Informing randomized clinical trials of respiratory syncytial virus vaccination during pregnancy to prevent recurrent childhood wheezing: A sample size analysis

Background: Early RSV illness is associated with wheeze-associated disorders in childhood. Candidate respiratory syncytial virus (RSV) vaccines may prevent acute RSV illness in infants. We investigated the feasibility of maternal RSV vaccine trials to demonstrate reductions in recurrent childhood wheezing in general paediatric populations. Methods: We calculated vaccine trial effect sizes that depended on vaccine efficacy, allocation ratio, rate of early severe RSV illness, risk of recurrent wheezing at age 3, and increased risk of RSV infection on recurrent wheezing. Model inputs came from systematic reviews and meta-analyses. For each combination of inputs, we estimated the sample size required to detect the effect of vaccination on recurrent wheezing. Results: There were 81 scenarios with 1:1 allocation ratio. Risk ratios between vaccination and recurrent wheezing ranged from 0.9 to 1.0 for 70% of the scenarios. Among the 57 more plausible scenarios, the lowest sample size required to detect significant reductions in recurrent wheezing was 6196 mother-infant pairs per trial arm; however, 75% and 47% of plausible scenarios required >31,060 and >100,000 mother-infant pairs per trial arm, respectively. Studies with asthma endpoints at age 5 will likely need to be larger. Discussion: Clinical efficacy trials of candidate maternal RSV vaccines undertaken for licensure are unlikely to demonstrate an effect on recurrent wheezing illness due to the large sample sizes likely needed to demonstrate a significant effect. Further efforts are needed to plan for alternative study designs to estimate the impact of maternal RSV vaccine programs on recurrent childhood wheezing in general populations.

We estimated the RCT sample size that would be required to demonstrate an effect of RSV vaccination during pregnancy on recurrent childhood wheezing through 3 years of age. We chose this endpoint because recurrent childhood wheezing is an important cause of paediatric morbidity [10], [12]. Additionally, RCTs with recurrent childhood wheezing as a primary endpoint will likely be more favourable than other childhood respiratory outcomes for several reasons. First, an outcome that could be assessed at a younger age is advantageous, because longer trials have higher rates of losses to follow-up and are more costly. Second, recurrent wheezing at 3 years is expected to have a higher prevalence than asthma at 5 years [12], [15], increasing the statistical power for a given sample size. Third, if some portion of asthma at 5 years is caused by early infant RSV exposure, it is plausible that recurrent wheezing at 3 years mediates this relationship. Fourth, atopy and environmental risk factors are more likely to be found in children with asthma than in children with recurrent wheeze [9], [14], [16], [17], indicating that early RSV illness may contribute a larger attributable fraction to recurrent childhood wheezing at 3 years than to asthma at 5 years. We based our approach on an earlier study that estimated the detectable risks in observational studies of potential foetal benefits of maternal influenza vaccination [18]. The approach is described in the following illustrative example. Suppose that 1000 mother-infant pairs were randomized to each arm of a placebo-controlled RCT of a maternal RSV vaccine. If the baseline risk of vaccine-preventable early infant RSV illness in the population is r, then approximately 1000 × r infants born to women in the placebo arm will acquire RSV illness during the vaccine-preventable period. In the vaccine arm, prevention of infant RSV illness is proportional to the vaccine efficacy (%), VE, implying that 1000 × r × (1 − VE) infants born to women immunized against RSV will acquire an RSV illness. Supposing an attack rate of 20% (r) and vaccine efficacy of 50% (VE), we would expect 200 cases (= 1000 × 0.2) of RSV illness among infants in the placebo arm and 100 cases (=1000 × 0.2 × 0.5) of RSV illness among infants in the vaccine arm. We considered infant RSV illness within the first months to be a mediator on the causal pathway of maternal RSV vaccine protection against recurrent childhood wheezing since these early RSV illnesses occur within the vaccine-preventable time frame and are associated with the highest acute morbidity (Fig. 1). That is, the ability of RSV vaccination during pregnancy to prevent later recurrent childhood wheezing would operate through the reduction in early RSV illness during the first months of life. If the baseline risk of recurrent wheezing at 3 years, w, is 5% among infants who do not experience an early RSV illness, and 20% among those who do, this would imply a 4-fold relative increase in the risk resulting in a risk ratio RRRW of 4. (The subscript RW is used to denote the risk ratio of RSV illness associated with wheezing illness). In the placebo arm, applying these risks to the 200 and 800 infants with and without RSV illness, respectively, results in 80 total cases of recurrent wheezing among infants in the placebo arm ((200 × 0.20 + (800 × 0.05)). Among infants born to RSV-vaccinated mothers, 900 will have a 5% risk of developing the recurrent wheezing, and 100 will have a 20% risk, for a total of 65 cases ((900 × 0.05) + (100 × 0.2)). In this hypothetical scenario, maternal RSV immunization reduced the risk of childhood wheezing from 80 per 1000 (placebo arm) to 65 per 1000 (vaccine