Determinants of physical activity during cancer treatment: a longitudinal exploration of psycho-cognitive variables and physician counseling

In the past decades, a vast number of scientific publications have highlighted the benefits of physical activity (PA) for individuals living with cancer. The state of research on exercise and cancer was reviewed in a roundtable of international experts in 2018, organized by the American College of Sports Medicine (ACSM). The consortium found substantial evidence that sufficient PA increases survival in individuals with breast, prostate, and colorectal cancer (Patel et al., 2019). A recent meta-analysis of randomized controlled exercise trials in cancer patients and survivors found a 26% relative reduction in mortality risk in favor of exercise interventions (Morishita et al., 2020). The ACSM roundtable further underlined robust associations for the ameliorative effects of PA on eight common health-related outcomes of cancer or cancer therapies (e.g., cancer-related fatigue or chemotherapy-induced peripheral neuropathy) (Campbell et al., 2019). This scientific evidence led to the recommendation by the ACSM that individuals with cancer should engage in 150–300 min of moderate-to-vigorous PA per week (in addition to bi-weekly resistance training), also during cancer therapy.

However, most individuals with cancer are not sufficiently physically active, with only 30–50% meeting guidelines for aerobic exercise according to self-reports (Coletta et al., 2019; Crawford et al., 2016; Eng et al., 2018; Ottenbacher et al., 2015; Steindorf et al., 2019; Yan et al., 2018) and less than 15% according to accelerometer-derived activity data (McDonald et al., 2019; Smith et al., 2019; Thraen-Borowski et al., 2017). An important reason for unfavorable PA behaviors of cancer survivors seems to lie in a decrease of PA during chemotherapy or radiotherapy (Huy et al., 2012; Kenkhuis et al., 2021), or in the first weeks after cancer surgery (Smith et al., 2019; Zhou et al., 2022). Given the benefits of PA, it is worthwhile to elucidate which factors are essential for the development of PA levels during cancer treatment or post-surgery.

During cancer treatment, physicians have the potential to motivate individuals with cancer to engage in sufficient levels of PA (Schmitz et al., 2019). PA levels were found to be increased in those individuals with cancer who recalled a PA advice (Fisher et al., 2015; Tarasenko et al., 2017), as well as in study participants who received an oncologist-led PA recommendation in a randomized controlled trial (Jones et al., 2004). A systematic review by Brunet et al. (2020) pointed out that the intensity of PA counseling can make a difference in the effect on cancer patients’ PA levels, and that a simple recommendation is not necessarily sufficient to achieve behavior changes. A PA counseling based on the 5As framework goes beyond a recommendation for PA as it contains counseling steps defining basic counseling (Assess PA levels and give Advice on PA) but also in-depth counseling (Agree on PA goals, Assist in achieving them, and Arrange follow-up monitoring) (Estabrooks & Glasgow, 2006). A previous cross-sectional study with 1206 cancer survivors found evidence that only in-depth PA counseling (i.e., Agree, Assist, and/or Arrange), not basic PA counseling (i.e., Assess and/or Advise), was associated with higher levels of PA (Haussmann et al., 2021). Therefore, to gain more insights into the effects of PA counseling during cancer treatment, its impact on PA levels should be analyzed longitudinally and separately for basic and in-depth PA counseling.

In addition to PA counseling, psycho-cognitive variables play an important role for the engagement in PA by individuals with cancer (Hirschey et al., 2020). The Theory of Planned Behavior (TPB) (Ajzen, 1991) is a widely applied social cognitive theory to explain health behavior. In the TPB, the intention is the central determinant for explaining the health behavior. The intention, in turn, is influenced by self-efficacy, attitudes, and the perceived expectations by others (i.e., subjective norm). A meta-analysis of longitudinal observational studies indicated that assumptions of the TPB adequately reflect PA behavior of individuals with cancer (Hirschey et al., 2020). In the current study, we complemented the TPB with physician PA counseling as an additional determinant for PA intention and behavior. In addition, we considered it useful to assess cancer patients’ attitudes towards PA not only covering agreements with benefits of PA (Hunt-Shanks et al., 2006; Speed-Andrews et al., 2014; Trinh et al., 2012; Ungar et al., 2015). Benefits of PA for individuals with cancer are now widely acknowledged (activity paradigm); however, the attitude that cancer patients should rest during cancer therapy (rest paradigm) has not yet been completely overcome among health care professionals (Ungar et al., 2019) or among patients themselves (Haussmann et al., 2022).

