A Model for Incorporating A Clinically-Feasible Exercise Test in Paraplegic Annual Reviews: A Tool for Stratified Cardiopulmonary Stress Performance Classification and Monitoring

Objectives: (i) To identify and characterize an exercise test for use in routine spinal cord injury clinical review, and (ii) to describe levels of, and factors affecting, cardiopulmonary stress performance during exercise in the chronic paraplegic population in Scotland, UK. 
 
Design: Cross-sectional study. 
 
Setting: Queen Elizabeth National Spinal Injuries Unit (Glasgow, Scotland). 
 
Patients: 48 subjects with chronic paraplegia resulting from spinal cord injury at neurological levels T2-L2. 
 
Methods: Peak oxygen uptake, peak power output, gas exchange threshold and peak heart rate were determined from an incremental arm-cranking exercise test. Using a general linear model, the effects of gender, high (injury level above T6) versus low paraplegia, time since injury, body mass and age on peak oxygen uptake and peak power output were investigated. 
 
Results: All 48 subjects completed the arm-cranking exercise test, which was shown to be practical for fitness screening in paraplegia. Men (n=38) had a peak oxygen uptake of 1.302 ± 0.326 l.min-1 (mean ± SD) and peak power output of 81.6 ± 23.2W, which was significantly higher than for women (n=10), at 0.832 ± 0.277 l.min-1 and 50.1 ± 27.8 W, respectively. There was large intersubject variability in cardiopulmonary performance during arm-cranking exercise testing, but the overall mean for the Scottish population was lower than reference values from other countries. 
 
Conclusions: Arm-cranking exercise tests are feasible in the clinical environment. The motivation for their implementation is threefold: (i) to determine cardiopulmonary stress performance of individual paraplegic patients, (ii)


Introduction
With an increase in life expectancy, patients with spinal cord injury (SCI) are exposed to a higher incidence of metabolic disorders (obesity, hypertension, diabetes and dyslipidaemias) [1] and a higher risk of developing cardiovascular disease than the general population [2]. hese long-term health problems have been attributed to the resultant sedentary lifestyle and lack of cardiopulmonary stressors following SCI [3,4].
In the able-bodied population, there has been a considerable drive to increase participation in regular physical activity aimed at improving cardiopulmonary itness. Regular physical activity using appropriate exercise modalities should also be encouraged in the SCI population [5]. In order to monitor the success or otherwise of exercise interventions, a practical tool for cardiopulmonary assessment is required.
he proactive approach of implementing annual reviews as part of long-term SCI management has already been shown to be beneicial [6]. Even so, such regular medical check-ups in chronic SCI patients have not been implemented internationally. In Scotland (U.K.), routine reviews of SCI patients currently consist of assessment mostly of bladder, bowel and sexual function, skin problems and spasticity. It is proposed that the annual reviews should also include some assessment of cardiopulmonary stress performance, as there are still too many deaths from cardiovascular disease in this population.
In this study we set out to determine the practicality of adding the incremental exercise test to routine SCI clinical review in paraplegia and to describe the typical level of cardiopulmonary stress performance in a cross-section of the national paraplegic population of Scotland. his would enable stratiication of patients into cardiovascular risk categories, and provide reference data against which the efects of targeted community and outpatient interventions could later be measured. Previous studies of cardiopulmonary itness in the SCI population were not available in the UK. In particular, such data were not available from Scotland, which continues to have the highest prevalence of ischemic heart disease in Western Europe [7][8][9][10]. he Scottish SCI population provides an ideal model to demonstrate the potential beneits of incorporating exercise tests in annual reviews precisely because there is still such a high incidence of deaths from cardiovascular disease in this group.

Subjects
Paraplegics attending the national spinal injuries unit for Scotland in a university teaching hospital (Queen Elizabeth National Spinal Injuries Unit, Southern General Hospital, Glasgow) for annual review were invited to take part in the study. Inclusion criteria were: (i) SCI at neurological level T2 -L2, (ii) grade A -C on the American Spinal Injuries Association (ASIA) Impairment Scale [11], (iii) age 18-65 years. Candidates were excluded if they had diagnosed cardiovascular disease. A physician carried out medical screening of potential candidates for inclusion in the study, ensuring that individuals with history of blood pressure, angina, cardiac hypertrophy or any other cardiac conditions were not included in the study cohort. Each subject on the study was also asked to complete a Health Questionnaire, detailing any medical conditions, medication, and other relevant information. hese data conirmed that participants on the study had no known cardiac conditions. he 48 subjects' details are summarized in Table 1. Ten women and 38 men took part. SCI level was categorized in relation to sympathetic innervations of the heart (T1-T5): SCI above T6 was classiied as HIGH paraplegia (n=10, all male), and SCI at T6 and below as LOW paraplegia (n=38).
he study was approved by the South Glasgow Research Ethics Committee. Written informed consent was obtained from each subject prior to participation.

