Patterns of Physical Activity Adherence by Adolescents With Diabetes or Obesity Enrolled in a Personalized Community-Based Intervention

Abstract

Purpose

The purpose of this study was to determine the feasibility of a personalized, 16-week community-based physical activity intervention for adolescents with diabetes or obesity and examine the weekly patterns of adherence to the intervention.

Methods

Physical activity adherence was evaluated throughout the intervention using accelerometers in 46 adolescents with type 1 diabetes (N = 22), type 2 diabetes (N = 12), or obesity (N = 12) (age, 14.4 ± 1.5 years; 56.5% female; 61% Hispanic). Of these, 39 completed the intervention, and 7 did not.

Results

There were no differences in baseline anthropometric characteristics or fitness between the completers versus noncompleters. Completers began above 1060 metabolic equivalent (MET) min/wk⁻¹ and stayed above 900 MET min/wk⁻¹ for ~4 weeks and declined 39 MET min/wk⁻¹ until end of study. Noncompleters began at 924 MET min/wk⁻¹ yet dropped below 800 MET min/wk⁻¹ by end of week 1 and declined an average of 151 MET min/wk⁻¹. Interestingly, self-report of barriers to activity were higher in completers versus noncompleters.

Conclusions

Findings highlight that adolescents completing the intervention could sustain a prescribed level of personalized activity for at least 1 month but had steadfast declines in weekly activity. Even with individualized programs, factors other than barriers to activity need to be considered when designing approaches to physical activity adherence for adolescents with diabetes or obesity.

For children with diabetes or obesity, physical activity is associated with improved cardiovascular function, bone health, body composition, mental health and well-being, as well as acute increases in insulin sensitivity.^1–5 Despite the numerous health benefits, many youth are not meeting the recommended guidelines for physical activity.^6–9 According to objectively measured levels of physical activity obtained by accelerometer data, only 42% of 6- to 11-year-old youth and 8% of adolescents meet the recommended guidelines of 60 minutes of moderate to vigorous physical activity (MVPA) on most days of the week.⁸ Furthermore, these low levels of physical activity result in a greater challenge down the road as physical activity is reported to decline from childhood into adolescence.^6–10 The youth most vulnerable to the decline in physical activity tend to be female^9–14 or those with obesity or diabetes,^14–19 and they are most physically inactive during weekends and outside of school.^13,20 Despite the design and implementation of eloquent physical activity programs aimed at improving and sustaining physical activity adherence, incorporating physical activity as a consistent lifestyle behavior has proven difficult for both youth and adults.^10,21 Gordon-Larsen et al¹⁰ examined longitudinal patterns of physical activity in teens as they transition into adulthood and found of those achieving 5 or more weekly sessions of MVPA, only 4.4% continued to achieve these favorable amounts of activity into adulthood.¹⁰ There are no studies to date looking at the trajectory of physical activity from childhood into adulthood in youth with diabetes, However, Bohn et al²² reported the percentage of adults with T1DM that exercised twice a week or more declined from 25% in people aged 18 to 29 to 13% and 10% of those aged 30 to 44 years and 45 to 79 years, respectively. High participant dropout rates and low exercise adherence clearly jeopardize the maintenance of the long-term health benefits associated with ongoing and consistent patterns of physical activity.

Emphasis on understanding the sustainment of physical activity adherence to minimize a wide range of future health-related consequences is of particular importance. However, little is known about the factors that may influence physical activity adherence in youth, even in those whose treatment and preventive health recommendations strongly indicate the value of physical activity for optimal health outcomes. Furthermore, to our knowledge, no previous study has included the use of consecutive daily accelerometry tracking throughout a physical activity intervention lasting several months. Indeed, few studies have even reported adherence levels, and most define adherence in terms of attendance to the program versus compliance.¹⁰ In those that do report adherence, most have focused on those with T1DM rather than T2DM or monitored adherence via step count charts,²³ interviews or text messaging,^23–25 or online logs²⁶ rather than objectively measured data. In studies that used objectively measured data to examine adherence to exercise, heart rate monitors were used during the supervised exercise session only rather than volume of exercise expressed in metabolic equivalent (MET) minutes.^27–29 Consistent with the heterogeneity among adherence measures in these studies, adherence levels were also quite variable (52%−100%).¹⁰ In our study, sedentary adolescents with diabetes or obesity adhered to 30 minutes of MVPA 70% of the time and 60 minutes of MVPA 38% of the time, respectively.³⁰ To describe patterns of physical activity adherence over the 16-week intervention period, a secondary analysis of the accelerometry data was conducted to better explain when additional behavioral intervention(s) may be necessary to improve exercise adherence. The aims were to describe weekly patterns of physical activity between completers versus noncompleters as well as timepoints when physical activity declines. Baseline anthropometric characteristics, fitness, and physical activity barriers between those who completed versus those not completing the intervention were also examined.

