Longitudinal study shows increasing obesity and hyperglycemia in micronesia


  • Disclosure: The authors declare no conflicts of interest.

  • Funding agencies: This material is based in part upon work supported by the Texas Norman Hackerman Advanced Research Program under Grant No. [THECB] 02006; NHARP proposal number 0049-0041-2009.

Correspondence: Paul Farand (pbonnen@bcm.edu)



Obesity and diabetes are particularly high in indigenous populations exposed to a Western diet and lifestyle. The prevalence of obesity, diabetes, hyperglycemia, dyslipidemia, and hypertension in one such population, the Micronesian island of Kosrae was described.

Design and Methods

Longitudinal screenings for metabolic traits were conducted on adult Kosraens ≥ 20 years of age in 1994 and again in 2001. Data was obtained on 3,106 Kosraens, comprising ∼80% of the adult population. Diabetes was diagnosed using World Health Organization guidelines. Prevalences of obesity, hyperglycemia, dyslipidemia, and hypertension were assessed.


The overall age-adjusted prevalence of diabetes increased from 14 to 21%. The most significant change observed in the population was increases in obesity and hyperglycemia, especially among young Kosraens and women. Obesity age-adjusted prevalence increased from 45 to 62%, and hyperglycemia age-adjusted prevalence increased from 19 to 44%. Of note, Kosraens as a group had unusually low high density lipoprotein (HDL) levels with 80% classified as low HDL by NCEP-ATPIII criteria, despite lacking the usually accompanying increase in triglycerides. Comparison to reports from other populations shows that Kosrae experiences one of the highest rates of obesity, hyperglycemia, and low HDL globally while maintaining relatively healthy levels of triglycerides.


Our study shows a dramatic increase in obesity and hyperglycemia in Kosrae in just 7 years and forebodes significantly increased health risks for this part of the world.


The prevalence of metabolic disease such as obesity and diabetes is thought to be particularly high among indigenous populations exposed to a Western diet and lifestyle. Kosrae is a Pacific Island member of the Federated States of Micronesia (FSM) and home to ∼7,700 individuals. Native diet remained largely intact until 1948, when the FSM became a US protectorate. Kosrae continues to receive aid from the US under a Compact of Free Association. US aid brought high fat, high salt food to the island which resulted in a loss of incentives for fishing and farming and consequently a sedentary lifestyle. In 1994 a population screening found a high prevalence of obesity and diabetes [1].

To understand the current status of obesity and diabetes as well as the rates of change of prevalence of metabolic disorders on Kosrae, we have conducted a longitudinal study of obesity, diabetes, hyperglycemia, dyslipidemia, and hypertension. This work shows an overall increase in obesity and hyperglycemia with the greatest increases in women and in the young adult members of the population.


Clinical data

The Kosrae cohort studied here has been previously described [2]. Briefly, DNA and clinical data containing anthropometric and serum chemistry measurements were obtained from 3,106 individuals on Kosrae. Two samplings were taken, one in 1994 and one in 2001, with 1,084 individuals sampled both years. For this study we analyzed height, weight, body mass index (BMI), waist circumference, systolic and diastolic blood pressures (SBP,DBP), fasting blood glucose (FBG), 2-h oral glucose tolerance test (OGTT), serum triglycerides (TG), apolipoprotein A-I (apoA-I, 1994 only), high-density lipoprotein (HDL, 2001 only) cholesterol, and microalbuminuria (2001 only).

In the 1994 Kosrae survey direct quantification of HDL was not possible and the major lipoprotein of HDL, apoA-I, was used as a substitute [1]. To calculate a low apoA-I threshold lipid distribution data from several studies in which both apoA-I and HDL were measured on the same individuals were examined [3]. Study participants were adult men and women from white and black US populations. We ascertained the percentiles within the HDL distribution where HDL levels of 40 mg dL−1 (men) and 50 mg dL−1 (women) fell. We then estimated corresponding apoA-I levels from the apoA-I distribution using those same percentiles. Averaging values across studies provided one gender-specific apoA-I threshold. Using this technique we established low apoA-I thresholds of 122 mg dL−1 (men) and 140 mg dL−1 (women) for the national cholesterol education program adult treatment panel III (NCEP-ATPIII) criteria. Throughout the remainder of the paper HDL refers to both apoA-I (1994 only) and HDL (2001 only).

Blood pressure measurements were taken while seated and followed an established protocol in which the first measurement is disregarded [6]. In 1994 three measurements were taken, however in 2001 only two were recorded. To have parity across sampled years, for both years we treated the data the same and discarded the first value and used the second value as the blood pressure measurement.

