1. Epidemiology and Global Health
  2. Microbiology and Infectious Disease
Download icon

Zika seroprevalence declines and neutralizing antibodies wane in adults following outbreaks in French Polynesia and Fiji

  1. Alasdair D Henderson
  2. Maite Aubry
  3. Mike Kama
  4. Jessica Vanhomwegen
  5. Anita Teissier
  6. Teheipuaura Mariteragi-Helle
  7. Tuterarii Paoaafaite
  8. Yoann Teissier
  9. Jean-Claude Manuguerra
  10. John Edmunds
  11. Jimmy Whitworth
  12. Conall H Watson
  13. Colleen L Lau
  14. Van-Mai Cao-Lormeau  Is a corresponding author
  15. Adam J Kucharski  Is a corresponding author
  1. London School of Hygiene and Tropical Medicine, United Kingdom
  2. Institut Louis Malardé, French Polynesia
  3. Fiji Centre for Communicable Disease Control, Fiji
  4. The University of the South Pacific, Fiji
  5. Institut Pasteur, France
  6. Direction de la Santé de la Polynésie française, French Polynesia
  7. Australian National University, Australia
Research Article
  • Cited 3
  • Views 854
  • Annotations
Cite this article as: eLife 2020;9:e48460 doi: 10.7554/eLife.48460

Abstract

It has been commonly assumed that Zika virus (ZIKV) infection confers long-term protection against reinfection, preventing ZIKV from re-emerging in previously affected areas for several years. However, the long-term immune response to ZIKV following an outbreak remains poorly documented. We compared results from eight serological surveys before and after known ZIKV outbreaks in French Polynesia and Fiji, including cross-sectional and longitudinal studies. We found evidence of a decline in seroprevalence in both countries over a two-year period following first reported ZIKV transmission. This decline was concentrated in adults, while high seroprevalence persisted in children. In the Fiji cohort, there was also a significant decline in neutralizing antibody titres against ZIKV, but not against dengue viruses that circulated during the same period.

eLife digest

Since the Zika virus first emerged in the Pacific Islands in 2007, it has caused many outbreaks in the Pacific and Latin America. Some scientists thought that after exposure to the virus people would develop long-term immunity to it, reducing the number of outbreaks in the future. Several studies supported this idea. These studies showed that many people recently infected with Zika developed antibodies in their blood that might protect them from becoming ill during future outbreaks. But it was not clear how long this protection would last.

To better understand how immunity to the Zika virus changes over time, Henderson, Aubry et al. combined data from eight surveys that collected blood samples at different time points during Zika outbreaks in French Polynesia and Fiji. The analysis showed that the proportion of people with detectable antibodies against the Zika virus increased in both countries after the outbreaks. In children these immune responses persisted for years, but antibody levels declined over time in adults. By contrast, antibodies to the closely related dengue virus did not wane over time in individuals tested for both viruses in Fiji in 2013, 2015 and 2017.

The data suggest that immunity against the Zika virus may not last as long as previously thought, which could affect the chances of future outbreaks. The findings may also have implications for researchers studying the virus, because the number of people with antibodies against the virus is not a good estimate of how many people were initially infected. More studies are needed to understand immunity to Zika virus over time and how it may affect future outbreaks.

Introduction

Zika virus (ZIKV), a Flavivirus primarily transmitted to humans by Aedes mosquitoes, was first reported in the Pacific region on Yap island (Federated States of Micronesia) in 2007 (Duffy et al., 2009). Six years later, there was a large ZIKV outbreak in French Polynesia (Cao-Lormeau et al., 2014) where an estimated 11.5% of the population visited healthcare facilities with clinical symptoms suggestive of ZIKV infection (Kucharski et al., 2016). Since then the virus has spread across the Pacific region (Musso et al., 2014), including to Fiji where cases of ZIKV infection were first detected in July 2015 (World Health Organisation, 2015). The same year, cases of ZIKV infection in Latin America were reported for the first time (Zammarchi et al., 2015). From February 1 to November 18, 2016, due to its rapid spread and association with birth defects, microcephaly in newborns and Guillain-Barré syndrome in adults (Cao-Lormeau et al., 2016) the WHO declared ZIKV a Public Health Emergency of International Concern (World Health Organisation, 2016). At the end of 2016, outbreaks had declined in most of the countries recently affected (O'Reilly et al., 2018). However, ZIKV was still circulating in 2018 in several countries, including Fiji and Tonga in the Pacific region (World Health Organisation, 2019).

In countries with known ZIKV outbreaks, the few serological surveys that have been published found a high level of ZIKV seroprevalence following the outbreak. In French Polynesia, a population-representative cross-sectional serological survey at the end of the outbreak in 2014 found a seroprevalence of 49% (Aubry et al., 2017). In Martinique, a study of blood donors showed a post-outbreak seroprevalence of 42% in 2015 (Gallian et al., 2017). In Salvador, Northeastern Brazil, a serosurvey in 2016 of prospectively sampled individuals including microcephaly and non-microcephaly pregnancies, HIV-infected patients, tuberculosis patients, and university staff, found a post-outbreak seroprevalence of 63% (Netto et al., 2017). Another study in Salvador, conducted in a long-term health cohort, also found a post-outbreak seroprevalence of 63% (Rodriguez-Barraquer et al., 2019). Finally, in paediatric and household cohort studies in Managua, Nicaragua, ZIKV seroprevalence was estimated to be 46% in households following the outbreak in 2016 (Zambrana et al., 2018).

It has been suggested that infection with ZIKV confers immunity that lasts several years; if so, the high level of seroprevalence in affected countries may reflect sufficient herd immunity for the current ZIKV epidemic to be over in many locations, with the virus unable to re-emerge for decades to come (Kucharski et al., 2016; O'Reilly et al., 2018; Netto et al., 2017; Ferguson et al., 2016). Recent evidence suggests that neutralizing antibodies can distinguish between ZIKV and dengue virus (DENV) – a closely related Flavivirus – and that the immune response following ZIKV infection can persist over a year (Montoya et al., 2018; Griffin et al., 2019). It has also been suggested that primary ZIKV infection may confer protective immunity (Osuna et al., 2016). However, ZIKV serosurveys conducted at the end of the outbreak in French Polynesia and 18 months later found a drop in seroprevalence in the Society Islands, the archipelago where over 85% of the inhabitants of French Polynesia reside (Aubry et al., 2017). Therefore, the long-term antibody response following a ZIKV outbreak remains unclear.

Here, we explore short- and long-term seroprevalence against ZIKV as well as neutralizing responses against ZIKV following two ZIKV outbreaks in the Pacific region. We compared results from five serological surveys in the Society Islands, French Polynesia, over a seven-year period, and three serial serological surveys in the same cohort of individuals in Central Division, Fiji, over a four-year period. These surveys span the pre- and post- outbreak period in each country, allowing us to examine temporal changes in antibody responses following a ZIKV outbreak.

Results

In French Polynesia, seroprevalence of IgG antibodies against domain III of the ZIKV envelope glycoprotein in blood donors recruited before October 2013 was <1% (0.3–2%), which confirmed that the virus had not previously circulated in the population (Table 1). Analysis of samples collected in the general population of the Society Islands of French Polynesia after the emergence of ZIKV showed a decrease in ZIKV seroprevalence from 37% (26–47%) to 22% (16–28%) between February-March 2014 and September-November 2015 (chi-squared test, p=0.03). In Fiji, analysis of the serum samples serially collected from a cohort of participants in the Central Division showed an increase in ZIKV seroprevalence from 6.3% (3.3–11%) in October-November 2013 to 24% (18–31%) in November 2015 (chi-squared test, p<0.0001), and then a decrease to 12% (7.9–18%) by June 2017 (chi-squared test, p=0.005). In this cohort, based on IgG results tested by microsphere immunoassay (MIA), 6 of the 189 participants seroconverted (from negative to positive) and 28 seroreverted (from positive to negative) to ZIKV between 2015 and 2017 (McNemar’s test, p=0.0003).

Table 1
Seroprevalence of ZIKV among participants in five serological surveys in French Polynesia and three serological surveys in Fiji, conducted between July 2011 and June 2018.
DateCountryPopulation and assay usedAge range (median)Total no. seropositive/total no. testedSeroprevalence % [95% CI]
French Polynesia - General Population
Jul 2011-Oct 2013Society Islands, French PolynesiaBlood donors, ELISA18–75 (36)5/5930.8 [0.3–2.0]
Nov 2013First confirmed local transmission of ZIKV in French Polynesia
Feb-Mar 2014Society Islands, French PolynesiaGeneral, ELISA13–77 (47)18/4937 [26-47]*
Sep-Nov 2015Society Islands, French PolynesiaGeneral, MIA4–88 (43)154/70022 [16-28]*
French Polynesia - schoolchildren
May-Jun 2014Society Islands, French PolynesiaSchool children,
ELISA
6–16 (11)312/47666 [60-71]*
Jun-2018Society Islands, French PolynesiaSchool children,
MIA
6–16 (11)291/45764 [58-69]*
Fiji
Oct-Nov 2013Central Division, FijiGeneral, MIA2–78 (24)12/1896.3 [3.3–11]
Jul 2015First confirmed local transmission of ZIKV in Fiji
Nov-2015Central Division, FijiGeneral, MIA4–80 (26)45/18924 [18-31]
Jun-2017Central Division, FijiGeneral, MIA6–82 (28)23/18912 [7.9–18]
  1. * CIs were calculated taking into account the cluster sampling design (Aubry et al., 2017) and using the Fisher exact test.

    MIA – microsphere immunoassay.

To investigate possible factors influencing the decline in seroprevalence, we compared the seroprevalence profiles in children (defined as ≤16 years) and adults (>16 years) in both settings (Table 1 and Figure 1). In French Polynesia, although ZIKV seroprevalence declined in the general population from the Society Islands over 18 months, there was no evidence of a significant decline in seroprevalence in two serosurveys conducted four years apart in schoolchildren aged 6 to 16 years, with 66% (60–71%) positive in 2014 and 64% (58–69%) in 2018 (chi-squared test, p=0.6) (Table 1). When stratifying the general population from the Society Islands by age (≤16 years and >16 years), there was a decline in adults in the two consecutive cross-sectional studies conducted in 2014 and 2015, from 35.4% (22.2–50.5%) to 21.3% (18.2–24.5%) (Figure 1). A decline in adults was still observed, albeit with larger uncertainty, when the two datasets were standardised according to the age distribution of the population, with age-adjusted seroprevalence decreasing from 32.0% (16.7–62.1%) to 26.0% (20.1–33.9%) (Table 2).

Figure 1 with 4 supplements see all
Dynamics of ZIKV seroprevalence following outbreaks in Fiji and French Polynesia.

