Asymptomatic Bordetella pertussis infections in a longitudinal cohort of young African infants and their mothers
Curation statements for this article:-
Curated by eLife
Evaluation Summary:
Overall the reviewers were positive about this manuscript and the importance of this analysis being in identifying the role of asymptomatic transmitters in populations. There are some revisions that will be required and a number of areas for additional analyses and clarifications that would help the reader better put this manuscript in context.
(This preprint has been reviewed by eLife. We include the public reviews from the reviewers here; the authors also receive private feedback with suggested changes to the manuscript. Reviewer #3 agreed to share their name with the authors.)
This article has been Reviewed by the following groups
Listed in
- Evaluated articles (eLife)
Abstract
Recent pertussis resurgence in numerous countries may be driven by asymptomatic infections. Most pertussis surveillance studies are cross-sectional and cannot distinguish asymptomatic from pre-symptomatic infections. Longitudinal surveillance could overcome this barrier, providing more information about the true burden of pertussis at the population level. Here we analyze 17,442 nasopharyngeal samples from a longitudinal cohort of 1320 Zambian mother/infant pairs. Our analysis has two elements. First, we demonstrate that the full range of IS481 qPCR CT values provides insight into pertussis epidemiology, showing concordance of low and high CT results over time, within mother/infant pairs, and in relation to symptomatology. Second, we exploit these full-range qPCR data to demonstrate a high incidence of asymptomatic pertussis, including among infants. Our results demonstrate a wider burden of pertussis infection than we anticipated in this population, and expose key limitations of threshold-based interpretation of qPCR results in infectious disease surveillance.
Article activity feed
-
-
Author Response to Public Reviews
We thank the reviewers and editors for their detailed and insightful comments. We believe the consequent revisions have greatly increased the overall clarity of the manuscript, and provide important additional context and analysis.
Reviewer #1 (Public Review):
We thank the reviewer for the detailed comments.
[...] Overall, the manuscript lacks substantial statistical support or clear evidence of some of the patterns they are stating and would require a substantial revision to justify their conclusions. The majority of the manuscript relies on 8 infant/mother pairs where they have evidence of pertussis infection and rely on the dense sampling to investigate infection dynamics. However, this is a very small sample size and further, based on the results displayed in Figure 1, it is not obvious that the …
Author Response to Public Reviews
We thank the reviewers and editors for their detailed and insightful comments. We believe the consequent revisions have greatly increased the overall clarity of the manuscript, and provide important additional context and analysis.
Reviewer #1 (Public Review):
We thank the reviewer for the detailed comments.
[...] Overall, the manuscript lacks substantial statistical support or clear evidence of some of the patterns they are stating and would require a substantial revision to justify their conclusions. The majority of the manuscript relies on 8 infant/mother pairs where they have evidence of pertussis infection and rely on the dense sampling to investigate infection dynamics. However, this is a very small sample size and further, based on the results displayed in Figure 1, it is not obvious that the data has a very pattern that warrant their assertions.
As noted in the introduction, we begin our results with “a descriptive analysis of eight mother/infant pairs where each symptomatic infant had definitive qPCR-based evidence of pertussis infection.” Our goal in this section is to use noteworthy examples to highlight salient epidemiological patterns, which we explore in further detail using data from the full cohort in subsequent sections. We note that the results presented in Fig 3 onwards in no way rely on any arguments and/or specific patterns described in Fig 2. In other words, the original eight pairs revealed several unanticipated findings (particularly the finding of repeated high CT values PCR findings in the mothers of a child with definite pertussis), that were intriguing and potentially relevant in terms of pertussis epidemiology. They are also unique – we have not seen any published time series data using qPCR in this way before. These early observations motivated us to conduct a more detailed and quantitative analysis of the cohort of >1,300 mother/infant pairs.
The sample size under consideration in the majority of the manuscript (i.e., all except for the above section) is 1,320 mother/infant pairs (2,640 subjects), as shown in Table 1 and 2. In the original submission, sample sizes were also clearly indicated in Figure 2B (assays per week), Fig 3B (subjects per group), Table 2 (subjects per group), Figures S1-2 (study profile), Figure S3 (NP samples per infant), and Table S1.
