Epidemiological Markers for Interactions Among Streptococcus pneumoniae, Haemophilus influenzae, and Staphylococcus aureus in Upper Respiratory Tract Carriage

Abstract

Background: Cocolonization by Streptococcus pneumoniae and Haemophilus influenzae among children has been noted in numerous studies, as has an inverse relationship involving colonization with these species and Staphylococcus aureus. Interactions among these pathogens could mediate unanticipated outcomes of clinical interventions, including changes in H. influenzae and S. aureus disease incidence following pneumococcal vaccine introduction. However, it remains unclear whether cocolonization patterns represent true interspecies interactions or whether they result from confounding factors.

Methods: We investigated polymicrobial carriage using longitudinal data from 369 Bedouin children and 400 Jewish children in Israel who were enrolled in a 7-valent pneumococcal conjugate vaccine (PCV7) trial. Children were swabbed 10 times between 2 and 30 months of age.

Results: The pathogens followed distinct age and seasonal distributions, but polymicrobial carriage associations persisted after controlling for these and other confounding factors. Receipt of PCV7 resulted in pneumococcal serotype replacement but did not influence total carriage of S. pneumoniae, H. influenzae, or S. aureus.

Conclusions: The fact that S. pneumoniae, H. influenzae, and S. aureus polymicrobial carriage patterns do not result from confounding by age and season supports the idea of active interspecies interactions. However, pneumococcal serotype replacement may prevent changes in H. influenzae and S. aureus carriage among PCV7 recipients.

Keywords: Haemophilus influenzae; Staphylococcus aureus; Streptococcus pneumoniae; pneumococcal conjugate vaccine; serotype replacement.

Figure 1.

Variation in Streptococcus pneumoniae, Haemophilus influenzae, and Staphylococcus aureus colonization, by age, among Jewish and Bedouin children. Top, Plots illustrate variation in carriage odds by age within each ethnic group, with age at which carriage odds were predicted to be highest as a reference (30 months, S. pneumoniae and H. influenzae; 2 months, S. aureus). We identified that a logarithmic trend in carriage odds across ages provided the optimal fit to data (in comparison to linear and quadratic and cubic polynomial trends) via the Bayesian information criterion. Odds ratios (ORs) are estimated via cluster-bootstrapped logistic regression adjusting for seasonal trend, antibiotic receipt, child contacts, and ethnicity. Shaded regions delineate 95% confidence intervals (CIs). Bottom, Plots illustrate prevalence of carriage for each pathogen by study visit age. Error bars delineate 95% CIs.

Figure 2.

Seasonal variation in Streptococcus pneumoniae, Haemophilus influenzae, and Staphylococcus aureus colonization among Bedouin and Jewish children. Top, Plots illustrate variation in carriage odds for each pathogen throughout the calendar year, with carriage during the first week of September as a reference. Seasonal trends are constructed from sine and cosine transformations of calendar week at which sampling occurred, considering 4-, 6-, and 12-month periods of oscillation. Odds ratios (ORs) are estimated via cluster-bootstrapped logistic regression adjusting for age, antibiotic receipt, child contacts, and ethnicity. Shaded regions delineate 95% confidence intervals (CIs). Bottom, Plots illustrate prevalence of carriage for each pathogen, aggregated with a 3-week moving average across calendar weeks. Shaded regions delineate 95% CIs.

Figure 3.

Vaccine and non–vaccine-serotype Streptococcus pneumoniae carriage among Bedouin and Jewish children. Plots illustrate prevalence of vaccine-serotype (top) and non–vaccine­-serotype (bottom) carriage across study visit ages. Children are aggregated across study arms with primary-series schedules (left, “3 + 1” and “3 + 0” for 3-dose primary series at 2, 4, and 6 months with or without booster at 12 months; and “2 + 1” for 2-dose primary series at 4 and 6 months with booster at 12 months) and delayed-dosing schedules (right, “0 + 2” for doses at 12 and 18 months and “0 + 1” for 1 dose at 18 months). Shaded areas delineate 95% confidence intervals estimated by a cluster bootstrap of individuals.