U.S. flag

An official website of the United States government

NCBI Bookshelf. A service of the National Library of Medicine, National Institutes of Health.

Little P, Stuart B, Stokes M, et al. Alexander technique and Supervised Physiotherapy Exercises in back paiN (ASPEN): a four-group randomised feasibility trial. Southampton (UK): NIHR Journals Library; 2014 Oct. (Efficacy and Mechanism Evaluation, No. 1.2.)

Cover of Alexander technique and Supervised Physiotherapy Exercises in back paiN (ASPEN): a four-group randomised feasibility trial

Alexander technique and Supervised Physiotherapy Exercises in back paiN (ASPEN): a four-group randomised feasibility trial.

Show details

Chapter 4Results

Feasibility of recruitment

Several methods of recruitment were piloted and all proved to be feasible.

Invited recruitment for patient attending within the last 5 years

The easiest method of recruitment was to write to patients who had attended with back pain in the previous 5 years and, if they still had pain and fulfilled eligibility requirements, to then invite them to a recruitment appointment with the practice nurse; this was the method of recruitment used in the ATEAM trial.12 A total of 1987 letters were sent from six practices. In total, 297 potentially eligible participants expressed an interest in participating and 170 replied that they were not interested; 68 were consented after receiving the invitation letter. This group included patients not currently seeking care for their back pain. However, in a significant number of cases this is likely to be because they felt that no further help was available. A typical comment from a patient to an Alexander technique teacher following the intervention was, ‘Why didn’t anyone tell me about this before?’ (also see qualitative report).

Opportunistic recruitment in surgery

Patients attending their GP with back pain were invited by the GP to participate during the consultation and were consented within the consultation. However, as might be expected, because of the time restrictions in an acute appointment, far fewer patients (n = 11) were recruited by this mechanism.

Invited recruitment based on attendance in the previous week

As already documented, the board was concerned that those who were not currently seeking care might be substantially different from those who are currently seeking care and so might not be a population relevant for policy-making in the NHS. It transpired that this might be an artificial distinction and a function of the lack of help perceived by patients (see comment above). Given the desire of the board for us to achieve meaningful recruitment at or near the time of presentation with an episode of care in the NHS, we also piloted a recruitment method based on a search of attendance with back pain in the previous week. This method of recruitment was piloted in two practices to demonstrate that the method was feasible. In one of the practices the yield from 6 weeks of brief searching was 57 invitations, with 10 patients replying that they were interested in participating and four patients consented, which suggests that the method is likely to be feasible and provide timely recruitment for any full trial, should a main trial be funded, based on recent presentation with back pain.

Feasibility of randomisation

There were no issues here except that there was some dropout of participants (n = 14) between consent and randomisation, as shown in the CONsolidated Standards Of Reporting Trials (CONSORT) diagram (Figure 1), mainly because of time pressures.

FIGURE 1. CONsolidated Standards Of Reporting Trials (CONSORT) diagram.


CONsolidated Standards Of Reporting Trials (CONSORT) diagram. AT, Alexander technique; DNA, did not attend; EC, exercise classes.

Even with the relatively small numbers, the groups were reasonably well balanced at baseline (Table 4).



Baseline characteristics

Feasibility of intervention lessons

Logistical issues

The Alexander technique lessons were implemented more easily than the exercise classes, as the Alexander technique teacher and the pupil arranged a mutually convenient time.

The exercise class intervention was more problematic logistically and some areas (e.g. Portsmouth practices) struggled to recruit and so the classes had few patients per group and were sometimes one-to-one rather than group sessions. There was a booking lag at one of the exercise class centres as patients could not get through on the telephone and so the study team took over the scheduling and the class administrators gave us slots to fill. There was a slight delay in starting classes while waiting for numbers to make a ‘group’. Patients found the limited flexibility in class times more difficult, but adherence was nevertheless reasonably good (> 50% attended ≥ 80% of classes; see the following section) compared with that seen in a previous large trial.9 Transport to and from sessions was problematic for those without a car and so we arranged taxi collection and delivery when necessary (and so travel expenses for patients were a little more than anticipated). The time commitment for the classes, and particularly for the combined intervention, was a large burden, which suggests that, if the combined intervention is to be used, the intervention should be largely serially and flexibly organised rather than sessions being organised in parallel.

