Temporal trends and epidemiological aspects of ciguatera in French Polynesia: a 10-year analysis

Authors


Corresponding Author Eric Dewailly, Public Health Research Unit, Laval University Hospital Research Centre (CHUQ), Édifice Delta 2, Suite 600, 2875 Boul. Laurier, Québec (QC) G1V2M2, Canada. Tel.: 418 656 4141 #46518; Fax: 418 654 2726; E-mail: eric.dewailly@inspq.qc.ca

Summary

Objectives  The purpose of this study was to report the temporal trends of the incidence of ciguatera poisoning from 1992 to 2001 in French Polynesia.

Methods  This retrospective study analysed 7842 cases of ciguatera disease recorded over a period of 10 years.

Results  The annual incidence varied from 26.3 to 41.9 per 10 000 person-years. An analysis of cases grouped by archipelago revealed differences in incidences (P < 0.0001) with the most remote archipelagos having the highest incidences. A detailed analysis on a sub-sample of recorded cases for which clinical information was available (n = 1824) confirmed the neurological and gastrointestinal nature of this seafood poisoning.

Conclusion  The incidence of ciguatera poisoning appeared relatively stable during the 10 years of the study period. However, the gradient of remoteness observed suggests an adaptation of management of ciguatera disease to each archipelago.

Abstract

Objectifs  Le but de cette étude était de rapporter les tendances temporelles de l'incidence de l'empoisonnement de ciguatera de 1992 à 2001 en Polynésie française.

Méthodes  Etude rétrospective de 7842 cas de ciguatera enregistrés sur une période de dix ans.

Résultats  L'incidence annuelle variait de 26,3 à 41,9 par 10000 personnes année. Une analyse des cas groupés par archipel a indiqué des différences dans les incidences (p < 0.0001); les archipels les plus éloignés ayant les incidences les plus élevées. Une analyse détaillée sur une partie de l’échantillon de cas enregistrés pour lesquels l'information clinique était disponible (n = 1824) a confirmé la nature neurologique et gastro-intestinale de cet empoisonnement par fruits de mer.

Conclusion  L'incidence de l'empoisonnement de ciguatera semble relativement stable au cours des dix années de la période d’étude. Cependant, le gradient observé de l’éloignement suggère une adaptation dans la prise en charge de la ciguatera dans chacun des archipels.

Abstract

Objetivos  El propósito de este estudio es reportar las tendencias temporales en la incidencia de envenenamiento por ciguatera entre 1992 y 2001 en la Polinesia Francesa.

Métodos  En este estudio retrospectivo se analizaron 7842 casos de ciguatera reportados durante un período de 10 años.

Resultados  La incidencia anual varió entre 26.3 y 41.9 por 10,000 personas-años. El análisis de los casos agrupados por archipiélago reveló diferencias entre las incidencias (p < 0.0001), con los archipiélagos más remotos presentando las mayores incidencias. Un análisis detallado de una sub-muestra de casos reportados para los cuales había información clínica disponible (n = 1824), confirmó la naturaleza neurológica y gastrointestinal de esta intoxicación por marisco.

Conclusión  La incidencia de ciguatera fue relativamente estable durante los diez años del período de estudio. Sin embargo, el gradiente de lejanía observado, sugiere una adaptación del manejo de la ciaguatera en cada archipiélago.

Introduction

Ciguatera is a seafood poisoning found throughout the tropical areas of the world and is considered to be a public health problem in many southern countries including French Polynesia (Lewis 1986,1992; Ting et al. 1998; Lehane 2000; Shuval 2000). In spite of an endemic status in tropical areas, ciguatera also affects northern people through spreading via tourism and the global increase of consumption of reef fish (Johnson & Jong 1983; Sanner et al. 1997). It is suspected that 50 000 to 500 000 people per year are potentially at risk of contracting ciguatera disease in tropical areas with an annual incidence fluctuation found somewhere between 0.78/10 000 person-years to 58.5/10 000 person-years according to location (Lewis 1986; Quod & Turquet 1996). Nevertheless, these estimations must be interpreted with caution in light of the great underreporting of the disease (Ruff 1989; Fleming et al. 2000). An example of this being in the South Pacific where the reported rate of ciguatera disease was estimated to be at 10–20% (Ruff 1989).