Characterization of metabolic profiles

Metabolic phenotypes of the cohort were characterized for mean and prevalence of metabolic disorders: hyperglycemia, dyslipidemia, obesity, and hypertension. To be included in the study individuals had to meet the following requirements. (1) Nonpregnant individuals aged ≥20 years. (2) Clinical data was available in every tested category. (3) For those with data in both 1994 and 2001 values could not change by ≥3 standard deviations during that time. For NCEP-ATPIII criteria a total of 1,892 individuals (796 Male, 1,096 Female) in 1994, and 1,610 (585 male, 1,025 female) in 2001 met these requirements and were included in the metabolic syndrome analysis; 954 were sampled in 1994 and 2001.

Metabolic disorders were defined as:

  1. Hyperglycemia: fasting blood glucose ≥ 100 mg dl−1
  2. Hypertriglyceridemia: triglycerides ≥ 150 mg dl−1
  3. Hypo-HDL: HDL < 40 mg dl−1 (men) or 50 mg dl−1 (women)*
  4. Obesity: waist circumference >102 cm (men) or > 88 cm (women)
  5. Hypertension: systolic blood pressure ≥ 130 or diastolic blood pressure ≥ 85

Determination of diabetic status

Diabetes was diagnosed among nonpregnant individuals aged ≥20 years using WHO guidelines in which either a fasting blood glucose ≥126 mg dl−1 or oral glucose tolerance test ≥200 mg dl−1 is sufficient for diagnosis. A diabetes diagnosis was also made if an individual had a previously known diagnosis and/or was prescribed antidiabetes medication (insulin or oral drugs). Prediabetes was defined using American diabetes association (ADA) definitions of impaired fasting glucose (FBG 100-125 mg dl−1) and impaired oral glucose tolerance (OGTT 140-199 mg dl−1).

Age-adjustment of clinical values

As the age distribution of the Kosrae population differed in 1994 and 2001, we used the direct method of age-adjustment to derive mean trait values and prevalence rates comparable across multiple years. Briefly, the population ≥20 years from each sampled year was separated by gender and divided into 10-year age groups. Each age group's mean trait value and prevalence rate was then calculated and multiplied by an age group-specific weight. This weight represented the proportion of the Kosrae population within that age group as derived from the 2000 Kosrae Census data [7]. The weighted prevalence rates were then summed over all age groups to give an overall age-adjusted mean trait value and prevalence.

Statistical analysis

Pearson's chi-square test (or Fisher's exact test for small group sizes) was calculated to test for statistical difference on the adjusted prevalence estimates from 1994 and 2001. Unpaired t test was used to determine statistical difference between years. Statistical tests were performed using SPSS 17.0 (SPSS, Chicago, IL). A threshold of P ≤ 0.001 was used as the standard for significance. In addition, the Bonferroni correction for multiple testing was applied.


Prevalence of obesity, hypertension, dyslipidemia, and insulin resistance on Kosrae

Age-adjusted mean values for the parameters measured in this study reveal the primary features of metabolic disease on Kosrae (Table 1). Waist circumference, fasting blood glucose levels, and HDL were at relatively unhealthy levels while triglycerides were observed in the normal range. Mean fasting blood glucose increased dramatically in men and women between 1994 and 2001. We observed a 24% increase in fasting blood glucose in men from 91 to 113 mg dl−1 and a 29% increase in women from 90 to 116 mg dl−1. Notably, the 2001 mean blood glucose levels for both genders were sufficient for ADA diagnosis of impaired fasting glucose. Waist circumference also increased significantly in both genders and especially in women. Waist circumference in men increased from 91 to 95 cm. For women the 1994 mean waist circumference was already considered elevated at 89 cm and increased to 99 cm. In 2001 mean HDL levels were low at 35 and 40 mg dl−1 for men and women, respectively. Despite high waist circumference, high glucose levels, and low HDL, mean triglyceride levels remained fairly stable over time at a relatively healthy level in men (100 mg dl−1) and women (82 mg dl−1) in 2001.

