Archived Content

The content on this page is provided for reference purposes only. This content has not been altered or updated since it was archived.

Vaccines, Blood & Biologics

Postmarketing Requirement (PMR) Review Memo - Cervarix

Date: October 15, 2009
FDA STN:125259
Sponsor: GlaxoSmithKline Biologicals (GSK)
Product:CervarixTM Human Papillomavirus (HPV) Bivalent (Types 16 and 18) Vaccine, Recombinant

Each 0.5mL dose contains: 20 mcg HPV 16 L1 protein

20 mcg HPV 18 L1 protein
500 mcg aluminum hydroxide
50 mcg 3-O-fesacyl-4’-monophosphoryl lipid A (MPL)
0.624 mg sodium dihydrogen phosphate dehydrate as buffer

Indication:Active immunization of girls and women aged 10 through 25 years for prevention of the following disease caused by the Human Papillomavirus (HPV) types 16 and 18 included in the vaccine (indication still under negotiation):
  • Cervical cancer
  • Cervical intraepithelial neoplasia (CIN) grade 2 or worse and adenocarcinoma in situ
  • CIN  grade 1
From:Michael Nguyen, MD
Medical Officer, Vaccine Safety Branch
Through:Rick Wilson, MD, MS, JD
Director, Division of Epidemiology Robert Ball, MD, MPH, ScM
Director, Office of Biostatistics and Epidemiology
CC: Robert Wise, MD, MPH
      Acting Deputy Director, Division of Epidemiology
     Andrea Sutherland, MD, MSc, MPH
     Acting Chief, Vaccine Safety Branch

1.         Approach to Safety Signal Evaluation

In evaluating the safety of any product the highest level of concern is generated by a “demonstrated safety risk” which is an adverse effect accepted as being caused by the product using established epidemiological and statistical methods.  The standard of evidence to demonstrate a safety risk is high, and usually involves a study specifically designed to evaluate whether a particular adverse event is caused by the product.  The next category for describing safety issues is a safety “signal”.  In the guidance document on Good Pharmacovigilance Practices and Pharmacoepidemiologic Assessment, the FDA defines a safety signal to be a “concern about an excess of adverse events compared to what would be expected to be associated with a product’s use”. 

This definition of “signal” is broad and allows for a range of levels of concern driven by a number of factors and the quality of the evidence available.  Factors that are typically considered in assessing the level of concern include: strength of the association (e.g. relative risk in a controlled study), temporal relationship of product and event, consistency of findings across available data, evidence of dose response effect, biologic plausibility and the susceptibility of the methods used to confounding, bias, and chance.  Levels of concern can range from low (an unexpected serious adverse event occurring after receipt of the product, in the setting of substantial methodological limitations in data or study design); to high (an unexpected serious adverse event after receipt of the product, with statistically significant increased relative risk and few limitations in the data and study design).  A low level of concern is referred to as a “weak” signal and a high level of concern as a “strong” signal.

The main value in identifying the levels of concern is in helping to establish what actions are required, with lower levels of concern typically handled by postmarketing surveillance and studies, and possibly with restrictions in indications until the risk can be clarified.  Whereas higher levels of concern might require additional studies before a product can be marketed.

In this evaluation we first present the analysis of the available data and then present our conclusions using this framework.

2.         Postmarketing Study Development Post FDAAA

            2.1        FDA Amendments Act (FDAAA)

Effective March 25, 2008, section 901 of FDA Amendments Act (FDAAA) created section 505(o), authorizing the FDA to require postmarketing studies or clinical trials at the time of approval, or after approval, if the FDA becomes aware of “new safety information”.  FDAAA at 505(o)(3)(B) states that studies and clinical trials may be required for one of three purposes:

  1. To assess a known serious risk related to the use of the drug;
  2. i.       To assess signals of serious risk related to the use of the drug; or
  3. ii.      To identify an unexpected serious risk when available data indicate the potential for a serious risk

2.2        Regulatory Application of FDAAA

Following the anticipated approval of GSK’s Cervarix vaccine, FDA has determined that GSK is required, pursuant to section 505(o)(3)(B)(iii) of FDAAA, to conduct a postmarketing study, “to identify an unexpected serious risk when available data indicate the potential for a serious risk.”

This postmarketing requirement is based upon clinical trial data that identified higher rates of spontaneous abortion among Cervarix recipients whose pregnancies occurred around the time of vaccination (defined as the last menstrual period occurring 30 days before until 45 days after vaccination), compared to control subjects. The findings were strengthened by exploratory analyses conducted by the National Cancer Institute (NCI). In these analyses, NCI identified higher rates of spontaneous abortion among 15–25 year olds who received Cervarix 0 to +90 days of the estimated date of conception.

