Tweet
Dear readers,
I am facing a problem which I am sure many people planning clinical
studies with gait analysis will have had before:
In a study monitoring surgical outcomes in a prospective randomized
trial of two groups the number of patients to be recruited in a
reasonable time typically is very limited (between 10 and 50). We are
thinking about a study comparing two groups with about 25 patients in
each group monitoring the gait pattern pre and post surgery.
Ideally one would have the groups starting at the same level of walking
ability and with comparable walking patterns. A simple randomization
scheme for this small number however, might by accident lead to
significant differences between the groups pre-op.
In the literature I found a method called "minimization" where strata
are defined with respect to relevant influencing factors as eg. the
walking ability pre-op. The probability for allocating a patient to a
certain group then could be driven by the individual patients' walking
ability. In small sample numbers however this could effect in a hidden
pre-determination for group allocation for the majority of the samples
(patients). Talking to a biometrics I learned that in this case such a
procedure is not acceptable.
However, nobody could explain to me why such a procedure is not correct
(even in the case that all patients might be pre-determined for
allocation). As long as this procedure guarantees that neighter the
surgeon nor the staff taking the gait analysis data nor any person
involved in the recuitment process has the possibility to interfere or
influence the allocation scheme I would think that it is a reasonable
procedure for group allocation.
Further, taking that this method is acceptable, I do not know which
feature to choose for stratification regarding walking ability and/or
walking pattern. One very simple parameter would be body height
normalized walking speed. Another possibility would be the Gillette Gait
Index (Schutte et al. Gait Posture 2000). Any other suggestions?
Any comment is welcome.
Regards
Sebastian
_____________________________________________
Dr. Sebastian Wolf
Gaitlab Manager
Department of Orthopedic Surgery
University of Heidelberg
Schlierbacher Landstr. 200a
D-69118 Heidelberg - GERMANY
phone: +49 6221-96 6724
fax: +49 6221-96 6725
_____________________________________________
I am facing a problem which I am sure many people planning clinical
studies with gait analysis will have had before:
In a study monitoring surgical outcomes in a prospective randomized
trial of two groups the number of patients to be recruited in a
reasonable time typically is very limited (between 10 and 50). We are
thinking about a study comparing two groups with about 25 patients in
each group monitoring the gait pattern pre and post surgery.
Ideally one would have the groups starting at the same level of walking
ability and with comparable walking patterns. A simple randomization
scheme for this small number however, might by accident lead to
significant differences between the groups pre-op.
In the literature I found a method called "minimization" where strata
are defined with respect to relevant influencing factors as eg. the
walking ability pre-op. The probability for allocating a patient to a
certain group then could be driven by the individual patients' walking
ability. In small sample numbers however this could effect in a hidden
pre-determination for group allocation for the majority of the samples
(patients). Talking to a biometrics I learned that in this case such a
procedure is not acceptable.
However, nobody could explain to me why such a procedure is not correct
(even in the case that all patients might be pre-determined for
allocation). As long as this procedure guarantees that neighter the
surgeon nor the staff taking the gait analysis data nor any person
involved in the recuitment process has the possibility to interfere or
influence the allocation scheme I would think that it is a reasonable
procedure for group allocation.
Further, taking that this method is acceptable, I do not know which
feature to choose for stratification regarding walking ability and/or
walking pattern. One very simple parameter would be body height
normalized walking speed. Another possibility would be the Gillette Gait
Index (Schutte et al. Gait Posture 2000). Any other suggestions?
Any comment is welcome.
Regards
Sebastian
_____________________________________________
Dr. Sebastian Wolf
Gaitlab Manager
Department of Orthopedic Surgery
University of Heidelberg
Schlierbacher Landstr. 200a
D-69118 Heidelberg - GERMANY
phone: +49 6221-96 6724
fax: +49 6221-96 6725
_____________________________________________