SARS-CoV-2 seroprevalence and associated factors of infection before and after the Delta wave in French Polynesia: a cross-sectional study

MATAEA is a cross-sectional survey conducted on a random sample of the general population aged 18–69 years and representative of the five archipelagos of FP, as previously published [13]. Briefly, the territory of FP was divided into three geographic strata: the WIs, the LIs, and the other archipelagos (i.e. Tuamotu, Austral, Gambier, and Marquesas), which respectively include 75.1%, 12.8%, and 12.0% of the population according to the 2017 census [3]. In each stratum, the most populated islands with a healthcare center and an airport with regular flight rotations were selected. Finally, in each island, households were randomly selected and one resident per household was randomly recruited. The number of participants to enroll was the same in the three strata (N = 700), proportional to the population size of the islands selected for the study, and equally divided into three age groups (18–29, 30–44, and 45–69), each including 50% males and 50% females. Participation in the study involved agreeing to answer a questionnaire (including questions on demographic characteristics, lifestyle habits, and medical history), undergo physical measurements (including height, weight, waist circumference, and arterial pressure), and provide biological samples (including blood) for laboratory analyses.

Inclusion visits for the MATAEA study took place from November 2019 to December 2021. Because of the COVID-19 epidemic waves and restriction measures implemented to prevent the spread of SARS-CoV-2, inclusions were interrupted between March and May 2020, then between July 2020 and April 2021, and finally between August and September 2021. Upon resumption in April 2021, COVID-19 was added to the scope of the study and the questionnaire was expanded with questions specific to this disease [13]. Consequently, only participants recruited from April 2021 were considered in the subsequent analyses.

Blood samples from participants collected in 5 mL serum separation tubes (Becton Dickinson, USA) were centrifuged for 15 min at 3,500 rpm to retrieve serum. Immunoglobulin class G (IgG) antibodies targeting the SARS-CoV-2 nucleocapsid (N) protein were detected in the serum with the MAGPIX NxTAG® system (Luminex, USA) using a recombinant antigen-based microsphere immunoassay (MIA), as previously described [14]. Recombinant antigen-coated beads used for the MIA were provided by the Cellule d’Intervention Biologique d’Urgence (CIBU, Institut Pasteur, Paris). For each serum tested, the median fluorescence intensity value obtained was compared to a cut-off value to determine the participant’s immunological status (i.e. positive or negative for the detection of anti-SARS-CoV-2-N IgG antibodies). The cut-off value was calculated using receiver operating characteristic (ROC) analysis performed from a panel of positive and negative control sera.

Past SARS-CoV-2 infection was defined as a positive anti-N serology, given that the two COVID-19 vaccines used in FP between January and December 2021 do not elicit any antibody response to the SARS-CoV-2-N protein. In a sensitivity analysis, participants with a negative anti-N serology, but who reported having been tested positive for COVID-19 by a health professional before inclusion in the MATAEA study (i.e. who answered yes to the following question: have you had a nasopharyngeal test, performed by a nurse, that was positive for Covid-19?), were also considered as previously infected.

Information on COVID-19 vaccination status, as well as the type of vaccine used and the date of each injection if applicable, were collected from all participants in the questionnaire. Participants who self-reported having received at least one injection were considered as vaccinated. However, those who self-reported having been tested positive for COVID-19 by a health professional, prior to or within the same month as vaccination, were considered as unvaccinated for the analysis to reflect the time required for the vaccine to induce immune protection [15].

In another sensitivity analysis, participants were categorized as having been vaccinated for more or less than three months before inclusion to assess the impact of misclassifying the timing of infection compared to vaccination.

In each archipelago, the prevalence of SARS-CoV-2 infections and associated 95% CI were estimated within the sample. Weighted prevalence was inferred from the prevalence in each stratum of the sample by weighting. Weights, defined by archipelago (except the Society archipelago, for which the WIs and LIs were treated separately, and the Tuamotu and Gambier archipelagos which were combined), were each participant’s inverse probability of selection, where the probability is the number of participants in each of the six age/sex strata divided by its total population.

Factors associated with anti-SARS-CoV-2 seroprevalence were identified using logistic regression models. Factors investigated were sociodemographic characteristics, medical history, behavior characteristics (physical activity and smoking habit), barrier gestures implemented, and vaccination status. First, a univariate analysis was performed. For the multivariate analysis, a stepwise forward procedure was applied. As the study was conducted sequentially on the different archipelagos between April and December 2021 [13] (Austral and Tuamotu: April–May, Gambier: May, Marquesas and LIs: May–June, WIs: July then October to December 21), while the second wave of COVID-19 started in mid-July and peaked in mid-August 2021, the month of inclusion showed a strong confusion effect with several factors. This motivated the choice of a stepwise forward procedure, starting from a model that included only the month of inclusion. At each step of the forward procedure, all factors were investigated without restriction, and the most significant one was added to the model. The procedure stopped when all additional factors investigated presented a p-value greater than 0.05.

