Early testing of insulin resistance: a tale of two lipid ratios in a group of 5th graders screened by the Coronary Artery Risk Detection in Appalachian Communities Project (CARDIAC Project)

Abstract

Background

In West Virginia (WV), 47% of fifth-grade children are either overweight or obese. There is no clear consensus regarding the definition of insulin resistance in children, and directly measuring insulin on the population level is costly. Two proposed measures examined further in this study include triglyceride (TRIG)/high-density lipoprotein cholesterol (HDL-C) ratio and TRIG/low-density lipoprotein (LDL-C) ratio. The purpose of this study is to examine the relationship between TRIG/HDL-C ratio, TRIG/LDL-C ratio and insulin resistance in fifth-graders with acanthosis nigricans (AN).

Methods

Between 2007 and 2016, 52,545 fifth-grade students in WV were assessed for AN. Fasting glucose and insulin levels were collected only for a sub-group of students who were AN-positive and was used to determine insulin resistance using the Homeostatic Model for Insulin Resistance (HOMA-IR) equation. Statistical analysis included t tests and logistic regression with receiver operating characteristic curves.

Results

Of the students assessed for AN, 4.5% (n = 2360) tested positive. The prevalence of insulin resistance was 79% (n = 814) among 1030 with AN and complete HOMA-IR. TRIG/HDL-C ratio and TRIG/LDL-C ratio were significantly associated with insulin resistance (TRIG/HDL-C:Est. = 0.36, P < 0.0001, AUC = 0.68; TRIG/LDL-C: Est. = 0.87, P < 0.0001, AUC = 0.69). Multivariate analysis showed that increased body mass index (Est. = 0.05, P < 0.0001), gender (Est. = 0.49, p < 0.0001) and TRIG/HDL-C ratio (Est. = 0.21, P < 0.0001) were significantly associated with insulin resistance.

Conclusions

TRIG/HDL-C is a better surrogate marker of insulin resistance in AN-positive children compared to TRIG/LDL-C ratio; so, on a population-level, cholesterol rather than insulin may be obtained for preliminary testing of early insulin resistance in children.

Introduction

Childhood obesity and insulin resistance is a growing problem in the United States over the last two decades [1]. In West Virginia, 47% of 5th grade students are either overweight or obese [2]. Childhood obesity comes with a myriad of complications ranging from psychosocial issues (poor self-esteem, depression and eating disorders) to endocrine complications, such as insulin resistance [3]. Although the role of insulin resistance in the development of diabetes mellitus, hypertension and dyslipidemia is well-established, there is currently no consensus in the definition of insulin resistance in children and adolescents [4–7]. Additionally, at a population-level, directly measuring insulin is costly [relative to more cheaply obtained lipid profile or assessment of acanthosis nigricans (AN)]. This has led to the search for surrogate (i.e., cheaper and more easily measured) markers with similar predictive validity of increased risk of diabetic conditions.

Among the many methods used to assess insulin resistance, the euglycemic, hyper-insulinemic clamp is the gold standard and is used mostly for research purposes only [4, 6, 8]. It is labor intensive, expensive and not suitable for epidemiological studies [4, 6, 9]. The Homeostasis Model Assessment of Insulin Resistance equation (HOMA-IR) on the other hand is a more convenient method to assess insulin resistance in large epidemiological studies [4, 10]. It was developed by Mathews and colleagues in 1985 [10] and has been shown to have a strong positive correlation with the gold standard [10]. Clinical markers of insulin resistance described in children and adults include AN (which is a dermatologic finding characterized by thickened, dark skin in the axilla, nape of the neck, and other surface areas exposed to friction), triglyceride (TRIG) levels and TRIG/high-density lipoprotein cholesterol (HDL-C) ratio levels [5, 11, 12]. Although AN is an indirect measure of insulin resistance, its ability to detect insulin resistance is limited. Ice and colleagues showed that the prevalence of insulin resistance measured by HOMA-IR in 667 5th graders with AN was 61% [5]. In addition, AN is associated with other medical conditions such as malignancies, genetic disorders, polycystic ovarian syndrome and medications (although rare in childhood) [13]. This underscores the need to search for other clinical surrogate markers such as TRIG/HDL-C and TRIG/LDL-C ratios.

