The HIT-AAA project was a three-site, two-arm, parallel-group, randomized controlled feasibility study (trial registration ISRCTN09433624) approved by the North East-Tyne & Wear South Research Ethics Committee (13/NE/0116). Study recruitment was undertaken from August 2013 to December 2015 and all participants provided written informed consent. The HIT intervention was performed in the 4 weeks preceding surgery and the exercise protocol was based on prior HIT programmes shown to be safe and effective for improving cardiopulmonary fitness in patients with heart failure (Wisløff et al., 2007) and coronary heart disease (Rognmo et al., 2012). While the study's protocol, inclusion and exclusion criteria have been published elsewhere (Tew et al., 2014), for the purpose of this paper the methods pertaining to the HIT intervention are described below.

A total of 27 (two female) participants (mean ± SD age: 74.3 ± 5.7 years, height: 172.3 ± 9.0 cm, body mass: 79.1 ± 15.9 kg, aneurysm size: 6.0 ± 0.4 cm) were randomized to the HIT intervention. Following study enrolment, all participants underwent a baseline cardiopulmonary exercise test (CPET) on a cycle ergometer. The mean baseline anaerobic threshold (AT) and peak oxygen consumption (VO₂peak) were 11.0 ± 2.1 mL/kg/min and 16.5 ± 3.7 mL/kg/min, respectively. Mean recorded baseline power output on the CPET was 54 ± 19 watts at AT and 98 ± 29 watts at VO₂peak.

During the first HIT session, all participants exercised to the power output observed at the AT determined on the baseline CPET. In subsequent sessions, power output was increased until the patient reported ratings of perceived exertion (RPE) of five (“hard”) to seven (“very hard”) on Borg's CR-10 scale (Borg, 1982) at the end of each interval. The precision of RPE data during exercise can, however, be enhanced by differentiating perceptual reports according to their specific mediators with local and central being regarded as the most important signals (Borg et al., 2010). As such, we collected separate (differential) RPE scores for the perceived exertion in the legs (RPE-L) and the perceived sense of breathlessness in the chest (RPE-C). Each patient was familiarized with the scale and the recommended researcher instructions for scale administration were used (Borg, 1982). The research nurse and physiotherapist supervising each HIT session recorded power output (watts), blood pressure (manually via sphygmomanometer) and RPE at the end of each interval. Heart rate data were recorded continuously at 5-s intervals throughout the entire exercise session (Polar RS400, Kempele, Finland) with data download procedures as per Taylor et al. (2015).

Participants completed three HIT sessions per week throughout the 4-week training period with 48 h recovery in between sessions (e.g., Monday, Wednesday, Friday). All exercise sessions were hospital based, supervised by a physiotherapist and research nurse and performed on a cycle ergometer (Optibike Med, Ergoline, Germany), with sessions performed >3 h after waking given the higher frequency of cardiovascular events during the morning hours (Thompson et al., 2007). Each HIT exercise session commenced and finished with a 10-min warm up and 5-min cool down of unloaded cycling. For HIT sessions one to three, all participants performed 8 × 2 min repetitions, with each interval interspersed with a 2-min period of active (unloaded cycling) or passive (rest) recovery. Following this, for HIT sessions 4–12 participants had the choice of either the 2-min protocol or a 4-min protocol which consisted of 4 × 4 min repetitions with each interval interspersed with a 4-min period of active (unloaded cycling) or passive (rest) recovery. Where surgery was delayed beyond week five, participants entered a maintenance phase of training consisting of one HIT session per week. Here, the participants had a choice of which HIT protocol to undertake, either 8 × 2 min or 4 × 4 min.

Power output in any exercise session was maintained or reduced (termed workload reduction) if systolic blood pressure (SBP) exceeded 180 mm Hg (Isselbacher, 2005) or if heart rate exceeded 95% of the maximum observed on baseline CPET.

In our protocol, a participant was deemed compliant if they completed ≥75% of the scheduled sessions, that is, at least 9/12 sessions for the 4-week intervention, plus all once-weekly maintenance sessions if surgery was delayed (Tew et al., 2014). Across the 4-week intervention period, 20 out of 27 participants attended a minimum of 75% of the scheduled intervention sessions. Fifteen of the 27 participants entered the maintenance phase of the study, with 36 out of a total of 40 (90%) prescribed sessions attended. The mean number of maintenance sessions offered per participant was 1 (range 0–9).

