Association between self-assessed preoperative level of physical activity and postoperative complications-An observational cohort analysis within a randomized controlled trial (PHYSSURG-C)

Introduction: Physical activity has been suggested as a modifiable risk factor for postoperative recovery after major surgery. We aimed to further define the association between self-reported leisure time physical activity and postoperative complications and recovery on a group level. Materials and methods: An observational cohort analysis was performed within a randomized controlled trial. Patients 20 years with colorectal cancer were eligible. Between January 2015 and May 2020, 761 participants were recruited. Leisure time physical activity was self-assessed on a four-grade scale. The primary outcome in this analysis was postoperative complications within 90 days, measured with the comprehensive complication index (CCI). Secondary outcomes were CCI for the index hospital stay, CCI 30 days postoperatively, specific types of complications, length of hospital stay, and self-assessed physical recovery. Analyses were adjusted for gender, age, study center, alcohol consumption, tumor stage, colon/rectal cancer, neoadjuvant therapy, and open or laparoscopic surgery. Results: Data on preoperative physical activity was available for 614 participants. Participants who reported some physical activity had on average a lower CCI 90 days postoperatively than sedentary participants (odds ratio 0.63, 95% CI 0.43e0.92). Similar patterns were shown for complications during hospital stay and within 30 days. There were trends towards lower risk for several types of complications, reoperations and length of hospital stay but the only difference shown was for respiratory insufficiency. Conclusion: Physically active participants experienced fewer postoperative complications, which suggests the information has a potential prognostic value. The work for increasing physical activity in the general population should continue. © 2021 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).


Introduction
Physical activity has several positive health effects, including reduced risk for cardiovascular disease, diabetes mellitus, mental health issues, and several types of cancer, including colorectal cancer [1]. Curative treatment for colorectal cancer includes surgery, with high risk for postoperative complications and morbidity [2], which has led to the development of programs where perioperative measures aimed at enhancing recovery are collected [3].
Preoperative physical activity has been identified as a possible modifiable risk factor for postoperative complications and recovery after colorectal cancer surgery. This has led to several studies where preoperative exercise interventions before colorectal cancer surgery have been tested [4e6], and to recommendations on such interventions [3]. However, results in interventional studies have varied from halved risk for postoperative complications [4] to no effect [5,6]. A non-modifiable risk factor is habitual physical activity. Underlying studies exploring the association between habitual physical activity and postoperative recovery would be of value to create evidence to support advice to the public. There are some reports in the literature indicating a positive effect of habitual physical activity before colorectal cancer surgery [7e12], but the effect size and effects on specific types of complications remain uncertain.
We have recently reported the main results from PHYSSURG-C, a randomized controlled trial where the effect of short-term preoperative exercise intervention on postoperative recovery after colorectal cancer surgery was studied [6]. We did not find any differences in postoperative recovery depending on preoperative intervention or not. In order to further define the possibilities of preoperative physical activity, we aimed to examine the average effect of preoperative level of leisure time physical activity on postoperative complications and recovery after colorectal cancer surgery in an observational analysis within this randomized controlled trial.

Materials and methods
This manuscript reports results from an observational cohort analysis on data from the randomized controlled trial PHYSSURG-C. Methods for PHYSSURG-C have been described in detail previously [13], and results have been reported on both primary outcome [6] and a subgroup analysis [14]. The analyses performed for the current manuscript were not part of the original design of the study and are explorative. Ethical permission was obtained from the Regional Ethics Board in Gothenburg (2014-10-30, DNR:597e14). The study protocol has been published and is also available at www. ssorg.net. The study was registered at ClinicalTrials.gov with trial registration number NCT02299596, date of first registry was November 17, 2014.

Participants and setting
Study participants were recruited at six hospitals in Sweden including both county and university hospitals, from January 22nd , 2015 until June 28th , 2020. Individuals 20 years old scheduled for elective curative surgery due to colorectal cancer were eligible. Exclusion criteria included inability to perform study-specific procedures, inability to understand given information, emergency surgery, local excision (e.g., transanal endoscopic microsurgery), and cytoreductive surgery with subsequent hyperthermic intraperitoneal chemotherapy. For the main purpose of the study, participants were randomized to intervention with home-based exercise before and after surgery or to usual care. In this observational study we included all participants who had reported on their baseline level of leisure time physical activity.