We have revised the panel order and axes labels of the current Figure 3 to more clearly illustrate the relationship between panels, and to clarify that the 6 example pairs shown in Fig 3A are unrelated to the 8 pairs shown in Figure 2. We hope this addresses any remaining confusion.
While there are some instances with a combination of higher/lower IS481 CT values, it does not appear to have a clear pattern. For example, what are possible explanations for time periods between samples with evidence of IS481 and those without (such as pair A, C, D, E, F and H)? There also does not appear to be a clear pattern of symptoms in any of these samples (aside from having fewer symptoms in the mothers than infants).
The ambiguity of these patterns played a role in guiding our analysis of the entire cohort, where we establish evidence for infection based on a preponderance of evidence from a large number of individuals.
Further, it is not obvious how similar these observed (such as a mixture of times of high or low values often preceded or followed by times when IS481 was not detected) is similar to different to the rest of the cohort (in contrast to those who have a definitive positive NP sample during a symptomatic visit).
The main results are primarily a descriptive analysis of these 8 mother/infant pairs with little statistical analyses or additional support.
We strongly disagree with this characterization of our results, where we state that “In this analysis, we focus on the 1,320 pairs with ≥4 NP samples per subject (Figure S3)”. We believe the reviewer’s confusion may stem, in part, from a mis-interpretation of Figure 2 (below), along with our erroneous reference to Figure 3 (we incorrectly stated Fig 2, adding to the confusion). With this in mind, we have revised the previous Figure 2 (now Figure 3) in the interest of clarity, and more carefully described exactly what the points displayed in Figure 3 represent.
The authors do not provide evidence or detail about what is known about the variability in IS481 CT values, amongst individuals, or over time, or pre/post vaccination. Without this information, it is not clear how informative some of this variability is versus how much variability in these values is expected.
We agree that this is important information, and we have added figures and results summarizing the observed impact of vaccination on CT values (see essential revision 1, above), and the patterns of transitions of CT values across adjacent samples within individuals throughout the study (see essential revision 2). This latter analysis is now summarized in Figure 6, and shows a clear tendency for step-wise transitions over time. The implication is that the data present structure rather than random noise. This supports our overall contention that full-range CT values can provide meaningful insights into pertussis epidemiology. We also note that Fig. 7A (previously Fig 3A) and Table 3 (previously Table S1) do indeed summarize the distribution of CT values, including variability amongst individuals. As noted above, we have also included an additional analysis summarizing the interdependence of CT value on both symptoms and antibiotics (Fig 8-figure supplement 1).
I think particularly in Figure 1, how many of the individuals have periods between times when IS481 evidence was observed when it was not observed, is concerning that these data (at this granular a level) are measuring true infection dynamics.
Adding in additional information about the distribution and patterns of these values for the other cohort members would also provide valuable insight into how Figure 1 should be interpreted in this context.
We believe our previous comments concerning the relationship between the current Figure 2 (illustrative example) and the remaining figures (cohort analysis) addresses this comment.
As it stands, the authors do not provide sufficient interpretation and evidence for having relevant infection arcs.
We have revised the manuscript to clarify that infection arcs are observed in other studies and expected in infected individuals, rather than directly observed and/or quantified in this study.
It appears that Figure 2A was created using only 8 data points (from the infant data values). If so, this level of extrapolation from such few data points does not provide enough evidence to support to the results in the text (particularly about evidence for fade-in/fade-out population-level dynamics). Also, in Figure 2, it is not clear to me the added value of Figure 2C and the main goal of this figure.
We believe our previous comments have addressed this point. As noted, we have revised the current Fig 3 for clarity. Figure 3A and 3C are intended to demonstrate the structure of the cohort across the study period. We have revised the caption to clarify this point.
The authors have created a measure called, evidence for infection (EFI), which is a summary measure of their IS481 CT values across the study. However, it is not clear why the authors are only considering an aggregated (sum) value which loses any temporality or relationship with symptoms/antibiotic use. For example, the values may have been high earlier in the study, but symptoms were unrelated to that evidence for infection - or visa versa.
