Influenza immunization during pregnancy: toward a balanced assessment of safety evidence

KEYWORDS:

• Influenza
• pregnancy
• maternal immunization

To the Editor:

In a recent issue of Human Vaccines and Immunotherapeutics, a commentary¹ on vaccine efficacy/effectiveness and safety in pregnancy was published. The commentary presented evidence originating from randomized controlled trials (RCTs) and observational studies, with a stated objective to “analyze new observational evidence…….and to add new evidence from the available RCTs, supporting the assertion that better evidence is needed about the safety of influenza vaccination during pregnancy”.¹ However, we have several concerns about the interpretations and assertions made in this commentary.

First, we agree with the author that RCTs are preferable to observational studies. Randomization, strict inclusion and exclusion criteria, and highly controlled experimental conditions ensure a high degree of internal study validity. Yet, RCTs are only statistically powered to evaluate the primary outcome(s), not necessarily rare safety outcomes. Moreover, they do not reflect the “real-world” environment; thus, alternative study designs are required for complementary assessment of less frequent and/or adverse outcomes, as well as for evaluating effectiveness and safety under “real-world” circumstances.^2,3 For the comprehensive evaluation of the effectiveness and safety of influenza immunization during pregnancy, observational studies using large population databases have indeed played an important role.

The author proposes that observational studies should be excluded from effectiveness and safety assessment because of bias introduced by self-selection of vaccination by healthier women (e.g., those of higher socio-economic status). He cites a recent cross-sectional survey of 1,771 pregnant women designed to assess influenza vaccination coverage in the United States to support that variability in influenza vaccination by socio-demographic characteristics is evidence of a healthy-vaccinee bias¹ – a form of confounding that could overestimate the beneficial effects of vaccination.⁴ It is well recognized that baseline differences in socio-demographic and other characteristics between vaccinated and unvaccinated women must be addressed in observational studies of vaccination during pregnancy.^4,5 Advanced statistical approaches, such as propensity score methods, are routinely used in such studies to address this bias,^6,7 and while application of these methods does not guarantee removal of all potential healthy-vaccinee bias, they have been shown to effectively reduce bias in studies using large administrative databases,⁸ including studies of influenza vaccine in other populations.⁹ The preponderance of evidence from observational studies indicates no increased risk of adverse birth outcomes following influenza vaccination in pregnancy.^10-13 Moreover, it has been empirically demonstrated that it would take an implausibly strong unmeasured confounder in order for a healthy-vaccinee bias to completely obscure a true increased risk.⁶

Second, the author makes a strong inference based on actual counts of individual adverse events monitored by the available RCTs.^14-17 As one example, the author reported a risk ratio for maternal death based on two deaths in the vaccination group and none in the placebo group,¹⁴ resulting in a 95% confidence interval of 0.24 to 103 (as reported by a Cochrane review¹⁸) – a range too wide to be meaningfully interpretable. Furthermore, the maternal deaths both occurred more than 5 months after the vaccination, and from causes unlikely to be related (see Table S13 in Madhi et al.¹⁴). In accordance with international standards, each of the trials on maternal influenza immunization^14-17 monitored post-immunization safety, including pregnancy and health outcomes for 6 months (see Omer et al. for a complete list of all outcomes ascertained by the three most-recently published RCTs¹⁹). However, it is important to recognize that these studies were not designed to have the statistical power to detect differences in non-primary outcomes between the treatment groups. In anticipation of such challenges, the three Bill & Melinda Gates Foundation-funded trials^14-16 were coordinated from the planning phase to permit later pooled analyses of other outcomes, which the individual trials were not powered to assess. The pre-specified protocol for the pooled analyses¹⁹ was published before two of the three trials were even completed,^15,16 and includes plans to evaluate safety outcomes in mothers and infants, with a particular focus on those outcomes that were too rare to be assessed in individual trials (e.g., stillbirth).¹⁹ These analyses are in progress and will contribute additional important evidence on this subject.

Finally, the author’s assertions that RCT investigators had conflicts of interest due to receipt of prior support from vaccine manufacturers is unjustified. Many large vaccine studies conducted throughout the world are funded by the manufacturers of the vaccine. The studies are carefully designed, safety assessments are comprehensively outlined a priori, independent Data Safety and Monitoring Boards are in place to monitor the study results, and the resultant data are presented in a transparent and comprehensive manner in peer-reviewed journals, with full disclosure of any potential conflicts of interest. Additionally, these RCTs^14-16 had a significant proportion of investigators without a history of receiving funding from vaccine manufacturers – further ensuring that “independent” voices were represented. They were entirely funded by the Bill & Melinda Gates Foundation, and the investigators at each of the study sites were academic researchers who are well known and respected for their scientific expertise.

Pregnant women have been immunized with influenza vaccines for several decades in the United States and Canada,²⁰ and independently conducted systematic reviews of the now large body of international evidence have not found any indication of increased risk of adverse maternal or infant health outcomes.^10-13,21,22 Nevertheless, there will continue to be an ongoing program of independent safety assessments of influenza vaccines used in pregnant women. This will include the pooled analyses of the RCT data,¹⁹ routine pharmacovigilance and observational studies using large population databases, and reporting systems for adverse events, such as the Vaccine Adverse Event Reporting System (VAERS) or the Vaccine Safety Datalink (VSD) systems that are in place in the United States. Given the substantial evidence of efficacy and safety of maternal immunization – including from four RCTs^14-17 – a sufficient clinical equipoise does not exist. Therefore, it would be unethical to conduct further maternal influenza vaccine RCTs unless a substantially different vaccine is developed. While we agree with the author that the safety of maternal influenza immunization is critically important, an adequately balanced perspective was not provided in the commentary.¹

Disclosure of potential conflicts of interest

No potential conflict of interest was reported by the authors.