Eligibility rate

The proportion of people with CRC who were assessed and found eligible in CRIB was 85% and the proportion of people with CRC who were found eligible from the total number of registered patients was 67%. As Table 37 shows, these figures compare favourably with other studies.

The studies reported in Table 37 have different inclusion and exclusion criteria, which will obviously influence the proportion of assessed participants who are judged as eligible to participate in a study. The ways in which these criteria are likely to impact eligibility rates and also likely to influence study bias and generalisability of results are discussed below.

Table 38 shows that 6 (including CRIB) out of 13 studies specified that only people aged ≥ 18 years were eligible;^{49,145,146,148} two studies excluded people who were aged > 75 years^143,150 and another excluded people who were aged > 65 years.¹⁴² Given that it is very rare for young people to have a CRC diagnosis, we can confidently assume that a lower age criterion does not have a decisive influence on eligibility rates. Nine studies (including CRIB) included language restriction as a criterion,^{44,49,142,143,145,146,149,150} suggesting a bias towards people who speak a country’s first language and thereby potentially limiting generalisability of the study findings to other groups of the population. In CRIB, no patients with CRC were excluded because they were unable to speak English, suggesting that this criterion did not have any bearing on the eligibility rate. However, language restriction may impact on eligibility rates in any future multicentred trials that are likely to include areas in which some people do not have English as their first language.

TABLE 38

Eligibility criteria studies

One inclusion criterion that is likely to influence generalisability of study results is level of physical activity. As Table 38 shows, two studies were specifically designed for people who were not currently physically active,^48,49 whereas other studies did not apply physical activity behaviour as part of inclusion/exclusion criteria. Thus, most studies may have included people who were physically active, although as we highlighted in Chapter 1, evidence suggests that most people with CRC are unlikely to be meeting recommended levels for physical activity, which makes it probable that most participants in these studies were not physically active when they started the trial. Nonetheless, our baseline data show that most participants were already meeting or nearly meeting recommended levels for physical activity, suggesting that CRIB attracted people who are already physically active and therefore less in need of an intervention designed to increase people’s level of physical activity associated with health benefits (i.e. 30 minutes per day of moderate physical activity).

Another inclusion criterion that is likely to influence generalisability of study results is diagnostic inclusion criteria, which varied across studies. As Table 38 shows, CRIB excluded people with advanced disease but, as we discussed in Chapter 5, this is an ambiguous term. CRIB did not exclude people with CRC who had metastatic disease. Six studies specifically mentioned that people with metastatic disease would be excluded or only people with local disease would be included.^{49,143,145,146,148–150} Thus, there appears to be a bias towards people with early-stage disease, thereby limiting generalisability of findings to people with metastatic disease.

Most studies included only people who had completed active treatment. Only one study was specifically designed for people who were on active treatment (i.e. receiving adjuvant chemotherapy);¹⁴⁸ one study (CRIB) included people on or off active treatment, seven studies made it clear that people on active treatment would be excluded,^{49,142–144,147,149,150} and it was not clear in four studies if people on active treatment were included or not.^44,144–146 We are aware of only one current physical activity intervention trial for people with CRC on active treatment.¹⁵¹ Thus, there appears to be a bias towards people who have completed active treatment. Our study suggests that there may well be good reason for only including people with CRC post treatment; one of the main reasons why participants who were interested in participating changed their minds was because they felt unable to partake in an exercise class while having chemotherapy. Thus, for practical reasons alone, a post-treatment trial may be more feasible.

Virtually all studies had exclusion criteria for contraindications for physical activity, which was the most common reason for exclusion. As shown in Chapter 7, most people with CRC who were excluded from CRIB fell into this category because they had a medical problem (46%). Moreover, as described in Chapters 4 and 9, CRC nurses were not approaching some people with CRC about the study because they believed that they were unable to attend, or would not be interested in attending, cardiac rehabilitation owing to poor health. The CHALLENGE trial has recently reported that staff did not approach people who ‘do not look like an exerciser’.¹⁵² Similarly, most people who were excluded from a study conducted by Courneya et al.⁴⁴ fell into the clinical category (55% excluded owing to a medical condition). It is likely, therefore, that this particular exclusion criterion explains why most people are excluded from physical activity trials.