(A) Seroprevalence by MIA in Fiji. Red, seroprevalence and 95% confidence intervals for children (aged ≤16 years). Orange, seroprevalence and 95% confidence intervals for adults (aged >16 years). Solid lines, trends in data collected from the same individuals. Dotted line indicates the first confirmed ZIKV case. (B) Epidemiological dynamics in Fiji between 2013 and 2018. Coloured bars show number of PCR-confirmed samples of different DENV serotypes and ZIKV in Fiji; black lines show reported prolonged fever in Fiji from the Pacific Syndromic Surveillance System (World Health Organization, 2019). There was a major outbreak of DENV-3 outbreak in 2013–14 (Kucharski et al., 2018a) with a smaller DENV-2 outbreak in early 2017 (Aubry et al., 2019). (C) Seroprevalence by MIA in French Polynesia. Dashed lines, trends in seroprevalence between population representative cross-sectional surveys. Note that the pre-outbreak samples were collected between July 2011 and October 2013; for brevity, the latest possible collection date is used in the plot. (D) Epidemiological dynamics in French Polynesia between 2013 and 2018. Solid black line shows reported symptomatic dengue cases; dashed lines showed reported symptomatic Zika cases. In French Polynesia, between the sampling periods, there were no reported DENV outbreaks for serotypes 2,3,4, and there was hyper-endemic DENV-1 circulation. In April 2019, a DENV-2 outbreak was declared, the first since 1997 (Aubry et al., 2019).

Table 2
Age-adjusted seroprevalence by MIA in participants aged over 16 in the general population of the Society Islands in French Polynesia, based on serosurveys conducted in 2014 (n = 48) and 2015 (n = 672).
Virus2014 seroprevalence (95% CI)2014 age-adjusted seroprevalence (95% CI)2015 seroprevalence (95% CI)2015 age-adjusted (95% CI)
DENV185 (72–94)83 (55–100)80 (77–83)80 (71–91)
DENV248 (33–62)50 (28–87)19 (16–22)21 (15–21)
DENV375 (60–86)72 (47–100)56 (52–60)55 (48–64)
DENV463 (47–76)65 (40–100)42 (38–46)45 (38–54)
ZIKV35 (22–50)32 (16–62)21 (18–25)26 (20–34)
  1. *chi-squared test comparing 2014 bootstrap estimates with 2015 results.

In Fiji, in the subset of individuals who were aged over 16 years (n = 122), there was a decrease in seroprevalence by MIA from 24% (17–33%) in 2015 to 7.3% (3.4–13%) 2017 (Figure 1). There were two seroconversions in the collected samples over this period but 23 seroreversions (McNemar’s test, p<0.0001) (Table 3). In contrast seroprevalence in participants aged 16 and under (n = 67) remained relatively stable over this period (Figure 1), with four seroconversions and five seroreversions (McNemar’s test, p=1) (Table 3).

Table 3
Detection of IgG by MIA against ZIKV in the paired samples from participants aged under and over 16 years recruited during October-November 2015 and June 2017 in the Central division in Fiji (n = 189).

Age groups are defined using age of participants when recruited to the study in 2013.

20152017
≤16 years>16 yearsTotal participants
ZIKV+ZIKV-ZIKV+ZIKV-ZIKV+ZIKV-
≤16 years
ZIKV+105
ZIKV-448
>16 years
ZIKV+723
ZIKV-290
Total Participants
ZIKV+1728
ZIKV-6138

In order to assess whether the decline in ZIKV seroprevalence was also observed for other circulating Flaviviruses, the MIA seroprevalence pattern against each of the four DENV serotypes was analyzed in both countries, by age group (Figure 1—figure supplements 14). In Fiji, seroprevalence for DENV-1, DENV-2 and DENV-4 increased in participants in both age groups between 2013 and 2017 (Figure 1—figure supplements 1, 2 and 4). DENV-3 seroprevalence also increased in both age groups between 2013 and 2015 following an outbreak in 2013–14 (Kucharski et al., 2018a) and then declined in 2017 from 44% (32–57%) to 40% (28–52%) in children (McNemar’s test, p=0.6) and from 59% (50–68%) to 49% (40–58%) in adults (McNemar’s test, p=0.01) (Figure 1—figure supplement 3). In French Polynesia between 2014 and 2018, seroprevalence in children aged under 16 years showed no evidence of a change for DENV-1 and DENV-2 (chi-squared test, p=0.1917 and p=1, respectively) (Figure 1—figure supplements 12) and decreased for DENV-3 and DENV-4 (chi-squared test, p<0.0001 and p=0.0085, respectively) (Figure 1—figure supplements 34). In adult participants from the general population, seroprevalence for all four DENV serotypes declined between 2014 and 2015.

The age-adjusted values for seroprevalence by MIA for the four DENV serotypes were similar to the raw values (Table 2), suggesting that the decline in French Polynesia could not be explained by differences in sampling by age. However, a higher proportion of the samples in 2014 tested positive by MIA for all four DENV serotypes (Table 4), suggesting that the sampling included a group at higher risk for arbovirus infection than those sampled in 2015. To check that the estimated decline in ZIKV seroprevalence was not an artefact of this sampling bias, we re-estimated seroprevalence for the four DENV serotypes and ZIKV using a bootstrap sample of the 2014 responses, with replacement, weighted by the DENV exposure profile (excluding the virus of interest) in the 2015 survey so that the bootstrap sample of the 2014 responses had a similar DENV exposure profile as in the 2015 responses. For example, when generating bootstrap estimates for DENV-1 in 2014, we resampled participants based on the distribution of number of exposures to DENV-2, DENV-3, and DENV-4 in the 2015 data (Table 5). After adjusting for prior exposure, there was no significant decline in seroprevalence for DENV-1, DENV-3, or DENV-4, which had all circulated in the five years preceding the 2014 data collection, whereas the decline in ZIKV was still present (chi-squared test, p=0.0047).

Table 4
Age distribution and profile of DENV exposure history in two cross-sectional surveys conducted in the general population from the Society Islands, French Polynesia, in 2014 and 2015.

While the age distribution is similar in both studies, the sample in 2014 has a higher proportion of individuals who have tested positive for infection from all four DENV serotypes by MIA.

Variable2014 (= 49)2015 (= 700)
Age distribution (median [IQR])47 [29-56]43 [29-57]
Number of DENV serotypes positive at time of sample collection (n [%])
03 [0.061]118 [0.17]
16 [0.12]163 [0.23]
211 [0.22]159 [0.23]
311 [0.22]154 [0.22]
418 [0.37]106 [0.15]
Table 5
Bootstrap estimated seroprevalence for each of the four DENV serotypes and ZIKV adjusted for sampling bias in two cross-sectional surveys conducted in the general population from the Society Islands, French Polynesia, in 2014 and 2015.

Results from the cross-sectional surveys in the Society Islands, French Polynesia, in 2014 and 2015 show a decline in seroprevalence by MIA against all 4 DENV serotypes and ZIKV. However, the 2014 sample included more individuals that tested positive for >1 DENV serotype and are assumed to be a higher risk group. We used a bootstrap method with 10,000 iterations which estimated seroprevalence from a sample of the 2014 dataset, taken with replacement, weighted by the exposure distribution to other DENV viruses in the 2015 survey. After adjusting for the sample bias, there was no evidence of a decline in seroprevalence for DENV-1, DENV-3, or DENV-4, which had circulated in the years preceding the 2014 sample collection (World Health Organisation, 2019), but there remained strong evidence that ZIKV seroprevalence declined between 2014–15.

Virus2014 seroprevalence (95% CI)
(n = 49)
2014 bootstrap estimates of seroprevalence (95% CI)2015 seroprevalence (95% CI)
(n = 700)
p-value*
DENV186 (73–94)74 (61–86)80 (77–83)0.36
DENV247 (33–62)38 (24–53)18 (15–21)0.0008
DENV376 (61–87)64 (51–78)55 (51–59)0.21
DENV463 (48–77)50 (37–65)42 (38–46)0.42
ZIKV37 (23–52)42 (29–55)22 (19–25)0.0047
  1. *chi-squared test comparing 2014 bootstrap estimates with 2015 results.

To explore dynamics of antibody waning at the individual level, we performed neutralization assays (NT) on a subset of 45 participants from Fiji for whom sufficient sera were available to test against ZIKV from all three collection periods, focusing on those who were seropositive to ZIKV by MIA in 2013 or 2015. We found that in the 31 individuals who were ZIKV seronegative by NT (i.e. log titre <2) in 2013 and had a rise in log titre ≥2 against ZIKV between 2013 and 2015, anti-ZIKV antibody responses waned significantly in 2017, with an average decline in log titre of −1.94 (t-test, p<0.0001) (Figure 2A and Table 6). In total, four participants seroreverted between 2015 and 2017; all had a log titre of 4 against ZIKV in 2015. We observed a similar effect when we analysed all participants who had a rise in log titre of at least 2 between 2013–15, regardless of serostatus in 2013 (Figure 2—figure supplement 1).

Figure 2 with 2 supplements see all
Waning of neutralizing antibody responses against ZIKV and DENV-3 in Fiji for participants who were seronegative to each virus in 2013 and seroconverted in 2015.

(A) Histogram of change in neutralization assay log titre against DENV-3 (= 19) and ZIKV (n = 31) between 2015–2017 for individuals who seroconverted to these respective viruses between 2013–2015 (i.e. log titre < 2 in 2013 and log titre ≥ 2 in 2015). (B) Histogram of change in log titre against DENV-3 and ZIKV for individuals who seroconverted to these respective viruses during 2013–2015.

Table 6
Change in neutralization titre between 2013–2017 in a cohort of 45 study participants in Fiji.

ZIKV and DENV-3 both circulated in Fiji between the collection of samples in 2013 and 2015, with ZIKV first reported in July 2015 and DENV-3 circulating between October 2013 and January 2015. Neutralization titre levels rose significantly over this period. Between 2015 and 2017, DENV-3 titre levels still increased with a mean change in tire of 0.89. By contrast, the mean change in ZIKV titre over this period decreased (−1.9).

Virus2013–2015 change,
Mean [95% CI]
p-value*2015–2017 change,
Mean [95% CI]
p-value*
ZIKV (n=31)5 [4.5, 5.5]<0.0001−1.9 [-2.4,–1.5]<0.0001
DENV3 (n=19)3.4 [2.9, 3.9]0.89 [0.046, 1.7]
  1. * t-test comparing change in neutralization titre for ZIKV and DENV-3 between 2013–2015, and 2015–2017.

