INTRODUCTION
Type 1 diabetes (T1D) is a chronic illness in which the pancreas stops producing insulin (American Diabetes Association, 2019). Management of T1D is complex and requires multiple daily adherence behaviors to keep blood glucose levels close to the normal range (e.g., checking blood glucose, counting carbohydrates, calculating and administering insulin, and maintaining healthy diet and exercise; Hood, Peterson, Rohan, & Drotar, Reference Hood, Peterson, Rohan and Drotar2009). The degree to which blood glucose range is maintained across a period of approximately 3 months is reflected in glycolated hemoglobin (HbA1c), with lower numbers reflecting more optimal control. In childhood and adolescence, better adherence and lower HbA1c are associated with greater parental involvement (Berg et al., Reference Berg, Butner, Wiebe, Lansing, Osborn, King, Palmer and Butler2017). However, as adolescents transition into emerging adulthood, parental involvement declines with concomitant increases in HbA1c (King, Berg, Butner, Butler, & Wiebe, Reference King, Berg, Butner, Butler and Wiebe2014; Majumder, Cogen, & Monaghan, Reference Majumder, Cogen and Monaghan2017), making this time period an important focus of research.
Because the T1D regimen is quite complex, it is not surprising that cognitive capacity, most notably executive functioning (EF), plays a role in adherence to T1D regimen and in maintenance of lower HbA1c (Berg et al., Reference Berg, Wiebe, Suchy, Turner, Butner, Munion, Lansing, White and Murray2018; Duke & Harris, Reference Duke and Harris2014; Goethals et al., Reference Goethals, de Wit, Van Broeck, Lemiere, Van Liefferinge, Böhler, De Wulf, Dello, Laridaen, Van Hecke, Van Impe, Casteels and Luyckx2018; McNally, Rohan, Pendley, Delamater, & Drotar, Reference McNally, Rohan, Pendley, Delamater and Drotar2010; Suchy et al., Reference Suchy, Turner, Queen, Durracio, Wiebe, Butner, Franchow, White, Murray, Swinyard and Berg2016, Reference Suchy, Queen, Huntbach, Wiebe, Turner, Butner, Kelly, White, Murray, Swinyard and Berg2017). EF is a complex construct, thought to be comprised of several subdomains (Lezak, Howieson, Bigler, & Tranel, Reference Lezak, Howieson, Bigler and Tranel2012; Miyake, Friedman, Emerson, Witzki, & Howerter, Reference Miyake, Friedman, Emerson, Witzki and Howerter2000; Stuss, Reference Stuss2011; Suchy, Reference Suchy2015), although no single model of EF has been universally adopted (Suchy, Reference Suchy2009, Reference Suchy2015) and none has been consistently replicated in factor-analytic studies (Karr et al., Reference Karr, Areshenkoff, Rast, Hofer, Iverson and Garcia-Barrera2018). Nevertheless, it is generally accepted that EF is comprised of (a) cognitive control processes (sometimes referred to as “executive cognitive functions”), which are responsible for higher order abilities such as planning, reasoning, and problem-solving (Köstering, Leonhart, Stahl, Weiller, & Kaller, Reference Köstering, Leonhart, Stahl, Weiller and Kaller2016; Swanson & Fung, Reference Swanson and Fung2016; Van Hoeck, Watson, & Barbey, Reference Van Hoeck, Watson and Barbey2015), and (b) behavioral control processes (closely related to the concept of “self- or behavioral regulation”), responsible for impulse control, initiation, and maintenance of adaptive actions over time, and selection of adaptive actions under competing emotionally salient contingencies (Panwar et al., Reference Panwar, Rutherford, Mencl, Lacadie, Potenza and Mayes2014; Steinberg, Reference Steinberg2005; Suchy, Reference Suchy2015). Such a dissociation between processes responsible for cognitive versus behavioral control has also been evidenced in research on T1D management. Specifically, Miller et al. (Reference Miller, Rohan, Delamater, Shroff-Pendley, Dolan, Reeves and Drotar2013) reported that behavioral regulation (measured via the self-report measure BRIEF; Guy, Isquith, & Gioia, Reference Guy, Isquith and Gioia2004) increased across early adolescence along with increases in adherence to T1D regimen, whereas no such associations were seen for self-reported cognitive control at that age.