Overall, a total of 17/27 participants (63%) met our compliance criteria. Of this cohort, the mean number of HIT sessions attended (per participant) during the 4-week intervention and subsequent maintenance phases was 11 (range 9–12) and 1 (range 0–4), respectively. By comparison, the 10 non-compliant participants attended a mean (range) of 6 (0–12) HIT sessions during the 4-week intervention and 2 (0–9) during the maintenance phase. Our CONSORT diagram (Figure 1) summarizes participant flow, reasons for all missed HIT sessions and number of participants included in the fidelity analysis for compliant and non-compliant groups. Given the relatively low number of HIT sessions performed by the non-compliant participants, our in-depth analysis of the exercise data is confined to the 17 compliant participants.

Differences in baseline fitness variables between compliant and non-compliant participants were examined using an unpaired t-test, with magnitude based inferences subsequently applied (Hopkins, 2007). Here, inferences were based on standardized thresholds for small, moderate and large differences of 0.2, 0.6, and 1.2 of the pooled between-subject standard deviations (Hopkins et al., 2009). The chance of the difference being substantial or trivial was interpreted using the following scale: 25–75%, possibly; 75–95%, likely; 95–99.5%, very likely; >99.5%, most likely (Batterham and Hopkins, 2006). Descriptive statistics were used to calculate the session attendance data. The difference in the proportion of workload reductions between compliant and non-complaint participants was calculated, with uncertainty in the estimates expressed as 90% Confidence Limits (CL). We determined the proportion of HIT repetitions that met our prespecified RPE criteria for high-intensity, along with the median and interquartile range (IQR) for these proportions. Linear mixed modeling (SPSS v.23, Armonk, NY: IBM Corp) was used to calculate the effect of compliance (compliant, non-compliant), interval duration (2 min, 4 min) and training phase (intervention, maintenance) on our measures of exercise intensity (power, RPE-L, RPE-C, heart rate) and also to examine the difference between differential RPE scores (RPE-L, RPE-C), with magnitude-based inferences subsequently applied. For all variables, we classified the magnitude of effects mechanistically, whereby if the 90% confidence limits overlapped the thresholds for the smallest worthwhile positive and negative effects the effect was deemed unclear (Hopkins et al., 2009). The within-subject variability (expressed as a standard deviation [SD]) in RPE-L and RPE-C was determined via a linear mixed model with a random intercept, with the SD doubled to interpret its magnitude (Smith and Hopkins, 2012). For the assessment of exercise progression in power, RPE-L, RPE-C, and heart rate across the 4-week HIT intervention we applied a linear mixed model. “Session” (1–12) was entered as fixed effect, with a random slope and intercept for session (unstructured covariance matrix). We used the slope of the relationship between session and outcome to derive the percentage change (power output) or raw change (RPE, heart rate) over the 4-week intervention.

Demographic and baseline fitness levels for all 27 patients randomized to the HIT-AAA exercise intervention are displayed in Table 1. When compared to non-compliers, study compliers had a higher baseline AT, baseline VO₂peak and power output recorded at AT and VO₂peak with the magnitude of all effects being likely small (possibly moderate). The between-group differences in age, height, body mass and aneurysm size were unclear.

Attendance and safety data for the exercise sessions are presented in Table 2. Overall session attendance was 74% for the 4-week intervention and 95% for the maintenance phase of the study (see Figure 1). Attendance was higher in the compliant participants and there were clear patient preferences for 2-min intervals. Thirteen (of 17) compliant and seven (of 10) non-compliant patients experienced workload reductions in-line with our safety criteria. The difference in the proportion of workload reductions between compliant and non-compliant participants was 7.5% (±90% Confidence Limits 2.5%). One participant (a complier) experienced an adverse event (angina episode) requiring the use of Glyceryl trinitrate spray 29 min into exercise session eight (session completed). This participant missed session nine due to cardiology review, but returned to complete the remaining three exercise sessions plus one maintenance session. One participant (a non complier) was investigated for prodromal symptoms of dizziness during exercise. Subsequent cardiac evaluation was normal but the participant was not able to resume training due to the time taken for referral.