To test whether the dynamics of anti-ZIKV antibody waning were different from the responses to DENV infection, we compared results for ZIKV to the neutralization response following a DENV-3 infection in the same cohort from Fiji. There was a large DENV-3 epidemic during 2013–14 in Fiji (Osuna et al., 2016), which meant most seroconversions to DENV-3 occurred between the collection of samples in 2013 and 2015. In those individuals that seroconverted to DENV-3 (n = 19) or ZIKV (n = 31) between 2013 and 2015, the initial rise in NT log titres against ZIKV was larger than for DENV-3, with a mean change of 5.0 and 3.37 respectively (Figure 2B and Table 6). All individuals who had seroconverted to DENV-3 remained seropositive to the virus in 2017, while four individuals who had seroconverted to ZIKV were seronegative in 2017. Although the NT log titres increased by a mean of 0.89 for DENV-3 between 2015 and 2017 (two-sided t-test, p=0.04), log titres against ZIKV declined by a mean of 1.94 over the same period (two-sided t-test, p<0.001) (Figure 2A and Table 6).

In Fiji, there was a delay of around 18 months between the end of the 2013–14 DENV-3 epidemic and collection of samples in 2015. As DENV titres can wane following infection, particularly in individuals with a prior DENV exposure (Clapham et al., 2016), titres against DENV-3 in Fiji may therefore have had more time to wane and reach a stable persistent level than titres against ZIKV, which may have circulated later than DENV-3. We therefore analysed changes in titre for participants who were initially seronegative to DENV-1 and DENV-2, which were circulating at low levels in Fiji between the two serological surveys in 2013 and 2015 (Figure 1). As with DENV-3, we found no evidence of a subsequent overall decline during 2015–17 for those participants who seroconverted to DENV-1 or DENV-2 during 2013–15 (Figure 2—figure supplement 2).

Of the 45 participants tested by neutralization assay, nine were initially seropositive to ZIKV by NT in 2013. Fitting a generalized additive model to these data, we found that higher baseline mean NT log titres against DENV were associated with an increased probability of seropositivity to ZIKV (Figure 3A). In contrast, higher baseline mean DENV titres were not associated with increased seropositivity by MIA in 2013. There was little difference between the assay results in the 2015 samples (Figure 3B), but we did find evidence of a difference in the 2017 results, with 15/45 participants positive by MIA and 31/45 positive by NT. This difference was associated with participants’ 2013 DENV titres: those with intermediate DENV titres in 2013 had a significantly lower probability of being seropositive in the MIA in 2017 compared to NT (Figure 3C).

Relationship between mean DENV log neutralization titre across the four serotypes in 2013 and ZIKV seroprevalence using different assays, in a subset of 45 participants.

(A) Seroprevalence by MIA, shown in grey, and neutralization test (NT), shown in orange, for sera collected in 2013. Line shows prediction from GAM fitted to each dataset, with shaded region showing 95% CI, and points show raw data. (B) Seroprevalence for sera collected from the same participants in 2015. (C) Seroprevalence for sera collected from the same participants in 2017.

Discussion

Analyzing data from serological surveys conducted in French Polynesia and Fiji at different time points after the first reported autochthonous ZIKV transmission, we found evidence of a decline in ZIKV seroprevalence. The high number of participants from the Fijian cohort that seroreverted between 2015 and 2017 suggested that anti-ZIKV antibody levels waned in these individuals to the point that they were no longer detectable by MIA. Using a neutralization assay to test longitudinal sera collected in Fiji, we found that the mean change in neutralizing antibody titres against ZIKV also decreased significantly between 2015 and 2017, showing that individual-level antibody titres against ZIKV as well as overall seroprevalence decreased over time. In contrast, over the same period, neutralizing antibody titres against DENV-3, a closely related Flavivirus which caused a large epidemic in Fiji in 2013–2014 (Kucharski et al., 2018a), remained stable.

In both countries we found seroprevalence against ZIKV in individuals aged over 16 declined over the two-year period following an outbreak, while the overall level of seroprevalence persisted in children. This pattern was unique to ZIKV compared to DENV in both countries. It is possible that this is related to the DENV immunological profile of individuals, given that the older population is likely to have experienced more DENV infections over their lifetime. If an individual has experienced prior DENV infections, high numbers of weakly neutralizing cross-reactive B cells may outcompete naïve B cells for ZIKV antigen (Midgley et al., 2011), leading to a short-term boost in antibody response against ZIKV following ZIKV infection (Robbiani et al., 2017) but not a persistent specific response; a similar phenomenon has been observed for other antigenically variable viruses like influenza (Kucharski et al., 2018b). In the 2017 samples, more participants remained seropositive in the neutralization assay – which measures the overall ability of sera to neutralize ZIKV – than in the MIA, which tests for IgG antibodies against domain III of the envelope glycoprotein. This difference was greatest for participants who had intermediate baseline titres to DENV in 2013 (Figure 3C), which would support the hypothesis that prior DENV exposure may result in a detectable short-term specific response against ZIKV following ZIKV infection (as measured by MIA), but not a persistent specific response.

To our knowledge, the only other study to date that has investigated the long-term persistence of neutralizing antibodies against ZIKV was conducted in 62 residents of Miami (Florida, USA), who had a confirmed ZIKV infection in 2016 (Griffin et al., 2019). This cross-sectional study found that all participants had neutralizing antibodies against ZIKV 12–19 months after infection. This study also found that at least 37% of the participants had no evidence of past DENV infection, which is consistent with the hypothesis that anti-ZIKV immune responses may persist longer in populations that have had less exposure to DENV. More data are therefore needed to test hypotheses about the potential impact of pre-existing anti-DENV immune response on anti-ZIKV antibody waning.

Although we found evidence of a decline in seroprevalence for antibodies against domain III of the envelope glycoprotein, as well as waning neutralizing antibody responses following two ZIKV outbreaks, the implications for susceptibility to future ZIKV infection remain unclear. Given the antigenic similarity of DENV and ZIKV (Priyamvada et al., 2016), it is commonly assumed that the immune response to ZIKV infection will be similar to that following DENV infection. High levels of neutralizing antibodies to DENV have been shown to correlate with protection from symptomatic infection (Katzelnick et al., 2016). Moreover, infection with a single DENV serotype can confer lifelong immunity to the infecting serotype as well as a transient period of cross-neutralization against heterologous serotypes (Wahala and Silva, 2011). However, it is unclear in the context of ZIKV what the relationship is between a specific titre value and susceptibility to further infection. A key aim for future work will be to establish how waning antibody levels as measured by MIA and neutralization assays may impact protective immunity, and hence susceptibility to reinfection in populations that have already experienced transmission of ZIKV.

There are some additional limitations to our analysis. First, we did not have reverse transcription polymerase chain reaction (RT-PCR) confirmation of ZIKV infection in individuals sampled in this study. We have presented analysis of representative serological surveys in two locations with known, RT-PCR-confirmed ZIKV outbreaks (Mallet et al., 2015; World Health Organisation, 2015). However, RT-PCR confirmation for ZIKV at the individual level remains difficult to obtain, in particular from blood samples, and there have been relatively few confirmations globally compared to the number of suspected cases (Ferguson et al., 2016), let alone analysis of long-term antibody dynamics in RT-PCR confirmed patients. In French Polynesia, there were approximately 32,000 reported clinical cases of ZIKV infection, but only 297 documented RT-PCR-confirmed cases (Mallet et al., 2015). As a result, antibody responses in RT-PCR-confirmed cases may not necessarily be representative of immune responses against ZIKV in the wider population, particularly following asymptomatic infection. Although MIA seropositivity in our study was defined using control sera collected over a year after RT-PCR-confirmed infection, our results suggest that this threshold may not detect long-term waning responses in individuals who had unreported, and likely less severe, infections.

Our analysis was also limited by study design. In French Polynesia, surveys were cross-sectional, so we were unable to examine temporal antibody dynamics at the individual level. However, both cross-sectional studies of the general population were conducted using population representative cluster sampling (Aubry et al., 2017) in the same remote island locations with stable population composition, which enabled robust comparisons of overall seroprevalence. We did identify one potential source of sampling bias with different DENV exposure profiles in the two surveys, but our conclusions of declining seroprevalence for ZIKV persisted once we adjusted for this bias. We also used a different serological testing method between the studies in French Polynesia in 2014 and 2015. However, both used the same recombinant antigens and it has been shown that there was good agreement between ELISA and MIA in the 2014 samples (see Materials and methods). In Fiji, a strength of our study was the collection of longitudinal samples from the same individuals at three time points. However, our sample size was limited given the logistical challenge of recontacting participants twice over a four-year period. These data provided strong evidence that ZIKV seroprevalence declined over the two-year period following first reports of circulation, but our sample size was insufficient to fully explore the potential effect of anti-DENV pre-existing immunity on anti-ZIKV antibody waning once we stratified individuals by previous DENV exposure. Although the outbreaks of DENV-3 in Fiji and ZIKV in French Polynesia were well-documented and occurred over a relatively brief period of time (Figure 1), it was harder to identify the likely time of infection for other viruses – such as ZIKV in Fiji or DENV in French Polynesia – in our study populations. Several participants in Fiji were seropositive to ZIKV by neutralization assay (NT) in 2013, but this result may be influenced by cross-reaction; participants who had high pre-existing titres to DENV in 2013 were more likely to be seropositive by NT (Figure 3A). In our main analysis of titre dynamics, we therefore focused on the subset of participants who were seronegative by NT in 2013 (Figure 2). However, we obtained the same conclusion when participants who were initially seropositive were also considered (Figure 2—figure supplement 1).

The global ZIKV epidemic began in the Pacific islands in 2013 before spreading in Central and South America from 2015. Seroprevalence studies following ZIKV epidemics in Latin America have been reported but data have either been non-representative (Netto et al., 2017) or not enough time had elapsed since the outbreak to observe long-term dynamics (Rodriguez-Barraquer et al., 2019; Zambrana et al., 2018). To our knowledge, these are the first studies of community seroprevalence over a long-term period following a ZIKV outbreak. Therefore, patterns observed in Pacific islands may be an early indication of what might happen to seroprevalence in Latin America where ZIKV outbreaks began two to three years after the French Polynesia epidemic (Cao-Lormeau et al., 2014; Bogoch et al., 2016).

In the short-term, our findings have implications for the design of follow up studies of ZIKV. Our results provide evidence that levels of seroprevalence one to two years following ZIKV circulation may be lower than previously expected and study designs may need to be adapted to reflect this, particularly in settings that exhibit long-term low level circulation of ZIKV as opposed to large sporadic outbreaks (Ruchusatsawat et al., 2019). For example, estimates of microcephaly risk may be inflated if derived from long-term seroprevalence data that underestimate the true extent of infection within the population, and results of clinical trials could also be biased if post-outbreak seroprevalence is used an indicator of infection within a population (Cohen, 2018). In the longer-term, our results demonstrate the value of longitudinal serological studies of flaviviruses, and analysis using multiple serological tests, including neutralization assays (Clapham et al., 2016). Such studies will be essential to understand different aspects of the short and long-term immune antibody response against ZIKV, and how prior exposures to DENV may influence these responses.

