There is a hundred year-old history of control of malaria in African urban enviroment 1. However, increasing attention is being devoted to this issue as a result of the rapid growth of cities and the fact that by 2025, more than 50% of the people in Africa are expected to be living in urban centres 2. Despite its broad usage, the word 'urban' has an amazingly varied set of definitions. Most studies investigating the health consequences of urbanization employ ecological and cultural context comparisons of individuals living in urban versus rural environments 34567. There is no simple association between urbanicity and health 6, but the rapid increase of the world's urban population has major implications for the transmission and epidemiology of malaria and other vector-borne diseases 89.

In most countries of sub-Saharan Africa, malaria transmission is generally intense in rural areas. In contrast, malaria transmission occurs at significantly lower levels in urban areas, where population density is high, and where the number of mosquito breeding sites is reduced 1011. Few data are available to prioritize strategies for malaria control. Many questions remain for public health specialists hoping to regain control over malaria in urban environments 111213.

In the central highlands of Madagascar, at the end of the French colonial era in the 1950s, malaria and its main vector Anopheles funestus has been suppressed to near or complete eradication. This was achieved by spraying Dichlorodiphenyltrichloroethane (DDT) mixed with gamma-hexachlorocyclohexane (γHCH) in over a million houses and mass chemoprophylaxis of schoolchildren with chloroquine given 3 times per week in 2,375 treatment centres 14. As a result, the percentage of positive blood slides in Antananarivo was reduced from 25–35% in the late 1940s to 0.18% (6 positive among 3,348 slides from children between two and 10 years of age) in 1955. An. funestus, which had been abundant in highland houses in the 1940s, could not be found in the area in 1955 14. Some spraying continued and there were centres at which chloroquine was available until the late 1970s, when this intervention was, regrettably, abandoned 15. An. funestus crept back into the area between 1976 and the 1980s. A sporozoite positive individual was found in 1987 and a plea was made for urgent introduction of control measures 16. This plea was not heeded in time and an explosive epidemic hit the highland region in the late 1980s, the prevalence of parasitaemia returned to the levels seen in the 1940s 171819 and tens of thousands died amongst the non-immune population 2021.

In 1988, to control the spread of the vector, a national control programme that was based on accessibility to chloroquine and DDT indoor spraying aimed at the foci where the most malaria cases were found. From 1993 to 1998, the programme was extended to all 27 districts of the central highlands, including all areas affected by malaria. The main targets of the control programme were the villages located in altitudes of 1,000–1,500 m, representing the majority of the human settlements in the central highlands. Five malaria vector control campaigns, the so-called "Operation de Pulvérisation Intra-Domiciliaire" (OPID) programmeme, were funded by the World Bank. The goal was to protect a population of 2.3 million inhabitants per year.

Five cycles of residual indoor treatment with DDT in the central highlands of Madagascar have resulted in low levels of malaria transmission in the study area. The low prevalence of the infection and the rarity of the main vector, An. funestus, observed at the end of the OPID programme were the basis for suspending the antivector interventions in 1998–1999. Instead, more targeted interventions, the "Campagne d'Aspersion Intra-Domiciliaire" (CAID), was introduced, which still uses 2 g DDT/m2. These latter interventions have consolidated the impact on the vector but have not succeeded in stopping transmission. Certain cities as Antananarivo are no longer included in the CAID 2223.

There is now full control of the epidemics and a considerable decrease in malaria transmission in the Highlands. As a result, febrile episodes are now often considered as malaria attacks based only on a clinical diagnosis, although the real contribution of malaria to febrile episodes remains unclear.

Two cross-surveys in February 2003 (rainy season) and in June 2003 (dry and cold season) in primary health centres in Antananarivo were conducted. The principal aim was to detect malaria attacks during febrile episodes in Antananarivo and to determine if there is local transmission in Antananarivo. The usefulness of malaria Rapid Diagnosis Test (RDT) was also assessed for fever cases management in heath care centres.

During the study, 771 of 5,428 outpatients were included in the 43 dispensaries during March/February for 739 of 5,314 outpatients in June/July. The populations characteristics (age, sex) were comparables between the two seasons (Table 1).

With regard to seasonal patterns, presumptive malaria rates varied from 59.2% (55.8–62.7) of all diagnoses in February to 32.9% (29.6–36.3) in July. Respiratory problems were diagnosed in 46.5% of the cases in July. However, the proportion of antimalarial drugs prescribed did not differ between the two seasons. Generally, there was a frequent prescription of an antimalarial and antibiotic together. In February, among 771 outpatients, 375 (48.7%) were treated with antimalarial drugs. Among these 375 subjects, 225 (60%) had both antimalarial drugs and antibiotics. In July, among 739 outpatients, 365 (49.4%) were treated with antimalarial drugs. Among these 365 subjects, 150 (41.1%) had both antimalarial drugs and antibiotics.

In February, in surveys made around the two index cases, five subjects were RDT positive. Thick and thin blood films confirmed the presence of P. falciparum for four subjects. They had not travelled out of urban Antananarivo during the previous four weeks. The two index cases and the four subjects are located in three houses (Table 2 and Figure 3).

The entomological survey detected three vectors: Anopheles arabiensis, An. funestus and Anopheles mascarensis. For the collection of adult anophelines, An. arabiensis was the most common species found during the two study seasons. Human-bait collection was the most successful method (Table 3).

Of the 338 anophelines (An. arabiensis, An. funestus, and An. mascarensis) collected and examined by ELISA to detect circum sporozoite protein, none was found positive. The biting rates of An. arabiensis, the most common species, were relatively low. An. arabiensis manifested exophagic tendencies (Table 4). Parity rates were 57% in February-March and 53% in April.