Table 1. Age adjusted mean values of metabolic parameters
NMean ± SEMNMean ± SEMP value
  1. The mean trait value in each year was compared and tested for whether there was a significant difference. *, Two tailed P value was significant. Bonferroni correction for multiple testing yields a significance level of P value < 0.0001. NA, HDL was not measured in both years.
Fasting blood glucoseMale82591 ± 0.9626113 ± 1.6*
Female1,13790 ± 0.81057115 ± 1.5*
TriglyceridesMale801107 ± 2.262799 ± 1.9-
Female1,09883 ± 1.2105782 ± 1.0-
apo A-I/HDLMale803109 ± 0.859235 ± 0.4NA
Female1,104123 ± 0.7103140 ± 0.3NA
Systolic blood pressureMale825120 ± 0.5622116 ± 0.6-
Female1,138113 ± 0.41056112 ± 0.4-
Diastolic blood pressureMale82578 ± 0.462277 ± 0.4-
Female1,13873 ± 0.3105675 ± 0.3-
WaistMale82191 ± 0.462495 ± 0.5*
Female1,13689 ± 0.4105899 ± 0.4*

Analysis of the individuals ascertained in both years of sampling, 1994 and 2001, present an opportunity directly observe the change of metabolic status over time. There were 329 men and 625 women sampled in both years who had complete phenotyping for all traits measured and met criteria for inclusion. The mean trait values in this subset of the cohort reflect those of the larger population (Table 2). A significant increase in fasting blood glucose levels and waist circumference was observed in both men and women.

Table 2. Mean values of metabolic parameters for individuals who participated in the study in both 1994 and 2001
CategorySexMean ± SEMAge-adjusted mean ± SEMP value
  1. There were 329 men and 625 women sampled in both years. *, Two tailed P value for the difference between adjusted mean values was significant. Bonferroni correction for multiple testing yields a significance level of P value < 0.0001. NA, HDL was not measured in both years.
Fasting blood glucoseMale93 ± 1.6130 ± 3.385 ± 0.9112 ± 2.3*
Female90 ± 1.1126 ± 2.486 ± 0.8116 ± 1.8*
TriglyceridesMale125 ± 10.8110 ± 3.0109 ± 5.6104 ± 2.5-
Female86 ± 1.689 ± 1.481 ± 1.482 ± 1.3-
apo A-I/HDLMale106 ± 1.235 ± 0.5102 ± 1.033 ± 0.4NA
Female123 ± 1.041 ± 0.4123 ± 0.940 ± 0.4NA
Systolic blood pressureMale122 ± 0.7121 ± 0.9120 ± 0.6117 ± 0.7-
Female115 ± 0.6177 ± 0.7112 ± 0.4112 ± 0.5-
Diastolic blood pressureMale81 ± 0.580 ± 0.580 ± 0.580 ± 0.5-
Female76 ± 0.478 ± 0.474 ± 0.374 ± 0.4-
WaistMale94 ± 0.699 ± 0.689 ± 0.698 ± 0.7*
Female92 ± 0.5103 ± 0.588 ± 0.5102 ± 0.5*

The prevalence estimates of obesity, hypertension, dyslipidemia, and insulin resistance on Kosrae were calculated (Table 3). The most common individual metabolic complication on Kosrae for both men and women is low HDL/apoA-I. The prevalence of hypo HDL/apoA-I is very high in men and women (2001, 72% M and 84% F) and is steady across years in both genders. Paradoxically hypertriglyceridemia and low HDL/apoA-I trend in opposite directions on Kosrae, with hypertriglyceridemia being the least common metabolic problem. Age-adjusted prevalence is higher in men than women (2001, 13% M and 6% F) and the overall prevalence is relatively stable from 1994 to 2001.

Table 3. Unadjusted and age-adjusted prevalence values for metabolic phenotypes
CategorySexUnadjusted prevalenceAdjusted prevalanceP value
1994 (%)2001 (%)1994 (%)2001 (%)
  1. Only individuals with values for every trait were included. Resultingly, the number of individuals sampled for each trait was the same across traits: Men 1994 = 796; Men 2001 = 585, Women 1994 = 1,096; Women 2001 = 1,025. The P value is from the test of whether age-adjusted prevalences for a trait were different from each other. Bonferroni correction for multiple testing yields a significance level of P value < 0.00005.
Low HDL/apo A-IMale75727672-
Abdominal obesityMale27352532-

Abdominal obesity is the second largest metabolic problem for women on Kosrae with a high prevalence of 79% in 2001. This is a significant increase from 59% in 1994. Abdominal obesity in men is lower than women in both years at 25% in 1994 and 32% in 2001. While an increase in men is apparent it is not nearly as significant as for women. Hypertension remained largely unchanged between the sampled years and is higher in men than women who had a prevalence of 34 and 27% respectively in 2001.

The largest change in the Kosraen metabolic profile is a dramatic increase in hyperglycemia that was observed in both men and women. From 1994 to 2001 the overall prevalence of hyperglycemia more than doubled on Kosrae from 19 to 44%. The increase was also extremely significant when measured in men (19-45%, P = 1.75E-26) and women (18-44%, P = 2.9E-36) independently.