Materials and methods

Study location and participants

French polynesia

Request a detailed protocol

Four separate ZIKV serosurveys were previously conducted in the Society Islands (Table 1). As reported previously (Aubry et al., 2017; Aubry et al., 2015a), a first serosurvey (n = 593) was conducted in adult blood donors recruited between July 2011 and October 2013, before the ZIKV outbreak that occurred between October 2013 and April 2014 (Cao-Lormeau et al., 2014). Two population-representative serosurveys were conducted among the general population, firstly between February and March 2014 (n = 196), and then between September and November 2015 (n = 700). The two studies in the general population both spanned a range of adult age groups (Table 7). An additional serosurvey was conducted among schoolchildren between May and June 2014 (n = 476). Finally, a fifth serosurvey was conducted among schoolchildren in the Society Islands in June 2018 (n = 457) using the same protocol as in 2014 (Aubry et al., 2017).

Table 7
Age distribution of study population in French Polynesia.

Overall population distribution shown, along with total samples collected in each age group in 2014 and 2015 serosurveys.

Age rangePopulation estimate (2017)Samples in 2014 studySamples in 2015 study
0–942,77000
10–1943,705322
20–2948,91410135
30–3942,1445131
40–4940,8868119
50–5934,47815128
60–6921,099285
70–7910,481546
80–89377309
90+41600

Fiji

Request a detailed protocol

Three serosurveys were conducted in Fiji (Table 1). Individuals were first recruited into a population-representative community-based typhoid/leptospirosis seroprevalence study between September and November 2013 (Watson et al., 2017) (n = 1,787), before autochthonous transmission of ZIKV was first detected in July 2015 (World Health Organisation, 2015). Briefly, nursing zones were randomly selected, from which one individual from 25 households in a randomly selected community was recruited. Participants who had consented to being contacted again for health research were subsequently recruited in November 2015 in 23 communities in Central Division through last known addresses, phone numbers and the assistance of local nurses (n = 327) (Kama et al., 2019). A third follow-up serosurvey was conducted in June 2017 using the same protocol as in 2015 (n = 321) (Kucharski et al., 2018a). Follow-up surveys were only performed in Central Division, which was the focus of a DENV-3 outbreak in 2013–14 (Kucharski et al., 2018a). Only blood samples serially collected from the same participants (n = 189) in 2013, 2015 and 2017 were analyzed in the main results presented in this study.

Informed consent and ethics approvals

French polynesia

Request a detailed protocol

The five serosurveys were approved by the Ethics Committee of French Polynesia (ref 61/CEPF 08/29/2013, 60/CEPF 06/27/2013, 74/CEPF 05/04/2018, and 75/CEPF 05/04/2018).

Fiji

Request a detailed protocol

The original 2013 study, and the 2015 and 2017 follow up studies were approved by the Fiji National Research Ethics Review Committee (ref 2013–03, 2015.111.C.D, 2017.20.MC) and the London School of Hygiene and Tropical Medicine Observational Research Ethics Committee (ref 6344, 10207, 12037).

Serological analysis

French polynesia

Request a detailed protocol

Serum samples collected from blood donors between July 2011 and October 2013 and samples collected from the general population and schoolchildren in 2014 were all tested for presence of IgG antibodies against ZIKV and each of the four DENV serotypes using a recombinant antigen-based indirect ELISA as reported previously (Aubry et al., 2017; Aubry et al., 2015a). Samples collected from the general population in 2015 and from schoolchildren in 2018 were tested by microsphere immunoassay (MIA) using the same recombinant antigens as for the ELISA (Cao-Lormeau et al., 2016; Aubry et al., 2017; Kama et al., 2019). Recombinant antigens used in both assays comprised domain III of the envelope glycoprotein of ZIKV, DENV-1, DENV-2, DENV-3, or DENV-4 strains (respective GenBank accession no. KJ776791, AF226686.1, FM986654, FJ44740.1, FM986672.1) and were produced using the Drosophila S2 expression system (Life Technologies, USA) as previously detailed (Aubry et al., 2015b). Serostatus was defined by a cut-off determined using positive and negative control sera analyzed by ROC curve. The sensitivity and specificity of the MIA assay were respectively 100% and 100% for DENV-1, 89.5% and 97.1% for DENV-2, 100% and 100% for DENV-3, 96.9% and 100% for DENV-4, and 79.6% and 94.9% for ZIKV. The positive control sera for ZIKV was collected 13 months after RT-PCR confirmed infection. In the serosurvey conducted among the general population of the five archipelagos in French Polynesia in 2014 (Aubry et al., 2017), 196 samples were tested using both ELISA and MIA: among the 97 serum samples that tested positive for anti-ZIKV IgG by ELISA, 78 (80%) were also found positive by MIA; and among the 99 serum samples that tested negative for anti-ZIKV IgG by ELISA, 70 (71%) were also found negative by MIA. This produced a value of Cohen’s κ = 0.51 (Aubry et al., 2017).

Fiji

Request a detailed protocol

All serum samples collected in Fiji were tested using MIA to detect IgG antibodies against ZIKV and all four DENV serotypes as previously reported (Cao-Lormeau et al., 2016; Aubry et al., 2017; Kucharski et al., 2018a). To follow the evolution of antibody titres at the individual level, a subset of samples collected from the same individuals in 2013, 2015 and 2017 were tested for the presence of neutralizing antibodies against ZIKV and each of the four DENV serotypes using a neutralization assay as previously described (Cao-Lormeau et al., 2016). This subset of samples was selected to include all participants who were seropositive to ZIKV by MIA in 2013 and 2015, as well as one participant just below the seropositivity threshold, and for whom we had sufficient longitudinal serum available from 2013, 2015 and 2017 to test by neutralization assay (n = 45). We also tested samples from an additional 24 participants from the same cohort who were seropositive to ZIKV by MIA in 2013 or 2015 and for whom we had sufficient serum from 2013 and 2015 to test by neutralization assay, but no matched sample from the 2017 follow up survey (i.e. 69 paired samples in total). ZIKV log titres in the neutralization assay followed a bimodal distribution, which supported the use of a log titre of ≥2 as a cutoff for seropositivity (Figure 4).

Figure 4 with 3 supplements see all
Distribution of ZIKV neutralization titres in the Fiji serosurveys.

Results shown for 45 participants who had samples available from 2013, 2015, and 2017. Dashed line shows the threshold used to define seropositivity.

Of the 9/45 participants with three samples who were seropositive to ZIKV by neutralization assay in 2013, all were seropositive to at least one DENV serotype (Figure 4—figure supplement 1). To assess the potential for cross-reactive antibody responses, we examined the correlation between changes in log titre to different viruses between 2013 and 2015. Among the 20/69 paired samples that tested seronegative against all five viruses in 2013 and were re-tested in 2015, there was no evidence of an association between changes in ZIKV titre and changes in titre against any of the DENV serotypes, suggesting that the changes in ZIKV titre were unlikely to be strongly influenced by DENV cross-reaction (Figure 4—figure supplement 2). However, the 49/69 participants who had a pre-2013 DENV exposure and a large rise against ZIKV between 2013–15 tended to exhibit a smaller rise against DENV viruses (Figure 4—figure supplement 3).

A previous study, which tested serological samples from Fiji across three divisions (Kama et al., 2019), found that of the samples reactive by MIA, 66/83 (79.5%) exhibited neutralizing activity for ZIKV (κ = 0.71) and 109/112 (97.3%) for DENV (κ = 0.80). In this study, we tested what proportion of samples for the 45 participants in the full dataset (i.e. 135 samples in total) that were seropositive or seronegative by MIA had the same result by the neutralization assay. We found that 54/68 (79.4%) samples that were positive to ZIKV by MIA were also positive in the neutralization assay, and 42/67 (62.7%) who were seronegative were also negative by neutralization assay (κ = 0.42). We also calculated the proportion positive by neutralization assay that had the same result by MIA. We found that 54/79 (68.4%) samples that were positive to ZIKV in the neutralization assay were also positive by MIA, and 42/56 (75%) who were seronegative were also negative by MIA.

Statistical analysis

Request a detailed protocol

For data from Fiji, where serial samples were collected from the same individual, changes in seroprevalence between studies were tested using McNemar’s test. In French Polynesia, chi-squared tests were performed to test for evidence of a change in seroprevalence between two cross-sectional surveys. Changes in mean log titre between groups were analyzed using a t-test. To analyse the potentially non-linear relationship between DENV neutralization titres and seroprevalence by MIA and neutralization test (Figure 3), we used a generalized additive model via the mcgv package in R (Wood, 2019). The model was of the form g(E(y))=b + f(x), where y was the binary outcome variable, x was the predictor (i.e. titre), g was the link function, b was the intercept, and f was a smooth function represented by a penalized regression spline. Mean DENV titre was calculated as the mean of log titres against the four DENV serotypes for each participant. All data and code used in the analysis are available at: https://github.com/a-henderson91/zika-sero-pacific/ (Henderson and Kucharski, 2019; copy archived at https://github.com/elifesciences-publications/zika-sero-pacific/settings).

References

  1. 1
  2. 2
  3. 3
  4. 4
  5. 5
  6. 6
  7. 7
  8. 8
  9. 9
  10. 10
  11. 11
  12. 12
  13. 13
  14. 14
  15. 15
  16. 16
  17. 17
  18. 18
  19. 19
  20. 20
    Bulletins d’informations sanitaires,épidémiologiques et statistiques No 13
    1. HP Mallet
    2. AL Vial
    3. D Musso
    (2015)
    Direction De La Santé.
  21. 21
  22. 22
  23. 23
  24. 24
  25. 25
  26. 26
  27. 27
  28. 28
  29. 29
  30. 30
  31. 31
  32. 32
  33. 33
    mgcv R package verion 1.8-29
    1. S Wood
    (2019)
    mgcv R package verion 1.8-29, https://CRAN.R-project.org/package=mgcv.
  34. 34
  35. 35
  36. 36
  37. 37
  38. 38
  39. 39

Decision letter

  1. Isabel Rodriguez-Barraquer
    Reviewing Editor; University of California, San Francisco, United States
  2. Neil M Ferguson
    Senior Editor; Imperial College London, United Kingdom
  3. Isabel Rodriguez-Barraquer
    Reviewer; University of California, San Francisco, United States
  4. Leah Katzelnick
    Reviewer; University of California, Berkeley, United States

In the interests of transparency, eLife publishes the most substantive revision requests and the accompanying author responses.