3.         Responses to Sponsor Interpretations of Prelicensure Data on Spontaneous Abortions

          3.1        Sponsor Interpretation

On September 9, 2009 GSK presented their interpretation of the safety data to the Vaccines and Related Biologic Products Advisory Committee (VRBPAC). Basing their safety assessment on the Bradford-Hill criteria[1], GSK concluded that there is no evidence of a causal association between Cervarix vaccination and spontaneous abortion. Their enumerated reasons are summarized here:

  1.          Consistency: The observed difference in risk of spontaneous abortions in study HPV-008 was not consistent across other clinical studies.
  2.       Strength of association: The criterion for strength of association was not met because the magnitude of risk was small (RR 2.30 in HPV-008 and 1.17 in HPV-009), was not statistically significant, and no dose-response relationship existed between pregnancy loss and the number of vaccinations.
  3.      Specificity: The criterion for specificity was not met because an increased risk was not apparent among women who became pregnant >90 days after vaccination, despite evidence that there is a sustained immune response in vaccine recipients.
  4.      Temporality: The temporal relationship with study vaccination was only partially met. There was no difference among study groups in estimated fetal gestational age at the time of pregnancy loss.
  5.        Plausible mechanism: There is no clear plausible biologic mechanism to explain the increased risk of pregnancy loss.  Prelicensure animal toxicology studies do not indicate that Cervarix is teratogenic, or that the novel adjuvant is mutagenic, even at doses 24–54 fold higher than the equivalent dose administered to humans. There were also no differences noted in human studies in the proportions of congenital abnormalities to suggest that the vaccine is teratogenic or differences in birthweight to indicate that the vaccine interferes with embryonic development.

Additionally, GSK put forth the following alternative hypotheses to explain the noted imbalance.

  1.          Chance: The observed differences may be attributable to chance alone.
  2.        Loss of blinding: Due to differences in the rates of reactogenicity between study and control groups, GSK hypothesized that pregnant women who had believed that they had received the experimental vaccine and who elected to terminate their pregnancies in countries where abortion is illegal, would report these elective terminations as spontaneous abortions.  As described in GSK’s September 2, 2009 discussion document:

“The potential exists for subjects determining that they are pregnant peri-vaccination to elect to terminate their pregnancy based on their perceived likelihood of having received the experimental vaccine (for example, the Study HPV-008 informed consent form [version 2 December 3, 2004] stated “there may be possible risks to a pregnant woman, embryo, fetus and nursing infant that are unknown at this time”).  If this were the case, it can be expected that this would occur particularly among women who reported a local or general reaction to the vaccine.  In other words, individuals who experience an adverse event of any severity may be more likely to elect for an abortion due to a perceived risk to the fetus.”


iii.      Background rate: GSK stated that because the observed rates of pregnancy loss were within the background rate in the general population, and below the rates observed in clinical trials of another US-licensed HPV vaccine, that the imbalance was not sufficient to signal a true deleterious effect of vaccination on pregnancy outcomes.

Finally, a GSK identified subject matter expert in Obstetrics and Gynecology stated that the totality of evidence does not demonstrate any discernible increased risk, beyond the expected rate, for spontaneous abortion in pregnant women. He further emphasized that the quality of the reproductive toxicology studies and the lack of a clear biological mechanism were compelling evidence that the vaccine did not portend an elevated risk for women who became pregnant around the time of vaccination.

3.2        OBE Response

OBE disagrees with GSK’s assertion that the criterion for strength of association is not met. As a prerequisite, comparison groups in clinical trials are randomized prior to intervention and blinded as to their treatment; therefore, potential confounders are presumed to be distributed equally and the results, as much as possible, are unbiased. These methods ensure that any detected differences are not only meaningful but internally valid. The detection of an elevated relative risk in the preclinical data, even if not statistically significant, should at a minimum prompt further scrutiny. Moreover, if well selected, the comparison group within the trial represents the best estimate for a background rate for an adverse event, not one derived from outside studies.

Furthermore, although a precise biologic mechanism does not exist, it is clear that the AS04 adjuvant is a powerful immunostimulant.  Because in pregnant women, the fetus represents a significant and novel antigenic exposure, it is not completely implausible that the stimulatory effects may adversely impact the balanced state of immune suppression and immune tolerance of the fetus.  Moreover, animal studies are not absolutely predictive of human responses to vaccination.

Additionally, the alternative hypotheses put forth by GSK to explain the data remain hypotheses. As stated in their September 2, 2009 “Discussion on the Observed Numerical Imbalance in Rates of Spontaneous Abortions During the Peri-Vaccination Period in Clinical Trials with Cervarix,”

“[T]his analysis cannot be further validated as it is not expected that the women will retrospectively admit to having had an elective termination instead of a spontaneous loss.”

