Analysis of Repeated Assessments Data

Sir, Many patients come with vertigo as their presenting symptom. They have been to physicians/neurologists/ENT specialists and have been fully investigated and treated symptomatically without any positive findings or relief. On detailed probing, these patients are found to be cases of depression-mostly of retarded variety. There is no specific age, sex, domicile or socioeconomic distribution of such cases The vertigo is of the classical rotational type and different from unsteadiness or dizziness. I put these cases on an antidepressant (usually imipramine), in addition to an antivertigo drug (meclizine or dimenhydrinate or cinnarizine). In fact, most of the patients are already on these drugs without having derived any benefit. With the addition of the antidepressant, there is a rapid and definite improvement in vertigo as well as depression. The antivertigo drug is gradually withdrawn and the antidepressant is continued. There is no recurrence of vertigo. In several cases, vertigo came back on stopping imipramine. All these strongly suggest that vertigo, in these cases, is a part of the overall symptomatology of depression. Although patients of vertigo (of organic origin) have been found to have a strong psycholo.gical component, along with an element of secondary gain (Kaplan etal., 1994), I could not find any mention of vertigo as a symptom of depression or of any primary psychiatric iilness for that matter. However, Srivastava (1998) has recently reported a case where deafness (of organic origin) and depression both responded to imipramine. These associations certainly deserve and need to be studied and reported upon. Sir, A frequent inferential statistical error in articles published in the Indian Journal of Psychiatry relates to the analysis of data obtained on the same measure from the same subjects at different points in time. In the January 1998 issue of the UP, for example, two articles (Malathi et al., 1998; Dalai et al., 1998) used this design; both should have been analysed using more appropriate statistical techniques. When repeated assessment data are obtained from one group only, the null hypothesis is that there is no change in the mean score of the group across occasions of assessment. To pair occasions of assessment, and to apply the paired t test is not the best way to analyse these data. This method is weak because it employs several paired t tests to test the single null hypothesis; therefore, it inflates the risk for a type I (false positive) statistical error. This …