I agree with Nakchbandi and colleagues [1] that voluntary, not mandatory, HIV testing in pregnant women is the superior policy option. For practicing clinicians to effectively use this analysis, however, epidemiologic and ethical factors not addressed in this analysis must also be considered.

Roughly 2 million births occur in the United States each year. With the conservative assumption that 80% of these are single births, 1.6 million pregnant women could be subject to mandatory HIV testing each year. Screening tests for HIV are excellent, with sensitivity and specificity both almost 99%. However, in the general U.S. population, where the seroprevalence of HIV disease is low (0.5%), we would expect approximately 24 000 positive test results, of which only one third would truly reflect disease (positive predictive value, 33.15%). Nakchbandi and colleagues' prediction that 1200 of 2000 HIV-infected infants might be saved by mandatory screening must be balanced against the huge number of pregnant women screened and a false-positive rate of 67%.

From an ethical standpoint, mandatory screening and subsequent treatment of pregnant women with zidovudine would violate patient autonomy. This doctrine states that everyone has a right to be free of interference and control by others. Autonomy safeguards the patient's right to make decisions that are consistent with his or her own values, even if those decisions are contrary to what others believe is in their best interest [2]. Incorporated in this construct is the idea that patients should not be required to receive information against their will. Testing a pregnant woman who does not want to be tested would certainly violate these rights.

Society currently requires HIV testing in certain population categories (military personnel, federal prisoners, some immigrant populations). Nakchbandi and colleagues' decision analysis should strongly dissuade us from making "pregnant women" the next category.