Methods

Research Design

The parent pretest-posttest feasibility trial, Personalized Exercise for Adolescents With Diabetes, funded by the National Institutes of Health (7R21 NR009267),³⁰ previously explored the feasibility of achieving the recommended goal of 60 minutes of MVPA on at least 5 days per week. Our current investigation employed a secondary analysis³¹ of existing data to answer new research questions and describe the daily and weekly patterns of exercise adherence to the personalized exercise program and factors that affected exercise adherence for adolescents with a diagnosis of diabetes or obesity. Against the backdrop of rising obesity prevalence rates in those with and without diabetes, this study approach was useful for gaining an understanding of how best to translate the findings into practice for motivating adolescents with known future cardiovascular risks to be more physically active at key timepoints when activity levels decline.

Key research questions included: (1) Were there differences in demographic characteristics, baseline anthropometric data, lipid and metabolic profiles, barriers to physical activity, and accelerometer outcomes between those completing versus not completing the study? Accelerometer outcomes reflected measures of adherence and included: daily hours worn, days of data, weeks of data, MET min/day⁻¹, and MET min/wk⁻¹. (2) Did accelerometer outcomes differ for completers versus noncompleters based on age, gender, race, ethnicity, or diagnosis (T1DM, T2DM, or obese) and baseline measures of A1C, BMI z-score, and VO_{2 peak} (expressed in ml/kg/min)?

Sample/Setting

Accelerometry data were available for 46 youth (22 with T1DM, 12 with T2DM, and 12 with obesity [BMI ≥95th percentile]). Of these, 39 completed the intervention that included wearing the accelerometer, pretesting, and posttesting (completers), and 7 did not, only completing pretesting and shortened accelerometer wear time (noncompleters). As noted in Figure 1A, noncompleters primarily ended their wear time of the accelerometer during the first month, with a couple maintaining some level of participation into the third month. Mean age of completers was 14.4 ± 1.5 years, 56.5% were female, and 61% were Hispanic, with the remainder non-Hispanic white. Participants were recruited from a pediatric diabetes clinic in the southwestern United States. Inclusion criteria consisted of a diagnosis of either T1DM or T2DM or obesity for at least 6 months, as determined by the treating pediatric endocrinologist. Rationale to include youth with diabetes or obesity was based on the guidelines for physical activity for all youth, particularly for those who have known potential risks for future cardiovascular disease due to their current health condition in addition to a sedentary lifestyle. Youth were between 12 and 19 years of age and had a parent willing to participate in regular, moderate activity such as walking. Patient education was provided to the participant (including those that were obese) and the parent prior to the start of the intervention. Topics included nutrition, interaction effects of intensity/duration of exercise on glucose level, importance of checking blood glucose before and after exercise, and risk of nocturnal hypoglycemia. Participants with diabetes remained in standardized care with their diabetes educator and endocrinologist throughout the intervention. The study was approved by the University of Arizona Institutional Review Board where the trial was conducted, and parental consent and child assent were obtained prior to study participation.

Figure 1.

(A) Weekly metabolic equivalent (MET) minutes⁻¹: Completers and noncompleters overall. (B) Weekly MET minutes⁻¹: Completers and noncompleters by diagnosis. Horizontal red dashed line drawn at 900 MET minutes per week; in Figure 1B: solid circles indicate subjects with T1DM, blue stars indicate subjects with T2DM, and red diamonds indicate subjects who are obese (≥95th percentile BMI). The lines drawn in each plot are locally weighted smoothed fit (loess) lines for each set of subjects, respectively. The x-axis “Weeks From Start” begins at 0 for statistical modeling purposes so that the model intercept estimates the average weekly MET minute levels for the first week of physical activity.

