Greater Improvement in Obesity Among Children With Prediabetes in a Clinical Weight Management Program

Abstract

Background:

There is a range of responses among individuals seen for medical management of their obesity. This retrospective analysis of longitudinal data considers the relationship between identified prediabetes and subsequent weight change among children (8–17 years) in a weight management clinic.

Methods:

Analysis included 733 patients (2687 visits in 2008–2016) with overweight and obesity (but not diabetes) whose referral laboratories included a hemoglobin A1c (HbA1c) within 90 days. Mixed-effects modeling examined the association between baseline prediabetes (serum HbA1c 5.7%–6.4%) and growth curve of percentage of the 95th percentile for BMI (%BMIp95). Random effects (individual growth curves) and fixed effects (prediabetes status, starting age and %BMIp95, sex, race/ethnicity, and linear slope and quadratic term of months since the initial visit) were modeled. Interactions between prediabetes and elapsed time estimated the influence of a recent prediabetic-range HbA1c on weight during the subsequent 12 months.

Results:

Mean %BMIp95 was 125.5% (SD 22.5), corresponding to severe obesity, and 35% had prediabetes. Adjusted monthly decrease in %BMIp95 was stronger for children with prediabetes compared with the peers in this clinic (slope: −0.62, standard error 0.10, p < 0.001).

Conclusion:

There was greater weight improvement among children with prediabetes compared with their peers with normal HbA1c.

Introduction

Weight improvement among children engaged in treatment for overweight and obesity, whether in the context of routine clinical care or a resource-intensive intervention, can be highly variable.^1,2 Even for children engaged in interventions that comprised 26 or more contact hours, as recommended by the US Preventive Services Task Force, some have no change and some even worsen with respect to their weight status.³ Psychosocial factors such as depression^4,5 and age (being older than 5 years) have been shown to be associated with less optimal weight change over time.

Baseline weight status also appears to be important. For example, in a randomized trial of 549 children, for the roughly third of the children in the overall sample who had severe obesity (at or above 120 percentage of the 95th percentile for BMI, or %BMIp95), there was no difference in BMI change whether they were randomized to the intensive intervention⁶ or the primary care control group. This highlights how particularly challenging BMI improvement can be among children with the highest degree of obesity.⁷

Some experts have advocated for a precision medicine approach with pediatric obesity that extends beyond traditional “one-size-fits-all” treatment recommendations to focus on understanding and more nimbly addressing this heterogeneity in response.^2,8 Notably, children and adolescents with obesity demonstrate a wide variation in their manifestation of obesity-related comorbidities (e.g., hypertension, fatty liver disease, and prediabetes). However, while sociodemographic factors are often examined as potential mediators and moderators of lifestyle change and resultant weight change, the presence of comorbidities has not generally been considered.^9,10 Recent estimates from the National Health and Nutrition Examination Survey (NHANES) suggest that 28% of adolescents have prediabetes.⁹ For many, having a history of exposure to gestational diabetes¹⁰ and/or a family history of type 2 diabetes¹¹ incurs extra risk for development of diabetes in the future.

The prevalence of prediabetes increases with a higher degree of obesity severity,^12,13 disproportionately affecting children and adolescents with severe obesity who have greater challenges with weight reduction.⁷ We conducted this longitudinal analysis to evaluate the potential influence of the presence of prediabetes as an identified comorbidity of weight trajectory within the first year after referral to a weight management program.

Objective

Examine whether prediabetes independently influenced weight trajectory in a clinical population of children 8 to 17 years seeking treatment for overweight and obesity.

Methods

Study Setting

The patient population consisted of patients seen in a Northern California pediatric weight management program based at the Children's Hospital and Research Center Oakland from November 2008 to November 2016, with follow-up until closure of that clinic in May 2017. Called Healthy Hearts between 2008 and 2013, and then the Healthy Eating Active Living Clinic afterward, it was a multidisciplinary pediatric weight management program in a very socioeconomically and racially diverse setting. Visits were conducted at four sites in Northern California (Oakland, Walnut Creek, Larkspur, and Fairfield), each located in a different county (Alameda, Contra Costa, Marin, and Solano Counties).

During an initial intake visit, the child/adolescent and their parent/caretaker met with a physician or pediatric nurse practitioner to obtain a medical history, perform a physical examination, and review laboratory tests. In general, although patients had already been notified of their prediabetic range results (fasting glucose or hemoglobin A1c [HbA1c]) by their pediatrician, part of the intake process also included review and interpretation of these test results with the family, and making plans to order follow-up testing. Subsequent visits were scheduled with other providers (dietitian, exercise specialist, or psychologist) who saw the patient alongside a physician or nurse practitioner with intervals of ∼6 to 8 weeks between visits. There were no programmatic differences in management of patients with prediabetes compared with patients without prediabetes.

