Resistance training is associated with spontaneous changes in aerobic physical activity but not overall diet quality in adults with prediabetes

Abstract

Background:

Aerobic exercise interventions have been shown to result in alterations to dietary intake and non-exercise physical activity (PA). To date, the ability for resistance training (RT) to influence other health-related behaviors has not been examined. This study aimed to determine if initiation and maintenance of RT is associated with spontaneous changes in dietary quality and non-RT PA in adults with prediabetes.

Methods:

Overweight/obese adults (n = 170, BMI = 32.9 ± 3.8 kg·m², age = 59.5 ± 5.5 years, 73% female) with prediabetes were enrolled in the 15-month Resist Diabetes trial. Participants completed a supervised 3-month RT initiation phase followed by a 6-month maintenance phase and a 6-month no-contact phase. Participants were not encouraged to change eating or non-RT PA behaviors. At baseline, and months 3, 9, and 15, three 24-hour diet recalls were collected to evaluate dietary intake and quality, the Aerobics Institute Longitudinal Study Questionnaire was completed to evaluate non-RT PA, and body mass, body composition (DXA), and muscular strength were measured. At months 3, 9, and 15 social cognitive theory (SCT) constructs were assessed with a RT Health Beliefs Questionnaire. Mixed effects models were used to assess changes in dietary intake and non-RT PA over the 15-month study period.

Results:

Energy and carbohydrate intake decreased with RT initiation and maintenance phases (baseline to month 9: β = −87.9, p = 0.015 and β = −16.3, p < 0.001, respectively). No change in overall dietary quality (Healthy Eating Index [HEI]-2010 score: β = −0.13, p = 0.722) occurred, but alterations in HEI-2010 sub-scores were detected. Maintenance of RT was accompanied by an increase in MET-min/week of total non-RT PA (month 3 to month 9: β = 146.2, p = 0.01), which was predicted by increased self-regulation and decreased negative outcome expectancies for RT (β = 83.7, p = 0.014 and β = −70.0, p = 0.038, respectively).

Conclusions:

Initiation and maintenance of RT may be a gateway behavior leading to improvements in other health-related behaviors. These results provide rationale for single-component lifestyle interventions as an alternative to multi-component interventions, when resources are limited.

1. Introduction

Prediabetes prevalence has increased in recent decades and current estimates indicate over 86 million adults in the United States are classified as having this state of intermediate hyperglycemia [1,2]. Prediabetes is the obligate precursor state to overt type 2 diabetes mellitus (T2DM), and the majority of individuals with prediabetes progress to T2DM [3]. Combined, these hyperglycemic conditions are associated with increased risk of macro- and micro-vascular complications [4] and increased medical costs [5]. Modifiable lifestyle factors such as physical inactivity, obesity, and poor dietary quality have been established as key factors in the both development of prediabetes, and the transition to T2DM [6]. Middle-aged and older adults are at increased risk for developing prediabetes due to age-related muscle loss, increased fat mass, and alterations in glucose handling [7]. In addition, this segment of the population is least likely to meet physical activity (PA) guidelines, particularly the resistance training (RT) recommendation of completing a whole body routine 2×/week [8,9]. Reversion to normoglycemia and/or preventing prediabetes from developing into T2DM is strongly indicated in this population. Thus, interventions that improve the initiation and maintenance of multiple health-related behaviors and decrease metabolic disease progression in this population are needed [10].

Improving health-related behaviors is challenging, and optimal strategies to promote and maintain changes have yet to be determined. Furthermore, effective behavior change interventions tend to be time-, cost-, and resource-intensive, limiting the ability for efficacious programs to be broadly translated into community or clinical settings [11]. In addition, multi-component interventions can be burdensome for participants, leading to increased barriers, reduced adherence, and greater susceptibility to relapse [11-13]. Evidence suggests that health-related behaviors, particularly diet and exercise habits, tend to cluster together; individuals that do not meet PA guidelines, also generally have poor dietary habits, and those who meet PA guidelines are more likely to have healthy dietary habits [14]. Thus, intervening on one behavior (e.g. PA) may lead to a “spillover effect” and result in alterations to other behaviors (e.g. dietary intake) [15,16]. If so, this phenomenon could be utilized in order to develop effective, yet time-, cost-, and resource-efficient interventions focusing on alterations of a single (vs. multiple) health behavior.

