We found three systematic reviews
and three subsequent RCTs.
The first review (search date 2004, 17 RCTs, see comment) included RCTs of back exercises versus placebo, no treatment, or other conservative treatments.
The second review (search date 2003, 6 RCTs, 518 people) included RCTs of McKenzie treatment
versus passive treatment, advice to stay active, flexion exercises, spinal manipulation, back school
, or strengthening.
The third review (3 RCTs, number of people included not clear) included RCTs of McKenzie treatment versus the NSAID ketoprofen, massage/advice, or passive movement/mobilisation.
The methodological quality of RCTs identified by the first review was assessed by the adequacy of 4 criteria: randomisation, allocation concealment, follow-up, and outcome blinding.
Studies were classed as high quality if they met all 4 criteria. The review identified 11 RCTs in people with acute back pain and 6 RCTs in people with subacute back pain; one RCT in each group was categorised as being of high quality. Methodological quality in the second and third reviews was based on the PEDro scale.
The second review identified 5 RCTs in people with acute low back pain, one RCT in people with subacute low back pain, and three RCTs in a mixed population of acute/subacute low back pain; all but one of the identified RCTs were high quality (score of 5+/10).
In the third review, two of the three RCTs identified were high quality (5+/10).
The first and second reviews identified 6 RCTs, one of which was also identified by the third review (see comment). The second and third reviews identified three RCTs assessing the effects of McKenzie treatment (see comment).
All three reviews defined the included RCTs as either acute (<6 weeks' duration), subacute (6–12 weeks' duration), or duration not subgrouped (<12 weeks). The first review used both a qualitative rating system and a quantitative pooling of data where possible.
The second review pooled data (only statistically homogeneous RCTs) to compare the McKenzie treatment versus passive therapy (combined data on educational booklet, ice packs, massage, and bed rest) and advice to stay active (random effects model).
The third review transformed pain and disability scores to a score ranging from 0 to 100. To describe treatment effect for individual studies, mean and 95% confidence intervals were calculated for between-group differences (see comment).
The second review pooled data based on treatments, whereas the third review pooled data based on outcomes, and so, here, we report meta-analyses from only the second review.
Generic back exercise versus usual care or no treatment for acute low back pain (<6 weeks' duration):
The first review reported that 10 of 11 RCTs identified had non-exercise comparisons.
The review found no significant difference between generic exercise
and no treatment in change in pain or function measured at the earliest follow-up (scale 0–100; pain: 3 RCTs, 491 people; WMD –0.59, 95% CI –12.9 to +11.51; function: 3 RCTs, 491 people; WMD –2.82, 95% CI –15.35 to +9.71; see comment). One high-quality RCT in an occupational setting found that mobilising home exercises were less effective than usual care, and one low-quality RCT in a healthcare setting found that a therapist-delivered endurance programme improved short-term functioning more than no treatment. Of the remaining 8 RCTs, 6 studies identified by the review found no statistically significant or clinically important difference between exercise therapy and usual care/no treatment, and the results of two RCTs were unclear.
Generic back exercise versus usual care or no treatment for subacute low back pain (6–12 weeks' duration):
The first review reported that, in 6 included RCTs, 7 exercise groups (total number of exercise groups not reported) had a non-exercise comparison.
One high-quality and one low-quality RCT found that a graded exercise intervention reduced absenteeism outcomes in the workplace compared with usual care, and one low-quality RCT found improved functioning with exercise plus behavioural therapy compared with usual care. Two poor-quality RCTs found no difference in outcomes between exercise and the comparative treatments (including usual care), and one poor-quality RCT reported unclear results. One subsequent RCT (134 people with low back pain for at least 4 weeks before inclusion) compared graded exercise versus usual care.
The RCT found no significant difference in pain severity (11-point visual analogue scale [VAS]: 0 = no pain to 10 = very severe pain) or functional status (Roland Disability Questionnaire) between graded exercise and usual care, although there were greater improvements in both outcomes with graded exercise (between-group difference at 12 months: pain severity [favours graded exercise]: –0.2, 95% CI –1.2 to +0.8; P = 0.67: functional status [favours graded exercise]: –0.6, 95% CI –2.8 to +1.5; P = 0.56). The RCT found that people assigned to the graded-exercise group returned to work faster than those assigned to usual care (median duration of first continuous period of sick leave after randomisation: 54 days with graded activity v
67 days with usual care; significance not assessed). Graded exercise consisted of twice-weekly exercise sessions lasting 60 minutes each until the people either achieved full return to work, or the maximum therapy duration of 3 months had been completed.
Generic back exercise versus non-exercise interventions for acute low back pain (<6 weeks' duration):
The first review found no significant difference between exercise and other conservative treatments (advice to stay active, education, and usual care) in change in pain or function measured at the earliest follow-up (scale 0–100; pain: 7 RCTs, 606 people; WMD +0.31, 95% CI –0.10 to +0.72; function: 6 RCTs, 534 people; WMD –1.34, 95% CI –5.5 to +2.81).
Results were similar at intermediate and long-term follow-up.
Generic back exercise versus non-exercise interventions for subacute low back pain (6–12 weeks' duration):
The first review found no significant difference between exercise and all other comparisons (including no treatment, usual care, advice to stay active, and education) in change in pain or function measured at the earliest follow-up (scale 0–100; pain: 5 RCTs, 608 people; WMD –1.89, 95% CI –4.91 to +1.13; function: 4 RCTs, 579 people; WMD –1.07, 95% CI –5.32 to +3.18).
Results were similar at intermediate follow-up. The review concluded that there was insufficient evidence to support or refute the effectiveness of exercise for pain or function in subacute low back pain.