Between April and December 2021, a total of 1,148 participants were enrolled in the MATAEA survey, including 1,120 (97.6%) for whom serological analysis results for the detection of anti-SARS-CoV-2-N antibodies were available. The sociodemographic characteristics of the 1,120 participants considered in the subsequent analyses are detailed in Supplementary Table 1. Most of the participants were recruited in the WIs (57.2%), followed by the Marquesas (19.4%), the LIs (9.2%), the Tuamotu and the Austral (5.3%), and the Gambier (3.7%). The proportion of men and mean age of the participants did not differ between archipelagos (p = 0.28 and p = 0.32, respectively), while a significant difference was found for the level of education (p < 0.001).

Anti-SARS-CoV-2-N IgG antibodies, indicating past natural infection, were detected in 503/1,120 participants (44.9% [95% CI: 42.0–47.9]), corresponding to a weighted seroprevalence of 56.8% [95% CI: 53.4–60.2] for the FP population aged 18–69 years (Table 1). Before the Delta wave, from April to June 2021, anti-SARS-CoV-2-N antibodies were detected in 105/479 participants (21.9% [95% CI: 18.3–25.9]), versus 398/641 participants (62.1% [95% CI: 58.2–65.9]) during and after the Delta wave from July to December 2021. The proportions of participants with evidence of past infection significantly differed between age groups (p = 0.03), archipelagos (p < 0.001), and months of inclusion (p < 0.001), with the highest values found in individuals aged 18–29 years (50.1% [95% CI: 44.7–55.6]), those living in the WIs (62.1% [95% CI: 58.2–65.9]), and those enrolled in November 2021 (76.8% [95% CI: 69.0–83.3]) (Fig. 2).

Moreover, the prevalence of anti-SARS-CoV-2-N antibodies was significantly higher in participants who self-reported at least one past SARS-CoV-2 infection than in those who did not (143/153 (93.5%) and 360/967 (37.2%), respectively; p < 0.001) (Table 1).

Participants were recruited in the five archipelagos at different periods: in the Tuamotu (Fakarava and Hao) and the Austral (Raivavae and Rimatara) during April–May 2021, the Gambier (Mangareva) in May, the Marquesas (Nuku Hiva, Hiva Oa, and Ua Pou) and the LIs (Huahine and Bora Bora) during May–June, and the WIs (Tahiti and Moorea) in July and then during October–December 2021 (Table 2 and Fig. 1). Due to collinearity between the month of inclusion and archipelago, related to occurrence of the Delta epidemic wave from July 2021, we only considered month of inclusion from these two variables for further analyses.

In the multivariate analysis, positivity for the detection of anti-SARS-CoV-2-N antibodies was significantly associated with the month of inclusion (p < 0.001). Indeed, participants recruited during the months of October and November 2021 were more likely to be positive than those recruited in July 2021, while participants recruited in April, May, and June were less likely to be positive (Table 3). Male participants were significantly less likely to be seropositive (p = 0.02). Regarding marital status, participants who had never been married (and were single at the time of inclusion) were significantly less likely to have been infected than those in a relationship (married or cohabiting) (p = 0.01). Finally, the prevalence of anti-N antibodies was significantly lower in participants with physician-diagnosed respiratory allergy (p = 0.03), those who smoked tobacco products (p = 0.04), and those who had been vaccinated (p < 0.001). In a sensitivity analysis, those vaccinated within the three months preceding inclusion, and those vaccinated more than three months before inclusion, were distinguished. Again, infection rate was significantly lower in the vaccinated, and the effect was stronger in those vaccinated more than three months before inclusion than in those vaccinated within three months of inclusion (p = 0.019). The effect of the other factors remained the same. In another sensitivity analysis where the ten participants with a negative anti-N serology, but who self-reported having been tested positive for COVID-19 by a health professional before inclusion, were also considered as previously infected; the same variables as above were found to be associated with prior infection, with similar odds ratios (data not shown).

The following variables were significantly associated with the prevalence of anti-N antibodies in the univariate analysis but were no longer associated after adjustment: INSEE classification of municipality (urban/rural, p < 0.001 in univariate), age group (p = 0.02), professional occupation (p < 0.001), number of people living in the household (p < 0.001), coughing into the elbow as a protective measure (p = 0.01), and social distancing as a protective measure (p < 0.001).