Analysing malaria drug trials on a per-individual or per-clone basis: a comparison of methods

Persistent and multiclonal malaria parasite dynamics despite extended artemether-lumefantrine treatment in children

Standard diagnostics used in longitudinal antimalarial studies are unable to characterize the complexity of submicroscopic parasite dynamics, particularly in high transmission settings. We use molecular markers and amplicon sequencing to characterize post-treatment stage-specific malaria parasite dynamics during a 42 day randomized trial of 3- versus 5 day artemether-lumefantrine in 303 children with and without HIV (ClinicalTrials.gov number NCT03453840). The prevalence of parasite-derived 18S rRNA is >70% in children throughout follow-up, and the ring-stage marker SBP1 is detectable in over 15% of children on day 14 despite effective treatment. We find that the extended regimen significantly lowers the risk of recurrent ring-stage parasitemia compared to the standard 3 day regimen, and that higher day 7 lumefantrine concentrations decrease the probability of ring-stage parasites in the early post-treatment period. Longitudinal amplicon sequencing reveals remarkably dynamic patterns of multiclonal infections that include new and persistent clones in both the early post-treatment and later time periods. Our data indicate that post-treatment parasite dynamics are highly complex despite efficacious therapy, findings that will inform strategies to optimize regimens in the face of emerging partial artemisinin resistance in Africa.

The many definitions of multiplicity of infection

The presence of multiple genetically different pathogenic variants within the same individual host is common in infectious diseases. Although this is neglected in some diseases, it is well recognized in others like malaria, where it is typically referred to as multiplicity of infection (MOI) or complexity of infection (COI). In malaria, with the advent of molecular surveillance, data is increasingly being available with enough resolution to capture MOI and integrate it into molecular surveillance strategies. The distribution of MOI on the population level scales with transmission intensities, while MOI on the individual level is a confounding factor when monitoring haplotypes of particular interests, e.g., those associated with drug-resistance. Particularly, in high-transmission areas, MOI leads to a discrepancy between the likelihood of a haplotype being observed in an infection (prevalence) and its abundance in the pathogen population (frequency). Despite its importance, MOI is not universally defined. Competing definitions vary from verbal ones to those based on concise statistical frameworks. Heuristic approaches to MOI are popular, although they do not mine the full potential of available data and are typically biased, potentially leading to misinferences. We introduce a formal statistical framework and suggest a concise definition of MOI and its distribution on the host-population level. We show how it relates to alternative definitions such as the number of distinct haplotypes within an infection or the maximum number of alleles detectable across a set of genetic markers. It is shown how alternatives can be derived from the general framework. Different statistical methods to estimate the distribution of MOI and pathogenic variants at the population level are discussed. The estimates can be used as plug-ins to reconstruct the most probable MOI of an infection and set of infecting haplotypes in individual infections. Furthermore, the relation between prevalence of pathogenic variants and their frequency (relative abundance) in the pathogen population in the context of MOI is clarified, with particular regard to seasonality in transmission intensities. The framework introduced here helps to guide the correct interpretation of results emerging from different definitions of MOI. Especially, it excels comparisons between studies based on different analytical methods.

Should deep-sequenced amplicons become the new gold-standard for analysing malaria drug clinical trials?

Regulatory clinical trials are required to ensure the continued supply and deployment of effective antimalarial drugs. Patient follow-up in such trials typically lasts several weeks, as the drugs have long half-lives and new infections often occur during this period. “Molecular correction” is therefore used to distinguish drug failures from new infections. The current WHO-recommended method for molecular correction uses length-polymorphic alleles at highly diverse loci but is inherently poor at detecting low-density clones in polyclonal infections. This likely leads to substantial underestimates of failure rates, delaying the replacement of failing drugs with potentially lethal consequences. Deep-sequenced amplicons (AmpSeq) substantially increase the detectability of low-density clones and may offer a new “gold standard” for molecular correction. Pharmacological simulation of clinical trials was used to evaluate the suitability of AmpSeq for molecular correction. We investigated the impact of factors such as the number of amplicon loci analyzed, the informatics criteria used to distinguish genotyping “noise” from real low-density signals, the local epidemiology of malaria transmission, and the potential impact of genetic signals from gametocytes. AmpSeq greatly improved molecular correction and provided accurate drug failure rate estimates. The use of 3 to 5 amplicons was sufficient, and simple, nonstatistical criteria could be used to classify recurrent infections as drug failures or new infections. These results suggest AmpSeq is strongly placed to become the new standard for molecular correction in regulatory trials, with potential extension into routine surveillance once the requisite technical support becomes established.