Mean exercise intensity data are presented in Table 3. When compared to non-compliers, power output (magnitude of effect—likely small/ possibly moderate), RPE-C (likely small) and RPE-L (possibly small) were higher for the compliers. The differences in heart rate and systolic blood pressure were unclear. For the study compliers, the percentage of HIT repetitions meeting the compliance criteria for high-intensity training was 30% (IQR 16%, 68%) for RPE-L and 16% (10%, 42%) for RPE-C. Analysis of the RPE scores revealed a most likely small difference (0.6 Arbitrary Units [AU]; 90% confidence limits ±0.1 AU) between RPE-L and RPE-C. The magnitude of the within-subject variability was large for both RPE-L and RPE-C.

The effect of interval duration (Table 4) on the exercise intensity of the compliant participants was higher RPE-L (likely small) and RPE-C (possibly small) for the 2-min intervals and higher SBP (very likely small) and heart rates (likely small) on the 4-min intervals. The effect on power output was trivial. Also presented in Table 4 is the effect of study phase on exercise intensity. Here, RPE-L (very likely small), RPE-C and heart rate (both likely small) were higher during the 4-week intervention phase of the study, with power output higher (likely small) during the study's maintenance phase. The effect on SBP was trivial.

Presented in Figure 2 are the mean exercise intensity scores per HIT session along with the individual data points (compliers only) to illustrate the variability around the mean score. The regression slope revealed an increment of 23% (90% Confidence Limits ±12.5%) in power output across the 12 HIT sessions with a between-subject variability of ±23% (±10%). The rate of change in RPE-L and RPE-C across the 12 HIT sessions was −0.7 AU (±0.5 AU) and −0.8 (±0.7 AU), respectively with a between-subject variability of 1.0 AU (±0.4 AU) and 1.3 AU (±0.5 AU). There was an increment in heart rate of 3.1 percentage points (±2.9 percentage points) with a between-subject variability of 5.4 percentage points (±2.6 percentage points).

Our overall session attendance for the 27 participants across the 4-week HIT intervention compares favorably with an attendance range of 58–77% for older adults undertaking exercise programmes (Picorelli et al., 2014). The figure is, however, considerably below the 94% reported by Tew et al. (2012) for an endurance exercise intervention performed by patients with small AAA disease in a surveillance programme. Given that greater cardiorespiratory endurance (Rhodes et al., 1999) and better physical function (Picorelli et al., 2014) are associated with better adherence, such discrepant findings may well be explained by our participants being older, having larger aneurysms and lower cardiorespiratory fitness. As expected, there were clear differences in the session attendance rates between compliant and non-compliant participants, which again could likely be a consequence of fitness given that our compliant participants had substantially higher cardiorespiratory fitness levels. Further, the higher level of cardiorespiratory fitness and ability to cycle to higher power outputs explains why our complaint participants were able to exercise at higher absolute and relative exercise intensities throughout the study and possibly why they recorded fewer safety workload reductions. It has been previously reported from meta-analyses that HIT favors the less fit (Weston M. et al., 2014; Milanović et al., 2015) which poses an interesting juxtaposition as lower fitness in the present study was associated with non-compliance; although participant fitness and health in the present study were substantially lower than that reported in the afore-mentioned meta-analyses.

Evaluations of fidelity in exercise training interventions should ideally address session attendance and compliance (meeting the prescribed exercise intensity), as this interaction constitutes the dose of the intervention and influences the physiological response to exercise training (Taylor et al., 2015). Furthermore, while it is difficult to classify intervention fidelity, Taylor et al. (2015) felt that high-intensity criterion attainment in a little over half of the HIT repetitions per participant, represented moderate intervention fidelity. With this in mind, the overall fidelity of our exercise intervention is probably best described as low, given the poor compliance to our predetermined RPE criteria for high-intensity (~23%) and the overall intensity of the HIT repetitions being lower than prescribed. Indeed, our mean RPE scores fell in the range of moderate to somewhat hard on the CR10 scale, when the lower limit of our target was hard. Although patient medication (e.g., beta blockers) blunt the exercise heart rate response, our training heart rates support the moderate to hard intensity of the exercise intervention as a mean peak heart rate of 82.2% is slightly below that required for HIT (e.g., 85–95% of peak heart rate; Weston K. S. et al., 2014).