To date, findings related to the potential for exercise interventions to produce changes in dietary intake and other forms of PA are mixed, and have focused primarily on aerobic exercise [17,18]. The potential for RT to exert a spillover effect on other health-related behaviors has received little attention. Preliminary results have provided encouraging evidence that RT may be a unique mode of exercise in its ability to influence dietary intake and non-RT PA [15,19-22]. We have previously reported that initiation of a RT intervention (e.g. Initiation Phase of the Resist Diabetes trial [23]) was associated with a reduction in reported energy, carbohydrate, total sugar, fruit and vegetable, and sweets and dessert intake over a 3 month period [15]. Similarly, Bales et al. have shown that an 8 month RT program is associated with decreased self-reported fat intake in overweight/obese adults with dyslipidemia [20]. This suggests that dietary modifications associated with successful initiation of an exercise intervention may be specific to the disease state individuals are at risk of developing. Supporting the notion that RT may increase non-RT PA, using objective measures of PA energy expenditure, Hunter et al. reported that free living PA increased in a group of women assigned to RT during a calorie-restricted weight loss intervention [21]. However, the participants were young and relatively healthy, and it is currently unknown if RT would alter non-RT PA in middle-aged and older, less healthy adults, or among those with prediabetes.

To date, no evaluation of changes to overall diet quality in response to exercise interventions have been conducted. The objective of this investigation was to determine if participation in a social cognitive theory (SCT)-based RT program targeting the initiation and maintenance of RT exerts a spillover effect on other health behaviors. Specifically, we aimed to determine if RT initiation and maintenance are associated with alterations in overall diet quality (Healthy Eating Index [HEI]-2010 scores), and total non-RT PA in a population at risk for T2DM. A secondary aim was to explore demographic, physiological, and psychosocial factors which may predict spontaneous changes in health behaviors, with RT adoption.

2. Materials and methods

2.1. Participants

Overweight/obese (BMI: 25–39.9 kg/m²), middle-aged and older (50–69 years) adults who were sedentary to recreationally active and had not engaged in RT for the previous year were recruited from the Roanoke, Virginia area. Detailed inclusion and exclusion criteria have previously been published [23]. Briefly, eligible individuals had prediabetes, defined as impaired fasting glucose (IFG; fasting plasma glucose between 95 and 125 mg/dl [24]) and/or impaired glucose tolerance (IGT; blood glucose between 140 and 199 mg/dl 2-hours following a 75 gram oral glucose tolerance test [OGTT]) and received clearance from their personal physician. Individuals were excluded if they had a diagnosed cardiovascular, metabolic, pulmonary, liver, or kidney disease. In addition, current smokers, individuals taking medications known to influence food choice or energy intake, and individuals with conditions that restricted their ability to be physically active and engage in RT (e.g. orthopedic limitations) were also excluded. Individuals taking other commonly prescribed medications, such as those used to treat dyslipidemia or hormone replacement therapy, were eligible for participation provided that they had been on a stable dose of the medication and weight stable for an extended period of time (e.g., > 1 year). Prior to enrollment, individuals were fully informed about the study procedures and provided written informed consent.

2.2. Design

Resist Diabetes was a 15-month randomized controlled trial focusing on the initiation and maintenance of meeting RT guidelines (training all major muscle groups 2×/week) [25]. All study procedures were approved by the Virginia Tech Institutional Review Board (see Marinik et al. [23] for detailed methods). A study overview is presented in Fig. 1. Following baseline testing, all participants followed the same 3-month Initiation Phase that consisted of 2×/week supervised RT with an American College of Sports Medicine (ACSM)-certified personal trainer in a lab-gym using Nautilus Nitro Plus resistance training equipment. The progressive RT protocol consisted of a whole-body routine targeting major muscle groups, with twelve exercises per session. Participants completed one set of each exercise to concentric failure. This protocol is consistent with RT guidelines from the ACSM for novice lifters, and particularly for older adults (e.g. ≥ 65 years of age) and those 50–64 years of age with clinical conditions (e.g. prediabetes), and has been shown to result in skeletal muscle hypertrophy and increased strength [26], while being time efficient (~35–45 min per session). Training records of each RT session were maintained.

Fig. 1.

Study overview.SCT: Social Cognitive Theory Group; STD: Standard Group.Note: For this analysis, SCT and STD groups were combined because there was no main effect of intervention group on either diet quality or non-RT PA. Furthermore, the intervention conditions were similar in that both groups received the same 3-month initiation phase.

Participants who successfully completed the initiation phase (e.g. completed ≥ 70% of scheduled RT sessions) entered the 6-month Maintenance Phase and were randomly assigned to one of two different intervention maintenance conditions: a social cognitive theory (SCT)-based intervention involving more frequent and personalized contact with study staff and enhanced RT-tracking features and tailored feedback on the Resist Diabetes website; or a standard, usual care condition involving minimal contact with study staff (STD) and limited RT-tracking features and generic feedback on the Resist Diabetes website. While less intensive, the STD condition was informed by the SCT (i.e. ability to schedule and track RT sessions are a form of self-regulation). During the maintenance phase, participants were to continue the RT program on their own in a community/public workout facility. Finally, participants in both conditions completed a 6-month No Contact Phase in which they did not interact with study staff, but were expected to continue the 2×/week RT protocol. Participants retained access to the same website features associated with their intervention conditions (SCT or STD) during this phase.

