Association Between Serum Bioavailable Testosterone Concentration and the Ratio of Glycated Albumin to Glycated Hemoglobin in Men With Type 2 Diabetes

Serum bioavailable testosterone concentrations were measured in 222 consecutive men with type 2 diabetes recruited from the outpatient clinic at the Kyoto Prefectural University of Medicine. Relationships between serum bioavailable testosterone and hemoglobin concentrations and between serum bioavailable testosterone concentration and GA-to-A1C ratio were investigated. In addition, we evaluated relationships between GA-to-A1C ratio and age, duration of diabetes, blood pressure, plasma lipid concentration, BMI, waist circumference, severity of diabetic retinopathy, severity of diabetic nephropathy defined by urinary albumin excretion, presence of cardiovascular disease (CVD), smoking status, and current treatment for diabetes.

Blood samples were obtained in the morning. Bioavailable testosterone was separated using precipitation of testosterone bound to globulins with 50% ammonium sulfate, and serum bioavailable testosterone concentrations were measured by the liquid chromatography–tandem mass spectrometry using a modification method based on the use of picolinoyl derivatization (10). Intra-assay and interassay coefficients of variation (CVs) for serum bioavailable testosterone concentrations at 1 pg/ml were 4.73 and 12.94%, respectively. Hemoglobin concentrations were analyzed within 4 h of blood drawing using a SYSMEX XE-2100 autoanalyzer (Sysmex Corporation, Kobe, Japan). A1C was measured by high-performance liquid chromatography using an ADAMS-A1c HA-8160 (Arkray, Kyoto, Japan). Interassay CVs determined using representative blood samples with 5.2 and 10.9% A1C were 0.63 and 0.45%, respectively. GA was determined by an enzymatic method using albumin-specific proteinase, ketoamine oxidase, and an albumin assay reagent (Lucica GA-L; Asahi Kasei Pharma, Tokyo, Japan), with a Hitachi 7600 autoanalyzer (Hitachi Instrument Service, Tokyo, Japan). Interassay CVs determined using representative serum samples with 17.2 and 26.9% GA were 1.08 and 1.37%, respectively. Plasma total cholesterol, HDL cholesterol, and triglyceride concentrations were assessed using standard enzymatic methods. Mean values for biochemical parameters obtained during the preceding year were used for statistical analysis. Urinary albumin and creatinine concentration was determined in early morning spot urine. Urinary albumin excretion was measured with an immunoturbidimetric assay. A mean value for urinary albumin excretion was determined from three urine collections.

Type 2 diabetes was diagnosed according to the Report of the Expert Committee on the Diagnosis and Classification of Diabetes Mellitus (11). Retinopathy was graded as no diabetic retinopathy, simple diabetic retinopathy, or proliferative diabetic retinopathy. Nephropathy was graded as normoalbuminuria (urinary albumin excretion <30 mg/g creatinine) or microalbuminuria (30–300 mg/g creatinine). Sitting blood pressure was measured after a 5-min rest. CVD was defined as a previous myocardial or cerebral infarction based on the clinical history or physical examination. Subjects were classified as nonsmokers, past smokers, or current smokers according to a self-administered questionnaire. Patients were excluded if they had undergone castration for treatment of testicular or prostate cancer or were taking any medications known to affect sex hormone concentrations (e.g., antiandrogenic agents for prostate cancer). In addition, patients with macroalbuminuria were excluded since advanced diabetic nephropathy can influence hemoglobin concentration. Moreover, patients with malignant disease, liver cirrhosis, thyroid disorders, hematologic disease, or infectious disease were excluded from this study. To minimize effects of diabetes treatment on time-dependent variations of A1C and GA levels, we selected patients whose A1C levels had been stable for at least the preceding 3 months. Approval for the study was obtained from the local research ethics committee, and informed consent was obtained from all participants.