The aims of this longitudinal observational study are to analyze (a) changes in PA behavior during cancer treatment or in the post-surgery phase, (b) perceived PA counseling provided by the attending physician taking into account basic and in-depth PA counseling, (c) differences in PA counseling according to different sociodemographic and disease-related characteristics, and (d) effects of physician counseling and psycho-cognitive variables on patients’ intention for PA and changes in PA levels.

This longitudinal observational study provided insights into how individuals with cancer experience PA counseling by their physician during their treatment, and into how this PA counseling and psycho-cognitive determinants based on the Theory of Planned Behavior (TPB) explain accelerometer-derived changes in PA. The majority of participants reported being asked or advised about PA by their physician (basic counseling) at least once during the study period, while only one in three reported receiving counseling beyond a PA recommendation (in-depth counseling). Participants in poorer health and those with lower education reported less frequently that they received in-depth PA counseling. Remarkably, only in-depth PA counseling, but not basic PA counseling, predicted the intention to engage in PA four weeks later. This intention for PA was additionally predicted by participants’ self-efficacy. Results further point to stable activity levels, with PA three months after baseline being most strongly predicted by PA at baseline. Physicians’ counseling (in-depth or basic PA counseling) neither predicted PA three months later nor moderated the effect of PA intention on actual PA levels two months later.

In this study, the majority of individuals with cancer received some kind of PA counseling by their physician during their (initial) treatment. PA counseling practices were assessed in detail according to the 5As framework (Estabrooks & Glasgow, 2006) and classified into basic PA counseling (assess and/or advise) and in-depth PA counseling (setting up goals, assist, and/or arrange follow-up monitoring). More participants reported that they received basic PA counseling (48%) than in-depth PA counseling (30%) at least once during the study. The percentage of participants who reported that they received a PA recommendation in this study was higher than in previous studies in cancer survivors (13–50%), which did not differentiate between the intensity of PA counseling (Eng et al., 2018; Fisher et al., 2015; Höh et al., 2018; Steele et al., 2021). In a previous cross-sectional study, a similar pattern was observed with respect to the intensity of PA counseling (basic PA counseling: 59%; in-depth PA counseling: 20%) (Haussmann et al., 2021). In line with this cross-sectional study, we found that nearly 1 in 4 participants received no PA counseling at all. Thus, despite numerous research findings on the benefits of PA, there remains a proportion of cancer survivors who do not receive (or at least do not remember) any PA counseling by their physician. For this subset of cancer survivors, structures should be created to guarantee counseling on PA at least once after cancer diagnosis, optimally through targeted PA screening at the beginning of therapy (as an example, cf. the Alberta Cancer Exercise PA screening form (Mina et al., 2018)).

Results of this study also indicate that cancer patients with more comorbidities are more likely to receive no PA counseling at all. In line with this result, cancer patients in poorer health (i.e., more side effects and diagnosis of metastases) and who were undergoing chemotherapy were advised less intensively about PA by their attending physician, i.e., they reported more basic but less in-depth PA counseling. In a previous cross-sectional study, a different physician counseling behavior was found for patients not meeting PA guidelines before diagnosis (more likely to receive no PA counseling at all) and patients who were currently under treatment (more likely to receive basic counseling). We are not aware of any study that has examined PA counseling longitudinally at the onset of cancer treatment, a time when patients’ health status often declines (Lewandowska et al., 2020; Yucel et al., 2014). Importantly, it is well-known that PA is associated with a reduced incidence of several comorbidities (Kang et al., 2018), and beneficial to reducing different side effects of cancer treatment (Campbell et al., 2019). Exemplarily, with regard to alleviating cancer-related fatigue, a meta-analysis of individual patient data indicated that patients with severe fatigue particularly benefited from exercise training (Buffart et al., 2018). In a previous cross-sectional study, cancer patients with metastases reported that they disproportionately increased their PA level from pre- to post-diagnosis when given in-depth PA counseling (Haussmann et al., 2021). Thus, providing fewer PA counseling to cancer patients in poor health seems counterintuitive, as these patients would strongly benefit from (in-depth) PA counseling. In our study, patients with a lower educational level were also less likely to report in-depth PA counseling. This finding is of high practical relevance as this group is at increased risk of not adhering to exercise guidelines anyway (Eng et al., 2018; Steindorf et al., 2019). Future research should focus more on how to reach individuals with a lower educational level, e.g., by analyzing which kind of information material is beneficial for which patient groups (Webb et al., 2019).