Thank you for submitting your article "Zika seroprevalence declines and neutralizing antibodies wane in adults following outbreaks in French Polynesia and Fiji" for consideration by eLife. Your article has been reviewed by three peer reviewers, including Isabel Rodriguez-Barraquer as the Reviewing Editor and Reviewer #1, and the evaluation has been overseen by Neil Ferguson as the Senior Editor. The following individual involved in review of your submission has agreed to reveal their identity:); Leah Katzelnick (Reviewer #3).

The reviewers have discussed the reviews with one another and the Reviewing Editor has drafted this decision to help you prepare a revised submission.

Summary:

Henderson and Aubry et al. have performed a comparative study of cross-sectional and longitudinal serosurveys conducted before, soon after, and multiple years after the arrival of ZIKV to French Polynesia and Fiji. They use multiple serological assays to measure ZIKV-specific antibodies and DENV-specific antibodies. Specifically, they find an overall decline in ZIKV-specific seroprevalence by ~2 years after the outbreaks in each country, which was primarily driven by a decline in adults (defined as >16 years of age). Children appeared to maintain seroprevalence. They also use longitudinal data from a small subset of individuals from Fiji to show waning in neutralizing antibody titers as well.

These results are worth publishing, but we have several concerns about their presentation and interpretation. In particular, it is not clear to us whether the waning captured by the ELISA/MIA assay is meaningful (as evidence of waning protection) or simply reflects the kinetics of the specific antibody response (anti-domain III) measured by this assay. Data from neutralization assays, the gold standard, is unfortunately scarce (n = 45), hard to interpret and not very conclusive because of the timing of sample collection (some waning is expected) and because the evidence of seroreversion is quite weak (only observed in four individuals with weak seroconversions).

Essential revisions:

1) The authors speculate that the reductions in ZIKV seroprevalence measured by ELISA/MIA could indicate waning population immunity to ZIKV. This would have profound implications for future ZIKV dynamics in these and other populations, as the authors discuss. However, we think that another possibility, that is more likely, is that the observed decreases might just reflect the kinetics of antibody responses measured by this specific assay (antibodies against domain III). As far as we understand, it hasn't yet been established whether this antibody response is a good marker of historical exposure, or just a marker of recent exposure. DENV seroprevalence also seems to wane (Supplementary Figure 1). In the absence of additional data describing the performance/kinetics of this assay, this potential explanation needs to be discussed.

2) More details should be provided about the serological assays used. Two references are provided about the ELISA/MIA assays but they also provide very little information. At the very least, the paper should be explicit that the assay uses recombinant antigens comprising domain III of the envelope glycoprotein. Ideally, data on performance characteristics of the assay should also be reported as this information is crucial for the interpretation of results.

3) Related to the point above, please include more detail on the agreement between the MIA and ELISA tests. As far as I can see reference 11 does not contain enough information to assess this. It states: "80% were positive by both tests". Were all the rest of the samples negative by both tests?

4) While ZIKV seroprevalence is observed to wane in both populations (Figure 1), DENV seroprevalence dynamics differ between the populations. It is not clear why, and how this relates to the observed decay for ZIKV seroprevalence. The results seem to be affected by the historical intensity of DENV transmission in French Polynesia and Fiji and may have been affected by DENV or ZIKV transmission that occurred after the main ZIKV epidemics. We recommend that Supplementary Figure 1 accompany Figure 1 in the main text. In Fiji, following the ZIKV epidemic, titers to DENV appeared to increase (except to DENV3). In French Polynesia, there was no DENV immunity to the four DENV before the arrival of ZIKV, and DENV titers did decline in adults after the ZIKV epidemic. Basically, it would be helpful to the reader if the authors include a figure on the epidemics of DENV and ZIKV in these populations before and after the ZIKV epidemic, ideally with serotype-specific epidemic curves if such data are available, to be able to interpret these results.

5) The age-standardised seroprevalence is the important metric here. As the authors note there is less decline in this metric. This should be the main result, as it seems that the other comparisons are not informative at all-as far as I can determine the differences are simply driven by the age distributions of the population sampled. Please explain if this is not the case.

Longitudinal data from Fiji tested using neutralization assays could certainly have added strength to the hypothesis of waning population immunity, but unfortunately the data is scarce (n = 45) and hard to interpret for multiple reasons:

6) In order to be able to compare the DENV and ZIKV results (Figure 2), it would be crucial to understand when sample collection occurred with respect to ZIKV and DENV outbreaks. At the end of the Results, from what we understand: the DENV outbreak was before the ZIKV outbreak, so the observed differences in decline rates between DENV and ZIKV could be explained by much of the DENV decline already happened before the measurement, whereas the ZIKV transmission was much closer to the time of the first measurement. Differences of just a few months could make a large difference here. This would be consistent with the lower increase observed from 2013-2015 for DENV compared to ZIKV. Similarly for the differences in seroreversions, there may have been individuals that were DENV positive closer to the transmission, but were not by 2015, and therefore would not have been captured by the measurements here. In addition, it would be useful to know if there was any DENV transmission in the 2015-2017 period?

7) Also, looking at the neutralization data in GitHub, it seems that 9/45 (20%) of individuals had positive (but low) titers to ZIKV in 2013 (titer greater of equal to 2). 7/9 of these positive values are in individuals who also have positive titers to 3 or more dengue serotypes. We find this concerning as this is supposed to be the pre-Zika time-point and this might suggest some cross-reactivity in this ZIKV PRNT assay. Has the sensitivity and specificity of this neutralization assay (and this cut-point) been calculated? More details need to be provided about this assay and its performance as well.

8) Related to the above, it would be good to show the correlation between neutralization and MIA results. I tried to do this myself (using the provided ids) but was only able to match the data of 16/45 individuals (the neut dataset includes 344 individuals, most with incomplete data). Unless there's a problem with the data (or provided IDs), the agreement seems poor. For example, these 16 individuals have positive neutralization titers in 2015 (>4 in the majority) and yet, only 5/16 are classified as positive by the MIA assay.

9) The four observed ZIKV seroreversions described in the last paragraph of the Results seem to be in individuals with relatively low titers at seroconversion (4). There are four additional "seroreversions" among individuals who were already positive in the 2013 sample. Are these seroreversions mainly capturing cross-reactive responses? It would be useful to show a figure (maybe supplementary) with the longitudinal neutralization titer data.

10) In light of these limitations/uncertainties, we think the manuscript needs to be reworked to emphasize the uncertainty regarding the meaning of these findings. We would emphasize the need for more studies using longitudinal data (rather than cross-sectional) data and a broader set of assays, including PRNTs. We would de-emphasize the discussion/conclusions around the implications of waning protection (last paragraph of the Discussion). While provocative, we are not convinced it's supported by the presented data.

11) It would be good to clarify which subset of samples were tested and the sample sizes that went into each analysis.

[Editors' note: further revisions were requested prior to acceptance, as described below.]

Thank you for submitting your article "Zika seroprevalence declines and neutralizing antibodies wane in adults following outbreaks in French Polynesia and Fiji" for consideration by eLife. Your article has been reviewed by a Reviewing Editor and Neil Ferguson as the Senior Editor.

The Reviewing Editor has drafted this decision to help you prepare a revised submission.

The authors have satisfactorily addressed most of our concerns. Thanks for providing the merged data! We have some additional comments:

1) The low agreement reported for the ZIKV MIA and PRNTs assays is a bit concerning, and not consistent with the sensitivity and specificity reported. There seem to be 65 samples positive according to the PRNT in 2017, and of these only 17 (26%) are positive according to the MIA. Assuming that the PRNT is the "gold standard", this would suggest a much lower sensitivity of MIA than was reported for old infections, suggesting time-dependent sensitivity (26% vs. 80%). If instead we question the performance of the PRNT (cross-reactivity?) then the waning results are also questionable. I think these discrepancies (and the difficulty interpreting results from these novel assays) should be explicitly discussed.

2) Related to the above, authors should explicitly state in the text that, while neutralization titers decayed over the observation period, there's little (or no) evidence of sero-reversion (decreasing seroprevalence) according to the neutralization data.

3) Please provide a reference for the sensitivities and specificities reported for the MIA assay. Also, please state whether these performances were evaluated using acute/early convalescent samples vs. samples from historical infections (infections occurring one or more years before).

4) I also find it concerning that 9/46 samples (~20%) were positive against ZIKV (low titers) according to the PRNT assay in 2013. The authors perform some analyses around cross-reactivity, but these are only described in the Materials and methods. I think cross-reactivity (and how it could impact the results) should be addressed in the Discussion.

[Editors' note: further revisions were requested prior to acceptance, as described below.]

Thank you for resubmitting your work entitled "Zika seroprevalence declines and neutralizing antibodies wane in adults following outbreaks in French Polynesia and Fiji" for further consideration by eLife. Your revised article has been evaluated by Neil Ferguson (Senior Editor) and a Reviewing Editor.

Most of our concerns have been satisfactorily addressed, but I have one remaining question.

- You report that the proportion seropositive (by MIA) was ~24% in Fiji in 2015, yet Figure 3B shows a seroprevalence of >90% among the 45 individuals that were also tested using neutralization assays. It could almost be seen as if individuals were selected for neutralization assays based on the MIA status (42/45 samples positive by MIA were tested) but the text reads that they were selected based on sample availability only. Please clarify how the samples were selected.

https://doi.org/10.7554/eLife.48460.sa1

Author response

Summary:

Henderson and Aubry et al. have performed a comparative study of cross-sectional and longitudinal serosurveys conducted before, soon after, and multiple years after the arrival of ZIKV to French Polynesia and Fiji. They use multiple serological assays to measure ZIKV-specific antibodies and DENV-specific antibodies. Specifically, they find an overall decline in ZIKV-specific seroprevalence by ~2 years after the outbreaks in each country, which was primarily driven by a decline in adults (defined as >16 years of age). Children appeared to maintain seroprevalence. They also use longitudinal data from a small subset of individuals from Fiji to show waning in neutralizing antibody titers as well.

These results are worth publishing, but we have several concerns about their presentation and interpretation. In particular, it is not clear to us whether the waning captured by the ELISA/MIA assay is meaningful (as evidence of waning protection) or simply reflects the kinetics of the specific antibody response (anti-domain III) measured by this assay. Data from neutralization assays, the gold standard, is unfortunately scarce (n=45), hard to interpret and not very conclusive because of the timing of sample collection (some waning is expected) and because the evidence of seroreversion is quite weak (only observed in four individuals with weak seroconversions).

Thank you for the positive view of the overall value of the results. Having read the reviewer comments, we appreciate that the manuscript would benefit from some additional clarifications and analyses, and these are described in our responses below.