Exposure
At inclusion in the study, approximately two weeks before surgery, participants were asked to complete a baseline questionnaire on important demographic variables. This included the Saltin-Grimby physical activity level scale [15], a single item instrument where participants report their level of leisure time physical activity on a four-grade scale. The instrument has been validated for reproducibility and concurrent validity [15] It has been reported to predict several aspects of morbidity and mortality [15], and has been used previously for colorectal cancer patients [9,10]. In PHYSSURG-C, participants were asked to report their level of leisure time physical activity during the preceding week. This variable was used as exposure in the analyses described in this manuscript. As previously described [9,10,16,17], regular physical activity and regular hard physical training were combined due to low number of participants reporting regular hard physical training. The answering alternatives were:

Outcomes
The primary outcome measure in this manuscript was postoperative complications 90 days postoperatively, measured with the comprehensive complication index (CCI) [18]. This is a scale 0e100 derived from the Clavien-Dindo classification of surgical complications [19]. The scale is used as a continuous variable in all analyses. Information on postoperative complications was collected from medical charts by a single investigator blinded in terms of exposure, as previously described in detail [6]. Postoperative complications were chosen as primary outcome measure in these analyses instead of self-assessed physical recovery due to the general interest in including measures of postoperative complications in prehabilitation trials. The definitions for postoperative complications can be seen in the supplementary material. Secondary outcome measures were: Postoperative complications during index hospital stay and 30 days postoperatively, measured with comprehensive complication index. Complications up to 90 days postoperatively: cardiovascular events, respiratory insufficiency, pneumonia, infection, surgical leakage, and nausea (including paralytic ileus and small bowel obstruction). Information from medical charts. Total length of hospital stay during 90 days postoperatively, information from medical charts. Measured in days. Risk for re-operation during 90 days postoperatively, information from medical charts. Definitions described in detail previously [6]. Self-assessed physical recovery four weeks postoperatively. Participants categorized themselves as not recovered at all, recovered to 25%, 50%, 75%, or completely recovered. Described in detail previously [6].

Confounders
Since these analyses are observational, bias could occur due to confounding. To minimize the risk for confounding all analyses were adjusted for variables deemed to be possible confounders. These were included in a directed acyclic graph (DAG) and were decided upon before any analyses were performed. In the statistical models, we included the following confounders: Age. Continuous variable, calculated from birth date until date of surgery. Gender. Male/female. Smoking status. Active smoker vs non active smoker, reported by participants in baseline questionnaire. Abandoned in statistical analyses due to low numbers of smokers in the population. Risk consumption of alcohol, reported by participants using AUDIT-C [20]. Dichotomized as risk drinking (5 points) or no risk drinking (<5 points). Colon or rectal cancer, retrieved from screening log. Tumor stage according to clinical staging, UICC stage I-IV. Information from the Swedish colorectal cancer registry. Open or laparoscopic surgery, information from the Swedish Colorectal Cancer Registry.
As physical activity is known to reduce the risk for cardiovascular disease and overweight [1] we considered comorbidity and BMI to be possible mediating factors in the total effect of leisure time physical activity on the outcomes in our main analysis. Comorbidity was reported by participants at baseline while BMI was derived from the research nurse case report form at inclusion. Adjustment for group assignment (according to randomization) was performed as a sensitivity analysis.

Statistical methods
No sample size calculation was performed for the analyses reported in this manuscript. The statistical analyses were performed according to a pre-specified statistical analysis plan. All analyses were performed by linear models with group assignment (level of leisure time physical activity) as a fixed factor and age, gender, alcohol consumption, tumor site (colon or rectum), neoadjuvant therapy (none, radiotherapy, or chemo/radiotherapy), and type of surgery (open or laparoscopic) as adjusting covariates and are presented with 95% confidence intervals and p-values. For the primary endpoint (CCI 0e90 days), and the secondary endpoints CCI 0e30 days, length of hospital stay, self-assessed physical recovery four weeks postoperatively an ordinal logistic regression (proportional odds model) was used since this model make no distributional shape assumption [21]. For the specific postoperative complications, a binary logistic regression was used. Results are presented as an odds ratios, 95% confidence intervals and p-values.
Due to the explorative nature of the analyses, no correction for multiplicity was made. Missing covariate values were replaced using multiple imputations by predictive mean matching and 10 imputations [22]. The computations were performed with R, version 4.1.0 and the packages MASS, mice, gtsummary and forestplot.