Staff offered individualized treatment plans that involved the family, and made community referrals to support lifestyle changes. The team had a high degree of racial/ethnic diversity, including three bilingual Spanish-speakers, two of whom were Hispanic and two African American staff, and efforts were made to maximize cultural sensitivity in clinical management. The team worked with subspecialists to coordinate medical management for obesity-related conditions such as hypertension and fatty liver disease.

Study Population

Patients were included in this analysis if they were 8–17 years old at the time of their first clinic visit, had a starting BMI at or above the 85th percentile (BMI z-score ≥1.04), did not have a diagnosis of diabetes at baseline, and had referral laboratory data that included an HbA1c that had been done within 90 days before the first visit.

The clinic database consisted of children whose parents had given informed consent at the time of entry to the clinical program, as per a protocol approved by the Institutional Review Board at Children's Hospital & Research Center Oakland.

Measures

Anthropometric measurements

Anthropometrics at visits within 12 months were evaluated. Weight and height were measured by clinic staff (to the nearest 0.1 kg and 0.1 cm, respectively) using a digital electronic scale and a wall-mounted stadiometer. The BMI z-score and the %BMIp95 were calculated using SAS codes provided by the CDC.¹⁴ The 95th percentile for BMI (BMIp95) is the threshold above which a child is considered to have obesity, and %BMIp95 further quantifies BMI in relationship to this threshold, and is a particularly useful metric in populations with severe obesity.^15,16

A child with overweight (e.g., BMI greater than 85th percentile but below 95th percentile) would have a %BMIp95 below 100%, and a child whose BMI is at the 95th percentile would have a %BMIp95 of exactly 100%. A %BMIp95 of 120% corresponds to severe obesity (Class II), and %BMIp95 at or above 140% indicates the highest degree of severe obesity (Class III), corresponding to an adult BMI at or above 40.0 kg/m².

Demographic characteristics

Demographic information was recorded at the intake visit, and self-reported by the caregivers. Based on responses for race/ethnicity (“check all”), race/ethnicity was classified into five nonoverlapping categories: non-Hispanic White, African American, Asian/Pacific Islander, Hispanic, and mixed/other race (reporting multiple racial/ethnic identities).

Laboratory data

Prediabetes has been defined by the American Diabetes Association as being an individual without a diagnosis of diabetes who has impaired fasting glucose (IFG) (fasting glucose between 100 and 125 mg/dL), impaired glucose tolerance (glucose 140–199 mg/dL 2 hours after a glucose challenge), and—since 2009—having a serum HbA1c measurement between 5.7% and 6.4%.¹⁷

As part of the process of referral to this weight management clinic, patients had laboratory testing ordered by their primary care pediatrician. Patients were included if they had an HbA1c within 90 days of the intake visit. Although serum glucose was also requested, this analysis is limited to prediabetes defined by abnormal HbA1c because of the challenge with assuring that documented glucose values were fasting. Oral glucose tolerance tests were not generally ordered by referring pediatricians.

Statistical Analyses

The t test or the chi-square test was used for descriptive statistics. To compare the trajectory of the change in outcome (%BMIp95) by baseline prediabetes status, we performed unadjusted and adjusted models. First, to illustrate change in %BMIp95 (Fig. 1), we used a polynomial regression model (without adjusting for covariates) of repeated measures to fit the change in %BMIp95 over time (months from the first visit) for groups with and without baseline prediabetes, modeling the slope of both linear and quadratic terms¹⁸ (months and months²), as is commonly done for growth modeling in children. Second, for the unadjusted and adjusted individual growth analysis, we used a mixed-model growth curve wth the main effects of prediabetes status and follow-up time, and their interactions (e.g., linear and quadratic terms) with and without adjustment for starting value of %BMIp95, starting age, sex, and race/ethnicity.

Figure 1.

Unadjusted overall trajectory changes in %BMIp95 among patients with prediabetes (solid line) and without prediabetes (dashed line) during the first 12 months after initial visit. Both groups show decreasing %BMIp95 over time. Here the trajectory for patients with prediabetes (hemoglobin A1c 5.7%–6.4%) shows notably steeper decrease (linear) with gradual rebound (quadratic growth). Confidence intervals are shown for both growth curves with dotted lines. %BMIp95, percentage of the 95th percentile for BMI.

