Mean hemoglobin A1c level by age and duration of diabetes mellitus. Error bars represent SE.
Carter JS, Gilliland SS, Perez GE, Skipper B, Gilliland FD. Public Health and Clinical Implications of High Hemoglobin A1c Levels and Weight in Younger Adult Native American People With Diabetes. Arch Intern Med. 2000;160(22):3471-3476. doi:10.1001/archinte.160.22.3471
Type 2 diabetes mellitus is a major public health issue for Native American people. Because glycemic levels are predictive of diabetes outcome, understanding determinants of high hemoglobin A1c (HbA1c) levels may provide targets for prevention efforts.
To investigate determinants of high HbA1c levels in Native American people.
We conducted a population-based, cross-sectional study of 206 participants with diabetes from 8 Native American communities in New Mexico. We used linear regression to assess the relationship of HbA1c level with age, body mass index (BMI), treatment type, duration of diabetes, physical activity, and diet.
Age, dietary pattern, and treatment type were determinants of HbA1c levels. Participants younger than 55 years had the highest adjusted HbA1c levels at 9.5% and those 65 years and older had the lowest levels at 7.8%. According to a participant's dietary intake, HbA1c levels were highest for those who consumed the most fat and sugar, and high consumption of fat and sugar affected HbA1c levels most among those younger than 55 years. Participants treated with insulin had the highest hemoglobin A1c levels. Physical activity was not associated with HbA1c level.
We found an increasing severity of diabetes among younger people. To avoid increased morbidity and mortality in the future, young Native American adults with diabetes need vigorous therapy to maintain tight glucose control.
TYPE 2 diabetes mellitus is a major health problem among Native Americans, with some communities experiencing 50% prevalence among adults.1 Prevalence of type 2 diabetes mellitus is rising nationwide, largely due to environmental influences on genetic susceptibility.1,2 In New Mexico, diabetes is a major concern for Native American communities: some tribes have a prevalence of 30% in persons 35 years and older.3 The rise in prevalence is worrisome because diabetes is associated with serious complications, poor quality of life, and reduced life span.4- 8 An early hypothesis that diabetes resulted in less morbidity and mortality in Native Americans than in other ethnic groups has been widely refuted.4,9- 15 Compared with the mortality rate of other ethnic groups, the mortality rate for Native Americans with diabetes is rising dramatically.5- 8
The risk factors of adverse outcomes among people with diabetes have been extensively studied.16- 22 Hemoglobin A1c (HbA1c) level is an important determinant of diabetes outcome. In populations such as the Pima, high HbA1c level is a strong predictor of retinopathy, nephropathy, and subsequent mortality.23- 25 Recognizing the determinants of high HbA1c level may contribute to a clearer understanding of modifiable antecedents of diabetes-related complications.
A number of factors affect glucose control among people with diabetes including age, body mass index (BMI), treatment type, duration of diabetes, physical activity, and diet. Although these factors are likely to be determinants of HbA1c level, they have not been studied in Native American populations other than the Pimas. To further investigate the determinants of HbA1c levels in Native Americans, we examined population-based data collected from 8 Pueblo communities in New Mexico by the Native American Diabetes Project (NADP) in 1994. We assessed the cross-sectional relationship of HbA1c level with age, BMI, treatment type, duration of diabetes, physical activity, and diet.
The target population was all Native American women and men diagnosed as having diabetes from 8 Pueblo communities in New Mexico. While no distinction was made between type 1 and type 2 diabetes mellitus for this study, very few Native American people have been identified with type 1 diabetes mellitus in this population. All communities share the Tanoan language group, and although they are not completely homogeneous with regard to culture and religion, the differences are small. The communities are served by 3 Indian Health Service (IHS) clinics.
From April 1994 to July 1994, we conducted a past-year review (April 1993 to April 1994) of the medical records of 514 people with diabetes served by the 3 participating IHS centers. We reviewed 100% of the records of people diagnosed as having diabetes in the IHS clinic database (patient care encounter forms). A standardized protocol previously described by IHS Diabetes Headquarters West Program was used.26 Measurements in the chart review included height, weight, and the 3 most recent blood glucose levels that were subsequently averaged. Hemoglobin A1c levels, if available, were recorded for all patients. Treatment type was recorded as diet, insulin, oral agent, or insulin and oral agent. Patients prescribed both insulin and oral agents were included in the insulin treatment category for this analysis.