Means or frequencies of potential confounding variables were calculated as appropriate. Unpaired Student's t tests or ANOVAs were conducted to assess statistical significance of differences between groups using Stat View software (version 5.0; SAS Institute, Cary, NC). Because plasma triglyceride concentration showed a skewed distribution, logarithmic (log) transformation was carried out before performing a correlation analysis. Correlations between serum bioavailable testosterone concentration and hemoglobin concentration or GA-to-A1C ratio, as well as between GA-to-A1C ratio and age, duration of diabetes, BMI, or other variables, were examined by Pearson's correlation analyses. All continuous variables are presented as means ± SD. Multiple linear regression analysis was performed to assess the combined influence of variables on hemoglobin concentration or GA-to-A1C ratio. To examine the effects of various factors on hemoglobin concentration or GA-to-A1C ratio, the following factors were considered independent variables: serum bioavailable testosterone concentration, age, duration of diabetes, BMI, systolic blood pressure, plasma total cholesterol concentration, and smoking status. A P value <0.05 was considered statistically significant.

Characteristics of the 222 men with type 2 diabetes enrolled in this study are shown in Table 1. Mean GA-to-A1C ratio was 2.94 ± 0.38.

Relationships between GA-to-A1C ratio and other variables are shown in Table 2. Age, duration of diabetes, and plasma HDL cholesterol concentration were positively associated with GA-to-A1C ratio, while BMI, waist circumference, systolic blood pressure, diastolic blood pressure, plasma total cholesterol concentration, and log(plasma triglyceride concentration) were negatively associated with GA-to-A1C ratio. A positive correlation was found between serum bioavailable testosterone and hemoglobin concentrations (r = 0.368, P < 0.0001 (Fig. 1A), while a negative correlation was found between serum bioavailable testosterone concentration and GA-to-A1C ratio (r = −0.278, P < 0.0001) (Fig. 1B). No significant correlation was found between serum bioavailable testosterone and albumin concentrations (r = 0.147, P = 0.0625). Multiple regression analyses identified serum bioavailable testosterone concentration (β = 0.187, P = 0.0062), age (β = −0.204, P = 0.0075), BMI (β = 0.151, P = 0.0302), systolic blood pressure (β = 0.173, P = 0.0090), and plasma total cholesterol (β = 0.155, P = 0.0141) as independent determinants of hemoglobin concentration; moreover, serum bioavailable testosterone concentration (β = −0.155, P = 0.0381) and plasma total cholesterol (β = −0.170, P = 0.0144) were identified as independent determinants of GA-to-A1C ratio (Table 3).

GA-to-A1C ratio did not differ between patients with normoalbuminuria and microalbuminuria (2.94 ± 0.40 vs. 2.97 ± 0.35, P = 0.5162). GA-to-A1C ratio was higher in patients with proliferative diabetic retinopathy (3.16 ± 0.25, P = 0.0128) or simple diabetic retinopathy (3.10 ± 0.28, P = 0.0142) than that in patients without diabetic retinopathy (2.91 ± 0.40). GA-to-A1C ratio did not differ between patients with or without CVD (2.97 ± 0.37 vs. 2.95 ± 0.38, P = 0.7238). GA-to-A1C ratio was higher in patients treated with than without insulin (3.15 ± 0.45 vs. 2.87 ± 0.32, P < 0.0001). GA-to-A1C ratio did not differ between patients treated with or without statin (2.88 ± 0.37 vs. 2.96 ± 0.38, P = 0.1821), angiotensin II receptor blocker, or ACE inhibitor (2.95 ± 0.36 vs. 2.94 ± 0.39, P = 0.7533). GA-to-A1C ratio was higher in patients treated with than without antiplatelet agent (3.05 ± 0.46 vs. 2.92 ± 0.35, P = 0.0416). GA-to-A1C ratio did not differ according to smoking status (2.90 ± 0.27, 2.93 ± 0.41, and 3.04 ± 0.46 for current smokers, past smokers, and nonsmokers, respectively).

Serum bioavailable testosterone concentration correlated positively with hemoglobin concentration in men with type 2 diabetes, which is compatible with previous findings in the general population. Serum bioavailable testosterone concentration correlated negatively with GA-to-A1C ratio. Multiple regression analysis also identified serum bioavailable testosterone as a determinant of hemoglobin concentration or GA-to-A1C ratio.