Results
From January 22nd , 2015 until May 28th , 2020, 761 participants were recruited to the PhysSURG-C trial. After exclusions and removal of participants with no information on baseline level of leisure time physical activity, the study population here consisted of 614 participants (Fig. 1).
Two thirds of the population reported some physical activity, while 1/6 reported sedentary and 1/6 regular physical activity (Table 1). Participants who were sedentary or performed some physical activity were comparable regarding age, gender, level of education and.
Tumor stage was similar between the groups, while participants who were sedentary more often had open surgery than the other two groups. Participants with regular physical activity were more often receiving neoadjuvant therapy and had rectal cancer more often than the other two groups.
Participants who were not included due to missing information on baseline level of leisure time physical activity were similar to those reporting level of physical activity in terms of age, sex, BMI, and had similar tumor and surgical characteristics.
Our primary analysis showed that participants with some physical activity had a lower postoperative complications burden than sedentary participants, measured by CCI 90 days postoperatively (odds ratio (OR) 0.63, 95% confidence interval (CI) 0.43e0.92) (Fig. 2). This was also the case for those with regular physical activity (OR 0.60, 95% CI 0.36e0.99). Similar results were seen for complications during index hospital stay and 30 days postoperatively. There were no apparent differences between participants who were regularly physically active and those who performed some physical activity in terms of postoperative complications in our analyses.
The analyses of specific types of postoperative complications showed a lower risk for experiencing a postoperative respiratory insufficiency for regularly active than for sedentary participants (OR 0.40, 95% CI 0.18e0.87). Our analyses did not reveal any statistically significant differences between the groups for other outcomes. There were however trends for dose-response relationships for cardiologic complications with 8.4%, 5.6%, and 2.1% of participants in the three groups experiencing a cardiologic complication. A similar trend was seen regarding need for intensive care (grade IVa & IVb) with 11.8%, 6.2%, and 4.5% in the three groups (data not shown). There seemed to be no dose-response trends for nausea & vomiting, surgical leakage, or infections. There were no statistically significant differences between the groups for length of hospital stay, reoperations, or self-assessed physical recovery, although the point estimates for all these outcomes showed trends towards positive associations. For unadjusted results, see supplementary table.

Discussion
This study found that patients who had regular physical activity on average had a lower risk for postoperative complications after colorectal cancer surgery than sedentary patients.
There are few previous studies on the association between baseline physical activity and postoperative complications and recovery after colorectal cancer surgery, as shown by a systematic review [7]. Dronkers et al. reported that physical activity was associated with lower levels of in-hospital mortality, discharge destination and length of hospital stay in 169 patients with colorectal cancer [8]. They did not examine the association with postoperative complications. You et al. performed a propensity score observational study with 4632 individuals with colorectal cancer in Taiwan [11]. Participants were categorized as performing <12 metabolic equivalents (MET) or 12 MET. Unfortunately, they did not include information on how physical activity was measured, and did not assess postoperative morbidity using any validated instrument, which reduces validity and hinders interpretation of the results. They reported that participants with MET 12 had lower risk for overall complications, but that this could not be explained by differences in any specific type of complications. Whelan et al. recently reported results from a study on 125 patients with colorectal cancer in South Africa [12]. They reported an association between baseline physical activity and length of hospital stay, but not postoperative complications. In summary, the only previous report of the association between habitual physical activity and postoperative complications where a validated instrument for measuring complications is our previous report from a similar but considerably smaller cohort.
We have previously found in a study with 115 patients that participants with regular physical activity were more than 3 times more likely to feel highly physically recovered three weeks postoperatively than sedentary participants, and that sedentary participants had a CCI 12 points higher than participants with some physical activity and 17 points higher than participants with regular physical activity [9,10]. The effect sizes seen in our previous study could not be confirmed by the present study, where the differences between the groups were smaller. The overall level of postoperative complications was also higher in the present study, which is probably explained by a more rigorous collection of information of postoperative complications. Interestingly, the difference between the groups in terms of absolute burden of postoperative complications is the same between the two studies. Since CCI is the result of a squared root function, a decrease from CCI 32 to CCI 24 corresponds to a decrease by a single grade II complication. The difference between the groups of exposure in the present study could perhaps be explained somewhat by the randomized intervention as 50% did exercise more than usual, but we found no average efficacy and it had no influence on the results as reported by the supplementary analysis (data not shown). This should reduce the risk for the intervention masking actual associations between baseline physical activity and postoperative complications and recovery.
There are several intervention studies on the effect of preoperative exercise interventions before colorectal cancer surgery. These have reported mixed results, with two recent trials reporting no effects on postoperative complications or other clinically relevant outcomes [5,6]. However, a previous trial including six weeks of high intensity training in before major abdominal surgery reported impressive results with a 50% decrease in postoperative complications [4]. Interestingly, they also reported effect mainly on medical complications such as cardiovascular complications and respiratory complications, while there was no effect on surgical complications except for paralytic ileus. This is in line with the results from our present study where physical activity was associated with a lower risk for cardiologic complications and respiratory insufficiency, but not surgical complications. We also saw a trend towards more ICU care in sedentary patients, possible driven by more medical than surgical complications. This information could be used when planning future interventional studies. It would be more rational to aim these at improving rate of cardiovascular and respiratory complications, and possibly the need for ICU than the combined burden of postoperative complications, as there are many other factors influencing surgical complications. Also, these results can be used to calculate risk for medical complications, which may influence the shared decision-making prior surgery.
This study has several limitations. The most important being that it is an ad hoc observational analysis within an RCT. This must be kept in mind when interpreting the results. However, the method of performing observational studies on associations between physical activity or functional capacity and postoperative outcomes within an RCT has been done before. Gillis et al. performed an observational analysis of frail participants with colorectal cancer who had low functional capacity within a RCT of preoperative exercise before colorectal cancer surgery [23]. They reported that participants with low functional capacity at the time of surgery were more likely to suffer postoperative complications. Meyerhardt et al. performed a similar observational study within a randomized controlled adjuvant chemotherapy trial [24]. They reported that physical activity six months after completion of therapy was associated with approximately a 50% reduction in cancer recurrence or death from any cause. Performing observational analyses within randomized trials has both pros and cons. Participants are generally followed up in detail. However, the intervention introduces a possible confounding. Since we have reported previously that groups were balanced for the intervention, and that the intervention had no significant effect on the outcomes reported in the present manuscript we believe that randomization should cause no serious bias in the present study. Another limitation in this study is that it is observational, which makes it harder to claim causality. This is even more true in the case of habitual physical activity usually a long habit where it is hard to conclude the effect  of short-term changes without performing interventional studies. Hence, these results should not be used for recommending preoperative exercise for colorectal cancer patients, but to guide policy makers developing recommendations and strategies for achieving a more physically active general population. This study also has several strengths. One is the large sample size in combination with prospectively collected data, improving statistical power and enabling analysis of different types of complications. Another is the use of validated and well recognized instruments for both physical activity and postoperative complications increasing the possibility to compare results with future studies. Collection of information on outcome measures in a structured way by a researcher blinded regarding exposure further increases validity.