2.3. Outcome assessments

Assessments occurred at baseline, and following the initiation, maintenance, and no-contact phases (e.g. months 3, 9, and 15, respectively). Testing occurred over 2 days. Following completion of each assessment phase, participants were mailed a packet containing their body mass and composition, strength testing results, and a summary of their reported dietary intake (as well as other assessments completed as part of the main Resist Diabetes trial [23,27]). In addition, a categorical chart of normative or ideal results for body mass and composition was provided as well as an educational handout containing standard nutrition information from the 2010 Dietary Guidelines for Americans (DGA) [28]. Participants were not encouraged to make alterations to their dietary intake or non-RT PA and no information on these health-related behaviors was provided by study staff.

2.3.1. Dietary intake

Habitual dietary intake was assessed using the average of three multiple-pass 24-hr recalls [29,30] collected at each assessment time point by a research dietitian/technician. The first recall was completed in-person at the assessment visit, using 2D food models to aid in serving size estimations. The 2nd and 3rd recall were unannounced, and completed by phone within 2 weeks, and included 1 weekend day. Previous trials have found no difference in reported energy intakes between 24-hr food recalls completed in-person or via telephone [31,32]. Food recalls were analyzed using the Nutrition Data System for Research software (NDS-R 2010, University of Minnesota, Minneapolis, MN). HEI-2010 total and component scores [33] were calculated from NDS-R output files following established protocols [34]. The HEI-2010 total score is a measure of overall diet quality as determined by adherence to the 2010 DGA [28,33]. The HEI-2010 total score is the sum of 12 component scores, each reflecting intake of specific food groups or nutrients. Nine components (total fruit, whole fruit, total vegetables, dark-green vegetables and beans, whole grains, dairy, total protein foods, seafood and plant proteins, and fatty acid ratio) are included in the adequacy category, meaning individuals should eat enough of these foods/nutrients. Three components (refined grains, sodium, and empty calories [e.g. solid fats, alcohol, and added sugar (SoFAAS)]) are included in the moderation category, meaning individuals should limit consumption of foods containing these nutrients [33]. Component scores range from 0–5 (total fruit, whole fruit, total vegetables, dark-green vegetables and beans, total protein foods, and seafood and plan proteins), 0–10 (whole grains, dairy, and fatty acid ratio), or 0–20 (empty calories), and an HEI-2010 total score of 100 indicates perfect conformity to 2010 DGA [28]. HEI-2010 scores can be divided into three categories based on dietary quality: good (> 80), needs improvement (51–80), or poor (< 51) [33].

2.3.2. Non-RT Physical Activity

Prior to each scheduled assessment phase, PA over the previous 3 month period was assessed online via the Aerobics Center Longitudinal Study Questionnaire [35]. This questionnaire consists of questions on type, frequency, and intensity of PA participants may have performed, and it and similar versions have been validated against a treadmill test of aerobic capacity, accelerometry, and PA records [35]. To assess non-RT PA, responses to the RT specific question were not included in the calculations of total non-RT PA. Non-RT PA was categorized as low (< 3 Metabolic Equivalents [METs]), moderate (3–6 METs), or vigorous-intensity (> 6 METs) [36]. Total MET-minutes/week of all non-RT PA, low-intensity PA, moderate-intensity PA, and vigorous-intensity PA were calculated by multiplying frequency (sessions/week) * duration (the number of minutes/session) * intensity (assigned MET-value) for each activity in order to determine if participants were meeting U.S. physical activity guidelines of 500–1000 MET-mins/week [37].

2.3.3. Anthropometrics

Height was measured without shoes to the nearest 0.1 cm using a wall-mounted stadiometer. Body mass was measured in light clothing, without shoes, to the nearest 0.1 kg using a digital scale (Healthometer ProPlus, Pelstar, McCook, IL). BMI was calculated as weight (kg)/height (m²). Fat mass (FM) and fat free mass (FFM) were assessed via dual energy X-ray absorptiometry (DXA; GE Lunar Prodigy, software version 11.40.004, Madison, WI).

2.3.4. Muscular strength

Three repetition maximum (3RM) on the chest press and horizontal leg press machines (Nautilus Nitro Plus) were used to evaluate upper and lower body strength, respectively. Testing was conducted in accordance with ACSM guidelines [25]. Briefly, following orientation and familiarization with movements, participants completed warm-up sets. Participants rested between 3RM trials and resistance was increased until rate of perceived exertion on the last repetition was rated as a 9 or 10 on the Borg CR10 scale [38] and participants were unable to perform 3 or more repetitions using proper form.