Essential revisions:

1) The authors speculate that the reductions in ZIKV seroprevalence measured by ELISA/MIA could indicate waning population immunity to ZIKV. This would have profound implications for future ZIKV dynamics in these and other populations, as the authors discuss. However, we think that another possibility, that is more likely, is that the observed decreases might just reflect the kinetics of antibody responses measured by this specific assay (antibodies against domain III). As far as we understand, it hasn't yet been established whether this antibody response is a good marker of historical exposure, or just a marker of recent exposure. DENV seroprevalence also seems to wane (Supplementary Figure 1). In the absence of additional data describing the performance/kinetics of this assay, this potential explanation needs to be discussed.

There is evidence that MIA can provide information about historical exposure, because high seroprevalence was observed in school children in French Polynesia in 2018, four years after the Zika epidemic (Figure 1). We also observed a higher level of seroprevalence against DENV serotypes in older groups in Fiji in 2013, consistent with a long-lasting response. However, we agree it is worth clarifying the antibody responses against domain III are not necessarily equivalent to protective immunity, and now include this in the Discussion:

“Although we found evidence of a decline in seroprevalence for antibodies against domain III of the envelope glycoprotein, as well as waning neutralizing antibody responses following two ZIKV outbreaks, the implications for susceptibility to future ZIKV infection remain unclear. […] A key aim for future work will be establish how waning antibody levels as measured by MIA and neutralization assays correspond to protective immunity, and hence susceptibility to reinfection in populations that have already experienced transmission of ZIKV.”

2) More details should be provided about the serological assays used. Two references are provided about the ELISA/MIA assays but they also provide very little information. At the very least, the paper should be explicit that the assay uses recombinant antigens comprising domain III of the envelope glycoprotein. Ideally, data on performance characteristics of the assay should also be reported as this information is crucial for the interpretation of results.

We have added these details to the Materials and methods:

“Serum samples collected from blood donors between July 2011 and October 2013 and samples collected from the general population and schoolchildren in 2014 were all tested for presence of IgG antibodies against ZIKV and each of the four DENV serotypes using a recombinant antigen-based indirect ELISA as reported previously

[Aubry et al., 2017; Aubry et al., 2015]. […] Recombinant antigens used in both assays comprised domain III of the envelope glycoprotein of ZIKV, DENV-1, DENV-2, DENV-3, or DENV-4 strains (respective GenBank accession no. KJ776791,

AF226686.1, FM986654, FJ44740.1, FM986672.1) and were produced using the

Drosophila S2 expression system (Life Technologies, USA) as previously detailed [Aubry et al., 2015].”

3) Related to the point above, please include more detail on the agreement between the MIA and ELISA tests. As far as I can see reference 11 does not contain enough information to assess this. It states: "80% were positive by both tests". Were all the rest of the samples negative by both tests?

We have added additional details to the Materials and methods:

"The sensitivity and specificity of the MIA assay were respectively 100% and 100% for DENV-1, 89.47% and 97.1% for DENV-2, 100% and 100% for DENV-3, 96.88% and 100% for DENV-4, and 79.59% and 94.87% for ZIKV. In the serosurvey conducted among the general population of the 5 archipelagos in French Polynesia in 2014 [Aubry et al., 2017], 196 samples were tested using both ELISA and MIA: among the 97 serum samples that tested positive for anti-ZIKV IgG by ELISA, 78 (80%) were also found positive by MIA; and among the 99 serum samples that tested negative for anti-ZIKV IgG by ELISA, 70 (71%) were also found negative by MIA. This produced a value of Cohen’s κ = 0.51 [Aubry et al., 2017]."

4) While ZIKV seroprevalence is observed to wane in both populations (Figure 1), DENV seroprevalence dynamics differ between the populations. It is not clear why, and how this relates to the observed decay for ZIKV seroprevalence. The results seem to be affected by the historical intensity of DENV transmission in French Polynesia and Fiji and may have been affected by DENV or ZIKV transmission that occurred after the main ZIKV epidemics. We recommend that Supplementary Figure 1 accompany Figure 1 in the main text. In Fiji, following the ZIKV epidemic, titers to DENV appeared to increase (except to DENV3). In French Polynesia, there was no DENV immunity to the four DENV before the arrival of ZIKV, and DENV titers did decline in adults after the ZIKV epidemic. Basically, it would be helpful to the reader if the authors include a figure on the epidemics of DENV and ZIKV in these populations before and after the ZIKV epidemic, ideally with serotype-specific epidemic curves if such data are available, to be able to interpret these results.

We agree it would be useful to include epidemiological context and have updated Figure 1 to include data on which serotypes were confirmed by PCR during this period, as well as temporal trends in symptom reporting. Note that in Supplementary Figure 1 (now Figure 1—figure supplements 1–4), the plots were only intended to show DENV data for 2014 and 2015 in French Polynesia; the apparent zeros in 2013 were the result of redundant plotting code. However, we realise that it would be helpful to show the 2013 data as well, as have updated the figures accordingly.

We also cover the advantages and challenges of the epidemiology in these locations in the updated Discussion:

“These data provided strong evidence that ZIKV seroprevalence declined over the two-year period following first reports of circulation, but our sample size was insufficient to fully explore the potential effect of anti-DENV pre-existing immunity on anti-ZIKV antibody waning once we stratified individuals by previous DENV exposure. Although the outbreaks of DENV-3 in Fiji and ZIKV in French Polynesia were well-documented and occurred over a relatively brief period of time (Figure 1), it was harder to identify the likely time of infection for other viruses – such as ZIKV in Fiji or DENV in French Polynesia – in our study populations.”

5) The age-standardised seroprevalence is the important metric here. As the authors note there is less decline in this metric. This should be the main result, as it seems that the other comparisons are not informative at all-as far as I can determine the differences are simply driven by the age distributions of the population sampled. Please explain if this is not the case.

We think it is important to show the results of age-standardisation, and now do so in Table 2 in the main text. However, we disagree that differences observed are simply the results of the age distributions sampled.

There are three main reasons for this:

First, the age distributions of the 2014 and 2015 study populations were not particularly different (Table 7); both had a similar median and IQR.

Second, age-adjusted DENV seroprevalence estimates for over 16s in French Polynesia were similar to the raw seroprevalence values (now shown in Table 2). Standardising for age increased uncertainty, but did not consistently shift estimates in either direction for these other viruses, suggesting no evidence of a systematic bias as a result of sampling.

Third, ZIKV risk seems to be higher in younger age groups; in French Polynesia, seroprevalence was higher in school children. Hence the 2015 study, which had a slightly younger median age (43 vs. 47), would if anything be expected to have a raw seroprevalence estimate that was biased upwards rather than downwards.

It is therefore likely that any changes in ZIKV seroprevalence estimates after age standardisation are the result of increased uncertainty owing to relatively low levels of seropositivity in adults, particularly in 2015, and small sample size in 2014, rather than a genuine age effect.

It is worth noting that we have identified some differences between the 2014 and 2015 study populations, particularly in terms of exposure risk. We examined the potential impact of these differences by bootstrap resampling the data according to prior exposure history (see response to point 15 below). We still observe a decline in ZIKV seroprevalence, but no decline for recently circulating DENV serotypes.

Longitudinal data from Fiji tested using neutralization assays could certainly have added strength to the hypothesis of waning population immunity, but unfortunately the data is scarce (n=45) and hard to interpret for multiple reasons:

6) In order to be able to compare the DENV and ZIKV results (Figure 2), it would be crucial to understand when sample collection occurred with respect to ZIKV and DENV outbreaks. At the end of the Results, from what we understand: the DENV outbreak was before the ZIKV outbreak, so the observed differences in decline rates between DENV and ZIKV could be explained by much of the DENV decline already happened before the measurement, whereas the ZIKV transmission was much closer to the time of the first measurement. Differences of just a few months could make a large difference here. This would be consistent with the lower increase observed from 2013-2015 for DENV compared to ZIKV. Similarly for the differences in seroreversions, there may have been individuals that were DENV positive closer to the transmission, but were not by 2015, and therefore would not have been captured by the measurements here. In addition, it would be useful to know if there was any DENV transmission in the 2015-2017 period?

We agree it is worth including more details about the potential relationship between epidemiological dynamics and sample collection timing. We now include information on flavivirus circulation in the new Figure 1 (see also response to point 4 above). Although the DENV-3 epidemic Fiji occurred around 18 months for the 2015 sample collection, there was also low-level circulation of DENV-1 and DENV-2 during this period. We therefore analysed log neutralisation titres for participants who were seronegative to these respective viruses in 2013 and seroconverted in 2015. Unlike ZIKV, we found no evidence of a subsequent overall decline during 2015–17 (Figure 2—figure supplement 2). We describe this analysis in the updated Results section:

“In Fiji, there was a delay of around 18 months between the end of the 2013/14 DENV-3 epidemic and collection of samples in 2015. […] Unlike ZIKV, we found no evidence of a subsequent overall decline during 2015–17 for those participants who seroconverted during 2013–15 (Figure 2—figure supplement 2).”

7) Also, looking at the neutralization data in GitHub, it seems that 9/45 (20%) of individuals had positive (but low) titers to ZIKV in 2013 (titer greater of equal to 2). 7/9 of these positive values are in individuals who also have positive titers to 3 or more dengue serotypes. We find this concerning as this is supposed to be the pre-Zika time-point and this might suggest some cross-reactivity in this ZIKV PRNT assay. Has the sensitivity and specificity of this neutralization assay (and this cut-point) been calculated? More details need to be provided about this assay and its performance as well.

The sensitivity and specificity of this assay using this cut-off has not been calculated. However, log-titres to ZIKV follow a bimodal distribution (Figure 3 in new manuscript), which supports a cut-off of 2 as an appropriate definition of seropositivity. We also examined the correlation between MIA and neutralisation assay results, finding 68–75% agreement between the two (see response to point 8 below).

As log titres may have been influenced by cross-reactivity, we also compared correlations in responses against different pairs of viruses. For participants who were seronegative to all five viruses in 2013, we found no evidence of cross-reactive responses against ZIKV and the four DENV serotypes (Figure 3—figure supplement 2 in new manuscript), although there was evidence of potential cross-reaction for participants who were already seropositive to DENV (Figure 3—figure supplement 3 in new manuscript). In our main analysis using neutralisation titres (Figure 2), we therefore conditioned on participants being seronegative to ZIKV in 2013.

These additional analyses of neutralisation data are now described in the ‘Serological analysis: Fiji’ section of the Materials and methods.