Lastly, to be able to offer comparison of our findings (which examined monthly change) to published data from a large database of similar patients (which reported follow-up after 10–12 months of participation in weight management),¹⁹ we also estimated the model-based changes in %BMIp95 at 12 months. The analyses used all available data, and missing data were handled directly through maximum likelihood estimation via mixed modeling.

The sample comprised many patients with severe obesity, for whom BMIp95 is a preferred metric. However, for a sensitivity analysis, we also conducted all analyses with BMI z-score. Because there can be a transient decrease in insulin sensitivity and HbA1c during puberty, and because height gain during the pubertal growth spurt could influence BMI trajectory during the peripubertal period, we also conducted a sensitivity analysis limiting to older children (14 years and older).^20–22

All analyses were conducted using SAS^®, version 9.4 (SAS Institute, Inc., Cary, NC). Statistical significance was defined as p < 0.05 (two sided).

Results

Study Population

We identified 733 patients without diabetes who met age, BMI, and baseline laboratory data criteria. Of these, 149 did not have a subsequent visit in the 12-month follow-up period. Demographic characteristics did not differ between children with and without prediabetes in the groups that were excluded because of lack of follow-up visits.

Baseline Demographic Characteristics

Mean age at baseline was 144.7 (SD 28.6) months (12 years) (Table 1). The mean BMI z-score was 2.2 (SD 0.4) and the mean %BMIp95 was 125% (SD 22.5%), indicating that this was largely a population with severe obesity. (Only 8% of participants had a starting %BMIp95 below 100%.) Roughly a third (35%) of the sample had referral laboratories indicating prediabetes. Starting %BMIp95 was higher for children with prediabetes (129%, SD 26%) compared with peers without prediabetes (123.4%, SD 20.1%), p < 0.001. More than half (52%) of the patients were Hispanic.

Table 1.

Sociodemographic Characteristics of the Sample (N = 733)

Characteristic	Overall	Prediabetes	Without prediabetes	p
	(N = 733)	(n = 255)	(n = 478)
Starting age months	144.7 ± 28.6	146.7 ± 27.3	143.7 ± 29.2	0.18
Female, n (%)	379 (51.7)	142 (55.7)	237 (50.0)	0.12
Race/ethnicity, n (%)a, n = 725				<0.001
Non-Hispanic White	101 (13.9)	53 (21.0)	48 (10.2)
African American	134 (18.5)	34 (13.5)	100 (21.1)
Asian/Pacific Islander	52 (7.2)	23 (9.1)	29 (6.1)
Hispanic	380 (52.4)	122 (48.4)	258 (54.6)
Mixed race	58 (8.0)	20 (8.0)	38 (8.0)
Starting BMI Z score	2.2 ± 0.4	2.2 ± 0.4	2.2 ± 0.4	0.02
Starting %BMIp95	125.5 ± 22.5	129.5 ± 26.0	123.4 ± 20.1	0.001
No. of visits within 12 months				0.40
1	149 (20.3)	41 (16.1)	108 (22.6)
2	137 (18.7)	48 (18.8)	89 (18.6)
3	97 (13.2)	33 (12.9)	64 (13.4)
4	89 (12.1)	34 (13.3)	55 (11.5)
5+	261 (35.6)	99 (38.8)	162 (33.9)

^a Race/ethnicity data were available for n = 725 participants, of which 252 and 473 were children with and without prediabetes, respectively.

%BMIp95, percentage of the 95th percentile for BMI.

Characteristics of Follow-Up Data

Among the patients with any follow-up data (n = 584), the number of visits did not differ between the two comparison groups (p = 0.76). Forty-five percent of patients had more than four follow-up visits within a 12-month window.

Trajectory of %BMIp95

Children with prediabetes compared with peers without prediabetes

When change in %BMIp95 was examined by prediabetes status, the linear slopes differed significantly in unadjusted and adjusted models. Quadratic slopes also differed significantly in both the unadjusted models and adjusted models; the positive quadratic terms indicate that the differences between those with and without prediabetes narrowed over time.

Children with prediabetes had significantly more reduction in %BMIp95 over the 12-month follow-up period than their peers without prediabetes, as demonstrated in Figure 1 by the separated trajectories and 95% confidence intervals between the two groups. The unadjusted analysis for individual growth curve trajectories also indicated greater reduction in linear slope for children with prediabetes (difference in the linear coefficients of slope [standard error, SE]: −0.64 [0.10], p < 0.001). Adjusting for other covariates (e.g., starting %BMIp95, starting age, sex, and race/ethnicity), the greater reduction in slope seen in children with prediabetes compared with their peers remained significant (difference in coefficients of slope [SE]: −0.62 [0.10], p < 0.001).

