OBJECTIVE

To evaluate the relationship between skin advanced glycation end products (sAGEs) with mean blood glucose (MBG), hemoglobin A1c (HbA1c), and MBG-independent, between-patient differences in HbA1c among children with type 1 diabetes.

RESEARCH DESIGN AND METHODS

Children aged 5 to 20 years with type 1 diabetes of at least 1 year duration participated. At a clinic visit, sAGE was estimated noninvasively by measurement of skin intrinsic fluorescence (SIF). SIF data were adjusted to correct for variation in skin pigmentation. MBG-independent, between-patient differences in HbA1c were examined by statistically controlling HbA1c for MBG or alternatively by use of a hemoglobin glycation index (HGI). Results were similar whether HbA1c, MBG, and HGI were analyzed as single values from the time of the SIF examination visit or as the mean values from all available visits of the patient.

RESULTS

HbA1c was correlated with MBG (r = 0.5; P < 0.001; n = 110). HbA1c and HGI, but not MBG, were statistically associated with SIF after adjustment for age, duration of diabetes, race, sex, and BMI z-score. SIF increased with age and duration of diabetes and was higher in girls than boys.

CONCLUSIONS

sAGE levels estimated by SIF increase with age, duration of diabetes, and female sex. sAGE is correlated with MBG-independent biological variation in HbA1c, but not with MBG itself. These results suggest that factors besides MBG that influence HbA1c levels also contribute to accumulation of sAGE.

Chronically elevated blood glucose levels are associated with the development of serious microvascular complications in patients with type 1 diabetes (1). One mechanism for development of complications is through enhanced nonenzymatic glycation of important proteins facilitated by hyperglycemia (2,3). As conventionally understood, the formation of glycated proteins is dependent on glucose concentration. For example, over short periods of time glucose can reversibly attach to hemoglobin and other proteins to form a Schiff base. With more time, the Schiff base can chemically rearrange in a nearly irreversible process to form an Amadori product, glycated hemoglobin (HbA1c) (4). Clinically, HbA1c is highly correlated with mean blood glucose (MBG) levels from the preceding weeks and months (5), and both have been shown to be predictors for the development and progression of microvascular diabetes complications (1,6).

Other proteins besides hemoglobin are also susceptible to nonenzymatic glycation. Structural proteins have a longer life span than the 120 days typical of hemoglobin. Over weeks to months, the glycated moieties of long-lived proteins can form covalent complexes known as advanced glycation end products (AGEs). Formation of AGEs in tissues causes alteration in the normal structure and function of the involved proteins, which can elicit pathological changes (2,3). Tissue burden of AGEs has been assessed by direct chemical measurement of these substances from biopsy samples. The Epidemiology of Diabetes Interventions and Complications (EDIC) study found that AGE levels from biopsied skin were predictive for the progression of microvascular complications in patients with type 1 diabetes (7,8). Furthermore, these investigators found that HbA1c levels were correlated with skin AGE burden (8). Thus concentrations of an early-stage, shorter-lived glycation product, HbA1c, are associated with the amount of AGEs accumulated in tissue over time. Thus both HbA1c and AGEs appear to be predictors for development and progression of microvascular diabetes complications (8).

The high correlation between HbA1c and MBG has led to the more easily measurable HbA1c becoming a widely used method to estimate glycemic control and guide therapy in patients with diabetes. In addition to MBG, HbA1c levels have been shown to be influenced by MBG-independent factors, which are manifested by consistent between-individual differences in HbA1c level regardless of the preceding background MBG (9,10). Previously, we developed a hemoglobin glycation index (HGI) to quantify such between-patient differences from readily available clinical data (911). We found that MBG-independent between-patient HbA1c differences are predictive for development of diabetes complications (10). As HbA1c and MBG are highly correlated, we hypothesized that tissue AGEs would be associated with both MBG from patient self-monitored capillary blood samples as well as corresponding HbA1c levels. Furthermore, we hypothesized that MBG-independent differences in HbA1c would also be associated with tissue accumulation of AGEs.

Recently, methods have been developed to noninvasively estimate skin AGEs in vivo by measurement of skin intrinsic fluorescence (SIF) (1215). SIF is highly correlated with AGEs assayed directly from skin biopsy (15). The technology can rapidly and noninvasively estimate skin AGEs from large numbers of individuals and thus is particularly suitable for use in children. We used the SIF technology to test our hypotheses in a well-characterized group of children with type 1 diabetes.