Triglycerides and cholesterol metabolism share a common pathway with insulin resistance. According to the third National Health and Nutrition and Examination (NHANES) survey, the most frequently encountered component of metabolic syndrome in adolescents was high TRIG and low HDL-C levels [14]. Insulin resistant patients generally have high TRIG, TRIG-enriched LDL-C, and decreased HDL-C levels [15]. Although TRIG/HDL-C varies with ethnicity, no significant gender difference in TRIG/HDL-C ratio has been demonstrated in most studies [4, 11, 16]. Given the close relationship between dyslipidemia and insulin resistance, and the quest to search for cheap and cost-effective measures, two proposed measures examined in the present study include TRIG/HDL-C and TRIG/LDL-C ratios. We selected our novel TRIG/LDL-C as a potential biomarker because of the close link between cholesterol synthesis and insulin resistance. Studies have shown an increase in cholesterol synthesis in insulin resistant patients with metabolic syndrome or obesity [17]. In addition, insulin resistant patients (especially with obesity) have elevated small dense lipoprotein cholesterol generated from TRIG-enriched LDL-C.

The purpose of this research is to examine the relationship between TRIG/HDL-C and TRIG/LDL-ratios and insulin resistance in 5th graders with AN screened by the Coronary Artery Risk Detection in Appalachian Communities Project (CARDIAC) between 2007 and 2016.

Methods

The CARDIAC Project started as a small school-based cardiovascular disease surveillance project piloted in three rural West Virginia (WV) counties in 1998. It has grown to an expanded multi-dimensional screening, research and intervention effort involving all 55 WV counties and over 480 schools. When the program stopped in 2017, more than 200,000 children from select grades had received free health screening through the program. This is particularly important given the limited per capita income of families living in the state and the need to provide health screenings to all children [18]. For nearly two decades, CARDIAC has provided information to participating families, communities, the state and nation [19] about chronic illnesses including hyperlipidemia [20], abnormal blood lipids and obesity [21, 22]; asthma [23]; decreasing cholesterol risk [24]; pre-diabetic conditions [5]; health behaviors [25]; and intervention factors [26, 27]. Average findings for the program period demonstrate that from 1998 to 2017, 47.1% of fifth grade students in WV were either overweight (BMI percentile 85–94.9th) or obese (BMI percentile > 95th). More information about CARDIAC can be found on the website at http://www.cardiacwv.org. Only fifth grade participants receive lipid profiles and thus are the only participants included in this study. Between 2007 and 2016, 55,575 fifth-grade students were enrolled in the CARDIAC study. Of the 52,545 students who were assessed for AN, 4.5% (n = 2360) tested positive. Of the 2360 who were tested positive, 1030 had complete HOMA-IR values and are the focus of the analyses included in this study.

West Virginia University Institutional Review Board approved the study protocol for the Protection of Human Subjects, and written consent forms were sent to parents. Individualized health screening reports from the project were sent home to all participating children and their parents. Aggregate data are available by county and year through http://www.cardiacwv.org. With this active consent process, response rates for fifth grade participants by year ranged from 25 to 49% since 1998. Our previous work has shown that the differences between participants and nonparticipants were minimal. Nonparticipants were less likely to have a primary care provider and to have health insurance, but there was no difference in BMI or any other demographic variables analyzed in the present study [28].

Measures

The comprehensive risk screening for fifth grade participants included clinical assessment for AN (Presence or Absence on the neck), calculation of BMI from height and weight, resting blood pressure (DBP and SBP), and either a fasting or non-fasting lipid profile (FLP), which included measured TC, HDL-C, LDL-C, and TRIG. In addition, CARDIAC performed fasting insulin and glucose tests only among those children who tested positive for AN due to the relative costs of glucose and insulin tests.

Children’s height (in.) and weight (lb) were measured using the SECA Road Rod stadiometer (78″/200 cm) and the SECA 840 Personal Digital Scale. Students were asked to remove shoes prior to height and weight measurements. These measurements were used to determine the child’s BMI and BMI percentile, calculated using CDC Epi Info version 3.5.4 software [29]. Percent above ideal BMI (BMI % above ideal) was calculated using 100 × log base e (BMI/median BMI) [30] to control for age, gender and height and avoid the ceiling effect seen when using BMI percentile. Insulin resistance was calculated using the HOMA-IR index, which is generally considered a reliable measure of insulin resistance in children [31]. Blood pressure was taken after the child has been resting for 5 min. The first Korotkoff sound was used to record systolic pressure and the fifth Korotkoff sound was used to record diastolic pressure, and appropriate cuff sizing for the child was used.

All cholesterol levels were obtained in either a private area of the school or children were given a voucher to have a lipid profile conducted in a local laboratory (LabCorp) or hospital. Prior to 2002, cholesterol was taken using a finger-stick method. If the values were high, vouchers were given to the student to have a fasting lipid profile (FLP) done at a local lab. Starting in 2002, vouchers and/or lipid profiles were conducted. Thus, the lipid data analyzed in this manuscript do not include any finger-stick obtained cholesterol levels (1998–2002) to avoid potential bias due to different methods of cholesterol measurements. Consistent blood specimens were taken since 2003, and all labs (i.e., hospital or LabCorp) used consistent methods to process the specimens. Some labs calculated LDL-C via the Friedewald index whereas other labs measured LDL-C directly.