In the literature, the number of reported cases of ciguatera poisoning has increased over the last three decades. This apparent global increase of incidences is mainly because of tourism expansion, economic exchanges as well as natural and human-related environmental disturbances (Van Dolah 2000). Climate changes such as ocean temperature increases were also proposed as potential influences on the incidence of this food poisoning (Epstein et al. 1993). However, one study focusing on temporal trends of incidences in South Pacific countries from 1975 to 1990, noticed different situations according to the country. Over a 15-year period, some countries recorded a steady decrease or increase in ciguatera incidence while others reported an increase followed by a decrease (Lewis 1992). Since then, no published study focused on temporal incidence trends of this disease in endemic areas. Documenting such patterns is, however, an important issue for surveillance in island countries where ciguatera has great economic and public health impacts.

To this day, ciguatera diagnosis depends on clinical symptoms and epidemiological inquiries. However, symptoms of this seafood poisoning appear complex and are easily misdiagnosed by physicians unfamiliar with the disease (Kraa et al. 1994). Related medical literature consists mainly of clinical case reports and few studies based on a large number of patients have focused on the epidemiological aspects of this poisoning (Bagnis et al. 1979; Bagnis 1979; Glaziou & Martin 1993; Katz et al. 1993). The severity of the disease tends to vary from one patient to another and from one area to another (Tosteson 1995; Van Dolah 2000), and factors that would help explain this variation are still poorly studied and therefore poorly understood. However, it has been suggested that severity varies according to age, gender and history of previous intoxications (Bagnis et al. 1979; Bagnis 1979; Glaziou & Martin 1993; Katz et al. 1993).

In this context, data from the public health directorate were used to explore changes over time in incidences of ciguatera poisoning in French Polynesia from 1992 to 2001. Furthermore, detailed medical information collected on a sub-sample of cases recorded during this study period allowed us to describe symptoms and investigate the contribution of certain factors that could be related to the severity of the disease.

Population and methods

Case records

French Polynesia is a territory located in the South Pacific Ocean. It is comprised of 118 islands scattered among five archipelagos: Society, Marquesas, Tuamotu, Gambier and Austral. In 2001, the Institut de Statistiques de Polynésie Française estimated the total population at 235 200 inhabitants, 67% living on the island of Tahiti (Society archipelago). French Polynesia has 82 health centres including public hospitals and dispensaries. Every month, the public health directorate receives reports of notifiable diseases such as ciguatera from all public health centres. These monthly reports comprise case count regrouped according to four age categories (less than 1 year old, 1 to 4 year old, 5 to 14 years old, fifteen years old and over). For the present study, we used monthly reports recorded from January 1992 to December 2001 and the yearly population census provided by the Institut de Statistiques de Polynésie Française to calculate yearly incidence rates by the archipelago. We also estimated the population in each age group by archipelago in order to obtain incidence rates by age group. These calculations were based on the distribution of the population recorded during the 1996 census, which was also the midpoint period of the study. In this analysis, we use the same case definition of ciguatera as used by the public health directorate. However, we paid some particular attention to the high number of cases, and we confirmed these counts with the public health directorate referees. If any doubts about diagnosis appeared, we excluded cases.

Since the middle of the 1970s, a parallel surveillance programme of ciguatera disease has been ongoing at the Institut Louis Malardé (ILM). This programme is based on the voluntary participation of physicians. This latest programme created with research goal focused on clinical information compared with governmental surveillance programme, which had an administrative focus (i.e. ciguatera case count). On a regular basis, the institute receives standardized individual clinical records sent by public health physicians of French Polynesia. For each ciguatera case, these physicians complete a standardized individual clinical record made by the ILM. These records provide a detailed description of cases according to gender, age, symptoms experienced, number of past ciguatera events and type of fish consumed. In this study, the ILM provided us with 2648 individual clinical records received between 1992 and 2001. A ciguatera case was defined as the acute onset of neurological, gastrointestinal and or systemic symptoms after (<12 hours) consuming a good-tasting local reef fish. Exclusion criteria were any one of the following such as missing information on the following: age, gender, archipelago of residency, fish species consumed, past ciguatera event and on symptoms experienced.

This study was submitted and approved by the ethical committee of French Polynesia (#38 CEPF).