Feasibility of delivering the intervention: compliance with attending intervention classes

The previous ATEAM trial12 defined good adherence as attending five out of six lessons or 20 out of 24 lessons. This is attendance of just over 80%. We have followed this example and defined good adherence as the attendance of at least eight of the 10 Alexander technique lessons and 10 of the 12 exercise class lessons.

Of those randomised to receive Alexander technique lessons (either Alexander technique lessons only or both Alexander technique lessons and exercise classes), we have data from the lesson booklets (which were returned by practitioners) for 31 out of 35 participants. Only three out of the 31 participants did not complete all 10 lessons. One stopped at lesson 3, one at lesson 8 and one at lesson 9. Thus, 93.5% of participants for whom we have booklets returned by the practitioners (29/31) achieved ‘good adherence’ to the Alexander technique lessons. If we assume that those who did not return the booklets did not achieve good adherence, which is a conservative assumption, the figure would be 82.9% (29/35).

Of those randomised to receive exercise classes (either exercise classes or both exercise classes and Alexander technique lessons), we have data from the lesson booklets for 25 out of 35 participants. Of these, 19 completed all 12 lessons and ‘good adherence’ was achieved by 84.0% (21/25). If we assume that those who did not return the booklet did not achieve good adherence this figure would be 60.0% (21/35).

Table 5 provides a summary of the data on adherence.




In terms of harms, one patient who had been a little dizzy before starting the exercises (but who had not warned the staff) fell backwards in the exercise class group, grazing the underside of her arm.

Feasibility of retention

We followed up 57 of the 69 participants (83%) at 3 months and 56 out of 69 (81%) at 6 months. Those withdrawing mostly had other medical problems or problems with attending (Table 6). There was no significant difference in RMDQ score at baseline between those who withdrew and those who did not (10.25 vs. 10.28; p = 0.9836).



Reasons for withdrawal

Feasibility of laboratory-based biomechanical and neuromuscular physiological measures

The board was concerned about the participant burden of the biomechanical measures. The feedback from patients was that this was one of the most interesting elements of the study, and follow-up rates for the measures were good despite the issues of time and transport. Pilot testing of all measures included in the research proposal revealed that measures of spine length and curvature using the Vicon 3D motion analysis system were complex and too time-consuming to warrant inclusion, as they were not key measures. Accurate alternative tools were not available.

All tests were well tolerated during the study, although a small number of participants (n = 14) reported mild sensations when standing in the trunk rotation testing device, including feeling nauseous, dizzy and unstable, but this did not prevent them being able to undergo testing. Surface EMG was used, which did not cause the discomfort that can occur with fine-wire EMG, but the drawback was poor-quality signals over the abdominal muscles, possibly because of subcutaneous fat. Data from the symptomatic side and during functional tasks (leg lifts for muscle thickness and EMG and changes in posture for Myoton measures) were more informative than data from the asymptomatic side and at rest. The duration of data collection sessions reduced as the study progressed (start approximately 120 minutes, end approximately 90–100 minutes) and proved to be acceptable to participants.

Sensitivity to change

Clinical outcome measures

Referees were keen to see a variety of potential major outcomes assessed as part of the feasibility trial (e.g. the Oswestry Disability Index and the Aberdeen pain and function scale). The standardised response mean has been calculated as the difference between the mean at follow-up and the mean at baseline divided by the SD of the difference. Most of the major questionnaire outcomes (RMDQ, Von Korff pain and disability scale, Oswestry Disability Index, Aberdeen pain and function scale) were equally sensitive to change (Table 7).



Sensitivity to change of clinical outcome measures

Change in laboratory-based measures

All laboratory-based biomechanical and neurophysiological tests showed changes at 6 months at the 5% level of significance in at least one of the intervention groups, apart from lumbar proprioception, which was significant only at the 10% level (Table 8). When changes were not significant they were nearly always in a beneficial direction, reducing the likelihood of type 1 errors.



Change in mean scores for the biomechanical and neuromuscular physiological measures from baseline to 6 months’ follow-up by group

Exploratory analyses of effectiveness

Main clinical results

All analyses are based on a linear regression model and control for baseline score on the scale being tested. The models were tested to assure adherence to the linear regression assumptions, that the residuals were normally distributed and that there was no significant heteroskedasticity. As expected given the very limited power, most outcomes did not reach significance at the 5% level and so the lack of significance should be interpreted very cautiously; some reached significance at the 10% level by 3 and 6 months, particularly in the combined group (Tables 9 and 10 respectively), and when changes were not significant they were nearly always in a beneficial direction, suggesting that type 1 errors are unlikely. The estimates suggest that clinically important improvements were probably occurring [e.g. the RMDQ documented clinically important changes (> 2.59) by 6 months]; of the other two major questionnaire outcomes assessed (Aberdeen pain and function scale and Oswestry Disability Index), the Aberdeen instrument perhaps differentiated between groups a little better and performed close to significance at both 3 and 6 months for the combined group.



