Assessment of the risk of malaria re-introduction in the Maremma plain (Central Italy) using a multi-factorial approach

The risk of malaria reintroduction was evaluated as the "malariogenic potential" of the study area by assessment of the three parameters that define it:

Approaches to evaluate these parameters were:

1) Field collection of further entomological data (bionomics, distribution, abundance) for mosquitoes of the An. maculipennis complex; 2) investigation of seasonal dynamics of the vector through the implementation of a weather-based statistical dynamic model; 3) production of a distribution/predictive map of An. labranchiae across the study area; 4) evaluation of the length of the possible transmission season for P. vivax and P. falciparum; 5) assessment of the vector competence of the species to P. falciparum by artificial infection; 6) evaluation of the vectorial capacity of An. labranchiae in the site where the species is most abundant; 7) risk assessment related to the possibility that the vector may feed on gametocyte carriers occasionally circulating in the study area.

Maremma is a coastal plain of Central Italy, that covers about 5,500 km², extending along the Tyrrhenian coast for about 180 km from mid Tuscany (northern limit 43,3°N) to upper Latium (southern limit 42,17°N). Inland, Maremma extends from the sea to the pre-Apennine foothills, with a width ranging from 16 to 50 km. The territory includes different biotopes where natural and anthropic environments coexist. A wide sandy seashore and dunes, covered by typical Mediterranean bush, gives way to extensive pine woods. Large pastures alternate with intensively cultivated areas up to a range of hills that represents the inner Eastern limit of the study area. About 44,000 ha, nearly 10% of the whole surface of the Maremma plain, is included in protected reservation parks. The study area corresponds to the "core" of this region, shared between the administrative Provinces of Grosseto, and Siena (Tuscany) and Viterbo (Latium) (Figure 1). Ten collection sites were selected on the basis of their environmental characteristics, including three large rice cultivation areas, five rural areas with intensive farming and two protected coastal areas (Table 1).

Collections of An. maculipennis s.l. mosquitoes were performed between 2005 and 2009 during 40 surveys. In particular, sites 1, 3, 4 and 6 were visited by 28 fortnightly surveys carried out from April 2005 to October 2006 [23], while in 2007-2009 these sites and the remaining ones were subject to sporadic surveys (1-4 times) in June-August. Mosquito larvae were collected by an enamel standard 500 ml dipper. The number of dipping stations and dips by station was adequate to the type and size of the breeding sites visited according to a standardized protocol [23, 29]. Adult collections, using manual or battery-powered aspirators, were mainly targeted at resting females in animal shelters; other kinds of premises, such as haylofts, woodshed, fodder and tools depots represented less than 5% of the premises inspected. A minor fraction of An. maculipennis s.l. females were collected by CDC light/CO2 traps. Three night catches on human bait were seasonally carried out in late June, mid July and late August, between 2005 and 2007, respectively in Principina, S. Donato, and Val di Merse. Catches were performed as described in Romi et al. [29]. Mosquitoes were analyzed for species identification, blood meal source and age population structure as described in Di Luca et al. [23]. Previous entomological data used for comparisons are from the Operative Unit of Environmental Zoology, AUSL 9, Grosseto.

A binomial model was developed for the less productive sites (4 and 6) categorizing the entomological data with the threshold of one adult ("A" absence category/"B" presence category), while the multinomial model, developed for the most productive site 1, was categorize with the thresholds of one and 60 adults ("A" absence category/"B" medium-low presence category/"C" medium-high presence category). The multi-logistic model input were derived from a principal component analysis that leads to an optimal dimensionality reduction of matrix predictors. Akaike Information Criterium (AIC) [35] was finally used for the statistical model parameters selection in order to optimize model performances and to give a better discrimination in the microclimatic variability among different sites. Concerning model selection the statistical function (i.e. STEPAIC) used is taken from MASS R package [36, 37]. Weekly model outputs consist essentially on the occurrence probability associated to each adult abundance category, hence the forecasted category was chosen as the one with highest probability. The length of development season was estimated as the number of weeks between the first and the last week with a presence category (B or C category) along the year [38‐41]. A verification procedure to assess the reliability of the models was implemented for each site using skill score indexes derived from contingency Tables with observed and forecasted values: the BIAS index, the POD index (Probability Of Detection) and the FAR index (False Alarm Ratio) [42].

The quantitative mapping of larvae/adult mosquito presence and abundance was implemented following the approach suggested by Tran et al. [43]. The logistic regression model for Anopheles hyrcanus in the Camargue region, that explained the presence of larvae as a function of biotope and distance to the nearest rice field, was modified in order to obtain a larval index (probability of observing one larva in a point of a biotope at least once during the mosquito season) consistent with the collection data reported in Table 1. The logistic regression model for An. labranchiae, used in this study and implemented with Builder tool from ESRI ArcGIS™, consider also the distance from farm with livestock to the main biotopes as a new explanatory variable in addition to those used by Tran et al. Assuming that the abundance of adult mosquitoes is influenced by the presence of breeding sites in the surroundings, the adult index map was derived from the larval index map. Corine Land Cover 2000 (CLC 2000, produced by the European Environment Agency, EEA) spatial data sets for Italy (scale 1:100,000) was used to describe the environmental characteristics likely to influence the spatial distribution of An. labranchiae. Also a natural colour aerial photos (May 2007) with a spatial resolution of 1 metre was used to detect the main biotopes where An. labranchiae larvae and adults were collected, such as rice fields, reed beds, marshes, temporarily flooded rush wetland and clear water. A photo-interpretation was carried out through a workstation with ESRI ArcGIS™ software. Geographical database in shape-file format was used for localizing livestock and intensive cattle breeding farms in Grosseto Province [44]. The highest abundance of An. labranchiae was assumed to be related to rice paddies, and thus the distance to the nearest rice field was computed for each pixel using Geographic Information System (GIS) functionality.