An important aim of this study was to identify determinants that can explain the intention for PA and the change in PA over the course of cancer treatment or post-surgery. The TPB set the framework for the investigated structural equation models, with patient-reported physician PA counseling as an additional variable. Overall, the structural equation models yielded a good model fit, confirming the applicability of the TPB for the longitudinal explanation of PA behavior (Hirschey et al., 2020). Within the model, participants’ intention for PA four weeks after baseline was predicted by patients’ self-efficacy for PA and physician PA counseling. With regard to PA counseling, however, this was only true for in-depth counseling, not PA counseling in general. The result may indicate that a simple recommendation may not be sufficient to achieve changes in patients’ intention to engage in PA. However, the effect was rather small (the lower limit of its CI was slightly above 0) and refers only to the intention after 4 weeks; future studies, optimally randomized controlled trials, should investigate whether the effect can be validated and sustained over a longer period.

Given their heavy workload, physicians treating oncological patients may struggle to provide this time-consuming PA counseling on a regular basis. Correspondingly, a lack of time was cited by healthcare professionals as one of the most major barriers to counsel cancer survivors on PA (Haussmann et al., 2018b; Nadler et al., 2017; Ramsey et al., 2022). For this reason, the ACSM (i.e., the Exercise is Medicine Initiative) recommends that oncologists refer cancer patients to outpatient rehabilitation specialists for further evaluation if they are unsure whether the patient can safely exercise (Schmitz et al., 2019). In this regard, a predefined pathway of PA promotion with clear procedures and responsibilities could facilitate physicians to integrate PA counseling. Physicians (or other qualified healthcare professionals) could be the starting point in this pathway by referring patients to exercise programs or, if not available, providing self-management tools (Mina et al., 2018). Recently published studies on intervention developments provide blueprints for how physicians and other healthcare professionals can be involved in promoting PA to oncology patients in a structured facilitating process (Kennedy et al., 2020; Rammant et al., 2021; Turner et al., 2021).

Self-efficacy to engage in sufficient PA was the only psycho-cognitive variable based on the TPB that showed a significant effect on patients’ intention for PA four weeks later. This is in line with the meta-analysis by Hirschey et al. (2020) in which self-efficacy turned out (together with attitudes) to be the strongest predictor of PA intention. Current studies on breast cancer patients during chemotherapy also showed a significant association between self-efficacy with PA levels (Auster-Gussman et al., 2022) and with better adherence to exercise programs (An et al., 2020).

Results of the structural path models further revealed that PA of study participants three months after baseline was most strongly predicted by PA at baseline. This finding indicates a high stability of PA behavior in our study participants, as also indicated by the high number of participants who either achieved (50%) or failed to achieve (30%) PA guidelines at each of the two accelerometer-based measurements. Previous research pointed to a decline in PA levels during cancer treatment (Huy et al., 2012; Kenkhuis et al., 2021) or post-surgery (Smith et al., 2019; Zhou et al., 2022), which suggests that the study population in this work was less likely to let the cancer therapy discourage them from their usual PA.

Intention to engage in sufficient PA was another significant predictor of PA behavior, although the regression coefficient in this study (β = 0.19) was not as high as in the meta-analysis of longitudinal studies by Hirschey et al. (2020) (β = 0.27). Contrary to our expectation, the association between participants’ intention for PA and PA levels eight weeks later was not moderated by physician counseling; thus, overall PA counseling did not contribute to turn PA intentions into action (nevertheless, exploratory analyses revealed that it did, though, for participants with a high health status). Although some studies point to an effect of a single PA recommendation on cancer survivors’ PA levels (e.g., Jones et al., 2004), there is also evidence that additional supporting material to an oncologist PA recommendation is beneficial to foster cancer patients’ PA (Park et al., 2015; Winters-Stone et al., 2018). Due to the small number of study participants that reported in-depth counseling between t2 and t3, we could not distinguish between basic and in-depth PA counseling. Therefore, we cannot exclude that specifically in-depth counseling may help to translate PA intention into actual PA. Behavior change techniques such as concrete goal setting and coping and action planning are effective to bridge the intention-behavior gap (Avishai et al., 2019; Sheeran & Webb, 2016), but there were contrary results for study populations with cancer patients (Finne et al., 2018). This suggests that individuals with cancer need closer guidance in using behavioral change techniques. Future studies may investigate whether physicians (or exercise professionals) can actually offer this required closer guidance by providing in-depth PA counseling.