Statistical analysis

The mean (SD) and frequency (percentage) were reported for continuous and categorical variables, respectively. Bivariate analysis using t test was conducted to assess the relationship between TRIG/HDL-C and TRIG/LDL-C ratios and participant characteristics (gender, diabetes in child, family history of diabetes, family history of abnormal insulin). t test was also used to assess factors (age, gender, BMI % above ideal, total cholesterol, TRIG, HDL-C, HDL-C, TRIG/HDL-C ratio, TRIG/LDL-C ratio, insulin and glucose levels) associated with the presence of insulin resistance. We analyzed the relationship between gender and insulin resistance using Chi square. Univariate logistic regression with receiver operating characteristic curve analysis was used to assess the performance of TRIG/HDL-C and TRIG/LDL-C ratios in detecting insulin resistance. Insulin resistance (presence or absence of insulin resistance) was the dependent variable. Multiple logistic regression was used to assess the contribution of gender, year of screening, BMI, TRIG/HDL-C and TRIG/LDL-C ratios as predictors of insulin resistance. Only participants with positive AN and complete HOMA-IR values were considered in our analysis. Analysis was performed using SAS JMP pro Version 13 (JMP 1998–2014) [32].

Results

The majority of the study population was white: 82.6% (n = 825), female: 63% (n = 649) and obese: 89.4% (n = 917) (Table 1). The proportion of students with a family history of diabetes and personal history of diabetes was 37.9% (n = 602/969) and 1.6% (n = 10/643), respectively. The mean age of participants was 11.08 ± 0.53 years. The mean HOMA-IR was 7.65 ± 10.89, and the TRIG/HDL-C ratio ranged from 0.22 to 39.9, with a mean of 3.42 ± 2.81. The TRIG/LDL-C ratio ranged from 0.12 to 22.05, with a mean of 1.44 ± 1.26.

Table 1.

Characteristics of 5th graders screened and assessed for acanthosis nigricans between 2007 and 2016 (N = 1030)

Variables	N (%)	Mean (SD)	Range
Age (y)	1029	11.08(0.55)	9.42, 13.61
Fasting glucose(mg/dL)	1030	90.96(11.02)	9,188
Insulin ((μIU/mL)	1030	32.38 (35.44)	1,300
BMI % above ideal	1026	54.15 (19.02)	−33.8, 102.5
95th percentile SBP (mmHg)	1017	121 (2.09)	115,127
95th percentile DBP (mmHg)	1017	80.09(1.37)	76,83
TC (mg/dL)	1022	163.14(29.9)	72, 282
TRIG levels (mg/dL)		163.14(29.90)	72, 282
HDL-C (mg/dL)	1022	42.45 (10.17)	15,95
LDL-C (mg/dL)	1019	96.64 (26.58)	11,228
HOMA-IR	1030	7.65 (10.89)	0.22, 196
TRIG/HDL-C	1022	3.42 (2.81)	0.22, 39.9
TRIG/LDL-C	1019	1.44(1.26)	0.12,22.05
Gender
Male	381 (37.4)
Female	649 (63.6)
Race
Black	70(7)
White	825 (82.6)
Asian	10(1)
Hispanic	20(2)
Bi-racial	66 (6.6)
Other	8 (0.8)
BMI category
Underweight/normal weight	37 (6.6)
Overweight	72 (7.0)
Obese	917(89.4)
Family history of diabetes
Yes	602(62.1)
No	367 (37.9)
Family history of abnormal cholesterol
Yes	361 (41.5)
No	509 (58.5)
Diabetes in child
Yes	10(1.6)
No	633 (98.4)
Asthma
Yes	235 (23.7)
No	756 (76.3)

Analysis was done for a total of n = 1030. Rows that do not add up indicate missing values for the index variable

Table 2 describes bivariate differences among those with mean levels of TRIG/HDL-C and TRIG/LDL-C ratios. Participants with insulin resistance had higher mean levels of TRIG/HDL-C ratio (p < 0.0001) and TRIG/LDL-C levels (p < 0.0001) compared to those without insulin resistance. Similarly, obese participants had higher TRIG/HDL-C ratio (p < 0.0001) and TRIG/LDL-C ratio levels (p < 0.0001) compared to non-obese subjects, respectively. Table 3 examined bivariate differences among those with insulin resistance. Age, total cholesterol and LDL-C had no significant relationship with the presence of insulin resistance. There was a significant association between gender and presence of insulin resistance (χ² (1) = 31.85 p = 0.05). A higher proportion of females (53.3%) had insulin resistance compared to males (25.73%).

