Impact of Mean Cell Hemoglobin on Hb A 1c Defined Glycemia Status

Transcription

1 Clinical Chemistry 62: (2016) Endocrinology and Metabolism Impact of Mean Cell Hemoglobin on Hb A 1c Defined Glycemia Status Santiago Rodriguez-Segade, 1,2* Javier Rodriguez Garcia, 2 José M. García-López, 3 Francisco Gude, 4 Felipe F. Casanueva, 3,5 Santiago RS-Alonso, 6 and Félix Camiña 1 BACKGROUND: Several hematological alterations are associated with altered hemoglobin A 1c (Hb A 1c ). However, there have been no reports of their influence on the rates of exceeding standard Hb A 1c thresholds by patients for whom Hb A 1c determination is requested in clinical practice. METHODS: The initial data set included the first profiles (complete blood counts, Hb A 1c, fasting glucose, and renal and hepatic parameters) of all adult patients for whom such a profile was requested between 2008 and 2013 inclusive. After appropriate exclusions, patients remained in the study. Linear and logistic regression models were adjusted for demographic, hematological, and biochemical variables excluded from the predictors. RESULTS: Mean corpuscular hemoglobin (MCH) and mean corpuscular volume (MCV) correlated negatively with Hb A 1c. Fasting glucose, MCH, and age emerged as predictors of Hb A 1c in a stepwise regression that discarded sex, hemoglobin, MCV, mean corpuscular hemoglobin concentration (MCHC), serum creatinine, and liver disease. Mean Hb A 1c in MCH interdecile intervals fell from 6.8% (51 mmol/mol) in the lowest ( 27.5 pg) to 6.0% (43 mmol/mol) in the highest ( 32.5 pg), with similar results for MCV. After adjustment for fasting glucose and other correlates of Hb A 1c,a1pgincrease in MCH reduced the odds of Hb A 1c defined dysglycemia, diabetes and poor glycemia control by 10% 14%. CONCLUSIONS: For at least 25% of patients, low or high MCH or MCV levels are associated with increased risk of an erroneous Hb A 1c based identification of glycemia status. Although causality has not been demonstrated, these parameters should be taken into account in interpreting Hb A 1c levels in clinical practice American Association for Clinical Chemistry Hemoglobin (Hb) 7 A 1c is a measure of time-averaged glycemia. High Hb A 1c levels are strongly predictive of diabetic complications (1, 2), and possibly also predictive of cardiovascular morbidity among both diabetic and nondiabetic patients (3, 4). The American Diabetes Association (ADA) regards Hb A 1c as a parameter that in the absence of abnormal red cell turnover allows diagnosis of diabetes and the screening of persons at high risk of diabetes (2). Conditions or events in which the mean age of red blood cells is altered include blood loss, blood transfusions, anemia, chronic kidney or liver disease, ingestion of high doses of vitamin C, erythropoietin treatment, and hematological conditions such as sickle-cell disease (5 7). Unusually low Hb A 1c levels are found when mean erythrocyte life span is shortened, as by glucose-6-phosphate dehydrogenase deficiency; unusually high levels when erythrocyte life span is lengthened, as by deficiency of vitamin B 12, folate, or iron; and both high and low levels may be found among patients with end-stage kidney disease (8). However, even among hematologically normal people, erythrocyte life span varies sufficiently to cause clinically relevant differences in Hb A 1c (9). Most studies of the impact of such effects have investigated the influence of iron deficiency on the Hb A 1c - based diagnosis of diabetes and prediabetes in community samples, finding that iron deficiency raises Hb A 1c levels, without any concomitant rise in glucose indices, even in the absence of anemia (10 15). A few studies have also reported negative correlation between Hb A 1c on the one hand and, on the other, hemoglobin, mean 1 Department of Biochemistry and Molecular Biology, 2 University Hospital Clinical Biochemistry Laboratory, 3 Division of Endocrinology, 4 Clinical Epidemiology Unit, Universidade de Santiago de Compostela, Santiago de Compostela, A Coruña, Spain; 5 Physiopathology of Obesity and Nutrition Biomedical Research Network Consortium; and 6 the Division of Pneumology, University Hospital Clinical Complex (CHUAC), A Coruña, Spain. * Address correspondence to this author at: Clinical Biochemistry Laboratory, Complejo Hospitalario Universitario de Santiago, Travesía de la Choupana s/n, Santiago de Compostela, Spain. Fax ssegade@telefonica.net. Received March 18, 2016; accepted July 27, Previously published online at DOI: /clinchem American Association for Clinical Chemistry 7 Nonstandard abbreviations: Hb A 1c, hemoglobin A 1c ; MCH, mean corpuscular hemoglobin; MCV, mean corpuscular volume; MCHC, mean corpuscular hemoglobin concentration; JDS, Japan Diabetes Society; NGSP, National Glycohemoglobin Standardization Program; egfr, estimated glomerular filtration rate; eag, estimated average glucose; JDS/JSCC, Japanese Diabetes Society and Japanese Society for Clinical Chemistry; NGSP, US National Glycohemoglobin Standardization Program. 1570

