Appraisal of Clinical Care Practices for Child Obesity Treatment. Part II: Comorbidities

Obesity is a common concern in pediatric practice. In caring for patients with obesity or patients who may be at risk for developing obesity, clinicians have many unanswered questions. Examples of these questions include: What is the best way to identify excess adiposity, and does the identification of obesity provide opportunities for treatment? If so, what evidence-based interventions for obesity treatment, delivered at least in part by clinicians in office-based settings, are most effective? Among children and adolescents identified as having obesity, does screening for comorbidities result in improved health outcomes?

Many previous studies, most notably conducted by the US Preventive Services Task Force, have synthesized research regarding the treatment of obesity.¹  Unfortunately, some important gaps remain unfilled. The US Preventive Services Task Force recommendation was that obesity treatment should include at least 26 hours of contact, including clinical care and other behavioral intervention (eg, guided physical activity). However, subsequent studies have failed to demonstrate a consistent hours-based dose-response. In addition, feasibility studies have clearly shown how unrealistic it is for primary care or tertiary care providers to deliver this many hours of treatment in real-world, clinical settings.²  Additional information is needed about resources or partnerships that help reach that contact hour goal, the essential components delivered during these contact hours, the period of time over which this care is delivered, and information about lower-intensity strategies with some effectiveness.

Of particular concern for primary care pediatricians is the need to understand how to approach recommendations for screening comorbidities in their patients with obesity. Although previous recommendations have supported screening for common comorbidities, such as dyslipidemia and diabetes, there has been conflicting evidence regarding timing and effectiveness of screening. Additional data are now available that provide clinicians and researchers with information about comorbidity prevalence and severity by obesity class. The intent is to help the clinician screen for comorbidities when there is a high likelihood of detecting an abnormality and when detection of that abnormality leads to treatment options that can improve child health. Obesity classifications, including a more granular categorization of obesity as classes I through III, might assist clinicians in determining for whom screening would be most useful rather than viewing screening as a homogeneous approach for anyone whose BMI is >95th percentile.

This review was designed to answer 2 overarching key questions: (KQ1) “What are effective clinic-based treatments for obesity?” and (KQ2) “What is the risk of comorbidities among children with obesity?” We developed this focus based on the needs of clinicians and the evidence required to inform the future development of clinical practice guidelines. This review will not attempt to quantify the magnitude of the effect of obesity on child or adult outcomes. It will also not attempt to address treatment strategies for comorbidities (eg, hypertension), as other guidelines and reviews are available to guide such treatment.

Clinicians are a regular source of trusted information for parents, including issues related to nutrition and activity, which are key components of obesity prevention and treatment. Clinicians need to know what strategies have high-quality evidence for effectiveness in preventing and treating obesity. Additionally, physicians need guidance on which treatments are effective for their patient population and how to use available resources. The full results of KQ1 are reported in an accompanying technical report.³ 

Previous recommendations have included assessments of comorbidities, including hypertension, dyslipidemia, glucose, and others. It is not clear whether these assessments identify important health conditions or lead to improved treatment strategies. Additionally, it is not clear whether conducting these assessments would result in an adverse patient outcomes, such as further investigation for false-positive screening results. We will examine specific conditions previously recommended or that would reasonably require screening, as identified by the authors: dyslipidemia, hypertension, diabetes, liver function, depression, sleep apnea, and asthma. This is not intended to be a comprehensive list of all conditions comorbid with obesity but represents those most common and for which screening is potentially helpful.

We searched Pubmed and CENTRAL (for trials), completing the final search on April 6, 2018. An additional search was conducted to update the review, covering the time period April 7, 2018, through February 15, 2020. We combined the searches for both key questions because of significant overlap and to more efficiently review studies. Because our focus was on interventions that are relevant to primary care, we did not search other databases, such as ERIC or PsycInfo.

The complete search strategies are included in Appendix 2. Briefly, we searched for studies of children or adolescents, with a focus on overweight, obesity, or weight status; involving clinicians, health care, or other treatment or screening (KQ1); and examining common comorbidities (KQ2). For both questions, we limited only using key words, not filters, to ensure we included the newest studies that were not yet fully indexed. No date limits were placed on searches. In practice, this meant we reviewed studies from 1950 to 2020, although <2% were published before 1980.

The complete inclusion criteria are included in Appendix 3.

