Inferring Pathogen Type Interactions Using Cross-sectional Prevalence Data: Opportunities and Pitfalls for Predicting Type Replacement

Pattern of multiple human papillomavirus infection and type competition: An analysis in healthy Chinese women aged 18-45 years

To assess the pattern of multiple human papillomavirus infection to predict the type replacement postvaccination. A total of 7372 women aged 18-45y from a phase III trial of an Escherichia coli-produced HPV-16/18 vaccine were analyzed at enrollment visit before vaccination. Hierarchical multilevel logistic regression was used to evaluate HPV vaccine type and nonvaccine-type interactions with age as a covariate. Binary logistic regression was construed to compare multiple infections with single infections to explore the impact of multiple-type infections on the risk of cervical disease. Multiple HPV infections were observed in 25.2% of HPV-positive women and multiple infections were higher than expected by chance. Statistically significant negative associations were observed between HPV16 and 52, HPV18 and HPV51/52/58, HPV31 and HPV39/51/52/53/54/58, HPV33 and HPV52/58, HPV58 and HPV52, HPV6 and HPV 39/51/52/53/54/56/58. Multiple HPV infections increased the risk of CIN2+ and HSIL+, with the ORs of 2.27(95%CI: 1.41, 3.64) and 2.26 (95%CI: 1.29, 3.95) for multiple oncogenic HPV infection separately. However, no significant evidence for the type-type interactions on risk of CIN2+ or HSIL+. There is possibility of type replacement between several pairs of vaccine and nonvaccine HPV type. Multiple HPV infection increased the risk of cervical disease, but coinfection HPV types seem to follow independent disease processes. Continued post-vaccination surveillance for HPV 51/52/58 types and HPV 39/51 types separately was essential after the first and second generation of HPV vaccination implementation in China.

Considerations for viral co-infection studies in human populations

When respiratory viruses co-circulate in a population, individuals may be infected with multiple pathogens and experience possible virus-virus interactions, where concurrent or recent prior infection with one virus affects the infection process of another virus. While experimental studies have provided convincing evidence for within-host mechanisms of virus-virus interactions, evaluating evidence for viral interference or potentiation using population-level data has proven more difficult. Recent studies have quantified the prevalence of co-detections using populations drawn from clinical settings. Here, we focus on selection bias issues associated with this study design. We provide a quantitative account of the conditions under which selection bias arises in these studies, review previous attempts to address this bias, and propose unbiased study designs with sample size estimates needed to ascertain viral interference. We show that selection bias is expected in cross-sectional co-detection prevalence studies conducted in clinical settings, except under a strict set of assumptions regarding the relative probabilities of being included in a study limited to individuals with clinical disease under different viral states. Population-wide studies that collect samples from participants irrespective of their clinical status would meanwhile require large sample sizes to be sufficiently powered to detect viral interference, suggesting that a study's timing, inclusion criteria, and the expected magnitude of interference are instrumental in determining feasibility.

Reconstructing multi-strain pathogen interactions from cross-sectional survey data via statistical network inference

Infectious diseases often involve multiple pathogen species or multiple strains of the same pathogen. As such, knowledge of how different pathogens interact is key to understand and predict the outcome of interventions targeting only a subset of species or strains involved in disease. Population-level data may be useful to infer pathogen strain interactions, but most previously used inference methods only consider uniform interactions between all strains or focus on marginal pairwise interactions. As such, these methods are prone to bias induced by indirect interactions through other strains. Here, we evaluated statistical network inference for reconstructing heterogeneous interactions from cross-sectional surveys detecting joint presence/absence patterns of pathogen strains within hosts. We applied various network models to simulated survey data, representing endemic infection states of multiple pathogen strains with potential interactions in acquisition or clearance of infection. Satisfactory performance was demonstrated by the estimators converging to the true interactions. Accurate reconstruction of interaction networks was achieved by regularization or penalization for sample size. Although performance deteriorated in the presence of host heterogeneity, this was overcome by correcting for individual-level risk factors. Our work demonstrates how statistical network inference could prove useful for detecting multi-strain pathogen interactions and may have applications beyond epidemiology.