8) Related to the above, it would be good to show the correlation between neutralization and MIA results. I tried to do this myself (using the provided ids) but was only able to match the data of 16/45 individuals (the neut dataset includes 344 individuals, most with incomplete data). Unless there's a problem with the data (or provided IDs), the agreement seems poor. For example, these 16 individuals have positive neutralization titers in 2015 (>4 in the majority) and yet, only 5/16 are classified as positive by the MIA assay.

The IDs in the two datasets provided in the original submission were independent, as the datasets were not analysed together. In the new submission, we include a single merged dataset (“dset5-fiji-mergedassaydata.csv”), and use this to assess the correlation between neutralization and MIA results. This is described in the updated Materials and methods:

“A previous study, which tested serological samples from Fiji across three divisions [Kama et al., 2019], found that of the samples reactive by MIA, 66/83 (79.5%) exhibited neutralizing activity for ZIKV (κ = 0.71) and 109/112 (97.3%) for DENV (κ = 0.80). […] We found that 54/79 (68.4%) samples that were positive to ZIKV in the neutralization assay were also positive by MIA, and 42/56 (75%) who were seronegative were also negative by MIA.”

While compiling the linked dataset, we also realised that two participants had multiple aliquots tested for MIA. We have therefore omitted these repeat MIA measurements from the dataset, leaving 189 unique samples. We have updated all the values in the text accordingly, and the overall conclusions remain the same.

9) The four observed ZIKV seroreversions described in the last paragraph of the Results seem to be in individuals with relatively low titers at seroconversion (4). There are four additional "seroreversions" among individuals who were already positive in the 2013 sample. Are these seroreversions mainly capturing cross-reactive responses? It would be useful to show a figure (maybe supplementary) with the longitudinal neutralization titer data.

Given the limited neutralisation results we had available, we aimed to focus on the overall titre dynamics, rather than focusing too heavily on results from simple cut-off metrics, which as suggested, may not fully reflect the range of the underlying responses. The main reason we conditioned on seronegativity in Figure 2 was to remove potential effects of cross-reactivity responses in 2013 (see response to point 7 above). We realise, however, that it would also be helpful to show the same analysis including participants who were already seropositive by neutralisation assay to ZIKV or DENV in 2013. We therefore have added a new figure (Figure 2—figure supplement 1) showing results for participants who had a rise in log titre of at least 2 between 2013–15, regardless of titre in 2013. We obtain the same conclusion about waning, actually with a larger average decline in log titre for ZIKV (panel A). This is further evidence that individuals can exhibit a notable decline in ZIKV response, even if they do not fall below the titre = 2 threshold.

Thank you also for the suggestion to show the raw titres. We agree this would be useful and have added this to the new manuscript (Figure 3—figure supplement 1).

10) In light of these limitations/uncertainties, we think the manuscript needs to be reworked to emphasize the uncertainty regarding the meaning of these findings. We would emphasize the need for more studies using longitudinal data (rather than cross-sectional) data and a broader set of assays, including PRNTs. We would de-emphasize the discussion/conclusions around the implications of waning protection (last paragraph of the Discussion). While provocative, we are not convinced it's supported by the presented data.

We have de-emphasised the potential implications for protective immunity in the Abstract and Discussion, and focus instead on the need for additional studies to address the questions raised and remaining. The final paragraph of the Discussion is now as follows:

“In the short-term, our findings have implications for the design of follow up studies of ZIKV. […] Such studies will be essential to understand different aspects of the short and long-term immune antibody response against ZIKV, and how prior exposures to DENV may influence these responses.”

11) It would be good to clarify which subset of samples were tested and the sample sizes that went into each analysis.

We have added the sample sizes at relevant points in the text, as well as to figure and table legends.

[Editors' note: further revisions were requested prior to acceptance, as described below.]

The authors have satisfactorily addressed most of our concerns. Thanks for providing the merged data! We have some additional comments:

1) The low agreement reported for the ZIKV MIA and PRNTs assays is a bit concerning, and not consistent with the sensitivity and specificity reported. There seem to be 65 samples positive according to the PRNT in 2017, and of these only 17 (26%) are positive according to the MIA. Assuming that the PRNT is the "gold standard", this would suggest a much lower sensitivity of MIA than was reported for old infections, suggesting time-dependent sensitivity (26% vs. 80%). If instead we question the performance of the PRNT (cross-reactivity?) then the waning results are also questionable. I think these discrepancies (and the difficulty interpreting results from these novel assays) should be explicitly discussed.

As MIA and NT measure different aspects of the immune response, we decided to analyse the two results in more detail. First, we compared ZIKV seroprevalence estimates according to pre-existing DENV titre in 2013 (new Figure 3). As already noted in the previous iteration of the manuscript, we found some evidence of cross-reactivity by NT for participants with higher initial DENV titres, which is why we conditioned on participants being seronegative by NT in 2013 in our main analysis of the titre data (Figure 2). We found little difference between assays in 2015. The difference in 2017 was greatest for participants with some pre-existing DENV titre, consistent with the B cell competition hypothesis presented in previous version of the Discussion.

These findings are described in the updated Results:

“Of the 45 participants tested by neutralization assay, 9 were initially seropositive to ZIKV by NT in 2013. […] This difference was associated with participants’ 2013 DENV titres: those with intermediate DENV titres in 2013 had a significantly lower probability of being seropositive in the MIA in 2017 compared to NT (Figure 3C).”

We have also expanded the hypothesis about immunological mechanism for these findings in the Discussion, proposing an explanation for why MIA responses did not persist:

“If an individual has experienced prior DENV infections, high numbers of weakly neutralizing cross-reactive B cells may outcompete naïve B cells for ZIKV antigen [Midgley et al., 2011], leading to a short-term boost in antibody response against ZIKV following ZIKV infection [Robbiani et al., 2017] but not a persistent specific response; a similar phenomenon has been observed for other antigenically variable viruses like influenza [Kucharski et al., 20181]. […] This difference was greatest for participants who had intermediate baseline titres to DENV in 2013 (Figure 3C), which would support the hypothesis that prior DENV exposure may result in a detectable short-term specific response against ZIKV following ZIKV infection (as measured by MIA), but not a persistent specific response.”

2) Related to the above, authors should explicitly state in the text that, while neutralization titers decayed over the observation period, there's little (or no) evidence of sero-reversion (decreasing seroprevalence) according to the neutralization data.

There is some limited evidence of sero-reversion according to neutralization titres, and we now include the numbers in the Results:

“In total, four participants seroreverted between 2015 and 2017; all had a log titre of 4 against ZIKV in 2015.”

3) Please provide a reference for the sensitivities and specificities reported for the MIA assay. Also, please state whether these performances were evaluated using acute/early convalescent samples vs. samples from historical infections (infections occurring one or more years before).

We now clarify in the Materials and methods that serostatus was defined by a cut-off determined using positive and negative control sera analyzed by ROC curve, with the quoted sensitivities and specificities for MIA calculated from these results. The positive control serum for ZIKV was collected 13 months after RT-PCR confirmed infection.

In the revised Discussion, we also note that the post-infection immune response in RT-PCR-confirmed cases may not be the same as in asymptomatic or unreported cases:

“antibody responses in RT-PCR-confirmed cases may not necessarily be representative of immune responses against ZIKV in the wider population, particularly following asymptomatic infection. Although MIA seropositivity in our study was defined using control sera collected over a year after RT-PCR-confirmed infection, our results suggest that this threshold may not detect long-term waning responses in individuals who had unreported, and likely less severe, infections.”

4) I also find it concerning that 9/46 samples (~20%) were positive against ZIKV (low titers) according to the PRNT assay in 2013. The authors perform some analyses around cross-reactivity, but these are only described in the Materials and methods. I think cross-reactivity (and how it could impact the results) should be addressed in the Discussion.

As noted in the response to point 1 above, participants who were seropositive by NT in 2013 typically had higher initial DENV titres, suggesting potential cross-reactivity. However, our main analysis focused on participants who were seronegative to ZIKV in 2013, so this should not affect our conclusions.

We include the additional analysis as Figure 3 in the new Results and address the issue of cross-reactivity in the updated Discussion:

“Several participants in Fiji were seropositive to ZIKV by neutralization assay (NT) in 2013, but this result may be influenced by cross-reaction; participants who had high pre-existing titres to DENV in 2013 were more likely to be seropositive by NT (Figure 3A). […] However, we obtained the same conclusion when participants who were initially seropositive were also considered (Figure 2—figure supplement 1).”

[Editors' note: further revisions were requested prior to acceptance, as described below.]

Most of our concerns have been satisfactorily addressed, but I have one remaining question.

- You report that the proportion seropositive (by MIA) was ~24% in Fiji in 2015, yet Figure 3B shows a seroprevalence of >90% among the 45 individuals that were also tested using neutralization assays. It could almost be seen as if individuals were selected for neutralization assays based on the MIA status (42/45 samples positive by MIA were tested) but the text reads that they were selected based on sample availability only. Please clarify how the samples were selected.

Thank you for raising this point, and we realise we should have been clearer here. The NT testing was indeed informed by the 2013 and 2015 MIA status, as well as sample availability (i.e. from loss to follow up in the longitudinal serosurvey and stored serum availability). However, this selection choice did not affect our conclusions about waning of ZIKV titres, because our analysis focused on individual-level changes in NT titre conditioned on serostatus, rather than absolute seroprevalence by NT.

We have updated the Materials and methods accordingly:

“To follow the evolution of antibody titres at the individual level, a subset of samples collected from the same individuals in 2013, 2015 and 2017 were tested for the presence of neutralizing antibodies against ZIKV and each of the four DENV serotypes using a neutralization assay as previously described [Cao-Lormeau et al., 2016]. […] We also tested samples from an additional 24 participants from the same cohort who were seropositive to ZIKV by MIA in 2013 or 2015 and for whom we had sufficient serum from 2013 and 2015 to test by neutralization assay, but no matched sample from the 2017 follow up survey (i.e. 69 paired samples in total).”