We believe that temporal patterns of CT values within subjects now described in Figure 6 deserve further detailed attention that is outside the scope of the current work. We believe the high-level empirical summaries presented here are strengthened by their reliance on a preponderance of evidence. In the current revision, we have also included additional analyses that we believe address some (if not all) of the reviewers concerns.
This seems to be an important factor - were these possible undiagnosed, asymptomatic, or mild symptomatic pertussis infections? It is not clear why the authors only focus on a sum value for EFI versus other measures (such as multiple values above or below certain thresholds, etc.) to provide additional support and evidence for their results.
Our approach seeks to use an objective statistical summary (geometric mean RCD proportion) to quantify the “signal” contained in IS481 assays within individuals across the course of the study. We note that, while both false positives and false negatives are likely in this study, the sample characteristics of the cohort mean that repeated false positives within individuals are unlikely based on chance alone. Further, a central aspect to our argument is that dichotomizing a continuous variable at an arbitrary threshold is reductive and unnecessarily introduces misclassification that reduces, rather than improves, statistical power.
It is not clear why the authors have emphasized the novelty and large proportion of asymptomatic infections observed in these data. For example, there have been household studies of pertussis (see https://academic.oup.com/cid/article-abstract/70/1/152/5525423?redirectedFrom=PDF which performed a systematic review that included this topic) that have also found such evidence.
We are aware of the paper above, which we had cited in the discussion. A key limitation of the referenced study is reliance on retrospective recall spanning many months. Since pertussis infections may be mild and non-specific, the fact that household contacts of an index case cannot recall a pertussis-like infection is consistent with asymptomatic infection, but far from definitive evidence. Moreover, the use of seroconversion as the measure of exposure is unreliable, since variations in antibody concentrations can be driven by a number of factors other than natural exposure.
While cross-sectional surveys may be commonly used in practice, it is not clear that there is no other type of study that provides any evidence for asymptomatic infections.
Our core argument is that it is impossible to know with certainty that a symptom-free patient with a detecting qPCR on Monday would not have become symptomatic if recontacted on Tuesday. By their nature, cross-sectional studies simply cannot parse asymptomatic from pre-symptomatic infections. To do that, one needs a longitudinal design, as reflected in the aforementioned longitudinal household contact studies. A key consideration addressed in the current work is the extent to which low and/or borderline CT values should be reinterpreted within the context of A) repeated sampling of individuals over time and B) epidemiological surveillance versus clinical diagnosis. We do not claim that our approach is the only one possible.
Further, it is not clear why the authors refer to widespread asymptomatic pertussis when a large proportion of individuals with evidence for pertussis infection had symptoms. Would it not be undiagnosed pertussis if it is associated with clinical symptomatology?
We have revised the text to highlight the significance of both asymptomatic and minimally symptomatic pertussis. As we describe both here and in Gill et al. 2016, only a handful of individuals meet the consensus criteria for clinical pertussis (Ct<35). In addition, qPCR results were not available to clinic staff in real-time. This, coupled with the relative absence of severe symptoms during study visits (especially in mothers), meant that only one study participant was diagnosed with pertussis at the time of their visit.
Reviewer #2 (Public Review):
We thank the reviewer for their supportive comments.
This study was done in a population with wP vaccine, I wonder if that's part of the reason many of the CT values are high. Can the authors speculate what this study would look like in a population having received aP for a long period? I'd appreciate more discussion around vaccination in general.
We have added results summarizing the possible interaction between IS481 assays with infant vaccination.
We also note that aP is widely used in high-resource settings where overall pertussis incidence is lower, while pertussis diagnosis and treatment are more widely available. Our results indicate that mothers in this population experience non-trivial pertussis incidence over time, yielding immunological profiles from repeated infection that we expect differ markedly from that of individuals who lack naturally-derived resistance to infection via, e.g., mucosal antibodies and tissue-resident T-cells. Recognizing that our study does not provide a direct comparison with aP-vaccinated populations, we nonetheless believe that directly comparable populations (urban poor in under-served communities) are both numerous and under-studied.