Statistical analysis

Incidence rates were standardized by direct method using the age composition of the entire French Polynesian population (Greeland & Rothman 1998). Ratios and differences were computed to compare the incidence rates according to month, season and location. Two-sided P values for trends of rates over time were calculated using Poisson maximum likelihood regression (Kleinbaum et al. 1998).

All continuous variables of the sub-sample for which clinical data were available are presented as arithmetic means ± SD. The means for continuous variables were compared by conventional t-test, and Fisher exact tests were performed to compare proportions. When the distribution of the variable was not normal, parameters of dispersion presented were median and interquartile range (IQR = Q75-Q25), and comparisons were realized with Mann-Whitney U-test.

Descriptive analysis highlighted four main classes of symptoms experienced by patients (i.e. neurological, gastrointestinal, cardiac and systemic symptoms) independently of the number of symptoms experienced. An index of the variability of symptoms (IVS) was then created based on these classes of symptoms. This index, categorized into three levels (‘light’, ‘moderate’ and ‘severe’), took into account if patients experienced only 1, 2, 3 or all classes of symptoms, respectively. The higher the index, the more extended the disease. Ordinal logistic regressions were performed to examine the relationship between the index of the disease and factors provided by patients. The goodness-of-fit of the model was examined by testing the proportional odds assumption. A small P-value suggested that the fitted model was an inadequate model.

All statistical analyses were performed with a significance level of α = 5% except for multiple comparisons in which α was adjusted by the Bonferroni method. All analyses were realized using SAS software release 8.2 (SAS Institute Inc., Cary, NC, USA).

Results

Incidence trend

Over the 10-year period of the study, 7963 cases of ciguatera disease were declared to the French Polynesian health directorate. After excluding 121 cases because of uncertain diagnosis, the final sample retained included 7842 incident cases of ciguatera for 2 207 104 person-years. As presented in Table 1, this number appears equally distributed among archipelagos. However, Society shows a significant lower incidence rate compared with other archipelagos (P < 0.0001 in all comparisons).

Table 1.   Incidence rates per 10 000 person-years of ciguatera disease recorded from 1992 to 2001 according to archipelago and the whole territory
 ArchipelagosWhole territory
AustralMarquesasTuamotuSociety
  1. *Value of incidence was rounded up to the nearest whole number.

Total cases recorded13012037225319517842
Crude Incidence rate*1972511651036
Age-standardized incidence rate*1932801561036
Specific Incidence rate* (by age group)
 <1 year41644550
 1–4 years31635331
 5–14 years36686964
 >15 years2954182161331

This difference between archipelagos was observed in all periods under study, as presented in Figure 1. Society archipelago shows the lowest yearly incidence and Marquesas the highest. Statistical trend analysis suggested a significant global decrease in the yearly incidence rate of ciguatera in French Polynesia (P, trend <0.001). In Society and Marquesas archipelagos, a significant decrease and increase are detected respectively (P, trend <0.001). In Tuamotu and Austral, no temporal trends reached the statistical significance with P values equal to 0.09 and 0.06, respectively. Nevertheless, a visual inspection of all the data did not support the significant trends observed (Figure 1). Indeed, the distribution of incidences over the 10-year period suggests that significant trends could have been triggered by cases recorded in 1993, 1997 and 2000.

Figure 1.

 Time trends of incidence of ciguatera in French Polynesia by archipelago, 1992–2001. Rates are age-standardized to the entire French Polynesian population.

Differences in incidence rates were observed according to the month of the year in all of the archipelagos varying from 96 to 306 per 10 000 person-months. July (cool season) presented the weakest incidence rate and January (hot season) the highest. However, no trends were discernible when the normal time course of the year (January–December) was considered (P, trend = 0.09; Figure 2). Furthermore, we observed significant differences between both seasons. The incidence rate during the cool season (April–August) appeared significantly lower than the incidence rate recorded during the hot season (September–March) (P < 0.001). Similar significant results between seasons were observed only in Tuamotu and Austral archipelagos (Figure 2).

Figure 2.

 Specific incidence rates of ciguatera by archipelago and month (trends) and by season (differences between cold and hot seasons are only significant in Tuamotu and Austral archipelagos, **P < 0.05).

Among all cases in the territory, less than 1% were <1 year of age, 13% were children aged between 1 and 14 years and the majority of cases were 15 years old and above (86.4%), the adult group presenting the highest incidence rate. Similar results were observed for each archipelago considered separately (Table 1). Notice that incidence rates in Marquesas archipelago are very high in each age group. This is particularly remarkable for babies compared with babies from other archipelagos.