Conclusions
Participants who engaged in preoperative leisure time physical activity had on average fewer and less severe postoperative complications after colorectal cancer surgery compared with patients with a sedentary life-style. Information about an individual's physical activity at baseline could be included into the considerations of possible treatment choices preoperatively. As yet another argument in the ongoing public efforts to achieve a physically active population, these results may fit somewhere in such campaigns, but should not be used as an argument for recommending preoperative exercise for patients diagnosed with colorectal cancer in order to improve postoperative recovery. , and the healthcare committee Region V€ astra G€ otaland (H€ also-och sjukvårdsstyrelsen) (VGFOUREG-309261, VGRFOUREG-659011, VGRFOUREG-837542, VGRFOUREG-739191). The funders of the study had no role in study design, data collection, data analysis, data interpretation, or writing of the report. The corresponding author had full access to all the data in the study and had final responsibility for the decision to submit for publication.

Data statement
Individual participant data that underlie the results reported in this article, after de-identification (text, tables, figures, and appendices), as well as study protocol, statistical analysis plan, and analytic code will be shared with investigators whose proposed use of the data has been approved by an independent review committee ("learned intermediary") identified for this purpose, and after assuring that this data-sharing complies with ethical permissions. Data may be used for individual participant data meta-analysis. Data sharing proposals should be sent to the corresponding author.

Declaration of competing interest
All authors have completed the ICMJE uniform disclosure form at www.icmje.org/coi_disclosure.pdf and declare: EA and DB declare no support from any organization for the submitted work: AO reports grants from the Swedish state under the agreement between the Swedish government and the county councils, the ALF-agreement, grants from Assar Gabrielsson's foundation, grants from Anna-Lisa and Bror Bj€ ornsson's foundation, grants from Dr Felix Neubergh's foundation, grants from Gothenburg medical society, grants from Lions cancer fund west during the conduct of the study: EH reports grants from AFA insurance and grants from the Swedish state under the agreement between the Swedish government and the county councils, the ALF-agreement, and grants from The Swedish Cancer Society during the conduct of the study; AO, EA, and DB declare no financial relationships with any organizations that might have an interest in the submitted work in the previous three years: EH reports grants from Mary von Sydow Foundation, grants from the Swedish Research Council, and grants from G€ oteborg Medical Society outside the submitted work; all authors report no other relationships or activities that could appear to have influenced the submitted work. All authors confirm that they had full access to all the data in the study and accept responsibility to submit for publication.