Patient recruitment

Children with type 1 diabetes of at least 1 year duration, attending diabetes clinics at Children’s Hospital New Orleans, were invited to participate. Informed consent was obtained from parents and consent/assent from the patient. Data regarding duration of diabetes, height, weight, date of birth, race, and sex were collected at the time of visit, and additional data regarding prior HbA1c and MBG levels were extracted from patient medical records. The study was approved by the institutional review board at Louisiana State University Health Sciences Center and the Children’s Hospital New Orleans.

Assessment of glycemic control

At the time of clinic visit, the patient’s glucose meter was inspected for proper operation. If no problems were found, the MBG from the 30 days prior to the visit was calculated and recorded. A blood sample at the time of the clinic visit was obtained for HbA1c and glucose (blood glucose at time of clinic visit [cBG], measured by Accu-Chek Inform, Model 2001201, Roche Diagnostics, Indianapolis, IN). HbA1c was measured using a National Glycohemoglobin Standardization Program–certified assay. The HGI quantifies between-individual biological variation in HbA1c not due to MBG. HGI at the time of the clinic visit was calculated as the difference between a patient’s observed HbA1c minus the HbA1c level predicted from the patient’s observed MBG based on the population regression of HbA1c on MBG (predicted HbA1c = 0.021 × MBG + 4.3) (16). We have previously shown the intra- and interindividual consistency over time of HbA1c, MBG, and HGI collected from clinic data (16).

Measurement of SIF

During one clinic visit, SIF levels were noninvasively measured from the volar surface of the left forearm from each subject using a SCOUT DS instrument (VeraLight, Inc.). The device sequentially excited the skin surface using different light-emitting diodes (LEDs) that had peak excitation wavelengths of 375, 405, and 420 nm. The LEDs excite varying degrees of fluorescence from different fluorophores such as AGEs. Fluorescence generated from each LED was detected over a 435–600-nm emission window. Skin reflectance was measured for each excitation LED, and a white light LED was used to measure skin reflectance over the emission region.

Spectral data collected by the instrument were transmitted directly to VeraLight, Inc. for further analysis to adjust for the impact of skin pigmentation, hemoglobin content, light scattering, and other dermal characteristics. Data were adjusted mathematically using two different kx/km parameter sets on the measured reflectance at the excitation and emission wavelengths. For the first kx/km set (designated as set u), kx and km were 1.0 and 0.0, respectively. For the second kx/km set (set c), kx and km were 0.5 and 0.5, respectively, which is useful for populations with a wide range of skin melanin (17) (further technical detail in Supplementary Data). SIF data are reported in arbitrary relative fluorescence units as a function of excitation wavelength and which kx/km adjustment (“u” or “c”) used. VeraLight, Inc. investigators were blinded to the clinical and laboratory status of the patients.

Statistical analysis

Data were analyzed using HbA1c, MBG, HGI, and cBG at the time of the visit when SIF was measured. In addition, the means of MBG, HbA1c, and HGI from all available prior clinic visits for the patient were also evaluated. As results were similar whether using data for that clinic visit or the mean of data from all available clinic visits, we report here the mean data. Simple correlation analyses between variables were performed by Pearson method. The influence of biological variation in HbA1c on SIF was evaluated in multivariate regression models using either HGI or by substituting HbA1c adjusted for MBG in the model. For evaluation purposes, cBG was substituted in the models for MBG for single clinic visits. Models were further adjusted for chronological age, duration of diabetes, sex, race, and BMI z-score (z-BMI) using the GLM procedure in SAS software (SAS Institute, Cary, NC). Statistical significance was considered to be at P < 0.05.

Characteristics of the patient population studied at the time of SIF measurement in clinic are presented in Table 1. In the patient population, mean HbA1c (mHbA1c) was correlated with mean MBG (mMBG) (r = 0.5; P < 0.001) but not the visit cBG (r = 0.07; P = 0.51). mMBG and cBG were not correlated with each other, or with mHGI. mHbA1c was correlated with mean HGI (mHGI) (r = 0.8; P < 0.001). The SIF data for the three excitation wavelengths were highly intercorrelated within the same adjustment set whether “u” or “c”. However, correlations between “u” and “c” data were considerably lower (Supplementary Table 3).