One possible explanation for the lack of an association between T1D management and cognitive control during adolescence may be that parents provide ample scaffolding for prospective planning and problem-solving toward higher adherence. Given that cognitive control is related to intelligence (Friedman et al., Reference Friedman, Miyake, Corley, Young, DeFries and Hewitt2006), and intelligence is partly heritable (Plomin & Deary, Reference Plomin and Deary2015), there may be a considerable overlap between parents’ and teens’ executive cognitive functions, which renders the teens’ cognitive control somewhat superfluous for managing T1D. However, parents have limited control over how well their teens follow through behaviorally, especially when away from home during the day. This relative lack of parental control makes the teens’ own behavioral control a key aspect of T1D management. Once teens transition to adulthood and receive less scaffolding from their parents, their T1D management may become increasingly dependent not only on their behavioral control, but on their cognitive control as well. Consistent with this notion, we have previously found that teens’ IQ (which relates to cognitive control; Friedman et al., Reference Friedman, Miyake, Corley, Young, DeFries and Hewitt2006) was more important than EF performance (comprised of a combination of cognitive and behavioral control measures) in predicting HbA1c among high school seniors living at home with their parents (Suchy et al., Reference Suchy, Turner, Queen, Durracio, Wiebe, Butner, Franchow, White, Murray, Swinyard and Berg2016). In contrast, after teens graduated from high school and became more independent, changes in their HbA1c values were predicted by EF performance beyond IQ. Specifically, we found that (a) the teens’ ability to make adaptive behavioral choices predicted, beyond IQ, changes in HbA1c from baseline to 1-year follow-up (Suchy et al., Reference Suchy, Queen, Huntbach, Wiebe, Turner, Butner, Kelly, White, Murray, Swinyard and Berg2017) and (b) the teen’s performance on an EF composite comprised of both cognitive and behavioral components predicted, beyond IQ, changes in HbA1c (though, surprisingly, not changes in self-reported adherence) 2 years later (Berg et al., Reference Berg, Wiebe, Suchy, Turner, Butner, Munion, Lansing, White and Murray2018).
Taken together, findings suggest (a) differential associations of cognitive versus behavioral control with T1D management and (b) changes in the nature of these associations over time. In other words, different aspects of EF might relate to HbA1c and adherence at different time points during emerging adulthood (i.e., ages 18–25 years). However, this question has not been tested directly. Specifically, cognitive versus behavioral components of EF performance have not been pitted against each other in a single analysis, nor have their influence been examined simultaneously across multiple time points.
The purpose of this study was to test the notion that different aspects of EF contribute to HbA1c and to T1D regimen adherence differentially at different time points of transition from adolescence to emerging adulthood. To that end, we simultaneously examined two aspects of EF (executive cognitive functions vs. behavioral control, assessed during the senior year of high school) as predictors of current and future HbA1c and adherence, measured at four yearly time points. Furthermore, because prior research has suggested that HbA1c increases up to the age of 19 years and then begins to decline (Miller et al., Reference Miller, Rohan, Delamater, Shroff-Pendley, Dolan, Reeves and Drotar2013), we examined EF as a predictor of both linear and curvilinear HbA1c and adherence trajectories from baseline to 3-year follow-up. Lastly, to better isolate discrete aspects of the EF construct, we examined if EF processes predicted HbA1c and adherence beyond IQ and lower order processes that are not executive but are necessary for performance of EF tests (e.g., processing speed and visual scanning). We hypothesized that behavioral control would be necessary for HbA1c and adherence at all four time points, whereas cognitive control would become increasingly important, as over time young adults tend to manage their illness more independently from their parents.