Clinical records

From 1992 to 2001, the ILM received clinical reports from 2647 ciguatera cases. One hundred and nine cases were excluded from the analysis as a result of missing information on gender or age. We also excluded cases with missing information on symptoms experienced (n = 715). Excluded patients did not appear statistically different from other patients according to mean age (P = 0.42), gender proportion (P = 0.52) and archipelagos of residence (P = 0.65). The final sample used included 1824 clinical records.

The gender ratio (M/F) was 1.89. Median age of men was 37.00 years (IQR = 20) and was significantly higher than women (median: 36.00; IQR: 21) (P = 0.01). As previously observed in public health records, babies and children under 15 years represented a low proportion of ciguatera case reports (5.4%). Young adults represented 24.4% of patients (15–24 years) and approximately 70% of patients were in their second to fourth decade. Exactly 29.8% of the patients were found in the third decade. People over 55 years represented only 13.8% of the total study group.

Patients reported the consumption of more than 83 fish species potentially related to ciguatera; most of them were carnivorous (67.5%) followed by herbivorous fish (29.1%) and invertebrate eaters (3.4%). Among the fish found to be associated with the disease, the predominant species was parrotfish (9.7% of the cases reported). Moreover, we observed that in Austral, Marquesas and Tuamotu, 99% of the time, the fish consumed was harvested locally. However, in the Society archipelago the situation was slightly different as only 53.1% of the fish consumed was caught locally; the remainder came from a remote archipelago.

Frequencies of symptoms reported by patients are presented in Table 2. Considering the great differences in proportion between genders, we decided to show frequencies of symptoms for whole population as well as for both genders separately. Descriptive analysis revealed the predominance of neurological symptoms (frequencies around 90%) over other symptoms (Table 2) and showed that patients may experience from one to eleven different symptoms simultaneously (median: 5.00, IQR: 2.00). The majority of patients (53.8%) experienced a combination of neurological, gastrointestinal and diffuse symptoms and less than 4% experienced only one type of symptom (neurological or gastrointestinal or systemic). Interestingly, none of the patients presented cardiac symptoms alone. According to the IVS, around 13% of the patients had experienced severe ciguatera, 57% had experienced a moderate form of the disease and 30% of the patients had experienced a ciguatera qualified as light.

Table 2.   Symptom frequency reported by patients (in percentage, by gender)
SymptomsGenderTotal %P value*
Men (%) n = 1195Women (%) n = 629
  1. *Difference of proportion between genders according to symptoms.

Cardiac symptoms
 Bradycardia14.011.313.00.118
 Hypotension6.99.17.70.108
Gastrointestinal symptoms
 Diarrhoea78.273.976.70.049
 Vomiting31.533.732.20.361
Neurological symptoms
 Paresthesia89.687.788.90.262
 Trouble with cold perception88.191.289.20.052
 Vertigo53.861.856.30.001
 Itching51.350.250.90.687
 Skin lesions9.213.010.50.017
Systemic symptoms
 Pain82.986.284.00.075
 Dysuria24.319.822.80.031

Among symptoms listed in Table 2, significant differences appeared between genders. Men seemed to experience diarrhoea and dysuria more often than women. On the other hand, men experienced vertigo, skin lesions because of intense itching, trouble with cold perception and generalized pain less frequently than women. No other statistical differences were found for other symptoms according to gender.

According to our IVS, being in a higher category of the index was neither associated with gender [odds ratio, OR: 1.11 for male vs female, 95% confidence interval (CI): 0.90–1.32] nor with the type of fish consumed by the patients (OR: 1.11 for carnivorous vs invertebrate eaters, 95% CI: 0.66–1.82 and OR: 1.13 for herbivorous vs invertebrate eaters, 95% CI: 0.67–1.92). Moreover, no association was observed between the IVS and the number of events of ciguatera disease per category (0, 1 and >1 event) (P = 0.82). We did find a higher IVS, with a 0.04% increase in odds with each year of age. However, being in a higher category of IVS was strongly associated with people aged in the third to fifth decade compared with young people (less than 19 years old) (OR: 1.43, 95% CI:1.15–1.79) or older people (50 and over) (OR: 1.86, 95% CI:1.38–2.51). Finally, we also observed a positive association between the SI and the archipelago of residency (remote archipelago vs living in Society) (OR: 1.23, 95% CI: 1.02–1.50]. All aforementioned analysis of the goodness-of-fit statistic indicated a good fit (P < 0.05).