Analysis of these five metabolic phenotypes with age demonstrates escalating metabolic dysfunction in the younger population (Tables 4 and 5). Women from 20 to 69 years exhibited a significant rise in hyperglycemia and abdominal obesity. Increases were most pronounced in younger age groups where prevalence in hyperglycemia tripled (age 20-29, 5-17%; age 30-39, 11-39%). Female abdominal obesity also showed the highest increases in the two younger groups (age 20-29, 37-59%; age 30-39, 59-86%). Similarly men exhibited a highly significant increase in hyperglycemia in the 20-49 year age groups with the largest increase in the youngest age group. Men displayed an increase in hyperglycemia in ages 20-29 from 1-20% (P = 1.1e-07) and for age group 30-39 an increase of 7-37% (P = 3.7e-12). The highest prevalence of abdominal obesity for both men and women was in the 50-59 age group which showed very high rates of obesity particularly for women and a great disparity between the genders with 54 and 94%, respectively. Current prevalences of abdominal obesity are high in general and are among the highest reported globally for women. Both hyperglycemia and obesity increased significantly on Kosrae especially in the younger members of the population. The remaining metabolic parameters did not show a significant age effect.

Table 4. Prevalence estimates for metabolic syndrome and its five components in each 10-year age group using NCEP-ATPIII criteria
CategoryAge groupMaleP valueaFemaleP valuea
1994 Prevalence (%)2001 Prevalence (%)1994 Prevalence (%)2001 Prevalence (%)
  1. aBonferroni correction for multiple testing yields a significance level of P value < 0.000033
Low HDL/apo A-I20-297869-7885-
Abdominal obesity20-291016-3759a
Table 5. Prevalence estimates for metabolic syndrome and number of individuals tested in each age group
Age groupMaleFemale
1994 (N)2001 (N)1994 (N)2001 (N)

Diabetes and prediabetes are increasing on Kosrae

Diabetes and prediabetes prevalence among the Kosraen population was estimated (Table 6). Overall adjusted prevalence of diabetes shows an upsurge of more than 40% between 1994 and 2001, going from 14 to 21%. Pre-diabetes overall prevalence showed a dramatic increase from 6 to 32%, an increment of more than 400%. No gender bias was observed with regard to diabetes or prediabetes prevalence.

Table 6. Unadjusted and age-adjusted prevalence of diabetes and prediabetes on Kosrae
CategorySex1994 (n)2001 (n)Unadjusted prevalenceAdjusted prevalenceP alue
1994 (%)2001 (%)1994 (%)2001 (%)
  1. The difference between adjusted prevalences for each year was tested for significance using threshold of P value < 0.00017.

Analyzing the population in 10-year age bins reveals a large increase in prediabetes occurring particularly in younger ages (Tables 7 and 8). For both genders, most cases of diabetes appear in individuals older than 40. However, prediabetes prevalence was dramatically increased in all age categories, with a surge of at least 200%. The largest increase in pre-diabetic status occurred in younger ages with the prevalence in 20-29 year olds increasing from 0 to 11% (men) and 1 to 14% (women).

Table 7. Prevalence values for diabetes and prediabetes on Kosrae population in each 10-year age group using WHO guidelines for epidemiological studies and ADA definitions
CategoryAge groupMaleP valueFemaleP value
1994 Prevalence (%)2001 Prevalence (%)1994 Prevalence (%)2001 Prevalence (%)
  1. Significance levels after correcting for multiple testing are P value < 0.000056.
Table 8. Prevalence values for diabetes and prediabetes on Kosrae population and the number of individuals tested in each age group
Age groupMaleFemale
1994 (N)2001 (N)1994 (N)2001 (N)


Pacific Islanders are one of the highest risk groups for metabolic disease. To the best of our knowledge this is the only report of obesity and diabetes in Micronesians living in Micronesia. We present here the result of an epidemiological study of obesity, diabetes, hyperglycemia, dislipidemia and hypertension on the Micronesian island of Kosrae. This longitudinal study supports several conclusions about the nature of metabolic syndrome on Kosrae. First, there is a dramatic rise in prevalence of hyperglycemia and abdominal obesity. Second, the increase impacts both genders and particularly women. Third, the effects are widespread acting on multiple age groups and increasingly younger member of the population. And finally, the pattern of metabolic syndrome on Kosrae includes unusual lipid profiles.