METHOD
Participants
High school seniors with T1D were recruited from outpatient pediatric endocrinology clinics in two southwestern U.S. cities for a longitudinal multisite study investigating the transition from adolescence into emerging adulthood. Youth were eligible if they carried T1D diagnosis for at least 1 year (M time since diagnosis = 7.35 years, SD = 3.88; 94% having diabetes 2 years or more), had English as their primary language, were in their final year of high school, lived with a parent or parent figure at baseline (68.4% lived at home with both, 27.1% with one biological parent; 4.5% with adoptive parents or grandparents), would have regular contact with parents over the subsequent 2 years, and had no condition that would prohibit study completion (e.g., severe intellectual disability and blindness). Adolescents who dropped out of high school were eligible if they met all other criteria. Of the qualifying 507 individuals approached, 301 (59%) agreed to participate. Of those who initially agreed, 247 (82%) enrolled in the study. Reasons for not participating included lack of interest (33%) or being too busy (34%); 20% declined to provide a reason. At one site, the Institutional Review Board (IRB) permitted comparison of those who did versus did not participate. Participants did not differ from nonparticipants in HbA1c values, disease duration, gender, or usage of insulin pump (p > .05). However, participants were slightly younger [M (SD) = 17.77 (.43) vs. 17.91 (.48) years, t(203) = 2.27, p = .02] and more likely to be Hispanic [21% vs. 11%, X 2 (1) = 3.88, p = .049] than nonparticipants. Consistent with the participating clinics, 75.2% of the full sample (N = 247) identified as non-Hispanic White, 14.2% as Hispanic, 4.8% as African American, and the remainder as Asian/Pacific Islander, American Indian, or more than one race. Patients were 17.76 years old on average (SD = .39) at baseline and 60% were female. Participants were assessed at baseline (Time 1) and at annual intervals 1 (Time 2), 2 (Time 3), and 3 years later (Time 4). At Times 2, 3, and 4, samples were 216, 211, and 179, respectively. The most common reason for withdrawing from the study or for skipping one of the yearly assessments was that participants were too busy. Mean disease duration at baseline ranged from 1.09 to 17.98 years (M = 7.35, SD = 3.94), with 19.6% considered early onset (prior to age 7). Illness duration was largely unrelated to other dependent and independent variables (Table 1). While at baseline 100% of participants lived at home, this decreased to 51%, 46.8%, and 47.7% for times 2, 3, and 4, respectively.
Table 1. Zero-order correlations and descriptive statistics for all dependent and independent variables
Notes: ECF, Executive cognitive function composite; BC, Behavioral control composite; PS, Psychomotor speed composite; Duration, disease duration (years).
N = 247.
*p<.05, ** p<.01.
Procedure
This study was approved by the University of Utah and Southwestern University IRBs. Parents and adolescents provided consent or assent; teens provided consent after they turned 18 years. At baseline, participants completed all cognitive measures, an HbA1c assay, and received instructions for completing annual online surveys (including self-report of adherence). Because extreme hyper- and hypoglycemia can affect cognitive performance, blood glucose levels were checked prior to completing cognitive performance measures. If levels were outside the range of 75–400 mg/dl, participants took steps to normalize blood glucose; testing was rescheduled for one participant who could not bring blood glucose into range. Blood glucose levels in the testing session were unrelated to cognition (r values < .13, all p values > .11). Teens were paid $50 for the first two and $75 for the last two annual assessments.
Measures
Executive functions
To assess EF, we administered (1) the Delis–Kaplan Executive Function System (D-KEFS; Delis, Kaplan, & Kramer, Reference Delis, Kaplan and Kramer2001) battery, (2) the Conners’ Continuous Performance Test (CPT-2; Conners, Reference Conners2000), and (3) the Iowa Gambling Task (IGT; Bechara, Reference Bechara2007), which together generated 10 scores: Letter–Number Sequencing Condition completion time (D-KEFS Trail Making Test); Inhibition and Inhibition/Switching completion times (D-KEFS Color-Word Interference Test); Letter and Category Conditions correct responses (D-KEFS Verbal Fluency Test); the total correct response from the three conditions for the D-KEFS Design Fluency Test; the total number of selections from the “advantageous” decks on the IGT; and an arithmetical mean of eight indices of inattention on the CPT-2. All scores were age-corrected scaled scores or age-corrected T-scores, per respective test manuals.