Protocol
Participants attended an exercise testing session (TEST 1) at the Queen Elizabeth National Spinal Injuries Unit. A subgroup of participants (n=11) repeated the test (TEST 2) on a diferent day, between 24 hours and two weeks later, to determine test-retest reliability.
Subjects were asked to refrain from eating, smoking, and drinking tea or cofee for at least two hours prior to the test, and from drinking alcohol for 24 hours prior to the test. hey were required to empty their bladder (or urine bag) before starting the test. he subject's body mass was measured to the nearest kilogram using wheelchair scales (DP 2400, Marsden).
An instrumented arm-crank ergometer (Viva, ReckMotomed, Germany) was used for the test, connected via RS232 cable to a PC interface running a customized program for control and recording of work rate (Matlab v.7.3, Mathworks).
he subject was set-up at the arm-crank ergometer with the center of the cranks at shoulder-joint height and a distance between the wheelchair and the ergometer that ensured the elbow remained slightly lexed at the furthest position of the cranks. Detailed test instructions were given to the subject. A low-dead-space Hans-Rudolph face mask was placed over the nose and mouth and tightened to ensure a good seal. A volume turbine and sample line were attached to the mask and connected to a portable breath-by-breath cardiopulmonary measurement system (Metamax 3B, Cortex, Germany) to record volumes of air, and inspired and expired concentrations of oxygen and carbon dioxide for each breath. Prior to each test, the volume transducer was calibrated using a 3-litre syringe, and the gas analyzers were calibrated using two diferent reference gases of known concentration. A pulse oximeter (3800, DatexOhmeda) was connected to the subject via an earlobe sensor to monitor oxygen saturation of the arterial blood (SpO2) and heart rate (HR). HR was recorded once per minute.
Following a period of unrecorded rest to allow the subject to settle before formal testing, resting cardiopulmonary data were collected for at least two minutes (until consecutive values for oxygen uptake ( 2 O V  ) over one minute remained within 5% of each other, and similarly for end-tidal CO 2 ) before starting the exercise.
Visual feedback of actual and target work rate was provided in real time on the PC screen so that, as the target work rate changed, the subject was required to increase his/her power output (PO) accordingly, as in previous studies using a similar set-up [12,13]. Exercise began with three minutes of unloaded arm-cranking at a cadence of 50 rpm, ater which the target work rate was increased every minute in equal steps.
Step size was pre-programed and varied between subjects from 4-10 W (taking into account self-reported physical activity levels, age and gender), with the aim of completing the incremental phase within the optimal 8-12 minutes [14]. he subjects were verbally encouraged to continue until volitional exhaustion. A 2-3 minute warm-down was included if tolerated by the subject. Cardiopulmonary monitoring continued for a further 4-5 minutes until heart rate stabilized.

Data analysis
Prior to analysis, the raw cardiopulmonary data were systematically edited to remove outliers, using a custom Graphical User Interface (Matlab v.7.3, Mathworks) [15][16][17][18]. Line-itting and estimation of cardiopulmonary outcome parameters were carried out using the edited data. he outcome measures from the exercise tests are given in Table 2.

Statistical analysis
Statistical analyses were carried out using SPSS sotware (v15.0). he signiicance level was set at 5%.

Relationships between absolute
2 O V  peak and other outcome measures (PO peak , HR peak , and gas exchange threshold (GET)) were assessed using Pearson's correlation coeicient.
A linear regression was itted to absolute A two-tailed paired-sample t-test (n=11) was carried out to determine whether each outcome measure for TEST 1 was diferent to that in TEST 2. he intraclass correlation coeicient (ICC) was also calculated for each outcome measure [19] to determine test-retest reliability.

Results
All 48 subjects completed the exercise test and 11 of these subjects repeated the test on a diferent day. Body mass measurements were available for 41 patients. For TEST 1, the incremental phase of the tests was achieved within the target 8-12 minutes in 26  V  peak was strongly related to PO peak , (R 2 =0.863, p<0.001, n=46) and with GET (R 2 =0.676, p<0.001, n=36). For 12 out of 48 subjects, the GET could not be identiied ( Figure 2).

Outcome variable Abbreviation (Unit) Calculation
Peak power output PO peak (W) Highest power output increment completed Absolute peak oxygen uptake Absolute Highest 30s-average oxygen uptake reached, following guidelines from Wasserman et al. [16].
Relative peak oxygen uptake Relative Peak oxygen uptake normalized to body mass.
Gas exchange threshold GET (l.min -1 ) [15,17,18] Peak heart rate HR peak (beats.min -1 ) Highest HR recorded during the incremental phase of the test

Test-retest reliability
For the subgroup of subjects who repeated the procedure on a diferent day (n=11), a 2-tailed paired-sample t-test revealed that only PO peak for TEST 2 was signiicantly diferent (higher) to that in TEST 1. he paired t-test results and ICCs are shown in table 3. For 2 out of 11 subjects, a GET could not be identiied in either TEST 1 or TEST 2. he highest ICC was 0.968, for PO peak , and the lowest ICC was 0.626 for GET.

