Strengths of the study include (i) its design, allowing to
appreciate both long-term effects of the exercise intervention
and the sustainability of changes, (ii) the long-term
follow-up, with objective blinded assessments carried-out
over a total period of 4-year, and iii) the assessment of a
‘‘real-world’’ exercise program already available in the
community. A number of limitations need to be considered.
First, some caution must be retained given the limited
sample size. However, participants whose data are reported
herein did not differ significantly in any measures at original
trial completion from those who did not take part to
the 4-year follow-up assessment visit and the long-term
extension study sample was representative of the targeted
original trial population. Also, it might be argued that
participants who maintained a 4-year exercise program
were more likely to be healthier. While significant differences
in the number of frailty components and Tinetti score
were evident between groups at original trial entry, differences
did not remain at original trial completion. Furthermore,
because linear mixed-effects models were used,
with random effects fitted, the heterogeneity of baseline
scores was carefully taken into account. Second, part of
falls data was collected retrospectively without validation
against medical records. Retrospective falls history
obtained through participants’ recollection of falls in the
past year and 3 years may not have been as accurate as
collecting falls data through prospective methods (e.g.,
calendar method used during the original trial) and may
have involved under-reporting of falls. Third, because
long-term outcomes were secondary endpoints, no formal a