Pharmacotherapy for the Treatment of Overweight and Obesity in Children, Adolescents, and Young Adults in a Large Health System in the US

Lifestyle modifications focused on diet, physical activity, and behavior have a modest impact on weight reduction in children, adolescents, and young adults (YA) with overweight and obesity. Several anti-obesity medications (AOMs) have been approved by the Food and Drug Administration (FDA) for use among adult patients with a body mass index (BMI) ≥27 kg/m2 and at least one obesity-related illness. However, only two FDA-approved AOMs are available for use in children and adolescents, which leads to the frequent off-label use of adult AOMs among this population. We sought to investigate current prescribing patterns of AOMs from school age through to young adulthood in a large unified health system. Using a centralized clinical data registry containing the health data of ~6.5 million patients, individuals aged 5–25 years old with overweight and obesity who were taking one of eight commonly prescribed AOMs from 2009 to 2018 were extracted. A total of 1,720 patients were identified, representing 2,210 medication prescribing instances. The cohort was further stratified as children (5–12 years old), adolescents (13–18 years old), and YA (19–25 years old). The mean BMI at the time of medication initiation was 34.0, 39.1, and 39.6 kg/m2, respectively, which corresponded to a BMI z-score (BMIz) of 2.4 and 2.3 for children and adolescents, respectively. Metformin was the most commonly prescribed medication across all ages, including off-label use for weight-loss among children and adolescents. The most commonly off-label prescribed AOM among YA was topiramate. Multivariable analyses demonstrated phentermine was the most effective AOM, with a 1.54% total body weight among YA (p = 0.05) and a 0.12 decrease in BMIz among adolescents (p = 0.003) greater final weight loss when compared to the respective overall frequency-weighted means. Our study demonstrates a statistically significant weight loss among adolescents and young adults on select pharmacotherapy. The small magnitude of this effect should be interpreted carefully, as it is likely an underestimate in the absence of a true control group. Pharmacotherapy should therefore be considered in conjunction with other multimodal therapies such as lifestyle modification and metabolic and bariatric surgery when treating overweight and obesity.

Lifestyle modifications focused on diet, physical activity, and behavior have a modest impact on weight reduction in children, adolescents, and young adults (YA) with overweight and obesity. Several anti-obesity medications (AOMs) have been approved by the Food and Drug Administration (FDA) for use among adult patients with a body mass index (BMI) ≥27 kg/m 2 and at least one obesity-related illness. However, only two FDA-approved AOMs are available for use in children and adolescents, which leads to the frequent off-label use of adult AOMs among this population. We sought to investigate current prescribing patterns of AOMs from school age through to young adulthood in a large unified health system. Using a centralized clinical data registry containing the health data of ∼6.5 million patients, individuals aged 5-25 years old with overweight and obesity who were taking one of eight commonly prescribed AOMs from 2009 to 2018 were extracted. A total of 1,720 patients were identified, representing 2,210 medication prescribing instances. The cohort was further stratified as children (5)(6)(7)(8)(9)(10)(11)(12) years old), adolescents (13-18 years old), and YA (19-25 years old). The mean BMI at the time of medication initiation was 34.0, 39.1, and 39.6 kg/m 2 , respectively, which corresponded to a BMI z-score (BMIz) of 2.4 and 2.3 for children and adolescents, respectively. Metformin was the most commonly prescribed medication across all ages, including off-label use for weight-loss among children and adolescents. The most commonly off-label prescribed AOM among YA was topiramate. Multivariable analyses demonstrated phentermine was the most effective AOM, with a 1.54% total body weight among YA (p = 0.05) and a 0.12 decrease in BMIz among adolescents (p = 0.003) greater final weight loss when compared to the respective overall frequency-weighted means. Our study demonstrates a statistically significant weight loss among adolescents and young adults on select pharmacotherapy. The small magnitude of this effect should be interpreted carefully, as it is likely an underestimate in the absence of a true control