All studies were required to include children ages 2 to 18 years, although studies could also include young adults up to age 25 if stratified from older adult participants, as long as children under 18 were also included. Children could have other conditions (eg, asthma) as long as they were not known to cause obesity, such as Prader-Willi syndrome, obesogenic medication (eg, antipsychotics), or known genetic mutations (eg, MC4R) associated with obesity. All studies had to originate from the Organization for Economic Cooperation and Development member countries and had to be available in English.

We included studies with a primary aim of comparing comorbidities among those with and without obesity or by severity of obesity. Obesity and the comorbidity had to be measured contemporaneously to reflect the practice of clinical screening. Obesity had to be categorized using a BMI-based measure into accepted categories (ie, healthy weight, overweight, class I obesity, class II obesity, class III obesity).

These categories could be based on percentiles or z-scores and could use the distributions relevant to the studied population (eg, World Health Organization [WHO] or the US Centers for Disease Control and Prevention [CDC]). Comorbidities had to include 1 or more of: lipids, blood pressure, liver function, glucose metabolism, obstructive sleep apnea, asthma, or depression.

See the other technical report for a detailed description of KQ1 inclusion criteria.³ 

We used Covidence (Melbourne, Australia) to manage the review process. Covidence is a program for online collaboration and management of systematic reviews. All abstracts were reviewed by 2 independent reviewers for inclusion in full-text review. Articles were reviewed by 2 reviewers, with conflicts discussed and resolved. Articles excluded at this stage were assigned an exclusion reason, with a hierarchy as shown in Appendix 4.

All articles deemed relevant for full text inclusion were categorized into different data extraction strategies. We did not include a specific quality assessment for the comorbidity studies.

All studies were extracted by 2 reviewers. Extraction of these studies included reporting prevalence of comorbidities or mean values of laboratory parameters by weight classification. We included healthy weight, overweight, class I obesity, class II obesity, and class III obesity. However, because all classes of obesity severity are not always reported, these classes may include higher groups. For example, reporting of ≥95th percentile would only be considered class I obesity, although children at higher levels may be included. (See other technical report for detailed description of KQ1 extraction procedures.)

Our primary method of data synthesis is narrative. To allow broad inclusion, we did not limit to specific designs or measures that would facilitate meta-analysis. We report on studies in each group, based on their type and design, and we report findings for outcomes other than BMI.

A total of 15 988 studies were screened in the title and abstract stage. Of these, 1642 were given a full-text review. Excluded studies (n = 1260) were most commonly not original research, did not compare comorbidities by obesity (KQ2), or were not health-care system based (KQ1). See Fig 1 for the complete PRISMA diagram. Of the 382 studies included, 215 were intervention studies and 167 were comorbidity studies. This paper focuses on the 167 comorbidity studies.

A total of 39 studies examined the prevalence of abnormal high-density lipoprotein (HDL),^4–42  whereas 49 provided mean values for HDL.^{5–8,10,13,18,22,24,32,33,35,36,40–74} Table 1 reports the prevalence of abnormal HDL. Different countries report significantly different prevalence of abnormal HDL, with Korea having the highest prevalence^18,30  and Japan the lowest.⁴²  The majority of the 39 reported studies reporting the prevalence of abnormal HDL were conducted in the United States (24 of 39). Abnormal HDL was defined variably as <35 mg/dL, <40 mg/dL, and <50 mg/dL or <1.0 mmol/L. The most consistent findings were seen when using the definition of <40 mg/dL and when larger sample sizes were included. There was consistency of an inverse dose-response relationship, with increasing weight category associated with lower HDL. Few studies provided detailed information by obesity class, so less could be concluded when examining the prevalence of abnormal HDL within samples of increasing severity of obesity status. In general, overall prevalence of abnormal HDL increases from about 10% to 40% when children’s weight category was healthy weight versus obesity. The prevalence varied by age, with younger ages associated with lower prevalence of abnormal HDL. For example, in a study of 9- to 13-year-olds, those who had healthy weight had a prevalence of abnormal HDL of 10.2%, whereas those with obesity had a prevalence of abnormal HDL of 32.5%.¹⁵  In a study of 14- to 18-year-olds, those who had healthy weight had a prevalence of abnormal HDL of 13.2% and those with obesity had a prevalence of abnormal HDL of 38.9%.⁴⁰  When studies report larger age ranges, it is difficult to see these distinctions, and the mean prevalence might be obfuscating the differences in prevalence at the younger versus older ages. A few studies stratified their findings by biological sex. In 2 US-based studies, there appears to be a higher prevalence of abnormal HDL in female children of both healthy weight and overweight, but the prevalence is similar regardless of sex once children are categorized as obese.^12,20  Studies conducted in other countries also report differences by biological sex, but not always in the same direction or to the same degree.^9,19,35,42  Caution should be used in interpreting these results when small sample sizes were used.