Markov chain Monte Carlo Gibbs sampler approach for estimating haplotype frequencies among multiple malaria infected human blood samples

Background

Malaria patients can have two or more haplotypes in their blood sample making it challenging to identify which haplotypes they carry. In addition, there are challenges in measuring the type and frequency of resistant haplotypes in populations. This study presents a novel statistical method Gibbs sampler algorithm to investigate this issue.

Results

The performance of the algorithm is evaluated on simulated datasets consisting of patient blood samples characterized by their multiplicity of infection (MOI) and malaria genotype. The simulation used different resistance allele frequencies (RAF) at each Single Nucleotide Polymorphisms (SNPs) and different limit of detection (LoD) of the SNPs and the MOI. The Gibbs sampler algorithm presents higher accuracy among high LoD of the SNPs or the MOI, validated, and deals with missing MOI compared to previous related statistical approaches.

Conclusions

The Gibbs sampler algorithm provided robust results when faced with genotyping errors caused by LoDs and functioned well even in the absence of MOI data on individual patients.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12936-021-03841-9.

A Computer Modelling Approach To Evaluate the Accuracy of Microsatellite Markers for Classification of Recurrent Infections during Routine Monitoring of Antimalarial Drug Efficacy

Antimalarial drugs have long half-lives, so clinical trials to monitor their efficacy require long periods of follow-up to capture drug failure that may become patent only weeks after treatment. Reinfections often occur during follow-up, so robust methods of distinguishing drug failures (recrudescence) from emerging new infections are needed to produce accurate failure rate estimates. Molecular correction aims to achieve this by comparing the genotype of a patient's pretreatment (initial) blood sample with that of any infection that occurs during follow-up, with matching genotypes indicating drug failure. We use an in silico approach to show that the widely used match-counting method of molecular correction with microsatellite markers is likely to be highly unreliable and may lead to gross under- or overestimates of the true failure rates, depending on the choice of matching criterion. A Bayesian algorithm for molecular correction was previously developed and utilized for analysis of in vivo efficacy trials. We validated this algorithm using in silico data and showed it had high specificity and generated accurate failure rate estimates. This conclusion was robust for multiple drugs, different levels of drug failure rates, different levels of transmission intensity in the study sites, and microsatellite genetic diversity. The Bayesian algorithm was inherently unable to accurately identify low-density recrudescence that occurred in a small number of patients, but this did not appear to compromise its utility as a highly effective molecular correction method for analyzing microsatellite genotypes. Strong consideration should be given to using Bayesian methodology to obtain accurate failure rate estimates during routine monitoring trials of antimalarial efficacy that use microsatellite markers.

PCR correction strategies for malaria drug trials: updates and clarifications

Malaria drug trials conducted in endemic areas face a major challenge in their analysis because it is difficult to establish whether parasitaemia in blood samples collected after treatment indicate drug failure or a new infection acquired after treatment. It is therefore vital to reliably distinguish drug failures from new infections in order to obtain accurate estimates of drug failure rates. This distinction can be achieved for Plasmodium falciparum by comparing parasite genotypes obtained at the time of treatment (the baseline) and on the day of recurring parasitaemia. Such PCR correction is required to obtain accurate failure rates, even for new effective drugs. Despite the routine use of PCR correction in surveillance of drug resistance and in clinical drug trials, limitations inherent to the molecular genotyping methods have led some researchers to question the validity of current PCR correction strategies. Here we describe and discuss recent developments in these genotyping approaches, with a particular focus on method validation and limitations of the genotyping strategies. Our aim is to update scientists from public and private bodies who are working on the development, deployment, and surveillance of new malaria drugs. We aim to promote discussion around these issues and argue for the adoption of improved standardised PCR correction methodologies.

Improving Methods for Analyzing Antimalarial Drug Efficacy Trials: Molecular Correction Based on Length-Polymorphic Markers msp-1, msp-2, and glurp

Drug efficacy trials monitor the continued efficacy of front-line drugs against falciparum malaria. Overestimating efficacy results in a country retaining a failing drug as first-line treatment with associated increases in morbidity and mortality, while underestimating drug effectiveness leads to removal of an effective treatment with substantial practical and economic implications.

