The Food and Drug Administration’s opinion that a progesterone gel 8% lacks statistically significant proof of efficacy in reducing preterm births among U.S. women with a short cervix, while stronger evidence in ex-U.S. populations is offset by the different racial makeup in the foreign sites, makes a difficult-to-surmount problem for the sponsor going into the Jan. 20 meeting of the Reproductive Health Drugs Advisory Committee.
The FDA reviewers also rejected the sponsor’s statistical analysis on a crucial point. The new drug application was based on a single pivotal phase III trial, "Study 302," a prospective randomized trial that enrolled women with a cervical length of 1-2 cm as measured by transvaginal ultrasound at 23 U.S. sites and 21 sites in nine foreign countries.
Columbia Laboratories Inc.’s progesterone gel 8% would be indicated for the reduction of risk of preterm birth in women with a singleton gestation and a short uterine cervical length in the midtrimester of pregnancy.
The 8% progesterone gel has been known by the proposed brand name Prochieve, but while the drug is marketed under that brand name for different indications in other countries, it is not cited as the brand name for the prevention of the preterm birth indication in Columbia’s background document.
Supporting evidence came from the earlier "Study 300," which had failed to find statistically significant evidence that progesterone worked in preventing preterm births in high-risk pregnant women but found evidence of efficacy in the subgroup of women with short cervixes. The sponsor and the FDA agreed on the design of Study 302 based on this latter result. The estimated user fee goal date for Columbia’s application is Feb. 26.
Columbia presented analyses based on pooled data from both studies, a Cochran Mantel Haenszel test stratified by primary pooled study site and risk strata, but the FDA said this was not statistically sound due to the differences in the study populations and initiation of treatment. Moreover, the supplemental information from Study 300 did not provide adequate efficacy evidence since the short cervix subgroups analyzed included too few subjects – 9-116 subjects in total, depending on the specific definition used for short cervix.
According to the FDA, a cervical length of 2.5 cm or less "is generally considered to represent a clinically meaningful short cervix and a risk factor for preterm birth; only 33 subjects in Study 300 fell into this range." At a July 2007 meeting between Columbia and the FDA’s Division of Reproductive and Urologic Products, following the failure of Study 300, the sponsor proposed defining a "short cervix" as less than 2.8 cm for what would become Study 302, and the division said that while the literature generally defines a short cervix as one measuring less than 2.5 cm, "a robust finding in a population at any prespecified shortened cervical length would be of clinical importance."
Lack of Evidence for U.S. Population
Most damagingly for Columbia, the FDA’s analyses indicate that progesterone gel was not associated with a reduction in preterm birth in the U.S. subjects at any gestational age, with a treatment difference of only 2.4% in favor of progesterone gel compared with placebo before 33 weeks’ gestation. Moreover, "the CI around this point estimate of the risk difference indicated that the difference is not statistically significant," the FDA says in a background document prepared for the advisory panel meeting. "In addition, the key secondary end point of neonatal mortality and morbidity did not show a statistically significant treatment effect of progesterone gel."
The FDA found the results were literally all over the map, with South Africa and Belarus having very high rates of preterm births in the placebo patients and none in the progesterone-treated group, a result not replicated elsewhere.
"The treatment effect also shows marked heterogeneity, even within the U.S., where the difference in the preterm birth rate in placebo-treated subjects compared to those treated with progesterone gel ranged from -12% (favoring progesterone gel) to +7% (favoring placebo)," the Division of Reproductive and Urologic Products noted. "Similarly, three large Indian sites showed pooled results that favored placebo by 9%, while other non-U.S. sites showed treatment effects favoring progesterone gel that ranged from -4% to -31%."
The racial makeup of the U.S. enrolled population in Study 302 also differed substantially from that enrolled outside the United States, which is bound to spell additional trouble for the application. The majority of U.S. subjects were black (60%) and white (29%), while the majority of subjects in non-U.S. sites were Asian (59%) and white (23%). According to the FDA, "The preterm birth rate in placebo-treated subjects varies considerably by race. This may also reflect regional differences, as race was differentially distributed by region. Overall, there was minimal efficacy in Asian subjects, possibly because the background rate of prematurity was quite low. Efficacy in [whites] and blacks favored progesterone gel. In [whites], the treatment benefit was greatest in the earlier gestational ages, while in blacks, efficacy increased as gestational age advanced."