It is noteworthy that both obesity and hyperglycemia are increasing across all ages on Kosrae including the youngest adults on the island. This trend has been noted in other populations but appears to be accelerated on Kosrae [6]. With increasingly younger Kosraens developing prediabetes and features of metabolic syndrome there is increased disease burden since early presentation will result in affected individuals having more time to develop associated complications such as diabetes, coronary heart disease, myocardial infarction, and stroke. Our study also shows that women are disproportionately affected by this increase. This finding is in line with other studies that have shown higher prevalence of metabolic abnormalities among women compared to men [6, 8, 9]. The combined effect of these two trends of metabolic abnormalities affecting more women and more young adult members of the population has serious consequences for future generations as in utero exposure to high levels of circulating glucose has been associated with childhood insulin resistance and obesity [10]. This continues the cycle of early onset of metabolic disorders in each subsequent generation on Kosrae.

This study establishes that Kosraens have unusual lipid profiles. Low HDL and hypertriglyceridemia are commonly observed together and in association with insulin resistance. Because Kosrae suffers from high rates of obesity and hyperglycemia it would be expected to also have both hypo-HDL and hypertriglyceridemia. However, while low HDL is rampant on Kosrae hypertriglyceridemia is the least common metabolic problem with 13% of men and 6% of women meeting the NCEP-ATPIII criteria of ≥150 mg dl−1. By comparison to another indigenous population with high rates of obesity and diabetes, a study of Navajo Native Americans showed hypertriglyceridemia is present at much higher rates than Kosrae with 48% of men and 46% of women affected [11]. Kosraens mean triglycerides levels (100 M, 82 F) are also lower than US non-Hispanic whites (134 M, 117 F) [12] and three different Native American communities where mean TG values ranged from 122 to 133 in men and 117 to 123 in women [13]. Perhaps counter-intuitively, Kosraens had lower TG levels (104 M, 87 F Kosraens ≥35 years) than Samoans (119 M, 111 F Samoans ≥35 years) a fellow Pacific Islander population that is Polynesian rather than Micronesian [14]. A population of individuals who share a similarly unusual relationship between TG levels and obesity and diabetes is US blacks [15]. While also having high rates of obesity and diabetes US non-Hispanic blacks are reported to have lower rates of hypertriglyceridemia (21% M, 14% F) [16] and lower mean TG levels (99 M, 90 F) [12].

In stark contrast to having healthy levels of triglycerides, Kosraens experience dramatically low levels of HDL. Our data show that in 2001 a very high percentage of Kosraen adults had clinically low levels of HDL (72% M, 84% F). Mean HDL levels in Kosrae (35 M, 40 F) are lower than US non-Hispanic white (46 M, 57 F), US non-Hispanic black (51 M, 57 F) [12], Native Americans (41 M, 48 F) [13], as well as Samoans (Samoan mean HDL is 43 M, 47 F, ≥35 years [14]; Kosrae ≥35 years mean HDL is 34 M, 39 F). One population that appears to have similarly low levels of HDL is Turkey. Mean HDL values were noted to be low regardless of diet and ranged from 34 to 38 in men and 37 to 45 in women [17]. Subsequent studies in the Turkish population have shown genetic effects associating with HDL levels [18, 19].

In the clinical setting TG and HDL levels are used as indicators of an individual's risk for coronary heart disease and stroke. Additionally, dyslipidemia as indicated by TG and HDL levels, is observed in association with insulin resistance. However, on Kosrae, while levels of hyperglycemia increased substantially between the years of this study, mean TG and HDL did not change. The prevalence of low HDL on Kosrae is observed relatively equally in women and men, does not change with increasing age, and did not change significantly over the time period of this study. We conclude that Kosrae has a population distribution of HDL levels that are lower overall than other populations and suggests a need for population specific thresholds for HDL in order for HDL levels to be indicative of risk for diabetes and metabolic syndrome. Similarly, TG levels were relatively unchanged over time and did not appear to have a correlation with diabetic status. Recent data suggest a 13 and 12% reduction in risk of cardiac death or non-fatal myocardial infarction for every 5 mg dl−1 increase in HDL and 50 mg dl−1 decrease in TG concentration, respectively [20]. However, more research is needed to determine the comorbidities associated with the distribution of HDL and TG levels in Micronesians. Thus while Kosraens demonstrate elevated rates of metabolic syndrome, their lipid profiles are distinct from other populations including those that show high rates of metabolic disease. It is important to acknowledge these population specific lipid distributions in screening members of this population for risk factors for diabetes, cardiovascular disease and stroke.


We thank Dr. Jan Bressler, Human Genetics Center, University of Texas Health Science Center at Houston, TX, for commenting on the manuscript and helpful advice. We acknowledge the invaluable contribution of Vita Skilling, Secretary of the Department of Health and Social Affairs for the Federated States of Micronesia, and thank her for her long-standing collaboration. We are deeply grateful to the people of Kosrae and the Kosrae Department of Health Services for making this study possible.