Table 1

Patient characteristics at time of clinic visit (n = 110)

Age (years) 13.2 ± 3.8 
Race  
 White 82 (71%) 
 Black 33 (29%) 
Sex (M/F) 60/55 
Duration of diabetes (years) 5.9 ± 3.6 
BMI (kg/m221.6 ± 4.5 
z-BMI 0.6 ± 0.9 
HbA1c (%) 9.5 ± 2.0 
MBG (mg/dL) 204.6 ± 39.6 
cBG (mg/dL) 239.9 ± 93.3 
Age (years) 13.2 ± 3.8 
Race  
 White 82 (71%) 
 Black 33 (29%) 
Sex (M/F) 60/55 
Duration of diabetes (years) 5.9 ± 3.6 
BMI (kg/m221.6 ± 4.5 
z-BMI 0.6 ± 0.9 
HbA1c (%) 9.5 ± 2.0 
MBG (mg/dL) 204.6 ± 39.6 
cBG (mg/dL) 239.9 ± 93.3 

Results from the single visit data were similar to the results using mean values from multiple visits. Data are reported as the mean from multiple visits. Simple correlation analyses were performed between “u” and “c” adjustments of SIF with mMBG, mHbA1c, mHGI, age, and duration of diabetes (Table 2). The average number of prior patient visits was 8.2 with a range of 1–21. mHGI and mHbA1c were correlated with all SIF data whether “u” or “c” was adjusted. mMBG, cBG, and z-BMI were not correlated with any of the SIF data. Age and duration of diabetes were consistently correlated with all “c”-adjusted SIF data.

Table 2

Pearson correlation between “u” and “c” set SIF adjustments with mMBG, mHbA1c, mHGI, age, and duration of diabetes

mMBGmHbA1cmHGIAgeDuration of diabetes
SIF375u 0.14 0.39 0.39 0.02 0.01 
 0.14 <0.0001 <0.0001 0.82 0.89 
SIF405u 0.12 0.42 0.44 0.22 0.14 
 0.22 <0.0001 <0.0001 0.02 0.14 
SIF420u 0.10 0.41 0.43 0.28 0.18 
 0.30 <0.0001 <0.0001 0.002 0.05 
SIF375c 0.15 0.38 0.34 0.45 0.39 
 0.12 <0.0001 0.0003 <0.0001 <0.0001 
SIF405c 0.05 0.26 0.26 0.53 0.39 
 0.60 0.005 0.007 <0.0001 <0.0001 
SIF420c 0.04 0.22 0.21 0.52 0.38 
 0.69 0.02 0.02 <0.0001 <0.0001 
mMBGmHbA1cmHGIAgeDuration of diabetes
SIF375u 0.14 0.39 0.39 0.02 0.01 
 0.14 <0.0001 <0.0001 0.82 0.89 
SIF405u 0.12 0.42 0.44 0.22 0.14 
 0.22 <0.0001 <0.0001 0.02 0.14 
SIF420u 0.10 0.41 0.43 0.28 0.18 
 0.30 <0.0001 <0.0001 0.002 0.05 
SIF375c 0.15 0.38 0.34 0.45 0.39 
 0.12 <0.0001 0.0003 <0.0001 <0.0001 
SIF405c 0.05 0.26 0.26 0.53 0.39 
 0.60 0.005 0.007 <0.0001 <0.0001 
SIF420c 0.04 0.22 0.21 0.52 0.38 
 0.69 0.02 0.02 <0.0001 <0.0001 

The top number in each cell is the correlation coefficient (r), and the bottom number is the significance level (P value); n = 110.

Multivariate regression analysis showed that mHbA1c adjusted for mMBG, or mHGI substituted in the model for mHbA1c, was consistently associated with all SIF excitation wavelengths, both “u” and “c”, after also controlling for presence of age, race, sex, z-BMI, and duration of diabetes (Table 3). mMBG and z-BMI were not statistically significant in relationship with any of the SIF measures. Figure 1 depicts the relationship between 405-nm excited SIF using set “c” adjustment versus mMBG, mHGI, or mHbA1c.