The 10 scores were submitted to an exploratory factor analysis with Varimax rotation, yielding two theoretically cogent factors. Factor 1 (41.5% variance) included all verbal and design fluency scores (factor loadings ranged from .551 to .809) and was interpreted as “Executive Cognitive Functions,” based on the fact that both verbal and design fluencies require reasoning and problem-solving and depend heavily on higher order cognitive processes such as goal-directed retrieval of information and mental flexibility, which are recognized aspects of executive cognitive functions (Suchy, Reference Suchy2015). Factor 2 (12.1% variance) included IGT,Footnote 1 CPT-2, the two Color-Word Interference scores, and the Letter–Number Sequencing score (factor loadings ranged from .556 to .677) and was interpreted as “Behavioral Control,” given that these tasks do not require reasoning and problem-solving and rely heavily on behavioral response selection, inhibition of prepotent or emotionally attractive responses, and initiation and maintenance of responses (Suchy, Reference Suchy2015).
Lower order cognitive processes
To assess lower order processes needed for performance of the above EF tasks (e.g., graphomotor speed, verbal speed, visual scanning, and sequencing), we used six tasks included in the D-KEFS battery for the purpose of controlling for lower order processes, including Word Reading and Color Naming from the Color-Word Interference Test, and Visual Scanning, Letter Sequencing, Number Sequencing, and Motor Speed from the Trail Making Test. As we did previously (Franchow & Suchy, Reference Franchow and Suchy2015, Reference Franchow and Suchy2017; Suchy et al., Reference Suchy, Turner, Queen, Durracio, Wiebe, Butner, Franchow, White, Murray, Swinyard and Berg2016), we created a psychomotor speed composite by computing the mean of age-corrected scaled scores (based on test manual). Cronbach’s α in this sample was .816.
General IQ
IQ estimate (IQ-est) was generated using the Vocabulary subtest from the Wechsler Adult Intelligence Scale-Fourth Edition (Wechsler, Reference Wechsler2008). This subtest correlates .91 to .92 with verbal IQ and .79 to .81 with Full-Scale IQ in the present age group. Reported split-half reliability for the Vocabulary subtest for ages 16 to 19 in the normative sample is .93. Norm-based age-corrected scaled scores were used in analyses.
HbA1c
Glycated hemoglobin (HbA1c) was obtained at each time point using HbA1c mail-in kitsFootnote 2 (provided and processed by CoreMedica Laboratories, accredited by the College of American Pathologists; www.coremedica.net). HbA1c reflects average levels of blood glucose over the preceding 3–4 months, with lower scores indicating better blood glucose levels. At baseline, the kit was completed by the teen after receiving instructions from a trained research assistant who observed test completion. This approach was chosen over obtaining HbA1c from medical records to ensure that the HbA1c measurement at baseline occurred on the day of cognitive testing for all participants and to obtain HbA1c values from participants at follow-ups, even if they were not routinely seeing a health care provider.
Adherence
Participants completed a revised seven-item version of the Self-Care Inventory (SCI; La Greca, Reference La Greca1990) to assess adherence to the diabetes regimen. With the assistance of a certified diabetes educator and a pediatric endocrinologist, items from the SCI that reflected current recommendations for daily diabetes behaviors were selected. Participants rated how well they followed recommendations for each behavior on a scale from 1 = “never did it” to 5 = “always did as recommended, without fail.” Scores were averaged to reflect participants’ adherence over the past month, such that higher scores indicate better adherence. In this sample, Cronbach’s α across the four time points ranged from .77 to .83).
Analytic Plan
As indicated in Participants section, between 72% and 87% of initially enrolled participants participated in follow-ups 2 to 4. To account for missing data, we generated ten datasets through multiple imputation (MI) (Graham, Young, & Penny, Reference Graham, Young and Penny2009), estimating missing data both for participants missing a variable at a particular time point and for individuals missing an entire time point (for individuals who remained in the study but skipped one of the four assessments). The MI procedure included variables beyond those used in present analyses to help ensure an adequate missing-at-random model.
As a preliminary analysis, we generated zero-order correlations among all dependent and independent variables and potential covariates. To examine cognitive and behavioral aspects of EF simultaneously, we initially constructed a linear and a quadratic growth curve model of the four measures of HbA1c and adherence over time in Mplus v8 using maximum likelihood with the Yuan and Bentler adjustment for nonnormality. We then examined whether the cognitive and behavioral aspects of EF uniquely predicted the intercept and the slopes of HbA1c and adherence. Lastly, we followed up with four general linear regressions using the HbA1c and adherence values at each of the four time points individually as the dependent variables and the cognitive variables as predictors. In these analyses, we examined the unique and overlapping variance contributions of each predictor by employing hierarchical variable entry.