Patients who were registered at the clinics with a diagnosis of diabetes were invited to participate in interviews. Interviewers contacted potential participants by telephone and/or in person. The protocol was approved by the institutional review boards of the IHS, the University of New Mexico, Albuquerque, and by the participating tribes and institutions. All participants signed informed consent forms. In 1994, we successfully recruited 206 of 514 adults with diabetes in the 8 communities. In 1995, 159 participants returned for second interviews, and an additional 66 new participants were recruited, for a total of 225 participants in 1995. The 1995 sample was used to initially develop the dietary scales. Otherwise, all data were reported for 1994 data only. For the chart review portion of the study, people who signed consent forms and participated in interviews were identified using NADP codes and were referred to as "participants"; all others were identified only as "nonparticipants" without any other identifiers.
Bilingual community members were trained in interviewing techniques, administering standardized questionnaires, and performing clinical measurements. The native language of the participants was used to explain the project and answer questions. Because of the difficulty of translating the questionnaire into unwritten languages, which might affect the questionnaire's validity,27 the questionnaire portion of the interviews was conducted in English; however, the use of native language was required for one elderly participant. Interviews took place in 1 of 3 offices in or near the communities of the participants. Participants were paid $10 for their time.
The height and weight of each participant were recorded to the nearest half inch and half pound, respectively, using standardized protocol and converted to centimeters and kilograms. Body mass index, calculated as weight in kilograms divided by the square of the height in meters, was used to define the overweight category (BMI, ≥27.8 to <32.3 for women; ≥27.3 to <31.1 for men) and the obesity category (BMI, ≥32.3 for women; ≥31.1 for men). Both overweight and obesity criteria followed National Health and Nutrition Examination Survey (NHANES) II references.28 The HbA1c level was measured by DCA 2000 using standardized protocol.29
A questionnaire was developed to evaluate dietary intake and levels of physical activity. The questionnaire development team consisted of university investigators, IHS professionals, a medical anthropologist, and community members. Existing measurement tools were assessed for clinical appropriateness for type 2 diabetes mellitus, cultural appropriateness for the target population, and ease of administration in face-to-face interviews. The questionnaire was pilot-tested by NADP staff in the intervention communities in the spring of 1994 with individuals who represented a range of ages, duration of diabetes, and diabetes control.
In 1994, the questionnaire was administered to 206 participants, and in 1995, to 225 participants (66 of whom were new to the study). The 1995 questionnaire had an additional 10 dietary questions added. From the 1995 questionnaire, a factor analysis was used to develop dietary fat and sugar scales from 21 dietary questions. Each question was evaluated on a 4-point scale for frequency of consuming the food (every day, several times a week, several times a month, or never) and scored from 0 to 3 (the lower score associated with a healthier diet). The scales were divided into tertiles by the total score. Fifteen of the 21 questions asked about the consumption of high- and low-fat foods, and the remaining 6 questions asked about the consumption of high- and low-sugar foods. Items were retained if the standardized regression coefficient was greater than 0.40. The analysis yielded scores for 5 items that were summed for the sugar scale and scores for 8 items that were summed for the fat scale. The Cronbach α values were .64 for the sugar scale and .72 for the fat scale.
Two of the 5 sugar items were not included in the 1994 form of the questionnaire; therefore, for the 1994 data the sugar scale consisted of 3 items: (1) Hawaiian Punch (Mott's Inc, Stamford, Conn), Kool-Aid (Kraft Foods Inc, Northfield, Ill), or a similar drink; (2) cookies or candy; and (3) doughnuts or sweet rolls. The fat scale items were (1) fried meats; (2) tortilla chips or potato chips; (3) Spam (Hormel Foods Corporation, Austin, Minn) or canned meat; (4) beans prepared with meat or fat; (5) fried potatoes cooked with fat; (6) pizza with meat on it; (7) fast foods such as Big Macs (McDonald's Corporation, Oakbrook, Ill), Whoppers (Burger King Corporation, Miami, Fla), Kentucky Fried Chicken (KFC Corporation, Louisville, Ky), or french fries; and (8) hot dogs, bologna, or bacon. The sugar and fat scale results were added together to obtain the dietary scale. The Cronbach α values for the 1994 questionnaire were .54 for the sugar scale, .67 for the fat scale, and .71 for the dietary scale. For all dietary scales, a higher score represents a worse diet based on higher consumption of foods that are high in sugar and fat.
