Analysis of multivariate longitudinal immuno-epidemiological data using a pairwise joint modelling approach

Background

Immuno-epidemiologists are often faced with multivariate outcomes, measured repeatedly over time. Such data are characterised by complex inter- and intra-outcome relationships which must be accounted for during analysis. Scientific questions of interest might include determining the effect of a treatment on the evolution of all outcomes together, or grouping outcomes that change in the same way. Modelling the different outcomes separately may not be appropriate because it ignores the underlying relationships between outcomes. In such situations, a joint modelling strategy is necessary. This paper describes a pairwise joint modelling approach and discusses its benefits over more simple statistical analysis approaches, with application to data from a study of the response to BCG vaccination in the first year of life, conducted in Entebbe, Uganda.

Methods

The study aimed to determine the effect of maternal latent Mycobacterium tuberculosis infection (LTBI) on infant immune response (TNF, IFN-γ, IL-13, IL-10, IL-5, IL-17A and IL-2 responses to PPD), following immunisation with BCG. A simple analysis ignoring the correlation structure of multivariate longitudinal data is first shown. Univariate linear mixed models are then used to describe longitudinal profiles of each outcome, and are then combined into a multivariate mixed model, specifying a joint distribution for the random effects to account for correlations between the multiple outcomes. A pairwise joint modelling approach, where all possible pairs of bivariate mixed models are fitted, is then used to obtain parameter estimates.

Results

Univariate and pairwise longitudinal analysis approaches are consistent in finding that LTBI had no impact on the evolution of cytokine responses to PPD. Estimates from the pairwise joint modelling approach were more precise. Major advantages of the pairwise approach include the opportunity to test for the effect of LTBI on the joint evolution of all, or groups of, outcomes and the ability to estimate association structures of the outcomes.

Conclusions

The pairwise joint modelling approach reduces the complexity of analysis of high-dimensional multivariate repeated measures, allows for proper accounting for association structures and can improve our understanding and interpretation of longitudinal immuno-epidemiological data.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12865-021-00453-5.

Maternal BCG scar is associated with increased infant proinflammatory immune responses

Introduction

Prenatal exposures such as infections and immunisation may influence infant responses. We had an opportunity to undertake an analysis of innate responses in infants within the context of a study investigating the effects of maternal mycobacterial exposures and infection on BCG vaccine-induced responses in Ugandan infants.

Material and methods

Maternal and cord blood samples from 29 mother-infant pairs were stimulated with innate stimuli for 24 h and cytokines and chemokines in supernatants were measured using the Luminex® assay. The associations between maternal latent Mycobacterium tuberculosis infection (LTBI), maternal BCG scar (adjusted for each other’s effect) and infant responses were examined using linear regression. Principal Component Analysis (PCA) was used to assess patterns of cytokine and chemokine responses. Gene expression profiles for pathways associated with maternal LTBI and with maternal BCG scar were examined using samples collected at one (n = 42) and six (n = 51) weeks after BCG immunisation using microarray.

Results

Maternal LTBI was positively associated with infant IP-10 responses with an adjusted geometric mean ratio (aGMR) [95% confidence interval (CI)] of 5.10 [1.21, 21.48]. Maternal BCG scar showed strong and consistent associations with IFN-γ (aGMR 2.69 [1.15, 6.17]), IL-12p70 (1.95 [1.10, 3.55]), IL-10 (1.82 [1.07, 3.09]), VEGF (3.55 [1.07, 11.48]) and IP-10 (6.76 [1.17, 38.02]). Further assessment of the associations using PCA showed no differences for maternal LTBI, but maternal BCG scar was associated with higher scores for principal component (PC) 1 (median level of scores: 1.44 in scar-positive versus −0.94 in scar-negative, p = 0.020) in the infants. PC1 represented a controlled proinflammatory response. Interferon and inflammation response pathways were up-regulated in infants of mothers with LTBI at six weeks, and in infants of mothers with a BCG scar at one and six weeks after BCG immunisation.

Conclusions

Maternal BCG scar had a stronger association with infant responses than maternal LTBI, with an increased proinflammatory immune profile.

Serologic diagnosis of tuberculosis by combining Ig classes against selected mycobacterial targets

INTRODUCTION

Accurate, simple and cost-effective diagnostic tests are needed for diagnosis of active tuberculosis (TB). Serodiagnosis is attractive as it can be harnessed for point-of-care tests.

METHODS

We evaluated, in a blinded fashion, the sensitivity and specificity of serologic immunoglobulin (Ig)G, IgA and/or IgM responses to Apa, heat shock protein (HSP) 16.3, HSP20, PE35, probable thiol peroxidase Tpx and lipoarabinomannan (LAM) in 42 HIV-negative South African pulmonary TB patients and 67 control individuals. The status of latent Mycobacterium tuberculosis infection (LTBI) among controls was defined through the TST and IFN-γ release assays (IGRAs). We evaluated 47 definite LTBI (IGRA(+)/LTBI), 8 putative LTBI (IGRA(-)/TST(+)) and 12 TB-uninfected (non-LTBI) subjects.

RESULTS

In contrast to anti-PE35 IgA, anti-PE35 IgG and particularly anti-Apa IgA, performances of anti-LAM IgG and selected anti-protein antibodies were less affected by inclusion of LTBI participants into the analysis. Anti-LAM IgG showed with a sensitivity/specificity of 71.4%/86.6% (p < 0.001) the best discrimination between TB and non-TB subjects. Selected five-antibody-combinations (including anti-LAM IgG, anti-LAM IgA and anti-Tpx IgG) further improved this performance to an accuracy exceeding 86%.

CONCLUSIONS

Antibody responses to some Mycobacterium tuberculosis antigens often also reflect latent infection explaining the poor performance of antibody-based tests for active TB in TB-endemic settings. Our results suggest that rather a combination of serological responses against selected protein and non-protein antigens and different Ig classes should be investigated for TB serodiagnostics.