RESULTS
Preliminary Analyses
Zero-order correlations and descriptive statistics presented in Table 1 show that all cognitive variables correlated with HbA1c at all four time points. Also, consistent with the notion of positive manifold (i.e., better cognition in one area tends to be related to better cognition in another area; Spearman, Reference Spearman1904), all cognitive variables evidenced small-to-moderate associations with each other, demonstrating the need to control for IQ-est and lower order processes to isolate the EF construct. Of note, as previously reported by others (Friedman et al., Reference Friedman, Miyake, Corley, Young, DeFries and Hewitt2006), stronger association was seen between IQ-est and Executive Cognitive Functions as compared to IQ-est and Behavioral Control (R 2Δ = .19, p < .001). Lastly, adherence did not correlate with any of the cognitive variables, with the exception of a small correlation between IQ-est and adherence at time 4.
Cognitive and Behavioral Control Predicting HbA1c and Adherence over Time
In the growth model using HbA1c as the dependent variable, time was centered between the 2nd and 3rd measures to minimize collinearity due to scaling of time. Both linear (−1.5, −0.5, 0.5, and 1.5) and quadratic (2.25, 0.25, 0.25, and 2.25) slopes were tested, and the residuals for HbA1c were equated over time. Overall, the model was a good fit for the data, Average X 2(df = 4, n = 233) = 7.567, SD X 2 = 2.535. For the unconditional model, significant parameters were observed on all the growth parameters (see Figure 1) suggesting that, on average, individuals were increasing in HbA1c over time while this increase attenuated as time continued. The variance components for the intercept, linear, and quadratic slopes suggested that these patterns varied substantially across individuals.
Fig. 1. The unconditional model significant parameters.
Executive Cognitive Functions, Behavioral Control, IQ-est, and Psychomotor Speed were then added to simultaneously predict the intercept, as well as the linear and quadratic slopes (see Table 2 for coefficients). Only Behavioral Control uniquely predicted the intercept of HbA1c, with higher Behavioral Control related to lower HbA1c. Furthermore, only Executive Cognitive Functions related to the linear slope, with better cognitive control related to a slower increase in HbA1c over time. None of the variables tested accounted uniquely for variation in the quadratic slope.Footnote 3Figure 2 shows trajectories at one standard deviation above and below the mean in cognitive and behavioral control scores. Table 3 shows norm-referenced scaled scores for each of the four cognitive variables used in analyses.
Table 2. Unstandardized coefficients (Z-values) in predicting the Quadratic Growth Model of HbA1c
* p < .05.
N = 247.
Fig. 2. Predicted growth curve of HbA1c at 1 SD above and below of Executive Cognitive Functions and Behavioral Control. As seen in the figure, best outcomes are seen for individuals who have both good Behavioral Control and good Executive Cognitive Functions, whereas the poorest outcomes are seen for individuals who have weaknesses in both areas. Of note, interaction effects were not tested, as they were not hypothesized.
Table 3. Means, standard deviations, and ranges for dependent variables by group expressed as norm-referenced scaled scores
Notes: Scaled score in normative sample M = 10, SD = 3; “Low” = 1 SD or more below the sample mean, “High” = 1 SD or more above the sample mean, “Middle” = within + 1 SD around the sample mean.
N = 247.
Using the same growth modeling procedure, we treated adherence (operationalized as the SCI score) as the dependent variable. Both the fixed effect and random effect for the quadratic slope failed to meet conventional significance (fixed = 0.000, p = .994, variance component = 0.199, p = .993). Furthermore, inclusion of Executive Cognitive Functions, Behavioral Control, IQ, and Psychomotor Speed failed to predict the linear slope (though Executive Cognitive Functions and Behavioral Control uniquely predicted the intercept with higher scores corresponding to higher adherence). These results did not change as a function of including or excluding the quadratic slope (nor did Executive Cognitive Functions, Behavioral Control, IQ, and Psychomotor Speed predict the quadratic slope).Footnote 4
Unique Contributions of Cognitive and Behavioral Control to HbA1c at Each Time Point
To examine specific contributions of each cognitive variable to HbA1c at each time point, we further conducted a series of general linear regressions, using HbA1c at each time point as the dependent variables and the four cognitive variables as predictors.