Physical activity was assessed using an adaptation of the Pima Physical Activity Questionnaire of Kriska et al30 and has been documented elsewhere.31 Metabolic equivalents (METs) were assessed for each activity at 3 levels of intensity using a compendium of physical activities as described by Ainsworth et al.32 Metabolic equivalents for each activity were added to obtain MET hours per week, which were converted to kilocalories per week by multiplying by body weight in kilograms, as documented previously.31
We first examined the differences in participant and nonparticipant characteristics. Then we examined the relationship between HbA1c level and potential predictors using descriptive analyses. Differences in means were tested using analysis of variance (ANOVA), and differences in categorical variables were tested using χ2 statistics. We used standard linear regression techniques for the group as a whole and for age-specific strata to estimate the mean HbA1c level for the predictors of interest. The mean HbA1c level was adjusted for other covariates including age in years (18-54, 55-64, and ≥65), BMI (normal weight, overweight, and obese), treatment type (diet, oral agent, insulin), fat scale score (0-4, 5-7, ≥8), sugar scale score (0, 1-2, ≥3), and physical activity (0-1999, 2000-5499, and ≥5500 kcal/wk [0-8367, 8368-23,011, ≥23,012 kJ/wk]). All P values were 2-tailed. All statistical analyses were done using SAS statistical software (SAS Institute Inc, Cary, NC; 1990).
People recruited as participants were representative of the target population (n = 514) as evaluated in the chart review. Demographic and diabetes-related characteristics were similar among participants (n = 206) and nonparticipants with diabetes (n = 308) who resided in the 8 study communities (Table 1). Most participants were women (68.9%), who were slightly overrepresented. Participants had a mean age of 58.4 years, were diagnosed at a mean age of 49.3 years, and had a mean duration of diabetes of 8.8 years. Most participants were overweight or obese with an average BMI of 31.2 kg/m2. Participants had higher BMI values than nonparticipants. Diabetes among participants was not under tight control, as reflected by a mean of the last 3 recorded blood glucose levels of approximately 13.9 mmol/L (250 mg/dL) and HbA1c level of 11.6%. Most participants were treated with oral agents (45.7%); the other participants were treated with diet therapy (27.7 %) or insulin (26.6 %). There were no statistically significant differences in diabetes-related characteristics between women and men.
Table 2 gives the ages of the participants at the time of the interview. Treatment type and dietary pattern were strong determinants of high HbA1c levels. The mean HbA1c level in the 18- to 39-year age group was 10.1% and decreased to 7.7% in the group aged 70 years or older. Those treated with insulin had higher HbA1c levels than those receiving diet therapy or oral medications. Participants in the highest tertile of fat and combined fat and sugar intake had higher HbA1c levels than those with lower intake. Physical activity showed a paradoxical trend toward higher HbA1c levels in the highest physical activity category. There was an inconsistent relationship between the duration of diabetes and high HbA1c level.
The HbA1c levels were also significantly higher for those younger than 55 years who were of normal weight or overweight compared with those who were obese. Hemoglobin A1c levels were the highest for those treated with insulin and lowest for those treated with diet therapy. For every treatment type, the participants in the youngest age group had the highest HbA1c levels. Even for a short duration of diabetes, HbA1c levels were higher for the youngest age group as illustrated in Figure 1. Hemoglobin A1c levels varied by fat and sugar consumption within each of 3 age groups (18-54, 55-64, and ≥65 years) (Table 3); however, the magnitude of variation was larger among participants younger than 55 years than for the older participants.
The results of the multivariable regression show that age was a strong and significant determinant of high HbA1c level (P<.001). After adjusting for BMI, treatment type, fat and sugar consumption, and physical activity, participants in the group younger than 55 years had the highest HbA1c level at 9.5%, participants aged 65 years or older had the lowest level at 7.8%, and participants in the 55- to 64-year age group had an intermediate value at 8.7% (P<.001). Both treatment type and fat and sugar consumption were associated with higher HbA1c levels. At the levels of physical activity reported by participants in this study, physical activity was not associated with HbA1c levels.
In the age-specific linear regression, the determinants of high HbA1c level varied in the 3 age groups. Younger age (<55 years), being of normal weight or overweight, and consumption of more high-fat and high-sugar food items were significant predictors of high HbA1c level. In contrast, treatment type was the sole significant predictor of high HbA1c levels in the 55- to 64-year age group. None of the available predictors was associated with high HbA1c levels in the group aged 65 years or older.
We found that age, treatment type, and dietary pattern were determinants of high HbA1c levels in our population-based study of Native American people with diabetes. After adjusting for BMI, treatment type, fat and sugar consumption, and physical activity, participants younger than 55 years surprisingly had the highest HbA1c levels, and those 70 years or older had the lowest. The effects of diet and treatment type varied by age. Hemoglobin A1c levels were highest for participants who consumed the most fat and sugar, and high consumption of fat and sugar affected HbA1c levels most among those younger than 55 years. Treatment type was also associated with HbA1c level. Participants treated with insulin had the highest HbA1c levels. This may be a result of the progressive nature of diabetes, with those in worse glycemic control being treated with insulin. But the results suggest that treatment with insulin is not producing glycemic control tight enough to improve HbA1c levels in those younger than 55 years. This suggests an opportunity to improve diabetes outcome by improving control of glucose levels. At the levels of physical activity reported by participants in this study, physical activity was not associated with HbA1c level after adjusting for age, BMI, treatment type, and fat and sugar consumption.