Since Grosseto Airport weather station (42.75°N-11.07°E, 7 m. a.s.l.) is the climate reference centre for the World Meteorological Organization [WMO], meteorological data from that station were considered representative of the whole study area. Thermo-pluviometric diagram with monthly mean temperature and rainfall amount respectively over a 30-years-period (1961-1990) and over the study period 2005-2009 were compared with the minimum temperature for development of P. vivax and P. falciparum, 15°C and 18°C respectively [45]. The average and absolute values on a yearly and seasonal (May-August) basis of temperature and rainfall were compared for 1961-1990 and 2005-2009 period in order to highlight the possible variation from a climatic point of view (Table 2). An evaluation of the Plasmodium spp possible transmission season along the years 2005 and 2006, for the survey period between 2005 and 2009 and for climatological period 1961-1990 was also performed through the Gradient Model Risk (GMR) index calculation [25, 37]. GMR consists on the monthly evolution of accumulated values of the index itself to gain insight into possible transmission periods along the year; this index considers only climatic parameters as the minimum mean temperature required for the development of the parasite inside the vector, as reported above, monthly precipitation (mm) and PET (mm.), calculated by the equation:

where GDD is growing degree-days with a base temperature of 15°C and 18°C respectively for P. vivax and P. falciparum, R is the rainfall and PET is potential evapotranspiration calculated with the empirical method of Thornthwaite 1948 [46] as a function of mean temperature and latitude. The GMR index shows that a transmission risk exists when its value is equal or higher than 116, that is the value required for one Plasmodium spp generation.

Field samples of An. labranchiae females, collected in site 1 (Figure 1, Table 1) were submitted to the Plateform CEPIA (Institute Pasteur, Paris, France) to artificial blood infection with gametocyte-containing cultures of the P. falciparum NF54 African strain in 2008 and 2009. Production of mature gametocytes and artificial blood infection were performed following procedures described in Mitri et al. [47]. A laboratory colony of Anopheles gambiae (Ngousso, Cameroon) was used as a positive control. Mosquitoes were dissected on 8^tand 15 days post-infection to determine prevalence and oocyst load in the midgut. For each experiment all An. labranchiae females and the control An. gambiae strain were starved 24 hours prior to blood feeding. The infected red blood cells containing P falciparum gametocytes complemented with fresh RBC and human AB serum, were deposited in a Parafilm^® membrane feeder previously warmed at 37°C. After 15 minutes feeding, unfed An. labranchiae were offered a second blood meal on the next day, when possible. Engorged females were kept at 26 ± 1°C inside small cages and were provided with 10% sucrose until dissection 8 days or 15 days post infection. In the 2009, the detection of sporozoites was carried out at the Institute pour la Recherche et le Développement in Montpellier (IRD, France), using the cut head-thorax from the 15^th day survived mosquitoes. DNA extraction was performed by a single-round, multiplex PCR, according to Padley et al. [48]. Legs of all infected females were used for species identification by Multiplex PCR [49].

Vectorial capacity was assessed according to the Macdonald formula [50] revised by Garret-Jones [51]. The experimental variables needed for estimating it were evaluated as follows: the human biting rate (ma) by night catches on human bait, the human blood index (HBI) by the origin of the blood meal of the engorged females collected early in the morning in different premises and the parity rate by ovarian dissection [52]. The factors temperature-dependent, such as the length of the sporogonic cycle (n) of P. falciparum and P. vivax and the duration of gonotrophic cycle (gc) were calculated according to Macdonald [50]. The host feeding preference An. labranchiae was estimated by considering different feeding preference indices in addition to the HBI, i.e. the forage ratio (FR) and the feeding index (FI). FR quantifies vector selection of a particular vertebrate host rather than other available hosts. It was calculated by dividing the percentage of females fed on a given host by the percentage which that host represented in the total census of available animals and humans at the collecting site [53]. FRs significantly > 1.0 indicate a selective bias and values < 1.0 indicate avoidance in favour of other hosts; FRs ≈ 1.0 show neither preference nor avoidance. FI is defined as the observed proportion of females fed on a certain animal host with respect to another one divided by the expected comparative proportion of feeds on these two hosts [54]. This crude index was adjusted by taking in account factors that affect feeding, such as host abundance, their size and their temporal and spatial concurrence with the mosquito species. FI = 1 indicates equal feeding on the two hosts, while smaller or larger values indicate a decrease or increase in feeding on the first host relative to the second, respectively. FI were calculated for each pair of hosts.

The number of gametocyte carriers that may have been circulating in the territory during the period favourable to malaria transmission (June-October) was obtained by the analysis of the cases of imported malaria in Italy in 2000-2009 (cases confirmed by the Malaria Reference Centre at Istituto Superiore di Sanità), selecting those reported from hospitals located into the study area.