Table 2 reports the mean HDL values. Mean HDL values corroborate the findings regarding the prevalence of abnormal HDL, highlighting that age, sex, and country affect the findings of mean HDL values. Also apparent is the importance of sample size to lead to a stable mean value. Several of these studies reported mean values for large age ranges. In almost all of these studies, mean HDL decreases as weight category increases, validating the association between the 2.

A total of 26 studies examined the prevalence of abnormal low-density lipoprotein (LDL),^{4–10,12,13,15,16, 18,19,21–24,26,28,31,34–38,41}  whereas 41 provided mean values for LDL.^{5–8, 10,13,18,22,24,33,35,36,40,41,43–46,50–52, 54–67,69,70,72–75} Table 3 reports the prevalence of abnormal LDL. Approximately half (n = 13) of the studies evaluated children 9 years or older, a time point associated with physiologic increases in LDL cholesterol.⁷⁶  The remaining studies included children as young as 3 and up to 19 years of age. Sample size varied from 101 to 29 286; 13 of 23 studies reported sample sizes of 1000 or greater. One challenge in interpreting these data are the variation in definition of and reported units for abnormal LDL. Authors defined abnormal LDL as >110 mg/dL or >2.6 mmol/L through >130 mg/dL or >3.4 mmol/L. In 1 instance, authors used >75th percentile of National Cholesterol Education Program (NCEP) standards.²⁴  In nearly all the studies, abnormal LDL was more prevalent in children with increasing BMI, and when comparing healthy weight with obesity, this difference consistently achieved statistical significance. The majority of studies did not include a significant number of children in each obesity classification; therefore, it is difficult to conclude whether abnormal LDL is more common by obesity classification. Among the 3 studies that reported male and female LDL separately, there was not a significant difference at any weight classification.^5,9,12,19,35  Similarly to the LDL prevalence studies, the most evidence for mean LDL in populations includes children of school age and older (Table 4). Only 1 of the identified studies exclusively included children younger than 5 years⁴⁶ ; therefore, it is difficult to draw conclusions in this younger age group. Sample size of the reported studies ranged from 41 to 2244. Several, but not all, studies reported male and female LDL levels separately. Mean LDL was reported in some cases in mg/dL and in other cases as mmol/L. Across all studies, mean LDL tended to increase with increasing BMI; however, only the difference between healthy weight and obese consistently achieved statistical significance. In 1 Korean study that evaluated mean LDL in 1998 and again in 2001, secular increases in mean LDL were also observed.¹⁷  Although in some cases, females have higher mean LDL than males at matched age and BMI, this difference was inconsistent and did not achieve statistical significance. However, the difference between mean LDL when comparing healthy weight and obesity was more pronounced in males than females. It is interesting to note that in all studies, even in the highest BMI subcategories, mean LDL values did not exceed commonly accepted definitions for normal.

A total of 38 studies examined the prevalence of abnormal triglycerides (TG),^{4–26,28–32,34–42}  whereas 48 provided mean values for TG.^{5–8,10,13,18,22,24,32,35,36,40–60,62–75,77} Table 5 reports the prevalence of abnormal TG. About half of the 38 reported studies were conducted in the United States (20 of 38). Country comparisons are not possible given the variety of cutoff values employed. However, there is consistency of a dose-response relationship with increasing weight category associated with higher TG prevalence in most settings studied. Few studies provide detailed information broken down by obesity class, so less can be concluded when examining the prevalence of abnormal TG and increasing severity of obesity status. When studies report larger age ranges, it is difficult to see these distinctions, and the mean prevalence might be masking any potential differences in prevalence at the younger versus older ages. A few studies stratified their findings by gender, but the pattern of high TG prevalence was not always in the same direction or to the same degree. Caution should be used in interpreting these results when small sample sizes were used.