Generic back exercise plus CBT versus no exercise or CBT alone:
The second subsequent RCT (106 men with low back pain during the 3 months before study enrolment) compared neuromuscular training plus CBT
versus no exercise or CBT.
At 12 months, the RCT found that neuromuscular training plus CBT significantly decreased pain intensity (visual analogue scale [VAS]) for the 7 days before assessment compared with no treatment (80 people; change in VAS from baseline: from 9.9 to 5.5 with neuromuscular training plus CBT v
from 11.8 to 10.2 with no treatment; P = 0.032). There was no significant difference between groups in intensity of back pain for the 2 months before assessment, although a greater improvement in pain was reported by the group receiving neuromuscular training plus CBT (80 people; change in VAS from baseline: from 15.3 to 8.6 with neuromuscular training plus CBT v
from 15.8 to 14.3 with no treatment; P = 0.052). The RCT found no significant difference between treatments in disability (Oswestry Disability Index [ODI]) at 12 months (84 people; change in ODI from baseline: from 5.6 to 4.8 with neuromuscular training plus CBT v
from 5.8 to 5.0 with no treatment; P = 0.88). Neuromuscular training plus CBT consisted of neuromuscular training plus counselling with cognitive-behavioural goals for improved lumbar stability (2 sessions/week, one of which was physiotherapist-led and the other independent): the exercise programme consisted of 10 generic exercises.
Generic exercise versus CBT plus exercise:
The third subsequent RCT (47 people with subacute low back pain) compared a control group who only received physiotherapy in the form of generic exercises versus an intervention group who received the exercise treatment plus a CBT programme.
The CBT programme was conducted by a psychologist and took place independently of the physiotherapy treatment. For pain intensity (measured by 10-point VAS), there were no significant differences between groups after the last physiotherapy session (P = 0.81), 3 months later (P = 0.12), or 6 months later (P = 0.075). For overall severity (measured by the question "how severe do you think your back problems are?" and scored 1–4), there was no significant difference between groups after the last physiotherapy session (P = 0.22), or at 3 months (P = 0.98), although there was a significant difference between groups at 6 months (P = 0.004).
Specific back exercise (McKenzie treatment) versus usual care or no treatment:
The reviews identified no RCTs for this comparison.
Specific back exercise (McKenzie treatment) versus passive treatments (combined analysis of educational booklets, bed rest, ice packs, and massage):
The second review (4 RCTs, 681 people) found that McKenzie treatment significantly decreased pain and disability at 1 week compared with passive therapy (combined data on educational booklets, bed rest, ice packs, and massage) (2 RCTs, 470 people; pain: WMD –4.16, 95% CI –7.12 to –1.20; disability: WMD –5.22, 95% CI –8.28 to –2.16; absolute numbers and P values not reported).
However, there was no significant difference between groups in disability at 4 weeks (3 RCTs, 495 people; WMD –1.06, 95% CI –3.21 to +1.10; absolute numbers and P value not reported).
Specific back exercise (McKenzie treatment) versus advice to stay active:
The second review found a significant increase in disability after 12 weeks' treatment with the McKenzie treatment compared with advice to stay active (2 RCTs, 261 people; WMD [0–100 point scale] 3.85, 95% CI 0.30 to 7.39; absolute numbers not reported; P value not reported).
There was no significant difference between groups in pain intensity at 12 weeks (WMD +5.02, 95% CI –1.19 to +11.22; absolute numbers not reported).
Specific back exercise (McKenzie treatment) versus flexion exercises:
The second review did not pool data for this comparison because of clinical and statistical heterogeneity among studies.
The review identified two RCTs that met Clinical Evidence
inclusion criteria. One high-quality RCT (149 people with acute low back pain with or without radiation) identified by the review found no significant difference between treatment groups in pain at 8 weeks (data presented graphically; reported as not significant; P value not reported).
One low-quality RCT (24 people)
identified by the review
found a greater improvement in mean disability scores (ODI) at 5 days' follow-up with McKenzie treatment compared with flexion exercise (data presented graphically in the RCT; no further details reported: mean difference [0 to 100-point scale] between groups reported in the review: –22 points, 95% CI –26 points to –18 points).
Specific back exercise (McKenzie treatment) versus back school:
The second review identified one RCT (100 people with acute or subacute low back pain and with or without radiating pain) that met Clinical Evidence
The RCT found that McKenzie treatment decreased pain at 1 year compared with back school
(absolute numbers not reported; P <0.001).
A 5-year follow-up study of the RCT identified by the review found that McKenzie treatment significantly decreased the proportion of people on sick leave at 5 years compared with back school (24/47 [51%] with McKenzie treatment v
31/42 [74%] with back school; P <0.03).
Specific back exercise (McKenzie treatment) versus spinal manipulation:
The second review identified one high-quality RCT (24 people with acute or subacute low back pain
) that met Clinical Evidence
The RCT did not carry out a statistical analysis.
The review found a significant increase in disability (ODI) with McKenzie treatment at 5 days and 4 weeks compared with spinal manipulation (mean difference [0 to 100-point scale]; 5 days: 17 points, 95% CI 8 points to 27 points; 4 weeks; 22 points, 95% CI 10 points to 33 points).
Specific back exercise (McKenzie treatment) versus NSAIDs:
The third review (1 RCT, 260 people) found no significant difference in short-term disability between McKenzie treatment and the NSAID ketoprofen (follow-up at <3 months), although results favoured McKenzie treatment (mean AR –4.2, 95% CI –9.8 to +1.4; absolute numbers not reported).