Other covariates in the adjusted model

Higher starting %BMIp95 was associated with a greater decrease (more improvement) in %BMIp95. Compared with non-Hispanic White children, Hispanic, African American, Asian/Pacific Islander, and mixed race children had more reduction (more improvement) in %BMIp95 over time %BMIp95 (difference in the coefficients: −2.21 [0.35], p < 0.001; −1.47 [0.24], p = 0.007; −1.21 [0.28], p < 0.001; and −2.82 [0.33], p < 0.001], respectively). Females also had more reduction (more improvement) than their male counterparts (−0.44 [0.15], p = 0.004) (Table 2).

Table 2.

The Coefficient (Standard Error) from the Mixed Model Growth Curves Fitting for the Outcome Variable % BMI Relative to the 95th Percentile

Variable	Unadjusted modela		Adjusted modelb
	Coefficient (SE)	p	Coefficient (SE)	p
Months from the first visit
Slope
No prediabetes group	−0.55 (0.08)	<0.001	−0.56 (0.07)	<0.001
With versus without prediabetes group	−0.64 (0.10)	<0.001	−0.62 (0.10)	<0.001
Quadratic term
No prediabetes group	0.03 (0.01)	0.01	0.03 (0.01)	0.01
With versus without prediabetes group	0.05 (0.02)	0.01	0.05 (0.02)	0.005
Covariate
Starting %BMIp95			−0.05 (0.003)	<0.001
Starting age months			0.004 (0.003)	0.18
Sex
Female			−0.44 (0.15)	0.004
Male			Ref
Race
Non-Hispanic White			Ref
African American			−1.47 (0.24)	0.007
Asian/Pacific Islander			−1.21 (0.28)	<0.001
Hispanic			−2.21 (0.35)	<0.001
Mixed race			−2.82 (0.33)	<0.001

^a Mixed model with repeated measures included the fixed effects of the slope and quadratic term of the months from the first visit and their interactions with prediabetes status accounting for the repeated measures for each individual. The random effects included the intercept, slope, and quadratic terms of the months from the first visit and their interactions with prediabetes status.

^b The unadjusted model+adjusting for covariates (starting value of %BMIp95, starting age, sex, and race/ethnicity).

SE, standard error.

Change in %BMIp95 at 12 months

For children with prediabetes, holding all covariates at their reference values, there was a borderline significant decrease in %BMIp95 (−2.63 [1.43], p = 0.07). For normoglycemic peers, the magnitude of change at 12 months was slightly smaller but significant (−2.43 [0.91], p = 0.008).

Sensitivity Analysis

Using BMI z-score instead of BMIp95 yielded similar results to our main findings. There was a significant slope difference for the main set (−0.01 [0.001], p = 0.001). Using z-score, the quadratic growth term was not significant (0.0003 [0.0004], p = 0.49).

When we limited analysis to the 252 children who were 14 years of age or older at baseline, the findings were also significant, and in fact of greater magnitude. These adolescents with prediabetes had a greater linear slope change than normoglycemic peers (difference in slope [SE]: %BMIp95 − 0.95 [0.22], p < 0.001). Quadratic terms remained significant for 12-month follow-up intervals (0.08 [0.04], p = 0.03).

Discussion

This longitudinal analysis of treatment-seeking children with obesity showed that compared with peers with normal HbA1c, children in a weight management program with prediabetes had greater monthly reduction in their %BMIp95, indicating more optimal weight trajectories.

This analysis adds value as the first study to compare weight trajectories of children with and without prediabetes being seen in real-world clinical care settings for weight management.

One plausible mechanism could be that recent identification of prediabetes magnifies motivation for improving health-related behaviors to mitigate risk of progression to diabetes. In the Pathobiology of Prediabetes in a Biracial Cohort, adults were identified on the basis of having had a parent with type 2 diabetes, and were followed over time in a natural history study. This study found that individuals who developed prediabetes during study observation had significant reductions in their glucose levels and notably improved diet and exercise habits when compared with the normoglycemic peers in the study.¹¹ This suggests that awareness of prediabetes may stimulate greater engagement in diabetes risk-reduction behaviors. Other researchers have examined this question about the relationship between awareness of prediabetes status and lifestyle evaluating cross-sectional data from the NHANES.