Physical Activity Intervention

A more detailed description of the personalized intervention model has been discussed previously.^30,32 In brief, youth were screened using the Physical Activity Recall³³ to ensure inclusion of those who were sedentary. Sedentary was defined as <36 MET/d as this is well below the current recommendations of daily MVPA for youth. Based on the Physical Activity Recall, adolescents accumulated an average of 9.2 ± 10.7 minutes per day of MVPA the week preceding enrollment. Upon enrollment, youth were instructed to participate in an individualized community-based aerobic physical activity intervention at 60% to 75% peak heart rate (derived from a graded exercise test) for 60 minutes a day for a minimum of 5 days per week. Desired activities were age-appropriate and specific to the participant. Examples of chosen activities include: biking, kick boxing, Dance Dance Revolution (Konami, Japan), and walking. Exercise programs were 16 weeks at length, and bimonthly home visits were conducted as an intervention fidelity check to review accelerometer data, including each participant’s level of physical activity adherence and intensity and any risks such as hypoglycemia. If youth were not achieving the recommended levels of physical activity prescribed, motivational strategies including discussion of barriers and complimenting success ensued.

Accelerometer and Adherence Outcomes

Youth were instructed to wear an Actigraph accelerometer (model GT1M, Pensacola, FL) on the right hip throughout the 16-week intervention, as described previously.^30,32,34 As our primary purpose was to assess trends in moderate to vigorous physical activity, accelerometer data were deduced consistent with National Health and Nutrition Examination Survey (NHANES).⁸ In brief, wear time was determined by subtracting non–wear time from 24 hours. Non–wear time was defined as 60 consecutive minutes of zero activity counts. Sixty-second epoch durations were downloaded approximately every 2 weeks. MET were used to calculate energy cost of physical activity and extrapolated using raw accelerometer counts per epoch as developed by Freedson et al.³⁵ MET are a simple and practical approach to expressing energy cost or caloric expenditure. The more calories expended, the higher the energy cost of the activity. For example, 1 MET is defined as the amount of oxygen consumed while sitting at rest and is equivalent to ~3.5 ml/kg/min of O₂. The volume of exercise was used to estimate the total energy expenditure of a participant’s MVPA in MET minutes per week (MET min/wk⁻¹). Expressing physical activity in terms of volume is more representative of the total amount of activity performed and is shown to play an important role in health and fitness outcomes.³⁶ Weekly MET minutes were calculated by multiplying the energy expenditure of an activity (in MET) by the physical activity duration (in minutes). The goal of the intervention was to achieve moderate (3–5.9 MET) to vigorous (≥6 MET) physical activity for 60 minutes 5 days per week. This would be equivalent to a minimum volume of 900 MET min/wk⁻¹. For example, an average 3.5 MET activity performed for 1 hour 5 days per week is equivalent to 1050 MET min/wk⁻¹.

Baseline Measures

Baseline measures were collected prior to the start of the intervention and included anthropometric measures, fasting blood samples, cardiovascular fitness, and barriers to engaging in physical activity. Posttesting parameters were not explored in the current analysis as all noncompleters opted out of posttesting. Replicate height (≥1.0 cm) and weight (≥0.5 kg) were determined to the nearest centimeter and kilogram, respectively, and repeat trials were conducted if necessary. Blood pressure was taken in the seated position on the right side of the body. Gender and age-adjusted BMI percentile was calculated according to the Centers for Disease Control growth charts (http://www.cdc.gov/growthcharts/data_tables.htm). Fasting laboratory blood samples for lipid profile (total cholesterol, low-density lipoprotein [LDL-c], high-density lipoprotein [HDL-c], and triglycerides), insulin, and glucose were collected and assessed via a clinical Cardio-check PA analyzer (Polymer Technology Systems, Indianapolis, IN). Glycosylated hemoglobin (A1C) was obtained using the DCA2000 (Bayer HealthCare LLC, Elkart, IN). Cardiovascular fitness (VO₂ peak) was assessed using the McMaster protocol on a cycle ergometer (Ergoselect 100, Ergoline, Bitz, Germany), as previously described.^30,34 Bioelectrical impedance was measured using the RJL Systems Analyzer (Quantum X, Clinton Twp, MI, USA) for the estimation of percentage body fat, fat mass, and fat free mass.