2.3.5. Prediabetes status

Following an overnight fast, blood samples were collected from the antecubital vein in the fasting state and 120 min following consumption of a 75 g orange-flavored glucose beverage (Fisherbrand, Fisher Scientific, Hanover Park, IL). Plasma was collected in EDTA BD vacutainers and immediately placed on ice until centrifuged at 2000g for 15 min (Model 5702R, Eppendorf, Hauppauge, NY) for sample separation. Plasma glucose concentrations were determined using a YSI 2700 Select glucose analyzer (YSI Life Sciences, Yellow Springs, OH) on the day of the assessment visit. Participants were classified according to pre-diabetes phenotype status as 1. IFG only, 2. IGT only, or 3. IFG and IGT [24].

2.3.6. RT health beliefs

Following the initiation phase, but prior to randomization to intervention maintenance conditions, as well as prior to the 9- and 15-month assessment visits, participants completed the RT Health Beliefs survey [39,40]. Measures included: outcome expectancies; behavioral resolve (e.g. self-efficacy for overcoming barriers); and self-regulation related to RT. The scales have demonstrated adequate internal consistency (α = 0.729–0.925) and moderate predictive validity of self-reported RT participation (r = 0.296–0.506) [39].

The outcome expectancy scale was scored as the mean of all positive items (n = 10) and negative items (n = 6), separately with a scale of 1–7 as well as the mean of the outcome expectancy scale times the outcome value (for both positive and negative items separately), which has a scale of 1–49. For the positive scale, a higher score indicates more positive outcome expectancies. For the negative scale, a higher score indicates more negative outcome expectancies associated with RT. The behavioral resolve scale was scored as the mean of all items (n = 12) with a scale of 1–100. A higher score indicates a greater perceived ability to overcome barriers associated with RT. The self-regulation scale was scored as the mean of all items (n = 5) with a scale of 1–7. A higher score indicates a greater number of strategies are often used in order to maintain a regular RT regimen.

2.3.7. Adherence to RT protocol

A timeline follow back (TLFB) approach [41] was used to assess RT adherence at the 9 and 15 month assessment visits [23,42]. This was selected as our primary measure of RT adherence rather than participant tracking on the website as our prior research indicates participants may decrease use of a website for planning and reporting activity over extended periods (11–15 months) and we wanted to ensure we were capturing RT adherence, not website adherence [43]. At the assessment visits participants were provided with a paper calendar in which they were to indicate each day a RT session occurred within the past 30 days. Since the purpose of the Resist Diabetes intervention was to initiate and maintain a 2×/week RT protocol, 8 completed sessions per 30-day period were expected.

2.4. Statistical analysis

Statistical analyses were performed with Stata, version SE 14 (StataCorp LP, College Station, TX). Analyses included descriptive statistics (means, standard deviations, and frequencies) for demographic characteristics. Mixed effects models, controlling for intervention group and sex, were used to assess changes in dietary intake and non-RT PA across assessment time points. This approach was also used to explore potential predictors of changes to dietary intake and non-RT PA. Data were processed using an intention-to-treat approach with missing data handled by using full information maximum likelihood estimation. Pairwise comparisons were not conducted and alpha was set a priori as p < 0.05 being accepted as statistically meaningful. Continuous data are presented as mean ± S.D. Confidence intervals (CI) reported are 95% CIs.

3. Results

No differences in our primary outcomes (HEI-2010 and non-RT PA) were found between SCT and STD maintenance intervention conditions. Therefore, and given the similarities between the SCT and STD treatments, data are presented as pooled between the two conditions.

3.1. Participant characteristics

Participant characteristics are presented in Table 1. The majority of participants were Caucasian (94%) and female (73%). Forty-eight percent of participants were classified as having IFG only, 12% as having IGT only, and 40% as having both IFG and IGT at baseline. Detailed glycemic outcomes and other primary results of the Resist Diabetes trial will not be addressed here as they are presented in the main outcomes analysis [27]. Retention was considered high; 76% of participants who entered supervised training completed the month 15 assessment visit

Table 1:

Participant characteristics.