INTRODUCTION
The most recent analysis of obesity prevalence using the National Health and Nutrition Examination Survey (NHANES) database shows that 1 in 5 children in the United States have obesity (1). Children in high-risk groups such as those with a genetic predisposition to obesity (2), and those with poor quality early lifestyle and dietary behaviors (3) are likely to develop childhood obesity which often propagates into adulthood along with obesity-related illnesses (1,(4)(5)(6)(7)(8)(9).
Current treatment modalities for childhood obesity are multidisciplinary in nature with a significant preference toward lifestyle modifications that target dietary and behavioral change as the foundation for treatment in most children, adolescents and young adults who present for care. However, a 2017 Cochrane Review found low quality evidence that lifestyle modification focused on diet, physical activity, and other behaviors reduces BMI in adolescents (10). According to this review, the mean change in BMI was −1.18 kg/m 2 with a weight loss (WL) of 3.67 kg (95% CI −5.21 to −2.13) across 28 randomized-clinical trials, encompassing 2,774 participants ages 12-17. The rise of obesity over the past 30 years, especially with high rates of severe obesity in children living in non-metropolitan areas (11), calls for more aggressive combination therapies that include behavioral modification, medications, and surgical interventions (12,13). According to the most recently available data from the NHANES, the prevalence of severe obesity from 2015 to 2016 was 1.9% among children and adolescents 2 to 19 years old, with 4.5% of adolescents age 16-19 years old affected by severe obesity (1). While strides have been made toward encouraging the utilization of metabolic and bariatric surgery (MBS) for pediatric patients with severe obesity, with definitive guidance from organizations such as the American Academy of Pediatrics (AAP) and the American Society of Metabolic and Bariatric Surgery (ASMBS) (14)(15)(16)(17), there remains a gap in care for patients who have been refractory to lifestyle modifications but do not meet criteria for MBS.
The current use of pharmacotherapy for the treatment of obesity in the pediatric population is limited (18). Compared to prescribing patterns for the treatment of other pediatric chronic diseases such as type 2 diabetes mellitus (T2DM), anti-obesity medications (AOMs) are disproportionately underutilized in relation to the disease burden (19), likely due to the lack of current guidelines addressing their use. Notably, the current Endocrine Society practice guidelines recommend that AOM use should be confined to clinical trials for pediatric patients with obesity (20). Furthermore, only two AOMs are currently approved by the Food and Drug Administration (FDA) for management of obesity in the pediatric population: orlistat for patients ≥12 years and phentermine in those >16 years (21,22). Since most medications with the potential benefit of WL are used either for a separate indication or used off-label for the treatment of obesity, there is limited reliable data available on the utilization of these medications and weight outcomes (23). The purpose of this study was to ascertain current prescribing practices of these AOMs among the pediatric population in a large unified health system, and second, to assess the effect of these medications on weight status.