The Perceived Barriers to Action Instrument measures reasons why one might not exercise, using 10 items on a Likert scale from not at all true (1) to very true (5). Sample items include barriers such as not having someone with whom to exercise, homework, chores, the weather, and not having a good location to exercise. Scores are added across all items, and a mean score is calculated. Cronbach’s alpha for internal reliability is reported as .77 in a large, multiracial sample of youth.³⁷

Statistical Analysis Methods

Data were reviewed for completeness, outliers, and normality distribution assumptions. Weekly MET minutes were right skewed, and a square root transformation was applied to normalize the data prior to analysis. Demographics and baseline anthropometric measures were summarized and compared between the 39 subjects who completed the study and the 7 who did not. Comparison tests were performed using t tests for continuous measures and Fisher’s exact tests for the normally distributed categorical data (which was small, with expected counts <5 for the noncompleters). BMI percentile was significantly left skewed, so the nonparametric Mann-Whitney test was performed to compare completers to noncompleters. To weight each individual’s data contribution by the varying number of days of accelerometer data, a multilevel mixed modeling (MLM) approach to analysis of variance (ANOVA)³⁸ was used to test for group differences for daily minutes worn and daily and weekly MET minutes. The available data included 3212 daily measurements from 46 subjects collected up to 20 weeks since a few subjects continued to collect PA past the 16-week intervention. All available data were utilized in the longitudinal models and in comparing the average daily minutes worn and average daily and weekly MET minutes. The MLM models included linear and quadratic time effects with random intercept and random slope for linear time. After running an initial model for the time effects, additional models were run to investigate potential covariates for their main effects and potential interactions with both linear and quadratic time. The covariates considered for inclusion in the models were run and tested one at a time since the sample size was not large enough to test all covariates at once. The covariates evaluated were age, gender, race, ethnicity, diagnosis (T1DM, T2DM, or obese), and baseline measures of A1C, BMI z-scores, and VO₂ peak (expressed in ml/kg/min). Age, A1C, and VO_{2 peak} were mean centered prior to computing the interaction with time.³⁸ It is also noted that diagnosis and A1C were confounded. T1DM had the highest A1C levels (8.9 ± 1.9; range, 5.2–12.8), followed by T2DM subjects (8.2 ± 1.9; range, 5.5–11.2) and obese subjects (5.4 ± 0.3; range, 5.0–6.0), which were significantly different, Brown-Forsythe F(2, 25.2) = 22.044, P < .001, with the obese subjects having levels significantly lower than either the subjects with T1DM (Games-Howell P < .001) or T2DM (Games-Howell P < .001). This finding was expected and provided confirmation that the subjects in the obese group did not have a diagnosis of diabetes. For the purposes of the final models, A1C was found to be a significant covariate and was a numerically more precise predictor than diagnosis category; however, it is noted that these are highly related (statistically and clinically). To illustrate interaction effects between time, A1C, and gender, percentiles for the upper (75th) and lower (25th) quartiles of A1C for males and females were used. Using percentiles is a common practice for visualizing interaction effects with continuous measures.³⁹

Results

There were no significant differences in age, race, ethnicity, gender, or baseline measures of BMI, waist circumference, body fat percentage, cardiovascular fitness, A1C, blood lipids, triglycerides, or systolic blood pressure between completers (N = 39) and noncompleters (N = 7) (Table 1). Completers had lower diastolic blood pressure compared with noncompleters at baseline (P = .042). Interestingly, those who completed the exercise intervention reported more barriers to exercise versus those who dropped (P < .007). Furthermore, a comparison of baseline barriers to exercise across the 3 groups (T1DM, T2DM, and obese) revealed significantly more barriers in those with T2DM (3.30 ± .63) versus T1DM (2.63 ± .63, P = .012) or obese (2.80 ± .68, P = .025).

Table 1.