	Baseline	Month 3	Month 9	Month 15
Total number of participants, n	170	159	138	129
Male, n (%)	46 (27%)	44 (28%)	38 (28%)	37 (29%)
Female, n (%)	124 (73%)	115 (72%)	100 (72%)	92 (71%)
Age, years	59.5 ± 5.5	59.6 ± 5.4	59.8 ± 5.4	60.2 ± 5.3
BMI, kg/m2	32.9 ± 3.8	33.0 ± 3.9	32.8 ± 3.9	32.7 ± 4.0
Body mass, kg	93.3 ± 13.8	93.3 ± 13.3	92.5 ± 13.7	92.2 ± 13.7
Fat mass
%	43.8 ± 7.0	43.2 ± 6.8	43.0 ± 6.8	42.8 ± 6.8
kg	40.6 ± 8.4	39.9 ± 8.3	39.4 ± 8.4	39.2 ± 8.5
Fat free mass
%	56.2 ± 7.0	56.8 ± 6.8	57.0 ± 6.8	57.2 ± 6.8
kg	52.1 ± 10.4	52.7 ± 10.7	52.3 ± 10.0	52.1 ± 10.0
Chest press, 3 RM, kg	33.7 ± 11.6	42.9 ± 14.9	43.0 ± 16.1	43.0 ± 16.2
Leg press, 3 RM, kg	141.3 ± 36.0	166.7 ± 39.4	165.6 ± 39.4	164.9 ± 39.7

BMI = body mass index; 3 RM = 3 repetition maximum.

Note: Evaluation of the above outcomes is presented in the Resist Diabetes main outcomes paper [27].

3.2. Dietary intake

Most participants completed all three 24-hr dietary recalls at each time point (baseline: 96%, month 3: 90%, month 9: 92%, and month 15: 90%), and average energy intake was 113 ± 29% of estimated energy needs (Mifflin-St. Joer equation [44]) across time points, suggesting reasonably accurate reporting. Energy, macronutrient, fiber, and HEI-2010 (total and sub-scores) are presented in Table 2. Energy intake decreased from baseline to month 9 (e.g. with the initiation and maintenance phases of the intervention; β = −87.9, p = 0.015, CI: −158.6 to −17.3) and was maintained through the no-contact phase (month 9 to 15; β = 1.3, p = 0.973, CI: −72.3 to 74.8). No differences over time were detected for either gender or intervention condition (both p > 0.05), though women did report lower energy intakes than men at all time points (p < 0.05 for all). The decrease in energy intake detected could be explained by the reduction in absolute carbohydrate intake (baseline to month 9, β = −16.3, p < 0.001, CI: −25.0 to −7.5)

Table 2.

Reported energy, macronutrient, fiber, and HEI-2010 total and sub-scores.

	Baseline	Month 3	Month 9	Month 15
	Initiation phase	Maintenance phase	No contact phase
Energy, kcal	1803 ± 514	1743 ± 462	1727 ± 454*	1736 ± 500*
Macronutrients Carbohydrate
%	44 ± 8	43 ± 8	42 ± 9	43 ± 10
g	201 ± 63	192 ± 62*	185 ± 61*	191 ± 64*
Fat
%	37 ± 7	37 ± 6	37 ± 7	36 ± 7
g	76 ± 30	74 ± 24	75 ± 25	73 ± 28
Protein
%	18 ± 4	19 ± 4	18 ± 5	19 ± 5
g	78 ± 23	78 ± 21	78 ± 23	78 ± 24
Fiber, g	19 ± 7	18 ± 6	18 ± 7	19 ± 7
Added sugar, g	52 ± 30	49 ± 31	47 ± 24*	50 ± 31
HEI-2010 scores
Total HEI-2010	61.2 ± 12.0	59.5 ± 13.1	61.7 ± 10.4	60.6 ± 12.8
Total fruita,d	2.3 ± 1.6	2.0 ± 1.7	2.2 ± 1.7	2.2 ± 1.6
Whole fruita,d	3.0 ± 1.8	2.5 ± 1.9*	2.7 ± 2.0	2.9 ± 1.9
Total vegetablesa,d	3.6 ± 1.3	3.6 ± 1.3	4.0 ± 1.1*	3.8 ± 1.3
Greens and beansa,d	2.9 ± 2.1	2.7 ± 2.1	3.4 ± 2.0*	3.2 ± 2.1
Whole grainsb,d	5.2 ± 3.5	4.5 ± 3.7*	4.3 ± 3.5*	4.4 ± 3.3*
Dairyb,d	5.5 ± 2.7	5.5 ± 2.6	5.3 ± 2.8	5.1 ± 2.6
Total protein foodsa,d	4.8 ± 0.6	4.8 ± 0.5	4.8 ± 0.7	4.8 ± 0.6
Seafood and plant proteinsa,d	3.0 ± 2.0	2.9 ± 2.1	3.2 ± 2.0	3.1 ± 2.0
Fatty acidsb,d	4.6 ± 3.0	5.0 ± 3.2	5.2 ± 3.1	5.4 ± 3.1*
Refined grainsb,e	7.7 ± 2.6	7.0 ± 3.2*	8.0 ± 2.4	7.4 ± 2.7
Sodiumb,e	3.3 ± 3.0	3.1 ± 3.0	2.7 ± 2.6*	3.0 ± 3.0
Empty caloriesc,e	15.4 ± 3.6	15.8 ± 3.7	15.9 ± 3.1	15.7 ± 3.8

^aScore ranges from 0 to 5.