-
Reviewer #3 (Public Review):
Gill et al. presents an extensive analysis of information/data collected as part of a pertussis vaccine study conducted in Zambia (the basis for an earlier publication, Gill et al., CID 2016). As part of the initial study, the investigators collected serial NP samples from mother/infant pairs at sequential follow-up clinic visits and analyzed them by PCR for the presence of IS481 and, in some cases, ptxS1. The results from these assays were evaluated in conjunction with clinical information on potential manifestations of respiratory illness in the infants and mothers. The authors found important patterns of PCR Ct values, which might not have been considered positive on a single sample PCR from a single patient PCR in a US clinical microbiology lab. Together, however, representing a collection of serial …
Reviewer #3 (Public Review):
Gill et al. presents an extensive analysis of information/data collected as part of a pertussis vaccine study conducted in Zambia (the basis for an earlier publication, Gill et al., CID 2016). As part of the initial study, the investigators collected serial NP samples from mother/infant pairs at sequential follow-up clinic visits and analyzed them by PCR for the presence of IS481 and, in some cases, ptxS1. The results from these assays were evaluated in conjunction with clinical information on potential manifestations of respiratory illness in the infants and mothers. The authors found important patterns of PCR Ct values, which might not have been considered positive on a single sample PCR from a single patient PCR in a US clinical microbiology lab. Together, however, representing a collection of serial samples from study subjects, they strongly support the proposal that asymptomatic infections occurred in these study subjects. The authors used multiple approaches, including determining a mathematical "Evidence For Infection" or EFI to analyze the data from individual subjects and infant/mother pairs. From the collective data and analytical approaches, the authors provide a compelling case for infections with B. pertussis that are not associated with significant clinical symptoms. This possibility has certainly been considered previously, but not possible to address in the absence of the enormous amount of quantitative data and analysis provided from this prospective study. Another important point made from these data is that PCR Ct values can be useful in other than an all-or-nothing (positive or negative) decision, as is done appropriately with single patient samples submitted to clinical microbiology labs for PCR analysis.
-
Reviewer #2 (Public Review):
In the current study Gill et al present a retrospective analysis of NP swabs of mother infant pairs taken longitudinally in Zambia. They use qPCR CT values to quantify the amount of IS431 in each sample to detect pertussis infection. They find strong evidence for asymptomatic pertussis infection in both mothers and infants, validating previous work identifying the role of asymptomatic transmitters in populations. This is a tremendously important study and is conducted and analyzed very well. The manuscript is well written, and I heartily recommend publication. Excellent work, well done.
Comments:
This study was done in a population with wP vaccine, I wonder if that's part of the reason many of the CT values are high. Can the authors speculate what this study would look like in a population having received aP …
Reviewer #2 (Public Review):
In the current study Gill et al present a retrospective analysis of NP swabs of mother infant pairs taken longitudinally in Zambia. They use qPCR CT values to quantify the amount of IS431 in each sample to detect pertussis infection. They find strong evidence for asymptomatic pertussis infection in both mothers and infants, validating previous work identifying the role of asymptomatic transmitters in populations. This is a tremendously important study and is conducted and analyzed very well. The manuscript is well written, and I heartily recommend publication. Excellent work, well done.
Comments:
This study was done in a population with wP vaccine, I wonder if that's part of the reason many of the CT values are high. Can the authors speculate what this study would look like in a population having received aP for a long period? I'd appreciate more discussion around vaccination in general.
-
Reviewer #1 (Public Review):
In "Asymptomatic Bordetella pertussis infections in a longitudinal cohort of young African infants and their mothers", the authors analyze longitudinal data from a cohort in Zambia of infant/mother pairs to investigate the evidence for subclinical and asymptomatic infections in both pairs as well as the use of IS481 qPCR cycle threshold (CT) values in providing evidence for pertussis infection. Overall, the manuscript lacks substantial statistical support or clear evidence of some of the patterns they are stating and would require a substantial revision to justify their conclusions. The majority of the manuscript relies on 8 infant/mother pairs where they have evidence of pertussis infection and rely on the dense sampling to investigate infection dynamics. However, this is a very small sample size and …
Reviewer #1 (Public Review):
In "Asymptomatic Bordetella pertussis infections in a longitudinal cohort of young African infants and their mothers", the authors analyze longitudinal data from a cohort in Zambia of infant/mother pairs to investigate the evidence for subclinical and asymptomatic infections in both pairs as well as the use of IS481 qPCR cycle threshold (CT) values in providing evidence for pertussis infection. Overall, the manuscript lacks substantial statistical support or clear evidence of some of the patterns they are stating and would require a substantial revision to justify their conclusions. The majority of the manuscript relies on 8 infant/mother pairs where they have evidence of pertussis infection and rely on the dense sampling to investigate infection dynamics. However, this is a very small sample size and further, based on the results displayed in Figure 1, it is not obvious that the data has a very pattern that warrant their assertions.
