Browsing by Author "Sempa, Joseph B."

Now showing 1 - 3 of 3
  • Thumbnail Image
    Item
    Cumulative viral load as a predictor of CD4+ Tcell response to antiretroviral therapy using Bayesian statistical models
    (PLoS, 2019-11-13) Sempa, Joseph B.; Rossouw, Theresa M.; Lesaffre, Emmanuel; Nieuwoudt, Martin
    Introduction: There are Challenges in statistically modelling immune responses to longitudinal HIV viral load exposure as a function of covariates. We define Bayesian Markov Chain Monte Carlo mixed effects models to incorporate priors and examine the effect of different distributional assumptions. We prospectively fit these models to an as-yet-unpublished data from the Tshwane District Hospital HIV treatment clinic in South Africa, to determine if cumulative log viral load, an indicator of long-term viral exposure, is a valid predictor of immune response. Methods: Models are defined, to express ‘slope’, i.e. mean annual increase in CD4 counts, and ‘asymptote’, i.e. the odds of having a CD4 count ≥500 cells/μL during antiretroviral treatment, as a function of covariates and random-effects. We compare the effect of using informative versus non-informative prior distributions on model parameters. Models with cubic splines or Skew-normal distributions are also compared using the conditional Deviance Information Criterion. Results: The data of 750 patients are analyzed. Overall, models adjusting for cumulative log viral load provide a significantly better fit than those that do not. An increase in cumulative log viral load is associated with a decrease in CD4 count slope (19.6 cells/μL (95% credible interval: 28.26, 10.93)) and a reduction in the odds of achieving a CD4 counts ≥500 cells/μL (0.42 (95% CI: 0.236, 0.730)) during 5 years of therapy. Using informative priors improves the cumulative log viral load estimate, and a skew-normal distribution for the random-intercept and measurement error results is a better fit compared to using classical Gaussian distributions. Discussion: We demonstrate in an unpublished South African cohort that cumulative log viral load is a strong and significant predictor of both CD4 count slope and asymptote. We argue that Bayesian methods should be used more frequently for such data, given their flexibility to incorporate prior information and non-Gaussian distributions.
  • Thumbnail Image
    Item
    Performance comparison of the Maxim and Sedia Limiting Antigen Avidity assays for HIV incidence surveillance
    (Public Library of Science, 2019-07-26) Sempa, Joseph B.; Welte, Alex; Busch, Michael P.; Hall, Jake; Hampton, Dylan; Facente, Shelley N.; Keating, Sheila M.; Marson, Kara; Parkin, Neil; Pilcher, Christopher D.; Murphy, Gary; Grebe, Eduard
    Background: Two manufacturers, Maxim Biomedical and Sedia Biosciences Corporation, supply CDC-approved versions of the HIV-1 Limiting Antigen Avidity EIA (LAg) for detecting ‘recent’ HIV infection in cross-sectional incidence estimation. This study assesses and compares the performance of the two assays for incidence surveillance. Methods: We ran both assays on a panel of 2,500 well-characterized HIV-1-infected specimens. We analysed concordance of assay results, assessed reproducibility using repeat testing and estimated mean durations of recent infection (MDRIs) and false-recent rates (FRRs) for a range of normalized optical density (ODn) thresholds, alone and in combination with viral load thresholds. We defined three hypothetical surveillance scenarios, similar to the Kenyan and South African epidemics, and a concentrated epidemic. These scenarios allowed us to evaluate the precision of incidence estimates obtained by means of various recent infection testing algorithms (RITAs) based on each of the two assays. Results: The Maxim assay produced lower ODn values than the Sedia assay on average, largely as a result of higher calibrator readings (mean OD of 0.749 vs. 0.643), with correlation of normalized readings lower (R2 = 0.908 vs. R2 = 0.938). Reproducibility on blinded control specimens was slightly better for Maxim. The MDRI of a Maxim-based algorithm at the ‘standard’ threshold (ODn ≤1.5 & VL >1,000) was 201 days (95% CI: 180,223) and for Sedia 171 (152,191). The difference Differences in MDRI were estimated at 32.7 (22.9,42.8) and 30.9 days (21.7,40.7) for the two algorithms, respectively. Commensurately, the Maxim algorithm had a higher FRR in treatment-naive subjects (1.7% vs. 1.1%). The two assays produced similar precision of incidence estimates in the three surveillance scenarios. Conclusions: Differences between the assays can be primarily attributed to the calibrators supplied by the manufacturers. Performance for surveillance was extremely similar, although different thresholds were optimal (i.e. produced the lowest variance of incidence estimates) and at any given ODn threshold, different estimates of MDRI and FRR were obtained. The two assays cannot be treated as interchangeable: assay and algorithm-specific performance characteristic estimates must be used for survey planning and incidence estimation.
  • Thumbnail Image
    Item
    Systematic review of statistically-derived models of immunological response in HIV-infected adults on antiretroviral therapy in Sub-Saharan Africa
    (Public Library of Science, 2017-02-15) Sempa, Joseph B.; Ujeneza, Eva L.; Nieuwoudt, Martin
    Introduction: In Sub-Saharan African (SSA) resource limited settings, Cluster of Differentiation 4 (CD4) counts continue to be used for clinical decision making in antiretroviral therapy (ART). Here, HIV-infected people often remain with CD4 counts <350 cells/μL even after 5 years of viral load suppression. Ongoing immunological monitoring is necessary. Due to varying statistical modeling methods comparing immune response to ART across different cohorts is difficult. We systematically review such models and detail the similarities, differences and problems. Methods: ‘Preferred Reporting Items for Systematic Review and Meta-Analyses’ guidelines were used. Only studies of immune-response after ART initiation from SSA in adults were included. Data was extracted from each study and tabulated. Outcomes were categorized into 3 groups: ‘slope’, ‘survival’, and ‘asymptote’ models. Wordclouds were drawn wherein the frequency of variables occurring in the reviewed models is indicated by their size and color. Results: 69 covariates were identified in the final models of 35 studies. Effect sizes of covariates were not directly quantitatively comparable in view of the combination of differing variables and scale transformation methods across models. Wordclouds enabled the identification of qualitative and semi-quantitative covariate sets for each outcome category. Comparison across categories identified sex, baseline age, baseline log viral load, baseline CD4, ART initiation regimen and ART duration as a minimal consensus set. Conclusion: Most models were different with respect to covariates included, variable transformations and scales, model assumptions, modelling strategies and reporting methods, even for the same outcomes. To enable comparison across cohorts, statistical models would benefit from the application of more uniform modelling techniques. Historic efforts have produced results that are anecdotal to individual cohorts only. This study was able to define ‘prior’ knowledge in the Bayesian sense. Such information has value for prospective modelling efforts.