Effect of 10-valent pneumococcal conjugate vaccine on the incidence of radiologically-confirmed pneumonia and clinically-defined pneumonia in Kenyan children: an interrupted time-series analysis

Micah Silaba, Michael Ooko, Christian Bottomley, Joyce Sande, Rachel Benamore, Kate Park, James Ignas, Kathryn Maitland, Neema Mturi, Anne Makumi, Mark Otiende, Stanley Kagwanja, Sylvester Safari, Victor Ochola, Tahreni Bwanaali, Evasius Bauni, Fergus Gleeson, Maria Deloria Knoll, Ifedayo Adetifa, Kevin MarshThomas N. Williams, Tatu Kamau, Shahnaaz Sharif, Orin S. Levine, Laura L. Hammitt, J. Anthony G. Scott

Research output: Contribution to journalArticlepeer-review

16 Scopus citations


Background: Pneumococcal conjugate vaccines (PCV) are highly protective against invasive pneumococcal disease caused by vaccine serotypes, but the burden of pneumococcal disease in low-income and middle-income countries is dominated by pneumonia, most of which is non-bacteraemic. We examined the effect of 10-valent PCV on the incidence of pneumonia in Kenya. Methods: We linked prospective hospital surveillance for clinically-defined WHO severe or very severe pneumonia at Kilifi County Hospital, Kenya, from 2002 to 2015, to population surveillance at Kilifi Health and Demographic Surveillance System, comprising 45 000 children younger than 5 years. Chest radiographs were read according to a WHO standard. A 10-valent pneumococcal non-typeable Haemophilus influenzae protein D conjugate vaccine (PCV10) was introduced in Kenya in January, 2011. In Kilifi, there was a three-dose catch-up campaign for infants (aged <1 year) and a two-dose catch-up campaign for children aged 1–4 years, between January and March, 2011. We estimated the effect of PCV10 on the incidence of clinically-defined and radiologically-confirmed pneumonia through interrupted time-series analysis, accounting for seasonal and temporal trends. Findings: Between May 1, 2002 and March 31, 2015, 44 771 children aged 2–143 months were admitted to Kilifi County Hospital. We excluded 810 admissions between January and March, 2011, and 182 admissions during nurses' strikes. In 2002–03, the incidence of admission with clinically-defined pneumonia was 2170 per 100 000 in children aged 2–59 months. By the end of the catch-up campaign in 2011, 4997 (61·1%) of 8181 children aged 2–11 months had received at least two doses of PCV10 and 23 298 (62·3%) of 37 416 children aged 12–59 months had received at least one dose. Across the 13 years of surveillance, the incidence of clinically-defined pneumonia declined by 0·5% per month, independent of vaccine introduction. There was no secular trend in the incidence of radiologically-confirmed pneumonia over 8 years of study. After adjustment for secular trend and season, incidence rate ratios for admission with radiologically-confirmed pneumonia, clinically-defined pneumonia, and diarrhoea (control condition), associated temporally with PCV10 introduction and the catch-up campaign, were 0·52 (95% CI 0·32–0·86), 0·73 (0·54–0·97), and 0·63 (0·31–1·26), respectively. Immediately before PCV10 was introduced, the annual incidence of clinically-defined pneumonia was 1220 per 100 000; this value was reduced by 329 per 100 000 at the point of PCV10 introduction. Interpretation: Over 13 years, admissions to Kilifi County Hospital for clinically-defined pneumonia decreased sharply (by 27%) in association with the introduction of PCV10, as did the incidence of radiologically-confirmed pneumonia (by 48%). The burden of hospital admissions for childhood pneumonia in Kilifi, Kenya, has been reduced substantially by the introduction of PCV10. Funding: Gavi, The Vaccine Alliance and Wellcome Trust.

Original languageEnglish (US)
Pages (from-to)e337-e346
JournalThe Lancet Global Health
Issue number3
StatePublished - Mar 2019

ASJC Scopus subject areas

  • Medicine(all)


Dive into the research topics of 'Effect of 10-valent pneumococcal conjugate vaccine on the incidence of radiologically-confirmed pneumonia and clinically-defined pneumonia in Kenyan children: an interrupted time-series analysis'. Together they form a unique fingerprint.

Cite this