2 MCH Affects Hb A 1c Status corpuscular hemoglobin (MCH), and mean corpuscular volume (MCV) (7, 14, 15). However, although the existence of these latter correlations is generally acknowledged, especially in relation to the use of Hb A 1c for diagnosis of diabetes (16), the magnitude of their consequences for diagnosis and management of dysglycemic conditions i.e., the proportion of patients that, because of these correlations, are misclassified by standard Hb A 1c cutoffs is uncertain. It is not even clear which hematological indices are most appropriate for identification of problematic patients. In this study we investigated these issues in a large group of patients for whom determination of Hb A 1c had been requested for any reason. Adjusting for the effects of other conditions known to alter Hb A 1c levels, we determined the magnitude of the influence of hematological indices on the overstepping of the Hb A 1c thresholds most commonly employed for diagnosis and management of diabetes and prediabetes. Materials and Methods STUDY PARTICIPANTS The University Hospital Complex, Santiago de Compostela (Spain), is a tertiary center serving an almost exclusively white population of approximately people. Its diabetic outpatient clinics are attended by almost entirely local patients requiring insulin or oral antidiabetic drugs. The present retrospective study was approved by its Ethical Review Board. The individuals initially included in the current retrospective study were all those patients aged 18 years for whom an analytical profile consisting of (a) the glycemic parameters, Hb A 1c, and fasting glucose, (b) the hematological indices, hemoglobin, mean corpuscular hemoglobin (MCH), mean corpuscular hemoglobin concentration (MCHC), mean corpuscular volume (MCV), and erythrocyte count, and (c) the renal and hepatic indices, creatinine, estimated glomerular filtration rate (egfr), aspartate aminotransferase (AST), alanine aminotransferase (ALT), -glutamyltransferase (GGT), bilirubin, and albumin was obtained between 2008 and 2013 inclusive (only the first such profile of each patient during this period was considered). Patients were excluded if they were acutely ill, had undergone blood transfusions 3 months before profiling, or were pregnant. Patients with no previous diagnosis of dysglycemia were classified as diabetic, prediabetic, or normoglycemic in accordance with the American Diabetes Association (ADA) criteria (2). Patients with hemoglobin 12g/dL (women) or 13g/dL (men) were deemed to be anemic (17), and patients with AST 30 U/L or ALT 31 U/L or GGT 38 U/L or bilirubin 1.20 mg/dl or albumin 3.5g/dL were deemed to have liver disease. ANALYTICAL METHODS Blood samples were collected in the morning following fasting for at least 8 hours; for any one patient, all samples were collected at the same visit. Hb A 1c values referenced as per the Japanese Diabetes Society (JDS) and Japanese Society for Clinical Chemistry (JSCC) (Hb A 1c JDS/JSCC ) were determined by high-performance liquid chromatography in a Menarini Diagnostics HA 8160 analyzer. These JDS/JSCC-referenced Hb A 1c values were converted to Diabetes Control and Complications Trial (DCCT)-aligned units in accordance with the US National Glycohemoglobin Standardization Program (NGSP) (17, 18). The interassay CV was 1.6% at an Hb A 1c level of 5.9% (41 mmol/mol) and 0.9% at an Hb A 1c level of 11% (97 mmol/mol). Fasting glucose was measured in serum by the glucose oxidase peroxidase method. Serum creatinine, AST, ALT, GGT, and albumin were determined using standard kits on an Advia 2400 analyzer (Siemens Healthcare Diagnostics). egfr was calculated using the Modification of Diet in Renal Disease formula (19). Also, estimated average glucose (eag) was calculated based on Hb A 1c (20). Hemoglobin, MCH, MCV, MCHC, and erythrocyte count were determined in an Advia 1200 counter (Siemens Healthcare Diagnostics). All analyses were performed on the day of collection in the Clinical Biochemistry Laboratory of the University Hospital Complex. STATISTICAL ANALYSES Data are presented as: means SDs for normally distributed variables; as medians, with first and third quartiles in parentheses for skewed continuous variables; and as numbers with percentages in parentheses for categorical variables. Comparisons between distribution locations were performed using ANOVA or Mann Whitney U-tests, as appropriate. Relationships between Hb A 1c and other variables were investigated via Pearson and partial correlation analyses and by stepwise multivariate linear regression. Additionally, patients were classified by