arm), for a risk ratio associated with vaccination (RRVW) of 0.81 (= 65 per 1000 ÷ 80 per 1000) (Fig. 2). (The subscript VW is used to denote the risk ratio of vaccine associated with wheezing illness, or the vaccine effect size). We used estimates from the literature supplemented by expert opinion to inform the following parameter inputs for our model: (1) baseline attack rate of severe early RSV, r; (2) vaccine efficacy, VE; (3) risk of recurrent wheezing during childhood, w; and (4) risk ratio for recurrent wheezing according to early RSV illness, RRRW (Fig. 1). We also considered two allocation schemes, 1:1 and 2:1. Theoretical causal diagram for the relationships between maternal RSV vaccination, severe early infant RSV- lower respiratory infections, and later recurrent childhood wheezing. The diagram illustrates how maternal vaccination against RSV may prevent the development of recurrent childhood wheezing (the endpoint) through preventing an early severe RSV-related lower respiratory infection (LRI) during infancy (the mediator). Links between elements in the diagram and parameters in the sample size study are described in the blue caption boxes. The sign labelling each arrow indicates the direction of association as positive (+) or negative (−) between the connecting nodes. If the relationship between early severe RSV-LRI and recurrent childhood wheezing is confounded by a predisposition to respiratory infections, then observational studies estimating the increased risk of recurrent childhood wheezing due to early RSV-LRI may be overestimated. Illustration of the parameters used to estimate risk ratios and sample size in clinical trials of maternal RSV immunization on development of later recurrent childhood wheezing. Each filled dot in this figure represents a mother-infant pair, with the colour representing their RSV infection status and black outline indicating infants who go on to develop recurrent childhood wheezing. There are 100 rows of 10 dots, to represent 1000 mother-infant pairs randomized each to placebo and vaccination. Following the in-text example, 20% of placebo (200 mother-infant pairs; purple dots) acquire early and severe infant RSV vs. 10% in the immunized arm (100 mother infant pairs; purple dots), for a vaccine efficacy of 50%. If early RSV illness increases the later development of recurrent childhood wheezing, then the proportion of children who develop recurrent wheezing will be higher among those with early RSV. This is shown by the higher density of recurrent wheezing cases (black outline) among those with RSV illness (purple dots) vs. those without (blue dots). Summing the wheezing cases, there are 60 cases among the 300 infants who acquired RSV (for a 20% risk) vs. 85 wheezing cases among the 1700 infants who did not acquire RSV (for a 5% risk) giving rise to four-fold increased risk of wheezing in children exposed vs. unexposed to early and severe infant RSV. This example shows 80 cases of childhood recurrent wheezing among the placebo arm vs. 65 among the immunized arm for a risk ratio between vaccination and childhood recurrent wheezing (RRVW) of 0·81. Up to two-thirds of infants acquire RSV during their first year of life [5]. Only a proportion of these illnesses is preventable by maternal immunization, as passive protection from maternal antibodies wanes with time and are not expected to exceed 6 months [1]. For simplicity, we assumed that subclinical or mild early RSV illness would have a very weak association, if any, with development of later recurrent childhood wheezing. We found the most stable estimates of severe RSV incidence during this risk period were for infants younger than 6 months of age. Thus, we used estimates of the baseline RSV attack rate during the first 6 months of life that resulted in the more severe outcomes of LRI or hospitalization. In this report, we use “early severe RSV illness” to define RSV illness during the first 6 months of life that resulted in LRI or hospitalization. A recent meta-analysis estimated that between 6.3% and 16.9% of infants acquired a RSV-related LRI in the first 6 months of life across world income strata [4]. The rate of hospitalization in countries with adequate access to medical care may best reflect the attack rate of the most severe illnesses. RSV-hospitalization estimates for most income strata were greater than 2% between 0 and 5 months of age, with the upper bound of 2.7%. We used values of 2.7%, 6.0%, and 17.0% as attack rates of early severe RSV illness. There is no information currently available on the efficacy of candidate maternal RSV vaccines. The WHO Preferred Product Characteristics for RSV Vaccines specifies that a vaccine with 50% efficacy would be favourable, while greater than 70% efficacy would be preferred [1], [2]. We used values of 50.0%, 70.0%, and 90.0%. Estimates for recurrent childhood wheezing vary widely across countries and represent the overall rate of recurrent wheezing in a population, independent of early RSV status. We were unable to find global estimates of recurrent childhood wheeze for children aged up to 3 years. As a proxy, we used the International Study of Asthma and Allergies in Childhood (ISAAC) survey which reports the most recent international estimates for “symptoms of severe asthma” among children aged 6–7 years, which was the youngest age group ISAAC studied. Children had these symptoms if they had current wheeze and reported any of the following: ≥4 attacks of wheeze, ≥1 night/week of sleep disturbance from wheeze, or wheeze-affected