Clinical outcome data at 3 months



Clinical outcome data at 6 months

The concern that the NICE guidance8 would result in an unstable control group was not borne out. The control group improved only a little (change in RMDQ score from 9 to 8 at 6 months), which was entirely consistent with the level of change observed in the control group (pre NICE guidance) in our previous trial.12

‘Dose response’: rating of pain and function by participants at Alexander technique lessons and exercise classes

The justification for assuming that we might be able to observe a ‘dosing’ effect (i.e. that the benefit apparent after lessons is maintained) was that after six lessons in the previous Alexander technique trial12 the benefit remained stable over at least 9 months, that is, the impact of the intervention was achieved quickly and was relatively stable. The weekly changes in RMDQ score are shown in Table 11 for participants receiving Alexander technique lessons.



Weekly mean (SD) RMDQ scores for participants receiving Alexander technique lessons

Accepting that the time in between lessons was not exactly the same for all participants, then treating this as panel data shows a significant effect of an increased number of lessons over time. For all Alexander technique participants, for each additional lesson, the RMDQ score decreases by a mean of 0.48 points (p < 0.001). This relationship is very similar regardless of the randomisation group (0.47 for the Alexander technique-only group and 0.48 for the Alexander technique + exercise class group; p < 0.0001 for both). There seems to be no levelling off of effectiveness after six lessons (the briefer intervention in the ATEAM trial12) and up to 10 lessons at least is likely to provide consistent improvement.

For exercise classes, for all exercise class participants, the RMDQ score decreases significantly over time, by a mean of 0.26 points for each additional lesson (p < 0.001) (Table 12). Here, the effect is significant in both randomisation groups but is slightly higher in the exercise class-only group (a mean decrease of 0.31 points compared with a mean decrease of 0.21 points in the Alexander technique + exercise class group). Once again, there seems to be no levelling off of effect, suggesting that 12 weeks of classes is likely to be beneficial.



Weekly mean (SD) RMDQ scores for participants receiving exercise classes

We also assessed the ‘dose–response’ relationship for days in pain recorded in the Alexander technique diaries (Table 13) and the exercise class diaries (Table 14) by those who completed them. Overall, for Alexander technique lessons, for all Alexander technique participants, there was a decrease in the mean number of days in pain of 0.15 per lesson [95% confidence interval (CI) 0.09 to 0.21; p < 0.0001]. The corresponding figure was 0.14 in the Alexander technique-only group (95% CI 0.05 to 0.22; p = 0.002) and 0.15 in the Alexander technique plus exercise class group (95% CI 0.07 to 0.24; p < 0.0001). Similarly, for all exercise class participants, each additional exercise class resulted in a reduction in mean number of days in pain of 0.10 (95% CI 0.05 to 0.15; p < 0.0001). The corresponding figure was 0.09 in the exercise class-only group (95% CI 0.02 to 0.17; p = 0.018) and 0.11 in the exercise class pus Alexander technique group (95% CI 0.06 to 0.16; p < 0.0001).



Weekly rating of mean (SD) days in pain during the previous week for participants receiving Alexander technique lessons



Weekly rating of mean (SD) days in pain during the previous week for participants receiving exercise classes

Exploratory analyses of intermediate markers (laboratory-based tests)

As with the clinical outcomes the analysis of intermediate measures is underpowered and so it is not very useful to comment on negative findings. Similarly, although we will briefly comment here on results significant at the 10% and 5% levels, extreme caution in the interpretation is needed given the danger of type 1 errors because of the number of intermediate markers present. The results should be viewed conservatively in the context of the hypothesised changes (Table 15) and the CHAIN model (see Figure 2 and Tables 19 and 20). The danger of type 1 errors is reduced by the fact that most biomarker and clinical changes, even if not significant, changed in a beneficial direction.