Of the 2694 adults in two cycles of the NHANES identified with prediabetes, only 11.8% appeared to have already been aware that they had prediabetes, based on their self-report on the questionnaire. The rest were identified as having prediabetes with laboratory data. Notably, being “prediabetes-aware” was associated with higher odds of BMI-appropriate weight-related behavior (e.g., reported intent to lose weight and BMI was >25), reporting that they engaged in at least 150 minutes of weekly physical activity, and reporting that they had 7% weight loss in the past year.²³ Furthermore, semistructured interviews with people having recently diagnosed prediabetes highlight descriptions of shock and a “strong determination not to develop diabetes.”²⁴

In 1975, Rogers proposed the Protection Motivation Theory, describing three crucial components of how a “fear appeal” leads to attitude change: magnitude of the noxiousness of an event, probability of occurrence, and efficacy of a protective response.²⁵ Future extension of the findings from this analysis could include deeper investigation about the “fear appraisal” of a diagnosis of prediabetes, patient estimation about the likelihood that they might progress to having type 2 diabetes, and supporting patient understanding about the efficacy of lifestyle change to reverse their prediabetes.

One key difference to consider is that in this analysis, the potential motivating health consequence of concern is a health outcome in a child/adolescent, rather than an adult. Notably, adolescence as a developmental stage is notably characterized as one which brings a sense of invincibility, or “it can't happen to me.”^26,27 Based on the data at hand, it is impossible to determine whether and how much the identification of prediabetes was a motivating factor for these adolescent patients, as well as the degree to which it rallied concern from their parents.

The findings on %BMIp95 change in this analysis offer a novel glimpse at heterogeneity of weight change in a study population with overall findings that are comparable with what has been shown so far in the literature. The Pediatric Obesity Weight Evaluation Registry is a database that includes data from children (2–18 years) seen in multidisciplinary pediatric obesity programs across the country.¹⁹ Pooled data from 6454 children (median %BMIp95 of 132%) showed a median change in %BMIp95 at 10–12 months of follow-up of −2.86 (IQR, −8.7 to 1.9), which is comparable with what was shown in this population. Also, while children with prediabetes have more %BMIp95 decrease compared with their peers, the significant positive quadratic term for growth is an indicator that there is “closing of the gap” over time.

It is likely that even despite the contribution that a prediabetes diagnosis might bring an initial weight change, there is also some eventual relaxation of lifestyle factors and motivation, in keeping with the well-documented phenomenon of weight regain over time.²⁸

There are limitations and strengths to note with these findings. First, without corresponding data to measure lifestyle behaviors (e.g., dietary intake, physical activity), it is not possible to specify the pathway underlying this differential trajectory. Furthermore, there were no available data regarding medications. However, because metformin (sometimes used for patients with prediabetes and which can lead to a small amount of weight loss²⁹) was rarely prescribed by these clinic providers during the sample period (less than 5% of eligible patients), and weight loss medications (e.g., phentermine and topiramate) were not being prescribed by these providers at the time, treatment bias from medication effect would have been small. Also, this analysis only considers HbA1c (and not serum glucose) to classify patients as having prediabetes. Potentially, some children who had been given a diagnosis of prediabetes by their physician based on IFG would have been misclassified in this analysis as normoglycemic because of their normal HbA1c. However, this misclassification would have likely biased findings toward the null. Despite these limitations, this analysis adds value as the first study to compare weight trajectories of children with and without prediabetes being seen in real-world clinical care settings for weight management. The very notable racial-ethnic diversity represented in this patient population that was drawn from sites in four different counties was also a particular strength. Future research could include exploration of weight trajectory given identification of other obesity-related comorbidities such as hypertension and hyperlipidemia

Conclusions

This analysis evaluates the independent influence of identified prediabetes on the trajectory of weight status (measured as %BMIp95) among children seen in a pediatric weight management program over an eight-year period. The growth curves for children with prediabetes had more decrease in weight status over time compared with peers with normal HbA1c. More research is needed to fully understand the impact of identification of prediabetes in childhood on weight change among children with obesity.

Impact Statement

This longitudinal analysis of children referred to a weight management clinic with identified prediabetes found a steeper decrease in obesity (%BMIp95) compared with their peers who did not have prediabetes. Identification of prediabetes may be an influential factor in weight management.

Authors' Contributions

J.M.T.: Conceptualization, investigation, methodology, and writing—original draft. L.X.: Formal analysis and writing—review and editing. C.A.C.: Writing—review and editing. L.T.-D.: Investigation and writing—review and editing. S.S.: Writing—review and editing. L.G.R.: Resources and writing—review and editing.

Funding Information

Data analysis for this article was supported by 1R01MD016738 National Institute on Minority Health and Health Disparities (NIMHD).

Author Disclosure Statement

No competing financial interests exist.