speech, within the previous year. The global prevalence of these symptoms among this age group was 4.9%, with the highest prevalence in the Oceanic region (9.5%), and the highest centre rate in Costa Rica (20.3%). We used parameter inputs of 4.9%, 9.5%, and 20.0%. We expect the actual prevalence of recurrent wheezing at 3 years to be higher than the prevalence of these symptoms reported among 6–7 years. Furthermore, these estimates do not account for early RSV exposure status, even though we require estimates among children unexposed to early severe RSV illness. These overall estimates will approximate the estimate among the unexposed when the exposed population is small (i.e., for small baseline attack rates of severe early RSV) or when RRRW is low. Otherwise, they would be overestimates. We therefore chose a wide range of recurrent wheezing prevalence estimates to reflect these opposing measurement uncertainties. If RSV vaccination during pregnancy leads to reduction in later recurrent childhood wheezing, we assume the effect occurs by preventing early severe RSV illness among infants. The gold standard study design for estimating a causal risk ratio between early severe RSV illness and recurrent childhood wheezing would be a RCT of RSV prevention with predefined outcomes, however there have been few such trials conducted [19], [20]. Several observational studies have assessed the risk of recurrent wheezing associated with RSV-related hospitalizations in infancy, but they are at risk of confounding bias, particularly from factors predisposing infants to severe RSV illness and later development of wheezing, such as poor lung function. A 2017 systematic review summarizing these studies reported risk ratios between 1.7 and 3.3 for recurrent wheezing occurring 3–5 years after RSV-hospitalization during infancy [21]. However, risk ratios as high as 4.3 were estimated for recurrent childhood wheezing after only one year of follow-up [21]. We used RRRW parameter inputs of 1.6, 2.6, and 4.0. For each combination of the parameter estimates (summarized in Table 1), we calculated the risk of recurrent childhood wheezing in each of the placebo and active vaccine arms. From these risks, we calculated RRVW, the corresponding risk ratio for maternal RSV vaccination and recurrent childhood wheezing. These risk ratios are the effect sizes that a trial would be designed to detect, with larger sample sizes needed to detect smaller effect sizes (i.e., RRVW closer to 1). We did not consider the time-varying relationships between gestational timing of maternal vaccination, birth, and RSV seasonality [22], as maternal vaccine RCTs with paediatric RSV endpoints are expected to time vaccination to maximize transplacental antibody transport and births occurring during the RSV season. Parameters used in sample size calculations. To calculate the sample size required to detect a difference between these risks, we performed a two-sided sample size calculation assuming 80% statistical power and 5% type I error, under a 1:1 or a 2:1 vaccine allocation to active vaccine and placebo arms, respectively. These calculations were performed using the power.prop.test function from the stats package, and the bsamsize function from the Hmisc package in R version 3.4.2 [23]. To estimate the number of pregnant women needed to be vaccinated to prevent one case of recurrent childhood wheezing, we calculated the absolute difference in risk of recurrent childhood wheezing in vaccinated and unvaccinated mother-infant pairs (i.e., the number of excess cases per 100 unvaccinated women) and took the inverse of this value. While the parameter inputs reflect a range of evidence-based estimates, some combinations of parameter values resulted in scenarios that were unlikely, particularly when multiple extreme parameters were considered simultaneously. Scenarios that combined a baseline risk of recurrent wheezing of 20.0% with a 4.0-fold increase in risk of recurrent wheezing following a severe early RSV illness were categorized as “least likely” since this would imply that 80% of children who had an early severe RSV illness would later develop recurrent wheezing, which we deemed unreasonably high. Furthermore, scenarios that combined a 2.6-fold or 4.0-fold increase in risk of recurrent wheezing and a 17.0% attack rate of early severe RSV illness were categorized as “less likely” or “least likely”, respectively. This is because these risk ratios estimates were based on the association between RSV-hospitalization and recurrent childhood wheezing, while 17.0% referred to all RSV-LRIs, not just hospitalizations (which is 6 times higher than the estimated RSV-hospitalization rate of 2.7%). All other scenarios were categorized as “more plausible”. All of the statistical code to perform these calculations and reproduce this manuscript is publicly available on GitHub: https://github.com/corinne-riddell/RSV-wheeze-sample-size/. The funders played no role in study design, interpretation of the data, the writing of the report, or the decision to submit the manuscript for publication. The corresponding author had full access to all the data in the study and had final responsibility for the decision to submit for publication.

• riddell, corinne a.
• bhat, niranjan
• bont, louis
• dupont, william d.
• feikin, daniel r.
• fell, deshayne b.
• gebretsadik, tebeb
• hartert, tina v.
• hutcheon, jennifer a.
• karron, ruth a.
• nair, harish
• reiner, robert c.
• shi, ting
• sly, peter d.
• stein, renato t.
• wu, pingsheng
• zar, heather j.
• ortiz, justin r.