hemoglobin, MCH and MCV deciles for evaluation of relationships between these variables and Hb A 1c or fasting glucose using a general linear model with adjustment for age, sex, creatinine, and liver disease. The influence of increasing MCH or MCV on the odds of exceeding the Hb A 1c thresholds most commonly employed for diagnosis of diabetes [6.5% (48 mmol/ mol)] and prediabetes [5.7% (39 mmol/mol)], and as goals in management [7% (53 mmol/mol) and 8% (64 mmol/mol)], was investigated using multiple logistic regression with and without adjustment for age, sex, hemoglobin, creatinine, and liver disease. The criterion for statistical significance was P All statistical analyses were performed using SPSS17 (SPSS Inc.). Clinical Chemistry 62:12 (2016) 1571

3 Table 1. Characteristics of the study participants. a Number Hematological characteristics Age, years 62.2 ± 17.6 Anemic, n (%) 5712 (26.1) Range Hemoglobin, g/dl 13.3 ± 1.8 Female, n (%) (46.0) Range Glycemic status, n (%) MCV, fl 88.1 ± 6.1 Diabetes (48.6) Range Prediabetes 6453 (29.5) MCH, pg 30.1 ± 2.3 Normal 4790 (21.9) Range Measures of glycemia MCHC, g/dl 34.1 ± 1.2 Hb A 1c, (%) 6.5 ± 1.8 Range Range Erythrocytes, 10 6 /μl 4.45 ± 0.59 Hb A 1c, mmol/mol 48 ± 17 Range Range Renal profile Fasting glucose, mg/dl 134 ± 60 Kidney disease, n (%) 2634 (12.1) Range Creatinine, mg/dl 0.9 ( ) Hepatic profile Range Liver disease, n (%) 9102 (41.7) egfr, ml min 1 (1.73 m 2 ) 1 77 (63 91) AST, U/L 15 (11 20) Range range 8 52 ALT, U/L 18 (13 27) Range GT, U/L 18 (10 36) Range 5 34 Bilirubin, mg/dl 0.60 ( ) Range Albumin, g/dl 4.1 ± 0.5 Range a Dataaremeans±SDs,n(%),or50th(25th 75thcentiles)andranges(5thto95thcentiles).Toconvertconcentrationsfrommg/dLtommol/L,multiplyby forglucose, for creatinine. To convert bilirubin from mg/dl to μmol/l, multiply by Results Of the patients initially considered for the study, 3974 were excluded: 1617 for pregnancy, 473 for acute illness, and 1884 for lack of a requisite analytical value; none had had any blood transfusions in the relevant period. The clinical characteristics of the remaining are summarized in Table 1. None had been reported to have variant hemoglobins or thalassemia. Of the 6505 with no prior diagnosis of dysglycemia, 180 were classified on the basis of their Hb A 1c values as diabetic, and 1535 as prediabetic patients. The hematological index most closely correlated with Hb A 1c in the whole study group was MCH, followed by MCV; in both cases the correlation was negative and was even more negative when adjustment was made for age, creatinine and fasting glucose (see Table 1 in the Data Supplement that accompanies the online version of this article at content/vol62/issue12). Similar results held for men, women, patients with anemia, patients without anemia, and normoglycemic, prediabetic and diabetic patients with the following exceptions: for diabetic patients MCV was more closely correlated with Hb A 1c than was MCH before adjustment for age, creatinine and fasting glucose; for anemic patients the same was true both before and after adjustment. Among normoglycemic patients, erythrocyte count was correlated with Hb A 1c almost as strongly as MCV after adjustment; among prediabetic patients both erythrocyte count and MCHC showed considerable correlation with Hb A 1c after adjustment; and among the three glycemia classes the correlation of Hb A 1c with MCH, MCV, and erythrocyte count decreased in the order normoglycemia prediabetes 1572 Clinical Chemistry 62:12 (2016)

4 MCH Affects Hb A 1c Status diabetes. No correlation was found between any of these hematological indices and fasting glucose. Some 26.1% of patients were anemic (Table 1). Anemia was associated with significant rises in Hb A 1c of about 0.15% among men, women and patients without kidney disease, and about 0.3% among patients with liver disease, but was not associated with any significant alteration of Hb A 1c among patients with kidney disease or patients without liver disease (see online Supplemental Table 2). Regardless of anemia status, men did not differ from women in regard to Hb A 1c, but kidney disease was associated with significantly increased Hb A 1c, and so was liver disease, except among nonanemic patients. In the whole study group and in the subgroups of diabetic, prediabetic, and normoglycemic patients, the proportions with MCH 28 pg were, respectively, 14.1%, 15.1%, 13.3%, and 12.9%; and the proportions with MCH 32 pg, respectively, 15.3%, 13.9%, 16.6%, and 16.8%. In the whole study group and in the subgroups defined by sex, anemia, kidney disease, and liver disease, patients with MCH 28 pg had significantly higher Hb A 1c levels than those with MCH 28 pg, and patients with MCH 32 pg had significantly higher Hb A 1c levels than patients with MCH 32 pg (see online Supplemental Table 3). Among