To explore whether using a single matched or composite outcome might impact the results
of previous randomised controlled trials (RCTs) testing exercise for non-specific
low back pain (NSLBP). The first objective was to explore whether a single matched
outcome generated a greater standardised mean differences (SMD) when compared to the
original unmatched primary outcome SMD. The second objective was to explore whether
a composite measure, comprised of matched outcomes, generated a greater SMD when compared
to the original primary outcome SMD.
We conducted exploratory secondary analyses of data.
Seven RCTs were included, of which two were based in the USA (University research
clinic, Veterans Affairs medical centre) and the UK (primary care clinics, nonmedical
centres). One each were based in Norway (clinics), Brazil (primary care), and Japan
The first analysis comprised 1) five RCTs (n=1,033) that used an unmatched primary
outcome but included (some) matched outcomes as secondary outcomes, and the second
analysis comprised 2) four RCTs (n=864) that included multiple matched outcomes by
developing composite outcomes.
Exercise compared to no exercise.
Main Outcome Measures
The composite consisted of standardised averaged matched outcomes. All analyses replicated
the RCTs’ primary outcome analyses.
Of five RCTs, three had greater SMDs with matched outcomes (pooled effect SMD 0.30
(95% CI 0.04, 0.56), p=0.02) compared to an unmatched primary outcome (pooled effect
SMD 0.19 (95% CI -0.03, 0.40) p=0.09). Of four composite outcome analyses, three RCTs
had greater SMDs in the composite outcome (pooled effect SMD 0.28 (95%CI 0.05, 0.51)
p=0.02) compared to the primary outcome (pooled effect SMD 0.24 (95%CI -0.04, 0.53)
These exploratory analyses suggest that using an outcome matched to exercise treatment
targets in NSLBP RCTs may produce greater SMDs than an unmatched primary outcome.
Composite outcomes could offer a meaningful way of investigating superiority of exercise
than single domain outcomes.