^bScore ranges from 0 to 10.

^cScore ranges from 0 to 20.

^dAdequacy component. Higher score indicates higher consumption.

^eModeration component. Higher score indicates lower consumption.

^*Indicates significant difference from baseline (p < 0.05).

The mixed effects model showed no change in total HEI-2010 score across intervention time points (β = −0.13, p = 0.722, CI: −0.82 to 0.57). Probing for changes in dietary quality by gender and intervention groups also revealed no change in total HEI-2010 scores within groups (p > 0.05 for all). Despite no change to overall HEI-2010 score, the mixed model analysis of specific HEI-2010 component scores revealed alterations to specific food groups. Initiation of RT was accompanied by a reduction in scores for the whole fruit, whole grains, and refined grains components (baseline to month 3: β = −0.52, p = 0.003, CI: −0.85 to −0.18; β = −0.66, p = 0.039, CI: −1.29 to −0.03; and β = −0.64, p = 0.019, CI: −1.17 to −0.11, respectively). The whole fruit and refined grain scores returned to baseline during the RT maintenance phase (baseline to month 9: β = −0.31, CI: = −0.67 to 0.05, p = 0.087 and β = 0.31, p = 0.279, CI: −0.25 to 0.86), but the reduction in whole grain score persisted for the remainder of the intervention (baseline to month 15: β = −0.78, p = 0.023, CI: −1.5 to −0.11). Maintenance of RT was accompanied by an increase in scores for the total vegetable and greens and beans components (month 3 to 9: β = 0.45, p = 0.002, CI: 0.29 to 1.1 and β = 0.7, p < 0.001, CI: 0.21 to 0.69, respectively), which returned to baseline during the no-contact phase (baseline to month 15: β = 0.19, p = 0.123, CI: −0.12 to 0.72 and β = 0.3, p = 0.157, CI: −0.52 to 0.43, respectively). Over the 15-month study period, the fatty acid ratio component score increased (baseline to month 15: β = 0.70, p = 0.036, CI: 0.04 to 1.35). Finally, the sodium component score decreased over the initiation and maintenance phases (baseline to month 9: β = −0.59, p = 0.048, CI: −1.18 to −0.005), but returned to baseline during the no-contact phase (baseline to month 15: β = −0.2, p = 0.515, CI: −0.8 to 0.4).

To explore physiological and psychosocial factors which may predict the reduction in energy intake reported, baseline prediabetes status (IFG, IGT, or IFG + IGT), fat mass (kg), leg press, outcome expectancies, behavioral resolve, and self-regulation were added to the model. None of these factors predicted the reduction in energy intake (all p > 0.05). Additionally, RT adherence at months 9 and 15 did not predict changes in energy intake (β = −107.1, p = 0.158).

3.3. Non-RT physical activity

Prevalence of meeting aerobic activity guidelines of 500–1000 MET-min/week [37] was 53% at baseline, 59% at month 3, 65% at month 9, and 61% at month 15. Average MET-minutes/week of self-reported participation in low, moderate, and/or vigorous-intensity non-RT PA at each study phase is presented in Fig. 2. Of participants reporting engagement in non-RT PA, walking and lawn and garden work were the most common modes of low and moderate-intensity PA, and running was the most common mode of vigorous-intensity PA reported, although very few participants (~6% across time points) reported engaging in vigorous-intensity PA.

Fig. 2.

Reported MET-minutes/week of low-, moderate-, and vigorous-intensity non-RT PA.MET: Metabolic Equivalents; min: minute; Non-RT PA: Non-Resistance Training Physical Activity.MET-min/week presented as mean ± SE.