Drug efficacy trials monitor the continued efficacy of front-line drugs against falciparum malaria. Overestimating efficacy results in a country retaining a failing drug as first-line treatment with associated increases in morbidity and mortality, while underestimating drug effectiveness leads to removal of an effective treatment with substantial practical and economic implications. Trials are challenging: they require long durations of follow-up to detect drug failures, and patients are frequently reinfected during that period. Molecular correction based on parasite genotypes distinguishes reinfections from drug failures to ensure the accuracy of failure rate estimates. Several molecular correction “algorithms” have been proposed, but which is most accurate and/or robust remains unknown. We used pharmacological modeling to simulate parasite dynamics and genetic signals that occur in patients enrolled in malaria drug clinical trials. We compared estimates of treatment failure obtained from a selection of proposed molecular correction algorithms against the known “true” failure rate in the model. Our findings are as follows. (i) Molecular correction is essential to avoid substantial overestimates of drug failure rates. (ii) The current WHO-recommended algorithm consistently underestimates the true failure rate. (iii) Newly proposed algorithms produce more accurate failure rate estimates; the most accurate algorithm depends on the choice of drug, trial follow-up length, and transmission intensity. (iv) Long durations of patient follow-up may be counterproductive; large numbers of new infections accumulate and may be misclassified, overestimating drug failure rate. (v) Our model was highly consistent with existing in vivo data. The current WHO-recommended method for molecular correction and analysis of clinical trials should be reevaluated and updated.

Statistical methods to derive efficacy estimates of anti-malarials for uncomplicated Plasmodium falciparum malaria: pitfalls and challenges

The Kaplan-Meier (K-M) method is currently the preferred approach to derive an efficacy estimate from anti-malarial trial data. In this approach event times are assumed to be continuous and estimates are generated on the assumption that there is only one cause of failure. In reality, failures are captured at pre-scheduled time points and patients can fail treatment due to a variety of causes other than the primary endpoint, commonly termed competing risk events. Ignoring these underlying assumptions can potentially distort the derived efficacy estimates and result in misleading conclusions. This review details the evolution of statistical methods used to derive anti-malarial efficacy for uncomplicated Plasmodium falciparum malaria and assesses the limitations of the current practices. Alternative approaches are explored and their implementation is discussed using example data from a large multi-site study.

Markov chain Monte Carlo and expectation maximization approaches for estimation of haplotype frequencies for multiply infected human blood samples

Background

Haplotypes are important in anti-malarial drug resistance because genes encoding drug resistance may accumulate mutations at several codons in the same gene, each mutation increasing the level of drug resistance and, possibly, reducing the metabolic costs of previous mutation. Patients often have two or more haplotypes in their blood sample which may make it impossible to identify exactly which haplotypes they carry, and hence to measure the type and frequency of resistant haplotypes in the malaria population.

Results

This study presents two novel statistical methods expectation–maximization (EM) and Markov chain Monte Carlo (MCMC) algorithms to investigate this issue. The performance of the algorithms is evaluated on simulated datasets consisting of patient blood characterized by their multiplicity of infection (MOI) and malaria genotype. The datasets are generated using different resistance allele frequencies (RAF) at each single nucleotide polymorphisms (SNPs) and different limit of detection (LoD) of the SNPs and the MOI. The EM and the MCMC algorithm are validated and appear more accurate, faster and slightly less affected by LoD of the SNPs and the MOI compared to previous related statistical approaches.

Conclusions

The EM and the MCMC algorithms perform well when analysing malaria genetic data obtained from infected human blood samples. The results are robust to genotyping errors caused by LoDs and function well even in the absence of MOI data on individual patients.

Electronic supplementary material

The online version of this article (doi:10.1186/s12936-016-1473-5) contains supplementary material, which is available to authorized users.

Why are two mistakes not worse than one? A proposal for controlling the expected number of false claims

Multiplicity is common in clinical studies and the current standard is to use the familywise error rate to ensure that the errors are kept at a prespecified level. In this paper, we will show that, in certain situations, familywise error rate control does not account for all errors made. To counteract this problem, we propose the use of the expected number of false claims (EFC). We will show that a (weighted) Bonferroni approach can be used to control the EFC, discuss how a study that uses the EFC can be powered for co-primary, exchangeable, and hierarchical endpoints, and show how the weight for the weighted Bonferroni test can be determined in this manner. ©2016 The Authors. Pharmaceutical Statistics Published by John Wiley & Sons Ltd.

Altering Antimalarial Drug Regimens May Dramatically Enhance and Restore Drug Effectiveness

There is considerable concern that malaria parasites are starting to evolve resistance to the current generation of antimalarial drugs, the artemisinin-based combination therapies (ACTs). We use pharmacological modeling to investigate changes in ACT effectiveness likely to occur if current regimens are extended from 3 to 5 days or, alternatively, given twice daily over 3 days. We show that the pharmacology of artemisinins allows both regimen changes to substantially increase the artemisinin killing rate. Malaria patients rarely contain more than 10(12) parasites, while the standard dosing regimens allow approximately 1 in 10(10) parasites to survive artemisinin treatment. Parasite survival falls dramatically, to around 1 in 10(17) parasites if the dose is extended or split; theoretically, this increase in drug killing appears to be more than sufficient to restore failing ACT efficacy. One of the most widely used dosing regimens, artemether-lumefantrine, already successfully employs a twice-daily dosing regimen, and we argue that twice-daily dosing should be incorporated into all ACT regimen design considerations as a simple and effective way of ensuring the continued long-term effectiveness of ACTs.