Table 3

Multivariate regression analysis between “u”- and “c”-adjusted SIF levels at each wavelength as dependent variables with patient age, duration of diabetes (DOD), sex, race, z-BMI, HbA1c or HGI, and MBG

SIF wavelength and adjustmentAge (years)DOD (years)Sex F > MRace B > Wz-BMIHbA1c or HGI (%)MBG (mg/dL)
375u      
375c   
405u     
405c    
420u     
420c    
SIF wavelength and adjustmentAge (years)DOD (years)Sex F > MRace B > Wz-BMIHbA1c or HGI (%)MBG (mg/dL)
375u      
375c   
405u     
405c    
420u     
420c    

Statistically significant relationships, P = 0.05 or less, for the covariates in the model are indicated by X.

Figure 1

The relationship between 405-nm excited SIF using set “c” adjustment with mMBG, mHGI, and mHbA1c.

Figure 1

The relationship between 405-nm excited SIF using set “c” adjustment with mMBG, mHGI, and mHbA1c.

Close modal

Age, duration of diabetes, and sex were statistically significant covariates for the “c” SIF data at all three wavelengths. SIF levels increased with age and were higher in girls than boys at any given age (Fig. 2). Race was a significant covariate for all SIF excitation wavelengths with the “u” adjustment, but only for 375 nm in the “c”-adjusted SIF data, blacks being higher than whites.

Figure 2

The relationship of age and sex in SIF405c data. SIF levels increased with age and were higher in girls than boys.

Figure 2

The relationship of age and sex in SIF405c data. SIF levels increased with age and were higher in girls than boys.

Close modal

To our knowledge this is the first in-depth study of the relationship of skin AGEs (sAGEs) with HbA1c, MBG, HGI, and other characteristics in a biracial population of children with type 1 diabetes. We used a novel technology to estimate sAGEs by measuring SIF using a SCOUT DS device at three different excitation wavelengths and two mathematical adjustment sets. Prior reports using this device have focused on its ability to identify adults with abnormalities of glucose tolerance (13,14) and the relationship of SIF with coronary artery calcifications in type 1 diabetes (12).

Spectral information read by the SCOUT device from the skin is a combination of reflected light and fluorescence excited in the tissue by LEDs. Because of variability in skin pigmentation, skin thickness, hemoglobin, and other factors, the excitation and emission data returning to the device must be adjusted to minimize the distortion of the skin fluorescence by the aforementioned factors. How adjustment can affect interpretation of data can be seen by comparing “u” and “c” adjustments of the SIF data (Table 3). With the “u” adjustment, all data from the three excitation SIF wavelengths evaluated were consistently higher in black than in white patients, consistent with differences in skin pigmentations. Such a difference was only found at the 375-nm SIF excitation wavelength with the “c” adjustment. Thus the “c” adjustment eliminated a considerable amount of artifact due to skin pigmentation. sAGEs and auto-fluorescence have been shown to increase with aging and duration of diabetes in dermal biopsy samples from patients with diabetes (15,18,19). However, the “u” adjustment from our SIF data was insensitive to differences related to duration of diabetes and less sensitive to age-related change. Our finding of a significant increase in “c”-adjusted SIF at all excitation wavelengths with patient age and duration of diabetes in children is in accordance with prior biopsy data, albeit from adults.

We found a difference in the “c”-adjusted SIF levels between girls and boys that was independent of age, MBG, or HbA1c level. There was no sex difference discernable using the “u” adjustment, which agrees with reported findings in adults using a similar method (20), with the exception of smokers (21). Sex differences are known to occur in skin (22). Potentially the detected sex difference in SIF might represent a biochemical difference in accumulation of fluorescent AGEs between the sexes during childhood and adolescence. Alternatively this difference may represent non-AGE differences and be an artifact of the adjustment method or other factors.

It is conventionally understood that formation and accumulation of glycated proteins such as HbA1c (4) and AGEs (2) is a concentration-dependent function of glucose concentration. The SIF data demonstrated a strong relationship with HbA1c or HGI whether “u” or “c” adjusted. Interestingly there was no influence of MBG or cBG on SIF. Lack of relationship of SIF with randomly obtained cBG is not surprising. Previous investigators have studied skin autofluorescence during oral glucose tolerance testing and noted no changes related to acute fluctuations in blood glucose (23). This makes sense from the conventional understanding of how proteins are glycated. Initial nonenzymatic attachment of glucose to a protein such as hemoglobin occurs quickly and reversibly, but the formation of the stable Amadori product takes much longer, and progression to AGEs even more time. Thus a randomly obtained glucose level, while correlated with the labile Schiff base precursor to HbA1c, is negligibly associated with stable HbA1c levels (24).