Hypothesised effects of the interventions on biomechanical and neuromuscular function: influences of Alexander technique lessons and physiotherapy exercise classes at 3 and 6 months post intervention

FIGURE 2. Graphical CHAIN model illustrating biomechanical and physiological variables associated with outcome 6 months after the intervention.


Graphical CHAIN model illustrating biomechanical and physiological variables associated with outcome 6 months after the intervention. AT, Alexander technique; TrA, transversus abdominus; US, ultrasound.

Differences between randomised groups at 3 months

Significant differences between randomised groups at 3 months were observed for lumbar proprioception and trunk extension strength (see Appendix 2, Table 16, for baseline measures and Appendix 2, Table 17, for 3-month measures).

For lumbar proprioception at 20° flexion, the Alexander technique and the Alexander technique plus exercise class groups had significantly lower errors (between the target position and the achieved position) than the control group (p = 0.026 and p = 0.035 respectively) after controlling for their baseline scores. The change in level of error for the exercise class group was in the same direction as for the other two groups and was significant at the 10% level (p = 0.08).

Trunk extension strength (measured using the Biodex system) in the exercise class-only group was significantly stronger (by 36 N/m; p = 0.027) than in the control group at 3 months. The changes in the other two groups were in the same direction.

The ASLR compression test was the only binary measure and so the odds of scoring a 1 on this test were compared by group. At 3 months the odds ratio was 1.03 (95% CI 0.18 to 5.82; p = 0.973) in the exercise class-only group, 2.10 (95% CI 0.24 to 17.78; p = 0.496) in the Alexander technique-only group and 0.59 (95% CI 0.10 to 3.40; p = 0.554) in the combined Alexander technique plus exercise class group.

Differences between randomised groups at 6 months (see Appendix 2, Table 18)

There were no longer any significant differences between randomised groups at 6 months for lumbar proprioception or trunk extension strength. There were additional differences that had not been seen at 3 months for multifidus muscle thickness at rest and contracted, the Myoton measurements of multifidus muscle mechanical properties, EMG onset of the lumbar paraspinal muscles and the ASLR test. The axial rotation testing device failed to function for a period (a replacement part had to be obtained from abroad) and so too few participants completed this test to enable reliable results to be obtained (n = 16 at 6 months, n = 9 at 3 months). The EMG recordings over the abdominal muscles were also too few to analyse because of difficulties in obtaining acceptable recordings.

Lumbar multifidus muscle thickness measured using USI was significantly greater in the exercise class-only group than in the control group in the relaxed and contracted muscle. These two variables were also significantly greater at the 10% level for the combined Alexander technique plus exercise class group. These observations were as expected for the exercise intervention, which included activities to improve the strength (and thus size) of the multifidus muscle as well as automatic activity of the muscle during functional tasks.

Multifidus muscle tone and mechanical properties were measured using the MyotonPRO device. There was lower elasticity (higher decrement values) in the treatment groups than in the control group, which reached significance below the 5% level only in the Alexander technique plus exercise class group in the prone lying position. The treatment groups had lower values than the control group for tone (frequency) and stiffness (N/m) and the difference reached significance at the 10% level for frequency in the prone position (p = 0.076). The mechanical properties of muscles measured by the MyotonPRO device differ according to the physiological composition and habitual activity of the muscle.41 Few muscles have been characterised at this relatively early stage in the use of Myoton technology to allow definitive interpretation of these findings for the multifidus muscle. A fully powered study would help elucidate the relationships between Myoton parameters and other measures of muscle characteristics and behaviours. It is known, however, that Myoton technology can detect subtle changes in muscle stiffness42 and that the changes in parameters during muscle contraction include increases in stiffness and tone (frequency) and reduction in elasticity (increase in decrement). The present observations suggest that the combined Alexander technique plus exercise class group had less muscle tension than the control group.

Compared with the control group, earlier onset of EMG activity of paraspinal muscles occurred on the symptomatic side during an ipsilateral leg lift in the Alexander technique-only group for the multifidus muscle, indicating better motor control.

During the ASLR compression test, all three intervention groups showed an odds ratio of < 1 and this was highly statistically significant in the Alexander technique group (p = 0.007), indicating a better ability to transfer load through the pelvis.