Our finding that HbA1c levels were highest in the youngest age group was unexpected. The relationship of high HbA1c level and age may have a number of explanations. First, young people in our study may have worse diabetes because they have the greatest degree of obesity and the highest insulin resistance. The age of onset of type 2 diabetes mellitus is becoming younger, and there is evidence of a cohort effect of diabetes-related mortality.33 These epidemiologic trends are consistent with a worsening spectrum of diabetes severity.8 Second, younger people with diabetes may not be treated as vigorously as older people. However, this is unlikely to completely account for the findings in this study because for every treatment type the youngest age category had the highest HbA1c level. Third, there could be a survival bias: the older people who had higher HbA1c levels may have died at a higher rate than those with lower HbA1c levels. We have no data to directly assess this explanation. However, as shown in Figure 1, younger people have higher HbA1c levels than older people, even in the group with a duration of diabetes of less than 5 years (a period in which diabetes-related mortality is likely to be low and any survival bias is likely to be small). A more likely explanation is that there is a cohort effect, with the younger cohort having higher HbA1c levels. A cohort effect has been observed among the Pima Indians in Arizona and is related to the increasing prevalence of obesity, increasingly sedentary lifestyles, and in utero exposure to maternal diabetes.33,34 Additional longitudinal studies are needed to investigate the factors associated with this apparent cohort effect.
We found that HbA1c levels were significantly higher for those in the group younger than 55 years who were of normal weight or overweight compared with those who were obese. Although we lack data for further investigation, this paradoxical finding may be related to worse glucose control in the younger age group and a catabolic state with an associated weight loss without improved glucose control.
Our finding that levels of physical activity were not associated with HbA1c levels in people with type 2 diabetes mellitus is consistent with other reports.35 Although physical activity is associated with lower blood glucose levels, the levels of physical activity reported by NADP study participants may not have been sufficient to affect HbA1c levels.31
In our data the lack of an association between physical activity and HbA1c level may be related to higher HbA1c levels in people in the youngest age group, who also have the highest level of physical activity.31 Further research is needed to determine the frequency, duration, and intensity of physical activity necessary to lower HbA1c levels in different age groups and populations.
In contrast to the lack of effect of physical activity, higher dietary intake of fat and sugar was associated with higher HbA1c levels. We are unaware of other studies that have investigated the relationship of high HbA1c levels and dietary intake among Native Americans. These findings emphasize a clear target for secondary prevention efforts. Contemporary diets include the consumption of large amounts of high-fat and high-caloric foods. Encouraging traditional lifestyles and foods, which are characterized by diets with less animal fat and more complex carbohydrates, may be helpful.
Our study has a number of strengths and limitations. Participants were a representative sample of the population with type 2 diabetes mellitus in the 8 Pueblo communities (Table 1). Innovative and culturally appropriate methods were developed and used to assess fat and sugar consumption and physical activity. An important limitation is the cross-sectional design, which does not allow for assessment of temporality and uses prevalent rather than incident cases. We also had a relatively small sample size to precisely estimate the magnitude of effects. Although selection bias may have arisen by differential participation among young people with diabetes, which could limit generalization, we found that HbA1c levels and mean blood glucose levels did not differ by age group for participants and nonparticipants. Furthermore, we found no significant modification of the effect of age on HbA1c level by participation status. Only people from Native American Pueblo communities in New Mexico were included in this study, so findings may not be generalizable to other tribes.
Much of what we know about type 2 diabetes mellitus in Native Americans comes from work done with the Pima tribe of Arizona.25,34,36- 38 Our finding of an increasing severity of diabetes at younger ages is consistent with a cohort effect and is similar to temporal trends in the Pima people.33 To avoid increased morbidity and mortality in the future, young people with diabetes need to be vigorously treated to maintain tight glucose control. Oral agents and insulin therapy should not be delayed for too long when HbA1c levels are high. Furthermore, developing effective primary prevention strategies to delay the onset of type 2 diabetes mellitus and reduce the risk from in utero exposure to maternal diabetes must have the highest priority. Further research is needed to see if the same patterns are apparent in Native American people with diabetes nationwide.
Accepted for publication June 16, 2000.
This study was supported by the Veterans Affairs Medical Center, Albuquerque, NM; grant DK 47096 from the National Institute of Diabetes, Digestive and Kidney Diseases, Bethesda, Md; and dedicated funds of the University of New Mexico, Albuquerque.
Reprints: Janette S. Carter, MD, Department of Internal Medicine, University of New Mexico School of Medicine, Albuquerque, NM 87131 (e-mail: firstname.lastname@example.org).