Andreas Andersen

In this article, we illustrate how to use mi impute chained with intreg to fit an analysis of covariance analysis of censored and nondetectable immunological concentrations measured in a randomized pretest–posttest design.

Randomised trials have shown that early Bacille Calmette-Guérin (BCG) vaccine reduces overall neonatal and infant mortality. However, no study has examined how BCG affects growth. We investigated the effect on infant growth of early BCG vaccine given to low-birth-weight (LBW) infants. Two-thousand three hundred forty-three LBW infants were randomly allocated 1:1 to “early BCG” (intervention group) or “late BCG” (current practice). Furthermore, a subgroup (N = 1717) were included in a two-by-two randomised trial in which they were additionally randomised 1:1 to vitamin A supplementation (VAS) or placebo. Anthropometric measurements were obtained 2, 6, and 12 months after enrolment. Overall there was no effect of early BCG on growth in the first year of life. The effect of early BCG on weight and mid-upper-arm circumference at 2 months tended to be beneficial among girls but not among boys (interaction between “early BCG” and sex: weight p = 0.03 and MUAC p = 0.04). This beneficial effect among girls was particularly seen among the largest infants weighing 2.0 kg or more at inclusion. Though BCG vaccination is not recommended to be given to LBW infants at birth in Guinea-Bissau, early BCG had no negative effect on infant growth and may have had a beneficial effect for girls. ClinicalTrials.gov (NCT00146302) .

The mechanisms behind heterologous immunity and non-specific effects of vaccines on mortality are not well understood. We examined associations between cytokine responses and subsequent mortality in low-birth-weight infants in Guinea-Bissau. A low-birth-weight trial randomized children to Bacille Calmette-Guérin (BCG) at birth or later according to local policy. Blood samples were obtained from a sub-group at age 6 weeks. Interleukin (IL)-5, IL-10, IL-13, interferon (IFN)-γ, and tumor necrosis factor (TNF)-α were measured in whole-blood cell cultures stimulated with lipopolysaccharide (LPS), phytohaemagglutinin (PHA), or purified protein derivative (PPD). The outcome was mortality between bleeding and 1 year of age. Non-linear associations between cytokine responses and mortality were examined. Cytokine measurements were available from 390 children. The mortality rate (MR) was high (6.8/100 person-years-observation (PYO)). Both low and high cytokine responses to LPS and PHA were associated with high mortality (MR up to 25/100 PYO in the lowest 10% and 9.2/100 PYO in the highest 10%). In BCG-vaccinated children, higher IFN-γ responses to PPD were associated with better survival (MR ratio = 0.43 (0.24–0.77)). Data presented a rare opportunity to explore associations between cytokine responses and mortality. Both low and high cytokine responses were associated with high mortality; a balanced response to invading pathogens seems preferable.

The World Health Organization (WHO) classifies Guinea-Bissau as having severe vitamin A deficiency (VAD). To date, no national survey has been conducted. We assessed vitamin A status among children in rural Guinea-Bissau to assess status and identify risk factors for VAD.

Interest in cytokines as markers for the function of the immune system is increasing. Methods quantifying cytokine concentrations are often subject to detection limits, which lead to non‐detectable observations and censored distributions. When distributions are skewed, geometric mean ratios (GMRs) can be used to describe the relative concentration between two cytokines, and the GMR ratio (GMRR) can be used to compare two groups. The problem is how to estimate GMRRs from censored distributions.We evaluated methods, including simple deletion and substitution, in simulated and real data. One method applies Tobit directly to the censored difference between the two cytokine log‐concentrations (Diff). However, censoring is correlated to the outcome and is therefore not independent. The correlation increases as the correlation between the two log‐concentrations decreases. We propose a Tobit stacking method that uses clustered variance–covariance estimation allowing homogeneous (Stackc) or inhomogeneous (Stackh) variances. We compare it with direct estimation of the bivariate Tobit likelihood function (Bitobit) and multiple imputation. We assess sensitivity to inhomogeneity and non‐normality. Simulations show that deletion and substitution are empirically biased and that Diff has an empirical bias, which increases as the correlation between the log‐concentrations decreases. Estimates from multiple imputation, Stackh and Bitobit are almost identical. The estimates exhibit small empirical bias for both homogeneous and inhomogeneous normal distributions. For skewed mixture and heavy‐tailed distributions, they perform reasonably well if censoring is less than 30%. We recommend these methods to estimate GMRRs. At least one of the methods is available in Stata, R or SAS. Copyright © 2012 John Wiley & Sons, Ltd.

Plasma levels of the inflammatory biomarker soluble urokinase plasminogen activator (suPAR) have been shown to carry prognostic information in various infectious and inflammatory diseases. The present study aimed to compare the prognostic value of urine suPAR (U-suPAR) to that of plasma suPAR (P-suPAR), thereby exploring the possibility of replacing the blood sample with an easy obtainable urine sample. We enrolled 1,007 adults, older than 15 years of age, with a negative TB diagnosis between April 2004 and December 2006. Levels of U-suPAR and P-suPAR were available in 863 individuals. U-suPAR was measured using a commercial ELISA (suPARnostic®). We found that U-suPAR carried significant prognostic information on mortality for HIV-infected subjects with an area under the ROC curve of 0.75. For HIV-negative individuals, little or no prognostic effect was observed. However, in both HIV positives and negatives, the predictive effect of U-suPAR was found to be inferior to that of P-suPAR.