Graphical CHAIN model

The graphical CHAIN model (Figure 2 and see Appendix 2, Table 19) demonstrates that the only significant predictor at baseline of RMDQ score at 6 months was number of days in pain. Three laboratory measurement variables at 6 months were associated with RMDQ score at 6 months. These were lumbar proprioception at 20° flexion (beta coefficient 1.48°, 95% CI 0.83° to 2.12°; p < 0.001) and the lumbar multifidus mechanical properties measured using the MyotonPRO of tone (frequency) (0.94 Hz, 95% CI 0.48 Hz to 1.40 Hz; p < 0.0001) and difference in elasticity (log decrement) from lying to sitting, that is, altered activity with change in posture (−4.86, 95% CI −0.01 to −9.72; p < 0.05). Even with the small number of participants, the Alexander technique was shown to influence the mechanical property of elasticity at 6 months, suggesting that this may be a key element of the effectiveness of the Alexander technique. Likewise, exercise classes were shown to influence change in muscle tone from lying to sitting, suggesting that improvement in postural tone is a key element in the effectiveness of physiotherapy.

The baseline variables that were predictive of 6-month variables were laboratory measures, mainly proprioception, and multifidus muscle mechanical properties (using the MyotonPRO) and thickness (from USI). Of these baseline variables, multifidus muscle elasticity (difference in decrement values from prone to sitting measured using the MyotonPRO device) predicted four variables at 6 months: itself, proprioception in upright and flexion postures and multifidus muscle tone (change in Myoton frequency with posture). Multifidus thickness (when contracted during a straight leg lift, measured using USI) predicted itself, the thickness of the transversus abdominus muscle when contracted (which was reciprocally predictive of the thickness of the multifidus muscle when contracted) and proprioception in flexion. Five baseline variables were predictive of two variables each at 6 months. Going from left to right in the top row of Figure 2, the upright proprioception test predicted the Myoton measures of postural changes in tone (frequency) and stiffness. Contracted transversus abdominis muscle thickness predicted itself and contracted multifidus muscle thickness. Change in multifidus tone (frequency) with posture predicted itself and recruitment of the multifidus muscle (EMG onset). Multifidus tone at rest predicted itself and proprioception in flexion.

At 6 months, multifidus tone difference (i.e. change in frequency between lying and sitting) was associated with itself at baseline, body mass index, lumbar proprioception in the upright and 20° flexion positions, difference in multifidus tone (frequency) from prone to sitting and the exercise class-only intervention. The only other variable with more than two associations was proprioception in flexion, which was associated with multifidus thickness (contracted), multifidus tone at rest (Myoton frequency) and multifidus elasticity with change in posture (difference in decrement).

Understanding key issues in the trial: qualitative substudy

The qualitative substudy is reported in Appendix 3. This substudy documented the need for the provision of more information about the Alexander technique prior to consent (given its limited penetration in care to date), the low expectations of care for chronic back pain in the NHS and the low expectations of tailored physiotherapy classes (participants perceived that they were getting something that they had already tried and were pleasantly surprised by the group sessions). Time commitments for the interventions were a problem for some patients, but mostly there were very positive comments about both Alexander technique lessons and exercise classes and surprisingly positive comments about the biomechanical measurements.

Feedback from Alexander technique teachers

Carolyn Nicholls asked for feedback from all of the teachers involved in the study and their comments are shown in detail in Appendix 4. Most teachers reported observing significant benefits from the Alexander technique but commented on difficulties that participants had in arranging appointments, especially those receiving both Alexander technique lessons and exercise classes. For those who were working and looking after a family, keeping the appointments was doable but challenging. Retired people fared better in this respect. Several teachers commented that participants who received both Alexander technique lessons and exercise classes found the exercise classes easier, and this had also been noted by the physiotherapists.

Copyright © Queen’s Printer and Controller of HMSO 2014. This work was produced by Little et al. under the terms of a commissioning contract issued by the Secretary of State for Health. This issue may be freely reproduced for the purposes of private research and study and extracts (or indeed, the full report) may be included in professional journals provided that suitable acknowledgement is made and the reproduction is not associated with any form of advertising. Applications for commercial reproduction should be addressed to: NIHR Journals Library, National Institute for Health Research, Evaluation, Trials and Studies Coordinating Centre, Alpha House, University of Southampton Science Park, Southampton SO16 7NS, UK.

Included under terms of UK Non-commercial Government License.

Bookshelf ID: NBK259310


  • PubReader
  • Print View
  • Cite this Page
  • PDF version of this title (21M)

Other titles in this collection

Recent Activity

Your browsing activity is empty.

Activity recording is turned off.

Turn recording back on

See more...