the general population show that individuals with low RDW had substantially higher risk of developing DM. High RDW, which is a risk factor for heart failure and mortality, was a protective factor for new-onset DM. This relationship remained after adjustment for a wide range of potential confounding factors, including HbA1c and glucose levels.

Another important finding was the significant and positive relationship between RDW and HbA1c, which is in accordance with recent data from non-diabetic participants in the NHANES study [11]. This finding might seem conterintuitive, given the inverse relationships between RDW and insulin, glucose, waist circumference and incidence of DM (Fig. 1). As expected, HbA1c, waist circumference and glucose and insulin concentrations were all major risk factors for DM in this study.

HbA1c and glucose were differently related to RDW. It is known that HbA1c provides an incomplete view of the historical glucose exposure [26, 27] and the results show that the discrepancy between HbA1c and glucose is related to RDW. HbA1c is formed by a slow irreversible reaction between haemoglobin and glucose [28, 29]. The percentage of HbA1c is mainly determined by plasma glucose levels and the time the RBCs have been exposed to glucose [28, 29]. The average lifespan of the RBC is considered to be approximately 120 days; during this time, the density of the cells increases and the surface area gradually decreases [30-33]. However, available data show that there are substantial differences in RBC lifespan. A study of 40 healthy volunteers reported a mean RBC survival of 123 days, with a standard deviation of 23 days [34]. Variations in RBC lifespan of this magnitude are sufficient to cause considerable alterations in HbA1c [28]. Hence, the positive correlation between RDW and HbA1c could reflect a relationship between high variability of the RBC volumes and a high proportion of old dense cells which have been exposed to glucose for a long time. If this hypothesis is true, it could explain why HbA1c and glucose have different relationships with RDW. Furthermore, it could be speculated that the previously reported relationships between high RDW, heart failure and mortality could be related to the properties and functions of senescent RBCs. The results seem to be of clinical importance for screening and diagnosis of DM, as HbA1c overestimates the degree of hyperglycaemia if RDW is high and underestimates the DM risk in subjects with low RDW.

Diabetes mellitus and hyperglycaemia have several additional effects on RBCs [13], besides formation of HbA1c. DM is associated with reduced deformability and changes in mechanical properties of the RBCs [13, 14], increased adhesion [35] and increased osmotic fragility [36]. Furthermore, some studies have reported that the average lifespan of RBCs is reduced in individuals with DM [15-17, 37]. In the present study, the group with the lowest quartile of RDW had the highest glucose, insulin and HOMA levels. It seems possible that the rise in fasting glucose and postprandial hyperglycaemia could be sufficient to change the mechanical properties of the RBCs, to reduce survival and create a more homogenous population of cells, even in individuals with fasting glucose in the pre-diabetic range. If so, this could explain the increased risk of DM in subjects with low RDW.

It should be acknowledged that several factors could increase RDW, besides the age distribution of the RBCs. Increased erythropoiesis, for example after bleeding episodes or due to haemolysis, is a common reason for increased RDW in the clinical setting. Other factors include nutritional deficiencies, especially of iron, folate and vitamin B12 [38]. However, deficiency of these nutrients is an unlikely cause of reduced incidence of DM in this population. Haemoglobin levels were similar in all participants irrespective of quartile of RDW, and adjustments for haemoglobin or for dietary intake of iron, folate and vitamin B12 did not change the results. High RDW has been associated with systemic inflammation [3, 39]. However, as inflammation is a risk factor for DM [40], this is an unlikely explanation for the reduced incidence of DM in subjects with high RDW.

The large population-based sample of middle-aged men and women and the prospective design are major strengths of this study. The RBC indices were measured using an automated cell counter in subjects aged 45–73 years, and the MDC cohort was followed over a mean time of 14 years. Much information about glucose metabolism and other biomarkers was available, and adjustment was made for many potential confounding factors. A limitation of the study is that information about 2-h post-load glucose was not available. It is possible that some individuals were misclassified at the baseline examination. We also lacked information about reticulocyte counts. Further, to what extent the RDW changed during the follow-up period is unclear. However, a study of healthy subjects demonstrated that yearly variations in RDW are comparable to those of RBC and other haematological measures and that the intra-individual variability is small [41]. It is also noteworthy that the RDW values from the baseline examination remained predictive of DM over a long time period. Furthermore, it is unclear whether the results can be generalized to non-Caucasian populations.

New cases of DM were identified from several independent data sources. The registers of out- and inpatients include nationwide data, covering hospital visits in the whole Swedish population. The register of prescribed drugs records all filled prescriptions from all pharmacies in Sweden since 2005. The HbA1c register covers the entire population in the city of Malmö. Many cases were detected during repeated surveys of the MDC cohort. Of note, the relationship between RDW and incidence of DM was essentially the same for each of these data sources, which strongly supports the validity of the endpoints. It is well known that DM often remains undetected for a long period, and cases that did not seek medical care might have been missed. However, the coverage of the registers is very good and we have no reason to question the case validity of the endpoints.

In summary, low RDW was associated with a markedly increased risk of developing DM. Low RDW was related to higher waist circumference, HOMA and glucose, insulin and triglyceride concentrations. By contrast, RDW was significantly and positively associated with HbA1c. We propose that RBC survival rates on average are higher in subjects high RDW, leading to higher HbA1c due to increased duration of glucose exposure. RDW is an easily measurable biomarker that could improve risk assessment for individuals at risk of developing DM.

Conflict of interest statement

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conflict of interest statement
  8. Acknowledgements
  9. References

The authors have no conflicts of interest to disclose.

Acknowledgements

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conflict of interest statement
  8. Acknowledgements
  9. References

This study was supported by the Swedish Heart and Lung Foundation (grant no. 20100244), the Swedish Research Council (grant no. 2011-3891), the Ernhold Lundström Foundation and funds from Lund University and Skåne University Hospital. The funding organizations had no role in the design and conduct of the study, the collection, management, analyses and interpretation of the data or the preparation or approval of the manuscript.

References

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conflict of interest statement
  8. Acknowledgements
  9. References