Discussion
We investigated whether it was feasible to add an exercise test to the standard set of health checks that are carried out in SCI annual reviews. An arm-crank ergometer was used, as it is a modality that is accessible in most spinal and other rehabilitation unit gyms. Other groups have suggested a range of tests and indicators that would need to be recorded to compile a complete assessment of physical capacity in SCI [20,21]. However, the use of a single test was evaluated here as it may be more easily adopted as a tool for itness screening in a clinical setting.
Measurable and reliable indicators of cardiopulmonary stress performance are ultimately of beneit to the patient, clinician and trainer only if reference values are available for comparison and can be used to set achievable training targets. To attain this goal, the proposed test and its outcome measures should identify variation in cardiopulmonary stress performance between patients. his would enable the stratiication of patients into cardiovascular disease risk categories, to highlight (or "red-lag") individuals in greatest need of exercise prescription. Based on this stratiication, targeted exercise intervention aimed at the highest-risk individuals would be implemented; repeat exercise tests would then be appropriate for monitoring changes in cardiopulmonary stress performance over time to determine the efectiveness of the exercise intervention.
We assessed cardiopulmonary stress performance using breathby-breath gas exchange measurements during an incremental armcranking exercise test, and successfully incorporated this test into 48 routine reviews of paraplegic patients, adding 30 minutes to the average consultation. Both the 2 O V  peak reached during the exercise test and the GET are key measures of cardiopulmonary stress performance. We were able to determine the 2 O V  peak in all tests where cardiopulmonary data were collected, but unable to detect a GET for 12 out of 48 subjects.
he results from our study can be compared to cross-sectional data from untrained paraplegic populations investigated by other groups worldwide. In the US, arm-crank ergometry exercise tests were used to assess eight American paraplegics (men and women) [22]. he reported mean Compared to published reference values for paraplegia, it appears that this study population is generally less it, falling within the lower end of the range of the pooled international data. A program of exercise prescription monitored for its efectiveness through annual reviews may be warranted in Scotland.   he long-term clinical implication is that routine cardiopulmonary stress performance assessment could become part of the development of a targeted health maintenance program in the SCI population. he linear spread of oxygen uptake data enables us to achieve itness screening and patient stratiication by identifying the most "at risk" individuals. In much the same way as Janssen et al. [23] suggest, the categorization of normative values into ive quintiles of physical capacity (Poor, Fair, Average, Good and Excellent), a similar deinition could be used for itness screening and patient stratiication here. hese "high risk" patients (in the "Poor" and "Fair" quintiles) could be primary targets for exercise prescription, and exercise regimes could be designed with the aim of gradually increasing cardiopulmonary stress performance and progressing into lower-risk categories. he success of the exercise interventions would be monitored through regular exercise testing during subsequent annual reviews.