Data Source
The centralized clinical data registry of a large unified health care system, consisting of two large academic medical centers and three community teaching hospitals was used to identify the study cohort. This clinical data registry contains electronic health records (EHR) data spanning over 30 years for ∼6.5 million patients, and allows for research cohort identification using a combination of user-defined characteristics such as patient demographics, diagnoses, medications, etc. Institutional Review Board (IRB) approval was obtained prior to extraction of any patient health data.
The date of medication initiation was available as a structured data field directly from the clinical data registry. On the other hand, the date of medication discontinuation, though also available as a structured data field, was deemed unreliable due to known discrepancies between actual medication discontinuation and documentation in patients' EHR. Instead, the date of medication discontinuation was determined by direct examination of provider clinic notes, which ensured its accuracy.
For patients who underwent MBS while on a study medication, the date of discontinuation was set as the date of surgery, and for patients who remained on a medication at the time of data extraction (Aug 27, 2019), the discontinuation date was set as the date of data extraction. Given the possibility that patients may have taken multiple AOMs simultaneously, the dates of initiation and discontinuation were recalculated to be exclusive of those of any other overlapping medications, which allowed for determination of the individual effects of each medication on a patient's weight. A medication instance was defined as the period during which a patient took a single medication exclusively without interruption. Hence, one patient could have had multiple instances of the same medication, as well as several instances of multiple medications.
Once the exclusive dates had been determined for each medication instance, all weights recorded for each patient were extracted and the following pertinent weights were defined: (1) Starting weight: the weight recorded closest to medication initiation (within 90 days pre and post); (2) Nadir weight: Lowest weight recorded during a medication instance; (3) End weight: the weight recorded closest to medication discontinuation (within 90 days pre and post). Maximum WL was defined as the starting weight minus the nadir weight. Final WL was defined as the patient's starting weight minus the end weight, hence a positive number indicates WL, while a negative number indicates weight gain. WL was represented as change in BMIz (i.e., BMIz) for children and adolescents, and as percentage of total body weight (%TBW) for young adults. After the initial cohort had been identified, only medications ‡ Strength of recommendation IIa indicates that the treatment is generally considered to be useful and is indicated in most cases (33).
Strength of evidence Category B is based on data derived from meta-analyses of randomized controlled trials (RCT) with conflicting conclusions with regard to the directions and degrees of results between individual studies. RCTs involved small numbers of patients or had significant methodological flaws (33,34). § Strength of recommendation IIb indicates that the treatment may be useful, and is indicated in some, but not most cases (34). with at least 5 medication instances with complete weight data (i.e., start, nadir, and end weights recorded in chart) were kept, which prompted the elimination of six medications: bupropion/naltrexone, canagliflozin, exenatide, lorcaserin, phentermine/topiramate, pramlintide. Table 1 contains the eight medications ultimately included in our study. Given that these medications are not all uniquely prescribed for weight-loss, we also flagged the following pertinent medical conditions as potential on-label indications for medication prescription: epilepsy, tobacco use/dependence, alcohol dependence/abuse, opioid dependence/abuse (Supplemental Table 1).

Statistical Analysis
Univariate statistics comparing patient demographics and medication details (i.e., medication duration, BMI and BMIz at time of medication initiation, nadir and WL, etc.) were obtained using Student's t-test and Pearson's chi 2 . Multivariate analyses to ascertain the effects of each medication on a patient's weight were also performed using linear regression clustered at the patient level and adjusted for patient age at the time of medication initiation, sex, race, primary insurance, obesity class at time of medication initiation, and number of obesity-related illnesses. Rather than arbitrarily selecting one of the medications as reference for analysis, the change in BMIz or %TBW associated with each medication was compared to the overall frequency-weighted mean (mean).
Lastly, in order to ascertain the effects of these medications when prescribed specifically for WL, all patients with a potential on-label indication for prescription were excluded and all aforementioned unadjusted and adjusted analyses were repeated. The following specific groups were excluded: (1) Patients with diabetes on metformin or liraglutide; (2) Patients with epilepsy on

RESULTS
A total of 1,720 patients were identified, consisting of 83 children, 492 adolescents, and 1,145 young adults, which together represented 2,210 medication instances. A starting weight and BMI were available for roughly 75% of patients ( Table 2). The mean BMI at the time of medication initiation among children, adolescents, and young adults was 34.0, 39.1, and 39.6kg/m 2 , respectively. This corresponded to a BMIz of 2.4 and 2.3 for children and adolescents respectively. Most patients (1,343; 78.1%) were prescribed a single medication during the study period. A higher proportion of young adults were female (82.3%) than among adolescents (70.9%) or children (54.2%). Most patients were white (59.1%) and privately insured (58.1%). A total of 268 patients underwent MBS during the study period, consisting of 181 (10.5%) sleeve gastrectomies, 87 (5.1%) laparoscopic Roux-en-Y gastric bypasses, and ≤10 laparoscopic adjustable gastric band placements. Information on obesity-related illnesses can be found in Table 2.

Children
Overall, metformin was the most commonly prescribed medication among children (41; 49.4%). After excluding patients with potential on-label indications, metformin remained the most commonly prescribed medication, along with topiramate both at 52.2% (two patients were on both medications; Table 3). Most children continued to take the study medication at the time of data extraction (58.5%), while 22.0% had discontinued it for unknown/unspecified reasons. See Table 4 for the other common reasons for medication discontinuation.

Unadjusted Analyses
A total of 90 medication instances were identified among children, but complete weight data (starting, nadir, and end weights) were available for only 41 (45.6%) of them. Compared to the mean starting BMIz of 2.42, children on bupropion had a lower starting BMIz of 2.21 (p = 0.03). No other statistically significant differences were identified on unadjusted analysis among this small cohort ( Table 5). See Figure 1 for a graphic representation of the final WL distribution in this cohort.