While mean exercise session data provide valuable information, the values do not quantify the degree of consistency in exercise intensity across the intervention (Taylor et al., 2015). In the present study, the observed magnitude of the within-subject variability in RPE was large. Such variability illustrates an inconsistent exercise dose across the intervention as it encompasses a vista of intensities from easy through to hard (Borg, 1982). For example, when the within-subject variability for RPE-C (1.6 AU) is added to and subtracted from the mean of 3.5 AU, the estimate for RPE-C typically varies from easy (1.9 AU) to hard (5.1 AU). Such variability further supports low fidelity as the exercise dose was not applied consistently across the intervention.

The success of our prescribed exercise intensity contrasts with previous exercise training studies undertaken on AAA patients. Tew et al. (2012) reported on a thrice weekly, 12-week exercise programme targeted to fall within the range of 12 to 14 on the Borg 6–20 scale. The authors concluded that moderate-intensity endurance exercise training (mean RPE, 11.8 ± 0.8 AU, mean heart rate, 72 ± 8% of age-predicted maximum heart rate) is feasible in patients with small abdominal aortic aneurysms (mean aneurysm size 4.1 ± 0.1 cm). In another study examining the effects of exercise training in AAA patients (aneurysm size 3.0–5.0 cm), Myers et al. (2013) reported a mean training intensity of 98 ± 7% of target heart rate, which was initially 60% of heart rate reserve, increasing to 80%, with perceived exertion was targeted to fall within the range of 12–14 on the Borg 6–20 scale. Kothmann et al. (2009) also used a moderate-intensity exercise dose (12–14 on the Borg scale) in their pilot study examining the effect of short-term exercise training on aerobic fitness in AAA patients (aneurysm size 3.0–5.1 cm); however, the authors did not report any exercise training data. While a full comparison of between-study exercise intensity is not possible given substantial inconsistency in the detail of exercise data presented, clear differences in prescribed exercise intensity, namely moderate versus high-intensity, are apparent, yet the reported intensities appear similar (e.g., a moderate to somewhat hard intensity).

Of interest in the Kothmann et al. (2009) study was that the authors prescribed moderate continuous exercise rather than high-intensity interval training for “safety reasons.” The safety concerns surrounding AAA patients when exercising are combined excessive rises in SBP and heart rate causing aneurysm rupture, which is a physiologically catastrophic insult carrying a mortality of about 75% (The Vascular Society, 2011). Precise exercise training safety criteria are, however, lacking from previous AAA patient exercise training studies (Kothmann et al., 2009; Tew et al., 2012; Myers et al., 2013). Furthermore, aneurysm size in these studies was small (~40 mm), whereas we recruited patients with large abdominal aortic aneurysms (55–70 mm diameter), which reinforced the need for our exercise training safety guidelines as SBP should not rise above 180 mm Hg for AAA patients wishing to engaging in vigorous aerobic exercise (e.g., running or cycling; Isselbacher, 2005; Myers et al., 2012).

There were no serious adverse events. There was one adverse event in one of the 17 compliant participants in a total of 55 person-hours of exercise—an event rate of 1.8 per 100 person-hours. The 95% simple Bayesian confidence interval (Ludbrook and Lew, 2009) for this rate is 0.4–9.6 events. The upper limit of the confidence interval may be viewed as an estimate of the maximum risk consistent with the data (Batterham et al., 2014); the chance of an adverse event could be as high as around one in every 10 person-hours. However, proportions or rates derived from small integer counts in small samples are unstable, as a small change in numerator can affect the estimates substantially. For example, if we had observed no adverse events in our study the upper confidence limit would be 6.4 events/100 person-hours, versus 14.9/100 person-hours with three adverse events. Clearly, much more data is required to evaluate the safety of the intervention properly in this patient group.

We found substantially higher RPE scores (~0.5 AU) following 2-min intervals when compared to 4-min intervals despite trivial or small differences in power output during the study intervention and maintenance phases, respectively. Shorter intervals have recently been perceived as requiring less effort, with frequent breaks from severe intensities possibly contributing to the lower perceived effort (Kilpatrick et al., 2015). In contrast, other authors have reported either no difference in RPE between 1- and 4-min intervals despite power output being significantly higher during the shorter intervals (Tucker et al., 2015) or that the RPE of interval training programmes is more strongly related to work intensity than accumulated duration (Green et al., 2009). Heterogeneous study populations with a variety of exercise experience (e.g., patients, obese/overweight, active) may contribute to these disparate findings. Further research on the interplay of interval duration, power output (or treadmill speed) and RPE in patient populations is therefore recommended. This recommendation is further emphasized by the juxtaposition of recent HIT research whereby shorter intervals are more palatable than longer intervals for novice exercisers (Kilpatrick et al., 2015), yet there is an increased adaptive response on VO_2max following longer intervals (Milanović et al., 2015).

