400
A study was conducted in 500 patients to measure the effectiveness of a new growth factor in
reducing healing time of distal radial fractures. The authors reported that average healing time
was reduced from 9.2 to 8.9 weeks (P < 0.0001). Because the difference was highly statistically
significant, they recommended routine clinical use of this drug despite its high cost. A more
appropriate interpretation of these results is that they are
1- clinically significant.
2- statistically significant but perhaps not clinically significant.
3- statistically and clinically significant.
4- not statistically or clinically significant.
5- nonconclusive.
PREFERRED RESPONSE: 2
DISCUSSION: The results are statistically significant (at the arbitrary level of P < 0.05). That is, they
indicate a probability of only 1/10,000 that the observation that the drug is effective in reducing healing
time by 0.3 weeks occurred by chance selection of the study subjects. However, because the
statistical power of a study increases with the number of subjects included (sample size), a difference
that is trivial clinically can occur with a very high level of statistical significance (a very small P-value) if
enough patients are included in the study. Because of this, the P-value alone, no matter how small,
does not establish clinical significance or importance. Rather, the clinical significance of the observed
difference must be assessed taking into consideration the medical importance of the difference if it is, in
fact, true in the general population. In this example, the reduction in healing time of only a few days is
probably clinically unimportant, particularly if the use of the new growth factor is expensive, complex,
and/or has substantial side effects.