The strength of this study is that it longitudinally analyzed a period in which PA and physician PA counseling are particularly important, i.e., in the (initial) treatment phase. In this context, physician counseling in the form of the 5As framework and psycho-cognitive variables were assessed in detail at several time points. In addition, PA was objectively measured twice using accelerometers, resulting in a high data quality that is not biased by over or underestimation of own PA in self-reports (Prince et al., 2008). Results of this study provide valuable insights into what PA counseling is needed from a physician to influence PA intentions of patients during cancer treatment. The development of future interventions to encourage physicians to improve their PA counseling to individuals with cancer may build on these findings.

However, the results of this study must be considered in light of some limitations. The study sample might be subject to a selection bias, meaning that individuals with a high affinity for PA in particular participated in the study. According to retrospective self-reports, our study participants were on average already very active before their cancer diagnosis. This would explain the high level of PA at both measurements, which far exceeds the results of previous studies that objectively measured PA in cancer survivors (McDonald et al., 2019; Smith et al., 2019; Thraen-Borowski et al., 2017). Interestingly, at baseline, the self-reported moderate-to-vigorous-PA during the last week (before they wore the accelerometer the first time) was considerably lower. This is in line with research showing that wearing accelerometers can be an impetus to increase PA with similar effects as PA interventions that include additional behavioral change techniques (Brickwood et al., 2019). Potentially, the wearing period of seven days (with the minimum criterion of four wearing days) in this study was set too short to compensate for this effect. High mean values were also found regarding the psycho-cognitive variables of the TPB. These indicate ceiling effects, which may have led to under-estimations of the effects on intention in particular. The sample size was relatively small for the SEM analyzed which increases the risk of false-negative results (type 2 error) or may have caused the magnitude of effects found to be overestimated. Therefore, results have to be interpreted with caution. To account for the uncertainty in the results, we provided confidence intervals that allow a more accurate assessment of the effects. Furthermore, we conducted a sensitivity analysis showing that the study had adequate power to detect small to medium effects. We decided not to include any control variables in the two calculated models to ensure parsimony. As a result, our N exceeds the rule of thumb of a minimum of a 10:1 ratio of sample size to parameters in SEM (Kline, 2010). On the other side, not including control variables—such as medical (e.g., diagnosis of metastases) or sociodemographic variables (e.g., education)—in the model may also have resulted in an over-estimation of the effects found (omitted variable bias). This would be the case if the omitted control variable was positively associated with another predictor and had a positive effect on intention (t2) or on moderate-to-vigorous PA (t3). Due to the small sample size, we also could not conduct subgroup comparisons (as originally intended with regard to different cancer entities). In explaining PA intention and PA behavior in our study using the TPB, it may also be that additional variables (e.g., health-related variables or illness perceptions) would have been significant predictors. We did not use a controlled but an observational study design. Therefore, there is the risk that self-reported information was subject to biases, such as recall errors or socially desirable response behavior. Time points of received PA counseling could only be specified to the time between questionnaires (or time since diagnosis) and not to an exact date. Moreover, advice on PA could have been given by persons (other healthcare professionals or friends/family) other than the treating physician, who, however, seems to be a central source for PA advice for individuals with cancer (Avancini et al., 2020).

In this longitudinal study, we identified PA behavior at baseline as the strongest predictor of moderate-to-vigorous PA behavior during cancer treatment. However, in-depth counseling of physicians appears to make a difference by prompting higher intentions to engage in PA. At the same time, further research is needed how physicians can help translate the intention for PA into sufficient activity levels. In-depth PA counseling that integrates individualized guidance as part of a structured pathway in clinical care, seems to be a promising way forward. Our study also revealed that patients in poorer health and patients with lower education are less likely to receive an in-depth PA counseling. Therefore, a focus should be placed on ensuring that patient subgroups with different sociodemographic and medical backgrounds receive the optimal supportive care regarding PA.