This criterion, therefore, ought to be clearly defined and explained to the clinicians and investigators involved in screening people with CRC for eligibility so that they know what the contraindications for physical activity are and can apply criteria competently. Training of recruiters is therefore required. Our study aimed to include people with a stoma, on active treatment and with a cardiac condition, even though these are known potential, albeit not automatic, contraindications for physical activity. Moreover, cardiac rehabilitation does accommodate people with poor mobility, wheelchair users and those who are very frail, and therefore we also did not set out to exclude people with these difficulties. If this exclusion criterion were to be applied too stringently, then people with CRC who can potentially benefit from the physical activity intervention would be excluded. As we showed in Chapter 7, our study suggests that cancer care nurses who screened patients were excluding people with poor mobility, suggesting that cancer care nurses may require additional information about the ways in which cardiac rehabilitation accommodates people who face mobility difficulties.

Excluding people from studies, however, is not inherently incorrect; what is important is that it is made clear which subgroups of the sample population are included and excluded and that reasons are given. To improve reporting of research and the ability to make reasonable judgements about the relevance of particular physical activity interventions for subgroups of the sample population, we recommend the development of standardised inclusion and exclusion criteria. Table 39 presents potential minimum inclusion and exclusion criteria for future trials of physical activity interventions for people diagnosed with CRC.

TABLE 39

Minimum criteria for trials of physical activity interventions for CRC survivors

Consent rate

Table 37 shows that our study consented 31% of all eligible patients, which is slightly lower than an average of 40% (range 8–98%). It is also lower than our estimated target of 66 patients across the three sites. However, if 31% is a proxy for the number of people with CRC likely to take up the offer of cardiac rehabilitation, should this service be established as part of routine cancer care, then attendance of people with CRC would be 12% lower than the number of patients with CHD who attended phase 3 cardiac rehabilitation in 2011–12 (43%).⁶² Given that cardiac rehabilitation for people with CHD is a well-established service that has been audited by the British Heart Foundation since 2004, a rate of 31% engenders optimism that uptake among people with CRC would eventually match attendance rates among people with CHD. Nevertheless, we acknowledge that consent rates for research and for an actual service are not directly comparable.

Table 37 suggests that consent rates are unrelated to research design or intervention mode. Not all studies included reasons why eligible participants declined to participate in a study, but, of those that did, ‘medical conditions’ and ‘not interested’ were the most common reasons given for eligible participants not consenting. Table 40 shows the reasons why eligible participants did not consent in those studies that provided these data.

TABLE 40

Reasons why eligible participants do not consent (%)

Based on our study and those of others, we anticipate that one of the main barriers to participation in physical activity interventions is a person’s health. In our study, for instance, some eligible participants were unwilling to participate because they were receiving adjuvant chemotherapy and did not perceive that they would feel well enough to attend an exercise class. Some of the health difficulties that people with CRC experience that may impede involvement in a physical activity intervention are discussed further in the next chapter, which reports and discusses the findings from our qualitative study.

Of note, many of the barriers that we found are not dissimilar to reasons why people with CHD do not attend cardiac rehabilitation. A recent audit of UK cardiac rehabilitation, for instance, found that 33% did not attend because they were not interested, 7% did not attend because of travel difficulties, 9% did not attend because of physical incapacity and 3% did not attend because they were too ill.⁶² Addressing some of the barriers to people attending rehabilitation will, therefore, be relevant to people with CHD as well as people with CRC.

Retention, completion rates and missing data

Loss of consenting participants to a study may be due to a combination of factors, including participants formally dropping out of the study (retention rate), failing to complete outcome measures (completion rate) or failing to provide valid data (missing data). Loss of participants during trial follow-up can introduce bias and reduce power, affecting the generalisability, validity and reliability of results.¹⁵⁴ Thus, information about retention and completion rates and missing data is important for assessing bias.¹⁵⁵ It has been estimated that a 20% loss can threaten trial validity.¹⁵⁴ Some missing data can be dealt with statistically; nevertheless, the risk of bias due to missing data can remain¹⁵⁶ and, therefore, should be reported alongside other rates.