Adjusted Analyses
Multivariable analyses demonstrated higher obesity class was significantly associated with greater maximum WL ( BMIz 0.17 for both Class II and III, p = 0.002 for both; Table 6). Black children were more likely to experience a final weight gain ( BMIz −0.27; p = 0.05) when compared to whites. Similarly, compared to patients with no obesity-related illnesses, children with one or more conditions were significantly more likely to experience a final weight gain ( BMIz −0.78 to −0.4, p < 0.03; Table 7).

Off-Label Analyses
After removing all on-label medication instances, only metformin and topiramate had 5 or more instances with complete weight data. No statistically significant findings were identified on unadjusted analysis ( Table 8), or on multivariable analysis related to maximum WL ( Table 9). The previously observed relationship between black race and obesity-related illnesses and final WL remained (Table 10).

Adolescents
Metformin was again the most commonly prescribed medication among adolescents (280; 56.9%), even after excluding patients with potential on-label indications (173; 57.9%; Table 3). Most adolescents continued to take the study medication at the time of data extraction (45.5%), while 31.7% had discontinued it for unknown/unspecified reasons. See Table 4 for the most common reasons for medication discontinuation.

Unadjusted Analyses
A total of 624 medication instances were identified among adolescent patients, 299 (47.9%) of which had complete weight data recorded in the chart ( Compared to the respective means, adolescents on phentermine experienced a greater maximum WL (0.24 vs. 0.09 BMIz; p = 0.003) and final WL (0.12 vs. −0.02 BMIz; p = 0.001), while patients on topiramate and zonisamide experienced a significantly lesser final WL (−0.08 BMIz for both; p = 0.02 and <0.001, respectively). A significant proportion of adolescents ultimately experienced weight gain by the time of medication discontinuation (Figure 2).

Adjusted Analyses
Multivariable analyses demonstrated adolescents on phentermine experienced a final WL 0.12 BMIz greater than the mean (p = 0.003), while those on topiramate experienced a final WL 0.07 BMIz lesser than the mean (p = 0.02). Higher weight class at the time of medication initiation was associated with greater final WL ( Table 7).

Off-Label Analyses
After removing all on-label indications, metformin, topiramate, and phentermine remained. Unadjusted analyses were largely unchanged from prior ( Table 8). Multivariable analyses demonstrated patients on phentermine experienced a 0. 15 BMIz greater maximum WL and final WL (p = 0.003 and <0.001, respectively) when compared to the respective means, while patients on metformin experienced a 0.04 (p = 0.006) and 0.02 (p = 0.03) BMIz lesser maximum WL and final WL respectively (Tables 9, 10).

Young Adults
In line with the other two cohorts, metformin was the most commonly prescribed medication among young adults overall (484; 42.3%), but topiramate was the most commonly prescribed off-label medication (298; 46.3%, Table 3). Similar to children and adolescents, a large number of young adults continued to take the medication at the time of data extraction (36.8%), while a slightly higher proportion had discontinued it for unknown/unspecified reasons (37.3%; Table 4).

Unadjusted Analyses
A total of 1,496 medication instances among young adult patients were identified, 665 (44.5%) of which had complete weight data recorded in the chart. Among these, patients on metformin and phentermine had a higher starting BMI compared to the mean, while this was lower for patients on topiramate, bupropion, and naltrexone (  Figure 3 contains the overall and medication-specific distribution of final WL among young adults.

Adjusted Analyses
Young adult patients on orlistat experienced a 2.83%TBW lesser maximum WL when compared to themean (p < 0.001), while patients on phentermine and topiramate experienced a 1.54 (p = 0.05) and 1.47 (p = 0.01) %TBW greater final WL compared to the mean, respectively (Tables 6, 7). Similar to the findings among adolescents, higher weight class at medication initiation was associated with higher final WL.