The interactions of SARS-CoV-2 with cocirculating pathogens: Epidemiological implications and current knowledge gaps

Despite the availability of effective vaccines, the persistence of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) suggests that cocirculation with other pathogens and resulting multiepidemics (of, for example, COVID-19 and influenza) may become increasingly frequent. To better forecast and control the risk of such multiepidemics, it is essential to elucidate the potential interactions of SARS-CoV-2 with other pathogens; these interactions, however, remain poorly defined. Here, we aimed to review the current body of evidence about SARS-CoV-2 interactions. Our review is structured in four parts. To study pathogen interactions in a systematic and comprehensive way, we first developed a general framework to capture their major components: sign (either negative for antagonistic interactions or positive for synergistic interactions), strength (i.e., magnitude of the interaction), symmetry (describing whether the interaction depends on the order of infection of interacting pathogens), duration (describing whether the interaction is short-lived or long-lived), and mechanism (e.g., whether interaction modifies susceptibility to infection, transmissibility of infection, or severity of disease). Second, we reviewed the experimental evidence from animal models about SARS-CoV-2 interactions. Of the 14 studies identified, 11 focused on the outcomes of coinfection with nonattenuated influenza A viruses (IAVs), and 3 with other pathogens. The 11 studies on IAV used different designs and animal models (ferrets, hamsters, and mice) but generally demonstrated that coinfection increased disease severity compared with either monoinfection. By contrast, the effect of coinfection on the viral load of either virus was variable and inconsistent across studies. Third, we reviewed the epidemiological evidence about SARS-CoV-2 interactions in human populations. Although numerous studies were identified, only a few were specifically designed to infer interaction, and many were prone to multiple biases, including confounding. Nevertheless, their results suggested that influenza and pneumococcal conjugate vaccinations were associated with a reduced risk of SARS-CoV-2 infection. Finally, fourth, we formulated simple transmission models of SARS-CoV-2 cocirculation with an epidemic viral pathogen or an endemic bacterial pathogen, showing how they can naturally incorporate the proposed framework. More generally, we argue that such models, when designed with an integrative and multidisciplinary perspective, will be invaluable tools to resolve the substantial uncertainties that remain about SARS-CoV-2 interactions.

Approximate likelihood-based estimation method of multiple-type pathogen interactions: An application to longitudinal pneumococcal carriage data

While the serotypes of Streptococcus pneumoniae are known to compete during colonization in human hosts, our knowledge of how competition occurs is still incomplete. New insights of pneumococcal between‐type competition could be generated from carriage data obtained by molecular‐based detection methods, which record more complete sets of serotypes involved in co‐carriage than when detection is done by culture. Here, we develop a Bayesian estimation method for inferring between‐type interactions from longitudinal data recording the presence/absence of the types at discrete observation times. It allows inference from data containing co‐carriage of two or more serotypes, which is often the case when pneumococcal presence is determined by molecular‐based methods. The computational burden posed by the increased number of types detected in co‐carriage is addressed by approximating the likelihood under a multi‐state model with the likelihood of only those trajectories with minimum number of acquisition and clearance events between observation times. The proposed method's performance was validated on simulated data. The estimates of the interaction parameters of acquisition and clearance were unbiased in settings with short sampling intervals between observation times. With less frequent sampling, the estimates of the interaction parameters became more biased, but their ratio, which summarizes the total interaction, remained unbiased. Confounding due to unobserved heterogeneity in exposure could be corrected by including individual‐level random effects. In an application to empirical data about pneumococcal carriage in infants, we found new evidence for between‐serotype competition in clearance, although the effect size was small.

The pitfalls of inferring virus-virus interactions from co-detection prevalence data: application to influenza and SARS-CoV-2

There is growing experimental evidence that many respiratory viruses-including influenza and SARS-CoV-2-can interact, such that their epidemiological dynamics may not be independent. To assess these interactions, standard statistical tests of independence suggest that the prevalence ratio-defined as the ratio of co-infection prevalence to the product of single-infection prevalences-should equal unity for non-interacting pathogens. As a result, earlier epidemiological studies aimed to estimate the prevalence ratio from co-detection prevalence data, under the assumption that deviations from unity implied interaction. To examine the validity of this assumption, we designed a simulation study that built on a broadly applicable epidemiological model of co-circulation of two emerging or seasonal respiratory viruses. By focusing on the pair influenza-SARS-CoV-2, we first demonstrate that the prevalence ratio systematically underestimates the strength of interaction, and can even misclassify antagonistic or synergistic interactions that persist after clearance of infection. In a global sensitivity analysis, we further identify properties of viral infection-such as a high reproduction number or a short infectious period-that blur the interaction inferred from the prevalence ratio. Altogether, our results suggest that ecological or epidemiological studies based on co-detection prevalence data provide a poor guide to assess interactions among respiratory viruses.

Concurrent infection with multiple human papillomavirus types among unvaccinated and vaccinated 17-year-old Norwegian girls

Background

Whether type-specific human papillomavirus (HPV) infection influences the risk of acquiring infections with other HPV types is unclear. We studied concurrent HPV infections in 17-year-old girls from 2 birth cohorts; the first vaccine-eligible cohort in Norway and a prevaccination cohort.

