This study tested the TPB in explaining aerobic PA and resistance training in a national population sample of T2D adults, and compared the mean scores of the TPB constructs between these two modes of PA.
The results from our path analyses provide partial support of the TPB's utility in predicting PA intention and behavior of T2D adults (main study objective). With aerobic PA, significant associations were found with attitude, intention, and gender explaining 10% of the variance with behavior. Relationships between aerobic intention with attitude and injunctive norm were also reported explaining 39% of the variance for intention. With resistance training, no significant associations with the TPB variables were reported with behavior, but the variables explained 8% of the variance. However, relationships existed between intention with attitude and descriptive norm, explaining 45% of the variance with resistance training intention. Age and gender differently affected the TPB constructs in both aerobic PA and resistance training.
Limited research has been conducted using the TPB to determine its predictive ability in explaining PA behavior in T2D. A study by Plotnikoff et al. [34], investigated the utility of the TPB in understanding aerobic PA in a large adult population with T1D (N = 697) and T2D (N = 1614). For both diabetes types, attitudes, subjective norms, and PBC were all significantly associated with intention, and intention was significantly associated with behavior [34]. Similar to that study, we found attitude and injunctive norm to be significantly associated with intention, and intention was significantly associated with aerobic PA. The magnitude of these relationships are also congruent with previous PA results from meta-analyses conducted [32, 33], where attitude appeared to be the strongest correlate of intention, closely followed by PBC and with subjective norm being the weakest predictor.
In addition, the variances explained by the TPB for aerobic PA in our study were consistent with previous TPB studies as reported in meta-analytical studies [32, 33]. The studies indicate that the TPB constructs explained approximately 30–46% of the variance for PA intention and 21–27% for PA behavior. The TPB's ability to explain behavior (10%) in our study was lower than that reported in previous TPB studies. This could be attributed to the unique study characteristics, the nature of the diabetes disease and related PA behavior. The adult diabetic population tends to be less active than the general population [4] most likely due to the physical limitations from the disease. Further, the majority of the TPB studies reviewed in the above meta-analyses employed cross-sectional designs [32, 33].
As far as we know, our study is the first to examine the determinants of resistance training in individuals with T2D. It is also the first study to apply a theory-driven approach, namely the TPB, in examining the determinants of resistance training in a T2D sample. However, there is a small but emerging literature on the predictors of resistance training in non-T2D adults. Dean et al. [40] examined the efficacy of the TPB in understanding the factors influencing older adults' participation in strength training through purposeful sampling (N = 200) of men and women age 55 years and older from seniors' centers. Cross-sectional results revealed that subjective norm and PBC, but not attitude, explained 42% of the variance in strength-training intention, while intention, but not PBC explained 40% of the variance in strength-training behavior. Consistent with our study, descriptive norm significantly explained 45% of the variance in resistance training intention, but it was the addition of attitude, and not PBC, that contributed to the explained variance of the total TPB model. With regards to resistance training behavior, our study revealed no significant associations with intention or PBC with behavior. These inconsistent findings may be due to the different types of samples employed in the two studies, where Dean et al. [40] employed a non-diabetic/chronic disease sample.
Jette et al. [41] identified factors associated with resistance training exercise participation and adherence in a sample of sedentary, functionally limited, community-dwelling adults aged 60 to 94 years (N = 102) who were part of a 26-week home-based resistance training program. Cross-sectional findings of this older population study revealed that predictors of the frequency of exercise participation (number of exercise sessions performed divided by number of exercise sessions possible) were not the same as those that predicted high levels of adherence (number of calendar periods that participants exercised at least half the number of desired sessions) to the home-based program. Those participants with higher functional mobility, weaker muscle strength, and fewer new medical problems during the intervention participated more frequently in the home-based strength training program. None of the demographic factors, comorbidities, or psychological factors were significant predictors of participation. However, participants' positive attitudes toward exercise and strong sense of control over exercise, lower levels of perceived confusion, and depressed moods were associated with higher adherence to their individual home exercise programs. Although physical health variables were the primary indicators of overall participation in the program, it was the psychological factors that were most important to adherence to the home-based program [41]. Although our study did not measure adherence to resistance training, similar with the findings of Jette et al. [41], our results did reveal that attitude had a significant association with intention.