Our primary focus has centered on exercise intensity, yet a detailed collection and analysis of exercise training data permits the objective appraisal of success in other aspects of a training study, namely exercise progression. To facilitate a positive adaptation to training, the prescription of exercise needs to advance over time. Very often training intensities are re-defined and re-prescribed following a mid-programme assessment of exercise capacity (West et al., 2015); however, this is at greater fiscal cost and the assessments can disrupt the training programme (Weston et al., 2016). In the present study we used relative measures of exercise intensity, as RPE scores provide a practical and valid means of ensuring training progression is inherent within programmes (Weston et al., 2016). Despite our exercise data showing a lower than prescribed exercise intensity, a substantial increase in power output across the 12 HIT sessions demonstrates clear exercise progression. We do, however, acknowledge that there was substantial between-participant variability in this estimate. While it is difficult to reconcile the magnitude of this progression with other clinical exercise trials—previous work largely focuses on training outcome as opposed to process—our data are higher than that reported for young sedentary subjects (18%) (Foster et al., 2015) and elite athletes (4.6%) (Purkhús et al., 2016). We also found a reduction in RPE across the duration of the intervention, a finding consistent with recent successful HIT programmes in healthy middle-aged men (Saanijoki et al., 2015), sedentary (Astorino et al., 2016) and overweight women (Smith-Ryan et al., 2016). Of these studies, however, only Saanijoki et al. (2015) quantified exercise progression over the same time period during which RPE decreased.

As heart rate can be an unreliable measure of exercise intensity in patients on medication, we used differential RPE scores to enhance the sensitivity of the exercise data we collected. Consistent with previous studies, we found peripheral ratings of exertion to be higher than central ratings (Pandolf et al., 1975; Green et al., 2009; Borg et al., 2010; McLaren et al., 2016). The magnitude of the difference in the two perceptual responses was small, which is consistent with recent reports in athletic populations (McLaren et al., 2016). This finding helps to explain the lower compliance for RPE-C as our high-intensity criterion was the same for both measures of RPE. Previously, Pandolf et al. (1975) reported substantially lower RPE-L following a 6-week period of cycling training, yet no reduction in RPE-C. The authors concluded that local factors dominate the exertional perception when cycling and that focus on reporting these sensations could interfere with the perception of central factors. In the present study, however, the magnitude of reduction in RPE-L and RPE-C over the duration of the study was consistent (e.g., −0.7 AU), suggesting the absence of any such reporting inference and showing that differential RPE do indeed offer a sensitive evaluation of exercise intensity.

A major limitation of the present study was that we were unable to exercise our participants to the prescribed exercise intensity, namely high intensity. We cannot, however, rule out the possibility of underreported RPE scores due to the influence of observer sex. Males report lower RPE values when a female observer, as opposed to male, is in the room (Winchester et al., 2012) and all but one of our participants were male and at each site RPE data were collected by female nurses and physiotherapists. As such, Halperin et al. (2015) recently recommended that the sex and number of observers should be standardized to limit the effects of this potentially confounding variable. We also acknowledge that less stringent safety criteria, especially for SBP, could have enabled patients to exercise at higher intensities. Nonetheless, as this was the first study attempting to exercise this high-risk patient population to high intensities, it was important that our safety criteria, and also the extent to which the patients were pushed during exercise, reflected a conservative approach. Given that pre-operative exercise therapy exerts beneficial effects on physical fitness and post-operative outcome measures (Pouwels et al., 2016), our data have clear clinical application by showing it is possible to progressively exercise this high-risk population at moderate to hard intensities. Finally, an in-depth assessment of fidelity in multi-center exercise interventions should examine the consistency of the exercise dose across the different sites, yet given the relatively low sample size for each of our three study sites we elected not to include between-site comparisons.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.