Discussion

In this paper, we examined the temporal trends of ciguatera poisoning over a decade in French Polynesia and our results indicate a relatively constant annual incidence. However, compared with older data (Lewis 1986), we observed a decrease in the incidence rates since 1973. It fell from 545 per 100 000 person-years (1973–1983) to 363 per 100 000 person-years in the 1992–2001 period (this study).

Throughout French Polynesia, we observed that the annual incidence rates varied considerably according to location, the time and also the magnitude. Furthermore, a gradient of remoteness was observed; the archipelago furthest from Tahiti, Marquesas archipelago, was the chain of islands for which the annual incidence was the highest at all periods. This apparent gradient may reflect a differential information bias, as the information recorded by the public health directorate is provided only by physicians of the public health sector. Public healthcare is more prevalent in remote archipelagos than in Society archipelagos where healthcare is also offered by the private sector. This bias would probably contribute to the lowest incidence rates observed in Society archipelago. Nevertheless, a similar gradient was also observed in a 25-year retrospective study (1960–1984) in French Polynesia (Bagnis et al. 1985). Compared with this previous finding, incidence rates found here were generally lower but the largest variation was observed for Society archipelago where a 10-fold decrease was found. Differences in diet related to the sociological and geographical context of French Polynesia might also play a significant role in the results reported before. The Society archipelago includes Tahiti Island, the main island of French Polynesia, which is connected on a daily basis, by plane or boat, to other industrialized parts of the world. This particular context leads to diverse dietary habits that could modify the ciguatera exposure between the different French Polynesian archipelagos. The diet in Society archipelago is probably less based on fish than that of other remote archipelagos. A transition from traditional to occidental dietary habits which probably occurs in Society archipelago was also proposed for Pacific islands (Coyne 1984; Lewis 1986). However, a complete dietary study would be required to examine reductions of fish consumption and their effect on ciguatera exposure.

In addition, we cannot rule out the influence of another relevant factor, the toxicity of fish which varies according to location (Tosteson 1995). Fish caught in Society have already been reported to be less toxic than those caught in remote archipelagos (Bagnis et al. 1985). This variability has been recognized as linked to many factors such as ecological disturbances, quality and quantity of corals, water temperature, turbidity, salinity and the like (Carlson 1984; Bagnis et al. 1990,1992; Chinain et al. 1999). Moreover, we found that incidence rates changed according to season in all archipelagos. Our results corroborate with previous findings (Lawrence et al. 1980; Tosteson 1995) and could be explained by fluctuations of dinoflagellate organisms that are growth temperature-dependant (Tosteson et al. 1988; Lewis 1992).

As noted in earlier studies (Bagnis et al. 1979; Glaziou & Martin 1993; Katz et al. 1993; Goodman et al. 2003), our results indicate a common increase of incidence rates with age. The adult group presents the highest incidence rate. French Polynesians have an extensive traditional knowledge of this seafood poisoning, know the potential risk of eating reef fish and would prefer to give babies and children deep water fish instead of reef fish (Lewis 1986). The different dietary habits among age groups may explain the age distribution of ciguatera poisoning. However, these results and particularly in those obtained for babies from Marquesas require to be clarified in future anthropological study.

As stated in previous studies, the incidence of ciguatera disease is suspected to be underestimated by physicians unfamiliar with the disease and by patients themselves (Lewis 1986; Ting et al. 1998). In addition to the differential bias discussed, our incidence rates may also suffer from selection bias related to individuals who choose to use traditional remedies until their state requires medical assistance (Lewis 1986). This suggests that cases of ciguatera reported in our study as well as in earlier studies were probably the cases were more severe; mild cases would not have been reported. However, nothing leads us to believe that under-declaration of patients would be different from one archipelago to another.