GA and A1C have a tendency to vary in parallel, although the normal range of variation of GA-to-A1C ratio has not yet been determined. GA and A1C are equivalent measures of glycemic control in general, although the former is a short-term and the latter a long-term marker. In certain diabetic patients in whom A1C measurement proves insufficient for clinical management (8,9), measurement of serum GA represents an alternative assessment of glycemic control. For example, anemia is an important factor affecting A1C levels. Thus, changes in the GA-to-A1C ratio would indicate artifactual change, which lends GA-to-A1C ratio clinical value.

Our results suggest that we may underestimate A1C level in male diabetic patients with hypogonadism because of a negative association between serum bioavailable testosterone concentration and GA-to-A1C ratio that arises partly from a positive association between serum bioavailable testosterone and hemoglobin concentrations. Men with diabetes have significantly lower plasma concentrations of endogenous androgen than nondiabetic men (12–15); endogenous androgen concentrations also decline with advancing age (16). Close attention should be paid to diabetic men with poor diabetes control or advanced age with regard to evaluation of A1C levels.

GA-to-A1C ratio was negatively associated with BMI or waist circumference in the present study. Koga et al. (17) made a similar observation, suggesting that increased serum albumin turnover in obese subjects could keep serum GA low relative to plasma glucose concentrations. We additionally found age to be positively associated with GA-to-A1C ratio and systolic blood pressure or plasma total cholesterol concentration to be negatively associated with GA-to-A1C ratio. Although the reasons for these associations are not clear, possible explanations may be that age (r = −0.390, P < 0.0001) is negatively associated with hemoglobin while systolic blood pressure (r = 0.253, P = 0.0002) or plasma total cholesterol concentration (r = 0.258, P = 0.0001) is positively associated with hemoglobin concentration. Cholesterol is a precursor of testosterone, and serum bioavailable testosterone concentration is positively associated with plasma total cholesterol concentration (r = 0.194, P = 0.0039). GA-to-A1C ratio was higher according to progression of diabetic retinopathy. Finally, ongoing treatment for diabetic patients such as insulin or antiplatelet agent was associated with GA-to-A1C ratio.

We excluded patients with macroalbuminuria in this study because advanced diabetic nephropathy can influence hemoglobin concentration. MacIsaac RJ et al. (18) demonstrated that patients with type 2 diabetes can progress to a significant degree of renal impairment while remaining normoalbuminuric. Certainly, estimated glomerular filtration rates were below 60 ml/min per 1.73 m² in 42 of 222 patients in this study. However, a significant correlation between serum bioavailable testosterone and hemoglobin concentrations (r = 0.343, P < 0.0001) or between serum bioavailable testosterone concentration and GA-to-A1C ratio (r = −0.292, P < 0.0001) persisted even after excluding patients with estimated glomerular filtration rate <60 ml/min per 1.73 m².

That we could not have control groups of nondiabetic men with and without hypogonadism and/or hypogonadal men before and after testosterone replacement constitutes a limitation to our study, as we cannot see whether GA-to-A1C ratio is different and varies in these populations.

To our knowledge, this is the first study to investigate the relationship between serum bioavailable testosterone concentration and GA-to-A1C ratio and the relationship between serum bioavailable testosterone and hemoglobin concentrations in men with type 2 diabetes. The results of this study have raised new issues concerning the pathophysiologic characteristics of anemia in male diabetic patients with hypogonadism. The mechanism through which testosterone stimulates erythropoiesis is unclear. Testosterone enhances proliferation of erythroid burst-forming units and colony-forming units by stimulating specific nuclear receptors (19). Medras et al. (20) demonstrated the significant decrease of aldolase and pyruvate kinase activity in erythrocytes in men with hypogonadism, which may decrease the lifespan of erythrocyte. Moreover, Solomon et al. (21) reported that androgen therapy prolongs red cell survival and increases red cell production. We suggest that low serum bioavailable testosterone concentration could be an underrecognized contributor to anemia, while a low serum bioavailable testosterone concentration could affect GA-to-A1C ratio and thus cause spuriously low A1C values. Our findings should be confirmed in larger populations and in other clinical settings before they can be considered for clinical applications.

In conclusion, serum bioavailable testosterone concentration correlated positively with hemoglobin concentration and negatively with GA-to-A1C ratio in men with type 2 diabetes, which may lead to underestimation of A1C in hypogonadal men with type 2 diabetes, although in this study we cannot clarify whether these findings are applicable to nondiabetic men.