Off-Label Analyses
After removing all patients with on-label indications, results were similar to prior. Patients on metformin had a higher starting BMI compared to the mean (41.5 vs. 40.2 kg/m 2 ; p = 0.005), while it was lower for patients on topiramate (38.3 kg/m 2 , p < 0.001; Table 8). Both patients on Orlistat (0.9%TBW; p = 0.003) and Bupropion (−4.91 %TBW; p = 0.001) had significantly lesser maximum WL compared to the mean ( Table 9). The only medication with a final WL significantly different than the mean (−0.12%TBW) was naltrexone (3.29%TBW; p < 0.001). Patients  on phentermine trended toward a greater final WL (1.17 %TBW) than the mean but this relationship did not reach statistical significance (p = 0.06; Table 10).

DISCUSSION
Our study describes the current prescribing patterns of weightmodifying pharmacotherapy amongst children, adolescents and young adults in a large unified health care system. This study also distinguishes between overall and off-label prescribing to assess the effectiveness of these medications in achieving weight modification, regardless of the specific indication. Due to sampling size, the strongest conclusions may be drawn from the adolescent and young adult cohorts. In these patient populations, phentermine appears to most consistently support WL, which is statistically significant in our study, though clinically small. Interpretation of phentermine's true weight modification benefits, as well as those effects of the other medications, however, are likely minimized based on their comparison to the mean rather than a control group not taking weight-modifying pharmacotherapy. Generally, higher obesity classification at medication initiation was associated with greater final WL. The analysis also reveals that metformin was overall the most commonly prescribed medication for both on-and off-label use among children and adolescents; however, in the adjusted analyses, it was associated with statistically significant negative final WL (i.e., weight gain) among adolescents and young adults for whom it was prescribed off-label compared to the mean. To our knowledge, this study is the first of its kind to explore this breadth of medication use with a focus on WL across a spectrum of ages from 5 to 25 years old.
The retrospective nature of our study gives rise to some important limitations. Though start and end dates of medications were manually verified by a team of clinicians, chart review is only as good as the quality of the data documented, which means that medication duration should be interpreted with caution. Additionally, the lack of a control group and our inability to account for confounding lifestyle modifications, such as diet quality and physical activity, make it difficult to fully interpret the absolute magnitude of effectiveness of the medications studied as mentioned above. Several other limitations are worth noting. First, given that we chose to limit our assessment of a medication's effect on WL to its exclusive dates of administration, we are unable to comment on the synergistic effect that multiple medications taken together may have. Second, our study did not capture medication dosing, which could play an important role and should be investigated further in future studies (35). Despite significant effort to determine reasons for medication discontinuation as part of the manual chart review, it is notable that this reason was unspecified for approximately one-third of patients. Lastly, this study did not account for confounders such as the simultaneous use of other medications associated with weight gain (18), which could be of particular importance given the high rates of anxiety (62.3%) and depression (55.7%) noted.
Despite these limitations, our study is the first to our knowledge to evaluate prescribing patterns and usage of a variety of medications among children, adolescents, and young adults,  in the context of overweight and obesity. Strengths of this study include its large sample size based out of a multicenter academic health system with an established tertiary care facility to treat pediatric obesity. Providers in this center include a multidisciplinary team (i.e., physician, dietitian, psychologist) trained specifically in the management of both pediatric and adult obesity and therefore are more inclined and familiar with prescribing these medications for the indication of WL. Obesity is a heterogeneous, multifactorial disease that is best treated with multimodal therapies, including lifestyle modification, pharmacotherapy, and surgery as an integrated continuum of care (36,37). No studies have demonstrated a durable, sustained WL through lifestyle modifications alone (15).
A systematic review by O'Connor and colleagues found that children and adolescents following a traditional lifestyle modification plan require a minimum of 26 h over a 6-12-months period of contact with providers to achieve an average 1.0 kg/m 2 BMI reduction (38). Another study which evaluated age and obesity class, showed that behavioral interventions are less effective in terms of BMI reduction among adolescents (14-16 years old) when compared to children (6-9 years) (39). For adolescents who have experienced insufficient weight control with lifestyle modification therapy and who do not meet criteria for MBS, pharmacotherapy could serve as an important treatment option (40). Our study corroborates this and suggests that phentermine could be a useful first-line pharmacologic option for adolescents 13-18 years old.