Intra-Class Correlation
Apart from identifying individual patients with low cardiopulmonary stress performance and potentially at high risk of developing cardiovascular diseases, it would also be useful to determine whether any subgroups of the population tend to have lower values and whether they should be targeted as a group for exercise prescription. Subgroup analyses of our data revealed that women tended to have lower cardiopulmonary stress performance levels than men. he efect of gender was signiicant for key outcome measures. Some women in our sample may not have given a maximal efort during the exercise tests. he inding that a GET could not be identiied in 5 out of 10 female subjects lends some support to this speculation. Alternatively, women in the study may have had weaker upper limb muscles than the men, leading to peripheral rather than central limitation during arm cranking. However, it should be noted that the statistical signiicance of the gender efect on 2 O V  peak was lost when diferences in body mass were taken into account to give relative . he small number of female participants is a clear limitation of our study, and the proposed diferences in performance between men and women with SCI in these exercise tests would need to be conirmed with larger numbers.
here are a number of additional limitations to our study. We developed a reference database for the Scottish paraplegic population which is currently relevant for SCI at T2-L2 only, but around half of SCIs occur at higher neurological levels and lead to tetraplegia. In the healthy neurologically-intact population, there is a linear relationship between heart rate and oxygen uptake during incremental exercise tests. However, a "blunted heart rate response" that results from reduced sympathetic activity in SCI can afect the coupling normally seen between the two parameters [24][25][26]. When lesion level in relation to the sympathetic innervation of the heart (T1-T5) was investigated as a potential predictor of 2 O V  peak , we found no signiicant diference between the HIGH and LOW paraplegia groups. his may be explained by other groups' indings that, even in cases where extensive sympathetic nervous system dysfunction is expected because of the high level of SCI, there may be some preserved sympathetic pathways [27]. his could be due to a discrepancy between clinical and neurophysiological deinitions of a "complete" SCI [28,29]. Alternatively, the numbers of patients with HIGH paraplegia included in the study may have been insuicient to demonstrate the efect of lesion level on cardiopulmonary stress performance that has been characterized in previous evaluations of physical capacity in SCI [20]. Furthermore, tetraplegics have been shown to have lower levels of cardiopulmonary stress performance than paraplegics [22], but tetraplegic subjects were not included in our study. In future work, cardiopulmonary stress performance levels should be investigated in a cross-section of the Scottish tetraplegic population.
As the method of cardiopulmonary stress performance assessment used here may not be applicable to tetraplegia, which includes paralysis of muscles of the upper limbs as well as the lower limbs, the test setup and protocol would need to be modiied to accommodate these factors. More speciically, due to the level of innervation of key muscles involved in arm-cranking exercise, we would not be able to implement arm-cranking for cardiopulmonary stress performance in patients with SCI above C5, without the addition of electrical stimulation to recruit key muscle groups [12,13]. In the absence of electrical stimulation of elbow lexors and extensors, alternative cardiopulmonary stressors would need to be investigated and evaluated in SCI at higher cervical levels.
In addition to feasibility, we investigated the repeatability of the proposed exercise tests. In our study, where subjects repeated the test on a diferent day (n=11), there was a higher PO peak on the second test (p=0.005), but at a similar Possible explanations include that the power measurements are not very repeatable or that subjects had developed a more eicient technique for arm-cranking in the second test (a "practice efect"), coupled with reduced anxiety due to familiarity of the test when repeated. his test-retest analysis suggests that, if exercise testing were to be incorporated into annual reviews, a familiarization test may be warranted to determine a representative PO peak , but not necessary to determine It may be argued that a diferent solution is to carry out the exercise tests using wheelchair ergometry (as described by Janssen et al. [23]). his exercise modality involves upper body movements that individuals with paraplegia (and low tetraplegia) use on a daily basis. he issue of technique and optimum positioning in terms of distances and angles between the wheelchair and the ergometer may not afect the wheelchair ergometry test to the same extent as the armcrank ergometry test, which may negate the need for a familiarization test when using this alternative modality. However, as the set-up for wheelchair ergometry testing requires a considerable amount of space and specialized equipment, it may not be a practical alternative to armcrank ergometry testing in many spinal and other rehabilitation units.
A possible implementation strategy of the tests described here might be that a full baseline cardiopulmonary stress performance assessment of arm-cranking exercise is made for each patient initially. Following this, an incremental power test without full breath-bybreath cardiopulmonary measurement could be suicient for routine review because, as shown here and well known from the wider exercise physiology literature, V  peak from the PO peak reached by the patient during the incremental power test. Peak power tests would be more manageable as a routine review assessment than the full cardiopulmonary assessment test, as the time (15-20mins) and expertise required by the clinician running the test would be reduced. hese peak power tests could be performed by physiotherapists, instead of exercise or sports scientists, who are not typically part of the clinical rehabilitation team (in the UK at least).
he identiication of a test (or choice of tests) that could be incorporated in routine clinical practice in a spinal injuries unit and accommodate diferent levels of injury may require a compromise between data quality and feasibility in a clinical setting. In terms of data quality, subjects would ideally have a suiciently long resting period and a minimum of 4 hours without cafeine, food, or smoking prior to the test, and no distractions during the test. Achieving an ideal testing environment in a clinical setting may not always be possible due to mitigating circumstances, but this should not preclude the exercise test from being considered as a tool for health assessment as part of the SCI annual review.
In conclusion, the 30-minute arm-cranking exercise test proposed here proved feasible in a clinical setting. Implementing the exercise test enabled us to calculate 2 O V  peak , PO peak and HR peak (and in 75% of cases, GET), and these outcome measures were used to describe cardiopulmonary stress performance levels for a cross-section of the untrained chronic paraplegic population. Compared with published reference values Scottish paraplegics, with a mean 2 O V  peak of 1.21 l.min -1 , have lower cardiopulmonary stress performance levels than typically seen for this patient group. Furthermore, the clinical management of cardiovascular disease risk in this patient group could beneit from stratiication of patients based on their cardiopulmonary stress performance, to inform targeted exercise prescription for those patients who would beneit the most from improvements in itness.

Clinical Messages
• he incorporation of exercise tests in paraplegic annual reviews would enable the clinical team to:(i) identify most unit paraplegic patients, through stratiication into "cardiovascular risk" categories, and (ii) monitor the efects of targeted exercise prescription.
• Systematic monitoring of cardiopulmonary stress performance can be achieved feasibly by adding a 30-minute arm-cranking exercise test to the annual reviews of paraplegic patients.