In light of these concerns, the VRBPAC member, Dr. Destefano, expressed ongoing reservations regarding the evidence on spontaneous abortions and articulated that further data are necessary to clarify the risk on pregnancy outcomes following vaccination.

“Relating to the spontaneous abortion, I guess I have similar sentiments as Dr. Noller.  It does look like the comparison group was low, but these are the results we have from the pivotal trial results.  That is what we have to work with.  We can try to speculate and explain, but right now this is what we have. I would like to hear more from the manufacturer about their postmarketing efforts and what they would try to do to add knowledge to this question.  I also would like to second the public comment that was made earlier that this finding for now I don't think would be something that would make me vote against the vaccine, but I think it is something that women need to be aware of before deciding to get the vaccine.”

As stated by the NCI committee, an “increase in risk of miscarriage seen in the subgroup of pregnancies conceived within three months of vaccination is a concern for a vaccine that many women of reproductive age are likely to receive.”

In conclusion, although the evidence does not amount to a demonstrated risk of spontaneous abortions, there is also not a complete absence of evidence to preclude a causal relationship.  Further study is warranted to address this remaining uncertainty. 
4.         Recommendations to the Sponsor for the Required Postmarketing Study

FDA is concerned that the current concept protocol for the US Phase IV trial, as submitted, would not be sufficient to address the potential risk posed by Cervarix vaccination on fetuses with a normal karyotype.  Also, given heightened public awareness of the safety concerns related to pregnancy outcomes, subject accrual may not be sufficient under the current study protocol.

FDA requests that GSK design an epidemiologic study to assess the risk of spontaneous abortion in women who become pregnant around the time of Cervarix administration (defined as the estimated date of conception occurring –30 to +90 days of vaccination). Fetal viability is complex and the majority of spontaneous abortions result from chromosomal abnormalities. To better determine the true effect of Cervarix administration on fetal viability, we request that at least a subgroup of subjects have a normal karyotype. Data collection should ideally be inclusive of factors that are thought to influence fetal viability including but not limited to age, smoking status (using plasma cotinine), number of previous pregnancies, history of spontaneous abortion, and alcohol consumption. The target population should contain females aged 15–25 years, corresponding to the subgroup where the imbalance was noted in clinical trials. For analysis, the comparison group should be female controls (or the cohort from which cases were derived) who did not have a spontaneous abortion. As previously mentioned in GSK’s concept protocol, the sample size should be sufficient to detect a relative risk of approximately 2.0 for spontaneous abortions with 80% power.

The required postmarketing study does not need to be completed in the United States (US), but the study findings should be generalizable to females in the US.  Additionally, we request that GSK propose a timeline for completion of the draft and final versions of the study protocol.  Key elements to incorporate into the study protocol include progress reports to the Office of Biostatistics and Epidemiology (OBE) every 6 months for the duration of the study, as well as a timeline for the final report no later than 6 months after study completion.

Because this is a dedicated study of spontaneous abortions, OBE recommends that a general pregnancy registry still be implemented. 

5.         Current Status of the PMR Study

Although the final study design will be determined after licensure a general timeline has been established for this study.  While still under discussion, all aspects of the final study protocol are subject to FDA review and ultimate approval pursuant to Section 505(o)(3) of FDAAA.

            5.1        Current Study Design

The primary study population will be comprised of women whose estimated date of conception lies between –30 days and +90 days from nearest Cervarix vaccination, relative to a comparison group. The study will include karyotype analysis in a subset of women in order to address the issue of background spontaneous abortions due to chromosomal abnormalities.

The timetable submitted on September 21, 2009, states that the trial will adhere to the following schedule:           

  • The draft protocol must be submitted by December 31, 2009 and the final protocol will be submitted by April 30, 2010.
  • Study initiation will preferably occur within six months but no later than 12 months after protocol submission.
  • Interim reports will be submitted to the FDA every 6 months for the duration of the study.
  • The study completion date will be subject to the final study design (case-control versus cohort study) which is still under discussion.  Additionally, the study completion date will be to subject to other factors impacting patient accrual, including overall vaccine uptake, date of study initiation, pregnancy avoidance behaviors among vaccine recipients, and size of the surveillance population. The study will be considered completed when subject enrollment is sufficient to detect an increased relative risk of approximately 2.0 for spontaneous abortions.

The final clinical study report will be submitted within 6 months after study completion

[1] Weekly Epidemiological Review 2001 76: 85-89.

Page Last Updated: 03/30/2015
Note: If you need help accessing information in different file formats, see Instructions for Downloading Viewers and Players.
Language Assistance Available: Español | 繁體中文 | Tiếng Việt | 한국어 | Tagalog | Русский | العربية | Kreyòl Ayisyen | Français | Polski | Português | Italiano | Deutsch | 日本語 | فارسی | English