Table 2.

Relationship between TRIG/HDL-C, TRIG/LDL-C ratio and participant characteristics (N = 1030)

Participant character- istics	TRIG/HDL-C ratio			TRIG/LDL-C ratio
	N	Mean (SD)	P value	N	Mean (SD)	P value
Gender
Male	380	3.55 (2.63)	0.07	636	1.37(1.39)	0.22
Female	642	3.21 (3.08)		397	1.47(1.17)
Diabetes in child
Yes	10	4.5 (2.5)	0.1	10	1.50	0.67
No	627	3.50 (2.3)		627	1.42
Family history of diabetes
Yes	596	3.22 (2.75)	0.04	10	1.49(0.52)	0.16
No	365	3.56 (2.95)		627	1.42(0.85)
Family history of abnormal lipids
Yes	15	3.17(2.1)	0.8	15	1.21 (0.68)	0.71
No	11	2.96 (2.07)		11	1.31 (0.89)
Asthma
Yes	235	3.47 (3.2)	0.86	234	1.45(1.35)	0.40
No	748	3.42 (2.70)		746	1.47(1.02)
Insulin resistance
Yes	808	3.70 (2.97)	< 0.0001	801	1.53 (1.36)	< 0.0001
No	214	2.38(1.79)		214	1.07 (0.67)
BMI categorya
Obese	981	3.45 (2.59)	< 0.001	978	1.45(1.09)	< 0.0001
Non-obese	37	1.61 (0.95)		37	0.88 (0.32)

^aBMI category—Obese: overweight, obese and morbidly obese children. Non-obese: underweight and normal weight children

Table 3.

Factors associated with insulin resistance in 5th graders with positive acanthosis nigricans

Variables	Insulin resistance	No insulin resistance	P value
	(n = 810)(mean±SD)	(n = 216)(mean±SD)
Age (y)	11.09 ±0.59	11.05 ±0.54	0.42
BMI % above ideal	57.85 ±17.06	40.20 ±19.59	< 0.0001
TC (mg/dL)	162.45 ±29.30	165.72±2.18	0.18
TRIG (mg/dL)	136.86±2.58	100.52 ±5.02	< 0.0001
HDL-C (mg/dL)	41.10±9.22	47.56 ±11.86	< 0.0001
LDL-C	96.12±26.31	98.57 ±27.52	0.24
TRIG/LDL-C ratio	1.5±1.37	1.07 ±0.06	< 0.0001
TRIG/HDL-C ratio	3.70 ±2.97	2.38 ±1.78	< 0.0001
Insulin, (μIU/mL)	38.52±37.51	9.2 ±3.49	< 0.0001
Glucose (mg/dL)	91.87± 11.04	87.55 ±10.28	< 0.0001

Although TRIG/HDL-C ratio and TRIG/LDL-C ratio were significantly associated with the presence of insulin resistance (TRIG/HDL-C: Est. = 0.36, P ≤ 0.0001, AUC = 0.68; TRIG/LDL-C: Est. = 0.87, P ≤ 0.0001, AUC = 0.69), TRIG/HDL-C seemed to have a better performance in detecting the presence of insulin resistance in our study population. For a cut-off between 3.0 and 1.37, the sensitivity and specificity of TRIG/HDL-C to detect insulin resistance in AN-positive children increased (sensitivity 0.48–0.90, specificity 0.32–0.73). In contrast, using the same cut-off, the sensitivity and specificity for TRIG/LDL-C decreased and increased, respectively (sensitivity 0.44–0.05, specificity 0.80–0.98).

Multivariate analysis showed that increased BMI % above ideal (Est. = 0.05, P < 0.0001), gender (Est. = 0.49, P < 0.0001), year of screening (Est. = 0.21, P < 0.0001) and TRIG/HDL-C ratio (β = 0.21, P < 0.0001) were significantly associated with insulin resistance. For each unit increase in BMI % above ideal, the odds of having insulin resistance increased (Est. = 0.05, OR 1.05, 95% CI (OR) 1.04, 1.07). Males had decreased odds of having insulin resistance compared to females (OR 0.37, 95% CI (OR) 0.26, 0.50). The odds of having insulin resistance increased for every one-unit increase in TRIG/HDL-C ratio (Est. = 0.21, OR 1.24, 95% CI 1.11–1.37). Similar results were also seen model 2 with TRIG/LDL-C ratio. Further ROC analysis using BMI % above ideal, year of screening, TRIG/HDL-C and gender had a good performance compared to TRIG/HDL-C or TRIG/LDL-C ratio alone (AUC 0.82 vs 0.69).