patients with MCH 28 pg, Hb A 1c levels were significantly affected by kidney and liver disease but not by sex or anemia, while among patients with MCH 32 pg Hb A 1c levels were significantly affected by sex and kidney disease but not by liver disease or anemia. The top panel of Table 2 shows, for each interdecile range of MCH, the number of patients in that range, the proportion of those patients that were female, anemic, with kidney disease and with liver disease, and the means and standard deviations of their hemoglobin, Hb A 1c, estimated average glucose (eag) and fasting glucose values. Mean Hb A 1c decreased in a progressive fashion from 6.8% (51 mmol/mol) for patients with MCH below the first decile (27.55 pg) to 6.0% (42 mmol/mol) for those with MCH above the ninth decile (32.55 pg). Mean fasting glucose remained fairly stable ( mg/dl) until the seventh decile, and then declined slightly to 131 mg/dl above the ninth. Mean eag, calculated from Hb A 1c fell progressively from mg/dl up until the seventh decile, and then fell further to 128 mg/dl above the ninth, making a total decrease of 20 mg/dl as against the 5 mg/dl fall of fasting glucose. Regardless of the order in which variables were entered into a stepwise multivariate regression, this analysis identified fasting glucose, MCH and age as predictors of Hb A 1c, discarding the variables sex, hemoglobin concentration, MCV, MCHC, erythrocyte count, creatinine, and liver disease (although MCV was identified as a predictor if MCH was excluded from the analysis, in which case the influence of MCV on Hb A 1c was similar to that of MCH when MCH was included); Hb A 1c increased with fasting glucose and age, and decreased with increasing MCH (Table 3). After adjustment for sex, age, serum creatinine, and liver disease, mean Hb A 1c in MCH interdecile intervals exhibited a significant negative gradient from 6.8% (51 mmol/mol) in the lowest to 6.0% (42 mmol/ mol) in the highest, and mean Hb A 1c in MCV interdecile intervals a significant negative gradient from 6.8% in the lowest to 6.1% (43 mmol/mol) in the highest, but there was no Hb A 1c gradient across hemoglobin interdecile intervals (Table 4). Nor was there any significant trend in mean fasting glucose across MCH, MCV, or hemoglobin interdecile intervals (results not shown). The lower panel of Table 2 shows that 42.2% of the patients in the bottom tenth of the distribution of MCH had Hb A 1c 6.5% (48 mmol/mol), i.e., in the diabetic range, as against only 26.1% of those in the top tenth, a fall of 16.1% (P 0.001). Contrariwise, the corresponding proportions of patients with Hb A 1c in the normoglycemic range [ 5.7% (39 mmol/mol)] were 31.5% and 51.8%, a rise of 20.3% (P 0.001). By contrast, the corresponding differences for fasting glucose were statistically nonsignificant for both the diabetic range ( 126 mg/dl; P 0.235) and the normoglycemic range ( 100 mg/dl; P 0.209). As regards glycemia control goals, the proportion of apparently well-controlled patients [Hb A 1c 7% (53 mmol/mol)] rose from 66.5% to 79.7% between the bottom and top tenths of the MCH distribution, whereas the proportion of apparently illcontrolled patients [Hb A 1c 8% (64 mmol/mol)] fell from 20.4% to 12.3% (P in both cases). As a consequence of the trends just noted, the MCH distributions of the groups of patients with discrepant Hb A 1c -based and fasting-glucose-based diabetes statuses were skewed in opposite directions depending on the type of discrepancy (Fig. 1). Thus 21.1% of patients who had diabetes according to Hb A 1c ( 6.5%) but not according to fasting glucose ( 126 mg/dl) had MCH 28 pg, as against only 9.2% with MCH 32 pg. In contrast only 10.8% of those with diabetes diagnosed by fasting glucose but not by Hb A 1c had MCH 28 pg, as against 19.7% with MCH 32 pg. Online Supplemental Fig. 1 shows that when the study group was crossclassified by the usual Hb A 1c -based and fasting-glucosebased criteria for normoglycemia, prediabetes, and diabetes, the proportion of patients with MCH 28 pg increased with Hb A 1C and fell with increasing fasting glucose, whereas the proportion of patients with MCH 32 pg showed the opposite trends. Table 5 shows odds ratios per picogram increase in MCH and per femtoliter increase in MCV, for Hb A 1c to equal or exceed the thresholds of 5.7%, 6.5%, 7%, and 8%. After adjustment for sex, age, fasting glucose, hemoglobin, creatinine and liver disease, the odds of exceeding these thresholds decreased by 10% 14% for each pico- Clinical Chemistry 62:12 (2016) 1573