PCR amplification of the msp2 P. falciparum for seven infected subjects (one subject was RDT positive but slide negative) showed polymorphism (Figure 4). Genetic analysis of P. falciparum strains allowed distinguishing three genotypes, aggregated by house. The three P. falciparum samples genotyped from the subjects 89, 91 and 94 (house A) showed an identical band with a size of 600 pb. The subject 92 (House B) showed a band of 550 pb size. The subjects 93 and 95 (House C) showed two identical bands of a size from approximately 600 pb and 650 pb. There was no amplification of DNA from the subject 90 who was RDT positive but slide negative. The nested PCR determined that all P. falciparum samples belonged to the 3D7 msp2 allelic family. Genotypes were also aggregated by house on the same mode as that obtained with the msp2 primers.

It was originally thought that urban areas do not support significant levels of malaria transmission. The concentration of human populations in small areas is normally accompanied by pollution and destruction of the clean water sources needed by the anopheline vectors of malaria. However, this is true for formal urban development but not for the informal development that occurs in sub-Saharan African countries and in Madagascar.

From the parasitological survey, 1.9% and 1.5% of the study population were found to be infected in February-March and in June-July, respectively. P. falciparum was the most common species followed by P. vivax and P. ovale. It is known that P. falciparum is the dominant species in Madagascar 141727. In the same epidemiological facies (highland malaria in an urban area), Woyessa and al. also showed a relatively low proportion of malaria cases in Akaki Town, Ethiopia. Among 2136 blood films, they found 3.7% malaria positive with 69% of them due to P. vivax and 31% due to P. falciparum 28. In N'Djaména, Chad (urban malaria in Sahelian towns with highly seasonal malaria transmission), Othnigué and al. found that among 1,658 patients 30% were positive for Plasmodium (all P. falciparum)29.

In the present study, there were very few P. falciparum malaria cases in both the low and high transmission seasons. The intensive epidemic control measures, such as treatment of confirmed cases and especially vector control activities, after the 1986 outbreaks are thought to have contributed to the low infection rate 233031.

This study confirms that malaria cases in Antananarivo are mostly associated with travel to areas with high levels of malaria transmission, particularly the highlands fringe areas or the coastal regions of Madagascar. Nevertheless, autochthonous malaria transmission is suspected in some locations. The surveys around the two index cases seem to confirm this observation:

(i) Six subjects diagnosed positive, around the index cases, confirming that they had not travelled out of urban Antananarivo during the previous four weeks.

(ii) Anopheline vectors (a great majority of An. arabiensis) were captured by pyrethrum spray collections in the bedrooms of the two index cases and in the surrounding houses and by human landing catches of mosquitoes. The principal malaria vector on the Madagascar Highlands is An. funestus. More than 95% of its breeding sites are in the rice fields just before the harvest and then afterwards in the fallow lands. The vector peak and corresponding peak of malaria cases occurs between February and May, depending on the farming calendar. The second but less important vector, An. arabiensis, breeds in rice fields just after seeding when the surface water is sunlit. Although rice fields are the principal breeding site for this vector, it also breeds in rainwater pods 27. From these findings, An. arabiensis may be the most important species involved in the transmission of malaria in Antananarivo, although it occurs in relatively small numbers compared to other areas 32. The absence of sporozoïtes infection in this species may also be due to the small sample size of mosquitoes. Only, one An. funestus was aggressive for human in February-March. However, more detailed studies are required on bionomics of An. arabiensis and An. funestus to determine their exact role in malaria transmission in urban Antananarivo.

(iii) The six thick positive subjects were located in three houses near rice fields, an environment shown to be favourable for vector breeding 2733.

(iv) Genetic analysis of P. falciparum strains allowed to distinguish three genotypes, aggregated by house.

Generally, malaria transmission is negatively associated with urbanization as shown in Kinshasa 34. This is primarily due to overcrowding, to elimination of breeding sites due to construction, or to pollution of the breeding places resulting in lower vector densities in urban centres. However, in Antananarivo, autochthonous malaria transmission could also be maintained for short periods of time 222733 in locations with rice fields. A low level of transmission may persist in these areas and may lead to further outbreaks of malaria in the future due to the presence of anopheline vectors. Thus, sustainable and integrated vector control measures and appropriate case management should play a crucial role in preventing further malaria epidemics. Moreover, studies need to be carried out to improve the epidemiological understanding of the cause and origin of the transmission.

The study demonstrates deficiencies in fever cases management in health centres, since many febrile episodes are wrongly identified as malaria, despite full control of the epidemics. Madagascar has now revised its national malaria treatment policy and is currently implementing a change from chloroquine to artemisinin combination treatment (ACT), particularly artesunate-amodiaquine. The low quality of malaria case management, including an over-prescription of antimalarials, mostly chloroquine but also quinine, highlights the problem of drug overuse and the associated risk of more rapid development of resistance. This study shows the importance of strengthening diagnosis and improving the treatment of indigenously transmitted malaria in areas of Antananarivo areas where diagnosis could be overlooked. The results show that RDTs would be useful and should be implemented in areas of transmission, particularly in circumstances that are not covered by traditional microscopic diagnosis of malaria.

RLP is the principal investigator of the study and led the drafting of this manuscript; AF conceived the study and participated in its design and coordination; MR is investigator of the study, participated in the design of the study, supervised the acquisition of the data, helped analyse the data; CS, RAL and RLH, supervised the acquisition of the data, carried out laboratory analysis and helped analyse the data;

Le B J, RM and RV all conceived the study, took part in its development and revision.

All authors read and approved the final manuscript.