Population attributable risk percentage (PAR%) estimates the proportion of disease outcomes in a population that can be attributed to a risk factor which was developed using univariable models. However, clinical and epidemiological research is often multifactorial, particularly in infectious disease settings, where several factors may impact the outcome of interest which requires methods that can account for their correlation structure. We developed a generalisable methodology to estimate both full and partial PAR% using multivariable logistic and Cox regression models. The approach incorporated delta-method variance estimation and Fisher’s Z transformation to derive the 95 % confidence intervals after accounting for their correlation structure. Methodology was implemented by developing a SAS macro applicable to both cross-sectional and longitudinal designs. This methodology was applied to nationally representative data from South Africa’s COVID-19 Vaccine Surveys (CVACS-1 and CVACS-2) conducted during the 2021–2022 vaccination rollout to estimate the population-level impacts of various beliefs and attitudes on vaccine uptake. After adjusting for demographic and socioeconomic covariates, we examined the population-level impact of vaccine-related beliefs and behaviours on two outcomes: receipt of at least one COVID-19 vaccine dose and intention to receive a booster. The belief “I intend to get vaccinated” had the highest partial PAR% for uptake (62 %), followed by belief in vaccine effectiveness (43 %) and safety (34 %). The combined impact of all factors, i.e. full PAR%s, were 96 % (95 % CI: 89 %, 99 %) and 74 % (95 % CI: 67 %, 80 %) for on developing trust and intention to receive booster, respectively. These estimates highlight the significant contribution of key behavioural factors and the need for model-based estimation, as simple addition of partial PAR%s would result in overestimation. This study presented a methodological framework for deriving full and partial PAR% in multifactorial infectious disease settings. By combining advanced statistical techniques with real-world data and developing a SAS macro, we provided an epidemiological tool using advanced statistical methodologies for evaluating modifiable risk factors and informing targeted interventions in public health.
Objectives The approach of using HIV recency assay to estimate the counterfactual incidence rate is being used as the primary efficacy method in a few ongoing large-scale HIV pre-exposure prophylaxis (PrEP) trials, and the current available approach for the inference is based on the Wald method that leverages the asymptotic distribution of the estimators. One issue with the Wald test is that it does not work well when the number of HIV infections are small in the active arm, and it fails to work when there are zero HIV infections. As future long-acting PrEP products are becoming more efficacious, it is very likely that a small or zero number of infections will be observed in HIV prevention trials, especially for subgroup analyses or interim analyses, hence there is a pressing need to develop inference methods that work under such scenarios. Methods It is well known that when the sample size is small to moderate, likelihood ratio tests are more reliable than Wald tests in terms of actual error probabilities coming close to matching nominal levels. In this manuscript we derive the likelihood ratio test and the likelihood-based confidence intervals for HIV prevention trials based on recency assays. Results Compared with the Wald test, the proposed method works when there are zero infections. Additionally, unlike the Wald test, the p-value from the likelihood ratio test is an increasing function with respect to the number of infections, which is a desirable property as otherwise it will cause confusions. Conclusions For HIV PrEP trials based on recency assay, the likelihood-based p-value and confidence interval can be preferable to the Wald based inference methods when the number of HIV infections is expected to be small.
Objectives:Vigorous discussions are ongoing about future efficacy trial designs of candidate human immunodeficiency virus (HIV) prevention interventions. The study design challenges of HIV prevention interventions are considerable given rapid evolution of the prevention landscape and evidence of multiple modalities of highly effective products; future trials will likely be 'active-controlled', i.e., not include a placebo arm. Thus, novel design approaches are needed to accurately assess new interventions against these highly effective active controls. Methods:To discuss active control design challenges and identify solutions, an initial virtual workshop series was hosted and supported by the International AIDS Enterprise (October 2020-March 2021). Subsequent symposia discussions continue to advance these efforts. As the non-inferiority design is an important conceptual reference design for guiding active control trials, we adopt several of its principles in our proposed design approaches. Results:We discuss six potential study design approaches for formally evaluating absolute prevention efficacy given data from an active-controlled HIV prevention trial including using data from: 1) a registrational cohort, 2) recency assays, 3) an external trial placebo arm, 4) a biomarker of HIV incidence/exposure, 5) an anti-retroviral drug concentration as a mediator of prevention efficacy, and 6) immune biomarkers as a mediator of prevention efficacy. Conclusions:Our understanding of these proposed novel approaches to future trial designs remains incomplete and there are many future statistical research needs. Yet, each of these approaches, within the context of an active-controlled trial, have the potential to yield reliable evidence of efficacy for future biomedical interventions.
Randomization inference is a powerful tool in early phase vaccine trials when estimating the causal effect of a regimen against a placebo or another regimen. Randomization-based inference often focuses on testing either Fisher's sharp null hypothesis of no treatment effect for any participant or Neyman's weak null hypothesis of no sample average treatment effect. Many recent efforts have explored conducting exact randomization-based inference for other summaries of the treatment effect profile, for instance, quantiles of the treatment effect distribution function. In this article, we systematically review methods that conduct exact, randomization-based inference for quantiles of individual treatment effects (ITEs) and extend some results to a special case where naïve participants are expected not to exhibit responses to highly specific endpoints. These methods are suitable for completely randomized trials, stratified completely randomized trials, and a matched study comparing two non-randomized arms from possibly different trials. We evaluate the usefulness of these methods using synthetic data in simulation studies. Finally, we apply these methods to HIV Vaccine Trials Network Study 086 (HVTN 086) and HVTN 205 and showcase a wide range of application scenarios of the methods. R code that replicates all analyses in this article can be found in first author's GitHub page at https://github.com/Zhe-Chen-1999/ITE-Inference.
Abstract Objectives An exceptional effort by the scientific community has led to the development of multiple vaccines against COVID-19. Efficacy estimates for these vaccines have been widely communicated to the general public, but are nonetheless challenging to compare because they are based on phase 3 trials that differ in study design, definition of vaccine efficacy and the handling of cases arising shortly after vaccination. We investigate the impact of these choices on vaccine efficacy estimates, both theoretically and by re-analyzing the Janssen and Pfizer COVID-19 trial data under a uniform protocol. We moreover study the causal interpretation that can be assigned to per-protocol analyses typically performed in vaccine trials. Finally, we propose alternative estimands to measure the intrinsic vaccine efficacy in settings with delayed immune response. Methods The data of the Janssen COVID-19 trials were recreated, based on the published Kaplan-Meier curves. An estimator for the alternative causal estimand was developed using a Structural Distribution Model. Results In the data analyses, we observed rather large differences between intention-to-treat and per-protocol effect estimates. In contrast, the causal estimand and the different estimators used for per-protocol effects lead approximately to the same estimates. Conclusions In these COVID-10 vaccine trials, per-protocol effects can be interpreted as the number of cases that can be avoided by vaccination, if the vaccine would immediately induce an immune response. However, it is unclear whether this interpretation also holds in other settings.
