To the Editor: Peter N. van Harten, M.D., Ph.D., and colleagues +(1) reported that patients whose neuroleptic therapy was interrupted more than twice were approximately three times more likely to develop tardive dyskinesia than those whose therapy was interrupted two times or fewer. This odds ratio was calculated for the lifetime intake of neuroleptics and anticholinergics, and both were not statistically significant. They conclude that neuroleptic therapy should not be interrupted to minimize the risk of tardive dyskinesia. We take exception to such a broad conclusion.

The authors make a rather common error of interpreting correlation as causation. It would seem likely that patients who experienced higher rates of extrapyramidal symptoms through the course of their neuroleptic treatment would be more likely to discontinue medications and thus have more frequent drug interruptions. Since extrapyramidal symptoms predict future development of tardive dyskinesia +(2–+5), the observed higher risk of tardive dyskinesia associated with antipsychotic drug interruptions is likely driven by the former association (i.e., between extrapyramidal symptoms and tardive dyskinesia). The linear regression model included lifetime anticholinergic use as a predictor variable and found no statistically significant contribution to the model (p=0.06). Although one may use the cumulative anticholinergic dose as a proxy of past extrapyramidal symptoms in the analysis, we believe that this does not adequately address the issue raised. This is because antipsychotic drug interruptions may have substituted for anticholinergic drug treatment. Considering this, the authors may want to enter the cumulative anticholinergic dose first in a stepwise regression model to account for the variance explained by extrapyramidal symptoms and subsequent anticholinergic drug treatment before examining the predictive value of drug interruptions. Albeit, this is informative to the extent that the cumulative anticholinergic dose reflects past extrapyramidal symptoms.

Furthermore, we note that total neuroleptic exposure (in chlorpromazine equivalents) has not always been a good predictor of tardive dyskinesia. This is likely because of the broad pharmacokinetic variability found among patients. A more common useful measure has been total time of neuroleptic exposure +(6). This variable was not addressed by the article.

In conclusion, the interpretation of a correlation as causation is misleading. We maintain that strategies to treat tardive dyskinesia should include cessation of antipsychotic drug treatment if otherwise clinically feasible.