5 Table 2. Characteristics of patients in each MCH interdecile interval. a < >9 P for trend All patients, n MCH range, pg Female, % <0.001 Hemoglobin, g/dl 11.7 (1.8) 12.8 (1.8) 13.2 (1.7) 13.4 (1.6) 13.5 (1.7) 13.6 (1.6) 13.7 (1.7) 13.8 (1.7) 13.9 (1.6) 13.9 (1.9) <0.001 Anemia, % <0.001 Liver disease, % <0.001 Kidney disease, % <0.001 Hb A 1c,% 6.79 (1.89) 6.74 (1.97) 6.65 (1.87) 6.57 (1.88) 6.57 (1.81) 6.45 (1.75) 6.52 (1.79) 6.40 (1.77) 6.33 (1.77) 6.09 (1.66) <0.001 Hb A 1c, mmol/mol 50.7 (18.3) 50.2 (19.2) 49.2 (18.1) 48.3 (18.2) 48.3 (17.4) 47.0 (16.8) 47.8 (17.2) 46.4 (17.0) 45.7 (17.0) 43.1 (15.8) <0.001 eag, b mg/dl 148 (54) 147 (57) 144 (54) 142 (54) 142 (52) 138 (50) 140 (51) 137 (51) 135 (51) 128 (48) <0.001 FPG, c mg/dl 135 (62) 136 (65) 135 (63) 136 (62) 136 (62) 134 (58) 135 (58) 133 (59) 133 (58) 131 (54) Patients satisfying criteria for, % Normoglycemia Hb A 1c <5.7% (39 mmol/mol) <0.001 FPG <100, mg/dl Prediabetes Hb A 1c % (39 46 mmol/mol) FPG mg/dl Diabetes Hb A 1c 6.5% (48 mmol/mol) <0.001 FPG 126 mg/dl Other Hb A 1c targets A 1c <7% (53 mmol/mol) <0.001 A 1c 8% (64 mmol/mol) <0.001 a Data are mean (SD) unless otherwise indicated. To convert glucose concentrations from mg/dl to mmol/l, multiply by b Calculated by the estimated average glucose formula [Nathan et al. (20)]. c FPG, fasting plasma glucose Clinical Chemistry 62:12 (2016)

6 MCH Affects Hb A 1c Status Table 3. Multivariate linear regression of Hb A 1c with hematological, demographic and metabolic indices. a Variable SE P value Adjusted R 2b Fasting glucose, mg/dl < % Mean cell hemoglobin, pg < % Age, years < % Constant < a Variables included and discarded in the regression analysis: hemoglobin, mean cell hemoglobin concentration, mean cell volume, erythrocytes, sex, serum creatinine, and liver disease. b Comulative variable effect. gram increase in MCH, and by 4% for each femtoliter increase in MCV. Discussion The utility of Hb A 1c measurements in clinical practice is limited by several nonglycemic influences on Hb A 1c, including several that affect erythrocyte turnover. The results of this study demonstrated that among patients for whom an Hb A 1c determination is requested for any reason, the hematological parameters with the greatest effect on Hb A 1c values independently of the glycemic environment, were MCH and MCV. From the lowest to the highest MCH interdecile interval, mean Hb A 1c fell from 6.8% (51 mmol/mol) to 6.0% (43 mmol/mol), causing the proportion of Hb A 1c defined normoglycemic patients (HbA 1c 5.7%) to rise from 31.5% to 51.8% and the proportion of Hb A 1c -defined diabetic patients to fall from 42.2% to 26.1%. Similar differences were obtained when examining the slightly larger extreme MCH groups defined by MCH 28 pg (14.1% of all patients) and MCH 32 pg (15.3% of all patients). Previous studies have also reported negative correlation between Hb A 1c and MCH, MCV, or MCHC among adults (12, 21), children (7), and premenopausal women (15). However, these other studies examined community populations from which persons with previously diagnosed diabetes were excluded. Alterations in erythrocyte turnover, and consequent alteration of Hb A 1c levels, can be caused by several common conditions, including iron deficiency, kidney failure, alcoholism, and chronic liver disease. In this study, a variation in Hb A 1c of % across MCH or MCV values was maintained after making allowance for sex, age, kidney disease, or liver disease. Similarly, a recent study in which 87% of the participants were normoglycemic (21) found an Hb A 1c difference of 0.5% between the extremes of MCHC (although in this study no adjustment for liver disease was made). In the present study, anemia was associated with an increase in units of Hb A 1c of about 0.15% among men, women and patients without kidney disease, and about 0.3% among patients with liver disease, but examination of the stepwise regression process leading to the results displayed in Table 3 shows that the influence of hemoglobin itself was overshadowed by that of MCH or MCV, as long as fasting glucose was included in the analysis (results not shown) and hemoglobin had no influence on Hb A 1c after adjustment for sex, age, creatinine and liver disease (Table 5). Simmons and Hlaing (21) similarly found that hemoglobin had no influence on Hb A 1c after adjustment for variables including sex, age and creatinine, and did not emerge as an independent predictor of Hb A 1c in analyses including fasting glucose and MCHC. In fact, the greatest deviation from the overall mean Hb A 1c of 6.51% in this study was not found among patients with low MCH ( 28 pg), whose mean Hb A 1c was 0.26% higher, but among patients with MCH 32 pg (15.3% of all patients), whose mean Hb A 1c of 6.16% was 0.35% lower than the overall mean despite 72% of the patients being male and 56% having liver disease, where both of the latter conditions tend to increase Hb A 1c in the whole study group (see online Supplemental Table 2). In the top tenth of MCH values, mean MCH was even lower (6.09%) and the proportions that were male and had liver disease