Improving the role and contribution of pharmacokinetic analyses in antimalarial drug clinical trials

It is now World Health Organization (WHO) policy that drug concentrations on day 7 be measured as part of routine assessment in antimalarial drug efficacy trials. The rationale is that this single pharmacological measure serves as a simple and practical predictor of treatment outcome for antimalarial drugs with long half-lives. Herein we review theoretical data and field studies and conclude that the day 7 drug concentration (d7c) actually appears to be a poor predictor of therapeutic outcome. This poor predictive capability combined with the fact that many routine antimalarial trials will have few or no failures means that there appears to be little justification for this WHO recommendation. Pharmacological studies have a huge potential to improve antimalarial dosing, and we propose study designs that use more-focused, sophisticated, and cost-effective ways of generating these data than the mass collection of single d7c concentrations.

Optimizing the programmatic deployment of the anti-malarials artemether-lumefantrine and dihydroartemisinin-piperaquine using pharmacological modelling

BACKGROUND

Successful programmatic use of anti-malarials faces challenges that are not covered by standard drug development processes. The development of appropriate pragmatic dosing regimens for low-resource settings or community-based use is not formally regulated, even though these may alter factors which can substantially affect individual patient and population level outcome, such as drug exposure, patient adherence and the spread of drug resistance and can affect a drug's reputation and its eventual therapeutic lifespan.

METHODS

An in silico pharmacological model of anti-malarial drug treatment with the pharmacokinetic/pharmacodynamic profiles of artemether-lumefantrine (AM-LF, Coartem®) and dihydroartemisinin-piperaquine (DHA-PPQ, Eurartesim®) was constructed to assess the potential impact of programmatic factors, including regionally optimized, age-based dosing regimens, poor patient adherence, food effects and drug resistance on treatment outcome at population level, and compared both drugs' susceptibility to these factors.

RESULTS

Compared with DHA-PPQ, therapeutic effectiveness of AM-LF seems more robust to factors affecting drug exposure, such as age- instead of weight-based dosing or poor adherence. The model highlights the sub-optimally low ratio of DHA:PPQ which, in combination with the narrow therapeutic dose range of PPQ compared to DHA that drives the weight or age cut-offs, leaves DHA at a high risk of under-dosing.

CONCLUSION

Pharmacological modelling of real-life scenarios can provide valuable supportive data and highlight modifiable determinants of therapeutic effectiveness that can help optimize the deployment of anti-malarials in control programmes.

Improving pharmacokinetic-pharmacodynamic modeling to investigate anti-infective chemotherapy with application to the current generation of antimalarial drugs

Mechanism-based pharmacokinetic-pharmacodynamic (PK/PD) modelling is the standard computational technique for simulating drug treatment of infectious diseases with the potential to enhance our understanding of drug treatment outcomes, drug deployment strategies, and dosing regimens. Standard methodologies assume only a single drug is used, it acts only in its unconverted form, and that oral drugs are instantaneously absorbed across the gut wall to their site of action. For drugs with short half-lives, this absorption period accounts for a significant period of their time in the body. Treatment of infectious diseases often uses combination therapies, so we refined and substantially extended the PK/PD methodologies to incorporate (i) time lags and drug concentration profiles resulting from absorption across the gut wall and, if required, conversion to another active form; (ii) multiple drugs within a treatment combination; (iii) differing modes of action of drugs in the combination: additive, synergistic, antagonistic; (iv) drugs converted to an active metabolite with a similar mode of action. This methodology was applied to a case study of two first-line malaria treatments based on artemisinin combination therapies (ACTs, artemether-lumefantrine and artesunate-mefloquine) where the likelihood of increased artemisinin tolerance/resistance has led to speculation on their continued long-term effectiveness. We note previous estimates of artemisinin kill rate were underestimated by a factor of seven, both the unconverted and converted form of the artemisinins kill parasites and the extended PK/PD methodology produced results consistent with field observations. The simulations predict that a potentially rapid decline in ACT effectiveness is likely to occur as artemisinin resistance spreads, emphasising the importance of containing the spread of artemisinin resistance before it results in widespread drug failure. We found that PK/PD data is generally very poorly reported in the malaria literature, severely reducing its value for subsequent re-application, and we make specific recommendations to improve this situation.