In the second part of the study, we analysed individual cases recorded by the ILM that correspond only to 23% of the cases recorded by the public health directorate during the same time period. Public health physicians provide these clinical records on a voluntary basis; therefore, a selection bias may have been introduced into our results. The direction of this bias is difficult to predict as case records are probably related to the workload of the physicians, interest for the programme and the like. Moreover, we observed that the highest percentage of these records came from Society archipelago (32%), the rest of the records was equally distributed across the three other archipelagos (around 20% each). Nevertheless, our results on types of symptoms reported and their frequencies are in accordance with previous results that have described the disease (Bagnis et al. 1979; Bagnis 1979; Katz et al. 1993; Glaziou & Legrand 1994; Lewis 2001). This suggests that our results are representative of ciguatera fish poisoning in the South Pacific as presented in the literature.

A detailed analysis of 11 symptoms experienced by patients reveals the excessive complexity of the ciguatera symptoms as observed by others (Morris et al. 1982a). It also reinforces previous results that presented cardiac symptoms as an indicator of the severity of ciguatera disease (Morris et al. 1982a; Katz et al. 1993) although cardiac troubles were always observed in association with other types of symptoms in patients from our study. Nevertheless, the complexity of the diagnosis means that there is a crucial need for confirmatory laboratory testing that would be very useful for physicians.

We observed that a great proportion of patients were in their second to fourth decades of age which is consistent with other studies (Bagnis et al. 1979; Glaziou & Martin 1993; Katz et al. 1993; Goodman et al. 2003). Moreover, we found that this age group suffered from more symptoms than other patients. This is also in accordance with previous results (Katz et al. 1993). As earlier findings suggest, it is possible that our results are related to the amount of fish ingested; children eating smaller portions than adults (Morris et al. 1982b; Goodman et al. 2003). Another possibility is the sensitization process that increases with age (Bagnis et al. 1979; Glaziou & Martin 1993; Katz et al. 1993; Goodman et al. 2003). However, our results did not corroborate this hypothesis. Older people (>65 years) did not appear more afflicted with the disease than other groups as proposed in other studies (Katz et al. 1993). More prospective follow-up work would be required to clarify to what extent age may be a prognosis of ciguatera disease.

Past studies have found that some symptoms vary according to gender and we observed similar results (Bagnis et al. 1979; Glaziou & Martin 1993). As proposed by Goodman et al. in 2003 and other researchers (Lewis 1986; Goodman et al. 2003), we believe that gender differences might be attributed to different dietary habits, leading to differential exposure to ciguatoxins. However, future epidemiological and anthropological inquiries would help to corroborate or contradict this latest hypothesis.

Regarding fish consumption, we found no association between the type of fish consumed (herbivorous vs carnivorous) and the IVS as well as the type of symptoms. As showed by Legrand and Bagnis (1991), we observed that Acanthuridæ, Serranidæ as well as Scaridæ were among the most high-risk species in French Polynesia (Legrand & Bagnis 1991; Glaziou & Martin 1993).

In summary, with regard to the question of whether there is an increasing trend of ciguatera over time, our results do not indicate an increase of the incidence. As presented by other researchers (Bagnis et al. 1985; Lewis 1986), ciguatera disease appears to be a complex problem in the South Pacific that needs to be handled differently according to the particular context of each island nation. Our results, then, suggest that risk management of ciguatera disease in French Polynesia might be adapted to the particular context of each archipelago. Since 1992, Lewis has deplored the inaccuracies of recorded data (Lewis 1986). Ten years later, our results still suffer from this information bias. As proposed earlier, with the objective to improve management of the disease in endemic areas, there is a need to establish a surveillance programme, which accurately quantifies the incidence of the disease. However, as trends observed in previous studies suffered from the same type of bias, it is possible that our conclusions on the stability of rates are true.

Lewis presented Pacific Island Countries as a mix of rural atolls and islands subject to western influence (Lewis 1986). From this point of view, our results may be generalized to Pacific Island Countries. We would speculate that current ciguatera incidence rate is stable for the pacific area compared with rates recorded previously, but with the less developed island countries having a higher incidence rate than those that are more developed.

The second part of this study allowed us to address the clinical profile of some patients who contracted ciguatera disease during the study period. Our results confirmed the gastro-neurological features of ciguatera. Further studies are needed to specify and clarify factors that influence the variability of symptoms between patients.

Acknowledgements

The authors thank the Centers for Disease Control and Prevention (CDC) and the Pan American Health organization (PAHO) for financial support of the study. We would also like to thank the Institut de Statistiques de Polynésie Française for providing us population data and Ms Suzanne Gingras for her expertise in statistics.

Ancillary