Dr. Olfson and Colleagues Reply

To the Editor: Drs. Rico-Villademoros and Calandre suggest that in our case control study confounding by indication may have attenuated associations of olanzapine with hyperlipidemia and exaggerated associations of risperidone and ziprasidone with hyperlipidemia. While we can not exclude this possibility, the assessment of pretreatment risk of hyperlipidemia among patients who receive these antipsychotic medications remains a challenge in clinical practice. In this context, we believe that by matching comparison subjects to individuals on age, sex, race, psychiatric diagnosis, and time period, we have created comparable groups in terms of underlying pretreatment risk for hyperlipidemia.

The two randomized controlled trials cited by Drs. Rico-Villademoros and Calandre were not adequately powered to detect differences among antipsychotic medications in the development of hyperlipidemia. In the study by Lindenmeyer et al., all of the cholesterol increases remained within the normal clinical range (1) . In the CATIE trial, the five study groups (olanzapine, quetiapine, risperidone, perphenazine, and ziprasidone) did not significantly differ from one another in the proportion started on a cholestatin (2) .

The study reported by Kingsbury et al. provided preliminary evidence of a reduction in serum cholesterol following switch from olanzapine, risperidone, or a first-generation antipsychotic to ziprasidone (3) . However, without an untreated comparison group, it remains unclear the extent to which this decrease was mediated by withdrawal of the previous medication or by initiation of ziprasidone. Comparison of untreated patients and ziprasidone-treated patients is an important feature of our case control study.

We agree that observational research is vulnerable to problems that flow from differences between patient groups in pretreatment risk factors for the disease outcome (e.g., hyperlipidemia) and that randomized controlled trials offer an unrivaled check against such bias. In the absence of adequately powered randomized controlled trials, however, carefully conducted observational research may help physicians in their evaluation of the risks of hyperlipidemia.