Further, Bryan and Rocheleau [39] found that all the TPB constructs were strongly associated with intention, and PBC was associated with resistance training behavior among college students (N = 210), of which 70% of the convenience sample was female. In our study, only attitude and descriptive norm were associated with intention, and none of the TPB variables had a direct effect on resistance training behavior. In addition, the TPB variables in Bryan and Rocheleau's [39] study, extroversion and perceived health accounted for 19% of the variance in aerobic exercise, while accounting for 40% of the variance in resistance training. Our study revealed 10% of the variance accounted for with aerobic PA behavior and 8% of the variance with resistance training behavior. It is important to note that these inconsistent results may be due to the differences in the study design and sample. Bryan and Rocheleau's [39] sample of college students were younger and disease-free, included other variables (i.e., extroversion and perceived health) in the TPB model, and employed an aggregated subjective norm measure.
The only study to present tests of the predictive validity of the TPB for aerobic-exercise behavior and resistance training was conducted among a convenience sample of young, healthy students [39]. An objective of the study examined whether extroverted personality and perceived health can be embedded in the TPB structure to improve the specificity of the model for exercise behaviors. The results revealed that the TPB constructs of attitude, norms, and PBC exhibited strong correlations with aerobic intentions, and PBC had a significant direct effect on aerobic behavior. Contrary to these findings, our study revealed that only attitude and injunctive norm exhibited associations with aerobic intention, and it was attitude that had a significant direct effect on aerobic behavior.
Age and gender in our study were found to act differently on the TPB constructs in both aerobic PA and resistance training. In our study, being younger and of the male gender were associated with higher mean scores in the TPB constructs. It is important to differentiate between age and gender when examining any population, as age and gender can be important determinants of PA [53]. For example, Bryan and Rocheleau [39] note that individuals who engage in aerobic activity versus resistance training often have different goals, and the difficulty of performing these two activities may be quite different.
In regards to resistance training, there was no significant relationship found between intention and behavior in our study. However, the bivariate correlation did reveal that resistance training intention had a significant association with resistance training behavior (r = 0.27, p < .01). Other TPB variables including attitude (r = 0.27, p < .01), injunctive norm (r = 0.26, p < .01), and descriptive norm (r = 0.23, p < .01) were also significantly correlated with resistance training behavior. Although injunctive norm (β = 0.14) and resistance training intention (β = 0.12) contributed to the 8% of explained variance for resistance training in the multiple regression model, neither variable reached significance at the .05 level. Relatively consistent with Plotnikoff [4] who reported only 12% of a large T2D population sample of adults performing any form of resistance training, our study reported 23% of the sample engaging in this behavior (of which only 17% were meeting guidelines). This may have limited the statistical power in our analyses to examine the determinants of this behavior.
Further, given the literature to explain resistance training is relatively embryonic, we conducted a set of additional analyses with different classifications of the resistance training dependent measure. These included multiple regression analyses using (1) frequency only (i.e., number of times individuals reported engaging in resistance training) which explained 4% of the variance; and (2) employing three resistance training categories [not engaging in any resistance training, engaging in some resistance training (i.e., 1–2 times/week), and meeting resistance training guidelines (i.e., = 3 times/week)] which accounted for 6% of the behavior. Two sets of logistic regression analyses also examined (1) those meeting (i.e., = 3 times/week) versus not meeting guidelines (i.e., < 3 times/week); and (2) those engaging in any resistance training (i.e., > 0 times/week) versus not engaging at all in this behavior (i.e., 0 times/week), explaining 11% and 16% of the variance respectively. However, none of these additional analyses produced any significant associations between the TPB variables with resistance training behavior.
Nevertheless, the TPB does hold partial utility for resistance training intention which is consistent with previous studies [39, 40]. Resistance training is a relatively novel behavior in the T2D population where the intention-behavior gap may be a realistic indication of how psychosocial cognitive factors directly influence intention, but does not necessarily translate into behavior. There may be other factors (e.g., lack of experience, lack of knowledge) for both aerobic PA and resistance training which may have impacted this relationship that were not examined in our study. Also, it is not surprising that many of the social cognitive measures used to predict general PA are more relevant for aerobic activity since the majority of exercisers are engaging in only aerobic forms of activity [4]. For example, PBC and attitude towards an aerobic activity such as walking, may be quite different from the barriers, perceived control and attitudes towards resistance training. Resistance training may be daunting for some and may be influenced by control factors including access to facilities and special equipment, and knowledge about what exercises to perform [4, 40]. Testing the TPB with additional constructs [e.g., environmental factors (costs, equipment/facilities), observational learning, behavioral capabilities (skill acquisition)] may further help to explain this behavior.