Participant Characteristics^a

	Completers n = 39	Noncompleters n = 7	P Value
Demographics
Age (y)	14.4 ± 1.6	14.1 ± 1.4	.689b
Race (white)	37 (94.9)	6 (85.7)	.398c
Ethnicity (Hispanic)	23 (59.0)	5 (71.4)	.688c
Gender (female)	23 (59.0)	3 (43.0)	.428c
Diagnosis			ned
T1DM	20 (51.3)	2 (28.6)
T2DM	9 (23.1)	3 (42.9)
Obese	10 (25.6)	2 (28.6)
Baseline anthropometric data
BMI (kg/m2)	30.1 ± 9.1	36.0 ± 10.8	.131b
BMI (z-score)	1.6 ± 1.1	2.3 ± 0.6	.124b
BMI percentile	85.7 ± 23.3	97.7 ± 2.9	.121d
Waist circumference (cm)	99.4 ± 23.8	111.0 ± 27.4	.251b
Fat mass (kg)	31.6 ± 19.1	39.7 ± 20.2	.609b
Body fat (%)	36.1 ± 12.7	38.6 ± 7.9	.609b
VO2 (ml/kg/min−1)e	27.6 ± 8.3	27.1 ± 8.4	.875b
Baseline metabolic and lipid profile
A1C (%; mmol/mol)	8.0 ± 2.2 (64 ± 24)	7.0 ± 2.1 (53 ± 23)	.243b
Cholesterol (mg/dL; mmol/L)	157.6 ± 43.9 (4.07 ± 1.14)	137.4 ± 38.8 (3.56 ± 1.00)	.262b
LDL-c (mg/dL; mmol/L)	99.2 ± 28.3 (2.57 ± 0.73)	93.5 ± 33.9 (2.42 ± 0.88)	.636b
HDL-c (mg/dL; mmol/L)	38.5 ± 12.8 (1.00 ± 0.33)	32.9 ± 7.7 (0.85 ± 0.20)	.264b
Triglycerides (mg/dL; mmol/L)	53.4 ± 37.1 (0.60 ± 0.42)	44.4 ± 20.7 (0.50 ± 0.23)	.540b
Systolic BP (mg/dL)	119.6 ± 10.1	113.2 ± 4.4	.172b
Diastolic BP (mg/dL)	70.0 ± 6.4	77.2 ± 11.2	.042b
Barriers to physical activity	2.86 ± .71	2.43 ± .26	.007b
Accelerometer outcomes
Daily minutes worn (min/d−1)	711.8 ± 279.9	559.1 ± 243.3	.046e
Daily hours worn (h/d−1)	11.9 ± 4.7	9.3 ± 4.1	.046e
Days of data	77.0 ± 25.6	29.7 ± 16.3	<.001b
Weeks of data	14.5 ± 3.1	7.1 ± 4.0	<.001b
MET min/day−1	187.9 ± 177.7	157.6 ± 166.7	.664e
MET min/wk−1	995.3 ± 808.5	652.7 ± 628.4	.348e

^aMean ± SD and n (%) reported. ned, not enough data to perform test.

^at test.

^bFisher’s exact test.

^cMann-Whitney test.

^dMultilevel mixed model ANOVA.

^eOne noncompleter missing.

Compared with noncompleters, completers wore the accelerometer for a longer duration throughout the 16-week intervention (14.5 ± 3.1 vs 7.1 ± 4.0 weeks, P < .001) and throughout the day (711.8 ± 279.9 vs 559.1 ± 243.3 min/d⁻¹, P = .046) (Table 1). There were no significant differences in the average daily (P = .664) or weekly (P = .348) MET minutes between groups, respectively (Table 1). In other words, the overall volume of physical activity between completers and noncompleters was similar. However, in terms of the pattern of activity, completers on average began the intervention above 1060 MET min/wk⁻¹ and stayed above 900 MET min/wk⁻¹ for an average of 4 weeks but declined thereafter by 39 MET min/wk⁻¹ until end of study (14.5 ± 3.1 wk) (Figure 1A). It should be noted that the x-axis in Figures 1A and 1B, “Weeks From Start,” begins at 0 for statistical modeling purposes so that the model intercept estimates the average MET min/wk⁻¹ levels for the first week of physical activity. Noncompleters started their first week at 924 MET min/wk⁻¹ but dropped below 800 MET min/wk⁻¹ by the end of the second week and declined an average of 151 MET min/wk⁻¹ until end of study (7.1 ± 4.0 wk). It is further noted that when viewing these data by diagnosis category (Figure 1B), the subjects who were obese started with the highest levels of weekly MET minutes (>1300 MET minutes, depicted by the red line) and maintained these levels until approximately 11 weeks from the start, dropping below 900 MET min/wk⁻¹ after 14 weeks from start. The subjects with T1DM (Figure 1B, represented by the black line) started a little lower than the obese but stayed above 900 MET min/wk⁻¹ until about 5 weeks from the start, after which their levels slowly declined. The subjects with T2DM (Figure 1B, represented by the blue line) started the lowest at approximately 800 MET min/wk⁻¹, which declined slowly until about 12 weeks from the start when their levels dropped off rapidly.