Total MET-min/week of non-RT PA increased during the RT maintenance phase (month 3 to month 9; β = 146.2, p = 0.01, CI: 34.7 to 257.7) and was maintained through the no-contact phase (month 9 to month 15; β = 34.4, p = 0.592, CI: −91.3 to 160.1). Changes in non-RT PA were not different according to gender or group assignment and were trimmed from the mixed effects model. To explore physiological and psychosocial factors which may predict the increase in non-RT PA seen from month 3 to months 9 and 15, baseline prediabetes status (IFG, IGT, or IFG + IGT), fat mass (kg), leg press 3RM, outcome expectancies, behavioral resolve, and self-regulation were added to the model, controlling for gender. Increases in RT self-regulation (β = 83.7, p = 0.014, CI: 16.6 to 150.8) and decreases in negative outcome expectancies of RT [e.g. participants were less likely to expect that RT would be accompanied by negative outcomes, such as “Make me feel embarrassed while I am resistance training”(45)] (β = −70.0, p = 0.038, CI: −135.9 to −4.0) predicted the increase in non-RT PA. Prediabetes status and behavioral resolve for RT were not predictive of the increase in non-RT PA. Trends were noted for decreases in fat mass (β = −9.7, p = 0.093, CI: −20.9 to 1.6) and increases in leg press 3RM (β = 1.1, p = 0.087, CI: −0.17 to 0.097) to predict an increase in non-RT PA. Adherence to the RT protocol at months 9 and 15 did not predict engagement in non-RT PA (β = 8.9, p = 0.949, CI: −263.2 to 281.0).

4. Discussion

The major findings of the present study are that participation in one of two SCT-informed RT interventions was accompanied by decreased energy and carbohydrate intake, alterations to HEI-2010 sub-scores, and increased non-RT PA in previously inactive, middle-aged and older, overweight/obese adults with prediabetes. Increased self-regulation for RT and decreased negative outcome expectancies of RT predicted the increase in non-RT PA. Trends were noted for decreased fat mass and increased lower body strength to be predictive of increased reported engagement in non-RT PA. Prediabetes phenotype and behavioral resolve for RT did not predict the increase seen in non-RT PA. None of our hypothesized predictors explained the alterations in dietary intake. We believe this is partially because the RT Health Beliefs survey assessed psychosocial constructs specific to RT, and not dietary habits. However, the predictive analysis was exploratory in nature and provides preliminary data for future work in this area (e.g., interventions could target these as possible mediators). Overall our findings add to the body of literature on health-related behavior change by showing that RT is a mode of exercise which may result in changes to dietary intake and increased participation in non-RT PA.

Previous trials evaluating changes to dietary habits in response to exercise interventions have produced conflicting results, and have primarily focused on aerobic exercise and examined only changes to total energy and macronutrient intake [17]. Studies which have assessed overall diet quality have been cross-sectional in nature, showing an association between increased HEI scores and increased participation in PA [45,46]. To our knowledge, no prior studies have evaluated changes in overall dietary quality with adoption of PA programs, particularly RT. While no change in overall HEI-2010 scores occurred with initiation or maintenance of RT in the current trial, energy and carbohydrate intake decreased and alterations in sub-component scores were detected, suggesting that adoption and maintenance of RT may influence the dietary habits of adults at risk for T2DM. We previously reported that initiation of RT was accompanied by reductions in energy, carbohydrate, fruit and vegetable, and sweets and dessert intake [15]. The current, longer-term analysis indicated that reductions in energy and carbohydrate intake are maintained with continued RT participation. This has important implications for overall diabetes risk, as decreased energy and carbohydrate intake are key recommendations from the American Diabetes Association [47]. Since exercise, particularly RT, is a challenging behavior to initiate [48], successful adoption may increase self-efficacy for additional health-related behavior changes and result in changes to dietary habits [16]. In support of this hypothesis Wycherly et al. showed that participants who completed a 16-week intervention involving caloric restriction and RT (vs caloric restriction without RT) reported the RT helped them adhere to the dietary prescription [49]. Therefore, RT may be a novel mode of exercise for adults, which is impetus enough for increased adherence to provided dietary prescriptions.

In addition to energy and carbohydrate intake, we detected changes in HEI-component scores for whole grains (decreased score indicating negative dietary change) and fatty acid ratio ([polyunsaturated fatty acids + monounsaturated fatty acids]/saturated fatty acids; increased score indicating positive dietary change) from baseline to month 15. Transient changes in other component scores (e.g. increased scores for total vegetable and dark green vegetables and beans components and decreased scores for whole fruit, refined grain, and sodium) as well as added sugar intake were noted over the course of the RT program, but returned to baseline values by month 15. It is possible that the transient decrease in the whole fruits component score and the persistent decrease in whole grains component score is due to lack of nutrition education related to what specific sources of carbohydrate should be limited in those at risk for diabetes. Furthermore, average total HEI-2010 score in our sample was ~60 (out of 100), indicating “needs improvement” in dietary intake quality. This is comparable to dietary quality reported by older US adults (60.5 ± 0.6) in population-wide investigations [50], and indicates that theoretically-based behavior change nutrition education and intervention is warranted in conjunction with, or after successful adoption of, RT.

