Reply: neuropsychological sequelae of bacterial meningitis: the influence of alcoholism and adjunctive dexamethasone therapy

Abstract

Van de Beek et al. present new important data concerning the clinical outcome of patients with alcoholism and pneumococcal meningitis. The data was derived from a subgroup analysis of the patient population of their comparative key study on neuropsychological sequelae after pneumococcal versus meningococcal meningitis (van de Beek et al., 2002). With respect to our paper, they raised three issues: (i) Absence of statistically significant differences between patients after Streptococcus pneumoniae and Neisseria meningitidis meningitis in our study. Van de Beek et al. reported that only 2 of 26 patients of the pneumococcal meningitis group who were actually examined neuropsychologically suffered from alcoholism. In contrast to our findings (Schmidt et al., 2006), van de Beek et al. found significant neuropsychological differences between patients who survived pneumococcal and meningococcal meningitis. Before we received the information from van de Beek et al. by this communication, we presumed that one of several possible reasons for this discrepancy was the different proportion of patients with alcoholism in our study versus those in van de Beek’s study. This now can be ruled out. Searching for further explanations, van de Beek et al. suggest that different Glasgow Outcome Scores (GOS) or differing intervals between hospital stay and re-examination of the respective meningitis group might have led to this discrepancy. Patients examined after pneumococcal meningitis in our study had a significantly lower mean GOS value than patients examined after meningococcal meningitis. Therefore, the differences should have been even more pronounced. The interval between hospital admission and neuropsychological testing was not significantly different in our pneumococcal and meningococcal patient groups. Van de Beek et al. pointed out that a group size of 16 patients per group might not yield enough statistical power to discriminate differences. This is true, especially for domains with only limited inter-group differences. We think that the reported difference between meningococcal and pneumococcal meningitis patients in van de Beek’s study (or in other terms the non-existing difference in our study) might be a result of different statistical approaches: In contrast to our colleagues’ work we compared the groups using two-sided tests (not assuming a priori one group to be more severely affected than the other). This approach results in an alpha error twice as high as with one-sided testing. If one-tailed tests are used, it is recommended to calculate with half the alpha error that would have been used with a twotailed study [according to the ICH topic E9 (EMEA, 1998)]. Applying a P-value of 0.025, only 3 out of 14 neuropsychological examinations for which van de Beek et al. gave mean T-values with standard deviations were significantly different (Visuospatial reasoning, WMS-R immediate memory, Trailmaking B) after their univariate analysis of covariance. Comparing the T-values of van de Beek’s groups in our way with two-sided t-tests for independent samples, only the items Visuospatioal reasoning and Trailmaking B yielded statistically significant differences in van de Beek’s study groups. Contrary to our analysis of alertness functions (performed after age-related transformation into z-values), reaction speed (dominant/non-dominant hand) and a two-choice paradigm in van de Beek’s paper were significantly slower for the pneumococcal than for the meningococcal meningitis group. In contrast to all other domains, for the comparison of the reaction speed, a transfer of the raw data into age-adapted zor T-values (derived from age-matched controls or with the standard values from the respective test manuals) was not