The lack of association of SIF with MBG is more surprising. Both MBG and HbA1c levels have been associated with the development of complications in type 1 diabetes (6). HbA1c is strongly correlated with MBG and widely used to estimate MBG (25). However in simple Pearson correlation analysis, HbA1c and HGI were correlated with the SIF levels whereas MBG was not. In the multivariate models, both HbA1c and HGI were significantly related to SIF at that visit, but the visit MBG as a covariate was not. Considering the possibility that the 30-day average MBG from the clinic visit might have been insufficiently long to be involved with detectable AGE formation, we substituted the average of all available prior MBG measurements from patients into the model (along with HGI or HbA1c). There was still no significant association between MBG and SIF. These findings suggest that MBG-independent, between-patient differences in HbA1c are predictive of AGE burden in the skin and potentially influence the development of diabetes complications. Thus factors besides MBG that can influence nonenzymatic glycation (24) may be more influential in formation of skin AGEs and development of diabetes complications than just MBG exposure alone.

N.M. and J.M. are employees of VeraLight, Inc. No other potential conflicts of interest relevant to this article were reported.

D.L.F. enrolled patients, collected data, maintained the study database, and cowrote the manuscript. J.M.H. helped analyze data, prepared figures, reviewed and edited the manuscript, and contributed to discussion. S.L. performed major statistical analysis of the dataset and reviewed and edited the manuscript. N.M. and J.M. analyzed SIF data, reviewed and edited the manuscript, and contributed to discussion. C.L. enrolled patients and collected data. S.A.C. performed overall supervision of the study, helped analyze data, and cowrote the manuscript.

The authors thank Kelly Allerton, Louisiana State University Health Sciences Center, for assistance in editorial preparation of the manuscript.