Methods

Urine samples were collected and tested for 37 HPV genotypes. This study was restricted to unvaccinated girls from the prevaccination cohort (n = 5245) and vaccinated girls from the vaccine-eligible cohort (n = 4904). Risk of HPV infection was modelled using mixed-effect logistic regression. Expected frequencies of concurrent infection with each pairwise combination of the vaccine types and high-risk types (6/11/16/18/31/33/35/39/45/51/52/56/58/59) were compared to observed frequencies.

Results

Infection with multiple HPV types was more common among unvaccinated girls than vaccinated girls (9.2% vs 3.7%). HPV33 and HPV51 was the only HPV pair that was detected together more often than expected among both unvaccinated (P = .002) and vaccinated girls (P < .001). No HPV pairs were observed significantly less often than expected.

Conclusions

HPV33 and HPV51 tended to be involved in coinfection among both unvaccinated and vaccinated girls. The introduction of HPV vaccination does not seem to have had an effect on the tendency of specific HPV types to cluster together.

Capturing multiple-type interactions into practical predictors of type replacement following human papillomavirus vaccination

Current HPV vaccines target a subset of the oncogenic human papillomavirus (HPV) types. If HPV types compete during infection, vaccination may trigger replacement by the non-targeted types. Existing approaches to assess the risk of type replacement have focused on detecting competitive interactions between pairs of vaccine and non-vaccine types. However, methods to translate any inferred pairwise interactions into predictors of replacement have been lacking. In this paper, we develop practical predictors of type replacement in a multi-type setting, readily estimable from pre-vaccination longitudinal or cross-sectional prevalence data. The predictors we propose for replacement by individual non-targeted types take the form of weighted cross-hazard ratios of acquisition versus clearance, or aggregate odds ratios of coinfection with the vaccine types. We elucidate how the hazard-based predictors incorporate potentially heterogeneous direct and indirect type interactions by appropriately weighting type-specific hazards and show when they are equivalent to the odds-based predictors. Additionally, pooling type-specific predictors proves to be useful for predicting increase in the overall non-vaccine-type prevalence. Using simulations, we demonstrate good performance of the predictors under different interaction structures. We discuss potential applications and limitations of the proposed methodology in predicting type replacement, as compared to existing approaches. This article is part of the theme issue 'Silent cancer agents: multi-disciplinary modelling of human DNA oncoviruses'.

Human Papillomavirus Genotype Replacement: Still Too Early to Tell?

Background

Although human papillomavirus (HPV) vaccines are highly efficacious in protecting against HPV infections and related diseases, vaccination may trigger replacement by nontargeted genotypes if these compete with the vaccine-targeted types. HPV genotype replacement has been deemed unlikely, based on the lack of systematic increases in the prevalence of nonvaccine-type (NVT) infection in the first decade after vaccination, and on the presence of cross-protection for some NVTs.

Methods

To investigate whether type replacement can be inferred from early postvaccination surveillance, we constructed a transmission model in which a vaccine type and an NVT compete through infection-induced cross-immunity. We simulated scenarios of different levels of cross-immunity and vaccine-induced cross-protection to the NVT. We validated whether commonly used measures correctly indicate type replacement in the long run.

Results

Type replacement is a trade-off between cross-immunity and cross-protection; cross-immunity leads to type replacement unless cross-protection is strong enough. With weak cross-protection, NVT prevalence may initially decrease before rebounding into type replacement, exhibiting a honeymoon period. Importantly, vaccine effectiveness for NVTs is inadequate for indicating type replacement.

Conclusions

Although postvaccination surveillance thus far is reassuring, it is still too early to preclude type replacement. Monitoring of NVTs remains pivotal in gauging population-level impacts of HPV vaccination.

Coinfections by noninteracting pathogens are not independent and require new tests of interaction

If pathogen species, strains, or clones do not interact, intuition suggests the proportion of coinfected hosts should be the product of the individual prevalences. Independence consequently underpins the wide range of methods for detecting pathogen interactions from cross-sectional survey data. However, the very simplest of epidemiological models challenge the underlying assumption of statistical independence. Even if pathogens do not interact, death of coinfected hosts causes net prevalences of individual pathogens to decrease simultaneously. The induced positive correlation between prevalences means the proportion of coinfected hosts is expected to be higher than multiplication would suggest. By modelling the dynamics of multiple noninteracting pathogens causing chronic infections, we develop a pair of novel tests of interaction that properly account for nonindependence between pathogens causing lifelong infection. Our tests allow us to reinterpret data from previous studies including pathogens of humans, plants, and animals. Our work demonstrates how methods to identify interactions between pathogens can be updated using simple epidemic models.

If pathogen species, strains, or clones do not interact, intuition suggests the proportion of coinfected hosts can be obtained by simply multiplying the individual prevalences. However, even simple epidemiological models show this to be untrue. This study develops new tests for interaction between pathogens that account for this surprising lack of statistical independence.