We have read with interest the article by Nicolaides et al. It is noteworthy that the authors reported a high rate of removal of the cervical pessary. Specifically, this occurred in 22.3% of cases before 34 weeks of gestation. However, it would be interesting to know the gestational age at delivery (mean, interquartile range, and range) of these patients.

Similarly, it would be relevant to know the percentage of patients with a short cervix; if this figure was high and the pessary was removed shortly thereafter, this could explain the differences in the results of this trial and the subgroup analysis reported by others.

In a previous study of pessary for prevent preterm birth in singletons, we did not remove the pessary in patients with preterm prelabor rupture of membranes, and this was not associated with differences in the frequency of clinical chorioamnionitis or neonatal sepsis between the pessary and control groups.

A strength of the study reported in the Journal was that there was systematic training in the measurement of cervical length, which is the method to identify the patient at risk. In contrast, for the key therapeutic intervention of the trial, many physicians responsible for inserting the pessary did not receive supervised training; this raises the question of whether inadequate placement of the pessary may explain the negative results in the subgroup analysis of patients with twin gestations and a short cervix.

We also noted a case in which cervical edema developed after insertion of the pessary, and this required removal under general anesthesia. It would be helpful to know if that patient had subclinical uterine contractions that may not have been identified prior to insertion of the pessary.

Our experience with insertion of a cervical pessary has led us to routinely ask patients whether they had developed pain and to monitor those with serial cervical length determinations. Patients whose cervix shortens may be developing preterm labor and benefit from tocolysis and corticosteroid administration. We would like to ask whether patients allocated to a cervical pessary and the control group underwent serial cervical length monitoring.

It is noteworthy that the authors reported the use of vaginal progesterone in 2 patients in the control group; in both cases, the gestational age at delivery was greater than 34 weeks. Given that some recent studies have shown that vaginal progesterone may reduce the rate of preterm birth by 30% and morbidity, could the administration of vaginal progesterone be a confounding factor in the results of the trial?

Finally, the authors used a random-sequence generation without stratification by center; is it possible that such a method could introduce biases in the composition of the study groups?