Pneumococcal vaccine and serotype replacement in England: the bias of increased reporting. (30th September 2020)

• Record Type: Journal Article
• Authors: Chiavenna, C; Presanis, A; Charlett, A; Ladhani, S; De Angelis, D
• Is Part Of: European journal of public health. Volume 30(2020)Supplement 5
• Publication Date: 2020-09-30
• Subjects: Epidemiology -- Europe -- Periodicals; Public health -- Europe -- Periodicals; 362.109405
• DOI: 10.1093/eurpub/ckaa165.1257
• Languages: English
• ISSNs: 1101-1262

Abstract

Abstract: Background: Increased incidence of invasive pneumococcal disease (IPD) attributable to non-vaccine serotypes (NVT) has been reported in several countries following introduction of PCV7 and PCV13 vaccines, concurrently with a reduction in vaccine-type IPD. Such serotype replacement has, importantly, emerged in England, offsetting the benefit of PCV introduction. We scrutinise most recent findings to assess if the estimated increase in NVT disease might result from surveillance artefacts. Methods: Using IPD surveillance for 2000-2018, we estimate the impact of PCV7 and PCV13 introduction on age-serotype-specific incidence rates through a synthetic control regression model, building counterfactuals by combining age-specific incidences reported for pathogens unaffected by PCVs. Results: Following the introduction of PCV7 and PCV13 (pre-2006 vs post-2011), total IPD incidence declined by 57% and by 76% in children younger than 5. PCV7-IPD decreased by 93% in all age groups, whereas PCV13-IPD declined by 68% since PCV13 was introduced. Importantly, NVT-IPD increased by 43% after PCV7, with non-significant statistical increases in most age groups. Conclusions: Through appropriate statistical modelling, we disentangled the impact of vaccine and improved surveillance on the changes in IPD incidence rates. By controlling for the confounding effects of improved surveillance, we are able to estimate a lower serotype replacement. Key messages: Pneumococcal vaccine has been beneficial despite serotype replacement. Adequate statistical methods are needed to disentangle the two phenomena.