were even higher, 74.3% and 60.6%, respectively. The fall in Hb A 1c with increasing MCH resulted in only 31.5% of the patients with the lowest tenth of MCH values having Hb A 1c 5.7% (cf. 51.8% of those with the highest tenth), whereas 42.2% had Hb A 1c 6.5% (cf. 26.1% of those with the highest tenth). With regard to glycemia management targets rather than diagnostic thresholds, an Hb A 1c value 8% was found in 20.4% of diabetic patients with MCH values below the first decile of all MCH values of diabetic patients, as against only 12.3% of those with MCH values among the highest tenth (results not shown). The fact that this latter difference, 8.1%, is only about half the difference of 16.1% between the proportions of patients with Hb A 1c 6.5% in the highest and lowest tenths of MCH values in the whole study group, together with the fact that Hb A 1c is Clinical Chemistry 62:12 (2016) 1575

7 Table 4. Mean Hb A 1c adjusted for sex, age, serum creatinine, and liver disease in each interdecile interval of MCH, MCV, and hemoglobin, with 95% CIs in parentheses. MCH Interdecile range, pg Hb A 1c,% 6.8 ( ) 6.8 ( ) 6.7 ( ) 6.6 ( ) 6.6 ( ) 6.5 ( ) 6.5 ( ) 6.4 ( ) 6.3 ( ) 6.0 ( ) Hb A 1c, mmol/mol 51 (50 52) 51 (50 52) 50 (49.51) 49 (48 50) 49 (48 50) 48 (46 48) 48 (46 49) 46 (45 48) 45 (44 46) 42 (42 43) MCV Interdecile range, fl Hb A 1c,% 6.8 ( ) 6.7 ( ) 6.7 ( ) 6.6 ( ) 6.6 ( ) 6.5 ( ) 6.4 ( ) 6.4 ( ) 6.3 ( ) 6.1 ( ) Hb A 1c, mmol/mol 51 (50 52) 50 (49 51) 50 (49 51) 49 (48 50) 49 (48 50) 48 (46 49) 46 (45 48) 46 (46 48) 45 (45 46) 43 (42 43) Hemoglobin Interdecile range, g/l Hb A 1c,% 6.5 ( ) 6.6 ( ) 6.5 ( ) 6.5 ( ) 6.5 ( ) 6.5 ( ) 6.5 ( ) 6.5 ( ) 6.5 ( ) 6.5 ( ) Hb A 1c,mmol/mol 48 (48 49) 49 (48 50) 48 (46 49) 48 (46 48) 48 (48 49) 48 (46 49) 48 (46 48) 48 (46 49) 48 (46 49) 48 (48 49) Fig. 1. MCH distributions between groups of patients with discrepant diabetes diagnosis criteria. FPG, fasting plasma glucose. a little more closely correlated with MCH among normoglycemic patients than among diabetic or prediabetic patients after adjusting for age, fasting glucose and creatinine, suggests that the influence of MCH on Hb A 1c decreases as glycemia increases. The question arises as to what extent the observed relationship between Hb A 1c and MCH is an effect of disease, for several pathologies that are not uncommon among hyperglycemic patients can cause low Hb A 1c. Hemolytic anemia lowers Hb A 1c levels because the resulting increased production of blood cells shortens the average erythrocyte life span and hence the exposure of hemoglobin to ambient glycemia (6, 22). In anemia of chronic disease, a number of factors can contribute to the lowering of Hb A 1c levels, including impaired proliferation and differentiation of erythroid precursors, blunted erythropoietin response, and altered erythrocyte life span (23). In patients with chronic liver disease, anemia, portal hypertension, and variceal bleeding are common complications that can alter erythrocyte survival and lead to low Hb A 1c levels (24). And in patients on hemodialysis, Hb A 1c levels can be lowered due to altered erythrocyte survival, erythropoietin therapy, and blood transfusion (25). However, in this study the exclusion of participants with anemia, kidney disease, or liver disease lowered average Hb A 1c levels (by about 0.1%) rather than raising them, and only lowered the difference between the groups with MCH 28 pg and 32 pg by 0.06%, to 0.55% (see online Supplemental Table 3). Among the 52.1% of patients with no kidney or liver disease (a group with an anemia rate of 16.8%), Hb A 1c ranged from an average 6.6% among those below the lowest MCH decile to 6.1% among those above the highest, while fasting 1576 Clinical Chemistry 62:12 (2016)

8 MCH Affects Hb A 1c Status Table 5. Odds ratios (per pg increase in MCH or fl MCV) for exceeding various Hb A 1c cutpoints. 5.7% (39 mmol/ mol) (95% CI) (n = ) 6.5% (48 mmol/ mol) (95% CI) (n = 7671) 7% (53 mmol/ mol) (95% CI) (n = 6070) 8% (64 mmol/ mol) (95% CI) (n = 3724) MCH Unadjusted 0.90 ( ) 0.93 ( ) 0.93 ( ) 0.93 ( ) Multivariable a 0.86 ( ) 0.90 ( ) 0.90 ( ) 0.90 ( ) MCV Unadjusted 0.97 ( ) 0.97 ( ) 0.97 ( ) 0.97 ( ) Multivariable a 0.96 ( ) 0.96 ( ) 0.96 ( ) 0.96 ( ) a Adjusted by age, sex, fasting glucose, hemoglobin, creatinine, and liver disease (defined as AST 30 U/L or ALT 31 U/L or GGT 38 U/L or bilirubin >1.20 mg/dl or albumin <3.5 g/dl). glucose rose only from mg/dl; among the 47.9% of patients with kidney or liver disease (36.3% of whom were anemic), Hb A 1c ranged from 6.1% to 7.0% and fasting glucose from mg/dl as MCH fell from the highest to the lowest decile. Both MCH and MCV decrease linearly during the life span of the red blood cell; loss of hemoglobin by the erythrocyte is attributable to its shedding hemoglobincontaining vesicles, and