Acquisition and clearance dynamics of Campylobacter spp. in children in low- and middle-income countries

Abstract

The prevalence of Campylobacter infection is generally high among children in low- and middle-income countries (LMIC), but the dynamics of its acquisition and clearance are understudied. We aim to quantify this process among children under two years old in eight LMIC using a statistical modeling approach, leveraging enzyme-immunoassay-based Campylobacter genus data and quantitative-PCR-based Campylobacter jejuni/coli data from the MAL-ED study. We developed a Markov model to compare the dynamics of acquisition and clearance of Campylobacter across countries and to explore the effect of antibiotic usage on Campylobacter clearance. Clearance rates were generally higher than acquisition rates, but their magnitude and temporal pattern varied across countries. For C. jejuni/coli, clearance was faster than acquisition throughout the two years at all sites. For Campylobacter spp., the acquisition rate either exceeded or stayed very close to the clearance rate after the first half year in Bangladesh, Pakistan and Tanzania, leading to high prevalence. Bangladesh had the shortest (28 and 57 days) while Brazil had the longest (328 and 306 days) mean times from last clearance to acquisition for Campylobacter spp. and C. jejuni/coli, respectively. South Africa had the shortest (10 and 8 days) while Tanzania had the longest (53 and 41 days) mean times to clearance for Campylobacter spp. and C. jejuni/col, respectively. The use of Macrolide accelerated clearance of C. jejuni/coli in Bangladesh and Peru and of Campylobacter spp. in Bangladesh and Pakistan. Fluoroquinolone showed statistically meaningful effects only in Bangladesh but for both Campylobacter groups. Higher prevalence of Campylobacter infection was mainly driven by a high acquisition rate that was close to or surpassing the clearance rate. Acquisition rate usually peaked in 11-17 months of age, indicating the importance of targeting the first year of life for effective interventions to reduce exposures.

Keywords: Acquisition; Antibiotic effect; Campylobacter; Clearance; Markov model.

Fig. 1

Site-specific relative frequencies of observed transitions of Campylobacter jejuni/coli (A) and Campylobacter spp. (B) between uncolonized and colonized states during 0–12 months (left column) and 13–24 months (right column) of age. Study sites include: Dhaka, Bangladesh (BGD); Vallore, India (INV); Bhaktapur, Nepal (NEB); Naushero Feroze, Pakistan (PKN); Venda, South Africa (SAV); Haydom, Tanzania (TZH); Fortaleza, Brazil (BRF); Loreto, Peru (PEL).

Fig. 2

Model-estimated site-specific daily acquisition (orange) and clearance (blue) probabilities of Campylobacter jejuni/coli (A) and Campylobacter spp. (B). 95% asymptotic confidence intervals are shown as gray error bands. These estimates are not adjusted for antibiotic use. Study sites include: Dhaka, Bangladesh (BGD); Vallore, India (INV); Bhaktapur, Nepal (NEB); Naushero Feroze, Pakistan (PKN); Venda, South Africa (SAV); Haydom, Tanzania (TZH); Fortaleza, Brazil (BRF); Loreto, Peru (PEL).

Fig. 3

Site-specific effects of the use of macrolides and fluoroquinolones during one week prior to specimen collection on the clearances of Campylobacter jejuni/coli and Campylobacter spp. Point estimates and 95% confidence intervals of the relative log time to clearance (RLTC) were shown, which is defined as the ratio of the log time to clearance with antibiotic use to the log time to clearance without. Study sites include: Dhaka, Bangladesh (BGD); Vallore, India (INV); Bhaktapur, Nepal (NEB); Naushero Feroze, Pakistan (PKN); Venda, South Africa (SAV); Haydom, Tanzania (TZH); Fortaleza, Brazil (BRF); Loreto, Peru (PEL).

Fig. 4

Site-specific observed and simulated longitudinal trends of prevalence of Campylobacter jejuni/coli (A) and Campylobacter spp. (B) among all and non-diarrheal samples. The simulated longitudinal trends were averaged from 100 realizations generated using fundamental parameters estimated from the corresponding observed data.