Evaluation (Validation and Verification) - Medical Device Design: Innovation from Concept to Market - page 247

Table 9.15: Results of a t-test Assuming Unequal Variances

Set 1 Set 2

Mean 9.10 9.08

Variance 0.09 57.22

Observations 20.00 20.00

Hypothesized mean difference 0.00

df 19.00

t stat 0.01

p (T <= t) one-tailed 0.49

t critical one-tailed 1.73

p (T <= t) two-tailed 0.99

t critical two-tailed 2.09

Table 9.16 illustrates that the value of p is 0.01 for one-tailed. The groups are significantly
different for a one-tailed test. The value of p is always double for a two-tailed test hence
p =0.02. This is still <<0.05, hence the groups are significantly different. Why give both? It is
down to you to select which you use. If you are absolutely sure your average can only move
in one direction, either higher or lower but not both, then a one-tailed test works. If you are
not sure which way it can move then use the two-tailed test. In most circumstances we are
looking for improvements, hence it would be a one-tailed test. If you have more than one
group you can do the same analysis for each group in turn.

9.6.6.5 Multivariant Analysis
In many clinical studies there is more than one variable that changes between subjects; this
can be body weight, hair color, date of birth, phase of the moon when injured, etc. If you are
to take these into account you need to perform a multivariant analysis. You can do this by
redefining your groups into these subsets. But the most efficient way is to find a statistician to
perform the analysis for you.

9.7 Literature Review
Since the recent changes in the Medical Devices Directive, the FDA and EC rules about
performing a review of literature have become much more similar. The form of presentation
may be different, but they are essentially the same:

l

Substantial equivalence : if you are to minimize the amount of testing described earlier in
this chapter then a proof of substantial equivalence is of paramount importance. In this
part of the review you need to demonstrate that your device has, in some form, been in
existence before. You then need to show how your device is equivalent.

l

Known issues : this is predominantly a review of recalls and notifications to either the
FDA or to any of the EC regulatory bodies. The aim here is to identify any recurring,