Results:

For non-human immunodeficiency virus (HIV)-infected children, the VE estimates for C-LRTI with CRP levels of≥ 40 mg/dL (VE 26.3%; P= 0.003) or CRP levels of≥ 120 mg/dL (VE 41.0%; P= 0.003) were 1.7-fold (P= 0.002) and 2.7-fold (P< 0.0001) greater, respectively, than that for CXR-AC (VE 15.1%; P= 0.15). The sensitivity of CXR-AC as an outcome to detect the burden of pneumonia prevented by PCV was 44%(95% confidence interval, 36–55%) in comparison with C-LRTI with CRP levels of≥ 40 mg/dL and 73%(95% confidence interval, 58–92%) in comparison with C-LRTI with CRP levels of≥ 120 mg/dL. CRP also helped to measure the PCV efficacy for children with C-LRTI but the absence of CXR-AC, for whom the outcome of C-LRTI with CRP levels of≥ 40 mg/dL (VE 31.5%; P= 0.007) increased the VE estimate 19.8-fold (P< 0.0001) in comparison with C-LRTI alone (VE 1.6%; P= 0.78) and 3.2-fold (P= 0.005) in comparison with WHO-defined severe pneumonia (VE 10.0%; P= 0.17). Although there was a significant correlation between CRP and procalcitonin levels (Spearman's ρ= 0.45; P< 0.0001), the use of procalcitonin levels did not improve either the specificity or sensitivity of measuring the effect of PCV against pneumonia for non-HIV-infected children. The observations were similar for HIV-infected children.

Conclusions:

CRP levels of≥ 40 mg/dL provide a better measure than chest radiographs to assess the effect of PCV in preventing pneumonia.