Table 37 shows completion rates for CRIB and other studies. Our completion rate was slightly below average in comparison with other studies. The ongoing CHALLENGE trial has also experienced losses to the study;¹⁵² at first follow-up, 189 out of 250 (75%) randomised participants completed outcomes measures, and at second follow-up it was 141 out of 250 (56%) randomised participants. Our completion rate for patient self-reported questionnaires (i.e. most studies used self-report to measure outcomes) at first follow-up was 31 out of 41 randomised participants (75.6%), whereas other studies’ completion rates ranged between 79% and 95%. Our completion rate at second follow-up was 25 out of 41 (61%). This difference may be a reflection of differences between pragmatic and explanatory trials. CRIB was a pragmatic trial of an already existing service (i.e. cardiac rehabilitation), whereas all of the others were explanatory trials. In CRIB the intervention was independent of the actual study and it may be that participants were committed to the service (evidenced by high intervention adherence) but not to the actual study. Any future trial of CRIB should, therefore, introduce strategies to improve completion rates. A 2014 systematic review¹⁵⁷ of 38 randomised retention trials evaluating six broad types of strategies to increase questionnaire response and retention in randomised trials concluded that no strategy had a clear impact on increasing the number of participants returning to sites for follow-up but found that the following strategies may improve questionnaire response: addition of monetary incentives for return of postal questionnaires, recorded delivery of questionnaires, and a ‘package’ of postal communication strategies with reminder letters.¹⁵⁷

Accelerometer validation

Our primary outcome was physical activity. Objective measures of physical activity and sedentary behaviour have been increasingly used to overcome limitations of self-report measures. Research conducted among the general population suggests that self-reported measures of physical activity and sedentary behaviour are inaccurate when compared with objective measurement from devices such as accelerometers.^158–160 A 2014 study comparing accelerometer-based and self-reported measures of recent moderate- to vigorous-intensity physical activity (MVPA) and sedentary time in colon cancer survivors found that total mean minutes per day spent in MVPA was 12 minutes based on accelerometer data and 26 minutes based on self-reported data (p < 0.01).⁹⁷ Correlation between the methods was fair (Spearman’s rank-order correlation = 0.51); however, agreement was poor [intraclass correlation coefficient (ICC) = 0.33]. Mean daily time spent sedentary was similar in both methods (≈ 8.5 hours); however, both correlation and agreement were poor (Spearman’s rank-order correlation = 0.19, ICC = 0.16). Pinto et al.⁴⁹ also found poor to fair agreement between self-reported and accelerometer-derived physical activity. Our research used objective (accelerometer) and self-reported measures of physical activity and also found poor agreement between the two measures.

Despite the advantages of obtaining an objective measure of physical activity and sedentary behaviour, there are few guidelines for using accelerometers in research^161,162 and little guidance on improving participant compliance.¹⁴⁷ Our study shows that 31% of accelerometer data sets were invalid, mainly because participants did not wear the device. Some of the challenges of using accelerometers in research involving people with CRC are described in Chapter 9 and include difficulties wearing the device around the waist after abdominal surgery. Recommended approaches for improving compliance include a daily monitoring log filled out by participants, reminder telephone calls, adequate education about the monitor and its proper wear, and the identification of potential barriers to wearing with each participant.¹⁶³

Intervention adherence

Intervention adherence refers to the extent to which participants randomised to the intervention group follow specific treatment therapy instruction as per intervention protocol. Low adherence increases the risk of policy and service commissioners rejecting physical activity interventions that may actually be effective should compliance levels be high. Addressing the problem of adherence is, therefore, important. As reported in Chapter 7, our study shows that 62% of participants randomised to the intervention group completed cardiac rehabilitation and that the main reason for either not starting cardiac rehabilitation or stopping it was poor physical health. Table 41 shows adherence rates of physical activity interventions for people with CRC. It is difficult to draw direct comparisons because studies use different ways of measuring intervention adherence. Nonetheless, CRIB compares favourably with most studies that use exercise classes as the intervention mode.

TABLE 41

Intervention adherence