Sir: Brookes et al (Psychiatric Bulletin, September 2000, 24, 329-330) report “a slight improvement in the proportion of electroconvulsive therapy (ECT) treatments considered therapeutic” following implementation of a dose titration protocol as recommended by the Royal College of Psychiatrists. Unfortunately their interesting paper did not include a statistical analysis. Comparing the proportion of therapeutic fits before and after the changes to administration (131/213 cases, 62% v. 92/132 cases, 70%) yields a non-significant finding using the chi-square test (χ²=2.39, d.f.=0.12).

However, ECT treatments where the fit duration was considered too short (under 20 seconds) did fall (from 79/213 cases, 37%, to 21/132 cases, 16%) and this result is highly statistically significant (χ²=17.8, d.f.=1, P<0.01, difference in proportions=21%, 95% confidence limits 12-30%). This backs up the authors' assertion that College guidelines reduced the likelihood of sub-therapeutic stimulations at the expense of increasing the number of prolonged (> 50 second) fits.

Although statistical significance is not the same as clinical significance, papers that report on practical service developments will be strengthened by an appropriate analysis. Statistics can enlighten as well as frighten!