For Time 1, general linear regression model was statistically significant [F(4,228) = 8.46, p ≤ .001] and accounted for 13.5% of variance in HbA1c. Behavioral Control (β = −.204, p = .009) and IQ-est (β = −.254, p = .001) emerged as a unique predictor beyond all other variables.
For Time 2, the model was statistically significant [F(4,228) = 5.56, p ≤ .001] and accounted for 8.6% of variance in HbA1c. Behavioral Control (β = −.228, p = .007) emerged as the sole unique predictor beyond all other variables.
For Time 3, the model was statistically significant [F(4,228) = 4.85, p = .001] and accounted for 7.6% of variance in HbA1c. Behavioral Control (β = −.181, p = .049) and Cognitive Control (β = −.191, p = .033) emerged as unique predictors above and beyond all other variables.
For Time 4, the model was statistically significant model [F(4,228) = 10.44, p < .001] and accounted for 14.7% of variance in HbA1c. Behavioral Control (β = −.257, p = .004) and Cognitive Control (β = −.354, p = .002) again emerged as unique predictors beyond all other variables.
Partitioning of variance is illustrated in Figure 3. As seen, Executive Cognitive Functions emerged with time as a progressively more important unique predictor, whereas Behavioral Control consistently contributed unique variance across time. Importantly, contributions of Executive Cognitive Functions and Behavioral control held above and beyond IQ-est and psychomotor speed, suggesting that the EF construct, not other aspects of cognition, explain the results.
Fig. 3. Percent of unique and overlapping variance accounting for in HbA1c by cognitive variables. As seen in the figure, IQ, executive cognitive functions, and behavioral control each contributed unique variance above and beyond the other variables at least at some time points. As would be expected, psychomotor speed did not account for any unique variance; it was included in the model simply to ensure that executive function scores are not confounded by lower order cognitive processes that are necessary for performance of most executive function tasks.
Lastly, the above analyses were not conducted with adherence as the dependent variable, given that lack of meaningful associations in the preliminary analyses.
DISCUSSION
The present study examined two components of EF (Executive Cognitive Functions and Behavioral Control) as predictors of HbA1c and adherence over four yearly assessments among young adults with T1D as they transitioned from the senior year of high school into early emerging adulthood. The results showed that Behavioral Control is consistently related to HbA1c over time, accounting for a small amount of unique variance beyond other cognitive variables. In contrast, Executive Cognitive Functions are related to the trajectory of HbA1c values from baseline to Time 4, becoming increasingly more prominent as a correlate of HbA1c with increasing transition to adulthood. Although IQ-est did not significantly factor into changes across time, it is notable that at baseline (i.e., in senior year of high school) it accounted for significant amount of unique variance beyond all the other cognitive variables in a linear regression. The present study contributes to the growing literature on the association between EF and HbA1c among youth with T1D and extends prior findings by demonstrating that different components of EF contribute differentially to HbA1c at different periods of development.
In this study, no association was found between the trajectories of self-reported adherence and cognition. Emerging adults’ reports of adherence did not vary greatly across the 3 years (see also Berg et al., Reference Berg, Wiebe, Suchy, Turner, Butner, Munion, Lansing, White and Murray2018), likely making it difficult to predict the small changes that did occur. Additionally, emerging adults’ self-reports of adherence are only modestly related to other metrics of adherence such as blood glucose checks (Berg et al., Reference Berg, Butner, Turner, Lansing, King and Wiebe2016). Future work utilizing additional metrics of adherence is needed to fully understand the link between EF and trajectories of adherence across time.