To model the longitudinal trends, all 46 subjects’ accelerometry data were combined to maximize the use of all available measurements. The MLM approach adjusted appropriately for the differing amounts of data not only among subjects completing the study but also between those who did and did not complete the study. Using the square root transformed weekly MET minutes, and initial MLM models testing each covariate separately indicated A1C and gender were the only two covariates that had significant interaction effects with both linear and quadratic time effects (P < .05). It is noted, however, that while diagnosis was confounded with A1C levels, as a covariate, diagnosis did not significantly interact with the linear and quadratic time effects. Thus, a final longitudinal model was computed including linear and quadratic time effects, plus the main effects of A1C and gender and their respective interaction effects (Table 2). With all effects included in the model, there was a significant difference between the genders at their first week (P = .030), with males beginning higher and staying consistently higher in their weekly MET minutes than females (Weeks × Gender P = .002; Weeks² × Gender P = .001). While there were no significant differences in weekly average MET minutes given their A1C levels (P = .155), subjects with lower A1C typically kept higher weekly MET minutes across time and declined at a slower rate than those with higher A1C (Weeks × A1C P = .025; Weeks² × A1C P = .012). Although the diagnosis of type of diabetes or obesity was not significant with linear or quadratic effects, the results of lower A1C levels with higher activity levels and gradual decline over time are reflective of those with obesity versus diabetes.

Table 2.

Longitudinal Multilevel Linear Mixed Models for MET min/wk⁻¹ by Gender or A1C^a

						95% CI
Parameter	B	SEB	df	t	P Value	LB	UB
Intercept	28.85	2.09	61.33	13.82	.000	24.68	33.03
Gender (male)	7.04	3.17	60.96	2.22	.030	.70	13.39
A1C	−1.04	.72	61.42	−1.44	.155	−2.48	.40
Weeks	.12	.37	347.76	.31	.756	−.62	.85
Weeks2	−.06	.02	561.96	−2.63	.009	−.11	−.02
Weeks × Gender	−1.77	.56	336.36	−3.13	.002	−2.88	−.66
Weeks2 × Gender	.11	.03	563.67	3.28	.001	.04	.18
Weeks × A1C	−.29	.13	353.35	−2.26	.025	−.55	−.04
Weeks2 × A1C	.02	.01	564.02	2.51	.012	.00	.04

^aOutcome = square root of weekly MET minutes. A1C mean centered at 7.9. Weeks² quadratic time effect. CI, confidence interval; LB, lower bound; UB, upper bound.

Discussion

This investigation is the first to examine 16 weeks of physical activity adherence patterns in adolescents with and without diabetes enrolled in a personalized, community-based feasibility trial. Our study is unique in that it incorporated 16 weeks of continuous accelerometer data and demonstrated that (1) for those who completed the intervention, youth were able to sustain on average at least 900 MET min/wk⁻¹ of aerobic physical activity for 1 month and the volume of moderate to vigorous physical activity declined by 39 MET min/wk⁻¹, (2) male youth acquired higher levels of physical activity compared with female youth, and (3) youth with lower A1C tended to acquire higher levels of physical activity compared with those with higher A1C.

Despite the abundance of evidence supporting the benefits of physical activity in youth,^1–3 estimates from the Centers for Disease Control and Prevention (CDC) self-report data indicate that only 29% of high school students accumulate 60 minutes of physical activity on a daily basis.⁴⁰ Estimates in youth with diabetes or obesity are much lower in some^41–43 but not all studies.^44–46 For example, youth with T1DM, surveyed using the Habitual Activity Estimation Scale, estimated similar activity levels although more time spent in sedentary activities compared with youth without T1DM.⁴¹ The SEARCH for diabetes in youth study demonstrated that approximately three-quarters of youth met physical activity guidelines according to self-report measures.⁴⁵ Using objectively measured heart rate monitors, Edmunds et al⁴⁶ showed that 37 youth with T1DM participated in 53.6 ± 31.4 minutes per day.⁴⁶ In contrast to youth with T1DM, youth with T2DM appear to engage in much lower levels (8.2 ± 8.9 to 35.0 ± 26.0 depending on age and gender).⁴³