loss of volume to both vesicle shedding and loss of water (26, 27). Thus large differences in MCH and MCV can be caused by differences in mean erythrocyte age, and this, rather than differences in the rate of hemoglobin or volume loss appears to be the usual cause of such differences in MCH or MCV (26). Accordingly, the Hb A 1c difference of % observed in this study between the extremes of the distributions of MCH and MCV seems likely to reflect a difference in mean erythrocyte age rather than a difference in glycemia (which is in any case belied by the fasting glucose data). In fact, it seems likely that several of the reported associations between Hb A 1c and hematological parameters (7, 10, 16) may result from differences in MCH that in turn reflect differences in erythrocyte survival; for hemoglobin, MCHC, MCV, and erythrocyte count were among the variables adjusted for in the stepwise analysis that returned only fasting glucose, age, and MCH as determinants of Hb A 1c. Our results therefore suggest that MCH, as an indirect measure of erythrocyte survival (9), can explain at least a large part of the reported discrepancies between the results of using Hb A 1c and fasting glucose criteria for diagnosis of diabetes and prediabetes (18, 28 31). This study, of course, had its weaknesses and limitations. Firstly, as regards the influence and variation of glycemia, our only direct glycemia measure was fasting glucose, which does not reflect mean glycemia. Thus our normoglycemic patients satisfied both the Hb A 1c and fasting glucose criteria of normoglycemia, but did not necessarily satisfy the glucose tolerance criterion, which may have affected observed differences among apparently normoglycemic, prediabetic and diabetic patients. Secondly, the study was limited in that we had no direct measurement of patients erythrocyte life span. On the other hand, the robustness of our conclusions as to the influence of MCH is supported by our large sample size, the uniformity of the laboratory measurements (all performed in a single laboratory using a single technique), and the internal consistency of our results. We conclude that in the population studied (patients for whom Hb A 1c determination is requested) the influence of MCH on HbA 1c :(a) largely explains the discrepancies between the results of using Hb A 1c -based or glucose-based criteria for diagnosis of diabetes and prediabetes that have been observed in this study (see Table 2) and in previous studies; and (b) results in considerable differences between diabetic patients with low and high MCH as regards the proportion achieving Hb A 1c -defined glycemia management goals. Accordingly, given the interrelationships among erythrocyte indices and the high prevalence of erythrocyte index alterations in this population, the proper interpretation of Hb A 1c values, whether for diagnostic or management purposes, would seem to require consideration of MCH (or in its absence, MCV), and possibly of full blood count data. This does not mean that MCH values could be used to correct Hb A 1c values their mutual correlation is too weak for this but that their influence should nevertheless be taken into account, for example by using different Hb A 1c thresholds for different MCH ranges. If subsequent studies confirm the present findings, it may be worthwhile investigating whether the use of such MCH-dependent Hb A 1c thresholds improves patient management. Clinical Chemistry 62:12 (2016) 1577

9 Author Contributions: All authors confirmed they have contributed to the intellectual content of this paper and have met the following 3 requirements: (a) significant contributions to the conception and design, acquisition of data, or analysis and interpretation of data; (b) drafting or revising the article for intellectual content; and (c) final approval of the published article. Authors Disclosures or Potential Conflicts of Interest: Upon manuscript submission, all authors completed the author disclosure form. Disclosures and/or potential conflicts of interest: References Employment or Leadership: F.F. Casanueva, SERGAS. Consultant or Advisory Role: None declared. Stock Ownership: None declared. Honoraria: None declared. Research Funding: None declared Expert Testimony: None declared. Patents: None declared. Role of Sponsor: The funding organizations played no role in the design of study, choice of enrolled patients, review and interpretation of data, and final approval of manuscript. 