The present results provide a theoretically cogent account of how cognition relates to HbA1c during this important developmental period. First, during the senior year of high school while all teens were still living at home with parents, HbA1c was, in part, explained by IQ-est. Teens’ IQ likely reflects a broad array of factors including the teens’ own self-control (Berg et al., Reference Berg, Hughes, King, Korbel, Fortenberry, Donaldson, Foster, Swinyard and Wiebe2014), along with the broader family environment and parental involvement. Research also shows that parental education and socioeconomic status (SES) are related to teens’ T1D management (Rechenberg, Whittemore, Grey, & Jaser, Reference Rechenberg, Whittemore, Grey and Jaser2016). Because children’s IQ is known to relate to parents’ IQ (Plomin & Deary, Reference Plomin and Deary2015) and maternal IQ is associated with children’s HbA1c beyond SES (Ross et al., Reference Ross, Frier, Kelnar and Deary2001), the association between teens’ IQ and their HbA1c may be more a reflection of the family environment than of the importance of teens’ IQ for their T1D management. This interpretation is particularly likely given that IQ drops out as a predictor once teens graduate from high school and begin to assume more independent lifestyle. Of note, although 100% of the teens lived at home at baseline, nearly half the sample continued to live at home for all three follow-ups, illustrating that living situation is but one aspect of independence. A more thorough assessment of independence beyond living situation was unfortunately not conducted.
Unlike IQ-est, Executive Cognitive Functions were not uniquely predictive of HbA1c while teens still lived at home. This likely reflects that parents provide structure for their teens that may simplify the cognitive demands of diabetes management. Such structure may consist of parents monitoring the teen’s glucose checks and food, or by helping them problem-solve around potential obstacles to adherence. However, as young people assume greater independence from their parents and become responsible for organizing their own daily life, Executive Cognitive Functions become progressively more important. Importantly, as is apparent from the trajectories illustrated in Figure 2, Executive Cognitive Functions are necessary but not sufficient for diabetes management. To execute intended plans, Behavioral Control is needed.
Behavioral Control reflects the capacity to (a) say no to attractive but unwise choices or to avoid giving in to impulses and urges (e.g., avoid consuming unhealthy foods or alcohol), (b) repeatedly and regularly execute daily tasks (e.g., glucose checks), and (c) follow through with intentions and plans (e.g., exercise). In line with the notion that Executive Cognitive Functions are necessary but not sufficient to effectively manage T1D, the present study suggests a fairly stable association between HbA1c and Behavioral Control across all four time points. This suggests that individual differences in impulse control, emotional decision-making, and the ability to initiate and complete planned tasks relate to how well teens are able to maintain good HbA1c. These results are in line with Steinberg’s (Steinberg, Reference Steinberg2005) model of the importance of impulse control for the successful execution of adaptive behaviors across this developmental period. In fact, Wiebe et al. (Reference Wiebe, Baker, Suchy, Stump and Berg2018) found that individuals who generally planned well for how to achieve diabetes management goals were derailed when others interfered with diabetes management in daily life, but this was primarily true for those with lower EF (measured as a combination of cognitive and behavioral control skills).
The different trajectories of HbA1c predicted by high and low cognitive and behavioral components of EF (see Figure 1) add to a growing literature indicating that individuals vary in their HbA1c across late adolescence and emerging adulthood. Longitudinal studies of HbA1c across emerging adulthood reveal that some individuals increase in HbA1c, whereas others maintain fairly stable levels (Helgeson et al., Reference Helgeson, Vaughn, Seltman, Orchard, Libman and Becker2018; Luyckx & Seiffge-Krenke, Reference Luyckx and Seiffge-Krenke2009; Schwandt et al., Reference Schwandt, Hermann, Rosenbauer, Boettcher, Dunstheimer, Grulich-Henn, Kuss, Rami-Merhar, Vogel and Holl2017). A large number of factors have distinguished those with more stable versus worsening HbA1c, including SES, psychological distress (Helgeson et al., Reference Helgeson, Vaughn, Seltman, Orchard, Libman and Becker2018), and family climate and teen self-concept (Luyckx & Seiffge-Krenke, Reference Luyckx and Seiffge-Krenke2009). Additionally, teens who leave home are at an increased risk for the development of depressive symptoms, which are themselves associated with poorer HbA1c (Baucom, Turner, Tracy, Berg, & Wiebe, Reference Baucom, Turner, Tracy, Berg and Wiebe2018) and with EF weaknesses (Rogers et al., Reference Rogers, Kasai, Koji, Fukuda, Iwanami, Nakagome, Fukuda and Kato2004). The present results indicate that those with poorer cognitive and behavioral control are also at risk for exhibiting detrimental HbA1c trajectories across the emerging adult years. Combined weaknesses in both the cognitive and behavioral components of EF may mean that such individuals lack both the cognitive skills to engage in the complex problem-solving required for more independent diabetes management and the impulse control to prevent oneself from engaging in high-risk behaviors (e.g., drinking, skipping blood glucose tests). We should note that, as seen in Table 3, our participants experiencing low cognitive and behavioral control were falling in a category that would be considered low average to mildly impaired (rather than clinically significantly impaired), with the mean values for the low groups falling at about 2/3 of SD below norms (i.e., mean scaled scores around 8), with some individuals falling more than 1 SD below norms (i.e., minimum scaled scores around 6). It is likely that individuals with clinically significant impairments in cognitive and behavioral components of EF would display even greater deterioration in HbA1c across this high-risk developmental period.