Even though physical activity interventions have been shown to be efficacious with initially increasing youth’s daily physical activity levels,^47,48 50% of individuals are reported to drop out within the first couple of months.⁴⁹ Appreciating the challenges youth with obesity and diabetes face with accumulating and adhering to the recommended guidelines, we designed a personalized community-based intervention^30,32 aimed at improving physical activity for 60 minutes per day, 5 days per week (at minimum ~900 MET min/wk⁻¹). We previously reported that youth enrolled in the personalized intervention achieved 900 MET min/wk⁻¹ approximately 38% of the time.^30,32 These data became the impetus to our current objectives to examine patterns in exercise adherence to determine potential key timepoints to intervene for promoting greater adherence.

This investigation demonstrates that despite a wide variability in the level of moderate to vigorous activity, youth with obesity or diabetes can sustain approximately an average of 900 MET min/wk⁻¹ for 1 month (depicted by the red dotted line in Figures 1A and 1B); however, the amount of physical activity declines by 39 MET min/wk⁻¹ (equivalent to ~15 min/wk⁻¹ at 3 METS) thereafter, until end of study (depicted by the solid black line in Figures 1A and 1B). Although it is important to note that the optimal exercise recommendations for improving health in youth with diabetes are unknown, studies in adults recommend volumes of 150 min/wk⁻¹ (or 450 MET min/wk⁻¹) to reduce the risk of developing T2DM⁵⁰ and volumes of 450 to 750 MET min/wk⁻¹ to significantly lower the risk of coronary heart disease or cardiovascular disease.³⁶ The consistent decline in activity beyond the first month of the intervention is evidence that youth eventually dropped below the recommended volumes (~450 MET min/wk⁻¹) for health benefits by the end of the 16-week intervention.

In considering individualized approaches to physical activity prescription, gender variation and level of glycemic control at the time of initiating the physical activity program were influential factors in physical activity adherence. Males engaged in a higher level of physical activity across the span of 16 weeks compared with females, confirming the numerous epidemiological studies demonstrating lower levels of physical activity and a greater decline in physical activity in females.^9–14,17 While there was no significant association of baseline level of glycemic control and amount of activity per week, youth with lower A1C had consistently higher levels of physical activity throughout the intervention, and their activity level declined more slowly than those with higher A1C. That is to say, obese youth that completed the intervention engaged in higher levels of physical activity compared to those with diabetes whose standards of care include daily moderate to vigorous aerobic activity. It is important to note that there is an influential association between muscle function, poor glycemic control, and lower aerobic fitness (ie, VO₂) that is further associated with lower exercise adherence.⁵¹ Nguyen et al⁵¹ assessed the relationship between physical activity levels and anaerobic and aerobic fitness in youth without diabetes (controls) and those with T1DM in good metabolic control versus poor metabolic control, defined by A1C. Anaerobic fitness, that is, the ability of muscles to synthesize adenosine triphosphate (ATP) without oxygen during short, high-intensity activities, was not different between groups. However, aerobic fitness, which describes the body’s ability to transport and utilize oxygen to create ATP during prolonged exercise was similar between controls and T1DM with good metabolic control but significantly lower in those with poor metabolic control. Poor aerobic fitness may be a reflection of poor oxidative muscle function. Therefore, it may be that poor metabolic control, defined as A1C, is reflective of poor oxidative muscle function that then translates into lower levels of physical activity endurance in terms of intensity and duration.