Pharmacotherapy, in particular, directly targets biologic adaptations and counterregulatory mechanisms through action on appetite and hunger (39,41). Note, that with any of the interventions utilized to treat obesity, there is an anticipated metabolic adaptation to WL. It is not uncommon for patients  with obesity to experience WL followed by weight regain (42). In a 2011 paper delving into the physiology of WL and regain, Maclean and colleagues summarize this phenomenon and conclude that treatments must therefore be wholly exhaustive and redundant to overcome this biology (43). Though weight reduction in our study was modest, studies have shown that initial WL with behavioral changes and adjunctive pharmacotherapy is enough to reduce significant health risks associated with many obesity-related conditions (44). In a systematic review for the US Preventive Services Task Force (USPSTF) of recent randomized-controlled trials, researchers found that the control groups were more likely to continue to gain excess weight compared to the pharmacologic treatment groups. They concluded that even an arrest in weight gain may prove clinically significant over a lifetime (38).
Unfortunately, there are several barriers to medication adherence including inadequate compliance, adverse effects, insurance coverage and other reasons for medication discontinuation as demonstrated in this study. There is also prescriber resistance to AOM use while awaiting results of FDA-approved long-term outcome trials and inadequate obesity education across the continuum from medical school to fellowship (19,45). Hence, most current prescribers of AOMs remain endocrinologists who are more comfortable with managing these medications for other indications despite availability of certification by the American Board of Obesity Medicine (ABOM). A recent study demonstrates that pediatricians are least likely to become certified by the ABOM to treat patients with obesity despite the consistent rise of obesity among the pediatric population (19,46). Still, off-label systemic medication prescribing for unapproved conditions in pediatrics has increased in recent years (47). As seen in this study, most of the medications on market and in use for separate indications are FDA-approved for adults for obesity and are often used off-label to treat obesity in the pediatric population.
A recent policy statement from the American Academy of Pediatrics (AAP) recommended improved access for pediatric patients "to multidisciplinary programs that provide high quality pediatric metabolic and bariatric surgery" with no lower age limit given emerging long-term outcome data (15,16,48). With this shift in policy comes an even greater role for pharmacotherapy treatment as prior studies have found that AOMs serve as a useful adjunct in postsurgical patients following bariatric surgery to treat inadequate WL or weight regain (49)(50)(51)(52)(53). Unfortunately, there also remain significant barriers to MBS for the management of pediatric obesity relating to racial disparities in that white adolescents are more likely to undergo MBS despite higher rates of obesity existing amongst black and Hispanic youth (54).
Medications are uniquely suited for use in the pediatric population because children with obesity have a greater degree of disease plasticity (4). The fact remains that there are far greater FDA-approved options available to adults with obesity than for children and adolescents. This study supports the need for more clinical trials and greater efforts which focus on FDA-approval of medications for WL in pediatrics to target obesity early.
In conclusion, our study demonstrates that AOMs may play a role in weight management for adolescents and young adults with obesity and should be considered an important component of a multimodal approach to managing this severe and deleterious condition. Their use may become particularly important as patients who undergo lifestyle modification or MBS continue to face challenges with attaining and maintaining appropriate, durable WL. Prospective studies should be conducted to better understand the impact of these medications in the pediatric population.

DATA AVAILABILITY STATEMENT
All datasets generated for this study are included in the article/Supplementary Material.

ETHICS STATEMENT
The studies involving human participants were reviewed and approved by the Partners Institutional Review Board (IRB) prior to written informed consent from the participants' legal guardian/next of kin was not required to participate in this study in accordance with the national legislation and the institutional requirements.

AUTHOR CONTRIBUTIONS
FS and KJC contributed to conception and design of the study. KJC and NP organized the database. KJC, SS, and KSC completed manual chart review. NP performed the statistical analysis. KSC wrote the first draft of the manuscript. NP and KSC wrote sections of the final manuscript. All authors contributed to manuscript revision, read, and approved the submitted version.

FUNDING
This work was supported by the Physician/Scientist Development Award (PSDA) granted by the Executive Committee on