The impact of 10-valent pneumococcal conjugate vaccine on the incidence of admissions to hospital with hypoxaemic and non-hypoxaemic pneumonia in Kenyan children

Observational evidence suggests that pneumococcal conjugate vaccines (PCVs) are more effective at reducing the incidence of hypoxaemic pneumonia compared to non-hypoxaemic pneumonia. We examined the impact of 10-valent PCV (PCV10, Synflorix) on hypoxaemic pneumonia hospital admissions using data from an existing PCV impact study of clinical and radiographic pneumonia. PCV10 was introduced, with catch-up among children under 5 years, in the Kilifi Health and Demographic Surveillance System (KHDSS) in January 2011. We undertook an interrupted time-series (ITS) analysis using seasonally-adjusted segmented Poisson regression of hypoxaemic and non-hypoxaemic pneumonia, accounting for secular trends, from January 2007 to December 2019 among KHDSS residents aged 2–59 months. The median monthly crude incidence of hypoxaemic pneumonia per 100,000 person-years was 95 (IQR 80–139) in the pre-PCV10 period. Time-series regression estimated PCV10 introduction was associated with an increase in the incidence of hypoxaemic pneumonia in children aged 2–59 months (IRR 1.63, 95% CI 1.10–2.41), whereas it was associated with a decrease in the incidence of non-hypoxaemic pneumonia (IRR 0.61, 95% CI 0.48–0.77). Despite the consistent pneumonia surveillance and robust ITS analysis which accounted for pre-PCV10 secular trend and seasonality, the apparent association with hypoxaemic pneumonia is likely due to unmeasured time-varying confounders around the time of vaccine introduction, such as the epidemiology of other respiratory pathogens. This study highlights limitations in the analysis and interpretation of observational data in vaccine impact studies against pneumonia.