The present findings need to be interpreted in the context of limitations. First, although this study conducted highly comprehensive assessments of EF, we did not assess all aspects of EF, especially the ability to manage multiple tasks simultaneously (Suchy, Reference Suchy2015). Relatedly, we did not assess prospective memory, which is known to relate to independent daily functioning, including management of complex medication regimens (Avci et al., Reference Avci, Sheppard, Tierney, Kordovski, Sullivan and Woods2018). Second, we only assessed EF at baseline, even though aspects of EF continue to mature across this transitional period, and deterioration in HbA1c as well as fluctuations in daily blood glucose may also affect cognition (Gonder-Frederick et al., Reference Gonder-Frederick, Zrebiec, Bauchowitz, Ritterband, Magee, Cox and Clarke2009). In fact, as recently reviewed, those with early T1D onset tend to have poorer cognition including EF, which is attributable to both hypo- and hyperglycemia (Ryan, van Duinkerken, & Rosano, Reference Ryan, van Duinkerken and Rosano2016). Future research needs to assess these components of EF across time to understand the complex interplay between developmental changes in EF as well as the deleterious impact of dysglycemia on the specific aspects of cognition examined in this study. Third, the majority of teens in the present study had parents with some college education, limiting the generalizability of the results. However, given that EF is negatively affected by low SES (Hackman, Gallop, Evans, & Farah, Reference Hackman, Gallop, Evans and Farah2015), the results may be even stronger in a more economically and educationally diverse sample. Fourth, the failure to find an association between adherence and cognition limits our interpretation that cognition affects HbA1c through health behaviors. Adherence behaviors are challenging to measure in the T1D self-management context, and different findings may occur with additional or alternative measures.
The findings of the present study have important implications for promoting T1D management as late adolescents transition into emerging adulthood. Assessment of both cognitive and behavioral aspects of EF in clinical care may be necessary to identify who is in need of additional support and skill-building interventions, including help with identifying the social contexts that pose barriers to T1D management and development of strategies for preventing lapses in diabetes self-management in those situations (e.g., if-then planning; Gollwitzer & Oettingen, Reference Gollwitzer, Oettingen, Vohs and Baumeister2011). These individuals may also benefit from efforts to identify new social support resources, especially in risky situations away from parents (Wasserman, Hilliard, Schwartz, & Anderson, Reference Wasserman, Hilliard, Schwartz and Anderson2015). Finally, the findings that cognitive control predicts changes in HbA1c across time and becomes increasingly important as emerging adults mature suggest the ongoing need to evaluate EF and to target interventions that compensate for lower cognitive skills (e.g., structured planning and problem-solving skills).
Lastly, the present findings have implications for neuropsychological practice, as related to the utility of EF assessment for prediction of daily functioning. Specifically, the present findings are consistent with the Contextually Valid Executive Assessment (ConVExA) model, which posits that univariate association between EF and outcomes cannot always be assumed, as this association may be moderated by a variety of contextual factors (Suchy, Reference Suchy2015, pp. 158–159; Reference Suchy2019). By identifying the specific contextual factors as they relate to individual patients, tailored predictions can identify who, and under what circumstances, is at particular risk for lapses in functioning.
ACKNOWLEDGEMENTS
This research was supported by a grant from the National Institute of Diabetes and Digestive and Kidney Diseases, R01 DK092939. We thank the staff at participating endocrinology clinics, the READY Study research team, and the late adolescents who took time from their busy schedules to participate.
CONFLICT OF INTEREST
The authors had no conflict of interest when conducting and reporting this research.