The inclusion of resistance training behavior in the TPB model still remains exploratory. Until resistance training behavior becomes more widely engaged in the population, the application of social cognitive models (including the TPB) cannot be fairly assessed at this point. Theoretical research is also needed on the intention-behavior gap with resistance training to determine the most salient predictors of this behavior to guide interventions by operationalizing appropriate theoretical constructs.
In conjunction with the consistent findings of the TPB in predicting PA behavior, significant differences were also found in the means and trends between aerobic PA and resistance training at both the global- and item-level of the TPB constructs (our second study objective). All the means for aerobic PA were significantly higher compared to the means for resistance training. This suggests the importance of how the relationships between the TPB variables vary in different PA settings. Bryan and Rocheleau [39] found that the TPB model for resistance training had greater predictability than that of aerobic exercise, with more variance explained for resistance training. The TPB's ability to account for a greater proportion of variance in resistance training was attributed to the stronger role of PBC. In other words, resistance training appears to be more strongly influenced by volitional control than aerobic activity due to increased equipment and training required, as well as knowledge needed to engage in it. Thus, PBC becomes a more significant direct predictor of resistance training.
Overall, our study results provide partial evidence towards the utility of the TPB for practitioners and researchers to develop and evaluate appropriate PA interventions for populations with T2D. While TPB appears to have good utility in predicting both aerobic and resistance training intention, its predictive ability was less evident for behavior in both modes of exercise. According to Ajzen [54], behavioral intentions must attempt to influence the beliefs that ultimately lead to the performance of the behavior. Fishbein, Von Haeften, and Appleyard [55] advocate identifying salient beliefs from the target population, developing persuasive messages around the beliefs, and then developing appropriate material based on the elicited beliefs. Based on our study results, these interventions would need to apply strategies for increasing the salience of attitude and descriptive norm for resistance training intention. For aerobic PA, specific emphasis should be placed on enhancing positive attitudes towards PA and having important others approve the PA behavior. On the other hand, specific emphasis should be placed on the importance of social norms and enhancing positive attitudes towards PA in interventions using resistance training. For example, for increasing the salience of attitude, interventions may focus on highlighting the benefits and enjoyment aspects of both aerobic PA and resistance training. For social norms, interventions may include messaging materials that encourage individuals with diabetes to perform aerobic PA and resistance training with a friend.
However, this study needs to be interpreted within the context of its limitations. The socio-cognitive and PA measures relied on self-report which can introduce measurement error such as recall error, social desirability and other reporting biases. Future research should use objective measures for assessing aerobic PA (i.e., pedometry, accelerometry) and consider employing validation studies (e.g., observation techniques) for resistance training measures. In addition, our study results need to be treated with some caution in terms of their generalizability to the national T2D population given the relatively low response rate. It may be that the respondents were more motivated for activity which may have led to an over-assessment of the true, more general predictive value of TPB variables for both aerobic PA and resistance training.
In conclusion, this study adds to the limited literature base on the TPB and aerobic PA in T2D, as there is currently only one study that has examined aerobic PA predictors in TPB [34]. We employed a national, random sample which is an additional strength in our study. Our study provides the first test of the TPB in the PA domain on a diabetic population, simultaneously examining the psychosocial cognitive factors (with parallel items) that influence both aerobic PA and resistance training. Also, it is the first of any social cognitive theory examining predictors of resistance training in this population. Further, the test-retest component in the study design added to the scale reliability of the cognitive measures. Since resistance training is relatively novel in the PA and diabetes literature, future research is needed to examine other predictors that may be present in understanding resistance training to further guide, develop, and evaluate theory-based interventions in this population. Future research is also warranted to broaden the TPB beyond its existing social-cognitive constructs by including social and environmental factors [56] for both aerobic PA and resistance training in this population.