Previous studies examining the spillover effect of exercise interventions to other forms of PA, while conflicting, suggest that older adults are prone to decreased engagement in non-exercise PA [18]. Similar to the literature regarding spontaneous changes to dietary intake, the majority of trials evaluating spontaneous changes to non-exercise PA have used endurance training interventions. Our study provides evidence that maintenance of an established RT program may promote increased PA outside of the prescribed RT. Adoption of new health-related behaviors may be more effective when initiated after a more challenging behavior (e.g., RT) has been successfully maintained [48]. Therefore, health-care professionals should consider promoting RT prior to alterations in dietary intake and/or aerobic exercise, as this may be a more efficacious intervention sequence. Furthermore, it is important to note that the increase in non-RT PA in the current trial (~160 MET-minutes/week) is clinically relevant for individuals at risk for T2DM. Glycemic-outcomes improve in a dose-dependent manner as PA volume increases, which could prevent the progression from prediabetes to overt diabetes diagnosis, or result in regression to normoglycemia [6,51].

Related to the potential for RT to influence other forms of PA, our findings add to the limited and inconclusive body of literature on this topic. In agreement with Hunter et al., participation in RT by our participants was also accompanied by increases in non-RT PA [21]. Hunter et al.'s findings provide evidence that increased ease of aerobic activity following RT explains the increased participation in non-RT PA. While economy of movement was not measured in the current study, detection of a trend for decreased fat mass and increased lower body strength to predict the increase in non-RT PA may be considered a similar explanatory mechanism, as ease of locomotion increases with loss of fat mass and increases in strength [52]. Contrary to our results, Church et al. saw no change in step counts outside of supervised exercise in the RT and combination (AT + RT) groups in the 9 month HART-D trial [53]. Similarly, a secondary analysis from Fielding et al. found that despite increases in muscular strength and functional ability following a 6-month RT program in elderly adults, non-RT PA did not increase [54]. While these studies suggest no spillover effect occurs when middle-aged and older individuals engage in RT, neither trial included a behaviorally-focused theoretical approach. In our analysis, increased self-regulation for RT and decreased negative outcome expectancies of RT were significant predictors of non-RT PA. Therefore, psychosocial factors are likely an important determinant of PA that should be addressed in interventions, particularly for older, inactive adults at risk for chronic diseases, as they have been shown to improve health-related behaviors [55].

To our knowledge, this is the first investigation to examine spontaneous alterations to overall diet quality in response to RT, and one of the few to date which evaluates if RT is associated with increased non-RT PA. Strengths of our study include a large sample size, minimal attrition, high dietary recall completion rates, and use of validated methods and instruments for assessing dietary intake [29], non-RT PA [35], and psychosocial factors [39]. The focus on HEI-2010 scores extends our previous findings [15] and addresses a gap in the literature [17,18] by providing novel information on alterations to specific dietary patterns which may spontaneously occur with RT adoption and maintenance. Our study is not without limitations. The observational nature and lack of a non-exercise control group is a significant limitation that diminishes our ability to determine if changes observed are the result of participation in RT, or instead due to knowledge of prediabetes status, interaction with study staff, and/or health-related information provided in the feedback packets. However, being at risk for diabetes was listed in recruitment materials and most participants were aware they had prediabetes prior to enrollment. Furthermore, if knowledge of prediabetes status was impetus enough for changes to dietary intake and increases in non-RT PA, changes would have occurred during the baseline testing period and not detected as different with initiation of RT, or occurring during the initiation phase only and not the maintenance phase. Although validated instruments and trained staff were utilized to assess dietary intake and non-RT PA, the reliance on self-reported rather than more objective measures is a limitation. Despite this, we are confident reporting errors due to social desirability biases were minimal, as not all dietary changes detected were beneficial (e.g. decreased whole fruit and whole grain scores) and only a small number of participants reported engagement in vigorous-intensity PA.

5. Conclusions

Findings from this observational trial suggest that RT may be a unique mode of PA in its ability to influence non-RT PA and dietary intake among previously inactive, overweight/obese adults with prediabetes. As single-component interventions are generally less costly and resource-intensive than multi-component interventions, our results support the use of single-component interventions, focused on RT, when time and resources are limited. Future research examining mechanisms by which RT may influence other health-related behaviors, particularly in individuals at risk for metabolic diseases, is warranted.

Abbreviations:

RT

resistance training

PA

physical activity

SCT

social cognitive theory

HEI

Healthy Eating Index

T2DM

type 2 diabetes mellitus

IFG

impaired fasting glucose

IGT

impaired glucose tolerance

OGTT

oral glucose tolerance test

STD

standard care

DGA

Dietary Guidelines for Americans

MET

Metabolic Equivalents

FM

fat mass

FFM

fat-free mass

DXA

dual-energy X-ray absorptiometry

3RM

three repetition maximum

TLFB

timeline follow back