While barriers to exercise would tend to be a negative influence on adherence, our findings were counterintuitive and indicated that perceptions about deterrents to engaging in physical activity were more prevalent in those who were most adherent. The report of greater barriers to exercise in those completing versus not completing the 16-week intervention may inherently have been influenced by our personalized design. The individualized program was designed to address common barriers to physical activity that have been reported (eg, time, cost, lack of transportation, interests, and family support).⁵² It is tantalizing to consider that the very nature of the personalized design appropriately addressed these barriers; however, adherence was far from ideal, and as such, we must consider other factors that may influence physical activity adherence. For example, studies addressing exercise in youth with T1DM or T2DM have recently shed light on diabetes-specific barriers to exercise such as variation in insulin management, timing of exercise, fear of hypoglycemia, and adverse exercise kinetics.^53,54 Furthermore, age may also have an independent influence, as Jabbour et al⁵⁵ recently reported differences in perceived barriers in younger versus older youth with T1DM. In this study,⁵⁵ the highest common barrier scores reported in a sample of 201 children with T1DM were loss of control of diabetes, fear of hypoglycemia, and external temperature, although work schedule (of parents) was a barrier in youth <12 years of age, whereas older children (≥12 years of age) reported low levels of fitness. Although our youth did not report any hypoglycemic events,³⁰ further studies are warranted to determine how influential the fear of hypoglycemia is on exercise adherence.

The most compelling finding, however, is that more attention is needed for personalizing approaches to improve exercise adherence in adolescents and recognizing individual benchmarks of achievement over time so that there is not the tendency to become disinterested. Of particular importance may be anticipating the decline in activity at 1 month following initiation of an activity program as our data suggest. Strategies to improve or maintain exercise adherence may include monthly incentives and exploration of individual motivational factors. Overall, the lack of attainment of recommended levels of activity from leading national organizations^56,57 is a concern for youth with diabetes or who are obese and have a high risk for developing prediabetes or T2DM. Reasons for the lower levels of activity may be related to poor aerobic capacity, a lack of incentives at key timepoints (ie, obese adolescents and those with T2DM may have expected to see evidence of weight loss), a greater perception of barriers to exercise in those with T2DM, limited family support, and fear of hypoglycemia in adolescents with T1DM. Roberts et al⁵⁸ noted the absence of insulin adjustment and an increased frequency of blood glucose testing in relation to exercise in youth with T1DM. Further study is warranted to provide best evidence for clinicians to become more proficient in providing instructions on monitoring self-management practices for balancing blood glucose checks, insulin regime (if prescribed), and food intake related to exercise and prescribing exercise aligned to personal profiles of adolescents with diabetes or obesity that have known risks for future health disparities in cardiovascular outcomes.

Some limitations must be acknowledged. First, the relatively small sample size of youth was recruited from a university-based endocrinology center, which may limit the generalizability of the study to more “treatment seeking” individuals. This assumption is supported by our lower dropout rate (15%) (noncompleters) compared with adults (≥27%) with or without diabetes.^49,59,60 Similarly, the degree of influence from parental, peer, or professional support on physical activity adherence was not assessed. MacMillan et al⁶¹ previously demonstrated that parent and peer support had the most influence on physical activity participation. Although we required a family member to participate by agreeing to walk regularly as a role model to improve support for the adolescent, we did not assess parent or peer support to physical activity. Lastly, it is important to note that the volume of moderate to vigorous physical activity required to reflect a healthy metabolic or cardiovascular change for those with diabetes and/or prevent diabetes or obesity-related co-morbidities as have been determined in adulthood have yet to be determined in youth. Therefore, one must consider that the evaluation of our adherence measures is based on the current physical activity guidelines.

In conclusion, our study is the first to our knowledge that tracked longitudinal data on adherence to an exercise intervention and examined patterns in adherence over time in adolescents with diabetes or obesity. Our findings highlight that most of these youth can sustain a prescribed level of physical activity for at least 1 month. Thereafter, declines of approximately 15 minutes per week are observed. These declines are more prominent in females and youth with poor metabolic control. Key implications for diabetes educators are to focus on the importance of individualizing the approach to incorporating regular physical activity into daily routines for adolescents who have existing vulnerabilities to more sedentary lifestyles. Although barriers to physical activity tend to be viewed as deterrents, those with the most barriers may in fact be the adolescents who respond best to personalized guidance in establishing a more active lifestyle that aligns with their choices and schedules. Adolescents with T1DM, T2DM, or obesity may begin their physical activity programs as different levels of intensity and duration, but counseling and education to motivate them to remain engaged over time and particularly after the first month is a major component of ongoing care in minimizing potential complications as they transition to adulthood. As there remains a dearth of well-devised clinical trials that identify best evidence for exercise adherence in youth with T1DM or T2DM or who are obese, investigations incorporating incentive and/or behavioral strategies at strategic timepoints to improve exercise adherence are warranted.