As well as the Results:

“To explore dynamics of antibody waning at the individual level, we performed neutralization assays (NT) on a subset of 45 participants from Fiji for whom sufficient sera were available to test against ZIKV from all three collection periods, focusing on those who were seropositive to ZIKV by MIA in 2013 or 2015.”

https://doi.org/10.7554/eLife.48460.sa2

Article and author information

Author details

  1. Alasdair D Henderson

    Centre for the Mathematical Modelling of Infectious Diseases, London School of Hygiene and Tropical Medicine, London, United Kingdom
    Contribution
    Data curation, Formal analysis, Writing—original draft
    Contributed equally with
    Maite Aubry
    Competing interests
    No competing interests declared
  2. Maite Aubry

    Institut Louis Malardé, Papeete, French Polynesia
    Contribution
    Data curation, Formal analysis, Writing—original draft
    Contributed equally with
    Alasdair D Henderson
    Competing interests
    No competing interests declared
  3. Mike Kama

    1. Fiji Centre for Communicable Disease Control, Suva, Fiji
    2. The University of the South Pacific, Suva, Fiji
    Contribution
    Data curation, Formal analysis, Writing—review and editing
    Competing interests
    No competing interests declared
  4. Jessica Vanhomwegen

    Institut Pasteur, Paris, France
    Contribution
    Methodology, Writing—review and editing
    Competing interests
    No competing interests declared
  5. Anita Teissier

    Institut Louis Malardé, Papeete, French Polynesia
    Contribution
    Data curation, Formal analysis, Writing—review and editing
    Competing interests
    No competing interests declared
  6. Teheipuaura Mariteragi-Helle

    Institut Louis Malardé, Papeete, French Polynesia
    Contribution
    Formal analysis, Writing—review and editing
    Competing interests
    No competing interests declared
  7. Tuterarii Paoaafaite

    Institut Louis Malardé, Papeete, French Polynesia
    Contribution
    Formal analysis, Writing—review and editing
    Competing interests
    No competing interests declared
  8. Yoann Teissier

    Direction de la Santé de la Polynésie française, Papeete, French Polynesia
    Contribution
    Data curation, Writing—review and editing
    Competing interests
    No competing interests declared
  9. Jean-Claude Manuguerra

    Institut Pasteur, Paris, France
    Contribution
    Methodology, Writing—review and editing
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-5202-6531
  10. John Edmunds

    Centre for the Mathematical Modelling of Infectious Diseases, London School of Hygiene and Tropical Medicine, London, United Kingdom
    Contribution
    Formal analysis, Writing—review and editing
    Competing interests
    No competing interests declared
  11. Jimmy Whitworth

    Centre for the Mathematical Modelling of Infectious Diseases, London School of Hygiene and Tropical Medicine, London, United Kingdom
    Contribution
    Formal analysis, Writing—review and editing
    Competing interests
    No competing interests declared
  12. Conall H Watson

    Centre for the Mathematical Modelling of Infectious Diseases, London School of Hygiene and Tropical Medicine, London, United Kingdom
    Contribution
    Data curation, Formal analysis, Writing—review and editing
    Competing interests
    No competing interests declared
  13. Colleen L Lau

    Australian National University, Canberra, Australia
    Contribution
    Data curation, Formal analysis, Writing—review and editing
    Competing interests
    No competing interests declared
  14. Van-Mai Cao-Lormeau

    Institut Louis Malardé, Papeete, French Polynesia
    Contribution
    Conceptualization, Data curation, Formal analysis, Writing—original draft
    Contributed equally with
    Adam J Kucharski
    For correspondence
    mlormeau@ilm.pf
    Competing interests
    No competing interests declared
  15. Adam J Kucharski

    Centre for the Mathematical Modelling of Infectious Diseases, London School of Hygiene and Tropical Medicine, London, United Kingdom
    Contribution
    Conceptualization, Data curation, Formal analysis, Writing—original draft
    Contributed equally with
    Van-Mai Cao-Lormeau
    For correspondence
    adam.kucharski@lshtm.ac.uk
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0001-8814-9421

Funding

Ministry for Europe and Foreign Affairs (Pacific funds 04917-19/07/17)

  • Van-Mai Cao-Lormeau

Horizon 2020 - Research and Innovation Framework Programme (ZIKAlliance grant 734548)

  • Van-Mai Cao-Lormeau

Investissement d'Avenir Program (Labex IBEID grant ANR-10-LABX-62- IBEID)

  • Jessica Vanhomwegen
  • Jean-Claude Manuguerra

Wellcome (107778/Z/15/Z)

  • Jimmy Whitworth

Wellcome (206250/Z/17/Z)

  • Adam J Kucharski

Medical Research Council (MR/J003999/1)

  • Conall H Watson

National Health and Medical Research Council (1109035)

  • Colleen L Lau

Medical Research Council (MR/N013638/1)

  • Alasdair D Henderson

Ministry of Europe and Foreign Affairs (Pacific funds 02918-18/06/18)

  • Van-Mai Cao-Lormeau

French and French Polynesia government MA’I’ORE program (03298/MTF/REC-17/05/18)

  • Van-Mai Cao-Lormeau

French and French Polynesia government MA’I’ORE program (HC/372/DIE/BPT- 18/05/18)

  • Van-Mai Cao-Lormeau

French and French Polynesia government MA’I’ORE program (HC/245/DIE/BPT-16/05/19)

  • Van-Mai Cao-Lormeau

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Acknowledgements

We are grateful to the Minister for Education of French Polynesia and to the directors, teachers, nurses and schoolchildren from the elementary and junior high schools selected for the serosurvey in 2018. We greatly thank all the participants and community leaders in Fiji who generously contributed to the study over the three visits. We would like to acknowledge the work of the field teams: Jessica Paka, Amele Ratevono, Warren Fong, Manisha Prakash, Jonetani Bola, Mosese Ligani, and Taina Naivalu (2017); Meredani Taufa, Adi Kuini Kadi, Jokaveti Vubaya, Colin Michel, Mereani Koroi, Atu Vesikula, and Josateki Raibevu (2015); Dr. Kitione Rawalai, Jeremaia Coriakula, Ilai Koro, Sala Ratulevu, Ala Salesi, Meredani Taufa, and Leone Vunileba (2013). We thank Rina Kumar, Sokoveti Covea, Taina Naivalu and Vinaisi Duituturaga for their assistance with sample preparation. We would also like to thank Eric J Nilles of the World Health Organization Western Pacific Region.

Ethics

Human subjects: The French Polynesia serosurveys were approved by the Ethics Committee of French Polynesia (ref 61/CEPF 08/29/2013, 60/CEPF 06/27/2013, 74/CEPF 05/04/2018, and 75/CEPF 05/04/2018). The 2013 Fiji study, and the 2015 and 2017 follow up studies were approved by the Fiji National Research Ethics Review Committee (ref 2013-03, 2015.111.C.D, 2017.20.MC) and the London School of Hygiene and Tropical Medicine Observational Research Ethics Committee (ref 6344, 10207, 12037).

Senior Editor

  1. Neil M Ferguson, Imperial College London, United Kingdom

Reviewing Editor

  1. Isabel Rodriguez-Barraquer, University of California, San Francisco, United States

Reviewers

  1. Isabel Rodriguez-Barraquer, University of California, San Francisco, United States
  2. Leah Katzelnick, University of California, Berkeley, United States

Publication history

  1. Received: May 14, 2019
  2. Accepted: November 20, 2019
  3. Version of Record published: January 28, 2020 (version 1)

Copyright

© 2020, Henderson et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.

Metrics

  • 854
    Page views
  • 111
    Downloads
  • 3
    Citations

Article citation count generated by polling the highest count across the following sources: Crossref, PubMed Central, Scopus.

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Download citations (links to download the citations from this article in formats compatible with various reference manager tools)

Open citations (links to open the citations from this article in various online reference manager services)

Further reading

    1. Epidemiology and Global Health
    2. Microbiology and Infectious Disease
    Fredrick M Mobegi et al.
    Research Article

    Poor diet and lifestyle exposures are implicated in substantial global increases in non-communicable disease burden in low-income, remote, and Indigenous communities. This observational study investigated the contribution of the fecal microbiome to influence host physiology in two Indigenous communities in the Torres Strait Islands: Mer, a remote island where a traditional diet predominates, and Waiben a more accessible island with greater access to takeaway food and alcohol. Counterintuitively, disease markers were more pronounced in Mer residents. However, island-specific differences in disease risk were explained, in part, by microbiome traits. The absence of Alistipes onderdonkii, for example, significantly (p=0.014) moderated island-specific patterns of systolic blood pressure in multivariate-adjusted models. We also report mediatory relationships between traits of the fecal metagenome, disease markers, and risk exposures. Understanding how intestinal microbiome traits influence response to disease risk exposures is critical for the development of strategies that mitigate the growing burden of cardiometabolic disease in these communities.

    1. Epidemiology and Global Health
    2. Medicine
    Janina Isabel Steinert et al.
    Research Article Updated

    Background:

    Since 2015, the World Health Organisation (WHO) recommends immediate initiation of antiretroviral therapy (ART) for all HIV-positive patients. Epidemiological evidence points to important health benefits of immediate ART initiation; however, the policy’s impact on the economic aspects of patients' lives remains unknown.

    Methods:

    We conducted a stepped-wedge cluster-randomised controlled trial in Eswatini to determine the causal impact of immediate ART initiation on patients’ individual- and household-level economic outcomes. Fourteen healthcare facilities were non-randomly matched into pairs and then randomly allocated to transition from the standard of care (ART eligibility at CD4 counts of <350 cells/mm3 until September 2016 and <500 cells/mm3 thereafter) to the ‘Early Initiation of ART for All’ (EAAA) intervention at one of seven timepoints. Patients, healthcare personnel, and outcome assessors remained unblinded. Data were collected via standardised paper-based surveys with HIV-positive adults who were neither pregnant nor breastfeeding. Outcomes were patients’ time use, employment status, household expenditures, and household living standards.

    Results:

    A total sample of 3019 participants were interviewed over the duration of the study. The mean number of participants approached at each facility per time step varied from 4 to 112 participants. Using mixed-effects negative binomial regressions accounting for time trends and clustering at the level of the healthcare facility, we found no significant difference between study arms for any economic outcome. Specifically, the EAAA intervention had no significant effect on non-resting time use (RR = 1.00 [CI: 0.96, 1.05, p=0.93]) or income-generating time use (RR = 0.94, [CI: 0.73,1.20, p=0.61]). Employment and household expenditures decreased slightly but not significantly in the EAAA group, with risk ratios of 0.93 [CI: 0.82, 1.04, p=0.21] and 0.92 [CI: 0.79, 1.06, p=0.26], respectively. We also found no significant treatment effect on households’ asset ownership and living standards (RR = 0.96, [CI 0.92, 1.00, p=0.253]). Lastly, there was no evidence of heterogeneity in effect estimates by patients’ sex, age, education, timing of HIV diagnosis and ART initiation.

    Conclusions:

    Our findings do not provide evidence that should discourage further investments into scaling up immediate ART for all HIV patients.

    Funding:

    Funded by the Dutch Postcode Lottery in the Netherlands, Alexander von Humboldt-Stiftung (Humboldt-Stiftung), the Embassy of the Kingdom of the Netherlands in South Africa/Mozambique, British Columbia Centre of Excellence in Canada, Doctors Without Borders (MSF USA), National Center for Advancing Translational Sciences of the National Institutes of Health and Joachim Herz Foundation.

    Clinical trial number:

    NCT02909218 and NCT03789448.