Major comments:
The main results and conclusions are highly reliant on details from eight mother/infant pairs. However, Figure 1 does not show a clear picture of the fade-in/fade-out. The authors go into great detail describing each of these 8 pairs, however based on the figure and text there does not appear to be clear evidence of an underlying pattern. While there are some instances with a combination of higher/lower IS481 CT values, it does not appear to have a clear pattern. For example, what are possible explanations for time periods between samples with evidence of IS481 and those without (such as pair A, C, D, E, F and H)? There also does not appear to be a clear pattern of symptoms in any of these samples (aside from having fewer symptoms in the mothers than infants). Further, it is not obvious how similar these observed (such as a mixture of times of high or low values often preceded or followed by times when IS481 was not detected) is similar to different to the rest of the cohort (in contrast to those who have a definitive positive NP sample during a symptomatic visit). The main results are primarily a descriptive analysis of these 8 mother/infant pairs with little statistical analyses or additional support.
The authors do not provide evidence or detail about what is known about the variability in IS481 CT values, amongst individuals, or over time, or pre/post vaccination. Without this information, it is not clear how informative some of this variability is versus how much variability in these values is expected. I think particularly in Figure 1, how many of the individuals have periods between times when IS481 evidence was observed when it was not observed, is concerning that these data (at this granular a level) are measuring true infection dynamics. Adding in additional information about the distribution and patterns of these values for the other cohort members would also provide valuable insight into how Figure 1 should be interpreted in this context. As it stands, the authors do not provide sufficient interpretation and evidence for having relevant infection arcs.
It appears that Figure 2A was created using only 8 data points (from the infant data values). If so, this level of extrapolation from such few data points does not provide enough evidence to support to the results in the text (particularly about evidence for fade-in/fade-out population-level dynamics). Also, in Figure 2, it is not clear to me the added value of Figure 2C and the main goal of this figure.
The authors have created a measure called, evidence for infection (EFI), which is a summary measure of their IS481 CT values across the study. However, it is not clear why the authors are only considering an aggregated (sum) value which loses any temporality or relationship with symptoms/antibiotic use. For example, the values may have been high earlier in the study, but symptoms were unrelated to that evidence for infection - or visa versa. This seems to be an important factor - were these possible undiagnosed, asymptomatic, or mild symptomatic pertussis infections? It is not clear why the authors only focus on a sum value for EFI versus other measures (such as multiple values above or below certain thresholds, etc.) to provide additional support and evidence for their results.
It is not clear why the authors have emphasized the novelty and large proportion of asymptomatic infections observed in these data. For example, there have been household studies of pertussis (see https://academic.oup.com/cid/article-abstract/70/1/152/5525423?redirectedFrom=PDF which performed a systematic review that included this topic) that have also found such evidence. While cross-sectional surveys may be commonly used in practice, it is not clear that there is no other type of study that provides any evidence for asymptomatic infections. Further, it is not clear why the authors refer to widespread asymptomatic pertussis when a large proportion of individuals with evidence for pertussis infection had symptoms. Would it not be undiagnosed pertussis if it is associated with clinical symptomatology?
-
Evaluation Summary:
Overall the reviewers were positive about this manuscript and the importance of this analysis being in identifying the role of asymptomatic transmitters in populations. There are some revisions that will be required and a number of areas for additional analyses and clarifications that would help the reader better put this manuscript in context.
(This preprint has been reviewed by eLife. We include the public reviews from the reviewers here; the authors also receive private feedback with suggested changes to the manuscript. Reviewer #3 agreed to share their name with the authors.)
-