Discussion

Currently, there is no clear consensus regarding the definition of insulin resistance in children. This has led to the search for cheaper and easily measured markers of insulin resistance in children. Our study examined the relationship between insulin resistance and simple surrogate markers such as TRIG/HDL-C and TRIG/LDL-C ratios. The main findings included: a 79% prevalence of insulin resistance among children with AN, females having higher TRIG/HDL-C and TRIG/LDL-C ratios, a significant association between TRIG/HDL-C and TRIG/LDL-C ratios and insulin resistance, with TRIG/HDL-C slightly outperforming TRIG/LDL-C ratio. In addition, TRIG/HDL-C and TRIG/LDL-C ratios were higher in participants with insulin resistance compared to those without insulin resistance.

The prevalence of insulin resistance in this study is higher than previously reported by Ice and colleagues in 2009 [5], where the prevalence of insulin resistance in 5th graders with AN was 61%. The large sample size (n = 1030) in the current study versus a smaller sample size (n = 676) in the 2009 study may help explain the discrepancy in the prevalence of insulin resistance. In a small cohort of obese (n = 76) Appalachian children with AN, the prevalence of insulin resistance was 33% [33].

The gender difference in TRIG/HDL-C ratio and TRIG/LDL-C ratios identified in our study is contrary to other reports [4, 34]. According to these studies, no significant gender difference in TRIG/HDL-C ratio in Caucasian children was observed. However, the participants in these studies did not have AN. Although, the current study cannot establish the clinical relevance of TRIG/LDL-C ratio, to the best of our knowledge, no study has actually established a gender difference in TRIG/HDL-C ratio or TRIG/LDL-C ratio in Caucasian children. Since majority of the participants were pre-pubertal, (mean age: 11.08 ± 0.53), the gender difference in TRIG/HDL-C and TRIG/LDL-C ratios observed in our study population should be studied in a different population and age group. In addition, although most females may have entered puberty by 11 years (but not completely), no statistical significant difference was observed [4, 11]. An important relationship demonstrated in this study was the significant association between insulin resistance and TRIG/LDL-C ratio. The relationship between TRIG/HDL-C ratio and insulin resistance has already been demonstrated in adults and children, however, to the best of our knowledge, no study has examined the relationship of TRIG/LDL-C ratio and insulin resistance in children (especially, children with AN and insulin resistant patients present with increased levels of VLDL, TRIG, and TRIG-enriched LDL-C levels. Despite absence of current clinical relevance, we believe that the significant relationship between TRIG/LDL-C ratio and insulin resistance in children with AN observed in our study may help us understand more the relationship between cholesterol metabolism and insulin resistance. Similar AUC values for the known TRIG/HDL-C (AUC: 0.69) ratio and TRIG/LDL-C ratio (AUC: 0.68) demonstrate the need to further examine the role of TRIG/LDL-C ratio in insulin resistance in children without AN, and in children from other races/ethnicities.

This study has limitations. First, the gold standard hyper-insulinemic euglycemic glucose clamp were not used to measure insulin resistance. However, studies have shown good correlations between HOMA-IR and hyper-insulinemic euglycemic glucose clamp (r = 0.88) [10]. HOMA-IR values have actually been shown to have a linear correlation with glucose clamp method in population studies [10]. Second, LDL-C levels was either calculated by Friedewald formula or directly measured LDL-C levels are generally affected when TRIG levels are greater than 200 mg/dl [35]. In this situation, it is preferable to measure LDL-C directly. However, the mean TRIG levels in the study was much less than 200 mg/dl (136.86 ± 2.58 mg/dl). Third, this study focused only on children who had AN. Although Young and colleagues reported increasing insulin resistance with severity of AN, the group without AN had lower HOMA-IR scores (less than the 3.0 cut-off used in the current study) [9]. The relationship between the two proposed ratios and insulin resistance in a sub-group without AN may have provided more information about their performance in detecting insulin resistance.

In conclusion, there was a significant relationship between insulin resistance and TRIG/HDL-C and TRIG/LDL-C ratios in 5th grade students with AN screened by the CARDIAC project. TRIG/HDL-C is a better surrogate marker of insulin resistance in children with AN compared to TRIG/LDL-C ratio. Finally, this study helps us understand the links between LDL-C metabolism and insulin resistance as expressed in this set of Appalachian children.