1.
The Diabetes Control and Complications Trial Research Group
.
The effect of intensive treatment of diabetes on the development and progression of long-term complications in insulin-dependent diabetes mellitus
.
N Engl J Med
1993
;
329
:
977
983
[PubMed]
2.
Ahmed
N
,
Thornalley
PJ
.
Advanced glycation endproducts: what is their relevance to diabetic complications?
Diabetes Obes Metab
2007
;
9
:
233
245
[PubMed]
3.
Monnier
VM
,
Sell
DR
,
Dai
Z
,
Nemet
I
,
Collard
F
,
Zhang
J
.
The role of the amadori product in the complications of diabetes
.
Ann N Y Acad Sci
2008
;
1126
:
81
88
[PubMed]
4.
Bunn
HF
.
Evaluation of glycosylated hemoglobin diabetic patients
.
Diabetes
1981
;
30
:
613
617
[PubMed]
5.
The DCCT Research Group
.
The Diabetes Control and Complications Trial (DCCT). Design and methodologic considerations for the feasibility phase
.
Diabetes
1986
;
35
:
530
545
[PubMed]
6.
Service
FJ
,
O’Brien
PC
.
The relation of glycaemia to the risk of development and progression of retinopathy in the Diabetic Control and Complications Trial
.
Diabetologia
2001
;
44
:
1215
1220
[PubMed]
7.
Monnier
VM
,
Bautista
O
,
Kenny
D
, et al
;
DCCT Skin Collagen Ancillary Study Group
.
Skin collagen glycation, glycoxidation, and crosslinking are lower in subjects with long-term intensive versus conventional therapy of type 1 diabetes: relevance of glycated collagen products versus HbA1c as markers of diabetic complications. Diabetes Control and Complications Trial
.
Diabetes
1999
;
48
:
870
880
[PubMed]
8.
Genuth
S
,
Sun
W
,
Cleary
P
, et al
;
DCCT Skin Collagen Ancillary Study Group
.
Glycation and carboxymethyllysine levels in skin collagen predict the risk of future 10-year progression of diabetic retinopathy and nephropathy in the diabetes control and complications trial and epidemiology of diabetes interventions and complications participants with type 1 diabetes
.
Diabetes
2005
;
54
:
3103
3111
[PubMed]
9.
Hempe
JM
,
Gomez
R
,
McCarter
RJ
 Jr
,
Chalew
SA
.
High and low hemoglobin glycation phenotypes in type 1 diabetes: a challenge for interpretation of glycemic control
.
J Diabetes Complications
2002
;
16
:
313
320
[PubMed]
10.
McCarter
RJ
,
Hempe
JM
,
Gomez
R
,
Chalew
SA
.
Biological variation in HbA1c predicts risk of retinopathy and nephropathy in type 1 diabetes
.
Diabetes Care
2004
;
27
:
1259
1264
[PubMed]
11.
McCarter
RJ
,
Hempe
JM
,
Chalew
SA
.
Mean blood glucose and biological variation have greater influence on HbA1c levels than glucose instability: an analysis of data from the Diabetes Control and Complications Trial
.
Diabetes Care
2006
;
29
:
352
355
[PubMed]
12.
Conway
B
,
Edmundowicz
D
,
Matter
N
,
Maynard
J
,
Orchard
T
.
Skin fluorescence correlates strongly with coronary artery calcification severity in type 1 diabetes
.
Diabetes Technol Ther
2010
;
12
:
339
345
[PubMed]
13.
Ediger
MN
,
Olson
BP
,
Maynard
JD
.
Noninvasive optical screening for diabetes
.
J Diabetes Sci Tech
2009
;
3
:
776
780
[PubMed]
14.
Maynard
JD
,
Rohrscheib
M
,
Way
JF
,
Nguyen
CM
,
Ediger
MN
.
Noninvasive type 2 diabetes screening: superior sensitivity to fasting plasma glucose and A1C
.
Diabetes Care
2007
;
30
:
1120
1124
[PubMed]
15.
Meerwaldt
R
,
Graaff
R
,
Oomen
PH
, et al
.
Simple non-invasive assessment of advanced glycation endproduct accumulation
.
Diabetologia
2004
;
47
:
1324
1330
[PubMed]
16.
Soros
AA
,
Chalew
SA
,
McCarter
RJ
,
Shepard
R
,
Hempe
JM
.
Hemoglobin glycation index: a robust measure of hemoglobin A1c bias in pediatric type 1 diabetes patients
.
Pediatr Diabetes
2010
;
11
:
455
461
[PubMed]
17.
Hull EL, Ediger MN, Brown CD, Maynard JD, Johnson RD. Determination of a measure of a glycation end-product or disease state using tissue fluorescence. Patent #7,139,598. 21 November 2006
18.
Monnier
VM
,
Vishwanath
V
,
Frank
KE
,
Elmets
CA
,
Dauchot
P
,
Kohn
RR
.
Relation between complications of type I diabetes mellitus and collagen-linked fluorescence
.
N Engl J Med
1986
;
314
:
403
408
[PubMed]
19.
Dominiczak
MH
,
Bell
J
,
Cox
NH
, et al
.
Increased collagen-linked fluorescence in skin of young patients with type I diabetes mellitus
.
Diabetes Care
1990
;
13
:
468
472
[PubMed]
20.
Meerwaldt
R
,
Lutgers
HL
,
Links
TP
, et al
.
Skin autofluorescence is a strong predictor of cardiac mortality in diabetes
.
Diabetes Care
2007
;
30
:
107
112
[PubMed]
21.
Koetsier
M
,
Lutgers
HL
,
de Jonge
C
,
Links
TP
,
Smit
AJ
,
Graaff
R
.
Reference values of skin autofluorescence
.
Diabetes Technol Ther
2010
;
12
:
399
403
[PubMed]
22.
Dao
H
 Jr
,
Kazin
RA
.
Gender differences in skin: a review of the literature
.
Gend Med
2007
;
4
:
308
328
[PubMed]
23.
Noordzij
MJ
,
Lefrandt
JD
,
Graaff
R
,
Smit
AJ
.
Skin autofluorescence and glycemic variability
.
Diabetes Technol Ther
2010
;
12
:
581
585
[PubMed]
24.
Chalew
SA
,
McCarter
RJ
,
Ory-Ascani
J
,
Hempe
JM
.
Labile A1C is inversely correlated with the hemoglobin glycation index in children with type 1 diabetes
.
Diabetes Care
2010
;
33
:
273
274
[PubMed]
25.
Service
FJ
.
Correlation between glycemia and glycated hemoglobin
.
Compr Ther
1990
;
16
:
33
40
[PubMed]

Supplementary data