1. Sacks DB, Arnold M, Bakris GL, Bruns DE, Horvath AR, Kirkman MS, et al. Position statement executive summary: Guidelinesandrecommendationsforlaboratory analysis in the diagnosis and management of diabetes mellitus. Diabetes Care 2011;34: American Diabetes Association. Standards of medical care in diabetes Diabetes Care 2014;37 Suppl 1:S14 S Selvin E, Steffes MW, Zhu H, Matsushita K, Wagenknecht L, Pankow J, et al. Glycated hemoglobin, diabetes, and cardiovascular risk in non-diabetic adults. N Engl J Med 2010;362: Khaw KT, Wareham N, Bingham S, Luben R, Welch A, Day N. Association of hemoglobin A1c with cardiovascular disease and mortality in adults: the European Prospective Investigation into Cancer in Norfolk. Ann Intern Med 2004;141: Cohen RM, Herman WH. Are glycated serum proteins ready for prime time? [Comment]. Lancet Diabetes Endocrinol 2014;2: Gallagher EJ, Bloomgarder ZT, Le Roith D. Review of hemoglobin A1c in the management of diabetes. J Diabetes 2009;1: Hardikar PS, Joshi SM, Bhat DS, Raut DA, Katre PA, Lubree HG, et al. Spuriously high prevalence of prediabetes diagnosed by HbA1c in young Indians partly explained by hematological factors and iron deficiency anemia. Diabetes Care 2012;35: Ly J, Marticorena R, Donnelly S. Red blood cell survival in chronic renal failure. Am J Kidney Dis 2004;44: Cohen RM, Franco RS, Khera PK, Smith EP, Lindsell CJ, Ciraolo PJ, et al. Red cell life span heterogeneity in hematologically normal people is sufficient to alter HbA 1c. Blood 2008;112: Coban E, Ozdogan M, Timuragaoglu A. Effect of iron deficiency anemia on the levels of hemoglobin A1c in nondiabetic patients. Acta Haematol 2004;112: Kim C, Bullard KM, Herman WH, Beckles GL. Association between iron deficiency and A1C levels among adults without diabetes in the National Health and Nutrition Examination Survey, Diabetes Care 2010;33: Ford ES, Cowie CC, Li C, Handelsman Y, Bloomgarden ZT. Iron-deficiency anemia, non-iron-deficiency anemia and HbA 1c among adults in the US. J Diabetes 2011;3: Son JI, Rhee SY, Woo JT, Hwang JK, Chin SO, Chon S, et al. Hemoglobin A1c may be an inadequate diagnostic tool for diabetes mellitus in anemic subjects. Diabetes Metab J 2013;37: Shanthi B, Revathy C, Manjula Devi AJ, Subhashree. Effect of iron deficiency on glycation of haemoglobin in nondiabetics. J Clin Diagn Res 2013;7: Koga M, Morita S, Saito H, Mukai M, Kasayama S. Association of erythrocyte indices with glycated haemoglobin in pre-menopausal women. Diabet Med 2007;24: English E, Idris I, Smith G, Dhatariya K, Kilpatrick ES, Garry John W. The effect of anaemia and abnormalities of erythrocyte indices on analysis: a systematic review. Diabetologia 2015;58: Hoelzel W, Weykamp C, Jeppsson JO, Miedema K, Barr JR, Goodall I, et al. 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: Rodriguez-Segade S, Rodriguez J, Garcia JM, Casanueva FF, Coleman IC, Alonso de la Peña C, Camiña F. Influence of the glycation gap on the diagnosis of type 2 diabetes. Acta Diabetol 2015;52: egfr Calculator [from UK CKD eguide]. renal.org (Accessed October 2015). 20. Nathan DM, Kuenen J, Borg R, Zheng H, Schoenfeld D, Heine RJ. Translating the A1C assay into estimated average glucose values. Diabetes Care 2008;31: Simmons D, Hlaing T. Interpretation of Hb A 1c : association with mean cell volume and hemoglobin concentration. Diabet Med 2014;31: Panzer S, Kronik G, Lechner K, Bettelheim P, Neumann E, Dudczak R. Glycosylated hemoglobins (Ghb): an index of red cell survival. Blood 1982;59: Weiss G, Goodnough LT. Anemia of chronic disease. N Engl J Med 2005;352(10): Blendea MC, Thompson MJ, Malkani S. Diabetes and chronic liver disease: etiology and pitfalls in monitoring. Clin Diabetes 2010;28: Inaba M, Okuno S, Kumeda Y, Yamada S, Imanishi Y, Tabata T, et al. Glycated albumin is a better glycemic indicator than glycated hemoglobin values in hemodialysis patients with diabetes: effect of anemia and erythropoietin injection. J Am Soc Nephrol 2007;18: Werre J, Willekens F, Bosch F, De Haab L, Van der Vegt S, Van den Bos A, Bosman J. The red cell revisited. Matters of life and death. Cell Mol Biol 2004;50: Gifford SC, Derganc J, Shevkoplyass SS, Yoshida T, Bitensky MW. A detailed study of time-dependent changes in human red blood cells: from reticulocyte maturation to erythrocyte senescence. Br J Haematol 2006;135: Lipska KJ, De Rekeneire N, Van Ness PH, Johnson KC, Kanaya A, Koster A, et al. Identifying dysglycemic states in older adults: implications of the emerging use of hemoglobin A1c. J Clin Endocrinol Metab 2010;95: Cowie CC, Rust KF, Byrd-Holt DD, Gregg EW, Ford ES, Geiss LS, et al. Prevalence of diabetes and high risk for diabetes using A1c criteria in the US population in Diabetes Care 2010;33: Guo F, Moellering DR, Garvey WT. Use of HbA1c for diagnoses of diabetes and prediabetes: comparison with diagnoses based on fasting and 2-hr glucose values and effects of gender, race, and age. Metab Syndr Relat Disord 2014;12: Wang W, Lee ET, Howard BV, Fabsitz RR, Devereux RB, Welty TK. Fasting plasma glucose and hemoglobin A1c in identifying and predicting diabetes: The Strong Heart Study. Diabetes Care 2011;34: Clinical Chemistry 62:12 (2016)